We dump device flags in few places (init_early, debugfs, gpu_error)
using different functions. Lets add reusable function to avoid
code duplication.
add/remove: 1/0 grow/shrink: 0/3 up/down: 1296/-3572 (-2276)
Function old new delta
intel_device_info_dump_flags - 1296 +1296
i915_capabilities 2435 1353 -1082
i915_error_state_to_str 6642 5507 -1135
intel_device_info_dump 1507 152 -1355
Total: Before=1287992, After=1285716, chg -0.18%
Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
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/20171219114346.26308-1-michal.wajdeczko@intel.com
drivers/gpu/drm/i915/intel_ddi.c:2098 intel_ddi_clk_select() warn: inconsistent indenting
References: 8edcda1266 ("drm/i915: Protect DDI port to DPLL map from theoretical race.")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171219112649.9388-1-chris@chris-wilson.co.uk
After a reset, the state of the CSB registers are scrubbed and not valid
until a powercontext is reloaded. We only know when a powercontext has
been reloaded once we see a CS-interrupt, before then we must ignore the
CSB registers within the execlists_submission_tasklet. However, glk is
sporadically dying with an illegal CSB pointer value (both in the HSWP
and mmio) suggesting that it is running with the CS-interrupt bit set
before the powercontext has been reloaded. Make sure the clearing of
that bit is serialised on reset with the re-enabling of the tasklet.
References: https://bugs.freedesktop.org/show_bug.cgi?id=104262
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171219090110.11153-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
In case we have multiple modesets for different connectors
happening in parallel we could have a race on the RMW on these
shared registers.
This possibility was initially raised by Paulo when reviewing
commit '555e38d27317 ("drm/i915/cnl: DDI - PLL mapping")'
but the original possibility comes from commit '5416d871136d
("drm/i915/skl: Set the eDP link rate on DPLL0")'. Or maybe
later when atomic commits entered into picture.
Apparently the discussion around this topic showed that the
right solution would be on serializing the atomic commits in
a way that we don't have the possibility of races here since
if that parallel modeset happenings apparently many other
things will be on fire.
Code is there since SKL and there was no report of issue,
but since we never looked back to that serialization possibility,
and also we don't have an igt case for that it is better to at
least protect this corner.
Suggested-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Fixes: 555e38d273 ("drm/i915/cnl: DDI - PLL mapping")
Fixes: 5416d87113 ("drm/i915/skl: Set the eDP link rate on DPLL0")
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst maarten.lankhorst@linux.intel.com
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Maarten Lankhorst maarten.lankhorst@linux.intel.com
Link: https://patchwork.freedesktop.org/patch/msgid/20171215224310.19103-1-rodrigo.vivi@intel.com
Now that we skip a per-engine reset on an idle engine, we need to update
the selftest to take that into account. In the process, we find that we
were not stressing the per-engine reset very hard, so add those missing
active resets.
v2: Actually test i915_reset_engine() by loading it with requests.
Fixes: f6ba181ada ("drm/i915: Skip an engine reset if it recovered before our preparations")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104313
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>
Link: https://patchwork.freedesktop.org/patch/msgid/20171217132852.30642-3-chris@chris-wilson.co.uk
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
When monitoring the GPU with i915 perf, reports are tagged with a hw
id. Gem context creation tracepoints already have a hw_id field,
unfortunately you only get this correlation between a process id and a
hw context id once when the context is created. It doesn't help if you
started monitoring after the process was initialized or if the drm fd
was transfered from one process to another.
This change adds the hw_id field to gem requests, so that correlation
can also be done on submission.
v2: Place hw_id at the end of the tracepoint to not disrupt too much
existing tools (Chris)
v3: Reorder hw_id field again (Chris)
v4: Add missing hw_id to i915_gem_request_wait_begin tracepoint (Chris)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171218151959.14073-3-lionel.g.landwerlin@intel.com
Let's make the order of the fields of the tracepoints involving gem
request match across i915. This makes userspace processing of
tracepoint a bit easier.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171218151959.14073-2-lionel.g.landwerlin@intel.com
A useful bit of information for inspecting GPU stalls from
intel_engine_dump() are the error registers, IPEIR and IPEHR.
v2: Fixup gen changes in register offsets (Tvrtko)
v3: Old FADDR location as well
v4: Use I915_READ64_2x32
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171218123914.19027-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
We have an existing helper for testing obj->mm.pages, so use it.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171218103855.25274-1-matthew.auld@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Inside i915_gem_reset(), we start touching the HW and so require the
low-level HW to be re-enabled, in particular the PCI BARs.
