public inbox for intel-gfx@lists.freedesktop.org
 help / color / mirror / Atom feed
From: Daniel Vetter <daniel@ffwll.ch>
To: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>,
	dri-devel@lists.freedesktop.org,
	Jason Ekstrand <jason.ekstrand@intel.com>,
	Akash Goel <akash.goel@intel.com>,
	stable@vger.kernel.org, intel-gfx@lists.freedesktop.org
Subject: Re: [PATCH 06/64] drm: Restore double clflush on the last partial cacheline
Date: Tue, 12 Jul 2016 15:57:44 +0200	[thread overview]
Message-ID: <20160712135744.GI23520@phenom.ffwll.local> (raw)
In-Reply-To: <1467880930-23082-6-git-send-email-chris@chris-wilson.co.uk>

On Thu, Jul 07, 2016 at 09:41:12AM +0100, Chris Wilson wrote:
> 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>
> Cc: stable@vger.kernel.org
> Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

It's duct-tape, but oh well. Applied to drm-misc.
-Daniel

> ---
>  drivers/gpu/drm/drm_cache.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/gpu/drm/drm_cache.c b/drivers/gpu/drm/drm_cache.c
> index 059f7c39c582..a7916e5f8864 100644
> --- a/drivers/gpu/drm/drm_cache.c
> +++ b/drivers/gpu/drm/drm_cache.c
> @@ -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;
>  	}
> -- 
> 2.8.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  reply	other threads:[~2016-07-12 13:57 UTC|newest]

