When decoding the semaphores inside hangcheck, we need to use the hw-id
and not the local array index.
Fixes: de1add3605 ("drm/i915: Decouple execbuf uAPI ...")
Testcase: igt/gem_exec_whisper/hang # gen6-7
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@vger.kernel.org
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161003124516.12388-3-chris@chris-wilson.co.uk
(cherry picked from commit 348b9b1192)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Following commit 821ed7df6e ("drm/i915: Update reset path to fix
incomplete requests") we no longer mark the context as lost on reset as
we keep the requests (and contexts) alive. However, RPS remains reset
and we need to restore the current state to match the in-flight
requests.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97824
Fixes: 821ed7df6e ("drm/i915: Update reset path to fix incomplete requests")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160921135108.29574-1-chris@chris-wilson.co.uk
(cherry picked from commit f2a91d1a6f)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
No functional changes; just renaming a bit, tweaking a datatype,
prettifying layout, and adding comments, in particular in the
GuC setup code that touches this data.
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1473711577-11454-2-git-send-email-david.s.gordon@intel.com
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If we are waiting upon an external fence, from the pov of hangcheck the
engine is stuck on the last submitted seqno. Currently we give a small
increment to the hangcheck score in order to catch a stuck waiter /
driver. Now that we both have an independent wait hangcheck and may be
stuck waiting on an external fence, resetting the GPU has little effect
on that external fence. As we cannot advance by resetting, skip
incrementing the hangcheck score.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-19-chris@chris-wilson.co.uk
Since we have a cooperative mode now with a direct reset, we can avoid
the contention on struct_mutex and instead try then sleep on the
I915_RESET_IN_PROGRESS bit. If the mutex is held and that bit is
cleared, all is fine. Otherwise, we sleep for a bit and try again. In
the worst case we sleep for an extra second waiting for the mutex to be
released (no one touching the GPU is allowed the struct_mutex whilst the
I915_RESET_IN_PROGRESS bit is set). But when we have a direct reset,
this allows us to clean up the reset worker faster.
v2: Remember to call wake_up_bit() after changing (for the faster wakeup
as promised)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-12-chris@chris-wilson.co.uk
If a waiter is holding the struct_mutex, then the reset worker cannot
reset the GPU until the waiter returns. We do not want to return -EAGAIN
form i915_wait_request as that breaks delicate operations like
i915_vma_unbind() which often cannot be restarted easily, and returning
-EIO is just as useless (and has in the past proven dangerous). The
remaining WARN_ON(i915_wait_request) serve as a valuable reminder that
handling errors from an indefinite wait are tricky.
We can keep the current semantic that knowing after a reset is complete,
so is the request, by performing the reset ourselves if we hold the
mutex.
uevent emission is still handled by the reset worker, so it may appear
slightly out of order with respect to the actual reset (and concurrent
use of the device).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-11-chris@chris-wilson.co.uk
In preparation for introducing a per-engine reset, we can first separate
the mixing of the reset state from the global reset counter.
The loss of atomicity in updating the reset state poses a small problem
for handling the waiters. For requests, this is solved by advancing the
seqno so that a waiter waking up after the reset knows the request is
complete. For pending flips, we still rely on the increment of the
global reset epoch (as well as the reset-in-progress flag) to signify
when the hardware was reset.
The advantage, now that we do not inspect the reset state during reset
itself i.e. we no longer emit requests during reset, is that we can use
the atomic updates of the state flags to ensure that only one reset
worker is active.
v2: Mika spotted that I transformed the i915_gem_wait_for_error() wakeup
into a waiter wakeup.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1470414607-32453-6-git-send-email-arun.siluvery@linux.intel.com
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160909131201.16673-7-chris@chris-wilson.co.uk
Rather than walk the full array of engines checking whether each is in
the mask in turn, we can use the mask to jump to the right engines. This
should quicker for a sparse array of engines or mask, whilst generating
smaller code:
text data bss dec hex filename
1251010 4579 800 1256389 132bc5 drivers/gpu/drm/i915/i915.ko
1250530 4579 800 1255909 1329e5 drivers/gpu/drm/i915/i915.ko
The downside is that we have to pass in a temporary, alas no C99
iterators yet.
