Nouveau - May 2020 - [RFC v4 02/12] kthread: Add kthread_(un)block_work_queuing() and kthread_work_queuable()

Currently, it's only possible to flush on a kthread_work in contexts
where it's possible to block. This can be kind of painful though when
trying to implement new types of delayed work which use kthread_work,
since it means we'd need to drop any spinlocks for new delayed work
implementations before we can actually call kthread_flush_work().

In the time between dropping locks and calling kthread_flush_work(), the
work might have already executed once and have gotten re-queued by the
time we're ready. This would mean that once the user finally executes
kthread_flush_work(), we'd accidentally wait for someone else's queued
work instead of our own.

For DRM vblank works it's preferable that we just return immediately
during such races, instead of blocking on the re-queue. Additionally, we
also want to be able to use kthread_flush_work structs in our own
contexts so that we can block until a vblank work's target vblank has
passed, _and_ said work has executed once since. And finally, we also
want to be able to finish flushing on a work early if it's been
cancelled at any point (e.g. before or after it's actually been queued
on the kthread_worker).

So, let's make all of these things possible by exposing struct
kthread_flush_work, and then splitting kthread_flush_work() into two
functions: kthread_queue_flush_work(); which handles possibly queuing up
the kthread_flush_work on the work's respective kthread_worker, and
kthread_flush_work(); which performs a simple synchronous flush just
like before. We also add a DEFINE_KTHREAD_FLUSH_WORK() macro, which
simply initializes a kthread_flush_work struct inline (I can't see
anyone needing to use a kthread_flush_work that gets used outside of the
scope of a single function, and that seems like it would be a bit
overkill anyway).

Cc: Daniel Vetter <daniel at ffwll.ch>
Cc: Tejun Heo <tj at kernel.org>
Cc: Ville Syrj?l? <ville.syrjala at linux.intel.com>
Cc: dri-devel at lists.freedesktop.org
Cc: nouveau at lists.freedesktop.org
Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 include/linux/kthread.h | 16 ++++++++
 kernel/kthread.c        | 87 ++++++++++++++++++++++++++---------------
 2 files changed, 71 insertions(+), 32 deletions(-)

diff --git a/include/linux/kthread.h b/include/linux/kthread.h
index 8bbcaad7ef0f..0006540ce7f9 100644
--- a/include/linux/kthread.h
+++ b/include/linux/kthread.h
@@ -105,6 +105,11 @@ struct kthread_delayed_work {
 	struct timer_list timer;
 };
 
+struct kthread_flush_work {
+	struct kthread_work	work;
+	struct completion	done;
+};
+
 #define KTHREAD_WORKER_INIT(worker)	{				\
 	.lock = __RAW_SPIN_LOCK_UNLOCKED((worker).lock),		\
 	.work_list = LIST_HEAD_INIT((worker).work_list),		\
@@ -122,6 +127,11 @@ struct kthread_delayed_work {
 				     TIMER_IRQSAFE),			\
 	}
 
+#define KTHREAD_FLUSH_WORK_INIT(fwork) { \
+	KTHREAD_WORK_INIT((fwork).work, __kthread_flush_work_fn), \
+	COMPLETION_INITIALIZER_ONSTACK((fwork).done), \
+	}
+
 #define DEFINE_KTHREAD_WORKER(worker)					\
 	struct kthread_worker worker = KTHREAD_WORKER_INIT(worker)
 
@@ -132,6 +142,9 @@ struct kthread_delayed_work {
 	struct kthread_delayed_work dwork =				\
 		KTHREAD_DELAYED_WORK_INIT(dwork, fn)
 
+#define DEFINE_KTHREAD_FLUSH_WORK(fwork) \
+	struct kthread_flush_work fwork = KTHREAD_FLUSH_WORK_INIT(fwork);
+
 /*
  * kthread_worker.lock needs its own lockdep class key when defined on
  * stack with lockdep enabled.  Use the following macros in such cases.
@@ -190,6 +203,9 @@ bool kthread_mod_delayed_work(struct kthread_worker *worker,
 			      struct kthread_delayed_work *dwork,
 			      unsigned long delay);
 
+bool kthread_queue_flush_work(struct kthread_work *work,
+			      struct kthread_flush_work *fwork);
+void __kthread_flush_work_fn(struct kthread_work *work);
 void kthread_flush_work(struct kthread_work *work);
 void kthread_flush_worker(struct kthread_worker *worker);
 
diff --git a/kernel/kthread.c b/kernel/kthread.c
index bfbfa481be3a..c1f8ec9d5836 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -943,52 +943,78 @@ bool kthread_queue_delayed_work(struct kthread_worker
*worker,
 }
 EXPORT_SYMBOL_GPL(kthread_queue_delayed_work);
 
-struct kthread_flush_work {
-	struct kthread_work	work;
-	struct completion	done;
-};
-
-static void kthread_flush_work_fn(struct kthread_work *work)
-{
-	struct kthread_flush_work *fwork -		container_of(work, struct
kthread_flush_work, work);
-	complete(&fwork->done);
-}
-
 /**
- * kthread_flush_work - flush a kthread_work
- * @work: work to flush
+ * kthread_queue_flush_work - try queuing a kthread_flush_work after a
+ * queued kthread_work to synchronize with later.
+ * @work: The &kthread_work to synchronize with later
+ * @fwork: The &kthread_flush_work to queue
+ *
+ * When working with &kthread_work structs in contexts where sleeping
isn't
+ * possible it may be desirable to synchronize with a &kthread_work
that's
+ * currently queued, but only after we've entered a context where it's
safe to
+ * sleep again, and while making sure we don't block on any later re-queues
of
+ * the work.
+ *
+ * If @work is queued or executing when kthread_queue_flush_work() is called,
+ * @fwork will be scheduled for execution immediately after @work. The caller
+ * can then later synchronize on @fwork.done, which will complete once @work
+ * has executed once or been cancelled since kthread_queue_flush_work() was
+ * called.
+ *
+ * Returns: %true% if @fwork was queued,and the caller needs to call
+ * wait_for_completion() on @fwork.done to finish synchronizing, %false%
+ * otherwise.
  *
- * If @work is queued or executing, wait for it to finish execution.
  */
-void kthread_flush_work(struct kthread_work *work)
+bool kthread_queue_flush_work(struct kthread_work *work,
+			      struct kthread_flush_work *fwork)
 {
-	struct kthread_flush_work fwork = {
-		KTHREAD_WORK_INIT(fwork.work, kthread_flush_work_fn),
-		COMPLETION_INITIALIZER_ONSTACK(fwork.done),
-	};
 	struct kthread_worker *worker;
-	bool noop = false;
+	unsigned long flags;
+	bool queued = true;
 
 	worker = work->worker;
 	if (!worker)
-		return;
+		return false;
 
-	raw_spin_lock_irq(&worker->lock);
+	raw_spin_lock_irqsave(&worker->lock, flags);
 	/* Work must not be used with >1 worker, see kthread_queue_work(). */
 	WARN_ON_ONCE(work->worker != worker);
 
 	if (!list_empty(&work->node))
-		kthread_insert_work(worker, &fwork.work, work->node.next);
+		kthread_insert_work(worker, &fwork->work, work->node.next);
 	else if (worker->current_work == work)
-		kthread_insert_work(worker, &fwork.work,
+		kthread_insert_work(worker, &fwork->work,
 				    worker->work_list.next);
 	else
-		noop = true;
+		queued = false;
 
-	raw_spin_unlock_irq(&worker->lock);
+	raw_spin_unlock_irqrestore(&worker->lock, flags);
+	return queued;
+}
+EXPORT_SYMBOL_GPL(kthread_queue_flush_work);
+
+void __kthread_flush_work_fn(struct kthread_work *work)
+{
+	struct kthread_flush_work *fwork +		container_of(work, struct
kthread_flush_work, work);
+	complete(&fwork->done);
+}
+EXPORT_SYMBOL_GPL(__kthread_flush_work_fn);
+
+/**
+ * kthread_flush_work - flush a kthread_work
+ * @work: work to flush
+ *
+ * If @work is queued or executing, wait for it to finish execution.
+ */
+void kthread_flush_work(struct kthread_work *work)
+{
+	bool queued;
+	DEFINE_KTHREAD_FLUSH_WORK(fwork);
 
-	if (!noop)
+	queued = kthread_queue_flush_work(work, &fwork);
+	if (queued)
 		wait_for_completion(&fwork.done);
 }
 EXPORT_SYMBOL_GPL(kthread_flush_work);
@@ -1170,10 +1196,7 @@ EXPORT_SYMBOL_GPL(kthread_cancel_delayed_work_sync);
  */
 void kthread_flush_worker(struct kthread_worker *worker)
 {
-	struct kthread_flush_work fwork = {
-		KTHREAD_WORK_INIT(fwork.work, kthread_flush_work_fn),
-		COMPLETION_INITIALIZER_ONSTACK(fwork.done),
-	};
+	DEFINE_KTHREAD_FLUSH_WORK(fwork);
 
 	kthread_queue_work(worker, &fwork.work);
 	wait_for_completion(&fwork.done);
-- 
2.25.4

Add some simple wrappers around incrementing/decrementing
kthread_work.cancelling under lock, along with checking whether queuing
is currently allowed on a given kthread_work, which we'll use want to
implement work cancelling with DRM's vblank work helpers.

Cc: Daniel Vetter <daniel at ffwll.ch>
Cc: Tejun Heo <tj at kernel.org>
Cc: Ville Syrj?l? <ville.syrjala at linux.intel.com>
Cc: dri-devel at lists.freedesktop.org
Cc: nouveau at lists.freedesktop.org
Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 include/linux/kthread.h | 19 +++++++++++++++++
 kernel/kthread.c        | 46 +++++++++++++++++++++++++++++++++++++++++
 2 files changed, 65 insertions(+)

diff --git a/include/linux/kthread.h b/include/linux/kthread.h
index 0006540ce7f9..c6fee200fced 100644
--- a/include/linux/kthread.h
+++ b/include/linux/kthread.h
@@ -211,9 +211,28 @@ void kthread_flush_worker(struct kthread_worker *worker);
 
 bool kthread_cancel_work_sync(struct kthread_work *work);
 bool kthread_cancel_delayed_work_sync(struct kthread_delayed_work *work);
+void kthread_block_work_queuing(struct kthread_worker *worker,
+				struct kthread_work *work);
+void kthread_unblock_work_queuing(struct kthread_worker *worker,
+				  struct kthread_work *work);
 
 void kthread_destroy_worker(struct kthread_worker *worker);
 
+/**
+ * kthread_work_queuable - whether or not a kthread work can be queued
+ * @work: The kthread work to check
+ *
+ * Checks whether or not queuing @work is currently blocked from queuing,
+ * either by kthread_cancel_work_sync() and friends or
+ * kthread_block_work_queuing().
+ *
+ * Returns: whether or not the @work may be queued.
+ */
+static inline bool kthread_work_queuable(struct kthread_work *work)
+{
+	return READ_ONCE(work->canceling) == 0;
+}
+
 struct cgroup_subsys_state;
 
 #ifdef CONFIG_BLK_CGROUP
diff --git a/kernel/kthread.c b/kernel/kthread.c
index c1f8ec9d5836..f8a5c5a87cc6 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -1187,6 +1187,52 @@ bool kthread_cancel_delayed_work_sync(struct
kthread_delayed_work *dwork)
 }
 EXPORT_SYMBOL_GPL(kthread_cancel_delayed_work_sync);
 
+/**
+ * kthread_block_work_queuing - prevent a kthread_work from being queued
+ * without actually cancelling it
+ * @worker: kthread worker to use
+ * @work: work to block queuing on
+ *
+ * Prevents @work from being queued using kthread_queue_work() and friends,
+ * but doesn't attempt to cancel any previous queuing. The caller must
unblock
+ * queuing later by calling kthread_unblock_work_queuing(). This call can be
+ * called multiple times.
+ *
+ * See also: kthread_work_queuable()
+ */
+void kthread_block_work_queuing(struct kthread_worker *worker,
+				struct kthread_work *work)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&worker->lock, flags);
+	work->canceling++;
+	raw_spin_unlock_irqrestore(&worker->lock, flags);
+}
+EXPORT_SYMBOL_GPL(kthread_block_work_queuing);
+
+/**
+ * kthread_unblock_work_queuing - unblock queuing on a kthread_work
+ * @worker: kthread worker to use
+ * @work: work to unblock queuing on
+ *
+ * Removes a request to prevent @work from being queued with
+ * kthread_queue_work() and friends, so that it may potentially be queued
+ * again.
+ *
+ * See also: kthread_work_queuable()
+ */
+void kthread_unblock_work_queuing(struct kthread_worker *worker,
+				  struct kthread_work *work)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&worker->lock, flags);
+	WARN_ON_ONCE(--work->canceling < 0);
+	raw_spin_unlock_irqrestore(&worker->lock, flags);
+}
+EXPORT_SYMBOL_GPL(kthread_unblock_work_queuing);
+
 /**
  * kthread_flush_worker - flush all current works on a kthread_worker
  * @worker: worker to flush
-- 
2.25.4

Since we'll be allocating resources for kthread_create_worker() in the
next commit (which could fail and require us to clean up the mess),
let's simplify the cleanup process a bit by registering a
drm_vblank_init_release() action for each drm_vblank_crtc so they're
still cleaned up if we fail to initialize one of them.

Changes since v3:
* Use drmm_add_action_or_reset() - Daniel Vetter

Cc: Daniel Vetter <daniel at ffwll.ch>
Cc: Ville Syrj?l? <ville.syrjala at linux.intel.com>
Cc: dri-devel at lists.freedesktop.org
Cc: nouveau at lists.freedesktop.org
Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/drm_vblank.c | 23 ++++++++++-------------
 1 file changed, 10 insertions(+), 13 deletions(-)

diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
index 758bf74e1cab..a4a9013584e2 100644
--- a/drivers/gpu/drm/drm_vblank.c
+++ b/drivers/gpu/drm/drm_vblank.c
@@ -491,16 +491,12 @@ static void vblank_disable_fn(struct timer_list *t)
 
 static void drm_vblank_init_release(struct drm_device *dev, void *ptr)
 {
-	unsigned int pipe;
-
-	for (pipe = 0; pipe < dev->num_crtcs; pipe++) {
-		struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
+	struct drm_vblank_crtc *vblank = ptr;
 
-		WARN_ON(READ_ONCE(vblank->enabled) &&
-			drm_core_check_feature(dev, DRIVER_MODESET));
+	WARN_ON(READ_ONCE(vblank->enabled) &&
+		drm_core_check_feature(dev, DRIVER_MODESET));
 
-		del_timer_sync(&vblank->disable_timer);
-	}
+	del_timer_sync(&vblank->disable_timer);
 }
 
 /**
@@ -510,7 +506,7 @@ static void drm_vblank_init_release(struct drm_device *dev,
void *ptr)
  *
  * This function initializes vblank support for @num_crtcs display pipelines.
  * Cleanup is handled automatically through a cleanup function added with
- * drmm_add_action().
+ * drmm_add_action_or_reset().
  *
  * Returns:
  * Zero on success or a negative error code on failure.
@@ -529,10 +525,6 @@ int drm_vblank_init(struct drm_device *dev, unsigned int
num_crtcs)
 
 	dev->num_crtcs = num_crtcs;
 
-	ret = drmm_add_action(dev, drm_vblank_init_release, NULL);
-	if (ret)
-		return ret;
-
 	for (i = 0; i < num_crtcs; i++) {
 		struct drm_vblank_crtc *vblank = &dev->vblank[i];
 
@@ -541,6 +533,11 @@ int drm_vblank_init(struct drm_device *dev, unsigned int
num_crtcs)
 		init_waitqueue_head(&vblank->queue);
 		timer_setup(&vblank->disable_timer, vblank_disable_fn, 0);
 		seqlock_init(&vblank->seqlock);
+
+		ret = drmm_add_action_or_reset(dev, drm_vblank_init_release,
+					       vblank);
+		if (ret)
+			return ret;
 	}
 
 	DRM_INFO("Supports vblank timestamp caching Rev 2 (21.10.2013).\n");
-- 
2.25.4

Add some kind of vblank workers. The interface is similar to regular
delayed works, and is mostly based off kthread_work. It allows for
scheduling delayed works that execute once a particular vblank sequence
has passed. It also allows for accurate flushing of scheduled vblank
works - in that flushing waits for both the vblank sequence and job
execution to complete, or for the work to get cancelled - whichever
comes first.

