2014-05-08 22:41:51 +08:00
|
|
|
/*
|
|
|
|
* drm_irq.c IRQ and vblank support
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
|
|
|
* \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.
|
|
|
|
*/
|
|
|
|
|
2012-10-03 01:01:07 +08:00
|
|
|
#include <drm/drmP.h>
|
2010-07-02 07:47:31 +08:00
|
|
|
#include "drm_trace.h"
|
2014-09-10 14:16:10 +08:00
|
|
|
#include "drm_internal.h"
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
#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>
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-09-21 12:33:58 +08:00
|
|
|
#include <linux/vgaarb.h>
|
2011-08-31 06:16:33 +08:00
|
|
|
#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, ×tamp).
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. */
|
2015-08-12 23:00:31 +08:00
|
|
|
#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, ×tamp).
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
|
|
|
|
|
2014-09-10 23:36:10 +08:00
|
|
|
static bool
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_get_last_vbltimestamp(struct drm_device *dev, unsigned int pipe,
|
2014-09-10 23:36:10 +08:00
|
|
|
struct timeval *tvblank, unsigned flags);
|
|
|
|
|
2014-09-10 14:16:10 +08:00
|
|
|
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);
|
2016-01-08 19:00:45 +08:00
|
|
|
MODULE_PARM_DESC(vblankoffdelay, "Delay until vblank irq auto-disable [msecs] (0: never disable, <0: disable immediately)");
|
|
|
|
MODULE_PARM_DESC(timestamp_precision_usec, "Max. error on timestamps [usecs]");
|
|
|
|
MODULE_PARM_DESC(timestamp_monotonic, "Use monotonic timestamps");
|
2014-09-10 14:16:10 +08:00
|
|
|
|
2015-09-15 03:43:42 +08:00
|
|
|
static void store_vblank(struct drm_device *dev, unsigned int pipe,
|
2015-08-07 18:31:17 +08:00
|
|
|
u32 vblank_count_inc,
|
2015-09-15 03:43:51 +08:00
|
|
|
struct timeval *t_vblank, u32 last)
|
2015-04-16 01:34:43 +08:00
|
|
|
{
|
2015-09-15 03:43:42 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2015-04-16 01:34:43 +08:00
|
|
|
u32 tslot;
|
|
|
|
|
|
|
|
assert_spin_locked(&dev->vblank_time_lock);
|
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
vblank->last = last;
|
|
|
|
|
2015-09-15 03:43:50 +08:00
|
|
|
/* 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;
|
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();
|
|
|
|
}
|
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
/**
|
|
|
|
* drm_reset_vblank_timestamp - reset the last timestamp to the last vblank
|
|
|
|
* @dev: DRM device
|
|
|
|
* @pipe: index of CRTC for which to reset the timestamp
|
|
|
|
*
|
|
|
|
* Reset the stored timestamp for the current vblank count to correspond
|
|
|
|
* to the last vblank occurred.
|
|
|
|
*
|
|
|
|
* Only to be called from drm_vblank_on().
|
|
|
|
*
|
|
|
|
* Note: caller must hold dev->vbl_lock since this reads & writes
|
|
|
|
* device vblank fields.
|
|
|
|
*/
|
|
|
|
static void drm_reset_vblank_timestamp(struct drm_device *dev, unsigned int pipe)
|
|
|
|
{
|
|
|
|
u32 cur_vblank;
|
|
|
|
bool rc;
|
|
|
|
struct timeval t_vblank;
|
|
|
|
int count = DRM_TIMESTAMP_MAXRETRIES;
|
|
|
|
|
|
|
|
spin_lock(&dev->vblank_time_lock);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* sample the current counter to avoid random jumps
|
|
|
|
* when drm_vblank_enable() applies the diff
|
|
|
|
*/
|
|
|
|
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) && --count > 0);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Only reinitialize corresponding vblank timestamp if high-precision query
|
|
|
|
* available and didn't fail. Otherwise reinitialize delayed at next vblank
|
|
|
|
* interrupt and assign 0 for now, to mark the vblanktimestamp as invalid.
|
|
|
|
*/
|
|
|
|
if (!rc)
|
|
|
|
t_vblank = (struct timeval) {0, 0};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* +1 to make sure user will never see the same
|
|
|
|
* vblank counter value before and after a modeset
|
|
|
|
*/
|
|
|
|
store_vblank(dev, pipe, 1, &t_vblank, cur_vblank);
|
|
|
|
|
|
|
|
spin_unlock(&dev->vblank_time_lock);
|
|
|
|
}
|
|
|
|
|
2014-08-06 19:49:47 +08:00
|
|
|
/**
|
|
|
|
* drm_update_vblank_count - update the master vblank counter
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: counter to update
|
2014-08-06 19:49:47 +08:00
|
|
|
*
|
|
|
|
* Call back into the driver to update the appropriate vblank counter
|
2015-09-15 03:43:42 +08:00
|
|
|
* (specified by @pipe). Deal with wraparound, if it occurred, and
|
2014-08-06 19:49:47 +08:00
|
|
|
* 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.
|
|
|
|
*/
|
2015-09-15 03:43:47 +08:00
|
|
|
static void drm_update_vblank_count(struct drm_device *dev, unsigned int pipe,
|
|
|
|
unsigned long flags)
|
2014-08-06 19:49:47 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2015-04-16 01:34:43 +08:00
|
|
|
u32 cur_vblank, diff;
|
2014-09-10 23:36:10 +08:00
|
|
|
bool rc;
|
2014-08-06 19:49:47 +08:00
|
|
|
struct timeval t_vblank;
|
2015-09-15 03:43:48 +08:00
|
|
|
int count = DRM_TIMESTAMP_MAXRETRIES;
|
2015-09-15 03:43:51 +08:00
|
|
|
int framedur_ns = vblank->framedur_ns;
|
2014-08-06 19:49:47 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Interrupts were disabled prior to this call, so deal with counter
|
|
|
|
* wrap if needed.
|
2015-05-26 16:53:38 +08:00
|
|
|
* NOTE! It's possible we lost a full dev->max_vblank_count + 1 events
|
2014-08-06 19:49:47 +08:00
|
|
|
* 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 {
|
2015-08-12 23:00:31 +08:00
|
|
|
cur_vblank = dev->driver->get_vblank_counter(dev, pipe);
|
2015-09-15 03:43:47 +08:00
|
|
|
rc = drm_get_last_vbltimestamp(dev, pipe, &t_vblank, flags);
|
2015-09-15 03:43:48 +08:00
|
|
|
} while (cur_vblank != dev->driver->get_vblank_counter(dev, pipe) && --count > 0);
|
2014-08-06 19:49:47 +08:00
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
if (dev->max_vblank_count != 0) {
|
|
|
|
/* trust the hw counter when it's around */
|
|
|
|
diff = (cur_vblank - vblank->last) & dev->max_vblank_count;
|
|
|
|
} else if (rc && framedur_ns) {
|
|
|
|
const struct timeval *t_old;
|
|
|
|
u64 diff_ns;
|
|
|
|
|
|
|
|
t_old = &vblanktimestamp(dev, pipe, vblank->count);
|
|
|
|
diff_ns = timeval_to_ns(&t_vblank) - timeval_to_ns(t_old);
|
2014-08-06 19:49:47 +08:00
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
/*
|
|
|
|
* Figure out how many vblanks we've missed based
|
|
|
|
* on the difference in the timestamps and the
|
|
|
|
* frame/field duration.
|
|
|
|
*/
|
|
|
|
diff = DIV_ROUND_CLOSEST_ULL(diff_ns, framedur_ns);
|
|
|
|
|
|
|
|
if (diff == 0 && flags & DRM_CALLED_FROM_VBLIRQ)
|
2015-10-10 03:57:37 +08:00
|
|
|
DRM_DEBUG_VBL("crtc %u: Redundant vblirq ignored."
|
|
|
|
" diff_ns = %lld, framedur_ns = %d)\n",
|
|
|
|
pipe, (long long) diff_ns, framedur_ns);
|
2015-09-15 03:43:51 +08:00
|
|
|
} else {
|
|
|
|
/* some kind of default for drivers w/o accurate vbl timestamping */
|
|
|
|
diff = (flags & DRM_CALLED_FROM_VBLIRQ) != 0;
|
2014-08-06 19:49:47 +08:00
|
|
|
}
|
|
|
|
|
2015-10-10 03:57:37 +08:00
|
|
|
DRM_DEBUG_VBL("updating vblank count on crtc %u:"
|
|
|
|
" current=%u, diff=%u, hw=%u hw_last=%u\n",
|
|
|
|
pipe, vblank->count, diff, cur_vblank, vblank->last);
|
2014-08-06 19:49:47 +08:00
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
if (diff == 0) {
|
|
|
|
WARN_ON_ONCE(cur_vblank != vblank->last);
|
2014-09-14 00:25:54 +08:00
|
|
|
return;
|
2015-09-15 03:43:51 +08:00
|
|
|
}
|
2014-09-14 00:25:54 +08:00
|
|
|
|
2015-04-16 01:34:43 +08:00
|
|
|
/*
|
|
|
|
* Only reinitialize corresponding vblank timestamp if high-precision query
|
2015-10-01 00:21:34 +08:00
|
|
|
* available and didn't fail, or we were called from the vblank interrupt.
|
|
|
|
* Otherwise reinitialize delayed at next vblank interrupt and assign 0
|
|
|
|
* for now, to mark the vblanktimestamp as invalid.
|
2014-08-06 19:49:47 +08:00
|
|
|
*/
|
2015-10-01 00:21:34 +08:00
|
|
|
if (!rc && (flags & DRM_CALLED_FROM_VBLIRQ) == 0)
|
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
|
|
|
t_vblank = (struct timeval) {0, 0};
|
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
store_vblank(dev, pipe, diff, &t_vblank, cur_vblank);
|
2014-08-06 19:49:47 +08:00
|
|
|
}
|
|
|
|
|
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, ×tamp).
