We currently using GuC as a proxy to the hardware. When Guc is used in
such mode, it consumes the bit 20 of the hw_id to indicate that the
workload was submitted by proxy.
So far we probably haven't seen the issue because we need to allocate
1048576+ contexts to hit this issue. Still, we should avoid allocating
the hw_id on that bit and restriction to bits [0:19] (i.e 20bits
instead of 21).
v2: Leave the max hw_id computation in i915_gem_context.c (Michel)
v3: Be consistent on if/else usage (Chris)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
BSpec: 1237
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180602112946.30803-2-lionel.g.landwerlin@intel.com
As we store the intel_context on the request (rq->hw_context), we can
simply compare that against the local intel_context for the
i915->kernel_context rather than using the rq->gem_context.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180601094002.13329-1-chris@chris-wilson.co.uk
We can reduce our exposure to random neutrinos by resting on the kernel
context having flushed out the user contexts to system memory and
beyond. The corollary is that we then we require two passes through the
idle handler to go to sleep, which on a truly idle system involves an
extra pass through the slow and irregular retire work handler.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180531082246.9763-1-chris@chris-wilson.co.uk
We were not very carefully checking to see if an older request on the
engine was an earlier switch-to-kernel-context before deciding to emit a
new switch. The end result would be that we could get into a permanent
loop of trying to emit a new request to perform the switch simply to
flush the existing switch.
What we need is a means of tracking the completion of each timeline
versus the kernel context, that is to detect if a more recent request
has been submitted that would result in a switch away from the kernel
context. To realise this, we need only to look in our syncmap on the
kernel context and check that we have synchronized against all active
rings.
v2: Since all ringbuffer clients currently share the same timeline, we do
have to use the gem_context to distinguish clients.
As a bonus, include all the tracing used to debug the death inside
suspend.
v3: Test, test, test. Construct a selftest to exercise and assert the
expected behaviour that multiple switch-to-contexts do not emit
redundant requests.
Reported-by: Mika Kuoppala <mika.kuoppala@intel.com>
Fixes: a89d1f921c ("drm/i915: Split i915_gem_timeline into individual timelines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180524081135.15278-1-chris@chris-wilson.co.uk
To ease the frequent and ugly pointer dance of
&request->gem_context->engine[request->engine->id] during request
submission, store that pointer as request->hw_context. One major
advantage that we will exploit later is that this decouples the logical
context state from the engine itself.
v2: Set mock_context->ops so we don't crash and burn in selftests.
Cleanups from Tvrtko.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180517212633.24934-3-chris@chris-wilson.co.uk
When switching to the kernel context, we force the switch to occur after
all currently active requests (so that we know the GPU won't switch
immediately away and the kernel context remains current as we work). To
do so we have to inspect all the timelines and add a fence from the
active work to queue our switch afterwards. We can use the tracked set
of active rings to shrink our search for active timelines.
v2: Use a local to shrink the list_for_each_entry()
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180515143149.4795-1-chris@chris-wilson.co.uk
We need to move to a more flexible timeline that doesn't assume one
fence context per engine, and so allow for a single timeline to be used
across a combination of engines. This means that preallocating a fence
context per engine is now a hindrance, and so we want to introduce the
singular timeline. From the code perspective, this has the notable
advantage of clearing up a lot of mirky semantics and some clumsy
pointer chasing.
By splitting the timeline up into a single entity rather than an array
of per-engine timelines, we can realise the goal of the previous patch
of tracking the timeline alongside the ring.
v2: Tweak wait_for_idle to stop the compiling thinking that ret may be
uninitialised.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180502163839.3248-2-chris@chris-wilson.co.uk
In the future, we want to move a request between engines. To achieve
this, we first realise that we have two timelines in effect here. The
first runs through the GTT is required for ordering vma access, which is
tracked currently by engine. The second is implied by sequential
execution of commands inside the ringbuffer. This timeline is one that
maps to userspace's expectations when submitting requests (i.e. given the
same context, batch A is executed before batch B). As the rings's
timelines map to userspace and the GTT timeline an implementation
detail, move the timeline from the GTT into the ring itself (per-context
in logical-ring-contexts/execlists, or a global per-engine timeline for
the shared ringbuffers in legacy submission.
The two timelines are still assumed to be equivalent at the moment (no
migrating requests between engines yet) and so we can simply move from
one to the other without adding extra ordering.
v2: Reinforce that one isn't allowed to mix the engine execution
timeline with the client timeline from userspace (on the ring).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180502163839.3248-1-chris@chris-wilson.co.uk
Make life easier in upcoming patches by moving the context_pin and
context_unpin vfuncs into inline helpers.
v2: Fixup mock_engine to mark the context as pinned on use.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180430131503.5375-2-chris@chris-wilson.co.uk
Today we only want to pass along the priority to engine->schedule(), but
in the future we want to have much more control over the various aspects
of the GPU during a context's execution, for example controlling the
frequency allowed. As we need an ever growing number of parameters for
scheduling, move those into a struct for convenience.
v2: Move the anonymous struct into its own function for legibility and
ye olde gcc.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180418184052.7129-3-chris@chris-wilson.co.uk
There are different kind of workarounds (those that modify registers that
live in the context image, those that modify global registers, those that
whitelist registers, etc...) and they have different requirements in terms
of where they are applied and how. Also, by splitting them apart, it should
be easier to decide where a new workaround should go.
v2:
- Add multiple MISSING_CASE
- Rebased
v3:
- Rename mmio_workarounds to gt_workarounds (Chris, Mika)
- Create empty placeholders for BDW and CHV GT WAs
- Rebased
v4: Rebased
v5:
- Rebased
- FORCE_TO_NONPRIV register exists since BDW, so make a path
for it to achieve universality, even if empty (Chris)
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[ickle: appease checkpatch]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/1523376767-18480-2-git-send-email-oscar.mateo@intel.com
Hardware may have specific restrictions on GuC WOPCM offset and size. On
Gen9, the value of the GuC WOPCM size register needs to be larger than the
value of GuC WOPCM offset register + a Gen9 specific offset (144KB) for
reserved GuC WOPCM. Fail to enforce such a restriction on GuC WOPCM size
will lead to GuC firmware execution failures. On the other hand, with
current static GuC WOPCM offset and size values (512KB for both offset and
size), the GuC WOPCM size verification will fail on Gen9 even if it can be
fixed by lowering the GuC WOPCM offset by calculating its value based on
HuC firmware size (which is likely less than 200KB on Gen9), so that we can
have a GuC WOPCM size value which is large enough to pass the GuC WOPCM
size check.
This patch updates the reserved GuC WOPCM size for RC6 context on Gen9 to
24KB to strictly align with the Gen9 GuC WOPCM layout. It also adds support
to verify the GuC WOPCM size aganist the Gen9 hardware restrictions. To
meet all above requirements, let's provide dynamic partitioning of the
WOPCM that will be based on platform specific HuC/GuC firmware sizes.
v2:
- Removed intel_wopcm_init (Ville/Sagar/Joonas)
- Renamed and Moved the intel_wopcm_partition into intel_guc (Sagar)
- Removed unnecessary function calls (Joonas)
- Init GuC WOPCM partition as soon as firmware fetching is completed
v3:
- Fixed indentation issues (Chris)
- Removed layering violation code (Chris/Michal)
- Created separat files for GuC wopcm code (Michal)
- Used inline function to avoid code duplication (Michal)
v4:
- Preset the GuC WOPCM top during early GuC init (Chris)
- Fail intel_uc_init_hw() as soon as GuC WOPCM partitioning failed
v5:
- Moved GuC DMA WOPCM register updating code into intel_wopcm.c
- Took care of the locking status before writing to GuC DMA
Write-Once registers. (Joonas)
v6:
- Made sure the GuC WOPCM size to be multiple of 4K (4K aligned)
v8:
- Updated comments and fixed naming issues (Sagar/Joonas)
- Updated commit message to include more description about the hardware
restriction on GuC WOPCM size (Sagar)
v9:
- Minor changes variable names and code comments (Sagar)
- Added detailed GuC WOPCM layout drawing (Sagar/Michal)
- Refined macro definitions to be reader friendly (Michal)
- Removed redundent check to valid flag (Michal)
- Unified first parameter for exported GuC WOPCM functions (Michal)
- Refined the name and parameter list of hardware restriction checking
functions (Michal)
v10:
- Used shorter function name for internal functions (Joonas)
- Moved init-ealry function into c file (Joonas)
- Consolidated and removed redundant size checks (Joonas/Michal)
- Removed unnecessary unlikely() from code which is only called once
during boot (Joonas)
- More fixes to kernel-doc format and content (Michal)
- Avoided the use of PAGE_MASK for 4K pages (Michal)
- Added error log messages to error paths (Michal)
v11:
- Replaced intel_guc_wopcm with more generic intel_wopcm and attached
intel_wopcm to drm_i915_private instead intel_guc (Michal)
- dynamic calculation of GuC non-wopcm memory start (a.k.a WOPCM Top
offset from GuC WOPCM base) (Michal)
- Moved WOPCM marco definitions into .c source file (Michal)
- Exported WOPCM layout diagram as kernel-doc (Michal)
v12:
- Updated naming, function kernel-doc to align with new changes (Michal)
v13:
- Updated the ordering of s-o-b/cc/r-b tags (Sagar)
- Corrected one tense error in comment (Sagar)
- Corrected typos and removed spurious comments (Joonas)
Bspec: 12690
Signed-off-by: Jackie Li <yaodong.li@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Cc: Sujaritha Sundaresan <sujaritha.sundaresan@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: John Spotswood <john.a.spotswood@intel.com>
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Sagar Arun Kamble <sagar.a.kamble@intel.com> (v8)
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> (v9)
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com> (v11)
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> (v12)
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1520987574-19351-2-git-send-email-yaodong.li@intel.com
Starting from Gen11 the context descriptor format has been updated in
the HW. The hw_id field has been considerably reduced in size and engine
class and instance fields have been added.
There is a slight name clashing issue because the field that we call
hw_id is actually called SW Context ID in the specs for Gen11+.
With the current size of the hw_id field we can have a maximum of 2k
contexts at any time, but we could use the sw_counter field (which is sw
defined) to increase that because the HW requirement is that
engine_id + sw id + sw_counter is a unique number.
GuC uses a similar method to support more contexts but does its tracking
at lrc level. To avoid doing an implementation that will need to be
reworked once GuC support lands, defer it for now and mark it as TODO.
v2: rebased, add documentation, fix GEN11_ENGINE_INSTANCE_SHIFT
v3: rebased, bring back lost code from i915_gem_context.c
v4: make TODO comment more generic
v5: be consistent with bit ordering, add extra checks (Chris)
Cc: Oscar Mateo <oscar.mateo@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Oscar Mateo <oscar.mateo@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180302161501.28594-3-mika.kuoppala@linux.intel.com
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
We want to de-emphasize the link between the request (dependency,
execution and fence tracking) from GEM and so rename the struct from
drm_i915_gem_request to i915_request. That is we may implement the GEM
user interface on top of requests, but they are an abstraction for
tracking execution rather than an implementation detail of GEM. (Since
they are not tied to HW, we keep the i915 prefix as opposed to intel.)
In short, the spatch:
@@
@@
- struct drm_i915_gem_request
+ struct i915_request
A corollary to contracting the type name, we also harmonise on using
'rq' shorthand for local variables where space if of the essence and
repetition makes 'request' unwieldy. For globals and struct members,
'request' is still much preferred for its clarity.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180221095636.6649-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The comment is very old and quite misleading now.
drivers/gpu/drm/i915/i915_gem_context.c:349: warning: No description found for parameter 'dev_priv'
drivers/gpu/drm/i915/i915_gem_context.c:349: warning: No description found for parameter 'file_priv'
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180208111559.32663-1-chris@chris-wilson.co.uk
Userspace provides a 64b value for the priority, we need to be careful
to preserve the full range before validation to prevent truncation (and
letting an illegal value pass).