Whatever hardware programming we do in the work must be fast (must at
least complete during the vblank or scanout period, sometimes during the
first few scanlines of the vblank). As such we use a high-priority
per-CRTC thread to accomplish this.

[based off patches from Ville Syrj?l? <ville.syrjala at linux.intel.com>,
change below to signoff later]

Changes since v3:
* Use our own spinlocks, don't integrate so tightly with kthread_works
Changes since v2:
* Use kthread_workers instead of reinventing the wheel.

Cc: Daniel Vetter <daniel at ffwll.ch>
Cc: Tejun Heo <tj at kernel.org>
Cc: Ville Syrj?l? <ville.syrjala at linux.intel.com>
Cc: dri-devel at lists.freedesktop.org
Cc: nouveau at lists.freedesktop.org
Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/drm_vblank.c | 259 +++++++++++++++++++++++++++++++++++
 include/drm/drm_vblank.h     |  35 +++++
 2 files changed, 294 insertions(+)

diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
index a4a9013584e2..5b10277553e4 100644
--- a/drivers/gpu/drm/drm_vblank.c
+++ b/drivers/gpu/drm/drm_vblank.c
@@ -25,7 +25,9 @@
  */
 
 #include <linux/export.h>
+#include <linux/kthread.h>
 #include <linux/moduleparam.h>
+#include <uapi/linux/sched/types.h>
 
 #include <drm/drm_crtc.h>
 #include <drm/drm_drv.h>
@@ -155,6 +157,7 @@
 static bool
 drm_get_last_vbltimestamp(struct drm_device *dev, unsigned int pipe,
 			  ktime_t *tvblank, bool in_vblank_irq);
+static int drm_vblank_get(struct drm_device *dev, unsigned int pipe);
 
 static unsigned int drm_timestamp_precision = 20;  /* Default to 20 usecs. */
 
@@ -496,9 +499,46 @@ static void drm_vblank_init_release(struct drm_device *dev,
void *ptr)
 	WARN_ON(READ_ONCE(vblank->enabled) &&
 		drm_core_check_feature(dev, DRIVER_MODESET));
 
+	if (vblank->worker && !IS_ERR(vblank->worker))
+		kthread_destroy_worker(vblank->worker);
 	del_timer_sync(&vblank->disable_timer);
 }
 
+/**
+ * drm_vblank_work_init - initialize a vblank work item
+ * @work: vblank work item
+ * @crtc: CRTC whose vblank will trigger the work execution
+ * @func: work function to be executed
+ *
+ * Initialize a vblank work item for a specific crtc.
+ */
+void drm_vblank_work_init(struct drm_vblank_work *work, struct drm_crtc *crtc,
+			  void (*func)(struct kthread_work *work))
+{
+	kthread_init_work(&work->base, func);
+	INIT_LIST_HEAD(&work->flush_work);
+	INIT_LIST_HEAD(&work->pending);
+	work->vblank = &crtc->dev->vblank[drm_crtc_index(crtc)];
+}
+EXPORT_SYMBOL(drm_vblank_work_init);
+
+static int vblank_worker_init(struct drm_vblank_crtc *vblank)
+{
+	struct sched_param param = {
+		.sched_priority = MAX_RT_PRIO - 1,
+	};
+
+	INIT_LIST_HEAD(&vblank->pending_work);
+	spin_lock_init(&vblank->work_lock);
+	vblank->worker = kthread_create_worker(0, "card%d-crtc%d",
+					       vblank->dev->primary->index,
+					       vblank->pipe);
+	if (IS_ERR(vblank->worker))
+		return PTR_ERR(vblank->worker);
+
+	return sched_setscheduler(vblank->worker->task, SCHED_FIFO, &param);
+}
+
 /**
  * drm_vblank_init - initialize vblank support
  * @dev: DRM device
@@ -538,6 +578,10 @@ int drm_vblank_init(struct drm_device *dev, unsigned int
num_crtcs)
 					       vblank);
 		if (ret)
 			return ret;
+
+		ret = vblank_worker_init(vblank);
+		if (ret)
+			return ret;
 	}
 
 	DRM_INFO("Supports vblank timestamp caching Rev 2 (21.10.2013).\n");
@@ -1878,6 +1922,45 @@ static void drm_handle_vblank_events(struct drm_device
*dev, unsigned int pipe)
 	trace_drm_vblank_event(pipe, seq, now, high_prec);
 }
 
+static void queue_vbl_work(struct drm_vblank_work *work)
+{
+	struct kthread_flush_work *fwork, *tmp;
+	bool busy, reinit = false;
+
+	busy = kthread_queue_work(work->vblank->worker, &work->base);
+	list_for_each_entry_safe(fwork, tmp, &work->flush_work, work.node) {
+		if (busy) {
+			list_del_init(&fwork->work.node);
+			busy = kthread_queue_flush_work(&work->base, fwork);
+			if (!busy)
+				complete(&fwork->done);
+		} else {
+			complete(&fwork->done);
+			reinit = true;
+		}
+	}
+
+	if (reinit)
+		INIT_LIST_HEAD(&work->flush_work);
+}
+
+static void drm_handle_vblank_works(struct drm_vblank_crtc *vblank)
+{
+	struct drm_vblank_work *work, *next;
+	u64 count = atomic64_read(&vblank->count);
+
+	spin_lock(&vblank->work_lock);
+	list_for_each_entry_safe(work, next, &vblank->pending_work, pending) {
+		if (!vblank_passed(count, work->count))
+			continue;
+
+		list_del_init(&work->pending);
+		drm_vblank_put(vblank->dev, vblank->pipe);
+		queue_vbl_work(work);
+	}
+	spin_unlock(&vblank->work_lock);
+}
+
 /**
  * drm_handle_vblank - handle a vblank event
  * @dev: DRM device
@@ -1919,6 +2002,7 @@ bool drm_handle_vblank(struct drm_device *dev, unsigned
int pipe)
 
 	spin_unlock(&dev->vblank_time_lock);
 
+	drm_handle_vblank_works(vblank);
 	wake_up(&vblank->queue);
 
 	/* With instant-off, we defer disabling the interrupt until after
@@ -2129,3 +2213,178 @@ int drm_crtc_queue_sequence_ioctl(struct drm_device
*dev, void *data,
 	kfree(e);
 	return ret;
 }
+
+/**
+ * drm_vblank_work_schedule - schedule a vblank work
+ * @work: vblank work to schedule
+ * @count: target vblank count
+ * @nextonmiss: defer until the next vblank if target vblank was missed
+ *
+ * Schedule @work for execution once the crtc vblank count reaches @count.
+ *
+ * If the crtc vblank count has already reached @count and @nextonmiss is
+ * %false the work starts to execute immediately.
+ *
+ * If the crtc vblank count has already reached @count and @nextonmiss is
+ * %true the work is deferred until the next vblank (as if @count has been
+ * specified as crtc vblank count + 1).
+ *
+ * If @work is already scheduled, this function will reschedule said work
+ * using the new @count.
+ *
+ * Returns:
+ * 0 on success, error code on failure.
+ */
+int drm_vblank_work_schedule(struct drm_vblank_work *work,
+			     u64 count, bool nextonmiss)
+{
+	struct drm_vblank_crtc *vblank = work->vblank;
+	struct drm_device *dev = vblank->dev;
+	u64 cur_vbl;
+	unsigned long irqflags;
+	bool passed, rescheduling = false;
+	int ret = 0;
+
+	spin_lock_irqsave(&vblank->work_lock, irqflags);
+	if (!kthread_work_queuable(&work->base))
+		goto out;
+
+	if (list_empty(&work->pending)) {
+		ret = drm_vblank_get(dev, vblank->pipe);
+		if (ret < 0)
+			goto out;
+	} else if (work->count == count) {
+		/* Already scheduled w/ same vbl count */
+		goto out;
+	} else {
+		rescheduling = true;
+	}
+
+	work->count = count;
+	cur_vbl = drm_vblank_count(dev, vblank->pipe);
+	passed = vblank_passed(cur_vbl, count);
+	if (passed)
+		DRM_ERROR("crtc %d vblank %llu already passed (current %llu)\n",
+			  vblank->pipe, count, cur_vbl);
+
+	if (!nextonmiss && passed) {
+		drm_vblank_put(dev, vblank->pipe);
+		if (rescheduling)
+			list_del_init(&work->pending);
+		queue_vbl_work(work);
+	} else if (rescheduling) {
+		list_move_tail(&work->pending, &vblank->pending_work);
+	} else {
+		list_add_tail(&work->pending, &vblank->pending_work);
+	}
+
+ out:
+	spin_unlock_irqrestore(&vblank->work_lock, irqflags);
+	return ret;
+}
+EXPORT_SYMBOL(drm_vblank_work_schedule);
+
+static bool vblank_work_cancel(struct drm_vblank_work *work)
+{
+	struct kthread_flush_work *pos, *tmp;
+
+	if (list_empty(&work->pending))
+		return false;
+
+	list_for_each_entry_safe(pos, tmp, &work->flush_work, work.node)
+		complete(&pos->done);
+
+	list_del_init(&work->pending);
+	INIT_LIST_HEAD(&work->flush_work);
+	return true;
+}
+
+/**
+ * drm_vblank_work_cancel - cancel a vblank work
+ * @work: vblank work to cancel
+ *
+ * Cancel an already scheduled vblank work.
+ *
+ * On return @work may still be executing, unless the return
+ * value is %true.
+ *
+ * Returns:
+ * True if the work was cancelled before it started to excute, false otherwise.
+ */
+bool drm_vblank_work_cancel(struct drm_vblank_work *work)
+{
+	struct drm_vblank_crtc *vblank = work->vblank;
+	bool cancelled;
+
+	spin_lock_irq(&vblank->work_lock);
+	cancelled = vblank_work_cancel(work);
+	spin_unlock_irq(&vblank->work_lock);
+	if (cancelled)
+		drm_vblank_put(vblank->dev, vblank->pipe);
+
+	return cancelled;
+}
+EXPORT_SYMBOL(drm_vblank_work_cancel);
+
+/**
+ * drm_vblank_work_cancel_sync - cancel a vblank work and wait for it to
+ * finish executing
+ * @work: vblank work to cancel
+ *
+ * Cancel an already scheduled vblank work and wait for its
+ * execution to finish.
+ *
+ * On return @work is no longer guaraneed to be executing.
+ *
+ * Returns:
+ * %True if the work was cancelled before it started to execute, %false
+ * otherwise.
+ */
+bool drm_vblank_work_cancel_sync(struct drm_vblank_work *work)
+{
+	struct drm_vblank_crtc *vblank = work->vblank;
+	bool vbl_put, cancelled;
+
+	spin_lock_irq(&vblank->work_lock);
+	cancelled = vblank_work_cancel(work);
+	vbl_put = cancelled;
+
+	kthread_block_work_queuing(vblank->worker, &work->base);
+	spin_unlock_irq(&vblank->work_lock);
+
+	cancelled |= kthread_cancel_work_sync(&work->base);
+	if (vbl_put)
+		drm_vblank_put(vblank->dev, vblank->pipe);
+
+	kthread_unblock_work_queuing(vblank->worker, &work->base);
+
+	return cancelled;
+}
+EXPORT_SYMBOL(drm_vblank_work_cancel_sync);
+
+/**
+ * drm_vblank_work_flush - wait for a scheduled vblank work to finish
+ * executing
+ * @work: vblank work to flush
+ *
+ * Wait until @work has finished executing once.
+ */
+void drm_vblank_work_flush(struct drm_vblank_work *work)
+{
+	struct drm_vblank_crtc *vblank = work->vblank;
+	bool scheduled;
+	DEFINE_KTHREAD_FLUSH_WORK(fwork);
+
+	spin_lock_irq(&vblank->work_lock);
+
+	scheduled = !list_empty(&work->pending);
+	if (scheduled)
+		list_add_tail(&fwork.work.node, &work->flush_work);
+	else
+		scheduled = kthread_queue_flush_work(&work->base, &fwork);
+
+	spin_unlock_irq(&vblank->work_lock);
+	if (scheduled)
+		wait_for_completion(&fwork.done);
+}
+EXPORT_SYMBOL(drm_vblank_work_flush);
diff --git a/include/drm/drm_vblank.h b/include/drm/drm_vblank.h
index dd9f5b9e56e4..020f48bc8473 100644
--- a/include/drm/drm_vblank.h
+++ b/include/drm/drm_vblank.h
@@ -27,6 +27,7 @@
 #include <linux/seqlock.h>
 #include <linux/idr.h>
 #include <linux/poll.h>
+#include <linux/kthread.h>
 
 #include <drm/drm_file.h>
 #include <drm/drm_modes.h>
@@ -203,8 +204,42 @@ struct drm_vblank_crtc {
 	 * disabling functions multiple times.
 	 */
 	bool enabled;
+
+	/**
+	 * @worker: The &kthread_worker used for executing vblank works.
+	 */
+	struct kthread_worker *worker;
+	/**
+	 * @work_lock: The spinlock that protects @pending_work
+	 */
+	spinlock_t work_lock;
+	/**
+	 * @pending_work: A list of scheduled &drm_vblank_work that are
+	 * waiting for a future vblank.
+	 */
+	struct list_head pending_work;
+};
+
+struct drm_vblank_work {
+	struct kthread_work base;
+	struct drm_vblank_crtc *vblank;
+	u64 count;
+
+	struct list_head pending;
+	struct list_head flush_work;
 };
 
+#define to_drm_vblank_work(_work) \
+	container_of((_work), struct drm_vblank_work, base)
+
+int drm_vblank_work_schedule(struct drm_vblank_work *work,
+			     u64 count, bool nextonmiss);
+void drm_vblank_work_init(struct drm_vblank_work *work, struct drm_crtc *crtc,
+			  void (*func)(struct kthread_work *work));
+bool drm_vblank_work_cancel(struct drm_vblank_work *work);
+bool drm_vblank_work_cancel_sync(struct drm_vblank_work *work);
+void drm_vblank_work_flush(struct drm_vblank_work *work);
+
 int drm_vblank_init(struct drm_device *dev, unsigned int num_crtcs);
 bool drm_dev_has_vblank(const struct drm_device *dev);
 u64 drm_crtc_vblank_count(struct drm_crtc *crtc);
-- 
2.25.4

We'll be rolling back more things in this function, and the way it's
structured is a bit confusing. So, let's clean this up a bit and just
unroll in the event of failure.

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/head.c | 33 +++++++++++++++++--------
 1 file changed, 23 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/head.c
b/drivers/gpu/drm/nouveau/dispnv50/head.c
index 8f6455697ba7..e29ea40e7c33 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/head.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/head.c
@@ -507,20 +507,28 @@ nv50_head_create(struct drm_device *dev, int index)
 
 	if (disp->disp->object.oclass < GV100_DISP) {
 		ret = nv50_base_new(drm, head->base.index, &base);
+		if (ret)
+			goto fail_free;
+
 		ret = nv50_ovly_new(drm, head->base.index, &ovly);
+		if (ret)
+			goto fail_free;
 	} else {
 		ret = nv50_wndw_new(drm, DRM_PLANE_TYPE_PRIMARY,
 				    head->base.index * 2 + 0, &base);
+		if (ret)
+			goto fail_free;
+
 		ret = nv50_wndw_new(drm, DRM_PLANE_TYPE_OVERLAY,
 				    head->base.index * 2 + 1, &ovly);
-	}
-	if (ret == 0)
-		ret = nv50_curs_new(drm, head->base.index, &curs);
-	if (ret) {
-		kfree(head);
-		return ERR_PTR(ret);
+		if (ret)
+			goto fail_free;
 	}
 
+	ret = nv50_curs_new(drm, head->base.index, &curs);
+	if (ret)
+		goto fail_free;
+
 	crtc = &head->base.base;
 	drm_crtc_init_with_planes(dev, crtc, &base->plane, &curs->plane,
 				  &nv50_head_func, "head-%d", head->base.index);
@@ -533,11 +541,16 @@ nv50_head_create(struct drm_device *dev, int index)
 
 	if (head->func->olut_set) {
 		ret = nv50_lut_init(disp, &drm->client.mmu, &head->olut);
-		if (ret) {
-			nv50_head_destroy(crtc);
-			return ERR_PTR(ret);
-		}
+		if (ret)
+			goto fail_crtc_cleanup;
 	}
 
 	return head;
+
+fail_crtc_cleanup:
+	drm_crtc_cleanup(crtc);
+fail_free:
+	kfree(head);
+
+	return ERR_PTR(ret);
 }
-- 
2.25.4

Currently, we modify the depth value stored in the atomic state when
performing a commit in order to workaround the fact we haven't
implemented support for depths higher then 10 yet. This isn't idempotent
though, as it will happen every atomic commit where we modify the OR
state even if the head's depth in the atomic state hasn't been modified.

