As preempt-to-busy leaves the request on the HW as the resubmission is
processed, that request may complete in the background and even cause a
second virtual request to enter queue. This second virtual request
breaks our "single request in the virtual pipeline" assumptions.
Furthermore, as the virtual request may be completed and retired, we
lose the reference the virtual engine assumes is held. Normally, just
removing the request from the scheduler queue removes it from the
engine, but the virtual engine keeps track of its singleton request via
its ve->request. This pointer needs protecting with a reference.
v2: Drop unnecessary motion of rq->engine = owner
Fixes: 22b7a426bb ("drm/i915/execlists: Preempt-to-busy")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190923152844.8914-1-chris@chris-wilson.co.uk
(cherry picked from commit b647c7df01)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Panfrost uses multiple schedulers (one for each slot, so 2 in reality),
and on a timeout has to stop all the schedulers to safely perform a
reset. However more than one scheduler can trigger a timeout at the same
time. This race condition results in jobs being freed while they are
still in use.
When stopping other slots use cancel_delayed_work_sync() to ensure that
any timeout started for that slot has completed. Also use
mutex_trylock() to obtain reset_lock. This means that only one thread
attempts the reset, the other threads will simply complete without doing
anything (the first thread will wait for this in the call to
cancel_delayed_work_sync()).
While we're here and since the function is already dependent on
sched_job not being NULL, let's remove the unnecessary checks.
Fixes: aa20236784 ("drm/panfrost: Prevent concurrent resets")
Tested-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191009094456.9704-1-steven.price@arm.com
Commit 4daa4fba3a ("gpu: drm: ttm: Adding new return type vm_fault_t")
broke TTM prefaulting. Since vmf_insert_mixed() typically always returns
VM_FAULT_NOPAGE, prefaulting stops after the second PTE.
Restore (almost) the original behaviour. Unfortunately we can no longer
with the new vm_fault_t return type determine whether a prefaulting
PTE insertion hit an already populated PTE, and terminate the insertion
loop. Instead we continue with the pre-determined number of prefaults.
Fixes: 4daa4fba3a ("gpu: drm: ttm: Adding new return type vm_fault_t")
Cc: Souptick Joarder <jrdr.linux@gmail.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: stable@vger.kernel.org # v4.19+
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/330387/
Pull xen fixes from Juergen Gross:
- correct panic handling when running as a Xen guest
- cleanup the Xen grant driver to remove printing a pointer being
always NULL
- remove a soon to be wrong call of of_dma_configure()
* tag 'for-linus-5.4-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
xen: Stop abusing DT of_dma_configure API
xen/grant-table: remove unnecessary printing
x86/xen: Return from panic notifier
On msm8998, vblank timeouts are observed because the DSI controller is not
reset properly, which ends up stalling the MDP. This is because the reset
logic is not correct per the hardware documentation.
The documentation states that after asserting reset, software should wait
some time (no indication of how long), or poll the status register until it
returns 0 before deasserting reset.
wmb() is insufficient for this purpose since it just ensures ordering, not
timing between writes. Since asserting and deasserting reset occurs on the
same register, ordering is already guaranteed by the architecture, making
the wmb extraneous.
Since we would define a timeout for polling the status register to avoid a
possible infinite loop, lets just use a static delay of 20 ms, since 16.666
ms is the time available to process one frame at 60 fps.
Fixes: a689554ba6 ("drm/msm: Initial add DSI connector support")
Cc: Hai Li <hali@codeaurora.org>
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Jeffrey Hugo <jeffrey.l.hugo@gmail.com>
Reviewed-by: Sean Paul <sean@poorly.run>
[seanpaul renamed RESET_DELAY to DSI_RESET_TOGGLE_DELAY_MS]
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191011133939.16551-1-jeffrey.l.hugo@gmail.com
- Fix CML display by adding a missing ID.
- Drop redundant list_del_init
- Only enqueue already completed requests to avoid races
- Fixup preempt-to-busy vs reset of a virtual request
- Protect peeking at execlists->active
- execlists->active is serialised by the tasklet
drm-intel-next-fixes-2019-09-19:
- Extend old HSW workaround to fix some GPU hangs on Haswell GT2
- Fix return error code on GEM mmap.
- White list a chicken bit register for push constants legacy mode on Mesa
- Fix resume issue related to GGTT restore
- Remove incorrect BUG_ON on execlist's schedule-out
- Fix unrecoverable GPU hangs with Vulkan compute workloads on SKL
drm-intel-next-fixes-2019-09-26:
- Fix concurrence on cases where requests where getting retired at same time as resubmitted to HW
- Fix gen9 display resolutions by setting the right max plane width
- Fix GPU hang on preemption
- Mark contents as dirty on a write fault. This was breaking cursor sprite with dumb buffers.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191010143039.GA15313@intel.com
[cherry-picked to drm-misc-fixes: drm-misc-next commit dfef959803]
Commit 554b3529fe ("thermal/drivers/core: Remove the module Kconfig's
option") changed the type of THERMAL from tristate to bool, so
THERMAL || !THERMAL is now always y. Remove the redundant dependency.
Discovered through Kconfiglib detecting a dependency loop. The C tools
simplify the expression to y before running dependency loop detection,
and so don't see it. Changing the type of THERMAL back to tristate makes
the C tools detect the same loop.
Not sure if running dep. loop detection after simplification can be
called a bug. Fixing this nit unbreaks Kconfiglib on the kernel at
least.
