One more complication of preempt-to-busy with respect to the virtual
engine is that we may have retired the last request along the virtual
engine at the same time as preparing to submit the completed request to
a new engine. That submit will be shortcircuited, but not before we have
updated the context with the new register offsets and marked the virtual
engine as bound to the new engine (by calling swap on ve->siblings[]).
As we may have just retired the completed request, we may also be in the
middle of calling virtual_context_exit() to turn off the power management
associated with the virtual engine, and that in turn walks the
ve->siblings[]. If we happen to call swap() on the array as we walk, we
will call intel_engine_pm_put() twice on the same engine.
In this patch, we prevent this by only updating the bound engine after a
successful submission which weeds out the already completed requests.
Alternatively, we could walk a non-volatile array for the pm, such as
using the engine->mask. The small advantage to performing the update
after the submit is that we then only have to do a swap for active
requests.
Fixes: 22b7a426bb ("drm/i915/execlists: Preempt-to-busy")
References: 6d06779e86 ("drm/i915: Load balancing across a virtual engine"
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: "Nayana, Venkata Ramana" <venkata.ramana.nayana@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200731154834.8378-3-chris@chris-wilson.co.uk
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Since the breadcrumb enabling/cancelling itself is serialised by the
breadcrumbs.irq_lock, with a bit of care we can remove the outer
serialisation with i915_request.lock for concurrent
dma_fence_enable_signaling(). This has the important side-effect of
eliminating the nested i915_request.lock within request submission.
The challenge in serialisation is around the unsubmission where we take
an active request that wants a breadcrumb on the signaling engine and
put it to sleep. We do not want a concurrent
dma_fence_enable_signaling() to attach a breadcrumb as we unsubmit, so
we must mark the request as no longer active before serialising with the
concurrent enable-signaling.
On retire, we serialise with the concurrent enable-signaling, but
instead of clearing ACTIVE, we mark it as SIGNALED.
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/20200731154834.8378-1-chris@chris-wilson.co.uk
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
[Joonas: Rebased and reordered into drm-intel-gt-next branch]
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Sometimes we have to be very careful not to allocate underneath a mutex
(or spinlock) and yet still want to track activity. Enter
i915_active_acquire_for_context(). This raises the activity counter on
i915_active prior to use and ensures that the fence-tree contains a slot
for the context.
v2: Refactor active_lookup() so it can be called again before/after
locking to resolve contention. Since we protect the rbtree until we
idle, we can do a lockfree lookup, with the caveat that if another
thread performs a concurrent insertion, the rotations from the insert
may cause us to not find our target. A second pass holding the treelock
will find the target if it exists, or the place to perform our
insertion.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Thomas Hellström <thomas.hellstrom@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200731085015.32368-3-chris@chris-wilson.co.uk
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
I915_GEM_THROTTLE dates back to the time before contexts where there was
just a single engine, and therefore a single timeline and request list
globally. That request list was in execution/retirement order, and so
walking it to find a particular aged request made sense and could be
split per file.
That is no more. We now have many timelines with a file, as many as the
user wants to construct (essentially per-engine, per-context). Each of
those run independently and so make the single list futile. Remove the
disordered list, and iterate over all the timelines to find a request to
wait on in each to satisfy the criteria that the CPU is no more than 20ms
ahead of its oldest request.
It should go without saying that the I915_GEM_THROTTLE ioctl is no
longer used as the primary means of throttling, so it makes sense to push
the complication into the ioctl where it only impacts upon its few
irregular users, rather than the execbuf/retire where everybody has to
pay the cost. Fortunately, the few users do not create vast amount of
contexts, so the loops over contexts/engines should be concise.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200728152010.30701-1-chris@chris-wilson.co.uk
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
TGL made stepping a litte mess, workarounds refer to the stepping of
the IP(GT or Display) not of the GPU stepping so it would already
require the same solution as used in commit 96c5a15f9f
("drm/i915/kbl: Fix revision ID checks").
But to make things even more messy it have a different IP stepping
mapping between SKUs and the same stepping revision of GT do not match
the same HW between TGL U/Y and regular TGL.
So it was required to have 2 different macros to check GT WAs while
for Display we are able to use just one macro that uses the right
revids table.
All TGL workarounds checked and updated accordingly.
v2:
- removed TODO to check if WA 14010919138 applies to regular TGL.
- fixed display stepping in regular TGL (Anusha)
BSpec: 52890
BSpec: 55378
BSpec: 44455
Reviewed-by: Anusha Srivatsa <anusha.srivtsa@intel.com>
Cc: Anusha Srivatsa <anusha.srivatsa@intel.com>
Cc: Penne Lee <penne.y.lee@intel.com>
Cc: Guangyao Bai <guangyao.bai@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200827233943.400946-1-jose.souza@intel.com
UAPI Changes:
- Introduce a mechanism to extend execbuf2 (Lionel)
- Add syncobj timeline support (Lionel)
Driver Changes:
- Limit stolen mem usage on the compressed frame buffer (Ville)
- Some clean-up around display's cdclk (Ville)
- Some DDI changes for better DP link training according
to spec (Imre)
- Provide the perf pmu.module (Chris)
- Remove dobious Valleyview PCI IDs (Alexei)
- Add new display power saving feature for gen12+ called
HOBL (Jose)
- Move SKL's clock gating w/a to skl_init_clock_gating() (Ville)
- Rocket Lake display additions (Matt)
- Selftest: temporarily downgrade on severity of frequency
scaling tests (Chris)
- Introduce a new display workaround for fixing FLR related
issues on new PCH. (Jose)
- Temporarily disable FBC on TGL. It was the culprit of random
underruns. (Uma).
- Copy default modparams to mock i915_device (Chris)
- Add compiler paranoia for checking HWSP values (Chris)
- Remove useless gen check before calling intel_rps_boost (Chris)
- Fix a null pointer dereference (Chris)
- Add a couple of missing i915_active_fini() (Chris)
- Update TGL display power's bw_buddy table according to
update spec (Matt)
- Fix couple wrong return values (Tianjia)
- Selftest: Avoid passing random 0 into ilog2 (George)
- Many Tiger Lake display fixes and improvements for Type-C and
DP compliance (Imre, Jose)
- Start the addition of PSR2 selective fetch (Jose)
- Update a few DMC and HuC firmware versions (Jose)
- Add gen11+ w/a to fix underuns (Matt)
- Fix cmd parser desc matching with mask (Mika)
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200826232733.GA129053@intel.com
We usually assume that increasing PCI device revision ID's translates to
newer steppings; macros like IS_KBL_REVID() that we use rely on this
behavior. Unfortunately this turns out to not be true on KBL; the
newer device 2 revision ID's sometimes go backward to older steppings.
The situation is further complicated by different GT and display
steppings associated with each revision ID.
Let's work around this by providing a table to map the revision ID to
specific GT and display steppings, and then perform our comparisons on
the mapped values.
v2:
- Move the kbl_revids[] array to intel_workarounds.c to avoid compiler
warnings about an unused variable in files that don't call the
macros (kernel test robot).
Bspec: 18329
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200811032105.2819370-1-matthew.d.roper@intel.com
Reviewed-by: Swathi Dhanavanthri <swathi.dhanavanthri@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Pull drm updates from Dave Airlie:
"New xilinx displayport driver, AMD support for two new GPUs (more
header files), i915 initial support for RocketLake and some work on
their DG1 (discrete chip).
The core also grew some lockdep annotations to try and constrain what
drivers do with dma-fences, and added some documentation on why the
idea of indefinite fences doesn't work.
The long list is below.
I do have some fixes trees outstanding, but I'll follow up with those
later.
core:
- add user def flag to cmd line modes
- dma_fence_wait added might_sleep
- dma-fence lockdep annotations
- indefinite fences are bad documentation
- gem CMA functions used in more drivers
- struct mutex removal
- more drm_ debug macro usage
- set/drop master api fixes
- fix for drm/mm hole size comparison
- drm/mm remove invalid entry optimization
- optimise drm/mm hole handling
- VRR debugfs added
- uncompressed AFBC modifier support
- multiple display id blocks in EDID
- multiple driver sg handling fixes
- __drm_atomic_helper_crtc_reset in all drivers
- managed vram helpers
ttm:
- ttm_mem_reg handling cleanup
- remove bo offset field
- drop CMA memtype flag
- drop mappable flag
xilinx:
- New Xilinx ZynqMP DisplayPort Subsystem driver
nouveau:
- add CRC support
- start using NVIDIA published class header files
- convert all push buffer emission to new macros
- Proper push buffer space management for EVO/NVD channels.