Normally this wouldn't matter, since we don't modify OR state outside of
modesets, but since the CRC capture region is implemented as part of the
OR state in hardware we'll want to make sure all commits modifying OR
state are idempotent so as to avoid changing the depth unexpectedly.

So, fix this by simply not writing the reduced depth value we come up
with to the atomic state.

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/headc37d.c | 11 +++++++----
 drivers/gpu/drm/nouveau/dispnv50/headc57d.c | 11 +++++++----
 2 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
b/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
index 00011ce109a6..68920f8d9c79 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
@@ -27,17 +27,20 @@ static void
 headc37d_or(struct nv50_head *head, struct nv50_head_atom *asyh)
 {
 	struct nv50_dmac *core =
&nv50_disp(head->base.base.dev)->core->chan;
+	u8 depth;
 	u32 *push;
+
 	if ((push = evo_wait(core, 2))) {
 		/*XXX: This is a dirty hack until OR depth handling is
 		 *     improved later for deep colour etc.
 		 */
 		switch (asyh->or.depth) {
-		case 6: asyh->or.depth = 5; break;
-		case 5: asyh->or.depth = 4; break;
-		case 2: asyh->or.depth = 1; break;
-		case 0:	asyh->or.depth = 4; break;
+		case 6: depth = 5; break;
+		case 5: depth = 4; break;
+		case 2: depth = 1; break;
+		case 0:	depth = 4; break;
 		default:
+			depth = asyh->or.depth;
 			WARN_ON(1);
 			break;
 		}
diff --git a/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
b/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
index 938d910a1b1e..0296cd1d761f 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
@@ -27,17 +27,20 @@ static void
 headc57d_or(struct nv50_head *head, struct nv50_head_atom *asyh)
 {
 	struct nv50_dmac *core =
&nv50_disp(head->base.base.dev)->core->chan;
+	u8 depth;
 	u32 *push;
+
 	if ((push = evo_wait(core, 2))) {
 		/*XXX: This is a dirty hack until OR depth handling is
 		 *     improved later for deep colour etc.
 		 */
 		switch (asyh->or.depth) {
-		case 6: asyh->or.depth = 5; break;
-		case 5: asyh->or.depth = 4; break;
-		case 2: asyh->or.depth = 1; break;
-		case 0:	asyh->or.depth = 4; break;
+		case 6: depth = 5; break;
+		case 5: depth = 4; break;
+		case 2: depth = 1; break;
+		case 0:	depth = 4; break;
 		default:
+			depth = asyh->or.depth;
 			WARN_ON(1);
 			break;
 		}
-- 
2.25.4

While we expose the ability to turn off hardware dithering for nouveau,
we actually make the mistake of turning it on anyway, due to
dithering_depth containing a non-zero value if our dithering depth isn't
also set to 6 bpc.

So, fix it by never enabling dithering when it's disabled.

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/head.c | 24 +++++++++++++-----------
 1 file changed, 13 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/head.c
b/drivers/gpu/drm/nouveau/dispnv50/head.c
index e29ea40e7c33..72bc3bce396a 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/head.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/head.c
@@ -84,18 +84,20 @@ nv50_head_atomic_check_dither(struct nv50_head_atom *armh,
 {
 	u32 mode = 0x00;
 
-	if (asyc->dither.mode == DITHERING_MODE_AUTO) {
-		if (asyh->base.depth > asyh->or.bpc * 3)
-			mode = DITHERING_MODE_DYNAMIC2X2;
-	} else {
-		mode = asyc->dither.mode;
-	}
+	if (asyc->dither.mode) {
+		if (asyc->dither.mode == DITHERING_MODE_AUTO) {
+			if (asyh->base.depth > asyh->or.bpc * 3)
+				mode = DITHERING_MODE_DYNAMIC2X2;
+		} else {
+			mode = asyc->dither.mode;
+		}
 
-	if (asyc->dither.depth == DITHERING_DEPTH_AUTO) {
-		if (asyh->or.bpc >= 8)
-			mode |= DITHERING_DEPTH_8BPC;
-	} else {
-		mode |= asyc->dither.depth;
+		if (asyc->dither.depth == DITHERING_DEPTH_AUTO) {
+			if (asyh->or.bpc >= 8)
+				mode |= DITHERING_DEPTH_8BPC;
+		} else {
+			mode |= asyc->dither.depth;
+		}
 	}
 
 	asyh->dither.enable = mode;
-- 
2.25.4

We refer to the armed hardware assembly as armh elsewhere in nouveau, so
fix the naming here to make it consistent.

This patch contains no functional changes.

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/wndw.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/wndw.c
b/drivers/gpu/drm/nouveau/dispnv50/wndw.c
index bb737f9281e6..39cca8eaa066 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/wndw.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/wndw.c
@@ -397,7 +397,7 @@ nv50_wndw_atomic_check(struct drm_plane *plane, struct
drm_plane_state *state)
 	struct nv50_wndw *wndw = nv50_wndw(plane);
 	struct nv50_wndw_atom *armw = nv50_wndw_atom(wndw->plane.state);
 	struct nv50_wndw_atom *asyw = nv50_wndw_atom(state);
-	struct nv50_head_atom *harm = NULL, *asyh = NULL;
+	struct nv50_head_atom *armh = NULL, *asyh = NULL;
 	bool modeset = false;
 	int ret;
 
@@ -418,9 +418,9 @@ nv50_wndw_atomic_check(struct drm_plane *plane, struct
drm_plane_state *state)
 
 	/* Fetch assembly state for the head the window used to belong to. */
 	if (armw->state.crtc) {
-		harm = nv50_head_atom_get(asyw->state.state, armw->state.crtc);
-		if (IS_ERR(harm))
-			return PTR_ERR(harm);
+		armh = nv50_head_atom_get(asyw->state.state, armw->state.crtc);
+		if (IS_ERR(armh))
+			return PTR_ERR(armh);
 	}
 
 	/* LUT configuration can potentially cause the window to be disabled. */
@@ -444,8 +444,8 @@ nv50_wndw_atomic_check(struct drm_plane *plane, struct
drm_plane_state *state)
 		asyh->wndw.mask |= BIT(wndw->id);
 	} else
 	if (armw->visible) {
-		nv50_wndw_atomic_check_release(wndw, asyw, harm);
-		harm->wndw.mask &= ~BIT(wndw->id);
+		nv50_wndw_atomic_check_release(wndw, asyw, armh);
+		armh->wndw.mask &= ~BIT(wndw->id);
 	} else {
 		return 0;
 	}
-- 
2.25.4

While we're not quite ready yet to add support for flexible wndw
mappings, we are going to need to at least keep track of the static wndw
mappings we're currently using in each head's atomic state. We'll
likely
use this in the future to implement real flexible window mapping, but
the primary reason we'll need this is for CRC support.

See: on nvidia hardware, each CRC entry in the CRC notifier dma context
has a "tag". This tag corresponds to the nth update on a specific
EVO/NvDisplay channel, which itself is referred to as the "controlling
channel". For gf119+ this can be the core channel, ovly channel, or base
channel. Since we don't expose CRC entry tags to userspace, we simply
ignore this feature and always use the core channel as the controlling
channel. Simple.

Things get a little bit more complicated on gv100+ though. GV100+ only
lets us set the controlling channel to a specific wndw channel, and that
wndw must be owned by the head that we're grabbing CRCs when we enable
CRC generation. Thus, we always need to make sure that each atomic head
state has at least one wndw that is mapped to the head, which will be
used as the controlling channel.

Note that since we don't have flexible wndw mappings yet, we don't
expect to run into any scenarios yet where we'd have a head with no
mapped wndws. When we do add support for flexible wndw mappings however,
we'll need to make sure that we handle reprogramming CRC capture if our
controlling wndw is moved to another head (and potentially reject the
new head state entirely if we can't find another available wndw to
replace it).

With that being said, nouveau currently tracks wndw visibility on heads.
It does not keep track of the actual ownership mappings, which are
(currently) statically programmed. To fix this, we introduce another
bitmask into nv50_head_atom.wndw to keep track of ownership separately
from visibility. We then introduce a nv50_head callback to handle
populating the wndw ownership map, and call it during the atomic check
phase when core->assign_windows is set to true.

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/atom.h     |  1 +
 drivers/gpu/drm/nouveau/dispnv50/disp.c     | 16 ++++++++++++++++
 drivers/gpu/drm/nouveau/dispnv50/head.h     |  2 ++
 drivers/gpu/drm/nouveau/dispnv50/headc37d.c | 10 ++++++++++
 drivers/gpu/drm/nouveau/dispnv50/headc57d.c |  2 ++
 5 files changed, 31 insertions(+)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/atom.h
b/drivers/gpu/drm/nouveau/dispnv50/atom.h
index 24f7700768da..62faaf60f47a 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/atom.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/atom.h
@@ -18,6 +18,7 @@ struct nv50_head_atom {
 
 	struct {
 		u32 mask;
+		u32 owned;
 		u32 olut;
 	} wndw;
 
diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c
b/drivers/gpu/drm/nouveau/dispnv50/disp.c
index 6be9df1820c5..5be036b3b533 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/disp.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c
@@ -2254,12 +2254,28 @@ static int
 nv50_disp_atomic_check(struct drm_device *dev, struct drm_atomic_state *state)
 {
 	struct nv50_atom *atom = nv50_atom(state);
+	struct nv50_core *core = nv50_disp(dev)->core;
 	struct drm_connector_state *old_connector_state, *new_connector_state;
 	struct drm_connector *connector;
 	struct drm_crtc_state *new_crtc_state;
 	struct drm_crtc *crtc;
+	struct nv50_head *head;
+	struct nv50_head_atom *asyh;
 	int ret, i;
 
+	if (core->assign_windows &&
core->func->head->static_wndw_map) {
+		drm_for_each_crtc(crtc, dev) {
+			new_crtc_state = drm_atomic_get_crtc_state(state,
+								   crtc);
+			if (IS_ERR(new_crtc_state))
+				return PTR_ERR(new_crtc_state);
+
+			head = nv50_head(crtc);
+			asyh = nv50_head_atom(new_crtc_state);
+			core->func->head->static_wndw_map(head, asyh);
+		}
+	}
+
 	/* We need to handle colour management on a per-plane basis. */
 	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
 		if (new_crtc_state->color_mgmt_changed) {
diff --git a/drivers/gpu/drm/nouveau/dispnv50/head.h
b/drivers/gpu/drm/nouveau/dispnv50/head.h
index c32b27cdaefc..c05bbba9e247 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/head.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/head.h
@@ -40,6 +40,7 @@ struct nv50_head_func {
 	void (*dither)(struct nv50_head *, struct nv50_head_atom *);
 	void (*procamp)(struct nv50_head *, struct nv50_head_atom *);
 	void (*or)(struct nv50_head *, struct nv50_head_atom *);
+	void (*static_wndw_map)(struct nv50_head *, struct nv50_head_atom *);
 };
 
 extern const struct nv50_head_func head507d;
@@ -86,6 +87,7 @@ int headc37d_curs_format(struct nv50_head *, struct
nv50_wndw_atom *,
 void headc37d_curs_set(struct nv50_head *, struct nv50_head_atom *);
 void headc37d_curs_clr(struct nv50_head *);
 void headc37d_dither(struct nv50_head *, struct nv50_head_atom *);
+void headc37d_static_wndw_map(struct nv50_head *, struct nv50_head_atom *);
 
 extern const struct nv50_head_func headc57d;
 #endif
diff --git a/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
b/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
index 68920f8d9c79..cf5a68f4021a 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
@@ -203,6 +203,15 @@ headc37d_view(struct nv50_head *head, struct nv50_head_atom
*asyh)
 	}
 }
 
+void
+headc37d_static_wndw_map(struct nv50_head *head, struct nv50_head_atom *asyh)
+{
+	int i, end;
+
+	for (i = head->base.index * 2, end = i + 2; i < end; i++)
+		asyh->wndw.owned |= BIT(i);
+}
+
 const struct nv50_head_func
 headc37d = {
 	.view = headc37d_view,
@@ -218,4 +227,5 @@ headc37d = {
 	.dither = headc37d_dither,
 	.procamp = headc37d_procamp,
 	.or = headc37d_or,
+	.static_wndw_map = headc37d_static_wndw_map,
 };
diff --git a/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
b/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
index 0296cd1d761f..65e3b60804c6 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
@@ -210,4 +210,6 @@ headc57d = {
 	.dither = headc37d_dither,
 	.procamp = headc57d_procamp,
 	.or = headc57d_or,
+	/* TODO: flexible window mappings */
+	.static_wndw_map = headc37d_static_wndw_map,
 };
-- 
2.25.4

In order to make sure that we flush disable updates at the right time
when disabling CRCs, we'll need to be able to look at the outp state to
see if we're changing it at the same time that we're disabling CRCs.

So, expose the struct in disp.h.

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/disp.c | 18 ------------------
 drivers/gpu/drm/nouveau/dispnv50/disp.h | 14 ++++++++++++++
 2 files changed, 14 insertions(+), 18 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c
b/drivers/gpu/drm/nouveau/dispnv50/disp.c
index 5be036b3b533..5d5661967d50 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/disp.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c
@@ -57,24 +57,6 @@
 
 #include <subdev/bios/dp.h>
 
-/******************************************************************************
- * Atomic state
- *****************************************************************************/
-
-struct nv50_outp_atom {
-	struct list_head head;
-
-	struct drm_encoder *encoder;
-	bool flush_disable;
-
-	union nv50_outp_atom_mask {
-		struct {
-			bool ctrl:1;
-		};
-		u8 mask;
-	} set, clr;
-};
-
 /******************************************************************************
  * EVO channel
  *****************************************************************************/
diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.h
b/drivers/gpu/drm/nouveau/dispnv50/disp.h
index d54fe00ac3a3..8935ebce8ab0 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/disp.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/disp.h
@@ -70,6 +70,20 @@ struct nv50_dmac {
 	struct mutex lock;
 };
 
+struct nv50_outp_atom {
+	struct list_head head;
+
+	struct drm_encoder *encoder;
+	bool flush_disable;
+
+	union nv50_outp_atom_mask {
+		struct {
+			bool ctrl:1;
+		};
+		u8 mask;
+	} set, clr;
+};
+
 int nv50_dmac_create(struct nvif_device *device, struct nvif_object *disp,
 		     const s32 *oclass, u8 head, void *data, u32 size,
 		     u64 syncbuf, struct nv50_dmac *dmac);
-- 
2.25.4

While most of the functionality on Nvidia GPUs doesn't require using an
explicit handle instead of the main VRAM handle + offset, there are a
couple of places that do require explicit handles, such as CRC
functionality. Since this means we're about to add another
nouveau-chosen handle, let's just go ahead and move any hard-coded
handles into a single header. This is just to keep things slightly
organized, and to make it a little bit easier if we need to add more
handles in the future.

This patch should contain no functional changes.