Reported-by: Antonio Argenziano <antonio.argenziano@intel.com>
Fixes: ac14fbd460 ("drm/i915/scheduler: Support user-defined priorities")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Antonio Argenziano <antonio.argenziano@intel.com>
Cc: Michal Winiarski <michal.winiarski@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180208085151.11480-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
If we remove some hardcoded assumptions about the preempt context having
a fixed id, reserved from use by normal user contexts, we may only
allocate the i915_gem_context when required. Then the subsequent
decisions on using preemption reduce to having the preempt context
available.
v2: Include an assert that we don't allocate the preempt context twice.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Winiarski <michal.winiarski@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Acked-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180207210544.26351-3-chris@chris-wilson.co.uk
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Rather than having the high level ioctl interface guess the underlying
implementation details, having the implementation declare what
capabilities it exports. We define an intel_driver_caps, similar to the
intel_device_info, which instead of trying to describe the HW gives
details on what the driver itself supports. This is then populated by
the engine backend for the new scheduler capability field for use
elsewhere.
v2: Use caps.scheduler for validating CONTEXT_PARAM_SET_PRIORITY (Mika)
One less assumption of engine[RCS] \o/
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tomasz Lis <tomasz.lis@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tomasz Lis <tomasz.lis@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180207210544.26351-2-chris@chris-wilson.co.uk
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
If a fence allocation fails in a blocking context, we will sleep on the
fence as a last resort. We can therefore allow ourselves to fail and
sleep on the fence instead of triggering a system-wide oom. This allows
us to throttle malicious clients that are consuming lots of system
resources by capping the amount of memory used by fences.
Testcase: igt/gem_shrink/execbufX
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171212180652.22061-2-chris@chris-wilson.co.uk
In the upcoming patch we will change the way how to recognize
when GuC is in use. Using helper macros will minimize scope
of that changes. While here, update dev_info message.
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171206135316.32556-3-michal.wajdeczko@intel.com
Even though all rendering should have been flushed at the end of the
previous requests, add an extra flush after switching to the
kernel_context. As the switch to the kernel_context is used when idling
the gpu (e.g. suspend), having an extra layer of paranoia to ensure
everything is flushed to memory seems sensible.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171126214856.23702-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The legacy i915_switch_context() is only applicable to the legacy
ringbuffer submission method, so move it from the general
i915_gem_context.c to intel_ringbuffer.c (rename pending!).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171123152631.31385-2-chris@chris-wilson.co.uk
The legacy context switch for ringbuffer submission is multistaged,
where each of those stages may fail. However, we were updating global
state after some stages, and so we had to force the incomplete request
to be submitted because we could not unwind. Save the global state
before performing the switches, and so enable us to unwind back to the
previous global state should any phase fail. We then must cancel the
request instead of submitting it should the construction fail.
v2: s/saved_ctx/from_ctx/; s/ctx/to_ctx/ etc.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171123152631.31385-1-chris@chris-wilson.co.uk
Having disabled the broken semaphores on Sandybridge, there is no need
for a modparam any more, so remove it in favour of a simple
HAS_LEGACY_SEMAPHORES() guard.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-5-chris@chris-wilson.co.uk
Since removing the module parameter to force selection of ringbuffer
emission for gen8, the code is defunct. Remove it.
To put the difference into perspective, a couple of microbenchmarks
(bdw i7-5557u, 20170324):
ring execlists
exec continuous nops on all rings: 1.491us 2.223us
exec sequential nops on each ring: 12.508us 53.682us
single nop + sync: 9.272us 30.291us
vblank_mode=0 glxgears: ~11000fps ~9000fps
Since the earlier submission, gen8 ringbuffer submission has fallen
further and further behind in features. So while ringbuffer may hold the
throughput crown, in terms of interactive latency, execlists is much
better. Alas, we have no convenient metrics for such, other than
demonstrating things we can do with execlists but can not using
legacy ringbuffer submission.
We have made a few improvements to lowlevel execlists throughput,
and ringbuffer currently panics on boot! (bdw i7-5557u, 20171026):
ring execlists
exec continuous nops on all rings: n/a 1.921us
exec sequential nops on each ring: n/a 44.621us
single nop + sync: n/a 21.953us
vblank_mode=0 glxgears: n/a ~18500fps
References: https://bugs.freedesktop.org/show_bug.cgi?id=87725
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Once-upon-a-time-Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-2-chris@chris-wilson.co.uk
Execlists and legacy ringbuffer submission are no longer feature
comparable (execlists now offer greater functionality that should
overcome their performance hit) and obsoletes the unsafe module
parameter, i.e. comparing the two modes of execution is no longer
useful, so remove the debug tool.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com> #i915_perf.c
Link: https://patchwork.freedesktop.org/patch/msgid/20171120205504.21892-1-chris@chris-wilson.co.uk
During request construction, after pinning the context we know whether
or not we have to emit a context switch. So move this common operation
from every caller into i915_gem_request_alloc() itself.
v2: Always submit the request if we emitted some commands during request
construction, as typically it also involves changes in global state.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171120102002.22254-2-chris@chris-wilson.co.uk
Take a copy of the HW state after a reset upon module loading by
executing a context switch from a blank context to the kernel context,
thus saving the default hw state over the blank context image.
We can then use the default hw state to initialise any future context,
ensuring that each starts with the default view of hw state.
v2: Unmap our default state from the GTT after stealing it from the
context. This should stop us from accidentally overwriting it via the
GTT (and frees up some precious GTT space).
Testcase: igt/gem_ctx_isolation
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110142634.10551-7-chris@chris-wilson.co.uk
When we close the VMA, we unbind it from the ppgtt and tear down the
page directory pointing at it. That may trigger us to return WC pages
back to the system, requiring conversion back to WB which itself may
sleep. That makes i915_vma_close() unsuitable for use inside the RCU
read lock, which we need to hold to iterate the radixtree.
The fix is quite simple, we can close all the VMA as we close the ppgtt,
we only need to do that instead of closing them during destruction of
the LUT.
v2: Order between closing the LUT and the ppgtt is important; we use the
vma inside the LUT as a means of retrieving the object, and so we must
clear the LUT before freeing the VMA when closing the ppgtt.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103638
Fixes: 547da76b57 ("drm/i915: Hold rcu_read_lock when iterating over the radixtree (vma idr)")
Fixes: d1b48c1e71 ("drm/i915: Replace execbuf vma ht with an idr")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171109085540.32264-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Kasan spotted
[IGT] gem_tiled_pread_pwrite: exiting, ret=0
==================================================================
BUG: KASAN: use-after-free in __i915_gem_object_reset_page_iter+0x15c/0x170 [i915]
Read of size 8 at addr ffff8801359da310 by task kworker/3:2/182
CPU: 3 PID: 182 Comm: kworker/3:2 Tainted: G U 4.14.0-rc6-CI-Custom_3340+ #1
Hardware name: Intel Corp. Geminilake/GLK RVP1 DDR4 (05), BIOS GELKRVPA.X64.0062.B30.1708222146 08/22/2017
Workqueue: events __i915_gem_free_work [i915]
Call Trace:
dump_stack+0x68/0xa0
print_address_description+0x78/0x290
? __i915_gem_object_reset_page_iter+0x15c/0x170 [i915]
kasan_report+0x23d/0x350
__asan_report_load8_noabort+0x19/0x20
__i915_gem_object_reset_page_iter+0x15c/0x170 [i915]
? i915_gem_object_truncate+0x100/0x100 [i915]
? lock_acquire+0x380/0x380
__i915_gem_object_put_pages+0x30d/0x530 [i915]
__i915_gem_free_objects+0x551/0xbd0 [i915]
? lock_acquire+0x13e/0x380
__i915_gem_free_work+0x4e/0x70 [i915]
process_one_work+0x6f6/0x1590
? pwq_dec_nr_in_flight+0x2b0/0x2b0
worker_thread+0xe6/0xe90
? pci_mmcfg_check_reserved+0x110/0x110
kthread+0x309/0x410
? process_one_work+0x1590/0x1590
? kthread_create_on_node+0xb0/0xb0
ret_from_fork+0x27/0x40
Allocated by task 1801:
save_stack_trace+0x1b/0x20
kasan_kmalloc+0xee/0x190
kasan_slab_alloc+0x12/0x20
kmem_cache_alloc+0xdc/0x2e0
radix_tree_node_alloc.constprop.12+0x48/0x330
__radix_tree_create+0x274/0x480
__radix_tree_insert+0xa2/0x610
i915_gem_object_get_sg+0x224/0x670 [i915]
i915_gem_object_get_page+0xb5/0x1c0 [i915]
i915_gem_pread_ioctl+0x822/0xf60 [i915]
drm_ioctl_kernel+0x13f/0x1c0
drm_ioctl+0x6cf/0x980
do_vfs_ioctl+0x184/0xf30
SyS_ioctl+0x41/0x70
entry_SYSCALL_64_fastpath+0x1c/0xb1
Freed by task 37:
save_stack_trace+0x1b/0x20
kasan_slab_free+0xaf/0x190
kmem_cache_free+0xbf/0x340
radix_tree_node_rcu_free+0x79/0x90
rcu_process_callbacks+0x46d/0xf40
__do_softirq+0x21c/0x8d3
The buggy address belongs to the object at ffff8801359da0f0
which belongs to the cache radix_tree_node of size 576
The buggy address is located 544 bytes inside of
576-byte region [ffff8801359da0f0, ffff8801359da330)
The buggy address belongs to the page:
page:ffffea0004d67600 count:1 mapcount:0 mapping: (null) index:0x0 compound_mapcount: 0
flags: 0x8000000000008100(slab|head)
raw: 8000000000008100 0000000000000000 0000000000000000 0000000100110011
raw: ffffea0004b52920 ffffea0004b38020 ffff88015b416a80 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffff8801359da200: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff8801359da280: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>ffff8801359da300: fb fb fb fb fb fb fc fc fc fc fc fc fc fc fc fc
^
ffff8801359da380: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
ffff8801359da400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
==================================================================
Disabling lock debugging due to kernel taint
which looks like the slab containing the radixtree iter was freed as we
traversed the tree, taking the rcu read lock across the loop should
prevent that (deferring all the frees until the end).
Reported-by: Tomi Sarvela <tomi.p.sarvela@intel.com>
Fixes: d1b48c1e71 ("drm/i915: Replace execbuf vma ht with an idr")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171026130032.10677-2-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
During evict, we wish to idle the GPU if we see that the GGTT is full.
However, our test for idle in i915_gem_evict_something() and in
i915_gem_switch_to_kernel_context() do not match leading to
disappointment - we never believe that we are idle and keep trying to
flush the GGTT ad infinitum.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103438
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171024220855.30155-2-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Use a priority stored in the context as the initial value when
submitting a request. This allows us to change the default priority on a
per-context basis, allowing different contexts to be favoured with GPU
time at the expense of lower importance work. The user can adjust the
context's priority via I915_CONTEXT_PARAM_PRIORITY, with more positive
values being higher priority (they will be serviced earlier, after their
dependencies have been resolved). Any prerequisite work for an execbuf
will have its priority raised to match the new request as required.
Normal users can specify any value in the range of -1023 to 0 [default],
i.e. they can reduce the priority of their workloads (and temporarily
boost it back to normal if so desired).
Privileged users can specify any value in the range of -1023 to 1023,
[default is 0], i.e. they can raise their priority above all overs and
so potentially starve the system.
Note that the existing schedulers are not fair, nor load balancing, the
execution is strictly by priority on a first-come, first-served basis,
and the driver may choose to boost some requests above the range
available to users.
This priority was originally based around nice(2), but evolved to allow
clients to adjust their priority within a small range, and allow for a
privileged high priority range.
For example, this can be used to implement EGL_IMG_context_priority
https://www.khronos.org/registry/egl/extensions/IMG/EGL_IMG_context_priority.txt
EGL_CONTEXT_PRIORITY_LEVEL_IMG determines the priority level of
the context to be created. This attribute is a hint, as an
implementation may not support multiple contexts at some
priority levels and system policy may limit access to high
priority contexts to appropriate system privilege level. The
default value for EGL_CONTEXT_PRIORITY_LEVEL_IMG is
EGL_CONTEXT_PRIORITY_MEDIUM_IMG."
so we can map
PRIORITY_HIGH -> 1023 [privileged, will failback to 0]
PRIORITY_MED -> 0 [default]
PRIORITY_LOW -> -1023
They also map onto the priorities used by VkQueue (and a VkQueue is
essentially a timeline, our i915_gem_context under full-ppgtt).
v2: s/CAP_SYS_ADMIN/CAP_SYS_NICE/
v3: Report min/max user priorities as defines in the uapi, and rebase
internal priorities on the exposed values.
Testcase: igt/gem_exec_schedule
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171003203453.15692-9-chris@chris-wilson.co.uk
Add another perma-pinned context for using for preemption at any time.
We cannot just reuse the existing kernel context, as first and foremost
we need to ensure that we can preempt the kernel context itself, so
require a distinct context id. Similar to the kernel context, we may
want to interrupt execution and switch to the preempt context at any
time, and so it needs to be permanently pinned and available.
To compensate for yet another permanent allocation, we shrink the
existing context and the new context by reducing their ringbuffer to the
minimum.
v2: Assert that we never allocate a request from the preemption context.
v3: Limit perma-pin to engines that may preempt.
v4: Onion cleanup for early driver death
v5: Onion ordering in main driver cleanup as well.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171003203453.15692-4-chris@chris-wilson.co.uk
Our global struct with params is named exactly the same way
as new preferred name for the drm_i915_private function parameter.
To avoid such name reuse lets use different name for the global.
v5: pure rename
v6: fix
Credits-to: Coccinelle
@@
identifier n;
@@
(
- i915.n
+ i915_modparams.n
)
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Ville Syrjala <ville.syrjala@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170919193846.38060-1-michal.wajdeczko@intel.com
This was the competing idea long ago, but it was only with the rewrite
of the idr as an radixtree and using the radixtree directly ourselves,
along with the realisation that we can store the vma directly in the
radixtree and only need a list for the reverse mapping, that made the
patch performant enough to displace using a hashtable. Though the vma ht
is fast and doesn't require any extra allocation (as we can embed the node
inside the vma), it does require a thread for resizing and serialization
and will have the occasional slow lookup. That is hairy enough to
investigate alternatives and favour them if equivalent in peak performance.
One advantage of allocating an indirection entry is that we can support a
single shared bo between many clients, something that was done on a
first-come first-serve basis for shared GGTT vma previously. To offset
the extra allocations, we create yet another kmem_cache for them.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170816085210.4199-5-chris@chris-wilson.co.uk
When switching between contexts using the aliasing_ppgtt, the VM is
shared. We don't need to reload the PD registers unless they are dirty.
Martin Peres reported an issue that looks like corruption between
Haswell context switches, bisecting to commit f9326be5f1 ("drm/i915:
Rearrange switch_context to load the aliasing ppgtt on first use").
Switching between the same mm (the aliasing_ppgtt is used for all
contexts in this case) should be a nop, but appears to trigger some
side-effects in the context switch. However, as we know the switch
is redundant in this case, we can skip it and continue to ignore the
issue until somebody feels strong enough to investigate full-ppgtt on
gen7 again!
