OpenCloudOS-Kernel/drivers/gpu/drm/drm_irq.c

1806 lines
53 KiB
C
Raw Normal View History

/*
* drm_irq.c IRQ and vblank support
*
* \author Rickard E. (Rik) Faith <faith@valinux.com>
* \author Gareth Hughes <gareth@valinux.com>
*/
/*
* Created: Fri Mar 19 14:30:16 1999 by faith@valinux.com
*
* Copyright 1999, 2000 Precision Insight, Inc., Cedar Park, Texas.
* Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* VA LINUX SYSTEMS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*/
#include <drm/drmP.h>
#include "drm_trace.h"
#include "drm_internal.h"
#include <linux/interrupt.h> /* For task queue support */
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
#include <linux/slab.h>
#include <linux/vgaarb.h>
#include <linux/export.h>
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Access macro for slots in vblank timestamp ringbuffer. */
#define vblanktimestamp(dev, pipe, count) \
((dev)->vblank[pipe].time[(count) % DRM_VBLANKTIME_RBSIZE])
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Retry timestamp calculation up to 3 times to satisfy
* drm_timestamp_precision before giving up.
*/
#define DRM_TIMESTAMP_MAXRETRIES 3
/* Threshold in nanoseconds for detection of redundant
* vblank irq in drm_handle_vblank(). 1 msec should be ok.
*/
#define DRM_REDUNDANT_VBLIRQ_THRESH_NS 1000000
static bool
drm_get_last_vbltimestamp(struct drm_device *dev, unsigned int pipe,
struct timeval *tvblank, unsigned flags);
static unsigned int drm_timestamp_precision = 20; /* Default to 20 usecs. */
/*
* Default to use monotonic timestamps for wait-for-vblank and page-flip
* complete events.
*/
unsigned int drm_timestamp_monotonic = 1;
static int drm_vblank_offdelay = 5000; /* Default to 5000 msecs. */
module_param_named(vblankoffdelay, drm_vblank_offdelay, int, 0600);
module_param_named(timestamp_precision_usec, drm_timestamp_precision, int, 0600);
module_param_named(timestamp_monotonic, drm_timestamp_monotonic, int, 0600);
static void store_vblank(struct drm_device *dev, unsigned int pipe,
u32 vblank_count_inc,
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
struct timeval *t_vblank)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
u32 tslot;
assert_spin_locked(&dev->vblank_time_lock);
if (t_vblank) {
/* All writers hold the spinlock, but readers are serialized by
* the latching of vblank->count below.
*/
tslot = vblank->count + vblank_count_inc;
vblanktimestamp(dev, pipe, tslot) = *t_vblank;
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
}
/*
* vblank timestamp updates are protected on the write side with
* vblank_time_lock, but on the read side done locklessly using a
* sequence-lock on the vblank counter. Ensure correct ordering using
* memory barrriers. We need the barrier both before and also after the
* counter update to synchronize with the next timestamp write.
* The read-side barriers for this are in drm_vblank_count_and_time.
*/
smp_wmb();
vblank->count += vblank_count_inc;
smp_wmb();
}
/**
* drm_update_vblank_count - update the master vblank counter
* @dev: DRM device
* @pipe: counter to update
*
* Call back into the driver to update the appropriate vblank counter
* (specified by @pipe). Deal with wraparound, if it occurred, and
* update the last read value so we can deal with wraparound on the next
* call if necessary.
*
* Only necessary when going from off->on, to account for frames we
* didn't get an interrupt for.
*
* Note: caller must hold dev->vbl_lock since this reads & writes
* device vblank fields.
*/
static void drm_update_vblank_count(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
u32 cur_vblank, diff;
bool rc;
struct timeval t_vblank;
/*
* Interrupts were disabled prior to this call, so deal with counter
* wrap if needed.
* NOTE! It's possible we lost a full dev->max_vblank_count + 1 events
* here if the register is small or we had vblank interrupts off for
* a long time.
*
* We repeat the hardware vblank counter & timestamp query until
* we get consistent results. This to prevent races between gpu
* updating its hardware counter while we are retrieving the
* corresponding vblank timestamp.
*/
do {
cur_vblank = dev->driver->get_vblank_counter(dev, pipe);
rc = drm_get_last_vbltimestamp(dev, pipe, &t_vblank, 0);
} while (cur_vblank != dev->driver->get_vblank_counter(dev, pipe));
/* Deal with counter wrap */
diff = cur_vblank - vblank->last;
if (cur_vblank < vblank->last) {
diff += dev->max_vblank_count + 1;
DRM_DEBUG("last_vblank[%u]=0x%x, cur_vblank=0x%x => diff=0x%x\n",
pipe, vblank->last, cur_vblank, diff);
}
DRM_DEBUG("updating vblank count on crtc %u, missed %d\n",
pipe, diff);
if (diff == 0)
return;
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
/*
* Only reinitialize corresponding vblank timestamp if high-precision query
drm: Zero out invalid vblank timestamp in drm_update_vblank_count. (v2) Since commit 844b03f27739135fe1fed2fef06da0ffc4c7a081 we make sure that after vblank irq off, we return the last valid (vblank count, vblank timestamp) pair to clients, e.g., during modesets, which is good. An overlooked side effect of that commit for kms drivers without support for precise vblank timestamping is that at vblank irq enable, when we update the vblank counter from the hw counter, we can't update the corresponding vblank timestamp, so now we have a totally mismatched timestamp for the new count to confuse clients. Restore old client visible behaviour from before Linux 3.18, but zero out the timestamp at vblank counter update (instead of disable as in original implementation) if we can't generate a meaningful timestamp immediately for the new vblank counter. This will fix this regression, so callers know they need to retry again later if they need a valid timestamp, but at the same time preserves the improvements made in the commit mentioned above. v2: Rebased on top of Daniel Vetter's fixup and documentation patch for timestamp updates. Drop request for stable kernel backport as this would be more difficult, unless the original patch would get applied to stable kernels. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Dave Airlie <airlied@redhat.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-05-04 12:29:46 +08:00
* available and didn't fail. Otherwise reinitialize delayed at next vblank
* interrupt and assign 0 for now, to mark the vblanktimestamp as invalid.
*/
drm: Zero out invalid vblank timestamp in drm_update_vblank_count. (v2) Since commit 844b03f27739135fe1fed2fef06da0ffc4c7a081 we make sure that after vblank irq off, we return the last valid (vblank count, vblank timestamp) pair to clients, e.g., during modesets, which is good. An overlooked side effect of that commit for kms drivers without support for precise vblank timestamping is that at vblank irq enable, when we update the vblank counter from the hw counter, we can't update the corresponding vblank timestamp, so now we have a totally mismatched timestamp for the new count to confuse clients. Restore old client visible behaviour from before Linux 3.18, but zero out the timestamp at vblank counter update (instead of disable as in original implementation) if we can't generate a meaningful timestamp immediately for the new vblank counter. This will fix this regression, so callers know they need to retry again later if they need a valid timestamp, but at the same time preserves the improvements made in the commit mentioned above. v2: Rebased on top of Daniel Vetter's fixup and documentation patch for timestamp updates. Drop request for stable kernel backport as this would be more difficult, unless the original patch would get applied to stable kernels. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Dave Airlie <airlied@redhat.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
2015-05-04 12:29:46 +08:00
if (!rc)
t_vblank = (struct timeval) {0, 0};
store_vblank(dev, pipe, diff, &t_vblank);
}
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/*
* Disable vblank irq's on crtc, make sure that last vblank count
* of hardware and corresponding consistent software vblank counter
* are preserved, even if there are any spurious vblank irq's after
* disable.
*/
static void vblank_disable_and_save(struct drm_device *dev, unsigned int pipe)
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
unsigned long irqflags;
u32 vblcount;
s64 diff_ns;
bool vblrc;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
struct timeval tvblank;
int count = DRM_TIMESTAMP_MAXRETRIES;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Prevent vblank irq processing while disabling vblank irqs,
* so no updates of timestamps or count can happen after we've
* disabled. Needed to prevent races in case of delayed irq's.
*/
spin_lock_irqsave(&dev->vblank_time_lock, irqflags);
/*
* If the vblank interrupt was already disabled update the count
* and timestamp to maintain the appearance that the counter
* has been ticking all along until this time. This makes the
* count account for the entire time between drm_vblank_on() and
* drm_vblank_off().
*
* But only do this if precise vblank timestamps are available.
* Otherwise we might read a totally bogus timestamp since drivers
* lacking precise timestamp support rely upon sampling the system clock
* at vblank interrupt time. Which obviously won't work out well if the
* vblank interrupt is disabled.
*/
if (!vblank->enabled &&
drm_get_last_vbltimestamp(dev, pipe, &tvblank, 0)) {
drm_update_vblank_count(dev, pipe);
spin_unlock_irqrestore(&dev->vblank_time_lock, irqflags);
return;
}
/*
* Only disable vblank interrupts if they're enabled. This avoids
* calling the ->disable_vblank() operation in atomic context with the
* hardware potentially runtime suspended.
*/
if (vblank->enabled) {
dev->driver->disable_vblank(dev, pipe);
vblank->enabled = false;
}
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* No further vblank irq's will be processed after
* this point. Get current hardware vblank count and
* vblank timestamp, repeat until they are consistent.
*
* FIXME: There is still a race condition here and in
* drm_update_vblank_count() which can cause off-by-one
* reinitialization of software vblank counter. If gpu
* vblank counter doesn't increment exactly at the leading
* edge of a vblank interval, then we can lose 1 count if
* we happen to execute between start of vblank and the
* delayed gpu counter increment.
*/
do {
vblank->last = dev->driver->get_vblank_counter(dev, pipe);
vblrc = drm_get_last_vbltimestamp(dev, pipe, &tvblank, 0);
} while (vblank->last != dev->driver->get_vblank_counter(dev, pipe) && (--count) && vblrc);
if (!count)
vblrc = 0;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Compute time difference to stored timestamp of last vblank
* as updated by last invocation of drm_handle_vblank() in vblank irq.
*/
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
vblcount = vblank->count;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
diff_ns = timeval_to_ns(&tvblank) -
timeval_to_ns(&vblanktimestamp(dev, pipe, vblcount));
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* If there is at least 1 msec difference between the last stored
* timestamp and tvblank, then we are currently executing our
* disable inside a new vblank interval, the tvblank timestamp
* corresponds to this new vblank interval and the irq handler
* for this vblank didn't run yet and won't run due to our disable.
* Therefore we need to do the job of drm_handle_vblank() and
* increment the vblank counter by one to account for this vblank.
*
* Skip this step if there isn't any high precision timestamp
* available. In that case we can't account for this and just
* hope for the best.
