drm: Restore double clflush on the last partial cacheline
authorChris Wilson <chris@chris-wilson.co.uk>
Thu, 7 Jul 2016 08:41:12 +0000 (09:41 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 20 Aug 2016 16:09:23 +0000 (18:09 +0200)
commit 396f5d62d1a5fd99421855a08ffdef8edb43c76e upstream.

This effectively reverts

commit afcd950cafea6e27b739fe7772cbbeed37d05b8b
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jun 10 15:58:01 2015 +0100

    drm: Avoid the double clflush on the last cache line in drm_clflush_virt_range()

as we have observed issues with serialisation of the clflush operations
on Baytrail+ Atoms with partial updates. Applying the double flush on the
last cacheline forces that clflush to be ordered with respect to the
previous clflush, and the mfence then protects against prefetches crossing
the clflush boundary.

The same issue can be demonstrated in userspace with igt/gem_exec_flush.

Fixes: afcd950cafea6 (drm: Avoid the double clflush on the last cache...)
Testcase: igt/gem_concurrent_blit
Testcase: igt/gem_partial_pread_pwrite
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92845
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: dri-devel@lists.freedesktop.org
Cc: Akash Goel <akash.goel@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1467880930-23082-6-git-send-email-chris@chris-wilson.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/gpu/drm/drm_cache.c

index 6743ff7dccfa30b2997d2529d97747d5261d5ad9..7f4a6c55031938c93b00bec6a954050cff9f347d 100644 (file)
@@ -136,6 +136,7 @@ drm_clflush_virt_range(void *addr, unsigned long length)
                mb();
                for (; addr < end; addr += size)
                        clflushopt(addr);
+               clflushopt(end - 1); /* force serialisation */
                mb();
                return;
        }