Thread overview: 91+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-07-07  8:41 [PATCH 01/64] drm/i915/breadcrumbs: Queue hangcheck before sleeping Chris Wilson
2016-07-07  8:41 ` [PATCH 02/64] drm/i915: Preserve current RPS frequency across init Chris Wilson
2016-07-07  8:41 ` [PATCH 03/64] drm/i915: Remove superfluous powersave work flushing Chris Wilson
2016-07-07  8:41 ` [PATCH 04/64] drm/i915: Defer enabling rc6 til after we submit the first batch/context Chris Wilson
2016-07-07  8:41 ` [PATCH 05/64] drm/i915: Remove temporary RPM wakeref assert disables Chris Wilson
2016-07-07  8:41 ` [PATCH 06/64] drm: Restore double clflush on the last partial cacheline Chris Wilson
2016-07-12 13:57   ` Daniel Vetter [this message]
2016-07-13  8:07     ` Mika Kuoppala
2016-07-07  8:41 ` [PATCH 07/64] drm/i915: Move GEM request routines to i915_gem_request.c Chris Wilson
2016-07-07  8:41 ` [PATCH 08/64] drm/i915: Retire oldest completed request before allocating next Chris Wilson
2016-07-07  9:41   ` Tvrtko Ursulin
2016-07-07  9:45     ` Chris Wilson
2016-07-07 10:03       ` Tvrtko Ursulin
2016-07-07 10:10         ` Chris Wilson
2016-07-08  9:21   ` Mika Kuoppala
2016-07-08  9:28     ` Chris Wilson
2016-07-07  8:41 ` [PATCH 09/64] drm/i915: Mark all current requests as complete before resetting them Chris Wilson
2016-07-07 16:09   ` Mika Kuoppala
2016-07-07  8:41 ` [PATCH 10/64] drm/i915: Derive GEM requests from dma-fence Chris Wilson
2016-07-07  8:41 ` [PATCH 11/64] drm/i915: Disable waitboosting for fence_wait() Chris Wilson
2016-07-07  8:41 ` [PATCH 12/64] drm/i915: Disable waitboosting for mmioflips/semaphores Chris Wilson
2016-07-07  8:41 ` [PATCH 13/64] drm/i915: Export our request as a dma-buf fence on the reservation object Chris Wilson
2016-07-07  8:41 ` [PATCH 14/64] drm/i915: Wait on external rendering for GEM objects Chris Wilson
2016-07-07  8:41 ` [PATCH 15/64] drm/i915: Mark imported dma-buf objects as being coherent Chris Wilson
2016-07-07  8:41 ` [PATCH 16/64] drm/i915: Rename request reference/unreference to get/put Chris Wilson
2016-07-07  8:41 ` [PATCH 17/64] drm/i915: Rename i915_gem_context_reference/unreference() Chris Wilson
2016-07-07  8:41 ` [PATCH 18/64] drm/i915: Wrap drm_gem_object_lookup in i915_gem_object_lookup Chris Wilson
2016-07-07  8:41 ` [PATCH 19/64] drm/i915: Wrap drm_gem_object_reference in i915_gem_object_get Chris Wilson
2016-07-07  8:41 ` [PATCH 20/64] drm/i915: Rename drm_gem_object_unreference in preparation for lockless free Chris Wilson
2016-07-07  8:41 ` [PATCH 21/64] drm/i915: Rename drm_gem_object_unreference_unlocked " Chris Wilson
2016-07-07  8:41 ` [PATCH 22/64] drm/i915: Treat ringbuffer writes as write to normal memory Chris Wilson
2016-07-07  8:41 ` [PATCH 23/64] drm/i915: Rename ring->virtual_start as ring->vaddr Chris Wilson
2016-07-07  8:41 ` [PATCH 24/64] drm/i915: Convert i915_semaphores_is_enabled over to early sanitize Chris Wilson
2016-07-07  8:41 ` [PATCH 25/64] drm/i915: Unify intel_logical_ring_emit and intel_ring_emit Chris Wilson
2016-07-07  8:41 ` [PATCH 26/64] drm/i915: Rename request->ringbuf to request->ring Chris Wilson
2016-07-07  8:41 ` [PATCH 27/64] drm/i915: Rename backpointer from intel_ringbuffer to intel_engine_cs Chris Wilson
2016-07-07  8:41 ` [PATCH 28/64] drm/i915: Rename intel_context[engine].ringbuf Chris Wilson
2016-07-07  8:41 ` [PATCH 29/64] drm/i915: Rename struct intel_ringbuffer to struct intel_ring Chris Wilson
2016-07-07  8:41 ` [PATCH 30/64] drm/i915: Rename residual ringbuf parameters Chris Wilson
2016-07-07  8:41 ` [PATCH 31/64] drm/i915: Rename intel_pin_and_map_ring() Chris Wilson
2016-07-07  8:41 ` [PATCH 32/64] drm/i915: Remove obsolete engine->gpu_caches_dirty Chris Wilson
2016-07-07  8:41 ` [PATCH 33/64] drm/i915: Simplify request_alloc by returning the allocated request Chris Wilson
2016-07-07  8:41 ` [PATCH 34/64] drm/i915: Unify legacy/execlists emission of MI_BATCHBUFFER_START Chris Wilson
2016-07-07  8:41 ` [PATCH 35/64] drm/i915: Convert engine->write_tail to operate on a request Chris Wilson