Changes since v3:
* Correct SPDX license identifier (checkpatch)

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv50/disp.c    |  7 +++++--
 drivers/gpu/drm/nouveau/dispnv50/handles.h | 15 +++++++++++++++
 drivers/gpu/drm/nouveau/dispnv50/wndw.c    |  3 ++-
 3 files changed, 22 insertions(+), 3 deletions(-)
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/handles.h

diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c
b/drivers/gpu/drm/nouveau/dispnv50/disp.c
index 5d5661967d50..568029825f62 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/disp.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c
@@ -26,6 +26,7 @@
 #include "core.h"
 #include "head.h"
 #include "wndw.h"
+#include "handles.h"
 
 #include <linux/dma-mapping.h>
 #include <linux/hdmi.h>
@@ -154,7 +155,8 @@ nv50_dmac_create(struct nvif_device *device, struct
nvif_object *disp,
 	if (!syncbuf)
 		return 0;
 
-	ret = nvif_object_init(&dmac->base.user, 0xf0000000, NV_DMA_IN_MEMORY,
+	ret = nvif_object_init(&dmac->base.user, NV50_DISP_HANDLE_SYNCBUF,
+			       NV_DMA_IN_MEMORY,
 			       &(struct nv_dma_v0) {
 					.target = NV_DMA_V0_TARGET_VRAM,
 					.access = NV_DMA_V0_ACCESS_RDWR,
@@ -165,7 +167,8 @@ nv50_dmac_create(struct nvif_device *device, struct
nvif_object *disp,
 	if (ret)
 		return ret;
 
-	ret = nvif_object_init(&dmac->base.user, 0xf0000001, NV_DMA_IN_MEMORY,
+	ret = nvif_object_init(&dmac->base.user, NV50_DISP_HANDLE_VRAM,
+			       NV_DMA_IN_MEMORY,
 			       &(struct nv_dma_v0) {
 					.target = NV_DMA_V0_TARGET_VRAM,
 					.access = NV_DMA_V0_ACCESS_RDWR,
diff --git a/drivers/gpu/drm/nouveau/dispnv50/handles.h
b/drivers/gpu/drm/nouveau/dispnv50/handles.h
new file mode 100644
index 000000000000..d1beeb9a444d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/dispnv50/handles.h
@@ -0,0 +1,15 @@
+// SPDX-License-Identifier: MIT
+#ifndef __NV50_KMS_HANDLES_H__
+#define __NV50_KMS_HANDLES_H__
+
+/*
+ * Various hard-coded object handles that nouveau uses. These are made-up by
+ * nouveau developers, not Nvidia. The only significance of the handles chosen
+ * is that they must all be unique.
+ */
+#define NV50_DISP_HANDLE_SYNCBUF                                       
0xf0000000
+#define NV50_DISP_HANDLE_VRAM                                          
0xf0000001
+
+#define NV50_DISP_HANDLE_WNDW_CTX(kind)                        (0xfb000000 |
kind)
+
+#endif /* !__NV50_KMS_HANDLES_H__ */
diff --git a/drivers/gpu/drm/nouveau/dispnv50/wndw.c
b/drivers/gpu/drm/nouveau/dispnv50/wndw.c
index 39cca8eaa066..cb67a715bd69 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/wndw.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/wndw.c
@@ -21,6 +21,7 @@
  */
 #include "wndw.h"
 #include "wimm.h"
+#include "handles.h"
 
 #include <nvif/class.h>
 #include <nvif/cl0002.h>
@@ -44,7 +45,7 @@ nv50_wndw_ctxdma_new(struct nv50_wndw *wndw, struct
nouveau_framebuffer *fb)
 	struct nouveau_drm *drm = nouveau_drm(fb->base.dev);
 	struct nv50_wndw_ctxdma *ctxdma;
 	const u8    kind = fb->nvbo->kind;
-	const u32 handle = 0xfb000000 | kind;
+	const u32 handle = NV50_DISP_HANDLE_WNDW_CTX(kind);
 	struct {
 		struct nv_dma_v0 base;
 		union {
-- 
2.25.4

This introduces support for CRC readback on gf119+, using the
documentation generously provided to us by Nvidia:

https://github.com/NVIDIA/open-gpu-doc/blob/master/Display-CRC/display-crc.txt

We expose all available CRC sources. SF, SOR, PIOR, and DAC are exposed
through a single set of "outp" sources: outp-active/auto for a CRC of
the scanout region, outp-complete for a CRC of both the scanout and
blanking/sync region combined, and outp-inactive for a CRC of only the
blanking/sync region. For each source, nouveau selects the appropriate
tap point based on the output path in use. We also expose an "rg"
source, which allows for capturing CRCs of the scanout raster before
it's encoded into a video signal in the output path. This tap point is
referred to as the raster generator.

Note that while there's some other neat features that can be used with
CRC capture on nvidia hardware, like capturing from two CRC sources
simultaneously, I couldn't see any usecase for them and did not
implement them.

Nvidia only allows for accessing CRCs through a shared DMA region that
we program through the core EVO/NvDisplay channel which is referred to
as the notifier context. The notifier context is limited to either 255
(for Fermi-Pascal) or 2047 (Volta+) entries to store CRCs in, and
unfortunately the hardware simply drops CRCs and reports an overflow
once all available entries in the notifier context are filled.

Since the DRM CRC API and igt-gpu-tools don't expect there to be a limit
on how many CRCs can be captured, we work around this in nouveau by
allocating two separate notifier contexts for each head instead of one.
We schedule a vblank worker ahead of time so that once we start getting
close to filling up all of the available entries in the notifier
context, we can swap the currently used notifier context out with
another pre-prepared notifier context in a manner similar to page
flipping.

Unfortunately, the hardware only allows us to this by flushing two
separate updates on the core channel: one to release the current
notifier context handle, and one to program the next notifier context's
handle. When the hardware processes the first update, the CRC for the
current frame is lost. However, the second update can be flushed
immediately without waiting for the first to complete so that CRC
generation resumes on the next frame. According to Nvidia's hardware
engineers, there isn't any cleaner way of flipping notifier contexts
that would avoid this.

Since using vblank workers to swap out the notifier context will ensure
we can usually flush both updates to hardware within the timespan of a
single frame, we can also ensure that there will only be exactly one
frame lost between the first and second update being executed by the
hardware. This gives us the guarantee that we're always correctly
matching each CRC entry with it's respective frame even after a context
flip. And since IGT will retrieve the CRC entry for a frame by waiting
until it receives a CRC for any subsequent frames, this doesn't cause an
issue with any tests and is much simpler than trying to change the
current DRM API to accommodate.

In order to facilitate testing of correct handling of this limitation,
we also expose a debugfs interface to manually control the threshold for
when we start trying to flip the notifier context. We will use this in
igt to trigger a context flip for testing purposes without needing to
wait for the notifier to completely fill up. This threshold is reset
to the default value set by nouveau after each capture, and is exposed
in a separate folder within each CRTC's debugfs directory labelled
"nv_crc".

Changes since v1:
* Forgot to finish saving crc.h before saving, whoops. This just adds
  some corrections to the empty function declarations that we use if
  CONFIG_DEBUG_FS isn't enabled.
Changes since v2:
* Don't check return code from debugfs_create_dir() or
  debugfs_create_file() - Greg K-H
Changes since v3:
  (no functional changes)
* Fix SPDX license identifiers (checkpatch)
* s/uint32_t/u32/ (checkpatch)
* Fix indenting in switch cases (checkpatch)

Signed-off-by: Lyude Paul <lyude at redhat.com>
---
 drivers/gpu/drm/nouveau/dispnv04/crtc.c     |  25 +-
 drivers/gpu/drm/nouveau/dispnv50/Kbuild     |   4 +
 drivers/gpu/drm/nouveau/dispnv50/atom.h     |  20 +
 drivers/gpu/drm/nouveau/dispnv50/core.h     |   4 +
 drivers/gpu/drm/nouveau/dispnv50/core907d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/core917d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/corec37d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/corec57d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/crc.c      | 715 ++++++++++++++++++++
 drivers/gpu/drm/nouveau/dispnv50/crc.h      | 125 ++++
 drivers/gpu/drm/nouveau/dispnv50/crc907d.c  | 139 ++++
 drivers/gpu/drm/nouveau/dispnv50/crcc37d.c  | 153 +++++
 drivers/gpu/drm/nouveau/dispnv50/disp.c     |  24 +-
 drivers/gpu/drm/nouveau/dispnv50/disp.h     |  10 +
 drivers/gpu/drm/nouveau/dispnv50/handles.h  |   1 +
 drivers/gpu/drm/nouveau/dispnv50/head.c     |  85 ++-
 drivers/gpu/drm/nouveau/dispnv50/head.h     |  11 +-
 drivers/gpu/drm/nouveau/dispnv50/head907d.c |  14 +-
 drivers/gpu/drm/nouveau/dispnv50/headc37d.c |   6 +-
 drivers/gpu/drm/nouveau/dispnv50/headc57d.c |   7 +-
 drivers/gpu/drm/nouveau/nouveau_display.c   |  60 +-
 21 files changed, 1341 insertions(+), 74 deletions(-)
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crc.c
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crc.h
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crc907d.c
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crcc37d.c

diff --git a/drivers/gpu/drm/nouveau/dispnv04/crtc.c
b/drivers/gpu/drm/nouveau/dispnv04/crtc.c
index 1f08de4241e0..fc178ffce8cd 100644
--- a/drivers/gpu/drm/nouveau/dispnv04/crtc.c
+++ b/drivers/gpu/drm/nouveau/dispnv04/crtc.c
@@ -44,6 +44,9 @@
 #include <subdev/bios/pll.h>
 #include <subdev/clk.h>
 
+#include <nvif/event.h>
+#include <nvif/cl0046.h>
+
 static int
 nv04_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y,
 			struct drm_framebuffer *old_fb);
@@ -755,6 +758,7 @@ static void nv_crtc_destroy(struct drm_crtc *crtc)
 	nouveau_bo_unmap(nv_crtc->cursor.nvbo);
 	nouveau_bo_unpin(nv_crtc->cursor.nvbo);
 	nouveau_bo_ref(NULL, &nv_crtc->cursor.nvbo);
+	nvif_notify_fini(&nv_crtc->vblank);
 	kfree(nv_crtc);
 }
 
@@ -1296,9 +1300,19 @@ create_primary_plane(struct drm_device *dev)
         return primary;
 }
 
+static int nv04_crtc_vblank_handler(struct nvif_notify *notify)
+{
+	struct nouveau_crtc *nv_crtc +		container_of(notify, struct nouveau_crtc,
vblank);
+
+	drm_crtc_handle_vblank(&nv_crtc->base);
+	return NVIF_NOTIFY_KEEP;
+}
+
 int
 nv04_crtc_create(struct drm_device *dev, int crtc_num)
 {
+	struct nouveau_display *disp = nouveau_display(dev);
 	struct nouveau_crtc *nv_crtc;
 	int ret;
 
@@ -1336,5 +1350,14 @@ nv04_crtc_create(struct drm_device *dev, int crtc_num)
 
 	nv04_cursor_init(nv_crtc);
 
-	return 0;
+	ret = nvif_notify_init(&disp->disp.object, nv04_crtc_vblank_handler,
+			       false, NV04_DISP_NTFY_VBLANK,
+			       &(struct nvif_notify_head_req_v0) {
+				    .head = nv_crtc->index,
+			       },
+			       sizeof(struct nvif_notify_head_req_v0),
+			       sizeof(struct nvif_notify_head_rep_v0),
+			       &nv_crtc->vblank);
+
+	return ret;
 }
diff --git a/drivers/gpu/drm/nouveau/dispnv50/Kbuild
b/drivers/gpu/drm/nouveau/dispnv50/Kbuild
index e0c435eae664..6fdddb266fb1 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/Kbuild
+++ b/drivers/gpu/drm/nouveau/dispnv50/Kbuild
@@ -10,6 +10,10 @@ nouveau-y += dispnv50/core917d.o
 nouveau-y += dispnv50/corec37d.o
 nouveau-y += dispnv50/corec57d.o
 
+nouveau-$(CONFIG_DEBUG_FS) += dispnv50/crc.o
+nouveau-$(CONFIG_DEBUG_FS) += dispnv50/crc907d.o
+nouveau-$(CONFIG_DEBUG_FS) += dispnv50/crcc37d.o
+
 nouveau-y += dispnv50/dac507d.o
 nouveau-y += dispnv50/dac907d.o
 
diff --git a/drivers/gpu/drm/nouveau/dispnv50/atom.h
b/drivers/gpu/drm/nouveau/dispnv50/atom.h
index 62faaf60f47a..3d82b3c67dec 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/atom.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/atom.h
@@ -2,6 +2,9 @@
 #define __NV50_KMS_ATOM_H__
 #define nv50_atom(p) container_of((p), struct nv50_atom, state)
 #include <drm/drm_atomic.h>
+#include "crc.h"
+
+struct nouveau_encoder;
 
 struct nv50_atom {
 	struct drm_atomic_state state;
@@ -115,9 +118,12 @@ struct nv50_head_atom {
 		u8 nhsync:1;
 		u8 nvsync:1;
 		u8 depth:4;
+		u8 crc_raster:2;
 		u8 bpc;
 	} or;
 
+	struct nv50_crc_atom crc;
+
 	/* Currently only used for MST */
 	struct {
 		int pbn;
@@ -135,6 +141,7 @@ struct nv50_head_atom {
 			bool ovly:1;
 			bool dither:1;
 			bool procamp:1;
+			bool crc:1;
 			bool or:1;
 		};
 		u16 mask;
@@ -150,6 +157,19 @@ nv50_head_atom_get(struct drm_atomic_state *state, struct
drm_crtc *crtc)
 	return nv50_head_atom(statec);
 }
 
+static inline struct drm_encoder *
+nv50_head_atom_get_encoder(struct nv50_head_atom *atom)
+{
+	struct drm_encoder *encoder = NULL;
+
+	/* We only ever have a single encoder */
+	drm_for_each_encoder_mask(encoder, atom->state.crtc->dev,
+				  atom->state.encoder_mask)
+		break;
+
+	return encoder;
+}
+
 #define nv50_wndw_atom(p) container_of((p), struct nv50_wndw_atom, state)
 
 struct nv50_wndw_atom {
diff --git a/drivers/gpu/drm/nouveau/dispnv50/core.h
b/drivers/gpu/drm/nouveau/dispnv50/core.h
index ff94f3f6f264..47470db0f154 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/core.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/core.h
@@ -2,6 +2,7 @@
 #define __NV50_KMS_CORE_H__
 #include "disp.h"
 #include "atom.h"
+#include "crc.h"
 
 struct nv50_core {
 	const struct nv50_core_func *func;
@@ -24,6 +25,9 @@ struct nv50_core_func {
 	} wndw;
 
 	const struct nv50_head_func *head;
+#if IS_ENABLED(CONFIG_DEBUG_FS)
+	const struct nv50_crc_func *crc;
+#endif
 	const struct nv50_outp_func {
 		void (*ctrl)(struct nv50_core *, int or, u32 ctrl,
 			     struct nv50_head_atom *);
diff --git a/drivers/gpu/drm/nouveau/dispnv50/core907d.c
b/drivers/gpu/drm/nouveau/dispnv50/core907d.c
index ef822f813435..fd6aaf629d02 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/core907d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/core907d.c
@@ -29,6 +29,9 @@ core907d = {
 	.ntfy_wait_done = core507d_ntfy_wait_done,
 	.update = core507d_update,
 	.head = &head907d,
+#if IS_ENABLED(CONFIG_DEBUG_FS)
+	.crc = &crc907d,
+#endif
 	.dac = &dac907d,
 	.sor = &sor907d,
 };
diff --git a/drivers/gpu/drm/nouveau/dispnv50/core917d.c
b/drivers/gpu/drm/nouveau/dispnv50/core917d.c
index 392338df5bfd..46debdce87ff 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/core917d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/core917d.c
@@ -29,6 +29,9 @@ core917d = {
 	.ntfy_wait_done = core507d_ntfy_wait_done,
 	.update = core507d_update,
 	.head = &head917d,
+#if IS_ENABLED(CONFIG_DEBUG_FS)
+	.crc = &crc907d,
+#endif
 	.dac = &dac907d,
 	.sor = &sor907d,
 };
diff --git a/drivers/gpu/drm/nouveau/dispnv50/corec37d.c
b/drivers/gpu/drm/nouveau/dispnv50/corec37d.c
index c03cb987856b..b9093a475bdb 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/corec37d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/corec37d.c
@@ -116,6 +116,9 @@ corec37d = {
 	.wndw.owner = corec37d_wndw_owner,
 	.head = &headc37d,
 	.sor = &sorc37d,
+#if IS_ENABLED(CONFIG_DEBUG_FS)
+	.crc = &crcc37d,
+#endif
 };
 
 int
diff --git a/drivers/gpu/drm/nouveau/dispnv50/corec57d.c
b/drivers/gpu/drm/nouveau/dispnv50/corec57d.c
index 147adcd60937..0ffd6286985c 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/corec57d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/corec57d.c
@@ -51,6 +51,9 @@ corec57d = {
 	.wndw.owner = corec37d_wndw_owner,
 	.head = &headc57d,
 	.sor = &sorc37d,
+#if IS_ENABLED(CONFIG_DEBUG_FS)
+	.crc = &crcc37d,
+#endif
 };
 