Except.. Martin was using full-ppgtt which is not supported as it
doesn't work correctly yet. So whilst the bisect did yield valuable
information about the failures, the fix should not have any user impact
under default settings, with the exception of a slightly lower
throughput on xcs as the VM would always be reloaded.
v2: Also remember to set the legacy_active_context following the switch
on xcs (commit e8a9c58fcd ("drm/i915: Unify active context tracking
between legacy/execlists/guc"))
Fixes: f9326be5f1 ("drm/i915: Rearrange switch_context to load the aliasing ppgtt on first use")
Fixes: e8a9c58fcd ("drm/i915: Unify active context tracking between legacy/execlists/guc")
Reported-by: Martin Peres <martin.peres@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Martin Peres <martin.peres@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170812152724.6883-1-chris@chris-wilson.co.uk
Since we make call i915_gem_context_mark_guilty() concurrently when
resetting different engines in parallel, we need to make sure that our
updates are safe for the unlocked access.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170721123238.16428-12-chris@chris-wilson.co.uk
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Currently, we move all unreferenced contexts to an RCU free list and
then onto a worker for eventual reaping. To compensate against this
growing into a long list with frequent allocations starving the system
of available memory, before we allocate a new context we reap all the
stale contexts. This puts all the cost of destroying the context into
the next allocator, which is presumably more sensitive to syscall
latency and unfair. We can limit the number of contexts being freed by
the new allocator to both keep the list trimmed and to allow the
allocator to be reasonably fast.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170705142634.18554-4-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Before we create a new context, we try and reap all the stale contexts
(i.e. those that are freed but waiting for a worker to come and return
their allocations to the system). Before we do this, we retire all
requests so that we clear any inflight no longer used contexts (who are
only being kept alived by those inflght requests). However, any context
that is finally unreferenced by this retirement is put onto an RCU list
and not available for immediately reaping, we stall for no immediate
benefit.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170705142634.18554-3-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
We need to reap the stale contexts for all new contexts, be they created
by user in i915_gem_context_ioctl or from opening a new file in
i915_gem_context_open. Both paths may be called very frequently
accumulating many stale contexts before any worker has a chance to run
and free their memory.
Fixes: 1acfc104cd ("drm/i915: Enable rcu-only context lookups")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170705142634.18554-2-chris@chris-wilson.co.uk
When iterating the list of contexts to free, we need to use a safe
iterator as we are freeing the link as we go. Pass an extra thick brown
paper bag.
Fixes: 5f09a9c8ab ("drm/i915: Allow contexts to be unreferenced locklessly")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170630230517.1938-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Whilst the contents of the context is still protected by the big
struct_mutex, this is not much of an improvement. It is just one tiny
step towards reducing our BKL.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170620110547.15947-3-chris@chris-wilson.co.uk
If we move the actual cleanup of the context to a worker, we can allow
the final free to be called from any context and avoid undue latency in
the caller.
v2: Negotiate handling the delayed contexts free by flushing the
workqueue before calling i915_gem_context_fini() and performing the final
free of the kernel context directly
v3: Flush deferred frees before new context allocations
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170620110547.15947-2-chris@chris-wilson.co.uk
If we take a reference to the object/vma when it is first used in an
execbuf, we can keep that reference until the object's file-local handle
is closed. Thereby saving a frequent ref/unref pair.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
The advent of full-ppgtt lead to an extra indirection between the object
and its binding. That extra indirection has a noticeable impact on how
fast we can convert from the user handles to our internal vma for
execbuffer. In order to bypass the extra indirection, we use a
resizable hashtable to jump from the object to the per-ctx vma.
rhashtable was considered but we don't need the online resizing feature
and the extra complexity proved to undermine its usefulness. Instead, we
simply reallocate the hastable on demand in a background task and
serialize it before iterating.
In non-full-ppgtt modes, multiple files and multiple contexts can share
the same vma. This leads to having multiple possible handle->vma links,
so we only use the first to establish the fast path. The majority of
buffers are not shared and so we should still be able to realise
speedups with multiple clients.
v2: Prettier names, more magic.
v3: Many style tweaks, most notably hiding the misuse of execobj[].rsvd2
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Lionel Landwerlin