Intel-GFX Archive on lore.kernel.org
 help / color / Atom feed
From: Daniel Vetter <daniel.vetter@ffwll.ch>
To: Intel Graphics Development <intel-gfx@lists.freedesktop.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Subject: [PATCH 5/6] drm/i915: clear up wedged transitions
Date: Wed, 14 Nov 2012 17:14:07 +0100
Message-ID: <1352909648-21514-6-git-send-email-daniel.vetter@ffwll.ch> (raw)
In-Reply-To: <1352909648-21514-1-git-send-email-daniel.vetter@ffwll.ch>

We have two important transitions of the wedged state in the current
code:

- 0 -> 1: This means a hang has been detected, and signals to everyone
  that they please get of any locks, so that the reset work item can
  do its job.

- 1 -> 0: The reset handler has completed.

Now the last transition mixes up two states: "Reset completed and
successful" and "Reset failed". To distinguish these two we do some
tricks with the reset completion, but I simply could not convince
myself that this doesn't race under odd circumstances.

Hence split this up, and add a new terminal state indicating that the
hw is gone for good.

Also add explicit #defines for both states, update comments.

v2: Split out the reset handling bugfix for the throttle ioctl.

v3: s/tmp/wedged/ sugested by Chris Wilson. Also fixup up a rebase
error which prevent this patch from actually compiling.

Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_drv.h | 17 ++++++++++++++++-
 drivers/gpu/drm/i915/i915_gem.c | 37 +++++++++++--------------------------
 drivers/gpu/drm/i915/i915_irq.c | 23 +++++++++++++++--------
 3 files changed, 42 insertions(+), 35 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 6958bb0..423541b 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -723,11 +723,26 @@ struct i915_gpu_error {
 	/* Protected by the above dev->gpu_error.lock. */
 	struct drm_i915_error_state *first_error;
 	struct work_struct work;
-	struct completion completion;
 
 	unsigned long last_reset;
 
+	/**
+	 * State variable controlling the reset flow
+	 *
+	 * 1 means a reset is in progress. This state will (presuming we don't
+	 * have any bugs) decay into either 0 (successful reset) or 2 (hw
+	 * terminally sour). All waiters on the reset_queue will be woken when
+	 * that happens.
+	 */
 	atomic_t wedged;
+#define I915_RESET_IN_PROGRESS	1
+#define I915_WEDGED		2
+
+	/**
+	 * Waitqueue to signal when the reset has completed. Used by clients
+	 * that wait for dev_priv->mm.wedged to settle.
+	 */
+	wait_queue_head_t reset_queue;
 
 	/* For gpu hang simulation. */
 	unsigned int stop_rings;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 55cdad9..6112a4f 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -89,36 +89,28 @@ static void i915_gem_info_remove_obj(struct drm_i915_private *dev_priv,
 static int
 i915_gem_wait_for_error(struct i915_gpu_error *error)
 {
-	struct completion *x = &error->completion;
-	unsigned long flags;
 	int ret;
 
 	if (!atomic_read(&error->wedged))
 		return 0;
 
+#define EXIT_COND (atomic_read(&error->wedged) != I915_RESET_IN_PROGRESS)
 	/*
 	 * Only wait 10 seconds for the gpu reset to complete to avoid hanging
 	 * userspace. If it takes that long something really bad is going on and
 	 * we should simply try to bail out and fail as gracefully as possible.
 	 */
-	ret = wait_for_completion_interruptible_timeout(x, 10*HZ);
+	ret = wait_event_interruptible_timeout(error->reset_queue,
+					       EXIT_COND,
+					       10*HZ);
 	if (ret == 0) {
 		DRM_ERROR("Timed out waiting for the gpu reset to complete\n");
 		return -EIO;
 	} else if (ret < 0) {
 		return ret;
 	}
+#undef EXIT_COND
 
-	if (atomic_read(&error->wedged)) {
-		/* GPU is hung, bump the completion count to account for
-		 * the token we just consumed so that we never hit zero and
-		 * end up waiting upon a subsequent completion event that
-		 * will never happen.
-		 */
-		spin_lock_irqsave(&x->wait.lock, flags);
-		x->done++;
-		spin_unlock_irqrestore(&x->wait.lock, flags);
-	}
 	return 0;
 }
 
@@ -943,23 +935,16 @@ int
 i915_gem_check_wedge(struct i915_gpu_error *error,
 		     bool interruptible)
 {
-	if (atomic_read(&error->wedged)) {
-		struct completion *x = &error->completion;
-		bool recovery_complete;
-		unsigned long flags;
-
-		/* Give the error handler a chance to run. */
-		spin_lock_irqsave(&x->wait.lock, flags);
-		recovery_complete = x->done > 0;
-		spin_unlock_irqrestore(&x->wait.lock, flags);
+	unsigned wedged = atomic_read(&error->wedged);
 
+	if (wedged) {
 		/* Non-interruptible callers can't handle -EAGAIN, hence return
 		 * -EIO unconditionally for these. */
 		if (!interruptible)
 			return -EIO;
 
-		/* Recovery complete, but still wedged means reset failure. */
-		if (recovery_complete)
+		/* Recovery complete, but the reset failed ... */
+		if (wedged == I915_WEDGED)
 			return -EIO;
 
 		return -EAGAIN;
@@ -1385,7 +1370,7 @@ out:
 		/* If this -EIO is due to a gpu hang, give the reset code a
 		 * chance to clean up the mess. Otherwise return the proper
 		 * SIGBUS. */
-		if (!atomic_read(&dev_priv->gpu_error.wedged))
+		if (atomic_read(&dev_priv->gpu_error.wedged) == I915_WEDGED)
 			return VM_FAULT_SIGBUS;
 	case -EAGAIN:
 		/* Give the error handler a chance to run and move the
@@ -4112,7 +4097,7 @@ i915_gem_load(struct drm_device *dev)
 		INIT_LIST_HEAD(&dev_priv->fence_regs[i].lru_list);
 	INIT_DELAYED_WORK(&dev_priv->mm.retire_work,
 			  i915_gem_retire_work_handler);
-	init_completion(&dev_priv->gpu_error.completion);
+	init_waitqueue_head(&dev_priv->gpu_error.reset_queue);
 