[P.S. Joonas doesn't like having to pass extra temporaries into the
macro, and even less that I called them tmp. As yet, we haven't found a
macro that avoids passing in a temporary that is smaller. We probably
will get C99 iterators first!]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160827075401.16470-2-chris@chris-wilson.co.uk
If the engine isn't being retired (worker starvation?) then it is
possible for us to repeatedly observe that between consecutive
hangchecks the seqno on the ring to be the same and there remain
unretired requests. Ignore these completely and only regard the engine
as busy for the purpose of hang detection (not stall detection) if there
are outstanding breadcrumbs.
In recent history we have looked at using both the request and seqno as
indication of activity on the engine, but that was reduced to just
inspecting seqno in commit cffa781e59 ("drm/i915: Simplify check for
idleness in hangcheck"). However, in commit dcff85c844 ("drm/i915:
Enable i915_gem_wait_for_idle() without holding struct_mutex"), I made
the decision to use the new common lockless function, under the
assumption that request retirement was more frequent than hangcheck and
so we would not have a stuck busy check. The flaw there was in
forgetting that we accumulate the hang score, and so successive checks
seeing a stuck request, albeit with the GPU advancing elsewhere and so
not necessary the same stuck request, would eventually trigger the hang.
Fixes: dcff85c844 ("drm/i915: Enable i915_gem_wait_for_idle()...")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20160820145408.32180-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
In commit 2529d57050 ("drm/i915: Drop racy markup of missed-irqs from
idle-worker") the racy detection of missed interrupts was removed when
we went idle. This however opened up the issue that the stuck waiters
were not being reported, causing a test case failure. If we move the
stuck waiter detection out of hangcheck and into the breadcrumb
mechanims (i.e. the waiter) itself, we can avoid this issue entirely.
This leaves hangcheck looking for a stuck GPU (inspecting for request
advancement and HEAD motion), and breadcrumbs looking for a stuck
waiter - hopefully make both easier to understand by their segregation.
v2: Reduce the error message as we now run independently of hangcheck,
and the hanging batch used by igt also counts as a stuck waiter causing
extra warnings in dmesg.
v3: Move the breadcrumb's hangcheck kickstart to the first missed wait.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97104
Fixes: 2529d57050 (waiter"drm/i915: Drop racy markup of missed-irqs...")
Testcase: igt/drv_missed_irq
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1470761272-1245-2-git-send-email-chris@chris-wilson.co.uk
No functional change. Instead of defining a new empty function
let's use what is available on drm.
It gets cleaner, and easy to read, and understand.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
The principal motivation for this was to try and eliminate the
struct_mutex from i915_gem_suspend - but we still need to hold the mutex
current for the i915_gem_context_lost(). (The issue there is that there
may be an indirect lockdep cycle between cpu_hotplug (i.e. suspend) and
struct_mutex via the stop_machine().) For the moment, enabling last
request tracking for the engine, allows us to do busyness checking and
waiting without requiring the struct_mutex - which is useful in its own
right.
As a side-effect of having a robust means for tracking engine busyness,
we can replace our other busyness heuristic, that of comparing against
the last submitted seqno. For paranoid reasons, we have a semi-ordered
check of that seqno inside the hangchecker, which we can now improve to
an ordered check of the engine's busyness (removing a locked xchg in the
process).
v2: Pass along "bool interruptible" as being unlocked we cannot rely on
i915->mm.interruptible being stable or even under our control.
v3: Replace check Ironlake i915_gpu_busy() with the common precalculated value
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/1470388464-28458-6-git-send-email-chris@chris-wilson.co.uk
The state stored in this struct is not only the information about the
buffer object, but the ring used to communicate with the hardware. Using
buffer here is overly specific and, for me at least, conflates with the
notion of buffer objects themselves.
s/struct intel_ringbuffer/struct intel_ring/
s/enum intel_ring_hangcheck/enum intel_engine_hangcheck/
s/describe_ctx_ringbuf()/describe_ctx_ring()/
s/intel_ring_get_active_head()/intel_engine_get_active_head()/
s/intel_ring_sync_index()/intel_engine_sync_index()/
s/intel_ring_init_seqno()/intel_engine_init_seqno()/
s/ring_stuck()/engine_stuck()/
s/intel_cleanup_engine()/intel_engine_cleanup()/
s/intel_stop_engine()/intel_engine_stop()/
s/intel_pin_and_map_ringbuffer_obj()/intel_pin_and_map_ring()/
s/intel_unpin_ringbuffer()/intel_unpin_ring()/
s/intel_engine_create_ringbuffer()/intel_engine_create_ring()/
s/intel_ring_flush_all_caches()/intel_engine_flush_all_caches()/
s/intel_ring_invalidate_all_caches()/intel_engine_invalidate_all_caches()/
s/intel_ringbuffer_free()/intel_ring_free()/
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/1469432687-22756-15-git-send-email-chris@chris-wilson.co.uk
Link: http://patchwork.freedesktop.org/patch/msgid/1470174640-18242-4-git-send-email-chris@chris-wilson.co.uk
Ringbuffers are now being written to either through LLC or WC paths, so
treating them as simply iomem is no longer adequate. However, for the
older !llc hardware, the hardware is documentated as treating the TAIL
register update as serialising, so we can relax the barriers when filling
the rings (but even if it were not, it is still an uncached register write
and so serialising anyway.).
For simplicity, let's ignore the iomem annotation.
v2: Remove iomem from ringbuffer->virtual_address
v3: And for good measure add iomem elsewhere to keep sparse happy
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> #v2
Link: http://patchwork.freedesktop.org/patch/msgid/1469005202-9659-8-git-send-email-chris@chris-wilson.co.uk
Link: http://patchwork.freedesktop.org/patch/msgid/1469017917-15134-7-git-send-email-chris@chris-wilson.co.uk
To allow the user finer control over waitboosting, allow them to set the
frequency we request for the boost. This also them allows to effectively
disable the boosting by setting the boost request to a low frequency.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1468397438-21226-5-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Some Kabylake SKUs are going to use Kabypoint PCH.
It is mainly for Halo and DT ones.
>From our specs it doesn't seem that KBP brings
any change on the display south engine. So let's consider
this as a continuation of SunrisePoint, i.e., SPT+.
Since it is easy to get confused by a letter change:
KBL = Kabylake - CPU/GPU codename.
KBP = Kabypoint - PCH codename.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96826
Link: http://patchwork.freedesktop.org/patch/msgid/1467418032-15167-1-git-send-email-rodrigo.vivi@intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
As we inspect both the tasklet (to check for an active bottom-half) and
set the irq-posted flag at the same time (both in the interrupt handler
and then in the bottom-halt), group those two together into the same
cacheline. (Not having total control over placement of the struct means
we can't guarantee the cacheline boundary, we need to align the kmalloc
and then each struct, but the grouping should help.)
v2: Try a couple of different names for the state touched by the user
interrupt handler.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467805142-22219-3-git-send-email-chris@chris-wilson.co.uk
Since drm_i915_private is now a subclass of drm_device we do not need to
chase the drm_i915_private->dev backpointer and can instead simply
access drm_i915_private->drm directly.
text data bss dec hex filename
1068757 4565 416 1073738 10624a drivers/gpu/drm/i915/i915.ko
1066949 4565 416 1071930 105b3a drivers/gpu/drm/i915/i915.ko
Created by the coccinelle script:
@@
struct drm_i915_private *d;
identifier i;
@@
(
- d->dev->i
+ d->drm.i
|
- d->dev
+ &d->drm
)
and for good measure the dev_priv->dev backpointer was removed entirely.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467711623-2905-4-git-send-email-chris@chris-wilson.co.uk
Remove some redundant kernel messages as we deduce a hung GPU and
capture the error state.
v2: Fix "hang" vs "no progress" message whilst I was there
v3: s/snprintf/scnprintf/
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467618513-4966-2-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
After Joonas complained about using READ_ONCE() on the only use of the
variable in the function, where the intent was to simply document that
the read was intentionally racy and unlocked, I switched the READ_ONCE()
over to lockless_dereference(). However, in linux-next that has a
stronger type-check to only allow pointers and is no longer
interchangeable with READ_ONCE(), see commit 331b6d8c7a
("locking/barriers: Validate lockless_dereference() is used on a pointer
type")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 67d97da349 ("drm/i915: Only start retire worker when idle")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1467705276-707-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Since we now subclass struct drm_device, we can save pointer dances by
noting the equivalence of struct drm_device and struct drm_i915_private,
i.e. by using to_i915().
text data bss dec hex filename
1073824 4562 416 1078802 107612 drivers/gpu/drm/i915/i915.ko
1068976 4562 416 1073954 106322 drivers/gpu/drm/i915/i915.ko
Created by the coccinelle script:
@@
expression E;
identifier p;
@@
- struct drm_i915_private *p = E->dev_private;
+ struct drm_i915_private *p = to_i915(E);
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467628477-25379-1-git-send-email-chris@chris-wilson.co.uk
Make sure that the RPS bottom-half is flushed before we set the idle
frequency when we decide the GPU is idle. This should prevent any races
with the bottom-half and setting the idle frequency, and ensures that
the bottom-half is bounded by the GPU's rpm reference taken for when it
is active (i.e. between gen6_rps_busy() and gen6_rps_idle()).
v2: Avoid recursively using the i915->wq - RPS does not touch the
struct_mutex so has no place being on the ordered i915->wq.
v3: Enable/disable interrupts for RPS busy/idle in order to prevent
further HW access from RPS outside of the wakeref.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
References: https://bugs.freedesktop.org/show_bug.cgi?id=89728
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467616119-4093-6-git-send-email-chris@chris-wilson.co.uk
The retire worker is a low frequency task that makes sure we retire
outstanding requests if userspace is being lax. We only need to start it
once as it remains active until the GPU is idle, so do a cheap test
before the more expensive queue_work(). A consequence of this is that we
need correct locking in the worker to make the hot path of request
submission cheap. To keep the symmetry and keep hangcheck strictly bound
by the GPU's wakelock, we move the cancel_sync(hangcheck) to the idle
worker before dropping the wakelock.
v2: Guard against RCU fouling the breadcrumbs bottom-half whilst we kick
the waiter.
v3: Remove the wakeref assertion squelching (now we hold a wakeref for
the hangcheck, any rpm error there is genuine).
v4: To prevent excess work when retiring requests, we split the busy
flag into two, a boolean to denote whether we hold the wakeref and a
bitmask of active engines.
v5: Reorder cancelling hangcheck upon idling to avoid a race where we
might cancel a hangcheck after being preempted by a new task
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
References: https://bugs.freedesktop.org/show_bug.cgi?id=88437
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467616119-4093-1-git-send-email-chris@chris-wilson.co.uk
Since the tests can and do explicitly check debugfs/i915_ring_missed_irqs
for the handling of a "missed interrupt", adding it to the dmesg at INFO
is just noise. When it happens for real, we still class it as an ERROR.
Note that I have chose to remove it entirely because when we detect the
"missed interrupt" is irrelevant and the message contains no more
information than we glean from looking in debugfs.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-20-git-send-email-chris@chris-wilson.co.uk
With only a single callsite for intel_engine_cs->irq_get and ->irq_put,
we can reduce the code size by moving the common preamble into the
caller, and we can also eliminate the reference counting.
For completeness, as we are no longer doing reference counting on irq,
rename the get/put vfunctions to enable/disable respectively and are
able to review the use of posting reads. We only require the
serialisation with hardware when enabling the interrupt (i.e. so we
cannot miss an interrupt by going to sleep before the hardware truly
enables it).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-18-git-send-email-chris@chris-wilson.co.uk
If we flag the seqno as potentially stale upon receiving an interrupt,
we can use that information to reduce the frequency that we apply the
heavyweight coherent seqno read (i.e. if we wake up a chain of waiters).
v2: Use cmpxchg to replace READ_ONCE/WRITE_ONCE for more explicit
control of the ordering wrt to interrupt generation and interrupt
checking in the bottom-half.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-14-git-send-email-chris@chris-wilson.co.uk
On Ironlake, there is no command nor register to ensure that the write
from a MI_STORE command is completed (and coherent on the CPU) before the
command parser continues. This means that the ordering between the seqno
write and the subsequent user interrupt is undefined (like gen6+). So to
ensure that the seqno write is completed after the final user interrupt
we need to delay the read sufficiently to allow the write to complete.
This delay is undefined by the bspec, and empirically requires 75us even
though a register read combined with a clflush is less than 500ns. Hence,
the delay is due to an on-chip buffer rather than the latency of the write
to memory.
Note that the render ring controls this by filling the PIPE_CONTROL fifo
with stalling commands that force the earliest pipe-control with the
seqno to be completed before the command parser continues. Given that we
need a barrier operation for BSD, we may as well forgo the extra
per-batch latency by using a common per-interrupt barrier.
Studying the impact of adding the usleep shows that in both sequences of
and individual synchronous no-op batches is negligible for the media
engine (where the write now is unordered with the interrupt). Converting
the render engine over from the current glutton of pie-controls over to
the per-interrupt delays speeds up both the sequential and individual
synchronous no-ops by 20% and 60%, respectively. This speed up holds
even when looking at the throughput of small copies (4KiB->4MiB), both
serial and synchronous, by about 20%. This is because despite adding a
significant delay to the interrupt, in all likelihood we will see the
seqno write without having to apply the barrier (only in the rare corner
cases where the write is delayed on the last required is the delay
necessary).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94307
Testcase: igt/gem_sync #ilk
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-12-git-send-email-chris@chris-wilson.co.uk
By using the same address for storing the HWS on every platform, we can
remove the platform specific vfuncs and reduce the get-seqno routine to
a single read of a cached memory location.
v2: Fix semaphore_passed() to look at the signaling engine (not the
waiter's)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-8-git-send-email-chris@chris-wilson.co.uk
One particularly stressful scenario consists of many independent tasks
all competing for GPU time and waiting upon the results (e.g. realtime
transcoding of many, many streams). One bottleneck in particular is that
each client waits on its own results, but every client is woken up after
every batchbuffer - hence the thunder of hooves as then every client must
do its heavyweight dance to read a coherent seqno to see if it is the
lucky one.
Ideally, we only want one client to wake up after the interrupt and
check its request for completion. Since the requests must retire in
order, we can select the first client on the oldest request to be woken.
Once that client has completed his wait, we can then wake up the
next client and so on. However, all clients then incur latency as every
process in the chain may be delayed for scheduling - this may also then
cause some priority inversion. To reduce the latency, when a client
is added or removed from the list, we scan the tree for completed
seqno and wake up all the completed waiters in parallel.
Using igt/benchmarks/gem_latency, we can demonstrate this effect. The
benchmark measures the number of GPU cycles between completion of a
batch and the client waking up from a call to wait-ioctl. With many
concurrent waiters, with each on a different request, we observe that
the wakeup latency before the patch scales nearly linearly with the
number of waiters (before external factors kick in making the scaling much
worse). After applying the patch, we can see that only the single waiter
for the request is being woken up, providing a constant wakeup latency
for every operation. However, the situation is not quite as rosy for
many waiters on the same request, though to the best of my knowledge this
is much less likely in practice. Here, we can observe that the
concurrent waiters incur extra latency from being woken up by the
solitary bottom-half, rather than directly by the interrupt. This
appears to be scheduler induced (having discounted adverse effects from
having a rbtree walk/erase in the wakeup path), each additional
wake_up_process() costs approximately 1us on big core. Another effect of
performing the secondary wakeups from the first bottom-half is the
incurred delay this imposes on high priority threads - rather than
immediately returning to userspace and leaving the interrupt handler to
wake the others.
To offset the delay incurred with additional waiters on a request, we
could use a hybrid scheme that did a quick read in the interrupt handler
and dequeued all the completed waiters (incurring the overhead in the
interrupt handler, not the best plan either as we then incur GPU
submission latency) but we would still have to wake up the bottom-half
every time to do the heavyweight slow read. Or we could only kick the
waiters on the seqno with the same priority as the current task (i.e. in
the realtime waiter scenario, only it is woken up immediately by the
interrupt and simply queues the next waiter before returning to userspace,
minimising its delay at the expense of the chain, and also reducing
contention on its scheduler runqueue). This is effective at avoid long
pauses in the interrupt handler and at avoiding the extra latency in
realtime/high-priority waiters.
v2: Convert from a kworker per engine into a dedicated kthread for the
bottom-half.
v3: Rename request members and tweak comments.
v4: Use a per-engine spinlock in the breadcrumbs bottom-half.
