drm/i915: Remove request retirement before each batch
authorChris Wilson <chris@chris-wilson.co.uk>
Thu, 4 Aug 2016 15:32:20 +0000 (16:32 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Thu, 4 Aug 2016 19:19:51 +0000 (20:19 +0100)
commit0340d9fd0f74309f435d5324b885c9ca1967262c
treefb3778dd54aa3e85313115fd30dc0fd799501815
parent115003e9ff0454687af35b9cb16ba970bf28dc61
drm/i915: Remove request retirement before each batch

This reimplements the denial-of-service protection against igt from
commit 227f782e4667 ("drm/i915: Retire requests before creating a new
one") and transfers the stall from before each batch into get_pages().
The issue is that the stall is increasing latency between batches which
is detrimental in some cases (especially coupled with execlists) to
keeping the GPU well fed. Also we have made the observation that retiring
requests can of itself free objects (and requests) and therefore makes
a good first step when shrinking.

v2: Recycle objects prior to i915_gem_object_get_pages()
v3: Remove the reference to the ring from i915_gem_requests_ring() as it
operates on an intel_engine_cs.
v4: Since commit 9b5f4e5ed6fd ("drm/i915: Retire oldest completed request
before allocating next") we no longer need the safeguard to retire
requests before get_pages(). We no longer see the huge latencies when
hitting the shrinker between allocations.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1470324762-2545-4-git-send-email-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/i915_drv.h
drivers/gpu/drm/i915/i915_gem_execbuffer.c
drivers/gpu/drm/i915/i915_gem_request.c