*/
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
if (vblrc && (abs64(diff_ns) > 1000000))
store_vblank(dev, pipe, 1, &tvblank);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
spin_unlock_irqrestore(&dev->vblank_time_lock, irqflags);
}
static void vblank_disable_fn(unsigned long arg)
{
struct drm_vblank_crtc *vblank = (void *)arg;
struct drm_device *dev = vblank->dev;
unsigned int pipe = vblank->pipe;
unsigned long irqflags;
if (!dev->vblank_disable_allowed)
return;
spin_lock_irqsave(&dev->vbl_lock, irqflags);
if (atomic_read(&vblank->refcount) == 0 && vblank->enabled) {
DRM_DEBUG("disabling vblank on crtc %u\n", pipe);
vblank_disable_and_save(dev, pipe);
}
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
}
/**
* drm_vblank_cleanup - cleanup vblank support
* @dev: DRM device
*
* This function cleans up any resources allocated in drm_vblank_init.
*/
void drm_vblank_cleanup(struct drm_device *dev)
{
unsigned int pipe;
/* Bail if the driver didn't call drm_vblank_init() */
if (dev->num_crtcs == 0)
return;
for (pipe = 0; pipe < dev->num_crtcs; pipe++) {
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/irq: Don't call ->get_vblank_counter directly from irq_uninstall/cleanup The pipe might already have been shut down, and then it's not a good idea to call hw accessor functions. Instead use the same logic as drm_vblank_off which has all the necessary checks to avoid troubles or inconsistency. Noticed by Imre while reviewing my patches to remove some sanity checks from ->get_vblank_counter. v2: Try harder. disable_and_save can still access the vblank stuff when vblank->enabled isn't set. It has to, since vlbank irq could be disable but the pipe is still on when being called from drm_vblank_off. But we still want to use that code for more code sharing. So add a check for vblank->enabled on top - if that's not set we shouldn't have anyone waiting for the vblank. If we have that's a pretty serious bug. The other issue that Imre spotted is drm_vblank_cleanup. That code again calls disable_and_save and so suffers from the same issues. But really drm_irq_uninstall should have cleaned that all up, so replace the code with WARN_ON. Note that we can't delete the timer cleanup since drivers aren't required to use drm_irq_install/uninstall, but can do their own irq handling. v3: Make it clear that all that gunk in drm_irq_uninstall is really just bandaids for UMS races between the irq/vblank code. In UMS userspace is in control of enabling/disabling interrupts in general and vblanks specifically. v4: Imre observed that KMS drivers all call drm_vblank_cleanup before drm_irq_uninstall (as they should), so again the code in there is dead for KMS (due to dev->num_crtcs == 0 after drm_vblank_cleanup). Or should be, so only WARN for KMS - with UMS userspace could try to do evil things. v5: After more discussion on irc we've gone back to v3: the del_timer_sync is required in all cases in drm_vblank_cleanup, but let's restrict the WARN_ON to kms drivers only. Imre was also concerned that bad things could happen without the disable_and_save call. But we immediately free vblank structures afterwards which makes the save useless. And drm_handle_vblank has a check for dev->num_crtcs to avoid surprises with ums. Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> Acked-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-22 22:11:19 +08:00
WARN_ON(vblank->enabled &&
drm_core_check_feature(dev, DRIVER_MODESET));
drm/irq: Don't call ->get_vblank_counter directly from irq_uninstall/cleanup The pipe might already have been shut down, and then it's not a good idea to call hw accessor functions. Instead use the same logic as drm_vblank_off which has all the necessary checks to avoid troubles or inconsistency. Noticed by Imre while reviewing my patches to remove some sanity checks from ->get_vblank_counter. v2: Try harder. disable_and_save can still access the vblank stuff when vblank->enabled isn't set. It has to, since vlbank irq could be disable but the pipe is still on when being called from drm_vblank_off. But we still want to use that code for more code sharing. So add a check for vblank->enabled on top - if that's not set we shouldn't have anyone waiting for the vblank. If we have that's a pretty serious bug. The other issue that Imre spotted is drm_vblank_cleanup. That code again calls disable_and_save and so suffers from the same issues. But really drm_irq_uninstall should have cleaned that all up, so replace the code with WARN_ON. Note that we can't delete the timer cleanup since drivers aren't required to use drm_irq_install/uninstall, but can do their own irq handling. v3: Make it clear that all that gunk in drm_irq_uninstall is really just bandaids for UMS races between the irq/vblank code. In UMS userspace is in control of enabling/disabling interrupts in general and vblanks specifically. v4: Imre observed that KMS drivers all call drm_vblank_cleanup before drm_irq_uninstall (as they should), so again the code in there is dead for KMS (due to dev->num_crtcs == 0 after drm_vblank_cleanup). Or should be, so only WARN for KMS - with UMS userspace could try to do evil things. v5: After more discussion on irc we've gone back to v3: the del_timer_sync is required in all cases in drm_vblank_cleanup, but let's restrict the WARN_ON to kms drivers only. Imre was also concerned that bad things could happen without the disable_and_save call. But we immediately free vblank structures afterwards which makes the save useless. And drm_handle_vblank has a check for dev->num_crtcs to avoid surprises with ums. Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> Acked-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-22 22:11:19 +08:00
del_timer_sync(&vblank->disable_timer);
}
kfree(dev->vblank);
dev->num_crtcs = 0;
}
EXPORT_SYMBOL(drm_vblank_cleanup);
/**
* drm_vblank_init - initialize vblank support
* @dev: DRM device
* @num_crtcs: number of CRTCs supported by @dev
*
* This function initializes vblank support for @num_crtcs display pipelines.
*
* Returns:
* Zero on success or a negative error code on failure.
*/
int drm_vblank_init(struct drm_device *dev, unsigned int num_crtcs)
{
int ret = -ENOMEM;
unsigned int i;
spin_lock_init(&dev->vbl_lock);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
spin_lock_init(&dev->vblank_time_lock);
dev->num_crtcs = num_crtcs;
dev->vblank = kcalloc(num_crtcs, sizeof(*dev->vblank), GFP_KERNEL);
if (!dev->vblank)
goto err;
for (i = 0; i < num_crtcs; i++) {
struct drm_vblank_crtc *vblank = &dev->vblank[i];
vblank->dev = dev;
vblank->pipe = i;
init_waitqueue_head(&vblank->queue);
setup_timer(&vblank->disable_timer, vblank_disable_fn,
(unsigned long)vblank);
}
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
drm: Push latency sensitive bits of vblank scanoutpos timestamping into kms drivers. A change in locking of some kms drivers (currently intel-kms) make the old approach too inaccurate and also incompatible with the PREEMPT_RT realtime kernel patchset. The driver->get_scanout_position() method of intel-kms now needs to aquire a spinlock, which clashes badly with the former preempt_disable() calls in the drm, and it also introduces larger delays and timing uncertainty on a contended lock than acceptable. This patch changes the prototype of driver->get_scanout_position() to require/allow kms drivers to perform the ktime_get() system time queries which go along with actual scanout position readout in a way that provides maximum precision and to return those timestamps to the drm. kms drivers implementations of get_scanout_position() are asked to implement timestamping and scanoutpos readout in a way that is as precise as possible and compatible with preempt_disable() on a PREMPT_RT kernel. A driver should follow this pattern in get_scanout_position() for precision and compatibility: spin_lock...(...); preempt_disable_rt(); // On a PREEMPT_RT kernel, otherwise omit. if (stime) *stime = ktime_get(); ... Minimum amount of MMIO register reads to get scanout position ... ... no taking of locks allowed here! ... if (etime) *etime = ktime_get(); preempt_enable_rt(); // On PREEMPT_RT kernel, otherwise omit. spin_unlock...(...); v2: Fix formatting of new multi-line code comments. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-10-30 12:13:06 +08:00
DRM_INFO("Supports vblank timestamp caching Rev 2 (21.10.2013).\n");
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Driver specific high-precision vblank timestamping supported? */
if (dev->driver->get_vblank_timestamp)
DRM_INFO("Driver supports precise vblank timestamp query.\n");
else
DRM_INFO("No driver support for vblank timestamp query.\n");
/* Must have precise timestamping for reliable vblank instant disable */
if (dev->vblank_disable_immediate && !dev->driver->get_vblank_timestamp) {
dev->vblank_disable_immediate = false;
DRM_INFO("Setting vblank_disable_immediate to false because "
"get_vblank_timestamp == NULL\n");
}
dev->vblank_disable_allowed = false;
return 0;
err:
dev->num_crtcs = 0;
return ret;
}
EXPORT_SYMBOL(drm_vblank_init);
static void drm_irq_vgaarb_nokms(void *cookie, bool state)
{
struct drm_device *dev = cookie;
if (dev->driver->vgaarb_irq) {
dev->driver->vgaarb_irq(dev, state);
return;
}
if (!dev->irq_enabled)
return;
if (state) {
if (dev->driver->irq_uninstall)
dev->driver->irq_uninstall(dev);
} else {
if (dev->driver->irq_preinstall)
dev->driver->irq_preinstall(dev);
if (dev->driver->irq_postinstall)
dev->driver->irq_postinstall(dev);
}
}
/**
* drm_irq_install - install IRQ handler
* @dev: DRM device
* @irq: IRQ number to install the handler for
*
* Initializes the IRQ related data. Installs the handler, calling the driver
* irq_preinstall() and irq_postinstall() functions before and after the
* installation.
*
* This is the simplified helper interface provided for drivers with no special
* needs. Drivers which need to install interrupt handlers for multiple
* interrupts must instead set drm_device->irq_enabled to signal the DRM core
* that vblank interrupts are available.
*
* Returns:
* Zero on success or a negative error code on failure.
*/
int drm_irq_install(struct drm_device *dev, int irq)
{
int ret;
unsigned long sh_flags = 0;
if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ))
return -EINVAL;
if (irq == 0)
return -EINVAL;
/* Driver must have been initialized */
if (!dev->dev_private)
return -EINVAL;
if (dev->irq_enabled)
return -EBUSY;
dev->irq_enabled = true;
DRM_DEBUG("irq=%d\n", irq);
/* Before installing handler */
if (dev->driver->irq_preinstall)
dev->driver->irq_preinstall(dev);
/* Install handler */
if (drm_core_check_feature(dev, DRIVER_IRQ_SHARED))
sh_flags = IRQF_SHARED;
ret = request_irq(irq, dev->driver->irq_handler,
sh_flags, dev->driver->name, dev);
if (ret < 0) {
dev->irq_enabled = false;
return ret;
}
if (!drm_core_check_feature(dev, DRIVER_MODESET))
vga_client_register(dev->pdev, (void *)dev, drm_irq_vgaarb_nokms, NULL);
/* After installing handler */
if (dev->driver->irq_postinstall)
ret = dev->driver->irq_postinstall(dev);
if (ret < 0) {
dev->irq_enabled = false;
if (!drm_core_check_feature(dev, DRIVER_MODESET))
vga_client_register(dev->pdev, NULL, NULL, NULL);
free_irq(irq, dev);
} else {
dev->irq = irq;
}
return ret;
}
EXPORT_SYMBOL(drm_irq_install);
/**
* drm_irq_uninstall - uninstall the IRQ handler
* @dev: DRM device
*
* Calls the driver's irq_uninstall() function and unregisters the IRQ handler.
* This should only be called by drivers which used drm_irq_install() to set up
* their interrupt handler. Other drivers must only reset
* drm_device->irq_enabled to false.
*
* Note that for kernel modesetting drivers it is a bug if this function fails.
* The sanity checks are only to catch buggy user modesetting drivers which call
* the same function through an ioctl.
*
* Returns:
* Zero on success or a negative error code on failure.
*/
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
int drm_irq_uninstall(struct drm_device *dev)
{
unsigned long irqflags;
bool irq_enabled;
int i;
if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ))
return -EINVAL;
irq_enabled = dev->irq_enabled;
dev->irq_enabled = false;
/*
drm/irq: Don't call ->get_vblank_counter directly from irq_uninstall/cleanup The pipe might already have been shut down, and then it's not a good idea to call hw accessor functions. Instead use the same logic as drm_vblank_off which has all the necessary checks to avoid troubles or inconsistency. Noticed by Imre while reviewing my patches to remove some sanity checks from ->get_vblank_counter. v2: Try harder. disable_and_save can still access the vblank stuff when vblank->enabled isn't set. It has to, since vlbank irq could be disable but the pipe is still on when being called from drm_vblank_off. But we still want to use that code for more code sharing. So add a check for vblank->enabled on top - if that's not set we shouldn't have anyone waiting for the vblank. If we have that's a pretty serious bug. The other issue that Imre spotted is drm_vblank_cleanup. That code again calls disable_and_save and so suffers from the same issues. But really drm_irq_uninstall should have cleaned that all up, so replace the code with WARN_ON. Note that we can't delete the timer cleanup since drivers aren't required to use drm_irq_install/uninstall, but can do their own irq handling. v3: Make it clear that all that gunk in drm_irq_uninstall is really just bandaids for UMS races between the irq/vblank code. In UMS userspace is in control of enabling/disabling interrupts in general and vblanks specifically. v4: Imre observed that KMS drivers all call drm_vblank_cleanup before drm_irq_uninstall (as they should), so again the code in there is dead for KMS (due to dev->num_crtcs == 0 after drm_vblank_cleanup). Or should be, so only WARN for KMS - with UMS userspace could try to do evil things. v5: After more discussion on irc we've gone back to v3: the del_timer_sync is required in all cases in drm_vblank_cleanup, but let's restrict the WARN_ON to kms drivers only. Imre was also concerned that bad things could happen without the disable_and_save call. But we immediately free vblank structures afterwards which makes the save useless. And drm_handle_vblank has a check for dev->num_crtcs to avoid surprises with ums. Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> Acked-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-22 22:11:19 +08:00
* Wake up any waiters so they don't hang. This is just to paper over
* isssues for UMS drivers which aren't in full control of their
* vblank/irq handling. KMS drivers must ensure that vblanks are all
* disabled when uninstalling the irq handler.
*/
if (dev->num_crtcs) {
spin_lock_irqsave(&dev->vbl_lock, irqflags);
for (i = 0; i < dev->num_crtcs; i++) {
struct drm_vblank_crtc *vblank = &dev->vblank[i];
drm/irq: Don't call ->get_vblank_counter directly from irq_uninstall/cleanup The pipe might already have been shut down, and then it's not a good idea to call hw accessor functions. Instead use the same logic as drm_vblank_off which has all the necessary checks to avoid troubles or inconsistency. Noticed by Imre while reviewing my patches to remove some sanity checks from ->get_vblank_counter. v2: Try harder. disable_and_save can still access the vblank stuff when vblank->enabled isn't set. It has to, since vlbank irq could be disable but the pipe is still on when being called from drm_vblank_off. But we still want to use that code for more code sharing. So add a check for vblank->enabled on top - if that's not set we shouldn't have anyone waiting for the vblank. If we have that's a pretty serious bug. The other issue that Imre spotted is drm_vblank_cleanup. That code again calls disable_and_save and so suffers from the same issues. But really drm_irq_uninstall should have cleaned that all up, so replace the code with WARN_ON. Note that we can't delete the timer cleanup since drivers aren't required to use drm_irq_install/uninstall, but can do their own irq handling. v3: Make it clear that all that gunk in drm_irq_uninstall is really just bandaids for UMS races between the irq/vblank code. In UMS userspace is in control of enabling/disabling interrupts in general and vblanks specifically. v4: Imre observed that KMS drivers all call drm_vblank_cleanup before drm_irq_uninstall (as they should), so again the code in there is dead for KMS (due to dev->num_crtcs == 0 after drm_vblank_cleanup). Or should be, so only WARN for KMS - with UMS userspace could try to do evil things. v5: After more discussion on irc we've gone back to v3: the del_timer_sync is required in all cases in drm_vblank_cleanup, but let's restrict the WARN_ON to kms drivers only. Imre was also concerned that bad things could happen without the disable_and_save call. But we immediately free vblank structures afterwards which makes the save useless. And drm_handle_vblank has a check for dev->num_crtcs to avoid surprises with ums. Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> Acked-by: Dave Airlie <airlied@redhat.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-22 22:11:19 +08:00
if (!vblank->enabled)
continue;
WARN_ON(drm_core_check_feature(dev, DRIVER_MODESET));
vblank_disable_and_save(dev, i);
wake_up(&vblank->queue);
}
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
}
if (!irq_enabled)
return -EINVAL;
DRM_DEBUG("irq=%d\n", dev->irq);
if (!drm_core_check_feature(dev, DRIVER_MODESET))
vga_client_register(dev->pdev, NULL, NULL, NULL);
if (dev->driver->irq_uninstall)
dev->driver->irq_uninstall(dev);
free_irq(dev->irq, dev);
return 0;
}
EXPORT_SYMBOL(drm_irq_uninstall);
/*
* IRQ control ioctl.
*
* \param inode device inode.
* \param file_priv DRM file private.
* \param cmd command.
* \param arg user argument, pointing to a drm_control structure.
* \return zero on success or a negative number on failure.
*
* Calls irq_install() or irq_uninstall() according to \p arg.
*/
int drm_control(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
struct drm_control *ctl = data;
int ret = 0, irq;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* if we haven't irq we fallback for compatibility reasons -
* this used to be a separate function in drm_dma.h
*/
if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ))
return 0;
if (drm_core_check_feature(dev, DRIVER_MODESET))
return 0;
/* UMS was only ever support on pci devices. */
if (WARN_ON(!dev->pdev))
return -EINVAL;
switch (ctl->func) {
case DRM_INST_HANDLER:
irq = dev->pdev->irq;
if (dev->if_version < DRM_IF_VERSION(1, 2) &&
ctl->irq != irq)
return -EINVAL;
mutex_lock(&dev->struct_mutex);
ret = drm_irq_install(dev, irq);
mutex_unlock(&dev->struct_mutex);
return ret;
case DRM_UNINST_HANDLER:
mutex_lock(&dev->struct_mutex);
ret = drm_irq_uninstall(dev);
mutex_unlock(&dev->struct_mutex);
return ret;
default:
return -EINVAL;
}
}
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/**
* drm_calc_timestamping_constants - calculate vblank timestamp constants
* @crtc: drm_crtc whose timestamp constants should be updated.
* @mode: display mode containing the scanout timings
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*
* Calculate and store various constants which are later
* needed by vblank and swap-completion timestamping, e.g,
* by drm_calc_vbltimestamp_from_scanoutpos(). They are
* derived from CRTC's true scanout timing, so they take
* things like panel scaling or other adjustments into account.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*/
void drm_calc_timestamping_constants(struct drm_crtc *crtc,
const struct drm_display_mode *mode)
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
{
struct drm_vblank_crtc *vblank = &crtc->dev->vblank[drm_crtc_index(crtc)];
int linedur_ns = 0, pixeldur_ns = 0, framedur_ns = 0;
int dotclock = mode->crtc_clock;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Valid dotclock? */
if (dotclock > 0) {
int frame_size = mode->crtc_htotal * mode->crtc_vtotal;
/*
* Convert scanline length in pixels and video
* dot clock to line duration, frame duration
* and pixel duration in nanoseconds:
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*/
pixeldur_ns = 1000000 / dotclock;
linedur_ns = div_u64((u64) mode->crtc_htotal * 1000000, dotclock);
framedur_ns = div_u64((u64) frame_size * 1000000, dotclock);
/*
* Fields of interlaced scanout modes are only half a frame duration.
*/
if (mode->flags & DRM_MODE_FLAG_INTERLACE)
framedur_ns /= 2;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
} else
DRM_ERROR("crtc %u: Can't calculate constants, dotclock = 0!\n",
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
crtc->base.id);
vblank->pixeldur_ns = pixeldur_ns;
vblank->linedur_ns = linedur_ns;
vblank->framedur_ns = framedur_ns;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
DRM_DEBUG("crtc %u: hwmode: htotal %d, vtotal %d, vdisplay %d\n",
crtc->base.id, mode->crtc_htotal,
mode->crtc_vtotal, mode->crtc_vdisplay);
DRM_DEBUG("crtc %u: clock %d kHz framedur %d linedur %d, pixeldur %d\n",
crtc->base.id, dotclock, framedur_ns,
linedur_ns, pixeldur_ns);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
}
EXPORT_SYMBOL(drm_calc_timestamping_constants);
/**
* drm_calc_vbltimestamp_from_scanoutpos - precise vblank timestamp helper
* @dev: DRM device
* @pipe: index of CRTC whose vblank timestamp to retrieve
* @max_error: Desired maximum allowable error in timestamps (nanosecs)
* On return contains true maximum error of timestamp
* @vblank_time: Pointer to struct timeval which should receive the timestamp
* @flags: Flags to pass to driver:
* 0 = Default,
* DRM_CALLED_FROM_VBLIRQ = If function is called from vbl IRQ handler
* @mode: mode which defines the scanout timings
*
* Implements calculation of exact vblank timestamps from given drm_display_mode
* timings and current video scanout position of a CRTC. This can be called from
* within get_vblank_timestamp() implementation of a kms driver to implement the
* actual timestamping.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*
* Should return timestamps conforming to the OML_sync_control OpenML
* extension specification. The timestamp corresponds to the end of
* the vblank interval, aka start of scanout of topmost-leftmost display
* pixel in the following video frame.
*
* Requires support for optional dev->driver->get_scanout_position()
* in kms driver, plus a bit of setup code to provide a drm_display_mode
* that corresponds to the true scanout timing.
*
* The current implementation only handles standard video modes. It
* returns as no operation if a doublescan or interlaced video mode is
* active. Higher level code is expected to handle this.
*
* Returns:
* Negative value on error, failure or if not supported in current
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* video mode:
*
* -EINVAL - Invalid CRTC.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* -EAGAIN - Temporary unavailable, e.g., called before initial modeset.
* -ENOTSUPP - Function not supported in current display mode.
* -EIO - Failed, e.g., due to failed scanout position query.
*
* Returns or'ed positive status flags on success:
*
* DRM_VBLANKTIME_SCANOUTPOS_METHOD - Signal this method used for timestamping.
* DRM_VBLANKTIME_INVBL - Timestamp taken while scanout was in vblank interval.
*
*/
int drm_calc_vbltimestamp_from_scanoutpos(struct drm_device *dev,
unsigned int pipe,
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
int *max_error,
struct timeval *vblank_time,
unsigned flags,
const struct drm_display_mode *mode)
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
{
struct timeval tv_etime;
ktime_t stime, etime;
int vbl_status;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
int vpos, hpos, i;
int delta_ns, duration_ns;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
bool invbl;
if (pipe >= dev->num_crtcs) {
DRM_ERROR("Invalid crtc %u\n", pipe);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
return -EINVAL;
}
/* Scanout position query not supported? Should not happen. */
if (!dev->driver->get_scanout_position) {
DRM_ERROR("Called from driver w/o get_scanout_position()!?\n");
return -EIO;
}
/* If mode timing undefined, just return as no-op:
* Happens during initial modesetting of a crtc.
*/
if (mode->crtc_clock == 0) {
DRM_DEBUG("crtc %u: Noop due to uninitialized mode.\n", pipe);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
return -EAGAIN;
}
/* Get current scanout position with system timestamp.
* Repeat query up to DRM_TIMESTAMP_MAXRETRIES times
* if single query takes longer than max_error nanoseconds.
*
* This guarantees a tight bound on maximum error if
* code gets preempted or delayed for some reason.
*/
for (i = 0; i < DRM_TIMESTAMP_MAXRETRIES; i++) {
drm: Push latency sensitive bits of vblank scanoutpos timestamping into kms drivers. A change in locking of some kms drivers (currently intel-kms) make the old approach too inaccurate and also incompatible with the PREEMPT_RT realtime kernel patchset. The driver->get_scanout_position() method of intel-kms now needs to aquire a spinlock, which clashes badly with the former preempt_disable() calls in the drm, and it also introduces larger delays and timing uncertainty on a contended lock than acceptable. This patch changes the prototype of driver->get_scanout_position() to require/allow kms drivers to perform the ktime_get() system time queries which go along with actual scanout position readout in a way that provides maximum precision and to return those timestamps to the drm. kms drivers implementations of get_scanout_position() are asked to implement timestamping and scanoutpos readout in a way that is as precise as possible and compatible with preempt_disable() on a PREMPT_RT kernel. A driver should follow this pattern in get_scanout_position() for precision and compatibility: spin_lock...(...); preempt_disable_rt(); // On a PREEMPT_RT kernel, otherwise omit. if (stime) *stime = ktime_get(); ... Minimum amount of MMIO register reads to get scanout position ... ... no taking of locks allowed here! ... if (etime) *etime = ktime_get(); preempt_enable_rt(); // On PREEMPT_RT kernel, otherwise omit. spin_unlock...(...); v2: Fix formatting of new multi-line code comments. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-10-30 12:13:06 +08:00
/*
* Get vertical and horizontal scanout position vpos, hpos,
* and bounding timestamps stime, etime, pre/post query.
*/
vbl_status = dev->driver->get_scanout_position(dev, pipe, flags,
&vpos, &hpos,
&stime, &etime,
mode);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Return as no-op if scanout query unsupported or failed. */
if (!(vbl_status & DRM_SCANOUTPOS_VALID)) {
DRM_DEBUG("crtc %u : scanoutpos query failed [%d].\n",
pipe, vbl_status);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
return -EIO;
}
drm: Push latency sensitive bits of vblank scanoutpos timestamping into kms drivers. A change in locking of some kms drivers (currently intel-kms) make the old approach too inaccurate and also incompatible with the PREEMPT_RT realtime kernel patchset. The driver->get_scanout_position() method of intel-kms now needs to aquire a spinlock, which clashes badly with the former preempt_disable() calls in the drm, and it also introduces larger delays and timing uncertainty on a contended lock than acceptable. This patch changes the prototype of driver->get_scanout_position() to require/allow kms drivers to perform the ktime_get() system time queries which go along with actual scanout position readout in a way that provides maximum precision and to return those timestamps to the drm. kms drivers implementations of get_scanout_position() are asked to implement timestamping and scanoutpos readout in a way that is as precise as possible and compatible with preempt_disable() on a PREMPT_RT kernel. A driver should follow this pattern in get_scanout_position() for precision and compatibility: spin_lock...(...); preempt_disable_rt(); // On a PREEMPT_RT kernel, otherwise omit. if (stime) *stime = ktime_get(); ... Minimum amount of MMIO register reads to get scanout position ... ... no taking of locks allowed here! ... if (etime) *etime = ktime_get(); preempt_enable_rt(); // On PREEMPT_RT kernel, otherwise omit. spin_unlock...(...); v2: Fix formatting of new multi-line code comments. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2013-10-30 12:13:06 +08:00
/* Compute uncertainty in timestamp of scanout position query. */
duration_ns = ktime_to_ns(etime) - ktime_to_ns(stime);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Accept result with < max_error nsecs timing uncertainty. */
if (duration_ns <= *max_error)
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
break;
}
/* Noisy system timing? */
if (i == DRM_TIMESTAMP_MAXRETRIES) {
DRM_DEBUG("crtc %u: Noisy timestamp %d us > %d us [%d reps].\n",
pipe, duration_ns/1000, *max_error/1000, i);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
}
/* Return upper bound of timestamp precision error. */
*max_error = duration_ns;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Check if in vblank area:
* vpos is >=0 in video scanout area, but negative
* within vblank area, counting down the number of lines until
* start of scanout.
*/
invbl = vbl_status & DRM_SCANOUTPOS_IN_VBLANK;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Convert scanout position into elapsed time at raw_time query
* since start of scanout at first display scanline. delta_ns
* can be negative if start of scanout hasn't happened yet.
*/
delta_ns = div_s64(1000000LL * (vpos * mode->crtc_htotal + hpos),
mode->crtc_clock);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
if (!drm_timestamp_monotonic)
etime = ktime_mono_to_real(etime);
/* save this only for debugging purposes */
tv_etime = ktime_to_timeval(etime);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Subtract time delta from raw timestamp to get final
* vblank_time timestamp for end of vblank.
*/
if (delta_ns < 0)
etime = ktime_add_ns(etime, -delta_ns);
else
etime = ktime_sub_ns(etime, delta_ns);
*vblank_time = ktime_to_timeval(etime);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
DRM_DEBUG("crtc %u : v %d p(%d,%d)@ %ld.%ld -> %ld.%ld [e %d us, %d rep]\n",
pipe, (int)vbl_status, hpos, vpos,
(long)tv_etime.tv_sec, (long)tv_etime.tv_usec,
(long)vblank_time->tv_sec, (long)vblank_time->tv_usec,
duration_ns/1000, i);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
vbl_status = DRM_VBLANKTIME_SCANOUTPOS_METHOD;
if (invbl)
vbl_status |= DRM_VBLANKTIME_IN_VBLANK;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
return vbl_status;
}
EXPORT_SYMBOL(drm_calc_vbltimestamp_from_scanoutpos);
static struct timeval get_drm_timestamp(void)
{
ktime_t now;
now = drm_timestamp_monotonic ? ktime_get() : ktime_get_real();
return ktime_to_timeval(now);
}
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/**
* drm_get_last_vbltimestamp - retrieve raw timestamp for the most recent
* vblank interval
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* @dev: DRM device
* @pipe: index of CRTC whose vblank timestamp to retrieve
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* @tvblank: Pointer to target struct timeval which should receive the timestamp
* @flags: Flags to pass to driver:
* 0 = Default,
* DRM_CALLED_FROM_VBLIRQ = If function is called from vbl IRQ handler
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*
* Fetches the system timestamp corresponding to the time of the most recent
* vblank interval on specified CRTC. May call into kms-driver to
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* compute the timestamp with a high-precision GPU specific method.
*
* Returns zero if timestamp originates from uncorrected do_gettimeofday()
* call, i.e., it isn't very precisely locked to the true vblank.
*
* Returns:
* True if timestamp is considered to be very precise, false otherwise.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*/
static bool
drm_get_last_vbltimestamp(struct drm_device *dev, unsigned int pipe,
struct timeval *tvblank, unsigned flags)
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
{
int ret;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Define requested maximum error on timestamps (nanoseconds). */
int max_error = (int) drm_timestamp_precision * 1000;
/* Query driver if possible and precision timestamping enabled. */
if (dev->driver->get_vblank_timestamp && (max_error > 0)) {
ret = dev->driver->get_vblank_timestamp(dev, pipe, &max_error,
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
tvblank, flags);
if (ret > 0)
return true;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
}
/* GPU high precision timestamp query unsupported or failed.
* Return current monotonic/gettimeofday timestamp as best estimate.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*/
*tvblank = get_drm_timestamp();
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
return false;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
}
/**
* drm_vblank_count - retrieve "cooked" vblank counter value
* @dev: DRM device
* @pipe: index of CRTC for which to retrieve the counter
*
* Fetches the "cooked" vblank count value that represents the number of
* vblank events since the system was booted, including lost events due to
* modesetting activity.
*
* This is the legacy version of drm_crtc_vblank_count().
*
* Returns:
* The software vblank counter.
*/
u32 drm_vblank_count(struct drm_device *dev, int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
if (WARN_ON(pipe >= dev->num_crtcs))
return 0;
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
return vblank->count;
}
EXPORT_SYMBOL(drm_vblank_count);
/**
* drm_crtc_vblank_count - retrieve "cooked" vblank counter value
* @crtc: which counter to retrieve
*
* Fetches the "cooked" vblank count value that represents the number of
* vblank events since the system was booted, including lost events due to
* modesetting activity.
*
* This is the native KMS version of drm_vblank_count().
*
* Returns:
* The software vblank counter.
*/
u32 drm_crtc_vblank_count(struct drm_crtc *crtc)
{
return drm_vblank_count(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_vblank_count);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/**
* drm_vblank_count_and_time - retrieve "cooked" vblank counter value and the
* system timestamp corresponding to that vblank counter value.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* @dev: DRM device
* @pipe: index of CRTC whose counter to retrieve
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* @vblanktime: Pointer to struct timeval to receive the vblank timestamp.
*
* Fetches the "cooked" vblank count value that represents the number of
* vblank events since the system was booted, including lost events due to
* modesetting activity. Returns corresponding system timestamp of the time
* of the vblank interval that corresponds to the current vblank counter value.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*/
u32 drm_vblank_count_and_time(struct drm_device *dev, unsigned int pipe,
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
struct timeval *vblanktime)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
u32 cur_vblank;
if (WARN_ON(pipe >= dev->num_crtcs))
return 0;
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
/*
* Vblank timestamps are read lockless. To ensure consistency the vblank
* counter is rechecked and ordering is ensured using memory barriers.
* This works like a seqlock. The write-side barriers are in store_vblank.
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
*/
do {
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
cur_vblank = vblank->count;
smp_rmb();
*vblanktime = vblanktimestamp(dev, pipe, cur_vblank);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
smp_rmb();
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
} while (cur_vblank != vblank->count);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
return cur_vblank;
}
EXPORT_SYMBOL(drm_vblank_count_and_time);
static void send_vblank_event(struct drm_device *dev,
struct drm_pending_vblank_event *e,
unsigned long seq, struct timeval *now)
{
WARN_ON_SMP(!spin_is_locked(&dev->event_lock));
e->event.sequence = seq;
e->event.tv_sec = now->tv_sec;
e->event.tv_usec = now->tv_usec;
list_add_tail(&e->base.link,
&e->base.file_priv->event_list);
wake_up_interruptible(&e->base.file_priv->event_wait);
trace_drm_vblank_event_delivered(e->base.pid, e->pipe,
e->event.sequence);
}
/**
* drm_send_vblank_event - helper to send vblank event after pageflip
* @dev: DRM device
* @pipe: CRTC index
* @e: the event to send
*
* Updates sequence # and timestamp on event, and sends it to userspace.
* Caller must hold event lock.
*
* This is the legacy version of drm_crtc_send_vblank_event().
*/
void drm_send_vblank_event(struct drm_device *dev, unsigned int pipe,
struct drm_pending_vblank_event *e)
{
struct timeval now;
unsigned int seq;
if (dev->num_crtcs > 0) {
seq = drm_vblank_count_and_time(dev, pipe, &now);
} else {
seq = 0;
now = get_drm_timestamp();
}
e->pipe = pipe;
send_vblank_event(dev, e, seq, &now);
}
EXPORT_SYMBOL(drm_send_vblank_event);
/**
* drm_crtc_send_vblank_event - helper to send vblank event after pageflip
* @crtc: the source CRTC of the vblank event
* @e: the event to send
*
* Updates sequence # and timestamp on event, and sends it to userspace.
* Caller must hold event lock.
*
* This is the native KMS version of drm_send_vblank_event().
*/
void drm_crtc_send_vblank_event(struct drm_crtc *crtc,
struct drm_pending_vblank_event *e)
{
drm_send_vblank_event(crtc->dev, drm_crtc_index(crtc), e);
}
EXPORT_SYMBOL(drm_crtc_send_vblank_event);
/**
* drm_vblank_enable - enable the vblank interrupt on a CRTC
* @dev: DRM device
* @pipe: CRTC index
*
* Returns:
* Zero on success or a negative error code on failure.
*/
static int drm_vblank_enable(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
int ret = 0;
assert_spin_locked(&dev->vbl_lock);
spin_lock(&dev->vblank_time_lock);
if (!vblank->enabled) {
/*
* Enable vblank irqs under vblank_time_lock protection.
* All vblank count & timestamp updates are held off
* until we are done reinitializing master counter and
* timestamps. Filtercode in drm_handle_vblank() will
* prevent double-accounting of same vblank interval.
*/
ret = dev->driver->enable_vblank(dev, pipe);
DRM_DEBUG("enabling vblank on crtc %u, ret: %d\n", pipe, ret);
if (ret)
atomic_dec(&vblank->refcount);
else {
vblank->enabled = true;
drm_update_vblank_count(dev, pipe);
}
}
spin_unlock(&dev->vblank_time_lock);
return ret;
}
/**
* drm_vblank_get - get a reference count on vblank events
* @dev: DRM device
* @pipe: index of CRTC to own
*
* Acquire a reference count on vblank events to avoid having them disabled
* while in use.
*
* This is the legacy version of drm_crtc_vblank_get().
*
* Returns:
* Zero on success or a negative error code on failure.
*/
int drm_vblank_get(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
unsigned long irqflags;
int ret = 0;
if (!dev->num_crtcs)
return -EINVAL;
if (WARN_ON(pipe >= dev->num_crtcs))
return -EINVAL;
spin_lock_irqsave(&dev->vbl_lock, irqflags);
/* Going from 0->1 means we have to enable interrupts again */
if (atomic_add_return(1, &vblank->refcount) == 1) {
ret = drm_vblank_enable(dev, pipe);
} else {
if (!vblank->enabled) {
atomic_dec(&vblank->refcount);
ret = -EINVAL;
}
}
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
return ret;
}
EXPORT_SYMBOL(drm_vblank_get);
/**
* drm_crtc_vblank_get - get a reference count on vblank events
* @crtc: which CRTC to own
*
* Acquire a reference count on vblank events to avoid having them disabled
* while in use.
*
* This is the native kms version of drm_vblank_get().
*
* Returns:
* Zero on success or a negative error code on failure.
*/
int drm_crtc_vblank_get(struct drm_crtc *crtc)
{
return drm_vblank_get(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_vblank_get);
/**
* drm_vblank_put - release ownership of vblank events
* @dev: DRM device
* @pipe: index of CRTC to release
*
* Release ownership of a given vblank counter, turning off interrupts
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
* if possible. Disable interrupts after drm_vblank_offdelay milliseconds.
*
* This is the legacy version of drm_crtc_vblank_put().
*/
void drm_vblank_put(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
if (WARN_ON(pipe >= dev->num_crtcs))
return;
if (WARN_ON(atomic_read(&vblank->refcount) == 0))
return;
/* Last user schedules interrupt disable */
if (atomic_dec_and_test(&vblank->refcount)) {
if (drm_vblank_offdelay == 0)
return;
else if (dev->vblank_disable_immediate || drm_vblank_offdelay < 0)
vblank_disable_fn((unsigned long)vblank);
else
mod_timer(&vblank->disable_timer,
jiffies + ((drm_vblank_offdelay * HZ)/1000));
}
}
EXPORT_SYMBOL(drm_vblank_put);
/**
* drm_crtc_vblank_put - give up ownership of vblank events
* @crtc: which counter to give up
*
* Release ownership of a given vblank counter, turning off interrupts
* if possible. Disable interrupts after drm_vblank_offdelay milliseconds.
*
* This is the native kms version of drm_vblank_put().
*/
void drm_crtc_vblank_put(struct drm_crtc *crtc)
{
drm_vblank_put(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_vblank_put);
/**
* drm_wait_one_vblank - wait for one vblank
* @dev: DRM device
* @pipe: CRTC index
*
* This waits for one vblank to pass on @pipe, using the irq driver interfaces.
* It is a failure to call this when the vblank irq for @pipe is disabled, e.g.
* due to lack of driver support or because the crtc is off.
*/
void drm_wait_one_vblank(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
int ret;
u32 last;
if (WARN_ON(pipe >= dev->num_crtcs))
return;
ret = drm_vblank_get(dev, pipe);
if (WARN(ret, "vblank not available on crtc %i, ret=%i\n", pipe, ret))
return;
last = drm_vblank_count(dev, pipe);
ret = wait_event_timeout(vblank->queue,
last != drm_vblank_count(dev, pipe),
msecs_to_jiffies(100));
WARN(ret == 0, "vblank wait timed out on crtc %i\n", pipe);
drm_vblank_put(dev, pipe);
}
EXPORT_SYMBOL(drm_wait_one_vblank);
/**
* drm_crtc_wait_one_vblank - wait for one vblank
* @crtc: DRM crtc
*
* This waits for one vblank to pass on @crtc, using the irq driver interfaces.
* It is a failure to call this when the vblank irq for @crtc is disabled, e.g.
* due to lack of driver support or because the crtc is off.
*/
void drm_crtc_wait_one_vblank(struct drm_crtc *crtc)
{
drm_wait_one_vblank(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_wait_one_vblank);
/**
* drm_vblank_off - disable vblank events on a CRTC
* @dev: DRM device
* @pipe: CRTC index
*
* Drivers can use this function to shut down the vblank interrupt handling when
* disabling a crtc. This function ensures that the latest vblank frame count is
* stored so that drm_vblank_on() can restore it again.
*
* Drivers must use this function when the hardware vblank counter can get
* reset, e.g. when suspending.
*
* This is the legacy version of drm_crtc_vblank_off().
*/
void drm_vblank_off(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
struct drm_pending_vblank_event *e, *t;
struct timeval now;
unsigned long irqflags;
unsigned int seq;
if (WARN_ON(pipe >= dev->num_crtcs))
return;
spin_lock_irqsave(&dev->event_lock, irqflags);
spin_lock(&dev->vbl_lock);
vblank_disable_and_save(dev, pipe);
wake_up(&vblank->queue);
/*
* Prevent subsequent drm_vblank_get() from re-enabling
* the vblank interrupt by bumping the refcount.
*/
if (!vblank->inmodeset) {
atomic_inc(&vblank->refcount);
vblank->inmodeset = 1;
}
spin_unlock(&dev->vbl_lock);
/* Send any queued vblank events, lest the natives grow disquiet */
seq = drm_vblank_count_and_time(dev, pipe, &now);
list_for_each_entry_safe(e, t, &dev->vblank_event_list, base.link) {
if (e->pipe != pipe)
continue;
DRM_DEBUG("Sending premature vblank event on disable: \
wanted %d, current %d\n",
e->event.sequence, seq);
list_del(&e->base.link);
drm_vblank_put(dev, pipe);
send_vblank_event(dev, e, seq, &now);
}
spin_unlock_irqrestore(&dev->event_lock, irqflags);
}
EXPORT_SYMBOL(drm_vblank_off);
/**
* drm_crtc_vblank_off - disable vblank events on a CRTC
* @crtc: CRTC in question
*
* Drivers can use this function to shut down the vblank interrupt handling when
* disabling a crtc. This function ensures that the latest vblank frame count is
* stored so that drm_vblank_on can restore it again.
*
* Drivers must use this function when the hardware vblank counter can get
* reset, e.g. when suspending.
*
* This is the native kms version of drm_vblank_off().