v5: Fix race in locklessly checking waiter status and kicking the task on
adding a new waiter.
v6: Fix deciding when to force the timer to hide missing interrupts.
v7: Move the bottom-half from the kthread to the first client process.
v8: Reword a few comments
v9: Break the busy loop when the interrupt is unmasked or has fired.
v10: Comments, unnecessary churn, better debugging from Tvrtko
v11: Wake all completed waiters on removing the current bottom-half to
reduce the latency of waking up a herd of clients all waiting on the
same request.
v12: Rearrange missed-interrupt fault injection so that it works with
igt/drv_missed_irq_hang
v13: Rename intel_breadcrumb and friends to intel_wait in preparation
for signal handling.
v14: RCU commentary, assert_spin_locked
v15: Hide BUG_ON behind the compiler; report on gem_latency findings.
v16: Sort seqno-groups by priority so that first-waiter has the highest
task priority (and so avoid priority inversion).
v17: Add waiters to post-mortem GPU hang state.
v18: Return early for a completed wait after acquiring the spinlock.
Avoids adding ourselves to the tree if the is already complete, and
skips the awkward question of why we don't do completion wakeups for
waits earlier than or equal to ourselves.
v19: Prepare for init_breadcrumbs to fail. Later patches may want to
allocate during init, so be prepared to propagate back the error code.
Testcase: igt/gem_concurrent_blit
Testcase: igt/benchmarks/gem_latency
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: "Rogozhkin, Dmitry V" <dmitry.v.rogozhkin@intel.com>
Cc: "Gong, Zhipeng" <zhipeng.gong@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: "Goel, Akash" <akash.goel@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> #v18
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-6-git-send-email-chris@chris-wilson.co.uk
Currently __i915_wait_request uses a per-engine wait_queue_t for the dual
purpose of waking after the GPU advances or for waking after an error.
In the future, we may add even more wake sources and require greater
separation, but for now we can conceptually simplify wakeups by separating
the two sources. In particular, this allows us to use different wait-queues
(e.g. one on the engine advancement, a global one for errors and one on
each requests) without any hassle.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-5-git-send-email-chris@chris-wilson.co.uk
The queue only ever contains at most one item and has no special flags.
It is just a very simple wrapper around the system-wq - a complication
with no benefits.
v2: Use the system_long_wq as we may wish to capture the error state
after detecting the hang - which may take a bit of time.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-3-git-send-email-chris@chris-wilson.co.uk
We can forgo queuing the hangcheck from the start of every request to
until we wait upon a request. This reduces the overhead of every
request, but may increase the latency of detecting a hang. However, if
nothing every waits upon a hang, did it ever hang? It also improves the
robustness of the wait-request by ensuring that the hangchecker is
indeed running before we sleep indefinitely (and thereby ensuring that
we never actually sleep forever waiting for a dead GPU).
As pointed out by Tvrtko, it is possible for a GPU hang to go unnoticed
for as long as nobody is waiting for the GPU. Though this rare, during
that time we may be consuming more power than if we had promptly
recovered, and in the most extreme case we may exhaust all memory before
forcing the hangcheck. Something to be wary off in future.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1467390209-3576-2-git-send-email-chris@chris-wilson.co.uk
On Loading, GuC sets PM interrupts routing (bit 31) and clears ARAT
expired interrupt (bit 9). Host turbo also updates this register
in RPS flows. This patch ensures bit 31 and bit 9 setup by GuC persists.
ARAT timer interrupt is needed in GuC for various features. It also
facilitates halting GuC and hence achieving RC6. PM interrupt routing
will not impact RPS interrupt reception by host as GuC will redirect
them.
This patch fixes igt test pm_rc6_residency that was failing with guc
load/submission enabled. Tested with SKL GuC v6.1 and BXT GuC v5.1 and v8.7.
v2: i915_irq/i915_pm decoupling from intel_guc. (ChrisW)
v3: restructuring the mask update and rebase w.r.t Ville's patch. (ChrisW)
v4: Updating the pm_intr_keep during direct_interrupts_to_guc. (Sagar)
Cc: Chris Harris <chris.harris@intel.com>
Cc: Zhe Wang <zhe1.wang@intel.com>
Cc: Deepak S <deepak.s@intel.com>
Cc: Satyanantha, Rama Gopal M <rama.gopal.m.satyanantha@intel.com>
Cc: Akash Goel <akash.goel@intel.com>
Testcase: igt/pm_rc6_residency
Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Tested-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1464683307-19475-1-git-send-email-sagar.a.kamble@intel.com
This reverts the following patches:
d55dbd06bb drm/i915: Allow nonblocking update of pageflips.