2016-07-07  8:41 ` [PATCH 36/64] drm/i915: Unify request submission Chris Wilson
2016-07-07  8:41 ` [PATCH 37/64] drm/i915: Stop passing caller's num_dwords to engine->semaphore.signal() Chris Wilson
2016-07-07  8:41 ` [PATCH 38/64] drm/i915: Reuse legacy breadcrumbs + tail emission Chris Wilson
2016-07-07  8:41 ` [PATCH 39/64] drm/i915/ringbuffer: Specialise SNB+ request emission for semaphores Chris Wilson
2016-07-07  8:41 ` [PATCH 40/64] drm/i915: Remove duplicate golden render state init from execlists Chris Wilson
2016-07-07  8:41 ` [PATCH 41/64] drm/i915: Unify legacy/execlists submit_execbuf callbacks Chris Wilson
2016-07-07  8:41 ` [PATCH 42/64] drm/i915: Simplify calling engine->sync_to Chris Wilson
2016-07-07  8:41 ` [PATCH 43/64] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Chris Wilson
2016-07-07  8:41 ` [PATCH 44/64] drm/i915: Split early global GTT initialisation Chris Wilson
2016-07-07  8:41 ` [PATCH 45/64] drm/i915: Store owning file on the i915_address_space Chris Wilson
2016-07-07  8:41 ` [PATCH 46/64] drm/i915: Count how many VMA are bound for an object Chris Wilson
2016-07-12 14:30   ` Tvrtko Ursulin
2016-07-12 14:38     ` Chris Wilson
2016-07-12 15:12       ` Tvrtko Ursulin
2016-07-12 16:08         ` Chris Wilson
2016-07-07  8:41 ` [PATCH 47/64] drm/i915: Be more careful when unbinding vma Chris Wilson
2016-07-12 15:04   ` Tvrtko Ursulin
2016-07-12 16:42     ` Chris Wilson
2016-07-13  8:53       ` Tvrtko Ursulin
2016-07-07  8:41 ` [PATCH 48/64] drm/i915: Kill drop_pages() Chris Wilson
2016-07-12 15:14   ` Tvrtko Ursulin
2016-07-07  8:41 ` [PATCH 49/64] drm/i915: Introduce i915_gem_active for request tracking Chris Wilson
2016-07-12 16:05   ` Tvrtko Ursulin
2016-07-12 16:30     ` Chris Wilson
2016-07-13  8:54       ` Tvrtko Ursulin
2016-07-07  8:41 ` [PATCH 50/64] drm/i915: Prepare i915_gem_active for annotations Chris Wilson
2016-07-13 15:40   ` Tvrtko Ursulin
2016-07-13 15:58     ` Chris Wilson
2016-07-14  9:32       ` Tvrtko Ursulin
2016-07-14 10:04         ` Chris Wilson
2016-07-07  8:41 ` [PATCH 51/64] drm/i915: Mark up i915_gem_active for locking annotation Chris Wilson
2016-07-07  8:41 ` [PATCH 52/64] drm/i915: Refactor blocking waits Chris Wilson
2016-07-07  8:41 ` [PATCH 53/64] drm/i915: Rename request->list to link for consistency Chris Wilson
2016-07-07  8:42 ` [PATCH 54/64] drm/i915: Remove obsolete i915_gem_object_flush_active() Chris Wilson
2016-07-07  8:42 ` [PATCH 55/64] drm/i915: Refactor activity tracking for requests Chris Wilson
2016-07-07  8:42 ` [PATCH 56/64] drm/i915: Convert intel_overlay to request tracking Chris Wilson
2016-07-07  8:42 ` [PATCH 57/64] drm/i915: Move the special case wait-request handling to its one caller Chris Wilson
2016-07-07  8:42 ` [PATCH 58/64] drm/i915: Double check activity before relocations Chris Wilson
2016-07-07  8:42 ` [PATCH 59/64] drm/i915: Move request list retirement to i915_gem_request.c Chris Wilson
2016-07-07  8:42 ` [PATCH 60/64] drm/i915: i915_vma_move_to_active prep patch Chris Wilson
2016-07-07  8:42 ` [PATCH 61/64] drm/i915: Track active vma requests Chris Wilson
2016-07-07  8:42 ` [PATCH 62/64] drm/i915: Release vma when the handle is closed Chris Wilson
2016-07-07  8:42 ` [PATCH 63/64] drm/i915: Mark the context and address space as closed Chris Wilson
2016-07-07  8:42 ` [PATCH 64/64] Revert "drm/i915: Clean up associated VMAs on context destruction" Chris Wilson
2016-07-07  9:12 ` ✗ Ro.CI.BAT: warning for series starting with [01/64] drm/i915/breadcrumbs: Queue hangcheck before sleeping Patchwork
2016-07-07  9:28   ` Chris Wilson
2016-07-08  8:44     ` Chris Wilson

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160712135744.GI23520@phenom.ffwll.local \
    --to=daniel@ffwll.ch \
    --cc=akash.goel@intel.com \
    --cc=chris@chris-wilson.co.uk \
    --cc=daniel.vetter@ffwll.ch \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=jason.ekstrand@intel.com \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox