drm/i915: Force CPU relocations if not GTT mapped
authorChris Wilson <chris@chris-wilson.co.uk>
Mon, 11 Aug 2014 10:00:12 +0000 (12:00 +0200)
committerDaniel Vetter <daniel.vetter@ffwll.ch>
Mon, 11 Aug 2014 10:01:29 +0000 (12:01 +0200)
commite6a844687cf929ec053c7578d5ecc794a8a6c5cf
treea8a9b1534bd862d48eba0f31282cac1cec6906f3
parentdc8cd1e790081a31ba4d86c3c0812c348eeec7fc
drm/i915: Force CPU relocations if not GTT mapped

Move the decision on whether we need to have a mappable object during
execbuffer to the fore and then reuse that decision by propagating the
flag through to reservation. As a corollary, before doing the actual
relocation through the GTT, we can make sure that we do have a GTT
mapping through which to operate.

Note that the key to make this work is to ditch the
obj->map_and_fenceable unbind optimization - with full ppgtt it
doesn't make a lot of sense any more anyway.

v2: Revamp and resend to ease future patches.
v3: Refresh patch rationale

References: https://bugs.freedesktop.org/show_bug.cgi?id=81094
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
[danvet: Explain why obj->map_and_fenceable is key and split out the
secure batch fix.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/i915_gem_execbuffer.c