 int
diff --git a/drivers/gpu/drm/nouveau/dispnv50/crc.c
b/drivers/gpu/drm/nouveau/dispnv50/crc.c
new file mode 100644
index 000000000000..e8fbed129917
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/dispnv50/crc.c
@@ -0,0 +1,715 @@
+// SPDX-License-Identifier: MIT
+#include <linux/string.h>
+#include <drm/drm_crtc.h>
+#include <drm/drm_atomic_helper.h>
+#include <drm/drm_vblank.h>
+
+#include <nvif/class.h>
+#include <nvif/cl0002.h>
+#include <nvif/timer.h>
+
+#include "nouveau_drv.h"
+#include "core.h"
+#include "head.h"
+#include "wndw.h"
+#include "handles.h"
+#include "crc.h"
+
+static const char * const nv50_crc_sources[] = {
+	[NV50_CRC_SOURCE_NONE] = "none",
+	[NV50_CRC_SOURCE_AUTO] = "auto",
+	[NV50_CRC_SOURCE_RG] = "rg",
+	[NV50_CRC_SOURCE_OUTP_ACTIVE] = "outp-active",
+	[NV50_CRC_SOURCE_OUTP_COMPLETE] = "outp-complete",
+	[NV50_CRC_SOURCE_OUTP_INACTIVE] = "outp-inactive",
+};
+
+static int nv50_crc_parse_source(const char *buf, enum nv50_crc_source *s)
+{
+	int i;
+
+	if (!buf) {
+		*s = NV50_CRC_SOURCE_NONE;
+		return 0;
+	}
+
+	i = match_string(nv50_crc_sources, ARRAY_SIZE(nv50_crc_sources), buf);
+	if (i < 0)
+		return i;
+
+	*s = i;
+	return 0;
+}
+
+int
+nv50_crc_verify_source(struct drm_crtc *crtc, const char *source_name,
+		       size_t *values_cnt)
+{
+	struct nouveau_drm *drm = nouveau_drm(crtc->dev);
+	enum nv50_crc_source source;
+
+	if (nv50_crc_parse_source(source_name, &source) < 0) {
+		NV_DEBUG(drm, "unknown source %s\n", source_name);
+		return -EINVAL;
+	}
+
+	*values_cnt = 1;
+	return 0;
+}
+
+const char *const *nv50_crc_get_sources(struct drm_crtc *crtc, size_t *count)
+{
+	*count = ARRAY_SIZE(nv50_crc_sources);
+	return nv50_crc_sources;
+}
+
+static void
+nv50_crc_program_ctx(struct nv50_head *head,
+		     struct nv50_crc_notifier_ctx *ctx)
+{
+	struct nv50_disp *disp = nv50_disp(head->base.base.dev);
+	struct nv50_core *core = disp->core;
+	u32 interlock[NV50_DISP_INTERLOCK__SIZE] = { 0 };
+
+	core->func->crc->set_ctx(head, ctx);
+	core->func->update(core, interlock, false);
+}
+
+static void nv50_crc_ctx_flip_work(struct kthread_work *kwork)
+{
+	struct drm_vblank_work *work = to_drm_vblank_work(kwork);
+	struct nv50_crc *crc = container_of(work, struct nv50_crc, flip_work);
+	struct nv50_head *head = container_of(crc, struct nv50_head, crc);
+	struct drm_crtc *crtc = &head->base.base;
+	struct nv50_disp *disp = nv50_disp(crtc->dev);
+	u8 new_idx = crc->ctx_idx ^ 1;
+
+	/*
+	 * We don't want to accidentally wait for longer then the vblank, so
+	 * try again for the next vblank if we don't grab the lock
+	 */
+	if (!mutex_trylock(&disp->mutex)) {
+		DRM_DEV_DEBUG_KMS(crtc->dev->dev,
+				  "Lock contended, delaying CRC ctx flip for head-%d\n",
+				  head->base.index);
+		drm_vblank_work_schedule(work,
+					 drm_crtc_vblank_count(crtc) + 1,
+					 true);
+		return;
+	}
+
+	DRM_DEV_DEBUG_KMS(crtc->dev->dev,
+			  "Flipping notifier ctx for head %d (%d -> %d)\n",
+			  drm_crtc_index(crtc), crc->ctx_idx, new_idx);
+
+	nv50_crc_program_ctx(head, NULL);
+	nv50_crc_program_ctx(head, &crc->ctx[new_idx]);
+	mutex_unlock(&disp->mutex);
+
+	spin_lock_irq(&crc->lock);
+	crc->ctx_changed = true;
+	spin_unlock_irq(&crc->lock);
+}
+
+static inline void nv50_crc_reset_ctx(struct nv50_crc_notifier_ctx *ctx)
+{
+	memset_io(ctx->mem.object.map.ptr, 0, ctx->mem.object.map.size);
+}
+
+static void
+nv50_crc_get_entries(struct nv50_head *head,
+		     const struct nv50_crc_func *func,
+		     enum nv50_crc_source source)
+{
+	struct drm_crtc *crtc = &head->base.base;
+	struct nv50_crc *crc = &head->crc;
+	u32 output_crc;
+
+	while (crc->entry_idx < func->num_entries) {
+		/*
+		 * While Nvidia's documentation says CRCs are written on each
+		 * subsequent vblank after being enabled, in practice they
+		 * aren't written immediately.
+		 */
+		output_crc = func->get_entry(head, &crc->ctx[crc->ctx_idx],
+					     source, crc->entry_idx);
+		if (!output_crc)
+			return;
+
+		drm_crtc_add_crc_entry(crtc, true, crc->frame, &output_crc);
+		crc->frame++;
+		crc->entry_idx++;
+	}
+}
+
+void nv50_crc_handle_vblank(struct nv50_head *head)
+{
+	struct drm_crtc *crtc = &head->base.base;
+	struct nv50_crc *crc = &head->crc;
+	const struct nv50_crc_func *func +	
nv50_disp(head->base.base.dev)->core->func->crc;
+	struct nv50_crc_notifier_ctx *ctx;
+	bool need_reschedule = false;
+
+	if (!func)
+		return;
+
+	/*
+	 * We don't lose events if we aren't able to report CRCs until the
+	 * next vblank, so only report CRCs if the locks we need aren't
+	 * contended to prevent missing an actual vblank event
+	 */
+	if (!spin_trylock(&crc->lock))
+		return;
+
+	if (!crc->src)
+		goto out;
+
+	ctx = &crc->ctx[crc->ctx_idx];
+	if (crc->ctx_changed && func->ctx_finished(head, ctx)) {
+		nv50_crc_get_entries(head, func, crc->src);
+
+		crc->ctx_idx ^= 1;
+		crc->entry_idx = 0;
+		crc->ctx_changed = false;
+
+		/*
+		 * Unfortunately when notifier contexts are changed during CRC
+		 * capture, we will inevitably lose the CRC entry for the
+		 * frame where the hardware actually latched onto the first
+		 * UPDATE. According to Nvidia's hardware engineers, there's
+		 * no workaround for this.
+		 *
+		 * Now, we could try to be smart here and calculate the number
+		 * of missed CRCs based on audit timestamps, but those were
+		 * removed starting with volta. Since we always flush our
+		 * updates back-to-back without waiting, we'll just be
+		 * optimistic and assume we always miss exactly one frame.
+		 */
+		DRM_DEV_DEBUG_KMS(head->base.base.dev->dev,
+				  "Notifier ctx flip for head-%d finished, lost CRC for frame
%llu\n",
+				  head->base.index, crc->frame);
+		crc->frame++;
+
+		nv50_crc_reset_ctx(ctx);
+		need_reschedule = true;
+	}
+
+	nv50_crc_get_entries(head, func, crc->src);
+
+	if (need_reschedule)
+		drm_vblank_work_schedule(&crc->flip_work,
+					 drm_crtc_vblank_count(crtc)
+					 + crc->flip_threshold
+					 - crc->entry_idx,
+					 true);
+
+out:
+	spin_unlock(&crc->lock);
+}
+
+static void nv50_crc_wait_ctx_finished(struct nv50_head *head,
+				       const struct nv50_crc_func *func,
+				       struct nv50_crc_notifier_ctx *ctx)
+{
+	struct drm_device *dev = head->base.base.dev;
+	struct nouveau_drm *drm = nouveau_drm(dev);
+	s64 ret;
+
+	ret = nvif_msec(&drm->client.device, 50,
+			if (func->ctx_finished(head, ctx)) break;);
+	if (ret == -ETIMEDOUT)
+		NV_ERROR(drm,
+			 "CRC notifier ctx for head %d not finished after 50ms\n",
+			 head->base.index);
+	else if (ret)
+		NV_ATOMIC(drm,
+			  "CRC notifier ctx for head-%d finished after %lldns\n",
+			  head->base.index, ret);
+}
+
+void nv50_crc_atomic_stop_reporting(struct drm_atomic_state *state)
+{
+	struct drm_crtc_state *crtc_state;
+	struct drm_crtc *crtc;
+	int i;
+
+	for_each_new_crtc_in_state(state, crtc, crtc_state, i) {
+		struct nv50_head *head = nv50_head(crtc);
+		struct nv50_head_atom *asyh = nv50_head_atom(crtc_state);
+		struct nv50_crc *crc = &head->crc;
+
+		if (!asyh->clr.crc)
+			continue;
+
+		spin_lock_irq(&crc->lock);
+		crc->src = NV50_CRC_SOURCE_NONE;
+		spin_unlock_irq(&crc->lock);
+
+		drm_crtc_vblank_put(crtc);
+		drm_vblank_work_cancel_sync(&crc->flip_work);
+
+		NV_ATOMIC(nouveau_drm(crtc->dev),
+			  "CRC reporting on vblank for head-%d disabled\n",
+			  head->base.index);
+
+		/* CRC generation is still enabled in hw, we'll just report
+		 * any remaining CRC entries ourselves after it gets disabled
+		 * in hardware
+		 */
+	}
+}
+
+void nv50_crc_atomic_prepare_notifier_contexts(struct drm_atomic_state *state)
+{
+	const struct nv50_crc_func *func +	
nv50_disp(state->dev)->core->func->crc;
+	struct drm_crtc_state *new_crtc_state;
+	struct drm_crtc *crtc;
+	int i;
+
+	for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) {
+		struct nv50_head *head = nv50_head(crtc);
+		struct nv50_head_atom *asyh = nv50_head_atom(new_crtc_state);
+		struct nv50_crc *crc = &head->crc;
+		struct nv50_crc_notifier_ctx *ctx = &crc->ctx[crc->ctx_idx];
+		int i;
+
+		if (asyh->clr.crc && asyh->crc.src) {
+			if (crc->ctx_changed) {
+				nv50_crc_wait_ctx_finished(head, func, ctx);
+				ctx = &crc->ctx[crc->ctx_idx ^ 1];
+			}
+			nv50_crc_wait_ctx_finished(head, func, ctx);
+		}
+
+		if (asyh->set.crc) {
+			crc->entry_idx = 0;
+			crc->ctx_changed = false;
+			for (i = 0; i < ARRAY_SIZE(crc->ctx); i++)
+				nv50_crc_reset_ctx(&crc->ctx[i]);
+		}
+	}
+}
+
+void nv50_crc_atomic_start_reporting(struct drm_atomic_state *state)
+{
+	struct drm_crtc_state *crtc_state;
+	struct drm_crtc *crtc;
+	int i;
+
+	for_each_new_crtc_in_state(state, crtc, crtc_state, i) {
+		struct nv50_head *head = nv50_head(crtc);
+		struct nv50_head_atom *asyh = nv50_head_atom(crtc_state);
+		struct nv50_crc *crc = &head->crc;
+		u64 vbl_count;
+
+		if (!asyh->set.crc)
+			continue;
+
+		drm_crtc_vblank_get(crtc);
+
+		spin_lock_irq(&crc->lock);
+		vbl_count = drm_crtc_vblank_count(crtc);
+		crc->frame = vbl_count;
+		crc->src = asyh->crc.src;
+		drm_vblank_work_schedule(&crc->flip_work,
+					 vbl_count + crc->flip_threshold,
+					 true);
+		spin_unlock_irq(&crc->lock);
+
+		NV_ATOMIC(nouveau_drm(crtc->dev),
+			  "CRC reporting on vblank for head-%d enabled\n",
+			  head->base.index);
+	}
+}
+
+int nv50_crc_atomic_check(struct nv50_head *head,
+			  struct nv50_head_atom *asyh,
+			  struct nv50_head_atom *armh)
+{
+	struct drm_atomic_state *state = asyh->state.state;
+	struct drm_device *dev = head->base.base.dev;
+	struct nv50_atom *atom = nv50_atom(state);
+	struct nv50_disp *disp = nv50_disp(dev);
+	struct drm_encoder *encoder;
+	struct nouveau_encoder *outp;
+	struct nv50_outp_atom *outp_atom;
+	bool changed = armh->crc.src != asyh->crc.src;
+
+	if (!armh->crc.src && !asyh->crc.src) {
+		asyh->set.crc = false;
+		asyh->clr.crc = false;
+		return 0;
+	}
+
+	/* While we don't care about entry tags, Volta+ hw always needs the
+	 * controlling wndw channel programmed to a wndw that's owned by our
+	 * head
+	 */
+	if (asyh->crc.src && disp->disp->object.oclass >=
GV100_DISP &&
+	    !(BIT(asyh->crc.wndw) & asyh->wndw.owned)) {
+		if (!asyh->wndw.owned) {
+			/* TODO: once we support flexible channel ownership,
+			 * we should write some code here to handle attempting
+			 * to "steal" a plane: e.g. take a plane that is
+			 * currently not-visible and owned by another head,
+			 * and reassign it to this head. If we fail to do so,
+			 * we shuld reject the mode outright as CRC capture
+			 * then becomes impossible.
+			 */
+			NV_ATOMIC(nouveau_drm(dev),
+				  "No available wndws for CRC readback\n");
+			return -EINVAL;
+		}
+		asyh->crc.wndw = ffs(asyh->wndw.owned) - 1;
+	}
+
+	if (drm_atomic_crtc_needs_modeset(&asyh->state) || changed ||
+	    armh->crc.wndw != asyh->crc.wndw) {
+		asyh->clr.crc = armh->crc.src && armh->state.active;
+		asyh->set.crc = asyh->crc.src && asyh->state.active;
+		if (changed)
+			asyh->set.or |= armh->or.crc_raster !+					asyh->or.crc_raster;
+
+		/*
+		 * If we're reprogramming our OR, we need to flush the CRC
+		 * disable first
+		 */
+		if (asyh->clr.crc) {
+			encoder = nv50_head_atom_get_encoder(armh);
+			outp = nv50_real_outp(encoder);
+
+			list_for_each_entry(outp_atom, &atom->outp, head) {
+				if (outp_atom->encoder == encoder) {
+					if (outp_atom->set.mask)
+						atom->flush_disable = true;
+					break;
+				}
+			}
+		}
+	} else {
+		asyh->set.crc = false;
+		asyh->clr.crc = false;
+	}
+
+	return 0;
+}
+
+static enum nv50_crc_source_type
+nv50_crc_source_type(struct nouveau_encoder *outp,
+		     enum nv50_crc_source source)
+{
+	struct dcb_output *dcbe = outp->dcb;
+
+	switch (source) {
+	case NV50_CRC_SOURCE_NONE: return NV50_CRC_SOURCE_TYPE_NONE;
+	case NV50_CRC_SOURCE_RG:   return NV50_CRC_SOURCE_TYPE_RG;
+	default:		   break;
+	}
+
+	if (dcbe->location != DCB_LOC_ON_CHIP)
+		return NV50_CRC_SOURCE_TYPE_PIOR;
+
+	switch (dcbe->type) {
+	case DCB_OUTPUT_DP:	return NV50_CRC_SOURCE_TYPE_SF;
+	case DCB_OUTPUT_ANALOG:	return NV50_CRC_SOURCE_TYPE_DAC;
+	default:		return NV50_CRC_SOURCE_TYPE_SOR;
+	}
+}
+
+void nv50_crc_atomic_set(struct nv50_head *head,
+			 struct nv50_head_atom *asyh)
+{
+	struct drm_crtc *crtc = &head->base.base;
+	struct drm_device *dev = crtc->dev;
+	struct nv50_crc *crc = &head->crc;
+	const struct nv50_crc_func *func = nv50_disp(dev)->core->func->crc;
+	struct nouveau_encoder *outp +	
nv50_real_outp(nv50_head_atom_get_encoder(asyh));
+
+	func->set_src(head, outp->or,
+		      nv50_crc_source_type(outp, asyh->crc.src),
+		      &crc->ctx[crc->ctx_idx], asyh->crc.wndw);
+}
+
+void nv50_crc_atomic_clr(struct nv50_head *head)
+{
+	const struct nv50_crc_func *func +	
nv50_disp(head->base.base.dev)->core->func->crc;
+
+	func->set_src(head, 0, NV50_CRC_SOURCE_TYPE_NONE, NULL, 0);
+}
+
+#define NV50_CRC_RASTER_ACTIVE   0
+#define NV50_CRC_RASTER_COMPLETE 1
+#define NV50_CRC_RASTER_INACTIVE 2
+
+static inline int
+nv50_crc_raster_type(enum nv50_crc_source source)
+{
+	switch (source) {
+	case NV50_CRC_SOURCE_NONE:
+	case NV50_CRC_SOURCE_AUTO:
+	case NV50_CRC_SOURCE_RG:
+	case NV50_CRC_SOURCE_OUTP_ACTIVE:
+		return NV50_CRC_RASTER_ACTIVE;
+	case NV50_CRC_SOURCE_OUTP_COMPLETE:
+		return NV50_CRC_RASTER_COMPLETE;
+	case NV50_CRC_SOURCE_OUTP_INACTIVE:
+		return NV50_CRC_RASTER_INACTIVE;
+	}
+
+	return 0;
+}
+
+/* We handle mapping the memory for CRC notifiers ourselves, since each
+ * notifier needs it's own handle
+ */
+static inline int
+nv50_crc_ctx_init(struct nv50_head *head, struct nvif_mmu *mmu,
+		  struct nv50_crc_notifier_ctx *ctx, size_t len, int idx)
+{
+	struct nv50_core *core = nv50_disp(head->base.base.dev)->core;
+	int ret;
+
+	ret = nvif_mem_init_map(mmu, NVIF_MEM_VRAM, len, &ctx->mem);
+	if (ret)
+		return ret;
+
+	ret = nvif_object_init(&core->chan.base.user,
+			       NV50_DISP_HANDLE_CRC_CTX(head, idx),
+			       NV_DMA_IN_MEMORY,
+			       &(struct nv_dma_v0) {
+					.target = NV_DMA_V0_TARGET_VRAM,
+					.access = NV_DMA_V0_ACCESS_RDWR,
+					.start = ctx->mem.addr,
+					.limit =  ctx->mem.addr
+						+ ctx->mem.size - 1,
+			       }, sizeof(struct nv_dma_v0),
+			       &ctx->ntfy);
+	if (ret)
+		goto fail_fini;
+
+	return 0;
+
+fail_fini:
+	nvif_mem_fini(&ctx->mem);
+	return ret;
+}
+
+static inline void
+nv50_crc_ctx_fini(struct nv50_crc_notifier_ctx *ctx)
+{
+	nvif_object_fini(&ctx->ntfy);
+	nvif_mem_fini(&ctx->mem);
+}
+
+int nv50_crc_set_source(struct drm_crtc *crtc, const char *source_str)
+{
+	struct drm_device *dev = crtc->dev;
+	struct drm_atomic_state *state;
+	struct drm_modeset_acquire_ctx ctx;
+	struct nv50_head *head = nv50_head(crtc);
+	struct nv50_crc *crc = &head->crc;
+	const struct nv50_crc_func *func = nv50_disp(dev)->core->func->crc;
+	struct nvif_mmu *mmu = &nouveau_drm(dev)->client.mmu;
+	struct nv50_head_atom *asyh;
+	struct drm_crtc_state *crtc_state;
+	enum nv50_crc_source source;
+	int ret = 0, ctx_flags = 0, i;
+
+	ret = nv50_crc_parse_source(source_str, &source);
+	if (ret)
+		return ret;
+
+	/*
+	 * Since we don't want the user to accidentally interrupt us as we're
+	 * disabling CRCs
+	 */
+	if (source)
+		ctx_flags |= DRM_MODESET_ACQUIRE_INTERRUPTIBLE;
+	drm_modeset_acquire_init(&ctx, ctx_flags);
+
+	state = drm_atomic_state_alloc(dev);
+	if (!state) {
+		ret = -ENOMEM;
+		goto out_acquire_fini;
+	}
+	state->acquire_ctx = &ctx;
+
+	if (source) {
+		for (i = 0; i < ARRAY_SIZE(head->crc.ctx); i++) {
+			ret = nv50_crc_ctx_init(head, mmu, &crc->ctx[i],
+						func->notifier_len, i);
+			if (ret)
+				goto out_ctx_fini;
+		}
+	}
+
+retry:
+	crtc_state = drm_atomic_get_crtc_state(state, &head->base.base);
+	if (IS_ERR(crtc_state)) {
+		ret = PTR_ERR(crtc_state);
+		if (ret == -EDEADLK)
+			goto deadlock;
+		else if (ret)
+			goto out_drop_locks;
+	}
+	asyh = nv50_head_atom(crtc_state);
+	asyh->crc.src = source;
+	asyh->or.crc_raster = nv50_crc_raster_type(source);
+
+	ret = drm_atomic_commit(state);
+	if (ret == -EDEADLK)
+		goto deadlock;
+	else if (ret)
+		goto out_drop_locks;
+
+	if (!source) {
+		/*
+		 * If the user specified a custom flip threshold through
+		 * debugfs, reset it
+		 */
+		crc->flip_threshold = func->flip_threshold;
+	}
+
+out_drop_locks:
+	drm_modeset_drop_locks(&ctx);
+out_ctx_fini:
+	if (!source || ret) {
+		for (i = 0; i < ARRAY_SIZE(crc->ctx); i++)
+			nv50_crc_ctx_fini(&crc->ctx[i]);
+	}
+	drm_atomic_state_put(state);
+out_acquire_fini:
+	drm_modeset_acquire_fini(&ctx);
+	return ret;
+
+deadlock:
+	drm_atomic_state_clear(state);
+	drm_modeset_backoff(&ctx);
+	goto retry;
+}
+
+static int
+nv50_crc_debugfs_flip_threshold_get(struct seq_file *m, void *data)
+{
+	struct nv50_head *head = m->private;
+	struct drm_crtc *crtc = &head->base.base;
+	struct nv50_crc *crc = &head->crc;
+	int ret;
+
+	ret = drm_modeset_lock_single_interruptible(&crtc->mutex);
+	if (ret)
+		return ret;
+
+	seq_printf(m, "%d\n", crc->flip_threshold);
+
+	drm_modeset_unlock(&crtc->mutex);
+	return ret;
+}
+
+static int
+nv50_crc_debugfs_flip_threshold_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, nv50_crc_debugfs_flip_threshold_get,
+			   inode->i_private);
+}
+
+static ssize_t
+nv50_crc_debugfs_flip_threshold_set(struct file *file,
+				    const char __user *ubuf, size_t len,
+				    loff_t *offp)
+{
+	struct seq_file *m = file->private_data;
+	struct nv50_head *head = m->private;
+	struct nv50_head_atom *armh;
+	struct drm_crtc *crtc = &head->base.base;
+	struct nouveau_drm *drm = nouveau_drm(crtc->dev);
+	struct nv50_crc *crc = &head->crc;
+	const struct nv50_crc_func *func +	
nv50_disp(crtc->dev)->core->func->crc;
+	int value, ret;
+
+	ret = kstrtoint_from_user(ubuf, len, 10, &value);
+	if (ret)
+		return ret;
+
+	if (value > func->flip_threshold)
+		return -EINVAL;
+	else if (value == -1)
+		value = func->flip_threshold;
+	else if (value < -1)
+		return -EINVAL;
+
+	ret = drm_modeset_lock_single_interruptible(&crtc->mutex);
+	if (ret)
+		return ret;
+
+	armh = nv50_head_atom(crtc->state);
+	if (armh->crc.src) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	NV_DEBUG(drm,
+		 "Changing CRC flip threshold for next capture on head-%d to %d\n",
+		 head->base.index, value);
+	crc->flip_threshold = value;
+	ret = len;
+
+out:
+	drm_modeset_unlock(&crtc->mutex);
+	return ret;
+}
+
+static const struct file_operations nv50_crc_flip_threshold_fops = {
+	.owner = THIS_MODULE,
+	.open = nv50_crc_debugfs_flip_threshold_open,
+	.read = seq_read,
+	.write = nv50_crc_debugfs_flip_threshold_set,
+};
+
+int nv50_head_crc_late_register(struct nv50_head *head)
+{
+	struct drm_crtc *crtc = &head->base.base;
+	const struct nv50_crc_func *func +	
nv50_disp(crtc->dev)->core->func->crc;
+	struct dentry *root;
+
+	if (!func || !crtc->debugfs_entry)
+		return 0;
+
+	root = debugfs_create_dir("nv_crc", crtc->debugfs_entry);
+	debugfs_create_file("flip_threshold", 0644, root, head,
+			    &nv50_crc_flip_threshold_fops);
+
+	return 0;
+}
+
+static inline void
+nv50_crc_init_head(struct nv50_disp *disp, const struct nv50_crc_func *func,
+		   struct nv50_head *head)
+{
+	struct nv50_crc *crc = &head->crc;
+
+	crc->flip_threshold = func->flip_threshold;
+	spin_lock_init(&crc->lock);
+	drm_vblank_work_init(&crc->flip_work, &head->base.base,
+			     nv50_crc_ctx_flip_work);
+}
+
+void nv50_crc_init(struct drm_device *dev)
+{
+	struct nv50_disp *disp = nv50_disp(dev);
+	struct drm_crtc *crtc;
+	const struct nv50_crc_func *func = disp->core->func->crc;
+
+	if (!func)
+		return;
+
+	drm_for_each_crtc(crtc, dev)
+		nv50_crc_init_head(disp, func, nv50_head(crtc));
+}
diff --git a/drivers/gpu/drm/nouveau/dispnv50/crc.h
b/drivers/gpu/drm/nouveau/dispnv50/crc.h
new file mode 100644
index 000000000000..fce3e20043a4
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/dispnv50/crc.h
@@ -0,0 +1,125 @@
+// SPDX-License-Identifier: MIT
+#ifndef __NV50_CRC_H__
+#define __NV50_CRC_H__
+
+#include <linux/mutex.h>
+#include <drm/drm_crtc.h>
+#include <drm/drm_vblank.h>
+
+#include <nvif/mem.h>
+#include <nvkm/subdev/bios.h>
+#include "nouveau_encoder.h"
+
+struct nv50_disp;
+struct nv50_head;
+
+#if IS_ENABLED(CONFIG_DEBUG_FS)
+enum nv50_crc_source {
+	NV50_CRC_SOURCE_NONE = 0,
+	NV50_CRC_SOURCE_AUTO,
+	NV50_CRC_SOURCE_RG,
+	NV50_CRC_SOURCE_OUTP_ACTIVE,
+	NV50_CRC_SOURCE_OUTP_COMPLETE,
+	NV50_CRC_SOURCE_OUTP_INACTIVE,
+};
+
+/* RG -> SF (DP only)
+ *    -> SOR
+ *    -> PIOR
+ *    -> DAC
+ */
+enum nv50_crc_source_type {
+	NV50_CRC_SOURCE_TYPE_NONE = 0,
+	NV50_CRC_SOURCE_TYPE_SOR,
+	NV50_CRC_SOURCE_TYPE_PIOR,
+	NV50_CRC_SOURCE_TYPE_DAC,
+	NV50_CRC_SOURCE_TYPE_RG,
+	NV50_CRC_SOURCE_TYPE_SF,
+};
+
+struct nv50_crc_notifier_ctx {
+	struct nvif_mem mem;
+	struct nvif_object ntfy;
+};
+
+struct nv50_crc_atom {
+	enum nv50_crc_source src;
+	/* Only used for gv100+ */
+	u8 wndw : 4;
+};
+
+struct nv50_crc_func {
+	void (*set_src)(struct nv50_head *, int or, enum nv50_crc_source_type,
+			struct nv50_crc_notifier_ctx *, u32 wndw);
+	void (*set_ctx)(struct nv50_head *, struct nv50_crc_notifier_ctx *);
+	u32 (*get_entry)(struct nv50_head *, struct nv50_crc_notifier_ctx *,
+			 enum nv50_crc_source, int idx);
+	bool (*ctx_finished)(struct nv50_head *,
+			     struct nv50_crc_notifier_ctx *);
+	short flip_threshold;
+	short num_entries;
+	size_t notifier_len;
+};
+
+struct nv50_crc {
+	spinlock_t lock;
+	struct nv50_crc_notifier_ctx ctx[2];
+	struct drm_vblank_work flip_work;
+	enum nv50_crc_source src;
+
+	u64 frame;
+	short entry_idx;
+	short flip_threshold;
+	u8 ctx_idx : 1;
+	bool ctx_changed : 1;
+};
+
+void nv50_crc_init(struct drm_device *dev);
+int nv50_head_crc_late_register(struct nv50_head *);
+void nv50_crc_handle_vblank(struct nv50_head *head);
+
+int nv50_crc_verify_source(struct drm_crtc *, const char *, size_t *);
+const char *const *nv50_crc_get_sources(struct drm_crtc *, size_t *);
+int nv50_crc_set_source(struct drm_crtc *, const char *);
+
+int nv50_crc_atomic_check(struct nv50_head *, struct nv50_head_atom *,
+			  struct nv50_head_atom *);
+void nv50_crc_atomic_stop_reporting(struct drm_atomic_state *);
+void nv50_crc_atomic_prepare_notifier_contexts(struct drm_atomic_state *);
+void nv50_crc_atomic_start_reporting(struct drm_atomic_state *);
+void nv50_crc_atomic_set(struct nv50_head *, struct nv50_head_atom *);
+void nv50_crc_atomic_clr(struct nv50_head *);
+
+extern const struct nv50_crc_func crc907d;
+extern const struct nv50_crc_func crcc37d;
+
+#else /* IS_ENABLED(CONFIG_DEBUG_FS) */
+struct nv50_crc {};
+struct nv50_crc_func {};
+struct nv50_crc_atom {};
+
+#define nv50_crc_verify_source NULL
+#define nv50_crc_get_sources NULL
+#define nv50_crc_set_source NULL
+
+static inline void nv50_crc_init(struct drm_device *dev) {}
+static inline int nv50_head_crc_late_register(struct nv50_head *) {}
+static inline void
+nv50_crc_handle_vblank(struct nv50_head *head) { return 0; }
+
+static inline int
+nv50_crc_atomic_check(struct nv50_head *, struct nv50_head_atom *,
+		      struct nv50_head_atom *) {}
+static inline void
+nv50_crc_atomic_stop_reporting(struct drm_atomic_state *) {}
+static inline void
+nv50_crc_atomic_prepare_notifier_contexts(struct drm_atomic_state *) {}
+static inline void
+nv50_crc_atomic_start_reporting(struct drm_atomic_state *) {}
+static inline void
+nv50_crc_atomic_set(struct nv50_head *, struct nv50_head_atom *) {}
+static inline void
+nv50_crc_atomic_clr(struct nv50_head *) {}
+
+#endif /* IS_ENABLED(CONFIG_DEBUG_FS) */
+#endif /* !__NV50_CRC_H__ */
diff --git a/drivers/gpu/drm/nouveau/dispnv50/crc907d.c
b/drivers/gpu/drm/nouveau/dispnv50/crc907d.c
new file mode 100644
index 000000000000..92e907de7645
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/dispnv50/crc907d.c
@@ -0,0 +1,139 @@
+// SPDX-License-Identifier: MIT
+#include <drm/drm_crtc.h>
+
+#include "crc.h"
+#include "core.h"
+#include "disp.h"
+#include "head.h"
+
+#define CRC907D_MAX_ENTRIES 255
+
+struct crc907d_notifier {
+	u32 status;
+	u32 :32; /* reserved */
+	struct crc907d_entry {
+		u32 status;
+		u32 compositor_crc;
+		u32 output_crc[2];
+	} entries[CRC907D_MAX_ENTRIES];
+} __packed;
+
+static void
+crc907d_set_src(struct nv50_head *head, int or,
+		enum nv50_crc_source_type source,
+		struct nv50_crc_notifier_ctx *ctx, u32 wndw)
+{
+	struct drm_crtc *crtc = &head->base.base;
+	struct nv50_dmac *core =
&nv50_disp(head->base.base.dev)->core->chan;
+	const u32 hoff = head->base.index * 0x300;
+	u32 *push;
+	u32 crc_args = 0xfff00000;
+
+	switch (source) {
+	case NV50_CRC_SOURCE_TYPE_SOR:
+		crc_args |= (0x00000f0f + or * 16) << 8;
+		break;
+	case NV50_CRC_SOURCE_TYPE_PIOR:
+		crc_args |= (0x000000ff + or * 256) << 8;
+		break;
+	case NV50_CRC_SOURCE_TYPE_DAC:
+		crc_args |= (0x00000ff0 + or) << 8;
+		break;
+	case NV50_CRC_SOURCE_TYPE_RG:
+		crc_args |= (0x00000ff8 + drm_crtc_index(crtc)) << 8;
+		break;
+	case NV50_CRC_SOURCE_TYPE_SF:
+		crc_args |= (0x00000f8f + drm_crtc_index(crtc) * 16) << 8;
+		break;
+	case NV50_CRC_SOURCE_NONE:
+		crc_args |= 0x000fff00;
+		break;
+	}
+
+	push = evo_wait(core, 4);
+	if (!push)
+		return;
+
+	if (source) {
+		evo_mthd(push, 0x0438 + hoff, 1);
+		evo_data(push, ctx->ntfy.handle);
+		evo_mthd(push, 0x0430 + hoff, 1);
+		evo_data(push, crc_args);
+	} else {
+		evo_mthd(push, 0x0430 + hoff, 1);
+		evo_data(push, crc_args);
+		evo_mthd(push, 0x0438 + hoff, 1);
+		evo_data(push, 0);
+	}
+	evo_kick(push, core);
+}
+
+static void crc907d_set_ctx(struct nv50_head *head,
+			    struct nv50_crc_notifier_ctx *ctx)
+{
+	struct nv50_dmac *core =
&nv50_disp(head->base.base.dev)->core->chan;
+	u32 *push = evo_wait(core, 2);
+
+	if (!push)
+		return;
+
+	evo_mthd(push, 0x0438 + (head->base.index * 0x300), 1);
+	evo_data(push, ctx ? ctx->ntfy.handle : 0);
+	evo_kick(push, core);
+}
+
+static u32 crc907d_get_entry(struct nv50_head *head,
+			     struct nv50_crc_notifier_ctx *ctx,
+			     enum nv50_crc_source source, int idx)
+{
+	struct crc907d_notifier __iomem *notifier = ctx->mem.object.map.ptr;
+
+	return ioread32_native(&notifier->entries[idx].output_crc[0]);
+}
+
+static bool crc907d_ctx_finished(struct nv50_head *head,
+				 struct nv50_crc_notifier_ctx *ctx)
+{
+	struct nouveau_drm *drm = nouveau_drm(head->base.base.dev);
+	struct crc907d_notifier __iomem *notifier = ctx->mem.object.map.ptr;
+	const u32 status = ioread32_native(&notifier->status);
+	const u32 overflow = status & 0x0000003e;
+
+	if (!(status & 0x00000001))
+		return false;
+
+	if (overflow) {
+		const char *engine = NULL;
+
+		switch (overflow) {
+		case 0x00000004: engine = "DSI"; break;
+		case 0x00000008: engine = "Compositor"; break;
+		case 0x00000010: engine = "CRC output 1"; break;
+		case 0x00000020: engine = "CRC output 2"; break;
+		}
+
+		if (engine)
+			NV_ERROR(drm,
+				 "CRC notifier context for head %d overflowed on %s: %x\n",
+				 head->base.index, engine, status);
+		else
+			NV_ERROR(drm,
+				 "CRC notifier context for head %d overflowed: %x\n",
+				 head->base.index, status);
+	}
+
+	NV_DEBUG(drm, "Head %d CRC context status: %x\n",
+		 head->base.index, status);
+
+	return true;
+}
+
+const struct nv50_crc_func crc907d = {
+	.set_src = crc907d_set_src,
+	.set_ctx = crc907d_set_ctx,
+	.get_entry = crc907d_get_entry,
+	.ctx_finished = crc907d_ctx_finished,
+	.flip_threshold = CRC907D_MAX_ENTRIES - 10,
+	.num_entries = CRC907D_MAX_ENTRIES,
+	.notifier_len = sizeof(struct crc907d_notifier),
+};
diff --git a/drivers/gpu/drm/nouveau/dispnv50/crcc37d.c
b/drivers/gpu/drm/nouveau/dispnv50/crcc37d.c
new file mode 100644
index 000000000000..940cefd5517d
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/dispnv50/crcc37d.c
@@ -0,0 +1,153 @@
+// SPDX-License-Identifier: MIT
+#include <drm/drm_crtc.h>
+
+#include "crc.h"
+#include "core.h"
+#include "disp.h"
+#include "head.h"
+
+#define CRCC37D_MAX_ENTRIES 2047
+
+struct crcc37d_notifier {
+	u32 status;
+
+	/* reserved */
+	u32 :32;
+	u32 :32;
+	u32 :32;
+	u32 :32;
+	u32 :32;
+	u32 :32;
+	u32 :32;
+
+	struct crcc37d_entry {
+		u32 status[2];
+		u32 :32; /* reserved */
+		u32 compositor_crc;
+		u32 rg_crc;
+		u32 output_crc[2];
+		u32 :32; /* reserved */
+	} entries[CRCC37D_MAX_ENTRIES];
+} __packed;
+
+static void
+crcc37d_set_src(struct nv50_head *head, int or,
+		enum nv50_crc_source_type source,
+		struct nv50_crc_notifier_ctx *ctx, u32 wndw)
+{
+	struct nv50_dmac *core =
&nv50_disp(head->base.base.dev)->core->chan;
+	const u32 hoff = head->base.index * 0x400;
+	u32 *push;
+	u32 crc_args;
+
+	switch (source) {
+	case NV50_CRC_SOURCE_TYPE_SOR:
+		crc_args = (0x00000050 + or) << 12;
+		break;
+	case NV50_CRC_SOURCE_TYPE_PIOR:
+		crc_args = (0x00000060 + or) << 12;
+		break;
+	case NV50_CRC_SOURCE_TYPE_SF:
+		crc_args = 0x00000030 << 12;
+		break;
+	default:
+		crc_args = 0;
+		break;
+	}
+
+	push = evo_wait(core, 4);
+	if (!push)
+		return;
+
+	if (source) {
+		evo_mthd(push, 0x2180 + hoff, 1);
+		evo_data(push, ctx->ntfy.handle);
+		evo_mthd(push, 0x2184 + hoff, 1);
+		evo_data(push, crc_args | wndw);
+	} else {
+		evo_mthd(push, 0x2184 + hoff, 1);
+		evo_data(push, 0);
+		evo_mthd(push, 0x2180 + hoff, 1);
+		evo_data(push, 0);
+	}
+
+	evo_kick(push, core);
+}
+
+static void crcc37d_set_ctx(struct nv50_head *head,
+			    struct nv50_crc_notifier_ctx *ctx)
+{
+	struct nv50_dmac *core =
&nv50_disp(head->base.base.dev)->core->chan;
+	u32 *push = evo_wait(core, 2);
+
+	if (!push)
+		return;
+
+	evo_mthd(push, 0x2180 + (head->base.index * 0x400), 1);
+	evo_data(push, ctx ? ctx->ntfy.handle : 0);
+	evo_kick(push, core);
+}
+
+static u32 crcc37d_get_entry(struct nv50_head *head,
+			     struct nv50_crc_notifier_ctx *ctx,
+			     enum nv50_crc_source source, int idx)
+{
+	struct crcc37d_notifier __iomem *notifier = ctx->mem.object.map.ptr;
+	struct crcc37d_entry __iomem *entry = &notifier->entries[idx];
+	u32 __iomem *crc_addr;
+
+	if (source == NV50_CRC_SOURCE_RG)
+		crc_addr = &entry->rg_crc;
+	else
+		crc_addr = &entry->output_crc[0];
+
+	return ioread32_native(crc_addr);
+}
+
+static bool crcc37d_ctx_finished(struct nv50_head *head,
+				 struct nv50_crc_notifier_ctx *ctx)
+{
+	struct nouveau_drm *drm = nouveau_drm(head->base.base.dev);
+	struct crcc37d_notifier __iomem *notifier = ctx->mem.object.map.ptr;
+	const u32 status = ioread32_native(&notifier->status);
+	const u32 overflow = status & 0x0000007e;
+
+	if (!(status & 0x00000001))
+		return false;
+
+	if (overflow) {
+		const char *engine = NULL;
+
+		switch (overflow) {
+		case 0x00000004: engine = "Front End"; break;
+		case 0x00000008: engine = "Compositor"; break;
+		case 0x00000010: engine = "RG"; break;
+		case 0x00000020: engine = "CRC output 1"; break;
+		case 0x00000040: engine = "CRC output 2"; break;
+		}
+
+		if (engine)
+			NV_ERROR(drm,
+				 "CRC notifier context for head %d overflowed on %s: %x\n",
+				 head->base.index, engine, status);
+		else
+			NV_ERROR(drm,
+				 "CRC notifier context for head %d overflowed: %x\n",
+				 head->base.index, status);
+	}
+
+	NV_DEBUG(drm, "Head %d CRC context status: %x\n",
+		 head->base.index, status);
+
+	return true;
+}
+
+const struct nv50_crc_func crcc37d = {
+	.set_src = crcc37d_set_src,
+	.set_ctx = crcc37d_set_ctx,
+	.get_entry = crcc37d_get_entry,
+	.ctx_finished = crcc37d_ctx_finished,
+	.flip_threshold = CRCC37D_MAX_ENTRIES - 30,
+	.num_entries = CRCC37D_MAX_ENTRIES,
+	.notifier_len = sizeof(struct crcc37d_notifier),
+};
diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.c
b/drivers/gpu/drm/nouveau/dispnv50/disp.c
index 568029825f62..91836d723262 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/disp.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/disp.c
@@ -776,6 +776,19 @@ struct nv50_msto {
 	bool disabled;
 };
 