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.
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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
|
|
|
{
|
2015-08-12 23:00:31 +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, ×tamp).
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;
|
|
|
|
|
|
|
|
/* 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);
|
|
|
|
|
2015-01-29 07:09:24 +08:00
|
|
|
/*
|
|
|
|
* 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) {
|
2015-08-12 23:00:31 +08:00
|
|
|
dev->driver->disable_vblank(dev, pipe);
|
2015-01-29 07:09:24 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
/*
|
|
|
|
* Always 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().
|
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, ×tamp).
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
|
|
|
*/
|
2015-09-15 03:43:51 +08:00
|
|
|
drm_update_vblank_count(dev, pipe, 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, ×tamp).
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);
|
|
|
|
}
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
static void vblank_disable_fn(unsigned long arg)
|
|
|
|
{
|
2014-02-20 01:36:08 +08:00
|
|
|
struct drm_vblank_crtc *vblank = (void *)arg;
|
|
|
|
struct drm_device *dev = vblank->dev;
|
2015-08-12 23:00:31 +08:00
|
|
|
unsigned int pipe = vblank->pipe;
|
2008-10-01 03:14:26 +08:00
|
|
|
unsigned long irqflags;
|
|
|
|
|
|
|
|
if (!dev->vblank_disable_allowed)
|
|
|
|
return;
|
|
|
|
|
2014-02-20 01:36:08 +08:00
|
|
|
spin_lock_irqsave(&dev->vbl_lock, irqflags);
|
|
|
|
if (atomic_read(&vblank->refcount) == 0 && vblank->enabled) {
|
2015-08-12 23:00:31 +08:00
|
|
|
DRM_DEBUG("disabling vblank on crtc %u\n", pipe);
|
|
|
|
vblank_disable_and_save(dev, pipe);
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
2014-02-20 01:36:08 +08:00
|
|
|
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
|
2014-05-08 22:41:51 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_cleanup - cleanup vblank support
|
|
|
|
* @dev: DRM device
|
|
|
|
*
|
|
|
|
* This function cleans up any resources allocated in drm_vblank_init.
|
|
|
|
*/
|
2008-11-19 01:30:25 +08:00
|
|
|
void drm_vblank_cleanup(struct drm_device *dev)
|
2008-10-01 03:14:26 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
unsigned int pipe;
|
2014-02-20 01:36:08 +08:00
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
/* Bail if the driver didn't call drm_vblank_init() */
|
|
|
|
if (dev->num_crtcs == 0)
|
|
|
|
return;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
for (pipe = 0; pipe < dev->num_crtcs; pipe++) {
|
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-08-06 19:49:50 +08:00
|
|
|
|
2015-02-22 22:11:19 +08:00
|
|
|
WARN_ON(vblank->enabled &&
|
|
|
|
drm_core_check_feature(dev, DRIVER_MODESET));
|
2014-08-06 09:22:46 +08:00
|
|
|
|
2015-02-22 22:11:19 +08:00
|
|
|
del_timer_sync(&vblank->disable_timer);
|
2014-02-20 01:36:08 +08:00
|
|
|
}
|
2008-10-01 03:14:26 +08:00
|
|
|
|
2013-10-04 19:53:36 +08:00
|
|
|
kfree(dev->vblank);
|
2008-10-01 03:14:26 +08:00
|
|
|
|
|
|
|
dev->num_crtcs = 0;
|
|
|
|
}
|
2010-01-07 22:39:13 +08:00
|
|
|
EXPORT_SYMBOL(drm_vblank_cleanup);
|
2008-10-01 03:14:26 +08:00
|
|
|
|
2014-05-08 22:41:51 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_init - initialize vblank support
|
2015-08-12 23:00:31 +08:00
|
|
|
* @dev: DRM device
|
|
|
|
* @num_crtcs: number of CRTCs supported by @dev
|
2014-05-08 22:41:51 +08:00
|
|
|
*
|
|
|
|
* This function initializes vblank support for @num_crtcs display pipelines.
|
|
|
|
*
|
|
|
|
* Returns:
|
|
|
|
* Zero on success or a negative error code on failure.
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
int drm_vblank_init(struct drm_device *dev, unsigned int num_crtcs)
|
2008-10-01 03:14:26 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
int ret = -ENOMEM;
|
|
|
|
unsigned int i;
|
2008-10-01 03:14:26 +08:00
|
|
|
|
|
|
|
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, ×tamp).
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);
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
dev->num_crtcs = num_crtcs;
|
|
|
|
|
2013-10-04 19:53:36 +08:00
|
|
|
dev->vblank = kcalloc(num_crtcs, sizeof(*dev->vblank), GFP_KERNEL);
|
|
|
|
if (!dev->vblank)
|
2008-10-01 03:14:26 +08:00
|
|
|
goto err;
|
|
|
|
|
2014-02-20 01:36:08 +08:00
|
|
|
for (i = 0; i < num_crtcs; i++) {
|
2014-08-06 19:49:50 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[i];
|
|
|
|
|
|
|
|
vblank->dev = dev;
|
2015-08-12 23:00:31 +08:00
|
|
|
vblank->pipe = i;
|
2014-08-06 19:49:50 +08:00
|
|
|
init_waitqueue_head(&vblank->queue);
|
|
|
|
setup_timer(&vblank->disable_timer, vblank_disable_fn,
|
|
|
|
(unsigned long)vblank);
|
2014-02-20 01:36:08 +08:00
|
|
|
}
|
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, ×tamp).
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
|
|
|
|
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, ×tamp).
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");
|
|
|
|
|
2015-05-04 12:29:45 +08:00
|
|
|
/* 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");
|
|
|
|
}
|
|
|
|
|
2013-10-04 19:53:33 +08:00
|
|
|
dev->vblank_disable_allowed = false;
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
err:
|
2014-08-06 09:22:44 +08:00
|
|
|
dev->num_crtcs = 0;
|
2008-10-01 03:14:26 +08:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_init);
|
|
|
|
|
2009-09-21 12:33:58 +08:00
|
|
|
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;
|
|
|
|
|
2011-08-04 13:41:01 +08:00
|
|
|
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);
|
2009-09-21 12:33:58 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/**
|
2014-05-08 22:41:51 +08:00
|
|
|
* drm_irq_install - install IRQ handler
|
|
|
|
* @dev: DRM device
|
|
|
|
* @irq: IRQ number to install the handler for
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2008-10-01 03:14:26 +08:00
|
|
|
* Initializes the IRQ related data. Installs the handler, calling the driver
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2013-11-04 04:09:27 +08:00
|
|
|
int drm_irq_install(struct drm_device *dev, int irq)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2012-05-17 19:27:21 +08:00
|
|
|
int ret;
|
2005-09-25 12:28:13 +08:00
|
|
|
unsigned long sh_flags = 0;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2013-12-16 18:21:15 +08:00
|
|
|
if (irq == 0)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/* Driver must have been initialized */
|
2013-11-04 03:27:05 +08:00
|
|
|
if (!dev->dev_private)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
2013-11-04 03:27:05 +08:00
|
|
|
if (dev->irq_enabled)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EBUSY;
|
2013-10-04 19:53:37 +08:00
|
|
|
dev->irq_enabled = true;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-12-16 18:21:15 +08:00
|
|
|
DRM_DEBUG("irq=%d\n", irq);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-09-25 12:28:13 +08:00
|
|
|
/* Before installing handler */
|
2011-08-04 13:41:01 +08:00
|
|
|
if (dev->driver->irq_preinstall)
|
|
|
|
dev->driver->irq_preinstall(dev);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-09-25 12:28:13 +08:00
|
|
|
/* Install handler */
|
2005-04-17 06:20:36 +08:00
|
|
|
if (drm_core_check_feature(dev, DRIVER_IRQ_SHARED))
|
2006-07-02 10:29:34 +08:00
|
|
|
sh_flags = IRQF_SHARED;
|
2005-09-25 12:28:13 +08:00
|
|
|
|
2013-12-16 18:21:15 +08:00
|
|
|
ret = request_irq(irq, dev->driver->irq_handler,
|
2013-11-04 04:48:48 +08:00
|
|
|
sh_flags, dev->driver->name, dev);
|
2008-09-16 06:00:33 +08:00
|
|
|
|
2005-09-25 12:28:13 +08:00
|
|
|
if (ret < 0) {
|
2013-10-04 19:53:37 +08:00
|
|
|
dev->irq_enabled = false;
|
2005-04-17 06:20:36 +08:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-09-21 12:33:58 +08:00
|
|
|
if (!drm_core_check_feature(dev, DRIVER_MODESET))
|
|
|
|
vga_client_register(dev->pdev, (void *)dev, drm_irq_vgaarb_nokms, NULL);
|
|
|
|
|
2005-09-25 12:28:13 +08:00
|
|
|
/* After installing handler */
|
2011-08-04 13:41:01 +08:00
|
|
|
if (dev->driver->irq_postinstall)
|
|
|
|
ret = dev->driver->irq_postinstall(dev);
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
if (ret < 0) {
|
2013-10-04 19:53:37 +08:00
|
|
|
dev->irq_enabled = false;
|
2011-08-04 13:41:00 +08:00
|
|
|
if (!drm_core_check_feature(dev, DRIVER_MODESET))
|
|
|
|
vga_client_register(dev->pdev, NULL, NULL, NULL);
|
2013-12-16 18:21:15 +08:00
|
|
|
free_irq(irq, dev);
|
|
|
|
} else {
|
|
|
|
dev->irq = irq;
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
return ret;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2008-10-01 03:14:26 +08:00
|
|
|
EXPORT_SYMBOL(drm_irq_install);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/**
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* Returns:
|
|
|
|
* Zero on success or a negative error code on failure.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
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, ×tamp).