Signed-off-by: Ulf Magnusson <ulfalizer@gmail.com>
Signed-off-by: Noralf Trønnes <noralf@tronnes.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20190927174218.GA32085@huvuddator
max_tu_symbol was programmed to TU_SIZE_RECOMMENDED - 1, which is not
what the spec says. The spec says:
roundup ((input active video bandwidth in bytes/output active video
bandwidth in bytes) * tu_size)
It is not quite clear what the above means, but calculating
max_tu_symbol = (input Bps / output Bps) * tu_size seems to work and
fixes the issues seen.
This fixes artifacts in some videomodes (e.g. 1024x768@60 on 2-lanes &
1.62Gbps was pretty bad for me).
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Tested-by: Jyri Sarha <jsarha@ti.com>
Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190924131702.9988-1-tomi.valkeinen@ti.com
Due to the nature of preempt-to-busy the execlists active tracking and
the schedule queue may become temporarily desync'ed (between resubmission
to HW and its ack from HW). This means that we may have unwound a
request and passed it back to the virtual engine, but it is still
inflight on the HW and may even result in a GPU hang. If we detect that
GPU hang and try to reset, the hanging request->engine will no longer
match the current engine, which means that the request is not on the
execlists active list and we should not try to find an older incomplete
request. Given that we have deduced this must be a request on a virtual
engine, it is the single active request in the context and so must be
guilty (as the context is still inflight, it is prevented from being
executed on another engine as we process the reset).
Fixes: 22b7a426bb ("drm/i915/execlists: Preempt-to-busy")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190923152844.8914-2-chris@chris-wilson.co.uk
(cherry picked from commit cb2377a919)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
If we are asked to submit a completed request, just move it onto the
active-list without modifying it's payload. If we try to emit the
modified payload of a completed request, we risk racing with the
ring->head update during retirement which may advance the head past our
breadcrumb and so we generate a warning for the emission being behind
the RING_HEAD.
v2: Commentary for the sneaky, shared responsibility between functions.
v3: Spelling mistakes and bonus assertion
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/20190923110056.15176-3-chris@chris-wilson.co.uk
(cherry picked from commit c0bb487dc1)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Polaris and vegam use count for the value rather than
level. This looks like a copy paste typo from when
the code was adapted from previous asics.
I'm not sure that the SMU actually uses this value, so
I don't know that it actually is a bug per se.
Bug: https://bugs.freedesktop.org/show_bug.cgi?id=108609
Reported-by: Robert Strube <rstrube@gmail.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
cleanup error handling code and make sure temporary info array
with the handles are freed by amdgpu_bo_list_put() on
idr_replace()'s failure.
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Force bonded requests to run on distinct engines so that they cannot be
shuffled onto the same engine where timeslicing will reverse the order.
A bonded request will often wait on a semaphore signaled by its master,
creating an implicit dependency -- if we ignore that implicit dependency
and allow the bonded request to run on the same engine and before its
master, we will cause a GPU hang. [Whether it will hang the GPU is
debatable, we should keep on timeslicing and each timeslice should be
"accidentally" counted as forward progress, in which case it should run
but at one-half to one-third speed.]
We can prevent this inversion by restricting which engines we allow
ourselves to jump to upon preemption, i.e. baking in the arrangement
established at first execution. (We should also consider capturing the
implicit dependency using i915_sched_add_dependency(), but first we need
to think about the constraints that requires on the execution/retirement
ordering.)
Fixes: 8ee36e048c ("drm/i915/execlists: Minimalistic timeslicing")
References: ee1136908e ("drm/i915/execlists: Virtual engine bonding")
Testcase: igt/gem_exec_balancer/bonded-slice
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190923152844.8914-3-chris@chris-wilson.co.uk
(cherry picked from commit e2144503bf)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
The officially validated plane width limit is 4k on skl+, however
we already had people using 5k displays before we started to enforce
the limit. Also it seems Windows allows 5k resolutions as well
(though not sure if they do it with one plane or two).
According to hw folks 5k should work with the possible
exception of the following features:
- Ytile (already limited to 4k)
- FP16 (already limited to 4k)
- render compression (already limited to 4k)
- KVMR sprite and cursor (don't care)
- horizontal panning (need to verify this)
- pipe and plane scaling (need to verify this)
So apart from last two items on that list we are already
fine. We should really verify what happens with those last
two items but I don't have a 5k display on hand atm so it'll
have to wait.
In the meantime let's just bump the limit back up to 5k since
several users have already been using it without apparent issues.
At least we'll be no worse off than we were prior to lowering
the limits.
Cc: stable@vger.kernel.org
Cc: Sean Paul <sean@poorly.run>
Cc: José Roberto de Souza <jose.souza@intel.com>
Tested-by: Leho Kraav <leho@kraav.com>
Fixes: 372b9ffb57 ("drm/i915: Fix skl+ max plane width")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111501
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190905135044.2001-1-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Sean Paul <sean@poorly.run>
(cherry picked from commit bed34ef544)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
As we may unwind incomplete requests (for preemption) prior to
processing the CSB and the schedule-out events, we may update rq->engine
(resetting it to point back to the parent virtual engine) prior to
calling execlists_schedule_out(), invalidating the assertion that the
request still points to the inflight engine. (The likelihood of this is
increased if the CSB interrupt processing is pushed to the ksoftirqd for
being too slow and direct submission overtakes it.)
Tvrtko summarised it as:
"So unwind from direct submission resets rq->engine and races with
process_csb from the tasklet which notices request has actually
completed."
Reported-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Fixes: df40306902 ("drm/i915/execlists: Lift process_csb() out of the irq-off spinlock")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190907105046.19934-1-chris@chris-wilson.co.uk
(cherry picked from commit d810583fc2)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
One tilcdc fix was left out in drm-misc-next-fixes and didn't make it
during the merge window. Let's bring it into drm-misc-fixes.
Signed-off-by: Maxime Ripard <mripard@kernel.org>
