- 08 Feb, 2018 11 commits
-
-
Chris Wilson authored
After we assert the reset request (and wait for 20us), when the device has been fully reset it asserts the reset-status bit. Before we stop requesting the reset and allow the device to return to normal, we should wait for the reset to be completed. (Similar to how we wait for the device to return to normal after deasserting the reset request.) v2: Rename i915_reset_completed() probe to not cause as much confusion. Signed-off-by:
Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180207222824.29864-1-chris@chris-wilson.co.ukReviewed-by:
Mika Kuoppala <mika.kuoppala@linux.intel.com>
-
Chris Wilson authored
Although the mmio are uncached and so should be flushed on every write, be paranoid and do a mmio read after setting the ring head/tail to be sure they have taken effect before moving on. v2: post tail to be pleasing to the eye Signed-off-by:
-
Chris Wilson authored
Reduce the window of opportunity for set-wedged being called concurrently with reset (after i915_reset() has performed the i915_gem_unset_wedged()) by moving the set_bit(I915_WEDGED) to before we complete the inflight requests. When i915_reset() is being blocked on a request, such completion may allow it to start and beginning resetting the GPU before i915_gem_set_wedged() has finished (and so before set-wedge will have marked the device as wedged). As such, i915_gem_init_hw() may see a wedged device even from inside i915_reset(). References: 36703e79 ("drm/i915: Break modeset deadlocks on reset") Signed-off-by:
Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by:
-
Chris Wilson authored
Userspace provides a 64b value for the priority, we need to be careful to preserve the full range before validation to prevent truncation (and letting an illegal value pass). Reported-by:
Antonio Argenziano <antonio.argenziano@intel.com> Fixes: ac14fbd4 ("drm/i915/scheduler: Support user-defined priorities") Signed-off-by:
-
Chris Wilson authored
We only need to wake up the RPS worker once when initially enabling the client boost, it remains in effect then until the last client no longer requires the boost. References: https://bugs.freedesktop.org/show_bug.cgi?id=102250 References: 7b92c1bd ("drm/i915: Avoid keeping waitboost active for signaling threads") Signed-off-by:
Michał Winiarski <michal.winiarski@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180206143137.15509-1-chris@chris-wilson.co.uk
-
Chris Wilson authored
drivers/gpu/drm/i915/i915_oa_cnl.c: In function ‘i915_perf_load_test_config_cnl’: drivers/gpu/drm/i915/i915_oa_cnl.c:99:2: error: ‘strncpy’ output truncated before terminating nul copying 36 bytes from a string of the same length [-Werror=stringop-truncation] v2: strlcpy Fixes: 95690a02 ("drm/i915/perf: enable perf support on CNL") Signed-off-by:
Lionel Landwerlin <lionel.g.landwerlin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180208102403.5587-2-chris@chris-wilson.co.uk
-
Chris Wilson authored
drivers/gpu/drm/i915/i915_oa_cflgt3.c: In function ‘i915_perf_load_test_config_cflgt3’: drivers/gpu/drm/i915/i915_oa_cflgt3.c:87:2: error: ‘strncpy’ output truncated before terminating nul copying 36 bytes from a string of the same length [-Werror=stringop-truncation] v2: strlcpy Fixes: 4407eaa9 ("drm/i915/perf: add support for Coffeelake GT3") Signed-off-by:
-
Daniele Ceraolo Spurio authored
Since commit 5896a5c8 (drm/i915: Always stop the rings before a missing GPU reset) we attempt to stop the engines during gem_sanitize even if reset=0 and nothing bad happened on the gpu. The specs says that the STOP_RINGS bit needs to be cleared to resume normal operation, but for some reason the value of the bit seems to be changing without us writing to it (maybe rc6 entry/exit?), so normal operation resumes correctly. However, it still feels incorrect to stop the engines if there hasn't been any issue so skip the whole reset call in gem_sanitize if i915.reset=0 Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by:
Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180207212440.13438-1-daniele.ceraolospurio@intel.comReviewed-by:
-
Chris Wilson authored
If we remove some hardcoded assumptions about the preempt context having a fixed id, reserved from use by normal user contexts, we may only allocate the i915_gem_context when required. Then the subsequent decisions on using preemption reduce to having the preempt context available. v2: Include an assert that we don't allocate the preempt context twice. Signed-off-by:
Michel Thierry <michel.thierry@intel.com>
-
Chris Wilson authored
Rather than having the high level ioctl interface guess the underlying implementation details, having the implementation declare what capabilities it exports. We define an intel_driver_caps, similar to the intel_device_info, which instead of trying to describe the HW gives details on what the driver itself supports. This is then populated by the engine backend for the new scheduler capability field for use elsewhere. v2: Use caps.scheduler for validating CONTEXT_PARAM_SET_PRIORITY (Mika) One less assumption of engine[RCS] \o/ Signed-off-by:Tomasz Lis <tomasz.lis@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180207210544.26351-2-chris@chris-wilson.co.ukReviewed-by:
-
Chris Wilson authored
In the next patch, we may only conditionally allocate the preempt-client if there is a global preempt context and so we need to be prepared in case the preempt-client itself is NULL. v2: Grep for more preempt_client. Signed-off-by:
-
- 07 Feb, 2018 8 commits
-
-
Chris Wilson authored
As we peek inside struct device to query members guarded by CONFIG_PM, so must be the code. Reported-by:
kbuild test robot <fengguang.wu@intel.com> Fixes: 1fe699e3 ("drm/i915/pmu: Fix sleep under atomic in RC6 readout") Signed-off-by:
Tvrtko Ursulin <tvrtko.ursulin@intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180207160428.17015-1-chris@chris-wilson.co.uk
-
Tvrtko Ursulin authored
We are not allowed to call intel_runtime_pm_get from the PMU counter read callback since the former can sleep, and the latter is running under IRQ context. To workaround this, we record the last known RC6 and while runtime suspended estimate its increase by querying the runtime PM core timestamps. Downside of this approach is that we can temporarily lose a chunk of RC6 time, from the last PMU read-out to runtime suspend entry, but that will eventually catch up, once device comes back online and in the presence of PMU queries. Also, we have to be careful not to overshoot the RC6 estimate, so once resumed after a period of approximation, we only update the counter once it catches up. With the observation that RC6 is increasing while the device is suspended, this should not pose a problem and can only cause slight inaccuracies due clock base differences. v2: Simplify by estimating on top of PM core counters. (Imre) Signed-off-by:
-
Chris Wilson authored
On blb and pnv, we are seeing sporadic i915 0000:00:02.0: Resetting chip after gpu hang [drm:intel_gpu_reset [i915]] rcs0: timed out on STOP_RING [drm:i915_reset [i915]] *ERROR* Failed hw init on reset -5 which notably lack the actual root cause of the error. Ostensibly it should be the init_ring_common() that failed, but it's error paths are covered by DRM_ERROR. Signed-off-by:
-
Chris Wilson authored
If we submit a request and see that the previous request on this timeline was already signaled, we first do not need to add the dependency tracker for that completed request and secondly we know that we there is then a large backlog in retiring requests affecting this timeline. Given that we just submitted more work to the HW, now would be a good time to catch up on those retirements. v2: Try to sum up the compromises involved in flushing the retirement queue after submission. Signed-off-by:
-
Chris Wilson authored
If the last request on the timeline is already complete, we do not need to emit the serialisation barriers. Signed-off-by:
-
Maarten Lankhorst authored
According to bspec, result_lines > 31 is only a maximum for latency level 1 through 7. For level 0 the number of lines is ignored, so always write 0 there to prevent overflowing the 5 bits value. This is required to make NV12 work. Changes since v1: - Rebase on top of GEN11 wm changes. It seems to use res_lines for level 0 limit calculations, but still doesn't appear to program it. Signed-off-by:
Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> #v1 Link: https://patchwork.freedesktop.org/patch/msgid/20180205105841.31634-1-maarten.lankhorst@linux.intel.com
-
Chris Wilson authored
When a request is preempted, it is unsubmitted from the HW queue and removed from the active list of breadcrumbs. In the process, this however triggers the signaler and it may see the clear rbtree with the old, and still valid, seqno, or it may match the cleared seqno with the now zero rq->global_seqno. This confuses the signaler into action and signaling the fence. Fixes: d6a2289d ("drm/i915: Remove the preempted request from the execution queue") Signed-off-by:
-
Joonas Lahtinen authored
Signed-off-by:
-
- 06 Feb, 2018 2 commits
-
-
Tvrtko Ursulin authored
Commit 99e48bf9 ("drm/i915: Lock out execlist tasklet while peeking inside for busy-stats") added a tasklet_disable call in busy stats enabling, but we failed to understand that the PMU enable callback runs as an hard IRQ (IPI). Consequence of this is that the PMU enable callback can interrupt the execlists tasklet, and will then deadlock when it calls intel_engine_stats_enable->tasklet_disable. To fix this, I realized it is possible to move the engine stats enablement and disablement to PMU event init and destroy hooks. This allows for much simpler implementation since those hooks run in normal context (can sleep). v2: Extract engine_event_destroy. (Chris Wilson) Signed-off-by:
-
Rafael Antognolli authored