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)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2009-01-07 02:21:24 +08:00
|
|
|
unsigned long irqflags;
|
2013-10-04 19:53:37 +08:00
|
|
|
bool irq_enabled;
|
|
|
|
int i;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (!drm_core_check_feature(dev, DRIVER_HAVE_IRQ))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
irq_enabled = dev->irq_enabled;
|
2013-10-04 19:53:37 +08:00
|
|
|
dev->irq_enabled = false;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-01-07 02:21:24 +08:00
|
|
|
/*
|
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.
|
2009-01-07 02:21:24 +08:00
|
|
|
*/
|
2011-07-04 10:52:27 +08:00
|
|
|
if (dev->num_crtcs) {
|
|
|
|
spin_lock_irqsave(&dev->vbl_lock, irqflags);
|
|
|
|
for (i = 0; i < dev->num_crtcs; i++) {
|
2014-08-06 19:49:50 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[i];
|
|
|
|
|
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);
|
2014-08-06 19:49:50 +08:00
|
|
|
wake_up(&vblank->queue);
|
2011-07-04 10:52:27 +08:00
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
|
2009-01-07 02:21:24 +08:00
|
|
|
}
|
|
|
|
|
2005-09-25 12:28:13 +08:00
|
|
|
if (!irq_enabled)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
2013-12-16 18:21:15 +08:00
|
|
|
DRM_DEBUG("irq=%d\n", dev->irq);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-09-21 12:33:58 +08:00
|
|
|
if (!drm_core_check_feature(dev, DRIVER_MODESET))
|
|
|
|
vga_client_register(dev->pdev, NULL, NULL, NULL);
|
|
|
|
|
2011-08-04 13:41:01 +08:00
|
|
|
if (dev->driver->irq_uninstall)
|
|
|
|
dev->driver->irq_uninstall(dev);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-12-16 18:21:15 +08:00
|
|
|
free_irq(dev->irq, dev);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_irq_uninstall);
|
|
|
|
|
2014-05-08 22:41:51 +08:00
|
|
|
/*
|
2005-04-17 06:20:36 +08:00
|
|
|
* IRQ control ioctl.
|
|
|
|
*
|
|
|
|
* \param inode device inode.
|
2007-08-25 18:23:09 +08:00
|
|
|
* \param file_priv DRM file private.
|
2005-04-17 06:20:36 +08:00
|
|
|
* \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.
|
|
|
|
*/
|
2007-09-03 10:06:45 +08:00
|
|
|
int drm_control(struct drm_device *dev, void *data,
|
|
|
|
struct drm_file *file_priv)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2007-09-03 10:06:45 +08:00
|
|
|
struct drm_control *ctl = data;
|
2013-11-04 04:09:15 +08:00
|
|
|
int ret = 0, irq;
|
2005-09-25 12:28:13 +08:00
|
|
|
|
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, ×tamp).
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
|
|
|
|
*/
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-11-04 03:02:50 +08:00
|
|
|
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;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2007-09-03 10:06:45 +08:00
|
|
|
switch (ctl->func) {
|
2005-04-17 06:20:36 +08:00
|
|
|
case DRM_INST_HANDLER:
|
2013-11-04 04:09:15 +08:00
|
|
|
irq = dev->pdev->irq;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
if (dev->if_version < DRM_IF_VERSION(1, 2) &&
|
2013-11-04 04:09:15 +08:00
|
|
|
ctl->irq != irq)
|
2005-04-17 06:20:36 +08:00
|
|
|
return -EINVAL;
|
2013-11-04 03:27:05 +08:00
|
|
|
mutex_lock(&dev->struct_mutex);
|
2013-11-04 04:09:27 +08:00
|
|
|
ret = drm_irq_install(dev, irq);
|
2013-11-04 03:27:05 +08:00
|
|
|
mutex_unlock(&dev->struct_mutex);
|
|
|
|
|
|
|
|
return ret;
|
2005-04-17 06:20:36 +08:00
|
|
|
case DRM_UNINST_HANDLER:
|
2013-11-04 03:27:05 +08:00
|
|
|
mutex_lock(&dev->struct_mutex);
|
|
|
|
ret = drm_irq_uninstall(dev);
|
|
|
|
mutex_unlock(&dev->struct_mutex);
|
|
|
|
|
|
|
|
return ret;
|
2005-04-17 06:20:36 +08:00
|
|
|
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, ×tamp).
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
|
|
|
/**
|
2014-05-08 22:41:51 +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, ×tamp).
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
|
|
|
*
|
2013-10-26 22:22:52 +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
|
2014-05-08 22:41:51 +08:00
|
|
|
* derived from CRTC's true scanout timing, so they take
|
2013-10-26 22:22:52 +08:00
|
|
|
* 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, ×tamp).
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
|
|
|
*/
|
2013-10-26 22:16:30 +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, ×tamp).
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
|
|
|
{
|
2015-11-12 20:34:18 +08:00
|
|
|
struct drm_device *dev = crtc->dev;
|
|
|
|
unsigned int pipe = drm_crtc_index(crtc);
|
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2015-09-15 03:43:45 +08:00
|
|
|
int linedur_ns = 0, framedur_ns = 0;
|
2013-10-28 03:22:58 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2015-11-12 20:34:18 +08:00
|
|
|
if (!dev->num_crtcs)
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
|
|
|
return;
|
|
|
|
|
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, ×tamp).
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) {
|
2013-10-26 22:11:01 +08:00
|
|
|
int frame_size = mode->crtc_htotal * mode->crtc_vtotal;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Convert scanline length in pixels and video
|
2015-09-15 03:43:45 +08:00
|
|
|
* dot clock to line duration and frame 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, ×tamp).
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
|
|
|
*/
|
2013-10-26 22:11:01 +08:00
|
|
|
linedur_ns = div_u64((u64) mode->crtc_htotal * 1000000, dotclock);
|
|
|
|
framedur_ns = div_u64((u64) frame_size * 1000000, dotclock);
|
2013-10-29 01:53:25 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* 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, ×tamp).
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
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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);
|
|
|
|
|
2015-09-15 03:43:43 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
DRM_DEBUG("crtc %u: hwmode: htotal %d, vtotal %d, vdisplay %d\n",
|
2013-10-26 22:16:30 +08:00
|
|
|
crtc->base.id, mode->crtc_htotal,
|
|
|
|
mode->crtc_vtotal, mode->crtc_vdisplay);
|
2015-09-15 03:43:45 +08:00
|
|
|
DRM_DEBUG("crtc %u: clock %d kHz framedur %d linedur %d\n",
|
|
|
|
crtc->base.id, dotclock, framedur_ns, linedur_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, ×tamp).
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);
|
|
|
|
|
|
|
|
/**
|
2014-05-08 22:41:51 +08:00
|
|
|
* drm_calc_vbltimestamp_from_scanoutpos - precise vblank timestamp helper
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: index of CRTC whose vblank timestamp to retrieve
|
2014-05-08 22:41:51 +08:00
|
|
|
* @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, ×tamp).
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.
|
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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, ×tamp).
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:
|
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* -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, ×tamp).
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.
|
|
|
|
*
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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,
|
2013-10-26 22:57:31 +08:00
|
|
|
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, ×tamp).
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
|
|
|
{
|
2012-10-24 02:53:25 +08:00
|
|
|
struct timeval tv_etime;
|
2014-07-17 05:04:26 +08:00
|
|
|
ktime_t stime, etime;
|
2015-09-15 03:43:49 +08:00
|
|
|
unsigned int vbl_status;
|
|
|
|
int ret = DRM_VBLANKTIME_SCANOUTPOS_METHOD;
|
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, ×tamp).
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;
|
2015-09-15 03:43:44 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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.
|
|
|
|
*/
|
2015-09-15 03:43:44 +08:00
|
|
|
if (mode->crtc_clock == 0) {
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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++) {
|
2013-10-30 12:13:06 +08:00
|
|
|
/*
|
|
|
|
* Get vertical and horizontal scanout position vpos, hpos,
|
|
|
|
* and bounding timestamps stime, etime, pre/post query.
|
|
|
|
*/
|
2015-09-15 03:43:44 +08:00
|
|
|
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, ×tamp).
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)) {
|
2015-09-15 03:43:49 +08:00
|
|
|
DRM_DEBUG("crtc %u : scanoutpos query failed [0x%x].\n",
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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;
|
|
|
|
}
|
|
|
|
|
2013-10-30 12:13:06 +08:00
|
|
|
/* Compute uncertainty in timestamp of scanout position query. */
|
2012-10-24 02:53:25 +08:00
|
|
|
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, ×tamp).
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. */
|
2013-10-26 22:38:52 +08:00
|
|
|
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, ×tamp).
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) {
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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. */
|
2013-10-26 22:38:52 +08:00
|
|
|
*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, ×tamp).
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.
|
|
|
|
*/
|
2015-09-15 03:43:49 +08:00
|
|
|
if (vbl_status & DRM_SCANOUTPOS_IN_VBLANK)
|
|
|
|
ret |= 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, ×tamp).