 	/* On GEN3 we really need to make sure the ARB C3 LP bit is set */
 	if (IS_GEN3(dev)) {
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 9d8921a..73d46f4 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -838,8 +838,10 @@ done:
  */
 static void i915_error_work_func(struct work_struct *work)
 {
-	drm_i915_private_t *dev_priv = container_of(work, drm_i915_private_t,
-						    gpu_error.work);
+	struct i915_gpu_error *error = container_of(work, struct i915_gpu_error,
+						    work);
+	drm_i915_private_t *dev_priv = container_of(error, drm_i915_private_t,
+						    gpu_error);
 	struct drm_device *dev = dev_priv->dev;
 	char *error_event[] = { "ERROR=1", NULL };
 	char *reset_event[] = { "RESET=1", NULL };
@@ -847,14 +849,19 @@ static void i915_error_work_func(struct work_struct *work)
 
 	kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, error_event);
 
-	if (atomic_read(&dev_priv->gpu_error.wedged)) {
+	if (atomic_read(&error->wedged) == I915_RESET_IN_PROGRESS) {
 		DRM_DEBUG_DRIVER("resetting chip\n");
 		kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, reset_event);
+
 		if (!i915_reset(dev)) {
-			atomic_set(&dev_priv->gpu_error.wedged, 0);
+			atomic_set(&error->wedged, 0);
 			kobject_uevent_env(&dev->primary->kdev.kobj, KOBJ_CHANGE, reset_done_event);
+		} else {
+			atomic_set(&error->wedged,
+				   I915_WEDGED);
 		}
-		complete_all(&dev_priv->gpu_error.completion);
+
+		wake_up_all(&dev_priv->gpu_error.reset_queue);
 	}
 }
 
@@ -1434,11 +1441,11 @@ void i915_handle_error(struct drm_device *dev, bool wedged)
 	i915_report_and_clear_eir(dev);
 
 	if (wedged) {
-		INIT_COMPLETION(dev_priv->gpu_error.completion);
-		atomic_set(&dev_priv->gpu_error.wedged, 1);
+		atomic_set(&dev_priv->gpu_error.wedged, I915_RESET_IN_PROGRESS);
 
 		/*
-		 * Wakeup waiting processes so they don't hang
+		 * Wakeup waiting processes so that the reset work item
+		 * doesn't deadlock trying to grab various locks.
 		 */
 		for_each_ring(ring, dev_priv, i)
 			wake_up_all(&ring->irq_queue);
-- 
1.7.11.4

  parent reply index

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-11-14 16:14 [PATCH 0/6] robustify reset transitions Daniel Vetter
2012-11-14 16:14 ` [PATCH 1/6] drm/i915: move dev_priv->mm out of line Daniel Vetter
2012-12-04 16:31   ` Damien Lespiau
2012-11-14 16:14 ` [PATCH 2/6] drm/i915: extract hangcheck/reset/error_state state into substruct Daniel Vetter
2012-12-04 17:20   ` Damien Lespiau
2012-11-14 16:14 ` [PATCH 3/6] drm/i915: move wedged to the other gpu error handling stuff Daniel Vetter
2012-12-04 17:24   ` Damien Lespiau
2012-11-14 16:14 ` [PATCH 4/6] drm/i915: fix reset handling in the throttle ioctl Daniel Vetter
2012-12-05 14:08   ` Damien Lespiau
2012-11-14 16:14 ` Daniel Vetter [this message]
2012-11-14 16:14 ` [PATCH 6/6] drm/i915: create a race-free reset detection Daniel Vetter
2012-11-15 16:17 ` [PATCH 1/2] drm/i915: clear up wedged transitions Daniel Vetter
2012-11-15 16:17   ` [PATCH 2/2] drm/i915: create a race-free reset detection Daniel Vetter
2012-12-05 16:35     ` Damien Lespiau
2012-12-06  8:01       ` [PATCH] " Daniel Vetter
2012-12-06 15:23       ` [PATCH] drm/i915: clarify concurrent hang detect/gpu reset consistency Daniel Vetter
2013-01-18 20:48       ` [PATCH 2/2] drm/i915: create a race-free reset detection Daniel Vetter
2013-01-21 12:06         ` Damien Lespiau
2013-01-21 19:15           ` Daniel Vetter
2012-12-05 14:54   ` [PATCH 1/2] drm/i915: clear up wedged transitions Damien Lespiau
2012-12-05 16:38   ` Damien Lespiau
2012-12-05 17:14     ` Daniel Vetter
2014-09-03 20:26   ` Chris Wilson
2014-09-04  6:03     ` Daniel Vetter
2014-09-04  6:11       ` 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=1352909648-21514-6-git-send-email-daniel.vetter@ffwll.ch \
    --to=daniel.vetter@ffwll.ch \
    --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

Intel-GFX Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/intel-gfx/0 intel-gfx/git/0.git
	git clone --mirror https://lore.kernel.org/intel-gfx/1 intel-gfx/git/1.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 intel-gfx intel-gfx/ https://lore.kernel.org/intel-gfx \
		intel-gfx@lists.freedesktop.org
	public-inbox-index intel-gfx

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.freedesktop.lists.intel-gfx


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git