*/
void drm_crtc_vblank_off(struct drm_crtc *crtc)
{
drm_vblank_off(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_vblank_off);
drm/irq: Add drm_crtc_vblank_reset At driver load we need to tell the vblank code about the state of the pipes, so that the logic around reject vblank_get when the pipe is off works correctly. Thus far i915 used drm_vblank_off, but one of the side-effects of it is that it also saves the vblank counter. And for that it calls down into the ->get_vblank_counter hook. Which isn't really a good idea when the pipe is off for a few reasons: - With runtime pm the register might not respond. - If the pipe is off some datastructures might not be around or unitialized. The later is what blew up on gen3: We look at intel_crtc->config to compute the vblank counter, and for a disabled pipe at boot-up that's just not there. Thus far this was papered over by a check for intel_crtc->active, but I want to get rid of that (since it's fairly race, vblank hooks are called from all kinds of places). So prep for that by adding a _reset functions which only does what we really need to be done at driver load: Mark the vblank pipe as off, but don't do any vblank counter saving or event flushing - neither of that is required. v2: Clarify the code flow slightly as suggested by Ville. v3: Fix kerneldoc spelling, spotted by Laurent. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> (v2) Acked-by: Dave Airlie <airlied@redhat.com> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-14 04:03:42 +08:00
/**
* drm_crtc_vblank_reset - reset vblank state to off on a CRTC
* @drm_crtc: CRTC in question
drm/irq: Add drm_crtc_vblank_reset At driver load we need to tell the vblank code about the state of the pipes, so that the logic around reject vblank_get when the pipe is off works correctly. Thus far i915 used drm_vblank_off, but one of the side-effects of it is that it also saves the vblank counter. And for that it calls down into the ->get_vblank_counter hook. Which isn't really a good idea when the pipe is off for a few reasons: - With runtime pm the register might not respond. - If the pipe is off some datastructures might not be around or unitialized. The later is what blew up on gen3: We look at intel_crtc->config to compute the vblank counter, and for a disabled pipe at boot-up that's just not there. Thus far this was papered over by a check for intel_crtc->active, but I want to get rid of that (since it's fairly race, vblank hooks are called from all kinds of places). So prep for that by adding a _reset functions which only does what we really need to be done at driver load: Mark the vblank pipe as off, but don't do any vblank counter saving or event flushing - neither of that is required. v2: Clarify the code flow slightly as suggested by Ville. v3: Fix kerneldoc spelling, spotted by Laurent. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> (v2) Acked-by: Dave Airlie <airlied@redhat.com> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-14 04:03:42 +08:00
*
* Drivers can use this function to reset the vblank state to off at load time.
* Drivers should use this together with the drm_crtc_vblank_off() and
* drm_crtc_vblank_on() functions. The difference compared to
* drm_crtc_vblank_off() is that this function doesn't save the vblank counter
* and hence doesn't need to call any driver hooks.
*/
void drm_crtc_vblank_reset(struct drm_crtc *drm_crtc)
{
struct drm_device *dev = drm_crtc->dev;
unsigned long irqflags;
unsigned int pipe = drm_crtc_index(drm_crtc);
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/irq: Add drm_crtc_vblank_reset At driver load we need to tell the vblank code about the state of the pipes, so that the logic around reject vblank_get when the pipe is off works correctly. Thus far i915 used drm_vblank_off, but one of the side-effects of it is that it also saves the vblank counter. And for that it calls down into the ->get_vblank_counter hook. Which isn't really a good idea when the pipe is off for a few reasons: - With runtime pm the register might not respond. - If the pipe is off some datastructures might not be around or unitialized. The later is what blew up on gen3: We look at intel_crtc->config to compute the vblank counter, and for a disabled pipe at boot-up that's just not there. Thus far this was papered over by a check for intel_crtc->active, but I want to get rid of that (since it's fairly race, vblank hooks are called from all kinds of places). So prep for that by adding a _reset functions which only does what we really need to be done at driver load: Mark the vblank pipe as off, but don't do any vblank counter saving or event flushing - neither of that is required. v2: Clarify the code flow slightly as suggested by Ville. v3: Fix kerneldoc spelling, spotted by Laurent. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Cc: Imre Deak <imre.deak@intel.com> Reviewed-by: Imre Deak <imre.deak@intel.com> (v2) Acked-by: Dave Airlie <airlied@redhat.com> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-02-14 04:03:42 +08:00
spin_lock_irqsave(&dev->vbl_lock, irqflags);
/*
* Prevent subsequent drm_vblank_get() from enabling the vblank
* interrupt by bumping the refcount.
*/
if (!vblank->inmodeset) {
atomic_inc(&vblank->refcount);
vblank->inmodeset = 1;
}
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
WARN_ON(!list_empty(&dev->vblank_event_list));
}
EXPORT_SYMBOL(drm_crtc_vblank_reset);
/**
* drm_vblank_on - enable vblank events on a CRTC
* @dev: DRM device
* @pipe: CRTC index
*
* This functions restores the vblank interrupt state captured with
* drm_vblank_off() again. Note that calls to drm_vblank_on() and
* drm_vblank_off() can be unbalanced and so can also be unconditionally called
* in driver load code to reflect the current hardware state of the crtc.
*
* This is the legacy version of drm_crtc_vblank_on().
*/
void drm_vblank_on(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
unsigned long irqflags;
if (WARN_ON(pipe >= dev->num_crtcs))
return;
spin_lock_irqsave(&dev->vbl_lock, irqflags);
/* Drop our private "prevent drm_vblank_get" refcount */
if (vblank->inmodeset) {
atomic_dec(&vblank->refcount);
vblank->inmodeset = 0;
}
/*
* sample the current counter to avoid random jumps
* when drm_vblank_enable() applies the diff
*
* -1 to make sure user will never see the same
* vblank counter value before and after a modeset
*/
vblank->last =
(dev->driver->get_vblank_counter(dev, pipe) - 1) &
dev->max_vblank_count;
/*
* re-enable interrupts if there are users left, or the
* user wishes vblank interrupts to be enabled all the time.
*/
if (atomic_read(&vblank->refcount) != 0 ||
(!dev->vblank_disable_immediate && drm_vblank_offdelay == 0))
WARN_ON(drm_vblank_enable(dev, pipe));
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
}
EXPORT_SYMBOL(drm_vblank_on);
/**
* drm_crtc_vblank_on - enable vblank events on a CRTC
* @crtc: CRTC in question
*
* This functions restores the vblank interrupt state captured with
* drm_vblank_off() again. Note that calls to drm_vblank_on() and
* drm_vblank_off() can be unbalanced and so can also be unconditionally called
* in driver load code to reflect the current hardware state of the crtc.
*
* This is the native kms version of drm_vblank_on().
*/
void drm_crtc_vblank_on(struct drm_crtc *crtc)
{
drm_vblank_on(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_vblank_on);
/**
* drm_vblank_pre_modeset - account for vblanks across mode sets
* @dev: DRM device
* @pipe: CRTC index
*
* Account for vblank events across mode setting events, which will likely
* reset the hardware frame counter.
*
* This is done by grabbing a temporary vblank reference to ensure that the
* vblank interrupt keeps running across the modeset sequence. With this the
* software-side vblank frame counting will ensure that there are no jumps or
* discontinuities.
*
* Unfortunately this approach is racy and also doesn't work when the vblank
* interrupt stops running, e.g. across system suspend resume. It is therefore
* highly recommended that drivers use the newer drm_vblank_off() and
* drm_vblank_on() instead. drm_vblank_pre_modeset() only works correctly when
* using "cooked" software vblank frame counters and not relying on any hardware
* counters.
*
* Drivers must call drm_vblank_post_modeset() when re-enabling the same crtc
* again.
*/
void drm_vblank_pre_modeset(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm: fix a use-after-free when GPU acceleration disabled When GPU acceleration is disabled, drm_vblank_cleanup() will free the vblank-related data, such as vblank_refcount, vblank_inmodeset, etc. But we found that drm_vblank_post_modeset() may be called after the cleanup, which use vblank_refcount and vblank_inmodeset. And this will cause a kernel panic. Fix this by return immediately if dev->num_crtcs is zero. This is the same thing that drm_vblank_pre_modeset() does. Call trace of a drm_vblank_post_modeset() after drm_vblank_cleanup(): [ 62.628906] [<ffffffff804868d0>] drm_vblank_post_modeset+0x34/0xb4 [ 62.628906] [<ffffffff804c7008>] atombios_crtc_dpms+0xb4/0x174 [ 62.628906] [<ffffffff804c70e0>] atombios_crtc_commit+0x18/0x38 [ 62.628906] [<ffffffff8047f038>] drm_crtc_helper_set_mode+0x304/0x3cc [ 62.628906] [<ffffffff8047f92c>] drm_crtc_helper_set_config+0x6d8/0x988 [ 62.628906] [<ffffffff8047dd40>] drm_fb_helper_set_par+0x94/0x104 [ 62.628906] [<ffffffff80439d14>] fbcon_init+0x424/0x57c [ 62.628906] [<ffffffff8046a638>] visual_init+0xb8/0x118 [ 62.628906] [<ffffffff8046b9f8>] take_over_console+0x238/0x384 [ 62.628906] [<ffffffff80436df8>] fbcon_takeover+0x7c/0xdc [ 62.628906] [<ffffffff8024fa20>] notifier_call_chain+0x44/0x94 [ 62.628906] [<ffffffff8024fcbc>] __blocking_notifier_call_chain+0x48/0x68 [ 62.628906] [<ffffffff8042d990>] register_framebuffer+0x228/0x260 [ 62.628906] [<ffffffff8047e010>] drm_fb_helper_single_fb_probe+0x260/0x314 [ 62.628906] [<ffffffff8047e2c4>] drm_fb_helper_initial_config+0x200/0x234 [ 62.628906] [<ffffffff804e5560>] radeon_fbdev_init+0xd4/0xf4 [ 62.628906] [<ffffffff804e0e08>] radeon_modeset_init+0x9bc/0xa18 [ 62.628906] [<ffffffff804bfc14>] radeon_driver_load_kms+0xdc/0x12c [ 62.628906] [<ffffffff8048b548>] drm_get_pci_dev+0x148/0x238 [ 62.628906] [<ffffffff80423564>] local_pci_probe+0x5c/0xd0 [ 62.628906] [<ffffffff80241ac4>] work_for_cpu_fn+0x1c/0x30 [ 62.628906] [<ffffffff802427c8>] process_one_work+0x274/0x3bc [ 62.628906] [<ffffffff80242934>] process_scheduled_works+0x24/0x44 [ 62.628906] [<ffffffff8024515c>] worker_thread+0x31c/0x3f4 [ 62.628906] [<ffffffff802497a8>] kthread+0x88/0x90 [ 62.628906] [<ffffffff80206794>] kernel_thread_helper+0x10/0x18 Signed-off-by: Huacai Chen <chenhc@lemote.com> Signed-off-by: Binbin Zhou <zhoubb@lemote.com> Cc: <stable@vger.kernel.org> Reviewed-by: Michel Dänzer <michel.daenzer@amd.com> Acked-by: Paul Menzel <paulepanter@users.sourceforge.net> Signed-off-by: Dave Airlie <airlied@gmail.com>
2013-05-21 14:23:43 +08:00
/* vblank is not initialized (IRQ not installed ?), or has been freed */
if (!dev->num_crtcs)
return;
if (WARN_ON(pipe >= dev->num_crtcs))
return;
/*
* To avoid all the problems that might happen if interrupts
* were enabled/disabled around or between these calls, we just
* have the kernel take a reference on the CRTC (just once though
* to avoid corrupting the count if multiple, mismatch calls occur),
* so that interrupts remain enabled in the interim.