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.
|
|
|
|
*/
|
2015-09-15 03:43:44 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2012-10-24 02:53:26 +08:00
|
|
|
if (!drm_timestamp_monotonic)
|
2014-07-17 05:04:26 +08:00
|
|
|
etime = ktime_mono_to_real(etime);
|
2012-10-24 02:53:26 +08:00
|
|
|
|
2012-10-24 02:53:25 +08:00
|
|
|
/* 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, ×tamp).
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.
|
|
|
|
*/
|
2013-06-12 17:58:44 +08:00
|
|
|
if (delta_ns < 0)
|
|
|
|
etime = ktime_add_ns(etime, -delta_ns);
|
|
|
|
else
|
|
|
|
etime = ktime_sub_ns(etime, delta_ns);
|
2012-10-24 02:53:25 +08:00
|
|
|
*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, ×tamp).
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
|
|
|
|
2015-10-10 03:57:37 +08:00
|
|
|
DRM_DEBUG_VBL("crtc %u : v 0x%x p(%d,%d)@ %ld.%ld -> %ld.%ld [e %d us, %d rep]\n",
|
|
|
|
pipe, 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, ×tamp).
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
|
|
|
|
2015-09-15 03:43:49 +08:00
|
|
|
return 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, ×tamp).
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_vbltimestamp_from_scanoutpos);
|
|
|
|
|
2012-10-24 02:53:26 +08:00
|
|
|
static struct timeval get_drm_timestamp(void)
|
|
|
|
{
|
|
|
|
ktime_t now;
|
|
|
|
|
2014-07-17 05:04:26 +08:00
|
|
|
now = drm_timestamp_monotonic ? ktime_get() : ktime_get_real();
|
2012-10-24 02:53:26 +08:00
|
|
|
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, ×tamp).
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
|
2014-12-10 20:03:34 +08:00
|
|
|
* 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, ×tamp).
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
|
2015-08-12 23:00:31 +08:00
|
|
|
* @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, ×tamp).
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:
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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, ×tamp).
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
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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, ×tamp).
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.
|
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* Returns:
|
2014-09-10 23:36:10 +08:00
|
|
|
* 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, ×tamp).
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
|
|
|
*/
|
2014-09-10 23:36:10 +08:00
|
|
|
static bool
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_get_last_vbltimestamp(struct drm_device *dev, unsigned int pipe,
|
2014-09-10 23:36:10 +08:00
|
|
|
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, ×tamp).
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
|
|
|
{
|
2012-05-17 19:27:21 +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, ×tamp).
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)) {
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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)
|
2014-09-10 23:36:10 +08:00
|
|
|
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, ×tamp).
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.
|
2012-10-24 02:53:26 +08:00
|
|
|
* 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, ×tamp).
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
|
|
|
*/
|
2012-10-24 02:53:26 +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, ×tamp).
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
|
|
|
|
2014-09-10 23:36:10 +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, ×tamp).
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
|
|
|
}
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_count - retrieve "cooked" vblank counter value
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: index of CRTC for which to retrieve the counter
|
2008-10-01 03:14:26 +08:00
|
|
|
*
|
|
|
|
* 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.
|
2014-05-08 22:41:51 +08:00
|
|
|
*
|
2014-12-16 20:08:47 +08:00
|
|
|
* This is the legacy version of drm_crtc_vblank_count().
|
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* Returns:
|
|
|
|
* The software vblank counter.
|
2008-10-01 03:14:26 +08:00
|
|
|
*/
|
2015-09-25 00:35:31 +08:00
|
|
|
u32 drm_vblank_count(struct drm_device *dev, unsigned int pipe)
|
2008-10-01 03:14:26 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-08-06 19:49:50 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return 0;
|
2015-08-12 23:00:30 +08:00
|
|
|
|
2015-04-16 01:34:43 +08:00
|
|
|
return vblank->count;
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_count);
|
|
|
|
|
2014-12-16 20:08:47 +08:00
|
|
|
/**
|
|
|
|
* 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, ×tamp).
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
|
|
|
/**
|
2015-08-12 23:00:31 +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, ×tamp).
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
|
2015-08-12 23:00:31 +08:00
|
|
|
* @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, ×tamp).
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
|
2014-05-08 22:41:51 +08:00
|
|
|
* of the vblank interval that corresponds to the current vblank counter value.
|
2015-09-25 00:35:36 +08:00
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_vblank_count_and_time().
|
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, ×tamp).
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
|
|
|
*/
|
2015-08-12 23:00:31 +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, ×tamp).
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)
|
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2015-09-15 03:43:48 +08:00
|
|
|
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, ×tamp).
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;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return 0;
|
|
|
|
|
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, ×tamp).
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 {
|
2015-04-16 01:34:43 +08:00
|
|
|
cur_vblank = vblank->count;
|
|
|
|
smp_rmb();
|
2015-08-12 23:00:31 +08:00
|
|
|
*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, ×tamp).
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();
|
2015-09-15 03:43:48 +08:00
|
|
|
} while (cur_vblank != vblank->count && --count > 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, ×tamp).
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);
|
|
|
|
|
2015-09-25 00:35:36 +08:00
|
|
|
/**
|
|
|
|
* drm_crtc_vblank_count_and_time - retrieve "cooked" vblank counter value
|
|
|
|
* and the system timestamp corresponding to that vblank counter value
|
|
|
|
* @crtc: which counter to retrieve
|
|
|
|
* @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.
|
|
|
|
*
|
|
|
|
* This is the native KMS version of drm_vblank_count_and_time().
|
|
|
|
*/
|
|
|
|
u32 drm_crtc_vblank_count_and_time(struct drm_crtc *crtc,
|
|
|
|
struct timeval *vblanktime)
|
|
|
|
{
|
|
|
|
return drm_vblank_count_and_time(crtc->dev, drm_crtc_index(crtc),
|
|
|
|
vblanktime);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_crtc_vblank_count_and_time);
|
|
|
|
|
2012-10-17 06:48:40 +08:00
|
|
|
static void send_vblank_event(struct drm_device *dev,
|
|
|
|
struct drm_pending_vblank_event *e,
|
|
|
|
unsigned long seq, struct timeval *now)
|
|
|
|
{
|
2015-11-11 00:37:31 +08:00
|
|
|
assert_spin_locked(&dev->event_lock);
|
|
|
|
|
2012-10-17 06:48:40 +08:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2015-11-11 00:37:31 +08:00
|
|
|
/**
|
|
|
|
* drm_arm_vblank_event - arm vblank event after pageflip
|
|
|
|
* @dev: DRM device
|
|
|
|
* @pipe: CRTC index
|
|
|
|
* @e: the event to prepare to send
|
|
|
|
*
|
|
|
|
* A lot of drivers need to generate vblank events for the very next vblank
|
|
|
|
* interrupt. For example when the page flip interrupt happens when the page
|
|
|
|
* flip gets armed, but not when it actually executes within the next vblank
|
|
|
|
* period. This helper function implements exactly the required vblank arming
|
|
|
|
* behaviour.
|
|
|
|
*
|
|
|
|
* Caller must hold event lock. Caller must also hold a vblank reference for
|
|
|
|
* the event @e, which will be dropped when the next vblank arrives.
|
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_arm_vblank_event().
|
|
|
|
*/
|
|
|
|
void drm_arm_vblank_event(struct drm_device *dev, unsigned int pipe,
|
|
|
|
struct drm_pending_vblank_event *e)
|
|
|
|
{
|
|
|
|
assert_spin_locked(&dev->event_lock);
|
|
|
|
|
|
|
|
e->pipe = pipe;
|
|
|
|
e->event.sequence = drm_vblank_count(dev, pipe);
|
|
|
|
list_add_tail(&e->base.link, &dev->vblank_event_list);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_arm_vblank_event);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* drm_crtc_arm_vblank_event - arm vblank event after pageflip
|
|
|
|
* @crtc: the source CRTC of the vblank event
|
|
|
|
* @e: the event to send
|
|
|
|
*
|
|
|
|
* A lot of drivers need to generate vblank events for the very next vblank
|
|
|
|
* interrupt. For example when the page flip interrupt happens when the page
|
|
|
|
* flip gets armed, but not when it actually executes within the next vblank
|
|
|
|
* period. This helper function implements exactly the required vblank arming
|
|
|
|
* behaviour.
|
|
|
|
*
|
|
|
|
* Caller must hold event lock. Caller must also hold a vblank reference for
|
|
|
|
* the event @e, which will be dropped when the next vblank arrives.
|
|
|
|
*
|
|
|
|
* This is the native KMS version of drm_arm_vblank_event().
|
|
|
|
*/
|
|
|
|
void drm_crtc_arm_vblank_event(struct drm_crtc *crtc,
|
|
|
|
struct drm_pending_vblank_event *e)
|
|
|
|
{
|
|
|
|
drm_arm_vblank_event(crtc->dev, drm_crtc_index(crtc), e);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_crtc_arm_vblank_event);
|
|
|
|
|
2012-10-17 06:48:40 +08:00
|
|
|
/**
|
|
|
|
* drm_send_vblank_event - helper to send vblank event after pageflip
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2012-10-17 06:48:40 +08:00
|
|
|
* @e: the event to send
|
|
|
|
*
|
|
|
|
* Updates sequence # and timestamp on event, and sends it to userspace.
|
|
|
|
* Caller must hold event lock.
|
2014-12-15 21:47:18 +08:00
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_send_vblank_event().