Fixes: 7b6da818d8 ("drm/i915: Restore the kernel context after a GPU reset on an idle engine")
References: 0db8c96120 ("drm/i915: Re-enable GTT following a device reset")
Testcase: igt/drv_hangman #i915g/i915gm
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171217132852.30642-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Instead of trying different seq_puts messages, lets use common
-ENODEV error code to indicate missing/unsupported feature.
v2: don't forget about guc_log_control fops (Sagar)
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
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>
Cc: Sujaritha Sundaresan <sujaritha.sundaresan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171215143635.17884-1-michal.wajdeczko@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As part of the system requirement for powersaving is that we always have
a context loaded. Upon boot and resume, we load the kernel_context to
ensure that some valid state is set before powersaving kicks in, we
should do so after a full GPU reset as well. We only need to do so for
an idle engine, as any active engines will restart by executing the
stuck request, loading its context. For the idle engine, we create a
new request to load the kernel_context instead.
For whatever reason, perfoming a dummy execute on the idle engine after
reset papers over a subsequent GPU hang in rare circumstances, even on
machines not using contexts (e.g. Pineview).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104259
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104261
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171216000334.8197-1-chris@chris-wilson.co.uk
At the beginning of a reset, we disable the submission method and find
the stuck request. We expect to find a stuck request for we have
declared the engine stalled. However, if we find no active request, the
engine must have recovered from its stall before we could issue a reset,
so let the engine continue on without a reset. If the engine is truly
stuck, we will back soon enough with the next reset attempt.
v2: Remove the stale debug message.
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: Michel Thierry <michel.thierry@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171216002206.31737-1-chris@chris-wilson.co.uk
Just printk the string, or at least do not double up on the newlines!
Fixes: eef57324d9 ("drm/i915: setup bridge for HDMI LPE audio driver")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Cc: Jerome Anand <jerome.anand@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171213182858.2159-1-chris@chris-wilson.co.uk
Internal objects consistent of scratch pages not subject to the
persistence guarantees of user facing objects. They are used for
example, in ring buffers where they are only required for temporary
storage of commands that will be rewritten every time. As they are
temporary constructs, quietly report -ENOMEM back along the callchain
rather than subject the system to oomkiller if an allocation fails.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171215101753.1519-1-chris@chris-wilson.co.uk
The code has an ifdef and uses two functions to either init the bare
spinlock or init it and set a lock-class. It is possible to do the same
thing without an ifdef.
With this patch (in debug case) we first use the "default" lock class
which is later overwritten to the supplied one. Without lockdep the set
name/class function vanishes.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171214131009.7479-1-joonas.lahtinen@linux.intel.com
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Knowing the state of the engine when hangcheck thinks it is stalling is
useful for both debugging hangcheck itself and the potential cause of an
unwanted stall.
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>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171214122613.26134-1-chris@chris-wilson.co.uk
We have the selftest that's checking doorbell create/destroy, so there's
no need to check all doorbells delaying the reset every time.
We do want to have that extra sanity check at module load/unload though.
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171213221352.7173-7-michal.winiarski@intel.com
We can now move the clients allocation to submission_init path, rather
than keeping the condition inside submission_enable called on every
reset.
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171213221352.7173-6-michal.winiarski@intel.com
Full GPU reset causes GuC to be reset. This means that every time we're
doing a reset, we need to talk to GuC and tell it about doorbells.
Let's separate the communication part (create_doorbell) from our
internal bookkeeping (reserve_doorbell) so that we can cleanly separate
the initialization done at module load from reinitialization done at
reset in the following patch.
While I'm here, let's also add a proper (although slightly asymetric)
cleanup that doesn't try to communicate with GuC after it's already
gone, getting rid of "expected" warnings caused by GuC action failures
on module unload.
Note that I've also removed one of the tests (bitmap out of sync), since
it doesn't make much sense anymore - bitmaps are now not expected to
change during the lifetime of a client.
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20171213221352.7173-5-michal.winiarski@intel.com
To make this operation a bit cleaner, we should make sure that the HW
can catch up by calling the new implementation right away.
Note that currently we're only touching the vfunc at module load time
(before GuC is even loaded), so this shouldn't cause any functional
changes.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Wajdeczko <michal.wajdeczko@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/20171213221352.7173-4-michal.winiarski@intel.com
After GPU reset, GuC HW needs to be reinitialized (with FW reload).
Unfortunately, we're doing some extra work there (mostly allocating stuff),
work that can be moved to guc_init and called once at driver load time.
As a side effect we're no longer hitting an assert in
i915_ggtt_enable_guc on suspend/resume.
v2: Do not duplicate disable_communication / reset_guc_interrupts
v3: Add proper teardown after rebase
References: 04f7b24ecc ("drm/i915/guc: Assert that we switch between known ggtt->invalidate functions")
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@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/20171213221352.7173-3-michal.winiarski@intel.com
Since Michal introduced new user controllable errors other than -EIO
during i915_gem_init(), we need to actually unwind on the error path as
we have to abort the module load (and we expect to do so cleanly!).
As we now teardown key state and then mark the driver as wedged (on
EIO), we have to be careful to not allow ourselves to resume and
unwedge, thus attempting to use the uninitialised driver.
v2: Try not to free driver state for the suppressed EIO
v3: Use load-fault-injection to test both error/recovery paths.
References: 8620eb1dbb ("drm/i915/uc: Don't use -EIO to report missing firmware")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171213134347.4608-1-chris@chris-wilson.co.uk
If we fail to allocate a request, we can reap the outstanding requests
and push them to the request's slab's freelist before trying again. This
forces us to ratelimit malicious clients that tie up all of the system
resources in requests, instead of causing a system-wide oom.
Testcase: igt/gem_shrink/execbuf1
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/20171212180652.22061-3-chris@chris-wilson.co.uk
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
As kmalloc is allowed to block (if given the right flags), mark up the
two i915_sw_fence routines that may call kmalloc as potential sleeping
routines.
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>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171212180652.22061-1-chris@chris-wilson.co.uk
It is illegal to perform an immediate free of the struct irq_work from
inside the irq_work callback (as irq_work_run_list modifies work->flags
after execution of the work->func()). As we use the irq_work to
coordinate the freeing of the callback from two different softirq paths,
we need to defer the kfree from inside our irq_work callback, for which
we can use kfree_rcu.
Fixes: 81c0ed21aa ("drm/i915/fence: Avoid del_timer_sync() from inside a timer")
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>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171213094802.28243-1-chris@chris-wilson.co.uk
If wait_for_engines() fails and we resort to declaring the HW wedged,
dump the engine state for debugging.
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>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211194135.27095-2-chris@chris-wilson.co.uk
Extract the timeout we use in i915_gem_idle_work_handler() and reuse it
for wait_for_engines() in i915_gem_wait_for_idle(). It too has the same
problem in sometimes having to wait for an extended period before the HW
settles, so make use of the same timeout.
References: 5427f20785 ("drm/i915: Bump wait-times for the final CS interrupt before parking")
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>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211194135.27095-1-chris@chris-wilson.co.uk
It never meant what it said, as it was always the total size of the
Global GTT and not a limit upon memory usage. Originally it served as a
quick guide to the largest batch that could be submitted by userspace,
an approximation to its maximum RSS, but was phrased badly. Today with
the 48b ppgtt, it is even more meaningless. Replace with a more specific
debug message; those wanting to know how much "video ram" they have
should consult the userspace libraries for the relevant approximation.
v2: Rebase
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171212113532.22574-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Since on gen2, we do not universally have a GPU reset implementation, we
fail i915_reset() at intel_has_gpu_reset(). However, this is also
intentionally disabled for CI testing and so it only has a debug
message. Promote that debug message to a user-facing error message that
should explain why their machine became unusable following the GPU hang.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211204040.22858-1-chris@chris-wilson.co.uk
Keeps things consistent now that we make use of struct resource. This
should keep us covered in case we ever get huge amounts of stolen
memory.
v2: bunch of missing conversions (Chris)
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-10-matthew.auld@intel.com
Kick it out of i915_ggtt and keep it grouped with dsm and dsm_reserved,
where it makes the most sense.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-9-matthew.auld@intel.com
Now that we are using struct resource to track the stolen region, it is
more convenient if we track the mappable region in a resource as well.
v2: prefer iomap and gmadr naming scheme
prefer DEFINE_RES_MEM
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-8-matthew.auld@intel.com
Now that we are using struct resource to track the stolen region, it is
more convenient if we track the reserved portion of that region in a
resource as well.
v2: s/<= end + 1/< end/ (Chris)
v3: prefer DEFINE_RES_MEM
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-7-matthew.auld@intel.com
Now that we are using struct resource to track the stolen region, it is
more convenient if we track dsm in a resource as well.
v2: check range_overflow when writing to 32b registers (Chris)
pepper in some comments (Chris)
v3: refit i915_stolen_to_dma()
v4: kill ggtt->stolen_size
v5: some more polish
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-6-matthew.auld@intel.com
We duplicate the stolen discovery code in early-quirks and in i915,
however now that the stolen region is exported as a resource from
early-quirks we can nuke the duplication.
v2: check overflows_type
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171211151822.20953-5-matthew.auld@intel.com
If we attempt to wake up a waiter, who is currently checking the seqno
it will be in the TASK_INTERRUPTIBLE state and ttwu will report success.
However, it is actually awake and functioning -- so delay reporting the
actual wake up until it sleeps. This fixes some spurious claims of
missed_breadcrumbs when running under heavy load; i.e. sufficient load to
preempt away the newly woken waiter before they complete their checks.
However, it does so at the cost of a rare false negative; where the
waiter changes between the check and ttwu -- the only way to fix that
would be to extend the reporting from ttwu where the check could be done
atomically.
v2: Defend against !CONFIG_SMP
v3: Don't filter out calls to wake_up_process
v4: Drop risky microoptimisation to skip wakeups
Testcase: igt/drv_missed_irq # sanity check we do detect missed_breadcrumb()
Testcase: igt/gem_concurrent_blit # for generating false positives
References: https://bugs.freedesktop.org/show_bug.cgi?id=100007
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/20171209124710.1606-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
The intent here was that we would be listening to
i915_gem_request_unsubmit in order to cancel the signaler quickly and
release the reference on the request. Cancelling the signaler is done
directly via intel_engine_cancel_signaling (called from unsubmit), but
that does not directly wake up the signaling thread, and neither does
setting the request->global_seqno back to zero wake up listeners to the
request->execute waitqueue. So the only time that listening to the
request->execute waitqueue would wake up the signaling kthread would be
on the request resubmission, during which time we would already receive
wake ups from rejoining the global breadcrumbs wait rbtree.
Trying to wake up to release the request remains an issue. If the
signaling was cancelled and no other request required signaling, then it
is possible for us to shutdown with the reference on the request still
held. To ensure that we do not try to shutdown, leaking that request, we
kick the signaling threads whenever we disarm the breadcrumbs, i.e. on
parking the engine when idle.
v2: We do need to be sure to release the last reference on stopping the
kthread; asserting that it has been dropped already is insufficient.
Fixes: d6a2289d9d ("drm/i915: Remove the preempted request from the execution queue")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171208121033.5236-1-chris@chris-wilson.co.uk
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Even for the mock i915 device, we need to initialise the
drm.mode_config, as we may ultimately query whether there are any KMS
users deep in the bowels of some paths (e.g. eviction). As we initialise
drm.mode_config we must cleanup after ourselves!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171209210835.32609-1-chris@chris-wilson.co.uk
In case function skl_format_to_drm returns -EINVAL, fmt turns into a huge
number as fmt is of type u32, hence there is an out-of-bounds read when
using fmt as an index for array skl_pixel_formats at line 225:
plane->bpp = skl_pixel_formats[fmt].bpp;
Fix this by comparing the value returned by function skl_format_to_drm
against the size of array skl_pixel_formats, so in case it is greater than
or equal to the number of items contained in skl_pixel_formats, print an
error message and return -EINVAL.
Addresses-Coverity-ID: 1462495
Addresses-Coverity-ID: 1462502 ("Out-of-bounds read")
Fixes: 9f31d1063b ("drm/i915/gvt: Add framebuffer decoder support")
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
These 2 functions are coded by multiple person in multiple patches. The
'return' and 'goto err' are mix-used in same place, which cause the
function looks disorder. Unify to use only 'goto' so that the gvt lock
is acquired in one place and released in one place.
Signed-off-by: Pei Zhang <pei.zhang@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