+struct nouveau_encoder *nv50_real_outp(struct drm_encoder *encoder)
+{
+	struct nv50_msto *msto;
+
+	if (encoder->encoder_type != DRM_MODE_ENCODER_DPMST)
+		return nouveau_encoder(encoder);
+
+	msto = nv50_msto(encoder);
+	if (!msto->mstc)
+		return NULL;
+	return msto->mstc->mstm->outp;
+}
+
 static struct drm_dp_payload *
 nv50_msto_payload(struct nv50_msto *msto)
 {
@@ -1899,6 +1912,7 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 	int i;
 
 	NV_ATOMIC(drm, "commit %d %d\n", atom->lock_core,
atom->flush_disable);
+	nv50_crc_atomic_stop_reporting(state);
 	drm_atomic_helper_wait_for_fences(dev, state, false);
 	drm_atomic_helper_wait_for_dependencies(state);
 	drm_atomic_helper_update_legacy_modeset_state(dev, state);
@@ -1908,6 +1922,7 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 
 	/* Disable head(s). */
 	for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i)
{
+		struct nv50_head_atom *armh = nv50_head_atom(old_crtc_state);
 		struct nv50_head_atom *asyh = nv50_head_atom(new_crtc_state);
 		struct nv50_head *head = nv50_head(crtc);
 
@@ -1920,7 +1935,8 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 		}
 
 		if (asyh->clr.mask) {
-			nv50_head_flush_clr(head, asyh, atom->flush_disable);
+			nv50_head_flush_clr(head, armh, asyh,
+					    atom->flush_disable);
 			interlock[NV50_DISP_INTERLOCK_CORE] |= 1;
 		}
 	}
@@ -1969,6 +1985,8 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 		}
 	}
 
+	nv50_crc_atomic_prepare_notifier_contexts(state);
+
 	/* Update output path(s). */
 	list_for_each_entry_safe(outp, outt, &atom->outp, head) {
 		const struct drm_encoder_helper_funcs *help;
@@ -1991,6 +2009,7 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 
 	/* Update head(s). */
 	for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i)
{
+		struct nv50_head_atom *armh = nv50_head_atom(old_crtc_state);
 		struct nv50_head_atom *asyh = nv50_head_atom(new_crtc_state);
 		struct nv50_head *head = nv50_head(crtc);
 
@@ -1998,7 +2017,7 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 			  asyh->set.mask, asyh->clr.mask);
 
 		if (asyh->set.mask) {
-			nv50_head_flush_set(head, asyh);
+			nv50_head_flush_set(head, armh, asyh);
 			interlock[NV50_DISP_INTERLOCK_CORE] = 1;
 		}
 
@@ -2082,6 +2101,7 @@ nv50_disp_atomic_commit_tail(struct drm_atomic_state
*state)
 		}
 	}
 
+	nv50_crc_atomic_start_reporting(state);
 	drm_atomic_helper_commit_hw_done(state);
 	drm_atomic_helper_cleanup_planes(dev, state);
 	drm_atomic_helper_commit_cleanup_done(state);
diff --git a/drivers/gpu/drm/nouveau/dispnv50/disp.h
b/drivers/gpu/drm/nouveau/dispnv50/disp.h
index 8935ebce8ab0..da72223277fe 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/disp.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/disp.h
@@ -1,10 +1,12 @@
 #ifndef __NV50_KMS_H__
 #define __NV50_KMS_H__
+#include <linux/workqueue.h>
 #include <nvif/mem.h>
 
 #include "nouveau_display.h"
 
 struct nv50_msto;
+struct nouveau_encoder;
 
 struct nv50_disp {
 	struct nvif_disp *disp;
@@ -89,6 +91,14 @@ int nv50_dmac_create(struct nvif_device *device, struct
nvif_object *disp,
 		     u64 syncbuf, struct nv50_dmac *dmac);
 void nv50_dmac_destroy(struct nv50_dmac *);
 
+/*
+ * For normal encoders this just returns the encoder. For active MST encoders,
+ * this returns the real outp that's driving displays on the topology.
+ * Inactive MST encoders return NULL, since they would have no real outp to
+ * return anyway.
+ */
+struct nouveau_encoder *nv50_real_outp(struct drm_encoder *encoder);
+
 u32 *evo_wait(struct nv50_dmac *, int nr);
 void evo_kick(u32 *, struct nv50_dmac *);
 
diff --git a/drivers/gpu/drm/nouveau/dispnv50/handles.h
b/drivers/gpu/drm/nouveau/dispnv50/handles.h
index d1beeb9a444d..27af7680294c 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/handles.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/handles.h
@@ -11,5 +11,6 @@
 #define NV50_DISP_HANDLE_VRAM                                          
0xf0000001
 
 #define NV50_DISP_HANDLE_WNDW_CTX(kind)                        (0xfb000000 |
kind)
+#define NV50_DISP_HANDLE_CRC_CTX(head, i) (0xfc000000 | head->base.index
<< 1 | i)
 
 #endif /* !__NV50_KMS_HANDLES_H__ */
diff --git a/drivers/gpu/drm/nouveau/dispnv50/head.c
b/drivers/gpu/drm/nouveau/dispnv50/head.c
index 72bc3bce396a..4668bcfb596c 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/head.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/head.c
@@ -24,27 +24,36 @@
 #include "core.h"
 #include "curs.h"
 #include "ovly.h"
+#include "crc.h"
 
 #include <nvif/class.h>
+#include <nvif/event.h>
+#include <nvif/cl0046.h>
 
 #include <drm/drm_atomic_helper.h>
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_vblank.h>
 #include "nouveau_connector.h"
+
 void
 nv50_head_flush_clr(struct nv50_head *head,
-		    struct nv50_head_atom *asyh, bool flush)
+		    struct nv50_head_atom *armh,
+		    struct nv50_head_atom *asyh,
+		    bool flush)
 {
 	union nv50_head_atom_mask clr = {
 		.mask = asyh->clr.mask & ~(flush ? 0 : asyh->set.mask),
 	};
+	if (clr.crc)  nv50_crc_atomic_clr(head);
 	if (clr.olut) head->func->olut_clr(head);
 	if (clr.core) head->func->core_clr(head);
 	if (clr.curs) head->func->curs_clr(head);
 }
 
 void
-nv50_head_flush_set(struct nv50_head *head, struct nv50_head_atom *asyh)
+nv50_head_flush_set(struct nv50_head *head,
+		    struct nv50_head_atom *armh,
+		    struct nv50_head_atom *asyh)
 {
 	if (asyh->set.view   ) head->func->view    (head, asyh);
 	if (asyh->set.mode   ) head->func->mode    (head, asyh);
@@ -61,6 +70,7 @@ nv50_head_flush_set(struct nv50_head *head, struct
nv50_head_atom *asyh)
 	if (asyh->set.ovly   ) head->func->ovly    (head, asyh);
 	if (asyh->set.dither ) head->func->dither  (head, asyh);
 	if (asyh->set.procamp) head->func->procamp (head, asyh);
+	if (asyh->set.crc    ) nv50_crc_atomic_set (head, asyh);
 	if (asyh->set.or     ) head->func->or      (head, asyh);
 }
 