|
2012-10-17 06:48:40 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_send_vblank_event(struct drm_device *dev, unsigned int pipe,
|
|
|
|
struct drm_pending_vblank_event *e)
|
2012-10-17 06:48:40 +08:00
|
|
|
{
|
|
|
|
struct timeval now;
|
|
|
|
unsigned int seq;
|
2014-12-10 20:03:34 +08:00
|
|
|
|
2015-08-12 23:00:24 +08:00
|
|
|
if (dev->num_crtcs > 0) {
|
2015-08-12 23:00:31 +08:00
|
|
|
seq = drm_vblank_count_and_time(dev, pipe, &now);
|
2012-10-17 06:48:40 +08:00
|
|
|
} else {
|
|
|
|
seq = 0;
|
2012-10-24 02:53:26 +08:00
|
|
|
|
|
|
|
now = get_drm_timestamp();
|
2012-10-17 06:48:40 +08:00
|
|
|
}
|
2015-08-12 23:00:31 +08:00
|
|
|
e->pipe = pipe;
|
2012-10-17 06:48:40 +08:00
|
|
|
send_vblank_event(dev, e, seq, &now);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_send_vblank_event);
|
|
|
|
|
2014-12-15 21:47:18 +08:00
|
|
|
/**
|
|
|
|
* 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);
|
|
|
|
|
2014-02-20 03:29:49 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_enable - enable the vblank interrupt on a CRTC
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2015-08-12 23:00:33 +08:00
|
|
|
*
|
|
|
|
* Returns:
|
|
|
|
* Zero on success or a negative error code on failure.
|
2014-02-20 03:29:49 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
static int drm_vblank_enable(struct drm_device *dev, unsigned int pipe)
|
2014-02-20 03:29:49 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-02-20 03:29:49 +08:00
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
assert_spin_locked(&dev->vbl_lock);
|
|
|
|
|
|
|
|
spin_lock(&dev->vblank_time_lock);
|
|
|
|
|
2014-08-06 19:49:50 +08:00
|
|
|
if (!vblank->enabled) {
|
2014-05-21 21:11:33 +08:00
|
|
|
/*
|
|
|
|
* Enable vblank irqs under vblank_time_lock protection.
|
2014-02-20 03:29:49 +08:00
|
|
|
* 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.
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
ret = dev->driver->enable_vblank(dev, pipe);
|
|
|
|
DRM_DEBUG("enabling vblank on crtc %u, ret: %d\n", pipe, ret);
|
2014-02-20 03:29:49 +08:00
|
|
|
if (ret)
|
2014-08-06 19:49:50 +08:00
|
|
|
atomic_dec(&vblank->refcount);
|
2014-02-20 03:29:49 +08:00
|
|
|
else {
|
2014-08-06 19:49:50 +08:00
|
|
|
vblank->enabled = true;
|
2015-09-15 03:43:47 +08:00
|
|
|
drm_update_vblank_count(dev, pipe, 0);
|
2014-02-20 03:29:49 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock(&dev->vblank_time_lock);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_get - get a reference count on vblank events
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: index of CRTC to own
|
2008-10-01 03:14:26 +08:00
|
|
|
*
|
|
|
|
* Acquire a reference count on vblank events to avoid having them disabled
|
|
|
|
* while in use.
|
|
|
|
*
|
2014-05-15 21:32:12 +08:00
|
|
|
* This is the legacy version of drm_crtc_vblank_get().
|
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* Returns:
|
2015-08-12 23:00:33 +08:00
|
|
|
* Zero on success or a negative error code on failure.
|
2008-10-01 03:14:26 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
int drm_vblank_get(struct drm_device *dev, unsigned int pipe)
|
2008-10-01 03:14:26 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2013-08-21 09:51:12 +08:00
|
|
|
unsigned long irqflags;
|
2008-10-01 03:14:26 +08:00
|
|
|
int ret = 0;
|
|
|
|
|
2015-03-30 18:50:50 +08:00
|
|
|
if (!dev->num_crtcs)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
spin_lock_irqsave(&dev->vbl_lock, irqflags);
|
|
|
|
/* Going from 0->1 means we have to enable interrupts again */
|
2014-08-06 19:49:50 +08:00
|
|
|
if (atomic_add_return(1, &vblank->refcount) == 1) {
|
2015-08-12 23:00:31 +08:00
|
|
|
ret = drm_vblank_enable(dev, pipe);
|
2009-11-09 12:51:22 +08:00
|
|
|
} else {
|
2014-08-06 19:49:50 +08:00
|
|
|
if (!vblank->enabled) {
|
|
|
|
atomic_dec(&vblank->refcount);
|
2009-11-09 12:51:22 +08:00
|
|
|
ret = -EINVAL;
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_get);
|
|
|
|
|
2014-05-15 21:32:12 +08:00
|
|
|
/**
|
|
|
|
* 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.
|
|
|
|
*
|
2015-02-13 23:53:31 +08:00
|
|
|
* This is the native kms version of drm_vblank_get().
|
2014-05-15 21:32:12 +08:00
|
|
|
*
|
|
|
|
* Returns:
|
2015-08-12 23:00:33 +08:00
|
|
|
* Zero on success or a negative error code on failure.
|
2014-05-15 21:32:12 +08:00
|
|
|
*/
|
|
|
|
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);
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
/**
|
2015-08-12 23:00:31 +08:00
|
|
|
* drm_vblank_put - release ownership of vblank events
|
2008-10-01 03:14:26 +08:00
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: index of CRTC to release
|
2008-10-01 03:14:26 +08:00
|
|
|
*
|
|
|
|
* 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, ×tamp).
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.
|
2014-05-15 21:32:12 +08:00
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_vblank_put().
|
2008-10-01 03:14:26 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_vblank_put(struct drm_device *dev, unsigned int pipe)
|
2008-10-01 03:14:26 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-08-06 19:49:50 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-11-08 23:16:19 +08:00
|
|
|
return;
|
2009-02-19 22:48:22 +08:00
|
|
|
|
2015-08-12 23:00:30 +08:00
|
|
|
if (WARN_ON(atomic_read(&vblank->refcount) == 0))
|
2014-08-07 01:16:59 +08:00
|
|
|
return;
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
/* Last user schedules interrupt disable */
|
2014-08-06 19:49:53 +08:00
|
|
|
if (atomic_dec_and_test(&vblank->refcount)) {
|
2014-09-10 23:36:08 +08:00
|
|
|
if (drm_vblank_offdelay == 0)
|
|
|
|
return;
|
|
|
|
else if (dev->vblank_disable_immediate || drm_vblank_offdelay < 0)
|
2014-08-06 19:49:53 +08:00
|
|
|
vblank_disable_fn((unsigned long)vblank);
|
2014-09-10 23:36:08 +08:00
|
|
|
else
|
2014-08-06 19:49:53 +08:00
|
|
|
mod_timer(&vblank->disable_timer,
|
|
|
|
jiffies + ((drm_vblank_offdelay * HZ)/1000));
|
|
|
|
}
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_put);
|
|
|
|
|
2014-05-15 21:32:12 +08:00
|
|
|
/**
|
|
|
|
* 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);
|
|
|
|
|
2014-07-26 05:34:03 +08:00
|
|
|
/**
|
|
|
|
* drm_wait_one_vblank - wait for one vblank
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2014-07-26 05:34:03 +08:00
|
|
|
*
|
2015-09-15 03:43:42 +08:00
|
|
|
* 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.
|
2014-07-26 05:34:03 +08:00
|
|
|
* due to lack of driver support or because the crtc is off.
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_wait_one_vblank(struct drm_device *dev, unsigned int pipe)
|
2014-07-26 05:34:03 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-07-26 05:34:03 +08:00
|
|
|
int ret;
|
|
|
|
u32 last;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2015-08-12 23:00:30 +08:00
|
|
|
return;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
ret = drm_vblank_get(dev, pipe);
|
|
|
|
if (WARN(ret, "vblank not available on crtc %i, ret=%i\n", pipe, ret))
|
2014-07-26 05:34:03 +08:00
|
|
|
return;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
last = drm_vblank_count(dev, pipe);
|
2014-07-26 05:34:03 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
ret = wait_event_timeout(vblank->queue,
|
|
|
|
last != drm_vblank_count(dev, pipe),
|
2014-07-26 05:34:03 +08:00
|
|
|
msecs_to_jiffies(100));
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
WARN(ret == 0, "vblank wait timed out on crtc %i\n", pipe);
|
2014-07-26 05:34:03 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2014-07-26 05:34:03 +08:00
|
|
|
}
|
|
|
|
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);
|
|
|
|
|
2012-10-17 06:48:40 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_off - disable vblank events on a CRTC
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2012-10-17 06:48:40 +08:00
|
|
|
*
|
2014-05-08 22:41:51 +08:00
|
|
|
* 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.
|
2014-05-15 21:32:12 +08:00
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_vblank_off().