- firmware loading fixes
- 2MiB system memory pages support on Pascal and newer
vkms:
- larger cursor support
i915:
- Rocketlake platform enablement
- Early DG1 enablement
- Numerous GEM refactorings
- DP MST fixes
- FBC, PSR, Cursor, Color, Gamma fixes
- TGL, RKL, EHL workaround updates
- TGL 8K display support fixes
- SDVO/HDMI/DVI fixes
amdgpu:
- Initial support for Sienna Cichlid GPU
- Initial support for Navy Flounder GPU
- SI UVD/VCE support
- expose rotation property
- Add support for unique id on Arcturus
- Enable runtime PM on vega10 boards that support BACO
- Skip BAR resizing if the bios already did id
- Major swSMU code cleanup
- Fixes for DCN bandwidth calculations
amdkfd:
- Track SDMA usage per process
- SMI events interface
radeon:
- Default to on chip GART for AGP boards on all arches
- Runtime PM reference count fixes
msm:
- headers regenerated causing churn
- a650/a640 display and GPU enablement
- dpu dither support for 6bpc panels
- dpu cursor fix
- dsi/mdp5 enablement for sdm630/sdm636/sdm66
tegra:
- video capture prep support
- reflection support
mediatek:
- convert mtk_dsi to bridge API
meson:
- FBC support
sun4i:
- iommu support
rockchip:
- register locking fix
- per-pixel alpha support PX30 VOP
mgag200:
- ported to simple and shmem helpers
- device init cleanups
- use managed pci functions
- dropped hw cursor support
ast:
- use managed pci functions
- use managed VRAM helpers
- rework cursor support
malidp:
- dev_groups support
hibmc:
- refactor hibmc_drv_vdac:
vc4:
- create TXP CRTC
imx:
- error path fixes and cleanups
etnaviv:
- clock handling and error handling cleanups
- use pin_user_pages"
* tag 'drm-next-2020-08-06' of git://anongit.freedesktop.org/drm/drm: (1747 commits)
drm/msm: use kthread_create_worker instead of kthread_run
drm/msm/mdp5: Add MDP5 configuration for SDM636/660
drm/msm/dsi: Add DSI configuration for SDM660
drm/msm/mdp5: Add MDP5 configuration for SDM630
drm/msm/dsi: Add phy configuration for SDM630/636/660
drm/msm/a6xx: add A640/A650 hwcg
drm/msm/a6xx: hwcg tables in gpulist
drm/msm/dpu: add SM8250 to hw catalog
drm/msm/dpu: add SM8150 to hw catalog
drm/msm/dpu: intf timing path for displayport
drm/msm/dpu: set missing flush bits for INTF_2 and INTF_3
drm/msm/dpu: don't use INTF_INPUT_CTRL feature on sdm845
drm/msm/dpu: move some sspp caps to dpu_caps
drm/msm/dpu: update UBWC config for sm8150 and sm8250
drm/msm/dpu: use right setup_blend_config for sm8150 and sm8250
drm/msm/a6xx: set ubwc config for A640 and A650
drm/msm/adreno: un-open-code some packets
drm/msm: sync generated headers
drm/msm/a6xx: add build_bw_table for A640/A650
drm/msm/a6xx: fix crashstate capture for A650
...
Preempt-to-busy introduces various fascinating complications in that the
requests may complete as we are unsubmitting them from HW. As they may
then signal after unsubmission, we may find ourselves having to cleanup
the signaling request from within the signaling callback. This causes us
to recurse onto the same i915_request.lock.
However, if the request is already signaled (as it will be before we
enter the signal callbacks), we know we can skip the signaling of that
request during submission, neatly evading the spinlock recursion.
