drm/i915/gt: Prevent queuing retire workers on the virtual engine
authorChris Wilson <chris@chris-wilson.co.uk>
Thu, 6 Feb 2020 20:49:12 +0000 (20:49 +0000)
committerChris Wilson <chris@chris-wilson.co.uk>
Fri, 7 Feb 2020 10:54:49 +0000 (10:54 +0000)
commitf91d8156ab8afb32447cd2bf3189219bab943f18
tree16ee71aa63599516c423ba5b579b8017707b12f6
parent4f7e6d22b545225e27580aeed8a7519307fab7f5
drm/i915/gt: Prevent queuing retire workers on the virtual engine

Virtual engines are fleeting. They carry a reference count and may be freed
when their last request is retired. This makes them unsuitable for the
task of housing engine->retire.work so assert that it is not used.

Tvrtko tracked down an instance where we did indeed violate this rule.
In virtual_submit_request, we flush a completed request directly with
__i915_request_submit and this causes us to queue that request on the
veng's breadcrumb list and signal it. Leading us down a path where we
should not attach the retire.

Reported-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: dc93c9b69315 ("drm/i915/gt: Schedule request retirement when signaler idles")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200206204915.2636606-1-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
drivers/gpu/drm/i915/gt/intel_gt_requests.c