|
2012-10-17 06:48:40 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_vblank_off(struct drm_device *dev, unsigned int pipe)
|
2009-11-09 12:51:22 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2011-04-27 14:10:57 +08:00
|
|
|
struct drm_pending_vblank_event *e, *t;
|
|
|
|
struct timeval now;
|
2009-11-09 12:51:22 +08:00
|
|
|
unsigned long irqflags;
|
2011-04-27 14:10:57 +08:00
|
|
|
unsigned int seq;
|
2009-11-09 12:51:22 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return;
|
|
|
|
|
2014-08-06 19:49:51 +08:00
|
|
|
spin_lock_irqsave(&dev->event_lock, irqflags);
|
|
|
|
|
|
|
|
spin_lock(&dev->vbl_lock);
|
2015-08-12 23:00:31 +08:00
|
|
|
vblank_disable_and_save(dev, pipe);
|
2014-08-06 19:49:50 +08:00
|
|
|
wake_up(&vblank->queue);
|
2011-04-27 14:10:57 +08:00
|
|
|
|
2014-08-06 19:49:51 +08:00
|
|
|
/*
|
|
|
|
* 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);
|
|
|
|
|
2011-04-27 14:10:57 +08:00
|
|
|
/* Send any queued vblank events, lest the natives grow disquiet */
|
2015-08-12 23:00:31 +08:00
|
|
|
seq = drm_vblank_count_and_time(dev, pipe, &now);
|
2012-11-02 19:30:50 +08:00
|
|
|
|
2011-04-27 14:10:57 +08:00
|
|
|
list_for_each_entry_safe(e, t, &dev->vblank_event_list, base.link) {
|
2015-08-12 23:00:31 +08:00
|
|
|
if (e->pipe != pipe)
|
2011-04-27 14:10:57 +08:00
|
|
|
continue;
|
2015-10-10 03:57:36 +08:00
|
|
|
DRM_DEBUG("Sending premature vblank event on disable: "
|
|
|
|
"wanted %d, current %d\n",
|
2011-04-27 14:10:57 +08:00
|
|
|
e->event.sequence, seq);
|
2012-10-17 06:48:40 +08:00
|
|
|
list_del(&e->base.link);
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2012-10-17 06:48:40 +08:00
|
|
|
send_vblank_event(dev, e, seq, &now);
|
2011-04-27 14:10:57 +08:00
|
|
|
}
|
2014-08-06 19:49:51 +08:00
|
|
|
spin_unlock_irqrestore(&dev->event_lock, irqflags);
|
2009-11-09 12:51:22 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_off);
|
|
|
|
|
2014-05-15 21:32:12 +08:00
|
|
|
/**
|
|
|
|
* 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);
|
|
|
|
|
2015-02-14 04:03:42 +08:00
|
|
|
/**
|
|
|
|
* drm_crtc_vblank_reset - reset vblank state to off on a CRTC
|
2015-09-25 00:35:30 +08:00
|
|
|
* @crtc: CRTC in question
|
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.
|
|
|
|
*/
|
2015-09-25 00:35:30 +08:00
|
|
|
void drm_crtc_vblank_reset(struct drm_crtc *crtc)
|
2015-02-14 04:03:42 +08:00
|
|
|
{
|
2015-09-25 00:35:30 +08:00
|
|
|
struct drm_device *dev = crtc->dev;
|
2015-02-14 04:03:42 +08:00
|
|
|
unsigned long irqflags;
|
2015-09-25 00:35:30 +08:00
|
|
|
unsigned int pipe = drm_crtc_index(crtc);
|
2015-09-15 03:43:42 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
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);
|
|
|
|
|
2014-02-20 03:29:49 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_on - enable vblank events on a CRTC
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2014-05-08 22:41:51 +08:00
|
|
|
*
|
|
|
|
* This functions restores the vblank interrupt state captured with
|
|
|
|
* drm_vblank_off() again. Note that calls to drm_vblank_on() and
|
2014-10-20 22:53:13 +08:00
|
|
|
* drm_vblank_off() can be unbalanced and so can also be unconditionally called
|
2014-05-08 22:41:51 +08:00
|
|
|
* in driver load code to reflect the current hardware state of the crtc.
|
2014-05-15 21:32:12 +08:00
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_vblank_on().
|
2014-02-20 03:29:49 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_vblank_on(struct drm_device *dev, unsigned int pipe)
|
2014-02-20 03:29:49 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-02-20 03:29:49 +08:00
|
|
|
unsigned long irqflags;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return;
|
|
|
|
|
2014-02-20 03:29:49 +08:00
|
|
|
spin_lock_irqsave(&dev->vbl_lock, irqflags);
|
2014-08-06 19:49:44 +08:00
|
|
|
/* Drop our private "prevent drm_vblank_get" refcount */
|
2014-08-06 19:49:50 +08:00
|
|
|
if (vblank->inmodeset) {
|
|
|
|
atomic_dec(&vblank->refcount);
|
|
|
|
vblank->inmodeset = 0;
|
2014-08-06 19:49:44 +08:00
|
|
|
}
|
2014-08-06 19:49:49 +08:00
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
drm_reset_vblank_timestamp(dev, pipe);
|
|
|
|
|
2014-08-06 19:49:56 +08:00
|
|
|
/*
|
|
|
|
* 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))
|
2015-08-12 23:00:31 +08:00
|
|
|
WARN_ON(drm_vblank_enable(dev, pipe));
|
2014-02-20 03:29:49 +08:00
|
|
|
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_on);
|
|
|
|
|
2014-05-15 21:32:12 +08:00
|
|
|
/**
|
|
|
|
* 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
|
2014-10-20 22:53:13 +08:00
|
|
|
* drm_vblank_off() can be unbalanced and so can also be unconditionally called
|
2014-05-15 21:32:12 +08:00
|
|
|
* 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);
|
|
|
|
|
2008-11-08 06:05:41 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_pre_modeset - account for vblanks across mode sets
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2008-11-08 06:05:41 +08:00
|
|
|
*
|
|
|
|
* Account for vblank events across mode setting events, which will likely
|
|
|
|
* reset the hardware frame counter.
|
2014-05-08 22:41:51 +08:00
|
|
|
*
|
|
|
|
* 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.
|
2008-11-08 06:05:41 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_vblank_pre_modeset(struct drm_device *dev, unsigned int pipe)
|
2008-11-08 06:05:41 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2014-08-06 19:49:50 +08:00
|
|
|
|
2013-05-21 14:23:43 +08:00
|
|
|
/* vblank is not initialized (IRQ not installed ?), or has been freed */
|
2010-01-07 22:39:13 +08:00
|
|
|
if (!dev->num_crtcs)
|
|
|
|
return;
|
2014-08-07 01:16:59 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return;
|
|
|
|
|
2008-11-08 06:05:41 +08:00
|
|
|
/*
|
|
|
|
* 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.
|
|
|
|
*/
|
2014-08-06 19:49:50 +08:00
|
|
|
if (!vblank->inmodeset) {
|
|
|
|
vblank->inmodeset = 0x1;
|
2015-08-12 23:00:31 +08:00
|
|
|
if (drm_vblank_get(dev, pipe) == 0)
|
2014-08-06 19:49:50 +08:00
|
|
|
vblank->inmodeset |= 0x2;
|
2008-11-08 06:05:41 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_pre_modeset);
|
|
|
|
|
2014-05-08 22:41:51 +08:00
|
|
|
/**
|
|
|
|
* drm_vblank_post_modeset - undo drm_vblank_pre_modeset changes
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: CRTC index
|
2014-05-08 22:41:51 +08:00
|
|
|
*
|
|
|
|
* This function again drops the temporary vblank reference acquired in
|
|
|
|
* drm_vblank_pre_modeset.
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe)
|
2008-11-08 06:05:41 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2008-11-08 06:05:41 +08:00
|
|
|
unsigned long irqflags;
|
|
|
|
|
2013-05-21 14:23:43 +08:00
|
|
|
/* vblank is not initialized (IRQ not installed ?), or has been freed */
|
|
|
|
if (!dev->num_crtcs)
|
|
|
|
return;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2015-08-12 23:00:30 +08:00
|
|
|
return;
|
|
|
|
|
2014-08-06 19:49:50 +08:00
|
|
|
if (vblank->inmodeset) {
|
2008-11-08 06:05:41 +08:00
|
|
|
spin_lock_irqsave(&dev->vbl_lock, irqflags);
|
2013-10-04 19:53:33 +08:00
|
|
|
dev->vblank_disable_allowed = true;
|
2008-11-08 06:05:41 +08:00
|
|
|
spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
|
2009-02-19 22:48:22 +08:00
|
|
|
|
2014-08-06 19:49:50 +08:00
|
|
|
if (vblank->inmodeset & 0x2)
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2009-02-19 22:48:22 +08:00
|
|
|
|
2014-08-06 19:49:50 +08:00
|
|
|
vblank->inmodeset = 0;
|
2008-11-08 06:05:41 +08:00
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_post_modeset);
|
|
|
|
|
2014-05-08 22:41:51 +08:00
|
|
|
/*
|
2008-10-01 03:14:26 +08:00
|
|
|
* 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;
|
2015-08-12 23:00:31 +08:00
|
|
|
unsigned int pipe;
|
2008-10-01 03:14:26 +08:00
|
|
|
|
|
|
|
/* If drm_vblank_init() hasn't been called yet, just no-op */
|
|
|
|
if (!dev->num_crtcs)
|
2012-05-17 19:27:21 +08:00
|
|
|
return 0;
|
2008-10-01 03:14:26 +08:00
|
|
|
|
2012-05-30 06:58:09 +08:00
|
|
|
/* KMS drivers handle this internally */
|
|
|
|
if (drm_core_check_feature(dev, DRIVER_MODESET))
|
|
|
|
return 0;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
pipe = modeset->crtc;
|
|
|
|
if (pipe >= dev->num_crtcs)
|
2012-05-17 19:27:21 +08:00
|
|
|
return -EINVAL;
|
2008-10-01 03:14:26 +08:00
|
|
|
|
|
|
|
switch (modeset->cmd) {
|
|
|
|
case _DRM_PRE_MODESET:
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_pre_modeset(dev, pipe);
|
2008-10-01 03:14:26 +08:00
|
|
|
break;
|
|
|
|
case _DRM_POST_MODESET:
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_post_modeset(dev, pipe);
|
2008-10-01 03:14:26 +08:00
|
|
|
break;
|
|
|
|
default:
|
2012-05-17 19:27:21 +08:00
|
|
|
return -EINVAL;
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
|
2012-05-17 19:27:21 +08:00
|
|
|
return 0;
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
static int drm_queue_vblank_event(struct drm_device *dev, unsigned int pipe,
|
2009-09-12 02:33:34 +08:00
|
|
|
union drm_wait_vblank *vblwait,
|
|
|
|
struct drm_file *file_priv)
|
|
|
|
{
|
2014-08-06 19:49:52 +08:00
|
|
|
struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
|
2009-09-12 02:33:34 +08:00
|
|
|
struct drm_pending_vblank_event *e;
|
|
|
|
struct timeval now;
|
|
|
|
unsigned long flags;
|
|
|
|
unsigned int seq;
|
2010-12-02 03:41:31 +08:00
|
|
|
int ret;
|
2009-09-12 02:33:34 +08:00
|
|
|
|
2014-12-10 20:03:40 +08:00
|
|
|
e = kzalloc(sizeof(*e), GFP_KERNEL);
|
2010-12-02 03:41:31 +08:00
|
|
|
if (e == NULL) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto err_put;
|
|
|
|
}
|
2009-09-12 02:33:34 +08:00
|
|
|
|
|
|
|
e->pipe = pipe;
|
2010-07-02 07:48:09 +08:00
|
|
|
e->base.pid = current->pid;
|
2009-09-12 02:33:34 +08:00
|
|
|
e->event.base.type = DRM_EVENT_VBLANK;
|
2014-12-10 20:03:40 +08:00
|
|
|
e->event.base.length = sizeof(e->event);
|
2009-09-12 02:33:34 +08:00
|
|
|
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);
|
|
|
|
|
2014-08-06 19:49:52 +08:00
|
|
|
/*
|
|
|
|
* 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;
|
|
|
|
}
|
|
|
|
|
2014-12-10 20:03:40 +08:00
|
|
|
if (file_priv->event_space < sizeof(e->event)) {
|
2010-12-02 03:41:31 +08:00
|
|
|
ret = -EBUSY;
|
|
|
|
goto err_unlock;
|
2009-09-12 02:33:34 +08:00
|
|
|
}
|
|
|
|
|
2014-12-10 20:03:40 +08:00
|
|
|
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, ×tamp).