unsubmit(ve.rq0) # timeslice expiration or other preemption
-> virtual_submit_request(ve.rq0)
dma_fence_signal(ve.rq0) # request completed before preemption ack
-> submit_notify(ve.rq1)
-> virtual_submit_request(ve.rq1) # sees that we have completed ve.rq0
-> __i915_request_submit(ve.rq0)
[ 264.210142] BUG: spinlock recursion on CPU#2, sample_multi_tr/2093
[ 264.210150] lock: 0xffff9efd6ac55080, .magic: dead4ead, .owner: sample_multi_tr/2093, .owner_cpu: 2
[ 264.210155] CPU: 2 PID: 2093 Comm: sample_multi_tr Tainted: G U
[ 264.210158] Hardware name: Intel Corporation CoffeeLake Client Platform/CoffeeLake S UDIMM RVP, BIOS CNLSFWR1.R00.X212.B01.1909060036 09/06/2019
[ 264.210160] Call Trace:
[ 264.210167] dump_stack+0x98/0xda
[ 264.210174] spin_dump.cold+0x24/0x3c
[ 264.210178] do_raw_spin_lock+0x9a/0xd0
[ 264.210184] _raw_spin_lock_nested+0x6a/0x70
[ 264.210314] __i915_request_submit+0x10a/0x3c0 [i915]
[ 264.210415] virtual_submit_request+0x9b/0x380 [i915]
[ 264.210516] submit_notify+0xaf/0x14c [i915]
[ 264.210602] __i915_sw_fence_complete+0x8a/0x230 [i915]
[ 264.210692] i915_sw_fence_complete+0x2d/0x40 [i915]
[ 264.210762] __dma_i915_sw_fence_wake+0x19/0x30 [i915]
[ 264.210767] dma_fence_signal_locked+0xb1/0x1c0
[ 264.210772] dma_fence_signal+0x29/0x50
[ 264.210871] i915_request_wait+0x5cb/0x830 [i915]
[ 264.210876] ? dma_resv_get_fences_rcu+0x294/0x5d0
[ 264.210974] i915_gem_object_wait_fence+0x2f/0x40 [i915]
[ 264.211084] i915_gem_object_wait+0xce/0x400 [i915]
[ 264.211178] i915_gem_wait_ioctl+0xff/0x290 [i915]
Fixes: 22b7a426bb ("drm/i915/execlists: Preempt-to-busy")
References: 6d06779e86 ("drm/i915: Load balancing across a virtual engine")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: "Nayana, Venkata Ramana" <venkata.ramana.nayana@intel.com>
Cc: <stable@vger.kernel.org> # v5.4+
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200713141636.29326-1-chris@chris-wilson.co.uk
If something has gone awry with the CSB processing, we need to pause,
unwind and restart the request submission and event processing. However,
currently we skip the engine reset if we raise an error but discover no
active context, in the mistaken belief that it was merely a glitch in
the matrix. The glitches are real enough, and we do need to unwind even
if the engine appears idle (as it has gone permanently idle!) The
simplest way to unwind and recover is simply do the engine reset, which
should be very fast and _safe_ as nothing is active.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200711091349.28865-1-chris@chris-wilson.co.uk
If the HW throws a curve ball and reports either en event before it is
possible, or just a completely impossible event, we have to grin and
bear it. The first few events, we will likely not notice as we would be
expecting some event, but as soon as we stop expecting an event and yet
they still keep coming, then we enter into undefined state territory.
In which case, bail out, stop processing the events, and reset the
engine and our set of queued requests to recover.
The sporadic hangs and warnings will continue to plague CI, but at least
system stability should not be compromised.
v2: Commentary and force the reset-on-error.
v3: Customised user facing message for forced resets from internal errors.
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2045
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/20200710133125.30194-1-chris@chris-wilson.co.uk
We removed retiring requests from the shrinker in order to decouple the
mutexes from reclaim in preparation for unravelling the struct_mutex.
The impact of not retiring is that we are much less agressive in making
global objects available for shrinking, as such objects remain pinned
until they are flushed by a heartbeat pulse following the last retired
request along their timeline. In order to ensure that pulse occurs in
time for memory reclamation, we should kick it from kswapd.
The catch is that we have added some flush_work() into the retirement
phase (to ensure that we reach a global idle in a timely manner), but
these flush_work() are not eligible (i.e do not belong to WQ_MEM_RELCAIM)
for use from inside kswapd. To avoid flushing those workqueues, we teach
the retirer not to do so unless we are actually waiting, and only do the
plain retire from inside the shrinker.
Note that for execlists, we already retire completed contexts as they
are scheduled out, so it should not be keeping global state
unnecessarily pinned. The legacy ringbuffer however...
References: 9e9539800d ("drm/i915: Remove waiting & retiring from shrinker paths")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200708173748.32734-1-chris@chris-wilson.co.uk