15c86bdb76 drm/i915: Check for unpin correctness.
95c2ccdc82 Reapply "drm/i915: Avoid stalling on pending flips for legacy cursor updates"
a6747b7304 drm/i915: Make unpin async.
03f476e1fc drm/i915: Prepare connectors for nonblocking checks.
2099deffef drm/i915: Pass atomic states to fbc update functions.
ee7171af72 drm/i915: Remove reset_counter from intel_crtc.
2ee004f7c5 drm/i915: Remove queue_flip pointer.
b8d2afae55 drm/i915: Remove use_mmio_flip kernel parameter.
8dd634d922 drm/i915: Remove cs based page flip support.
143f73b3bf drm/i915: Rework intel_crtc_page_flip to be almost atomic, v3.
84fc494b64 drm/i915: Add the exclusive fence to plane_state.
6885843ae1 drm/i915: Convert flip_work to a list.
aa420ddd8e drm/i915: Allow mmio updates on all platforms, v2.
afee4d8707 Revert "drm/i915: Avoid stalling on pending flips for legacy cursor updates"
"drm/i915: Allow nonblocking update of pageflips" should have been
split up, misses a proper commit message and seems to cause issues in
the legacy page_flip path as demonstrated by kms_flip.
"drm/i915: Make unpin async" doesn't handle the unthrottled cursor
updates correctly, leading to an apparent pin count leak. This is
caught by the WARN_ON in i915_gem_object_do_pin which screams if we
have more than DRM_I915_GEM_OBJECT_MAX_PIN_COUNT pins.
Unfortuantely we can't just revert these two because this patch series
came with a built-in bisect breakage in the form of temporarily
removing the unthrottled cursor update hack for legacy cursor ioctl.
Therefore there's no other option than to revert the entire pile :(
There's one tiny conflict in intel_drv.h due to other patches, nothing
serious.
Normally I'd wait a bit longer with doing a maintainer revert, but
since the minimal set of patches we need to revert (due to the bisect
breakage) is so big, time is running out fast. And very soon
(especially after a few attempts at fixing issues) it'll be really
hard to revert things cleanly.
Lessons learned:
- Not a good idea to rush the review (done by someone fairly new to
the area) and not make sure domain experts had a chance to read it.
- Patches should be properly split up. I only looked at the two
patches that should be reverted in detail, but both look like the
mix up different things in one patch.
- Patches really should have proper commit messages. Especially when
doing more than one thing, and especially when touching critical and
tricky core code.
- Building a patch series and r-b stamping it when it has a built-in
bisect breakage is not a good idea.
- I also think we need to stop building up technical debt by
postponing atomic igt testcases even longer. I think it's clear that
there's enough corner cases in this beast that we really need to
have the testcases _before_ the next step lands.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
Cc: John Harrison <John.C.Harrison@Intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
We've never actually enabled or unmasked the GSE interrupt on BDW+,
even though the interrupt handler was always prepared for it.
Let's enable it and see what happens.
Credit to Mark Kettenis who fixed this in the OpenBSD fork of the
driver. He reports that it fixed the "ACPI _BCM/_BCQ-based
brightness mechanism on a MacBookPro12,1 and a 3rd gen Lenovo X1
Carbon" for them.
Mark says:
"FWIW, this *is* needed if you want ACPI-based backlight control to
work. On Linux you probably don't notice, since "hardware" backlight
control is preferred over "firmware" or "platform" backlight control.
It would help me if this did land in the Linux tree though, as it will
make future imports of the i915 driver into OpenBSD easier."
