drm/i915: Restore the wait for idle engine after flushing interrupts
authorChris Wilson <chris@chris-wilson.co.uk>
Fri, 10 Nov 2017 11:25:50 +0000 (11:25 +0000)
committerChris Wilson <chris@chris-wilson.co.uk>
Fri, 10 Nov 2017 12:55:57 +0000 (12:55 +0000)
commit30b29406d9374989f34bce0eadaa630813049808
tree223cb385b5e70a64156e5b377d17add34c335d2d
parenta5266db4d31410fbfb4f40118e58085994e83dbc
drm/i915: Restore the wait for idle engine after flushing interrupts

So it appears that commit 5427f207852d ("drm/i915: Bump wait-times for
the final CS interrupt before parking") was a little over optimistic in
its belief that it had successfully waited for all residual activity on
the engines before parking. Numerous sightings in CI since then of

<7>[   52.542886] [IGT] core_auth: executing
<3>[   52.561013] [drm:intel_engines_park [i915]] *ERROR* vcs0 is not idle before parking
<7>[   52.561215] intel_engines_park vcs0
<7>[   52.561229] intel_engines_park  current seqno 98, last 98, hangcheck 0 [-247449 ms], inflight 0
<7>[   52.561238] intel_engines_park  Reset count: 0
<7>[   52.561266] intel_engines_park  Requests:
<7>[   52.561363] intel_engines_park  RING_START: 0x00000000 [0x00000000]
<7>[   52.561377] intel_engines_park  RING_HEAD:  0x00000000 [0x00000000]
<7>[   52.561390] intel_engines_park  RING_TAIL:  0x00000000 [0x00000000]
<7>[   52.561406] intel_engines_park  RING_CTL:   0x00000000
<7>[   52.561422] intel_engines_park  RING_MODE:  0x00000200 [idle]
<7>[   52.561442] intel_engines_park  ACTHD:  0x00000000_00000000
<7>[   52.561459] intel_engines_park  BBADDR: 0x00000000_00000000
<7>[   52.561474] intel_engines_park  Execlist status: 0x00000301 00000000
<7>[   52.561489] intel_engines_park  Execlist CSB read 5 [5 cached], write 5 [5 from hws], interrupt posted? no
<7>[   52.561500] intel_engines_park  ELSP[0] idle
<7>[   52.561510] intel_engines_park  ELSP[1] idle
<7>[   52.561519] intel_engines_park  HW active? 0x0
<7>[   52.561608] intel_engines_park Idle? yes
<7>[   52.561617] intel_engines_park

on Braswell, which indicates that the engine just needs that little bit
longer after flushing the tasklet to settle. So give it a few more
milliseconds before declaring an err and applying the emergency brake.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103479
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>
Link: https://patchwork.freedesktop.org/patch/msgid/20171110112550.28909-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
drivers/gpu/drm/i915/intel_engine_cs.c