All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Subject: [PATCH v2 18/21] drm/i915: Track the previous pinned context inside the request
Date: Sun, 24 Apr 2016 08:31:38 +0100	[thread overview]
Message-ID: <1461483101-10618-18-git-send-email-chris@chris-wilson.co.uk> (raw)
In-Reply-To: <1461483101-10618-1-git-send-email-chris@chris-wilson.co.uk>

As the contexts are accessed by the hardware until the switch is completed
to a new context, the hardware may still be writing to the context object
after the breadcrumb is visible. We must not unpin/unbind/prune that
object whilst still active and so we keep the previous context pinned until
the following request. We can generalise the tracking we already do via
the engine->last_context and move it to the request so that it works
equally for execlists and GuC.

v2: Drop the execlists double pin as that exposes a race inside the lrc
irq handler as it tries to access the context after it may be retired.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
 drivers/gpu/drm/i915/i915_drv.h  | 11 +++++++++++
 drivers/gpu/drm/i915/i915_gem.c  |  8 ++++----
 drivers/gpu/drm/i915/intel_lrc.c | 14 ++++++++------
 3 files changed, 23 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index a26a026ef8e2..515b8badce61 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2302,6 +2302,17 @@ struct drm_i915_gem_request {
 	struct intel_context *ctx;
 	struct intel_ringbuffer *ringbuf;
 
+	/**
+	 * Context related to the previous request.
+	 * As the contexts are accessed by the hardware until the switch is
+	 * completed to a new context, the hardware may still be writing
+	 * to the context object after the breadcrumb is visible. We must
+	 * not unpin/unbind/prune that object whilst still active and so
+	 * we keep the previous context pinned until the following (this)
+	 * request is retired.
+	 */
+	struct intel_context *previous_context;
+
 	/** Batch buffer related to this request if any (used for
 	    error state dump only) */
 	struct drm_i915_gem_object *batch_obj;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 7793f0086523..d16680b77e13 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1413,13 +1413,13 @@ static void i915_gem_request_retire(struct drm_i915_gem_request *request)
 	list_del_init(&request->list);
 	i915_gem_request_remove_from_client(request);
 
-	if (request->ctx) {
+	if (request->previous_context) {
 		if (i915.enable_execlists)
-			intel_lr_context_unpin(request->ctx, request->engine);
-
-		i915_gem_context_unreference(request->ctx);
+			intel_lr_context_unpin(request->previous_context,
+					       request->engine);
 	}
 
+	i915_gem_context_unreference(request->ctx);
 	i915_gem_request_unreference(request);
 }
 
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 23bc5fca6595..3293dd29e274 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -798,12 +798,14 @@ intel_logical_ring_advance_and_submit(struct drm_i915_gem_request *request)
 	if (intel_engine_stopped(engine))
 		return 0;
 
-	if (engine->last_context != request->ctx) {
-		if (engine->last_context)
-			intel_lr_context_unpin(engine->last_context, engine);
-		intel_lr_context_pin(request->ctx, engine);
-		engine->last_context = request->ctx;
-	}
+	/* We keep the previous context alive until we retire the following
+	 * request. This ensures that any the context object is still pinned
+	 * for any residual writes the HW makes into it on the context switch
+	 * into the next object following the breadcrumb. Otherwise, we may
+	 * retire the context too early.
+	 */
+	request->previous_context = engine->last_context;
+	engine->last_context = request->ctx;
 
 	if (dev_priv->guc.execbuf_client)
 		i915_guc_submit(dev_priv->guc.execbuf_client, request);
-- 
2.8.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  parent reply	other threads:[~2016-04-24  7:32 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-04-24  7:31 [PATCH v2 01/21] drm/i915/fbdev: Call intel_unpin_fb_obj() on release Chris Wilson
2016-04-24  7:31 ` [PATCH v2 02/21] drm/i915/overlay: Replace i915_gem_obj_ggtt_offset() with the known flip_addr Chris Wilson
2016-04-24  7:31 ` [PATCH v2 03/21] io-mapping: Specify mapping size for io_mapping_map_wc() Chris Wilson
2016-04-24  7:31   ` Chris Wilson
2016-04-24  7:31   ` Chris Wilson
2016-04-24  7:31 ` [PATCH v2 04/21] drm/i915: Introduce i915_vm_to_ggtt() Chris Wilson
2016-04-24  7:31 ` [PATCH v2 05/21] drm/i915: Move ioremap_wc tracking onto VMA Chris Wilson
2016-04-24  7:31 ` [PATCH v2 06/21] drm/i915: Use i915_vma_pin_iomap on the ringbuffer object Chris Wilson
2016-04-24  7:31 ` [PATCH v2 07/21] drm/i915: Mark the current context as lost on suspend Chris Wilson
2016-04-24  7:31 ` [PATCH v2 08/21] drm/i915: L3 cache remapping is part of context switching Chris Wilson
2016-04-24  7:31 ` [PATCH v2 09/21] drm/i915: Consolidate L3 remapping LRI Chris Wilson
2016-04-24  7:31 ` [PATCH v2 10/21] drm/i915: Remove early l3-remap Chris Wilson
2016-04-24  7:31 ` [PATCH v2 11/21] drm/i915: Rearrange switch_context to load the aliasing ppgtt on first use Chris Wilson
2016-04-24  7:31 ` [PATCH v2 12/21] drm/i915: Assign every HW context a unique ID Chris Wilson
2016-04-24  7:31 ` [PATCH v2 13/21] drm/i915: Replace the pinned context address with its " Chris Wilson
2016-04-24  7:31 ` [PATCH v2 14/21] drm/i915: Refactor execlists default context pinning Chris Wilson
2016-04-24  7:31 ` [PATCH v2 15/21] drm/i915: Move context initialisation to first-use Chris Wilson
2016-04-24  7:31 ` [PATCH v2 16/21] drm/i915: Move the magical deferred context allocation into the request Chris Wilson
2016-04-24  7:31 ` [PATCH v2 17/21] drm/i915: Move releasing of the GEM request from free to retire/cancel Chris Wilson
2016-04-24  7:31 ` Chris Wilson [this message]
2016-04-24  7:31 ` [PATCH v2 19/21] drm/i915: Store LRC hardware id in the request Chris Wilson
2016-04-24  7:31 ` [PATCH v2 20/21] drm/i915: Stop tracking execlists retired requests Chris Wilson
2016-04-24  7:31 ` [PATCH v2 21/21] drm/i915: Unify GPU resets upon shutdown Chris Wilson
2016-04-24 11:51 ` ✗ Fi.CI.BAT: failure for series starting with [v2,01/21] drm/i915/fbdev: Call intel_unpin_fb_obj() on release Patchwork
2016-04-24 12:40   ` 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=1461483101-10618-18-git-send-email-chris@chris-wilson.co.uk \
    --to=chris@chris-wilson.co.uk \
    --cc=intel-gfx@lists.freedesktop.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.