So even though we don't really need this, let's put it in to help Mark
with future porting efforts. Should be harmless to have it enabled in
any case.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
References: http://lists.freedesktop.org/archives/intel-gfx/2015-December/081799.html
Reported-by: Mark Kettenis <mark.kettenis@xs4all.nl>
Cc: Mark Kettenis <mark.kettenis@xs4all.nl>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463649283-28698-1-git-send-email-ville.syrjala@linux.intel.com
Acked-by: Jani Nikula <jani.nikula@intel.com>
Rename intel_unpin_work to intel_flip_work and use it for mmio flips
and unpinning. Use flip_queued_req to hold the wait request in the
mmio case, and the vblank counter from intel_crtc_get_vblank_counter.
MMIO flips get their own path through intel_finish_page_flip_mmio,
handled on vblank. CS page flips go through *_cs.
Changes since v1:
- Clean up destinction between MMIO and CS flips.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463490484-19540-7-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
Instead of calling prepare_flip right before calling finish_page_flip
do everything from prepare_page_flip in finish_page_flip.
Putting prepare and finish page_flip in a single step removes the need
for INTEL_FLIP_COMPLETE, so it can be removed. This simplifies the code
slightly.
Changes since v1:
- Invert if case to simplify code.
- Add missing barrier.
- Reword commit message.
Changes since v2:
- intel_page_flip_plane is removed.
- work->pending is turned into a bool.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463490484-19540-5-git-send-email-maarten.lankhorst@linux.intel.com
Reviewed-by: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
This way optimization from a previous patch works even better.
v2: Rebase.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Pass drm_i915_private to the uncore init/fini routines and their
subservients as it is their native type.
text data bss dec hex filename
6309978 3578778 696320 10585076 a183f4 vmlinux
6309530 3578778 696320 10584628 a18234 vmlinux
a modest 400 bytes of saving, but 60 lines of code deleted!
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/1462885804-26750-1-git-send-email-chris@chris-wilson.co.uk
text data bss dec hex filename
6309351 3578714 696320 10584385 a18141 vmlinux
6308391 3578714 696320 10583425 a17d81 vmlinux
Almost 1KiB of code reduction.
v2: More s/INTEL_INFO()->gen/INTEL_GEN()/ and IS_GENx() conversions
text data bss dec hex filename
6304579 3578778 696320 10579677 a16edd vmlinux
6303427 3578778 696320 10578525 a16a5d vmlinux
Now over 1KiB!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462545621-30125-3-git-send-email-chris@chris-wilson.co.uk
I have noticed some of our interrupt handlers use both dev and
dev_priv while they could get away with only dev_priv in the
huge majority of cases.
Tidying that up had a cascading effect on changing functions
prototypes, so relatively big churn factor, but I think it is
for the better.
For example even where changes cascade out of i915_irq.c, for
functions prefixed with intel_, genX_ or <plat>_, it makes more
sense to take dev_priv directly anyway.
This allows us to eliminate local variables and intermixed usage
of dev and dev_priv where only one is good enough.
End result is shrinkage of both source and the resulting binary.
i915.ko:
- .text 000b0899
+ .text 000b0619
Or if we look at the Gen8 display irq chain:
-00000000000006ad t gen8_irq_handler
+0000000000000663 t gen8_irq_handler
-0000000000000028 T intel_opregion_asle_intr
+0000000000000024 T intel_opregion_asle_intr
-000000000000008c t ilk_hpd_irq_handler
+000000000000007f t ilk_hpd_irq_handler
-0000000000000116 T intel_check_page_flip
+0000000000000112 T intel_check_page_flip
-000000000000011a T intel_prepare_page_flip
+0000000000000119 T intel_prepare_page_flip
-0000000000000014 T intel_finish_page_flip_plane
+0000000000000013 T intel_finish_page_flip_plane
-0000000000000053 t hsw_pipe_crc_irq_handler
+000000000000004c t hsw_pipe_crc_irq_handler
-000000000000022e t cpt_irq_handler
+0000000000000213 t cpt_irq_handler
So small shrinkage but it is all fast paths so doesn't harm.
Situation is similar in other interrupt handlers as well.
v2: Tidy intel_queue_rps_boost_for_request as well. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Looks like DPF was not implemented for gen8+ but the IER and IMR
are still enabled on initialization.
Since there is no code to handle this interrupt, gate the irq
enablement behind HAS_L3_DPF in case the feature gets enabled
in the future.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Chris Wilson