This workaround should prevent a bug that can be hit on a context restore. To avoid the issue, we must emit a PIPE_CONTROL with CS stall (0x7a000004 0x00100000 0x00000000 0x00000000) followed by 12DW's of NOOP(0x0) in the indirect context batch buffer, to ensure the engine is idle prior to programming 3DSTATE_SAMPLE_PATTERN. It's also not clear whether we should add those extra dwords because of the workaround itself, or if that's just padding for the WA BB (and next commands could come right after the PIPE_CONTROL). We keep them for now. References: HSD#1939868 v2: More descriptive changelog and comments. v3: Explain that PIPE_CONTROL is actually 6 dwords, and that we advance 10 more dwords because of that. Signed-off-by:Rafael Antognolli <rafael.antognolli@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Acked-by:
-
- 05 Feb, 2018 14 commits
-
-
Michal Srb authored
The command MEDIA_VFE_STATE checks bits at offset +2 dwords. However, it is possible to have MEDIA_VFE_STATE command with length = 0 + LENGTH_BIAS = 2. In that case check_cmd will read bits from the following command, or even past the end of the buffer. If the offset ends up outside of the command length, reject the command. Fixes: 351e3db2 ("drm/i915: Implement command buffer parsing logic") Signed-off-by:
Michal Srb <msrb@suse.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180205151745.29292-1-msrb@suse.comReviewed-by:
-
Michal Srb authored
The find_reg function was assuming that there is always at least one table in reg_tables. It is not always true. In case of VCS or VECS, the reg_tables is NULL and reg_table_count is 0, implying that no register-accessing commands are allowed. However, the command tables include commands such as MI_STORE_REGISTER_MEM. When trying to check such command, the find_reg would dereference NULL pointer. Now it will just return NULL meaning that the register was not found and the command will be rejected. Fixes: 76ff480e ("drm/i915/cmdparser: Use binary search for faster register lookup") Signed-off-by:
-
Ville Syrjälä authored
Deprecate the silly I915_SET_COLORKEY_NONE flag. The obvious way to disable colorkey is to just set flags to 0, which is exactly what the intel ddx has been doing all along. Currently when userspace sets the flags to 0, we end up in a funny state where colorkey is disabled, but various colorkey vs. scaling checks still consider colorkey to be enabled, and thus we don't allow plane scaling to kick in. In case there is some other userspace out there that actually uses this flag (unlikely as this is an i915 specific uapi) we'll keep on accepting it. Signed-off-by:
Ville Syrjälä <ville.syrjala@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/20180202204231.27905-1-ville.syrjala@linux.intel.comReviewed-by:
-
Chris Wilson authored
Since commit 7b6da818 ("drm/i915: Restore the kernel context after a GPU reset on an idle engine") we submit a request following the engine reset. The intent is that we don't submit a request if the engine is busy (as it will restart active by itself) but we only checked to see if there were remaining requests in flight on the hardware and skipped checking to see if there were any ready requests that would be immediately submitted on restart (the same time as our new request would be). Having convinced the engine to appear idle in the previous patch, we can use intel_engine_is_idle() as a better test to only submit a new request if there are no pending requests. As it happens, this is tripping up igt/drv_selftest/live_hangcheck in CI as we overfill the kernel_context ringbuffer trigger an infinite recursion from within the reset. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104786 References: 7b6da818 ("drm/i915: Restore the kernel context after a GPU reset on an idle engine") Signed-off-by:
-
Chris Wilson authored
In preparation for the next patch, we want the engine to appear idle after a reset (if there are no requests in flight). For execlists, this entails clearing the active status on reset, it will be regenerated on restarting the engine after the reset. In the process, note that a couple of other status flags and checks could be moved into the describing function. Signed-off-by:
-
Chris Wilson authored
Avoid injecting hangs in to the i915->kernel_context in case the GPU reset leaves corruption in the context image in its wake (leading to continual failures and system hangs after the selftests are ostensibly complete). Use a sacrificial kernel_context instead. v2: Closing a context is tricky; export a function (for selftests) from i915_gem_context.c to get it right. Signed-off-by:
-
Chris Wilson authored
When injecting rapid resets, we must be careful to at least wait for the previous reset to have taken effect and the engine restarted. If we perform a second reset before that has happened, we will notice that the engine hasn't recovered and declare it lost, wedging the device and failing. In practice, since we wait for each hanging batch to start before injecting the reset, this too-fast-reset condition can only be triggered when moving onto the next engine in the test, so we need only wait for the existing reset to complete before switching engines. v2: Wrap up the wait inside a safety net to bail out in case of angry hw. Signed-off-by:
-
Chris Wilson authored
If we remember to cancel the signaler on a request when retiring it (after we know that the request has been signaled), we do not need to carry an additional request in the signaler itself. This prevents an issue whereby the signaler threads may be delayed and hold on to thousands of request references, causing severe memory fragmentation and premature oom (most noticeable on 32b snb due to the limited GFP_KERNEL and frequent use of inter-engine fences). v2: Rename first_signal(), document reads outside of locks. Signed-off-by:
-
Chris Wilson authored
During testing, we trigger a lot of resets on an unbannable context leading to massive amounts of irrelevant debug spam. Remove the ban_score accounting and message for the unbannable context so that we improve the signal:noise in the log messages for when the unexpected occurs. Signed-off-by:
-
Chris Wilson authored
Execlists is now enabled by default and included in the list of capabilities printed out to dmesg and beyond. We do not need to mention it again, every time we restart the engine, so kill the spam. Signed-off-by:
-
Chris Wilson authored
Dump each engine state when i915_gem_set_wedged() is called to give us some more clues as to why we had to terminate the GPU. Signed-off-by:
-
Chris Wilson authored
Provide the reason why we call intel_fbc_deactivate() so that debugging issues with FBC being delayed is clearer. Signed-off-by:
Jani Nikula <jani.nikula@intel.com>
-
Chris Wilson authored
The headers should be on a separate line for consistency, so add the missing trailing newline in a few intel_engine_dump() callers. Reported-by:
-
Chris Wilson authored
Since unbannable contexts are special and supposed not to be causing GPU hangs in the first place, make it clear when they are implicated in said hang. In practice, most unbannable contexts are those created by igt for the express purpose of throwing untold thousands of hangs at the GPU and wish to keep doing so to finish the test. Normally they are cleaned up, but it's when they or the other unbannable kernel contexts stay stuck in an erroneous state that we need to worry and so need highlighting. Suggested-by:
-
- 02 Feb, 2018 5 commits
-
-
Chris Wilson authored
As we ourselves cancel interrupts during reset by clearing the GTIIR, it is possible for the master IIR to indicate a pending IRQ for which we have already cleared from the GTIIR. In this case, the DRM_ERROR are intended and should not be flagged as an error. Signed-off-by:
-
Chris Wilson authored
Be paranoid and flush the GTIIR after clearing the CS interrupt to be sure it has taken before we re-enable the interrupt handler. We still see early interrupts following reset, the tasklet handling the mmio read before it has been written by the CS. This hopefully reduces the frequency to 0... References: https://bugs.freedesktop.org/show_bug.cgi?id=104262Signed-off-by:
-
Jani Nikula authored
Save some horizontal space. Reviewed-by:
David Weinehall <david.weinehall@linux.intel.com> Signed-off-by:
-
Michal Wajdeczko authored
We're using i915_inject_load_failure() to inject dummy faults during driver load, but since this is debug utility we shouldn't expose it in default config as it consumes both code and data. add/remove: 0/1 grow/shrink: 0/2 up/down: 0/-302 (-302) Function old new delta __i915_inject_load_failure 61 - -61 i915_gem_init 1331 1268 -63 i915_driver_load 5923 5745 -178 Total: Before=1177454, After=1177152, chg -0.03% add/remove: 0/1 grow/shrink: 0/0 up/down: 0/-4 (-4) Data old new delta i915_load_fail_count 4 - -4 Total: Before=56762, After=56758, chg -0.01% add/remove: 4/8 grow/shrink: 0/1 up/down: 245/-591 (-346) RO Data old new delta __param_str_inject_load_failure 20 - -20 __UNIQUE_ID_inject_load_failuretype200 34 - -34 __param_inject_load_failure 40 - -40 __func__ 4998 4896 -102 __UNIQUE_ID_inject_load_failure201 150 - -150 Total: Before=119095, After=118749, chg -0.29% Signed-off-by:
Michal Wajdeczko <michal.wajdeczko@intel.com> Cc: Chris Wilson <chris@chris-wilson.co.uk> Link: https://patchwork.freedesktop.org/patch/msgid/20180201173248.3912-1-michal.wajdeczko@intel.comReviewed-by:
-
Jani Nikula authored
We have the max DP link rate info available in VBT since BDB version 216, included in child device config since commit c4fb60b9 ("drm/i915/bios: add DP max link rate to VBT child device struct"). Parse it and use it. Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com> Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com> Reviewed-by:
Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by:
-