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);
|
|
|
|
|
2009-09-12 02:33:34 +08:00
|
|
|
if ((vblwait->request.type & _DRM_VBLANK_NEXTONMISS) &&
|
|
|
|
(seq - vblwait->request.sequence) <= (1 << 23)) {
|
|
|
|
vblwait->request.sequence = seq + 1;
|
2009-11-10 16:21:25 +08:00
|
|
|
vblwait->reply.sequence = vblwait->request.sequence;
|
2009-09-12 02:33:34 +08:00
|
|
|
}
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
DRM_DEBUG("event on vblank count %d, current %d, crtc %u\n",
|
2009-09-12 02:33:34 +08:00
|
|
|
vblwait->request.sequence, seq, pipe);
|
|
|
|
|
2010-07-02 07:48:09 +08:00
|
|
|
trace_drm_vblank_event_queued(current->pid, pipe,
|
|
|
|
vblwait->request.sequence);
|
|
|
|
|
2009-09-12 02:33:34 +08:00
|
|
|
e->event.sequence = vblwait->request.sequence;
|
|
|
|
if ((seq - vblwait->request.sequence) <= (1 << 23)) {
|
2010-12-09 13:35:40 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2012-10-17 06:48:40 +08:00
|
|
|
send_vblank_event(dev, e, seq, &now);
|
2010-12-10 23:00:19 +08:00
|
|
|
vblwait->reply.sequence = seq;
|
2009-09-12 02:33:34 +08:00
|
|
|
} else {
|
2011-11-01 01:11:57 +08:00
|
|
|
/* drm_handle_vblank_events will call drm_vblank_put */
|
2009-09-12 02:33:34 +08:00
|
|
|
list_add_tail(&e->base.link, &dev->vblank_event_list);
|
2010-12-10 23:00:19 +08:00
|
|
|
vblwait->reply.sequence = vblwait->request.sequence;
|
2009-09-12 02:33:34 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&dev->event_lock, flags);
|
|
|
|
|
|
|
|
return 0;
|
2010-12-02 03:41:31 +08:00
|
|
|
|
|
|
|
err_unlock:
|
|
|
|
spin_unlock_irqrestore(&dev->event_lock, flags);
|
|
|
|
kfree(e);
|
|
|
|
err_put:
|
2010-12-09 13:35:40 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2010-12-02 03:41:31 +08:00
|
|
|
return ret;
|
2009-09-12 02:33:34 +08:00
|
|
|
}
|
|
|
|
|
2014-05-08 22:41:51 +08:00
|
|
|
/*
|
2005-04-17 06:20:36 +08:00
|
|
|
* Wait for VBLANK.
|
|
|
|
*
|
|
|
|
* \param inode device inode.
|
2007-08-25 18:23:09 +08:00
|
|
|
* \param file_priv DRM file private.
|
2005-04-17 06:20:36 +08:00
|
|
|
* \param cmd command.
|
|
|
|
* \param data user argument, pointing to a drm_wait_vblank structure.
|
|
|
|
* \return zero on success or a negative number on failure.
|
|
|
|
*
|
2009-01-28 13:19:41 +08:00
|
|
|
* This function enables the vblank interrupt on the pipe requested, then
|
|
|
|
* sleeps waiting for the requested sequence number to occur, and drops
|
2014-05-08 22:41:51 +08:00
|
|
|
* the vblank interrupt refcount afterwards. (vblank IRQ disable follows that
|
2009-01-28 13:19:41 +08:00
|
|
|
* after a timeout with no further vblank waits scheduled).
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2008-10-01 03:14:26 +08:00
|
|
|
int drm_wait_vblank(struct drm_device *dev, void *data,
|
|
|
|
struct drm_file *file_priv)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2014-08-06 19:49:50 +08:00
|
|
|
struct drm_vblank_crtc *vblank;
|
2007-09-03 10:06:45 +08:00
|
|
|
union drm_wait_vblank *vblwait = data;
|
2012-05-17 19:27:21 +08:00
|
|
|
int ret;
|
2015-08-12 23:00:31 +08:00
|
|
|
unsigned int flags, seq, pipe, high_pipe;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-08-28 21:02:37 +08:00
|
|
|
if (!dev->irq_enabled)
|
|
|
|
return -EINVAL;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-01-28 13:19:41 +08:00
|
|
|
if (vblwait->request.type & _DRM_VBLANK_SIGNAL)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2007-09-03 10:06:45 +08:00
|
|
|
if (vblwait->request.type &
|
2011-03-19 05:58:04 +08:00
|
|
|
~(_DRM_VBLANK_TYPES_MASK | _DRM_VBLANK_FLAGS_MASK |
|
|
|
|
_DRM_VBLANK_HIGH_CRTC_MASK)) {
|
2006-10-24 20:24:38 +08:00
|
|
|
DRM_ERROR("Unsupported type value 0x%x, supported mask 0x%x\n",
|
2007-09-03 10:06:45 +08:00
|
|
|
vblwait->request.type,
|
2011-03-19 05:58:04 +08:00
|
|
|
(_DRM_VBLANK_TYPES_MASK | _DRM_VBLANK_FLAGS_MASK |
|
|
|
|
_DRM_VBLANK_HIGH_CRTC_MASK));
|
2006-10-24 20:24:38 +08:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2007-09-03 10:06:45 +08:00
|
|
|
flags = vblwait->request.type & _DRM_VBLANK_FLAGS_MASK;
|
2015-08-12 23:00:31 +08:00
|
|
|
high_pipe = (vblwait->request.type & _DRM_VBLANK_HIGH_CRTC_MASK);
|
|
|
|
if (high_pipe)
|
|
|
|
pipe = high_pipe >> _DRM_VBLANK_HIGH_CRTC_SHIFT;
|
2011-03-19 05:58:04 +08:00
|
|
|
else
|
2015-08-12 23:00:31 +08:00
|
|
|
pipe = flags & _DRM_VBLANK_SECONDARY ? 1 : 0;
|
|
|
|
if (pipe >= dev->num_crtcs)
|
2006-10-24 20:24:38 +08:00
|
|
|
return -EINVAL;
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
vblank = &dev->vblank[pipe];
|
2014-08-06 19:49:50 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
ret = drm_vblank_get(dev, pipe);
|
2008-10-01 03:14:26 +08:00
|
|
|
if (ret) {
|
2009-08-09 10:24:01 +08:00
|
|
|
DRM_DEBUG("failed to acquire vblank counter, %d\n", ret);
|
2008-10-01 03:14:26 +08:00
|
|
|
return ret;
|
|
|
|
}
|
2015-08-12 23:00:31 +08:00
|
|
|
seq = drm_vblank_count(dev, pipe);
|
2006-10-24 20:24:38 +08:00
|
|
|
|
2007-09-03 10:06:45 +08:00
|
|
|
switch (vblwait->request.type & _DRM_VBLANK_TYPES_MASK) {
|
2005-04-17 06:20:36 +08:00
|
|
|
case _DRM_VBLANK_RELATIVE:
|
2007-09-03 10:06:45 +08:00
|
|
|
vblwait->request.sequence += seq;
|
|
|
|
vblwait->request.type &= ~_DRM_VBLANK_RELATIVE;
|
2005-04-17 06:20:36 +08:00
|
|
|
case _DRM_VBLANK_ABSOLUTE:
|
|
|
|
break;
|
|
|
|
default:
|
2008-10-01 03:14:26 +08:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto done;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2011-11-01 01:11:57 +08:00
|
|
|
if (flags & _DRM_VBLANK_EVENT) {
|
|
|
|
/* must hold on to the vblank ref until the event fires
|
|
|
|
* drm_vblank_put will be called asynchronously
|
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
return drm_queue_vblank_event(dev, pipe, vblwait, file_priv);
|
2011-11-01 01:11:57 +08:00
|
|
|
}
|
2009-09-12 02:33:34 +08:00
|
|
|
|
2006-10-24 21:34:18 +08:00
|
|
|
if ((flags & _DRM_VBLANK_NEXTONMISS) &&
|
2007-09-03 10:06:45 +08:00
|
|
|
(seq - vblwait->request.sequence) <= (1<<23)) {
|
|
|
|
vblwait->request.sequence = seq + 1;
|
2006-10-24 21:34:18 +08:00
|
|
|
}
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
DRM_DEBUG("waiting on vblank count %d, crtc %u\n",
|
|
|
|
vblwait->request.sequence, pipe);
|
2014-08-06 19:49:50 +08:00
|
|
|
vblank->last_wait = vblwait->request.sequence;
|
|
|
|
DRM_WAIT_ON(ret, vblank->queue, 3 * HZ,
|
2015-08-12 23:00:31 +08:00
|
|
|
(((drm_vblank_count(dev, pipe) -
|
2009-01-28 13:19:41 +08:00
|
|
|
vblwait->request.sequence) <= (1 << 23)) ||
|
2014-08-06 19:49:50 +08:00
|
|
|
!vblank->enabled ||
|
2009-01-28 13:19:41 +08:00
|
|
|
!dev->irq_enabled));
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2009-01-28 13:19:41 +08:00
|
|
|
if (ret != -EINTR) {
|
|
|
|
struct timeval now;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
vblwait->reply.sequence = drm_vblank_count_and_time(dev, pipe, &now);
|
2009-01-28 13:19:41 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2009-01-28 13:19:41 +08:00
|
|
|
DRM_DEBUG("returning %d to client\n",
|
|
|
|
vblwait->reply.sequence);
|
2005-04-17 06:20:36 +08:00
|
|
|
} else {
|
2009-01-28 13:19:41 +08:00
|
|
|
DRM_DEBUG("vblank wait interrupted by signal\n");
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
done:
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2005-04-17 06:20:36 +08:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
static void drm_handle_vblank_events(struct drm_device *dev, unsigned int pipe)
|
2009-09-12 02:33:34 +08:00
|
|
|
{
|
|
|
|
struct drm_pending_vblank_event *e, *t;
|
|
|
|
struct timeval now;
|
|
|
|
unsigned int seq;
|
|
|
|
|
2014-08-06 19:49:51 +08:00
|
|
|
assert_spin_locked(&dev->event_lock);
|
2009-09-12 02:33:34 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
seq = drm_vblank_count_and_time(dev, pipe, &now);
|
2009-09-12 02:33:34 +08:00
|
|
|
|
|
|
|
list_for_each_entry_safe(e, t, &dev->vblank_event_list, base.link) {
|
2015-08-12 23:00:31 +08:00
|
|
|
if (e->pipe != pipe)
|
2009-09-12 02:33:34 +08:00
|
|
|
continue;
|
|
|
|
if ((seq - e->event.sequence) > (1<<23))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
DRM_DEBUG("vblank event on %d, current %d\n",
|
|
|
|
e->event.sequence, seq);
|
|
|
|
|
2012-10-17 06:48:40 +08:00
|
|
|
list_del(&e->base.link);
|
2015-08-12 23:00:31 +08:00
|
|
|
drm_vblank_put(dev, pipe);
|
2012-10-17 06:48:40 +08:00
|
|
|
send_vblank_event(dev, e, seq, &now);
|
2009-09-12 02:33:34 +08:00
|
|
|
}
|
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
trace_drm_vblank_event(pipe, seq);
|
2009-09-12 02:33:34 +08:00
|
|
|
}
|
|
|
|
|
2008-10-01 03:14:26 +08:00
|
|
|
/**
|
|
|
|
* drm_handle_vblank - handle a vblank event
|
|
|
|
* @dev: DRM device
|
2015-08-12 23:00:31 +08:00
|
|
|
* @pipe: index of CRTC where this event occurred
|
2008-10-01 03:14:26 +08:00
|
|
|
*
|
|
|
|
* Drivers should call this routine in their vblank interrupt handlers to
|
|
|
|
* update the vblank counter and send any signals that may be pending.
|
2014-12-15 21:47:18 +08:00
|
|
|
*
|
|
|
|
* This is the legacy version of drm_crtc_handle_vblank().
|
2008-10-01 03:14:26 +08:00
|
|
|
*/
|
2015-08-12 23:00:31 +08:00
|
|
|
bool drm_handle_vblank(struct drm_device *dev, unsigned int pipe)
|
2008-10-01 03:14:26 +08:00
|
|
|
{
|
2015-08-12 23:00:31 +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, ×tamp).
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;
|
|
|
|
|
2015-02-22 22:11:20 +08:00
|
|
|
if (WARN_ON_ONCE(!dev->num_crtcs))
|
drm/i915: Suppress spurious vblank interrupts
Hugh Dickins found that characters in xterm were going missing and oft
delayed. Being the curious type, he managed to associate this with the
new high-precision vblank patches; disabling these he found, restored
the orderliness of his characters.
The oddness begins when one realised that Hugh was not using vblanks at
all on his system (fvwm and some xterms). Instead, all he had to go on
were warning of a pipe underrun, curiously enough at around 60Hz. He
poked and found that in addition to the underrun warning, the hardware
was flagging the start of a new frame, a vblank, which in turn was
kicking off the pending vblank processing code.
There is little we can do for the underruns on Hugh's machine, a
Crestline [965GM], which must have its FIFO watermarks set to 8.
However, we do not need to process the vblank if we know that they are
disabled...
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
2011-01-31 18:48:04 +08:00
|
|
|
return false;
|
2009-09-12 02:33:34 +08:00
|
|
|
|
2015-08-12 23:00:31 +08:00
|
|
|
if (WARN_ON(pipe >= dev->num_crtcs))
|
2014-08-07 01:16:59 +08:00
|
|
|
return false;
|
|
|
|
|
2014-08-06 19:49:51 +08:00
|
|
|
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, ×tamp).
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.
|
|
|
|
*/
|
2014-08-06 19:49:51 +08:00
|
|
|
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, ×tamp).
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. */
|
2014-08-06 19:49:50 +08:00
|
|
|
if (!vblank->enabled) {
|
2014-08-06 19:49:51 +08:00
|
|
|
spin_unlock(&dev->vblank_time_lock);
|
|
|
|
spin_unlock_irqrestore(&dev->event_lock, irqflags);
|
drm/i915: Suppress spurious vblank interrupts
Hugh Dickins found that characters in xterm were going missing and oft
delayed. Being the curious type, he managed to associate this with the
new high-precision vblank patches; disabling these he found, restored
the orderliness of his characters.
The oddness begins when one realised that Hugh was not using vblanks at
all on his system (fvwm and some xterms). Instead, all he had to go on
were warning of a pipe underrun, curiously enough at around 60Hz. He
poked and found that in addition to the underrun warning, the hardware
was flagging the start of a new frame, a vblank, which in turn was
kicking off the pending vblank processing code.
There is little we can do for the underruns on Hugh's machine, a
Crestline [965GM], which must have its FIFO watermarks set to 8.
However, we do not need to process the vblank if we know that they are
disabled...
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
2011-01-31 18:48:04 +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, ×tamp).
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
|
|
|
}
|
|
|
|
|
2015-09-15 03:43:51 +08:00
|
|
|
drm_update_vblank_count(dev, pipe, 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, ×tamp).
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
|
|
|
|
2014-08-06 19:49:51 +08:00
|
|
|
spin_unlock(&dev->vblank_time_lock);
|
|
|
|
|
2014-08-06 19:49:50 +08:00
|
|
|
wake_up(&vblank->queue);
|
2015-08-12 23:00:31 +08:00
|
|
|
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, ×tamp).
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
|
|
|
|
2014-08-06 19:49:51 +08:00
|
|
|
spin_unlock_irqrestore(&dev->event_lock, irqflags);
|
|
|
|
|
drm/i915: Suppress spurious vblank interrupts
Hugh Dickins found that characters in xterm were going missing and oft
delayed. Being the curious type, he managed to associate this with the
new high-precision vblank patches; disabling these he found, restored
the orderliness of his characters.
The oddness begins when one realised that Hugh was not using vblanks at
all on his system (fvwm and some xterms). Instead, all he had to go on
were warning of a pipe underrun, curiously enough at around 60Hz. He
poked and found that in addition to the underrun warning, the hardware
was flagging the start of a new frame, a vblank, which in turn was
kicking off the pending vblank processing code.
There is little we can do for the underruns on Hugh's machine, a
Crestline [965GM], which must have its FIFO watermarks set to 8.
However, we do not need to process the vblank if we know that they are
disabled...
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
2011-01-31 18:48:04 +08:00
|
|
|
return true;
|
2008-10-01 03:14:26 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_handle_vblank);
|
2014-12-15 21:47:18 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* 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);
|
2015-09-30 21:46:48 +08:00
|
|
|
|
|
|
|
/**
|
|
|
|
* drm_vblank_no_hw_counter - "No hw counter" implementation of .get_vblank_counter()
|
|
|
|
* @dev: DRM device
|
|
|
|
* @pipe: CRTC for which to read the counter
|
|
|
|
*
|
|
|
|
* Drivers can plug this into the .get_vblank_counter() function if
|
|
|
|
* there is no useable hardware frame counter available.
|
|
|
|
*
|
|
|
|
* Returns:
|
|
|
|
* 0
|
|
|
|
*/
|
|
|
|
u32 drm_vblank_no_hw_counter(struct drm_device *dev, unsigned int pipe)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(drm_vblank_no_hw_counter);
|