*/
if (!vblank->inmodeset) {
vblank->inmodeset = 0x1;
if (drm_vblank_get(dev, pipe) == 0)
vblank->inmodeset |= 0x2;
}
}
EXPORT_SYMBOL(drm_vblank_pre_modeset);
/**
* drm_vblank_post_modeset - undo drm_vblank_pre_modeset changes
* @dev: DRM device
* @pipe: CRTC index
*
* This function again drops the temporary vblank reference acquired in
* drm_vblank_pre_modeset.
*/
void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
unsigned long irqflags;
drm: fix a use-after-free when GPU acceleration disabled When GPU acceleration is disabled, drm_vblank_cleanup() will free the vblank-related data, such as vblank_refcount, vblank_inmodeset, etc. But we found that drm_vblank_post_modeset() may be called after the cleanup, which use vblank_refcount and vblank_inmodeset. And this will cause a kernel panic. Fix this by return immediately if dev->num_crtcs is zero. This is the same thing that drm_vblank_pre_modeset() does. Call trace of a drm_vblank_post_modeset() after drm_vblank_cleanup(): [ 62.628906] [<ffffffff804868d0>] drm_vblank_post_modeset+0x34/0xb4 [ 62.628906] [<ffffffff804c7008>] atombios_crtc_dpms+0xb4/0x174 [ 62.628906] [<ffffffff804c70e0>] atombios_crtc_commit+0x18/0x38 [ 62.628906] [<ffffffff8047f038>] drm_crtc_helper_set_mode+0x304/0x3cc [ 62.628906] [<ffffffff8047f92c>] drm_crtc_helper_set_config+0x6d8/0x988 [ 62.628906] [<ffffffff8047dd40>] drm_fb_helper_set_par+0x94/0x104 [ 62.628906] [<ffffffff80439d14>] fbcon_init+0x424/0x57c [ 62.628906] [<ffffffff8046a638>] visual_init+0xb8/0x118 [ 62.628906] [<ffffffff8046b9f8>] take_over_console+0x238/0x384 [ 62.628906] [<ffffffff80436df8>] fbcon_takeover+0x7c/0xdc [ 62.628906] [<ffffffff8024fa20>] notifier_call_chain+0x44/0x94 [ 62.628906] [<ffffffff8024fcbc>] __blocking_notifier_call_chain+0x48/0x68 [ 62.628906] [<ffffffff8042d990>] register_framebuffer+0x228/0x260 [ 62.628906] [<ffffffff8047e010>] drm_fb_helper_single_fb_probe+0x260/0x314 [ 62.628906] [<ffffffff8047e2c4>] drm_fb_helper_initial_config+0x200/0x234 [ 62.628906] [<ffffffff804e5560>] radeon_fbdev_init+0xd4/0xf4 [ 62.628906] [<ffffffff804e0e08>] radeon_modeset_init+0x9bc/0xa18 [ 62.628906] [<ffffffff804bfc14>] radeon_driver_load_kms+0xdc/0x12c [ 62.628906] [<ffffffff8048b548>] drm_get_pci_dev+0x148/0x238 [ 62.628906] [<ffffffff80423564>] local_pci_probe+0x5c/0xd0 [ 62.628906] [<ffffffff80241ac4>] work_for_cpu_fn+0x1c/0x30 [ 62.628906] [<ffffffff802427c8>] process_one_work+0x274/0x3bc [ 62.628906] [<ffffffff80242934>] process_scheduled_works+0x24/0x44 [ 62.628906] [<ffffffff8024515c>] worker_thread+0x31c/0x3f4 [ 62.628906] [<ffffffff802497a8>] kthread+0x88/0x90 [ 62.628906] [<ffffffff80206794>] kernel_thread_helper+0x10/0x18 Signed-off-by: Huacai Chen <chenhc@lemote.com> Signed-off-by: Binbin Zhou <zhoubb@lemote.com> Cc: <stable@vger.kernel.org> Reviewed-by: Michel Dänzer <michel.daenzer@amd.com> Acked-by: Paul Menzel <paulepanter@users.sourceforge.net> Signed-off-by: Dave Airlie <airlied@gmail.com>
2013-05-21 14:23:43 +08:00
/* vblank is not initialized (IRQ not installed ?), or has been freed */
if (!dev->num_crtcs)
return;
if (WARN_ON(pipe >= dev->num_crtcs))
return;
if (vblank->inmodeset) {
spin_lock_irqsave(&dev->vbl_lock, irqflags);
dev->vblank_disable_allowed = true;
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
if (vblank->inmodeset & 0x2)
drm_vblank_put(dev, pipe);
vblank->inmodeset = 0;
}
}
EXPORT_SYMBOL(drm_vblank_post_modeset);
/*
* drm_modeset_ctl - handle vblank event counter changes across mode switch
* @DRM_IOCTL_ARGS: standard ioctl arguments
*
* Applications should call the %_DRM_PRE_MODESET and %_DRM_POST_MODESET
* ioctls around modesetting so that any lost vblank events are accounted for.
*
* Generally the counter will reset across mode sets. If interrupts are
* enabled around this call, we don't have to do anything since the counter
* will have already been incremented.
*/
int drm_modeset_ctl(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
struct drm_modeset_ctl *modeset = data;
unsigned int pipe;
/* If drm_vblank_init() hasn't been called yet, just no-op */
if (!dev->num_crtcs)
return 0;
/* KMS drivers handle this internally */
if (drm_core_check_feature(dev, DRIVER_MODESET))
return 0;
pipe = modeset->crtc;
if (pipe >= dev->num_crtcs)
return -EINVAL;
switch (modeset->cmd) {
case _DRM_PRE_MODESET:
drm_vblank_pre_modeset(dev, pipe);
break;
case _DRM_POST_MODESET:
drm_vblank_post_modeset(dev, pipe);
break;
default:
return -EINVAL;
}
return 0;
}
static int drm_queue_vblank_event(struct drm_device *dev, unsigned int pipe,
union drm_wait_vblank *vblwait,
struct drm_file *file_priv)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
struct drm_pending_vblank_event *e;
struct timeval now;
unsigned long flags;
unsigned int seq;
int ret;
e = kzalloc(sizeof(*e), GFP_KERNEL);
if (e == NULL) {
ret = -ENOMEM;
goto err_put;
}
e->pipe = pipe;
e->base.pid = current->pid;
e->event.base.type = DRM_EVENT_VBLANK;
e->event.base.length = sizeof(e->event);
e->event.user_data = vblwait->request.signal;
e->base.event = &e->event.base;
e->base.file_priv = file_priv;
e->base.destroy = (void (*) (struct drm_pending_event *)) kfree;
spin_lock_irqsave(&dev->event_lock, flags);
/*
* drm_vblank_off() might have been called after we called
* drm_vblank_get(). drm_vblank_off() holds event_lock
* around the vblank disable, so no need for further locking.
* The reference from drm_vblank_get() protects against
* vblank disable from another source.
*/
if (!vblank->enabled) {
ret = -EINVAL;
goto err_unlock;
}
if (file_priv->event_space < sizeof(e->event)) {
ret = -EBUSY;
goto err_unlock;
}
file_priv->event_space -= sizeof(e->event);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
seq = drm_vblank_count_and_time(dev, pipe, &now);
if ((vblwait->request.type & _DRM_VBLANK_NEXTONMISS) &&
(seq - vblwait->request.sequence) <= (1 << 23)) {
vblwait->request.sequence = seq + 1;
vblwait->reply.sequence = vblwait->request.sequence;
}
DRM_DEBUG("event on vblank count %d, current %d, crtc %u\n",
vblwait->request.sequence, seq, pipe);
trace_drm_vblank_event_queued(current->pid, pipe,
vblwait->request.sequence);
e->event.sequence = vblwait->request.sequence;
if ((seq - vblwait->request.sequence) <= (1 << 23)) {
drm_vblank_put(dev, pipe);
send_vblank_event(dev, e, seq, &now);
vblwait->reply.sequence = seq;
} else {
/* drm_handle_vblank_events will call drm_vblank_put */
list_add_tail(&e->base.link, &dev->vblank_event_list);
vblwait->reply.sequence = vblwait->request.sequence;
}
spin_unlock_irqrestore(&dev->event_lock, flags);
return 0;
err_unlock:
spin_unlock_irqrestore(&dev->event_lock, flags);
kfree(e);
err_put:
drm_vblank_put(dev, pipe);
return ret;
}
/*
* Wait for VBLANK.
*
* \param inode device inode.
* \param file_priv DRM file private.
* \param cmd command.
* \param data user argument, pointing to a drm_wait_vblank structure.
* \return zero on success or a negative number on failure.
*
* This function enables the vblank interrupt on the pipe requested, then
* sleeps waiting for the requested sequence number to occur, and drops
* the vblank interrupt refcount afterwards. (vblank IRQ disable follows that
* after a timeout with no further vblank waits scheduled).
