drm/i915: Prevent timeline updates whilst performing reset

As the fence may be signaled concurrently from an interrupt on another
device, it is possible for the list of requests on the timeline to be
modified as we walk it. Take both (the context's timeline and the global
timeline) locks to prevent such modifications.

Fixes: 80b204bce8 ("drm/i915: Enable multiple timelines")
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>
Cc: <drm-intel-fixes@lists.freedesktop.org>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161223145804.6605-10-chris@chris-wilson.co.uk
(cherry picked from commit 00c25e3f40)
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
This commit is contained in:
Chris Wilson 2016-12-23 14:58:04 +00:00 committed by Jani Nikula
parent 64d1461ce0
commit 2471eb5fb6

View file

@ -2730,6 +2730,7 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
struct drm_i915_gem_request *request;
struct i915_gem_context *incomplete_ctx;
struct intel_timeline *timeline;
unsigned long flags;
bool ring_hung;
if (engine->irq_seqno_barrier)
@ -2765,13 +2766,20 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
if (i915_gem_context_is_default(incomplete_ctx))
return;
timeline = i915_gem_context_lookup_timeline(incomplete_ctx, engine);
spin_lock_irqsave(&engine->timeline->lock, flags);
spin_lock(&timeline->lock);
list_for_each_entry_continue(request, &engine->timeline->requests, link)
if (request->ctx == incomplete_ctx)
reset_request(request);
timeline = i915_gem_context_lookup_timeline(incomplete_ctx, engine);
list_for_each_entry(request, &timeline->requests, link)
reset_request(request);
spin_unlock(&timeline->lock);
spin_unlock_irqrestore(&engine->timeline->lock, flags);
}
void i915_gem_reset(struct drm_i915_private *dev_priv)