diff options
| author | Chris Wilson <chris@chris-wilson.co.uk> | 2018-10-24 11:49:39 +0100 |
|---|---|---|
| committer | Chris Wilson <chris@chris-wilson.co.uk> | 2018-10-26 12:37:13 +0100 |
| commit | 1e016a8693368c163b18104a28c4edc71858f2d2 (patch) | |
| tree | 935922c99b03ff83ad4d1472f351d9b0ecb3eab5 /drivers/gpu/drm/i915/i915_request.c | |
| parent | 8dd0f8d37eaec2bcb4c73b840f0390fee88583ac (diff) | |
drm/i915: Park signaling thread while wrapping the seqno
A danger encountered when resetting the seqno (using
debugfs/i915_next_seqno) is that as we change the breadcrumb stored in
the HWSP, it may be inspected by the signaler thread leading to
confusion in our sanity checks.
<0> [136.331342] i915/sig-347 3..s1 136336154us : execlists_submission_tasklet: rcs0 awake?=1, active=5
<0> [136.331373] i915/sig-347 3d.s2 136336155us : process_csb: rcs0 cs-irq head=5, tail=0
<0> [136.331402] i915/sig-347 3d.s2 136336155us : process_csb: rcs0 csb[0]: status=0x00000018:0x00000002, active=0x5
<0> [136.331434] i915/sig-347 3d.s2 136336156us : process_csb: rcs0 out[0]: ctx=2.1, global=219 (fence 46:8455) (current 219), prio=0
<0> [136.331466] i915/sig-347 3d.s2 136336156us : process_csb: rcs0 completed ctx=2
<0> [136.332027] gem_exec-1049 0.... 136336246us : reset_all_global_seqno.part.5: rcs0 seqno 219 (current 219) -> -43
<0> [136.332056] gem_exec-1049 0.... 136336251us : reset_all_global_seqno.part.5: bcs0 seqno 183 (current 183) -> -43
<0> [136.332085] gem_exec-1049 0.... 136336255us : reset_all_global_seqno.part.5: vcs0 seqno 191 (current 191) -> -43
<0> [136.332114] gem_exec-1049 0.... 136336259us : reset_all_global_seqno.part.5: vcs1 seqno 180 (current 180) -> -43
<0> [136.332143] gem_exec-1049 0.... 136336262us : reset_all_global_seqno.part.5: vecs0 seqno 212 (current 212) -> -43
<0> [136.332174] i915/sig-347 3.... 136336280us : intel_breadcrumbs_signaler: intel_breadcrumbs_signaler:673 GEM_BUG_ON(!i915_request_completed(rq))
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181024104939.2861-1-chris@chris-wilson.co.uk
Diffstat (limited to 'drivers/gpu/drm/i915/i915_request.c')
| -rw-r--r-- | drivers/gpu/drm/i915/i915_request.c | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c index 28819f8c4da6..71107540581d 100644 --- a/drivers/gpu/drm/i915/i915_request.c +++ b/drivers/gpu/drm/i915/i915_request.c @@ -136,6 +136,8 @@ static int reset_all_global_seqno(struct drm_i915_private *i915, u32 seqno) intel_engine_get_seqno(engine), seqno); + kthread_park(engine->breadcrumbs.signaler); + if (!i915_seqno_passed(seqno, engine->timeline.seqno)) { /* Flush any waiters before we reuse the seqno */ intel_engine_disarm_breadcrumbs(engine); @@ -150,6 +152,8 @@ static int reset_all_global_seqno(struct drm_i915_private *i915, u32 seqno) /* Finally reset hw state */ intel_engine_init_global_seqno(engine, seqno); engine->timeline.seqno = seqno; + + kthread_unpark(engine->breadcrumbs.signaler); } list_for_each_entry(timeline, &i915->gt.timelines, link) |