*/
int drm_wait_vblank(struct drm_device *dev, void *data,
struct drm_file *file_priv)
{
struct drm_vblank_crtc *vblank;
union drm_wait_vblank *vblwait = data;
int ret;
unsigned int flags, seq, pipe, high_pipe;
if (!dev->irq_enabled)
return -EINVAL;
if (vblwait->request.type & _DRM_VBLANK_SIGNAL)
return -EINVAL;
if (vblwait->request.type &
~(_DRM_VBLANK_TYPES_MASK | _DRM_VBLANK_FLAGS_MASK |
_DRM_VBLANK_HIGH_CRTC_MASK)) {
DRM_ERROR("Unsupported type value 0x%x, supported mask 0x%x\n",
vblwait->request.type,
(_DRM_VBLANK_TYPES_MASK | _DRM_VBLANK_FLAGS_MASK |
_DRM_VBLANK_HIGH_CRTC_MASK));
return -EINVAL;
}
flags = vblwait->request.type & _DRM_VBLANK_FLAGS_MASK;
high_pipe = (vblwait->request.type & _DRM_VBLANK_HIGH_CRTC_MASK);
if (high_pipe)
pipe = high_pipe >> _DRM_VBLANK_HIGH_CRTC_SHIFT;
else
pipe = flags & _DRM_VBLANK_SECONDARY ? 1 : 0;
if (pipe >= dev->num_crtcs)
return -EINVAL;
vblank = &dev->vblank[pipe];
ret = drm_vblank_get(dev, pipe);
if (ret) {
DRM_DEBUG("failed to acquire vblank counter, %d\n", ret);
return ret;
}
seq = drm_vblank_count(dev, pipe);
switch (vblwait->request.type & _DRM_VBLANK_TYPES_MASK) {
case _DRM_VBLANK_RELATIVE:
vblwait->request.sequence += seq;
vblwait->request.type &= ~_DRM_VBLANK_RELATIVE;
case _DRM_VBLANK_ABSOLUTE:
break;
default:
ret = -EINVAL;
goto done;
}
if (flags & _DRM_VBLANK_EVENT) {
/* must hold on to the vblank ref until the event fires
* drm_vblank_put will be called asynchronously
*/
return drm_queue_vblank_event(dev, pipe, vblwait, file_priv);
}
if ((flags & _DRM_VBLANK_NEXTONMISS) &&
(seq - vblwait->request.sequence) <= (1<<23)) {
vblwait->request.sequence = seq + 1;
}
DRM_DEBUG("waiting on vblank count %d, crtc %u\n",
vblwait->request.sequence, pipe);
vblank->last_wait = vblwait->request.sequence;
DRM_WAIT_ON(ret, vblank->queue, 3 * HZ,
(((drm_vblank_count(dev, pipe) -
vblwait->request.sequence) <= (1 << 23)) ||
!vblank->enabled ||
!dev->irq_enabled));
if (ret != -EINTR) {
struct timeval now;
vblwait->reply.sequence = drm_vblank_count_and_time(dev, pipe, &now);
vblwait->reply.tval_sec = now.tv_sec;
vblwait->reply.tval_usec = now.tv_usec;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
DRM_DEBUG("returning %d to client\n",
vblwait->reply.sequence);
} else {
DRM_DEBUG("vblank wait interrupted by signal\n");
}
done:
drm_vblank_put(dev, pipe);
return ret;
}
static void drm_handle_vblank_events(struct drm_device *dev, unsigned int pipe)
{
struct drm_pending_vblank_event *e, *t;
struct timeval now;
unsigned int seq;
assert_spin_locked(&dev->event_lock);
seq = drm_vblank_count_and_time(dev, pipe, &now);
list_for_each_entry_safe(e, t, &dev->vblank_event_list, base.link) {
if (e->pipe != pipe)
continue;
if ((seq - e->event.sequence) > (1<<23))
continue;
DRM_DEBUG("vblank event on %d, current %d\n",
e->event.sequence, seq);
list_del(&e->base.link);
drm_vblank_put(dev, pipe);
send_vblank_event(dev, e, seq, &now);
}
trace_drm_vblank_event(pipe, seq);
}
/**
* drm_handle_vblank - handle a vblank event
* @dev: DRM device
* @pipe: index of CRTC where this event occurred
*
* Drivers should call this routine in their vblank interrupt handlers to
* update the vblank counter and send any signals that may be pending.
*
* This is the legacy version of drm_crtc_handle_vblank().
*/
bool drm_handle_vblank(struct drm_device *dev, unsigned int pipe)
{
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
u32 vblcount;
s64 diff_ns;
struct timeval tvblank;
unsigned long irqflags;
if (WARN_ON_ONCE(!dev->num_crtcs))
return false;
if (WARN_ON(pipe >= dev->num_crtcs))
return false;
spin_lock_irqsave(&dev->event_lock, irqflags);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Need timestamp lock to prevent concurrent execution with
* vblank enable/disable, as this would cause inconsistent
* or corrupted timestamps and vblank counts.
*/
spin_lock(&dev->vblank_time_lock);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Vblank irq handling disabled. Nothing to do. */
if (!vblank->enabled) {
spin_unlock(&dev->vblank_time_lock);
spin_unlock_irqrestore(&dev->event_lock, irqflags);
return false;
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
}
/* Fetch corresponding timestamp for this vblank interval from
* driver and store it in proper slot of timestamp ringbuffer.
*/
/* Get current timestamp and count. */
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
vblcount = vblank->count;
drm_get_last_vbltimestamp(dev, pipe, &tvblank, DRM_CALLED_FROM_VBLIRQ);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Compute time difference to timestamp of last vblank */
diff_ns = timeval_to_ns(&tvblank) -
timeval_to_ns(&vblanktimestamp(dev, pipe, vblcount));
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
/* Update vblank timestamp and count if at least
* DRM_REDUNDANT_VBLIRQ_THRESH_NS nanoseconds
* difference between last stored timestamp and current
* timestamp. A smaller difference means basically
* identical timestamps. Happens if this vblank has
* been already processed and this is a redundant call,
* e.g., due to spurious vblank interrupts. We need to
* ignore those for accounting.
*/
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
if (abs64(diff_ns) > DRM_REDUNDANT_VBLIRQ_THRESH_NS)
store_vblank(dev, pipe, 1, &tvblank);
drm/vblank: Fixup and document timestamp update/read barriers This was a bit too much cargo-culted, so lets make it solid: - vblank->count doesn't need to be an atomic, writes are always done under the protection of dev->vblank_time_lock. Switch to an unsigned long instead and update comments. Note that atomic_read is just a normal read of a volatile variable, so no need to audit all the read-side access specifically. - The barriers for the vblank counter seqlock weren't complete: The read-side was missing the first barrier between the counter read and the timestamp read, it only had a barrier between the ts and the counter read. We need both. - Barriers weren't properly documented. Since barriers only work if you have them on boths sides of the transaction it's prudent to reference where the other side is. To avoid duplicating the write-side comment 3 times extract a little store_vblank() helper. In that helper also assert that we do indeed hold dev->vblank_time_lock, since in some cases the lock is acquired a few functions up in the callchain. Spotted while reviewing a patch from Chris Wilson to add a fastpath to the vblank_wait ioctl. v2: Add comment to better explain how store_vblank works, suggested by Chris. v3: Peter noticed that as-is the 2nd smp_wmb is redundant with the implicit barrier in the spin_unlock. But that can only be proven by auditing all callers and my point in extracting this little helper was to localize all the locking into just one place. Hence I think that additional optimization is too risky. Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mario Kleiner <mario.kleiner.de@gmail.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Michel Dänzer <michel@daenzer.net> Cc: Peter Hurley <peter@hurleysoftware.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-and-tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
2015-04-16 01:34:43 +08:00
else
DRM_DEBUG("crtc %u: Redundant vblirq ignored. diff_ns = %d\n",
pipe, (int) diff_ns);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
spin_unlock(&dev->vblank_time_lock);
wake_up(&vblank->queue);
drm_handle_vblank_events(dev, pipe);
drm/vblank: Add support for precise vblank timestamping. The DRI2 swap & sync implementation needs precise vblank counts and precise timestamps corresponding to those vblank counts. For conformance to the OpenML OML_sync_control extension specification the DRM timestamp associated with a vblank count should correspond to the start of video scanout of the first scanline of the video frame following the vblank interval for that vblank count. Therefore we need to carry around precise timestamps for vblanks. Currently the DRM and KMS drivers generate timestamps ad-hoc via do_gettimeofday() in some places. The resulting timestamps are sometimes not very precise due to interrupt handling delays, they don't conform to OML_sync_control and some are wrong, as they aren't taken synchronized to the vblank. This patch implements support inside the drm core for precise and robust timestamping. It consists of the following interrelated pieces. 1. Vblank timestamp caching: A per-crtc ringbuffer stores the most recent vblank timestamps corresponding to vblank counts. The ringbuffer can be read out lock-free via the accessor function: struct timeval timestamp; vblankcount = drm_vblank_count_and_time(dev, crtcid, &timestamp). The function returns the current vblank count and the corresponding timestamp for start of video scanout following the vblank interval. It can be used anywhere between enclosing drm_vblank_get(dev, crtcid) and drm_vblank_put(dev,crtcid) statements. It is used inside the drmWaitVblank ioctl and in the vblank event queueing and handling. It should be used by kms drivers for timestamping of bufferswap completion. The timestamp ringbuffer is reinitialized each time vblank irq's get reenabled in drm_vblank_get()/ drm_update_vblank_count(). It is invalidated when vblank irq's get disabled. The ringbuffer is updated inside drm_handle_vblank() at each vblank irq. 2. Calculation of precise vblank timestamps: drm_get_last_vbltimestamp() is used to compute the timestamp for the end of the most recent vblank (if inside active scanout), or the expected end of the current vblank interval (if called inside a vblank interval). The function calls into a new optional kms driver entry point dev->driver->get_vblank_timestamp() which is supposed to provide the precise timestamp. If a kms driver doesn't implement the entry point or if the call fails, a simple do_gettimeofday() timestamp is returned as crude approximation of the true vblank time. A new drm module parameter drm.timestamp_precision_usec allows to disable high precision timestamps (if set to zero) or to specify the maximum acceptable error in the timestamps in microseconds. Kms drivers could implement their get_vblank_timestamp() function in a gpu specific way, as long as returned timestamps conform to OML_sync_control, e.g., by use of gpu specific hardware timestamps. Optionally, kms drivers can simply wrap and use the new utility function drm_calc_vbltimestamp_from_scanoutpos(). This function calls a new optional kms driver function dev->driver->get_scanout_position() which returns the current horizontal and vertical video scanout position of the crtc. The scanout position together with the drm_display_timing of the current video mode is used to calculate elapsed time relative to start of active scanout for the current video frame. This elapsed time is subtracted from the current do_gettimeofday() time to get the timestamp corresponding to start of video scanout. Currently non-interlaced, non-doublescan video modes, with or without panel scaling are handled correctly. Interlaced/ doublescan modes are tbd in a future patch. 3. Filtering of redundant vblank irq's and removal of some race-conditions in the vblank irq enable/disable path: Some gpu's (e.g., Radeon R500/R600) send spurious vblank irq's outside the vblank if vblank irq's get reenabled. These get detected by use of the vblank timestamps and filtered out to avoid miscounting of vblanks. Some race-conditions between the vblank irq enable/disable functions, the vblank irq handler and the gpu itself (updating its hardware vblank counter in the "wrong" moment) are fixed inside vblank_disable_and_save() and drm_update_vblank_count() by use of the vblank timestamps and a new spinlock dev->vblank_time_lock. The time until vblank irq disable is now configurable via a new drm module parameter drm.vblankoffdelay to allow experimentation with timeouts that are much shorter than the current 5 seconds and should allow longer vblank off periods for better power savings. Followup patches will use these new functions to implement precise timestamping for the intel and radeon kms drivers. Signed-off-by: Mario Kleiner <mario.kleiner@tuebingen.mpg.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-23 10:20:23 +08:00
spin_unlock_irqrestore(&dev->event_lock, irqflags);
return true;
}
EXPORT_SYMBOL(drm_handle_vblank);
/**
* drm_crtc_handle_vblank - handle a vblank event
* @crtc: where this event occurred
*
* Drivers should call this routine in their vblank interrupt handlers to
* update the vblank counter and send any signals that may be pending.
*
* This is the native KMS version of drm_handle_vblank().
*
* Returns:
* True if the event was successfully handled, false on failure.
*/
bool drm_crtc_handle_vblank(struct drm_crtc *crtc)
{
return drm_handle_vblank(crtc->dev, drm_crtc_index(crtc));
}
EXPORT_SYMBOL(drm_crtc_handle_vblank);