@@ -313,7 +323,7 @@ nv50_head_atomic_check(struct drm_crtc *crtc, struct
drm_crtc_state *state)
 	struct nouveau_conn_atom *asyc = NULL;
 	struct drm_connector_state *conns;
 	struct drm_connector *conn;
-	int i;
+	int i, ret;
 
 	NV_ATOMIC(drm, "%s atomic_check %d\n", crtc->name,
asyh->state.active);
 	if (asyh->state.active) {
@@ -408,6 +418,10 @@ nv50_head_atomic_check(struct drm_crtc *crtc, struct
drm_crtc_state *state)
 		asyh->set.curs = asyh->curs.visible;
 	}
 
+	ret = nv50_crc_atomic_check(head, asyh, armh);
+	if (ret)
+		return ret;
+
 	if (asyh->clr.mask || asyh->set.mask)
 		nv50_atom(asyh->state.state)->lock_core = true;
 	return 0;
@@ -446,6 +460,7 @@ nv50_head_atomic_duplicate_state(struct drm_crtc *crtc)
 	asyh->ovly = armh->ovly;
 	asyh->dither = armh->dither;
 	asyh->procamp = armh->procamp;
+	asyh->crc = armh->crc;
 	asyh->or = armh->or;
 	asyh->dp = armh->dp;
 	asyh->clr.mask = 0;
@@ -467,10 +482,18 @@ nv50_head_reset(struct drm_crtc *crtc)
 	__drm_atomic_helper_crtc_reset(crtc, &asyh->state);
 }
 
+static int
+nv50_head_late_register(struct drm_crtc *crtc)
+{
+	return nv50_head_crc_late_register(nv50_head(crtc));
+}
+
 static void
 nv50_head_destroy(struct drm_crtc *crtc)
 {
 	struct nv50_head *head = nv50_head(crtc);
+
+	nvif_notify_fini(&head->base.vblank);
 	nv50_lut_fini(&head->olut);
 	drm_crtc_cleanup(crtc);
 	kfree(head);
@@ -488,8 +511,38 @@ nv50_head_func = {
 	.enable_vblank = nouveau_display_vblank_enable,
 	.disable_vblank = nouveau_display_vblank_disable,
 	.get_vblank_timestamp = drm_crtc_vblank_helper_get_vblank_timestamp,
+	.late_register = nv50_head_late_register,
+};
+
+static const struct drm_crtc_funcs
+nvd9_head_func = {
+	.reset = nv50_head_reset,
+	.gamma_set = drm_atomic_helper_legacy_gamma_set,
+	.destroy = nv50_head_destroy,
+	.set_config = drm_atomic_helper_set_config,
+	.page_flip = drm_atomic_helper_page_flip,
+	.atomic_duplicate_state = nv50_head_atomic_duplicate_state,
+	.atomic_destroy_state = nv50_head_atomic_destroy_state,
+	.enable_vblank = nouveau_display_vblank_enable,
+	.disable_vblank = nouveau_display_vblank_disable,
+	.get_vblank_timestamp = drm_crtc_vblank_helper_get_vblank_timestamp,
+	.verify_crc_source = nv50_crc_verify_source,
+	.get_crc_sources = nv50_crc_get_sources,
+	.set_crc_source = nv50_crc_set_source,
+	.late_register = nv50_head_late_register,
 };
 
+static int nv50_head_vblank_handler(struct nvif_notify *notify)
+{
+	struct nouveau_crtc *nv_crtc +		container_of(notify, struct nouveau_crtc,
vblank);
+
+	if (drm_crtc_handle_vblank(&nv_crtc->base))
+		nv50_crc_handle_vblank(nv50_head(&nv_crtc->base));
+
+	return NVIF_NOTIFY_KEEP;
+}
+
 struct nv50_head *
 nv50_head_create(struct drm_device *dev, int index)
 {
@@ -497,7 +550,9 @@ nv50_head_create(struct drm_device *dev, int index)
 	struct nv50_disp *disp = nv50_disp(dev);
 	struct nv50_head *head;
 	struct nv50_wndw *base, *ovly, *curs;
+	struct nouveau_crtc *nv_crtc;
 	struct drm_crtc *crtc;
+	const struct drm_crtc_funcs *funcs;
 	int ret;
 
 	head = kzalloc(sizeof(*head), GFP_KERNEL);
@@ -507,6 +562,11 @@ nv50_head_create(struct drm_device *dev, int index)
 	head->func = disp->core->func->head;
 	head->base.index = index;
 
+	if (disp->disp->object.oclass < GF110_DISP)
+		funcs = &nv50_head_func;
+	else
+		funcs = &nvd9_head_func;
+
 	if (disp->disp->object.oclass < GV100_DISP) {
 		ret = nv50_base_new(drm, head->base.index, &base);
 		if (ret)
@@ -531,9 +591,10 @@ nv50_head_create(struct drm_device *dev, int index)
 	if (ret)
 		goto fail_free;
 
-	crtc = &head->base.base;
+	nv_crtc = &head->base;
+	crtc = &nv_crtc->base;
 	drm_crtc_init_with_planes(dev, crtc, &base->plane, &curs->plane,
-				  &nv50_head_func, "head-%d", head->base.index);
+				  funcs, "head-%d", head->base.index);
 	drm_crtc_helper_add(crtc, &nv50_head_help);
 	/* Keep the legacy gamma size at 256 to avoid compatibility issues */
 	drm_mode_crtc_set_gamma_size(crtc, 256);
@@ -547,8 +608,22 @@ nv50_head_create(struct drm_device *dev, int index)
 			goto fail_crtc_cleanup;
 	}
 
+	ret = nvif_notify_init(&disp->disp->object,
nv50_head_vblank_handler,
+			       false, NV04_DISP_NTFY_VBLANK,
+			       &(struct nvif_notify_head_req_v0) {
+				    .head = nv_crtc->index,
+			       },
+			       sizeof(struct nvif_notify_head_req_v0),
+			       sizeof(struct nvif_notify_head_rep_v0),
+			       &nv_crtc->vblank);
+	if (ret)
+		goto fail_lut_fini;
+
 	return head;
 
+fail_lut_fini:
+	if (head->func->olut_set)
+		nv50_lut_fini(&head->olut);
 fail_crtc_cleanup:
 	drm_crtc_cleanup(crtc);
 fail_free:
diff --git a/drivers/gpu/drm/nouveau/dispnv50/head.h
b/drivers/gpu/drm/nouveau/dispnv50/head.h
index c05bbba9e247..2e57910ba450 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/head.h
+++ b/drivers/gpu/drm/nouveau/dispnv50/head.h
@@ -1,22 +1,29 @@
 #ifndef __NV50_KMS_HEAD_H__
 #define __NV50_KMS_HEAD_H__
 #define nv50_head(c) container_of((c), struct nv50_head, base.base)
+#include <linux/workqueue.h>
+
 #include "disp.h"
 #include "atom.h"
+#include "crc.h"
 #include "lut.h"
 
 #include "nouveau_crtc.h"
+#include "nouveau_encoder.h"
 
 struct nv50_head {
 	const struct nv50_head_func *func;
 	struct nouveau_crtc base;
+	struct nv50_crc crc;
 	struct nv50_lut olut;
 	struct nv50_msto *msto;
 };
 
 struct nv50_head *nv50_head_create(struct drm_device *, int index);
-void nv50_head_flush_set(struct nv50_head *, struct nv50_head_atom *);
-void nv50_head_flush_clr(struct nv50_head *, struct nv50_head_atom *, bool y);
+void nv50_head_flush_set(struct nv50_head *, struct nv50_head_atom *armh,
+			 struct nv50_head_atom *asyh);
+void nv50_head_flush_clr(struct nv50_head *, struct nv50_head_atom *armh,
+			 struct nv50_head_atom *asyh, bool y);
 
 struct nv50_head_func {
 	void (*view)(struct nv50_head *, struct nv50_head_atom *);
diff --git a/drivers/gpu/drm/nouveau/dispnv50/head907d.c
b/drivers/gpu/drm/nouveau/dispnv50/head907d.c
index 3002ec23d7a6..63a0b45d96d6 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/head907d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/head907d.c
@@ -19,8 +19,15 @@
  * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
  * OTHER DEALINGS IN THE SOFTWARE.
  */
+#include <drm/drm_connector.h>
+#include <drm/drm_mode_config.h>
+#include <drm/drm_vblank.h>
+#include "nouveau_drv.h"
+#include "nouveau_bios.h"
+#include "nouveau_connector.h"
 #include "head.h"
 #include "core.h"
+#include "crc.h"
 
 void
 head907d_or(struct nv50_head *head, struct nv50_head_atom *asyh)
@@ -29,9 +36,10 @@ head907d_or(struct nv50_head *head, struct nv50_head_atom
*asyh)
 	u32 *push;
 	if ((push = evo_wait(core, 3))) {
 		evo_mthd(push, 0x0404 + (head->base.index * 0x300), 2);
-		evo_data(push, 0x00000001 | asyh->or.depth  << 6 |
-					    asyh->or.nvsync << 4 |
-					    asyh->or.nhsync << 3);
+		evo_data(push, asyh->or.depth  << 6 |
+			       asyh->or.nvsync << 4 |
+			       asyh->or.nhsync << 3 |
+			       asyh->or.crc_raster);
 		evo_data(push, 0x31ec6000 | head->base.index << 25 |
 					    asyh->mode.interlace);
 		evo_kick(push, core);
diff --git a/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
b/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
index cf5a68f4021a..87dd11b2fe96 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/headc37d.c
@@ -46,10 +46,10 @@ headc37d_or(struct nv50_head *head, struct nv50_head_atom
*asyh)
 		}
 
 		evo_mthd(push, 0x2004 + (head->base.index * 0x400), 1);
-		evo_data(push, 0x00000001 |
-			       asyh->or.depth << 4 |
+		evo_data(push, depth << 4 |
 			       asyh->or.nvsync << 3 |
-			       asyh->or.nhsync << 2);
+			       asyh->or.nhsync << 2 |
+			       asyh->or.crc_raster);
 		evo_kick(push, core);
 	}
 }
diff --git a/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
b/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
index 65e3b60804c6..e6cef3593aec 100644
--- a/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
+++ b/drivers/gpu/drm/nouveau/dispnv50/headc57d.c
@@ -46,10 +46,11 @@ headc57d_or(struct nv50_head *head, struct nv50_head_atom
*asyh)
 		}
 
 		evo_mthd(push, 0x2004 + (head->base.index * 0x400), 1);
-		evo_data(push, 0xfc000001 |
-			       asyh->or.depth << 4 |
+		evo_data(push, 0xfc000000 |
+			       depth << 4 |
 			       asyh->or.nvsync << 3 |
-			       asyh->or.nhsync << 2);
+			       asyh->or.nhsync << 2 |
+			       asyh->or.crc_raster);
 		evo_kick(push, core);
 	}
 }
diff --git a/drivers/gpu/drm/nouveau/nouveau_display.c
b/drivers/gpu/drm/nouveau/nouveau_display.c
index 700817dc4fa0..7a0bd9a720a0 100644
--- a/drivers/gpu/drm/nouveau/nouveau_display.c
+++ b/drivers/gpu/drm/nouveau/nouveau_display.c
@@ -43,15 +43,7 @@
 #include <nvif/class.h>
 #include <nvif/cl0046.h>
 #include <nvif/event.h>
-
-static int
-nouveau_display_vblank_handler(struct nvif_notify *notify)
-{
-	struct nouveau_crtc *nv_crtc -		container_of(notify, typeof(*nv_crtc),
vblank);
-	drm_crtc_handle_vblank(&nv_crtc->base);
-	return NVIF_NOTIFY_KEEP;
-}
+#include <dispnv50/crc.h>
 
 int
 nouveau_display_vblank_enable(struct drm_crtc *crtc)
@@ -135,50 +127,6 @@ nouveau_display_scanoutpos(struct drm_crtc *crtc,
 					       stime, etime);
 }
 
-static void
-nouveau_display_vblank_fini(struct drm_device *dev)
-{
-	struct drm_crtc *crtc;
-
-	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
-		struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
-		nvif_notify_fini(&nv_crtc->vblank);
-	}
-}
-
-static int
-nouveau_display_vblank_init(struct drm_device *dev)
-{
-	struct nouveau_display *disp = nouveau_display(dev);
-	struct drm_crtc *crtc;
-	int ret;
-
-	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
-		struct nouveau_crtc *nv_crtc = nouveau_crtc(crtc);
-		ret = nvif_notify_init(&disp->disp.object,
-				       nouveau_display_vblank_handler, false,
-				       NV04_DISP_NTFY_VBLANK,
-				       &(struct nvif_notify_head_req_v0) {
-					.head = nv_crtc->index,
-				       },
-				       sizeof(struct nvif_notify_head_req_v0),
-				       sizeof(struct nvif_notify_head_rep_v0),
-				       &nv_crtc->vblank);
-		if (ret) {
-			nouveau_display_vblank_fini(dev);
-			return ret;
-		}
-	}
-
-	ret = drm_vblank_init(dev, dev->mode_config.num_crtc);
-	if (ret) {
-		nouveau_display_vblank_fini(dev);
-		return ret;
-	}
-
-	return 0;
-}
-
 static void
 nouveau_user_framebuffer_destroy(struct drm_framebuffer *drm_fb)
 {
@@ -545,9 +493,12 @@ nouveau_display_create(struct drm_device *dev)
 	drm_mode_config_reset(dev);
 
 	if (dev->mode_config.num_crtc) {
-		ret = nouveau_display_vblank_init(dev);
+		ret = drm_vblank_init(dev, dev->mode_config.num_crtc);
 		if (ret)
 			goto vblank_err;
+
+		if (disp->disp.object.oclass >= NV50_DISP)
+			nv50_crc_init(dev);
 	}
 
 	INIT_WORK(&drm->hpd_work, nouveau_display_hpd_work);
@@ -574,7 +525,6 @@ nouveau_display_destroy(struct drm_device *dev)
 #ifdef CONFIG_ACPI
 	unregister_acpi_notifier(&nouveau_drm(dev)->acpi_nb);
 #endif
-	nouveau_display_vblank_fini(dev);
 
 	drm_kms_helper_poll_fini(dev);
 	drm_mode_config_cleanup(dev);
-- 
2.25.4

Nvidia released some documentation on how CRC support works on their
GPUs, hooray!

So: this patch series implements said CRC support in nouveau, along with
adding some special debugfs interfaces for some relevant igt-gpu-tools
tests (already on the ML).

First - we add some new functionality to kthread_work in the kernel, and
then use this to add a new feature to DRM that Ville Syrj?l? came up
with: vblank workers. Basically, this is just a generic DRM interface
that allows for scheduling high-priority workers that start on a given
vblank interrupt. Note that while we're currently only using this in
nouveau, Intel has plans to use this for i915 as well (hence why they
came up with it!).

And finally: in order to implement the last feature, we expose some new
functions in the kernel's kthread_worker infrastructure so that we can
de-complicate our implementation of this.

Anyway-welcome to the future! :)

Major changes since v3:
* Style fixes on nouveau patches from checkpatch, no functional changes
* Don't integrate so tightly with kthread_work (and use our own lock),
  instead introduce some new functions for doing simple async flushing
  and cancelling. I think this interface looks a lot more acceptable
  then what I was previously trying.
* Apply some changes requested by danvet
Major changes since v2:
* Use kthread_worker instead of kthreadd for vblank workers
* Don't check debugfs return values

Lyude Paul (12):
  kthread: Add kthread_queue_flush_work()
  kthread: Add kthread_(un)block_work_queuing() and
    kthread_work_queuable()
  drm/vblank: Register drmm cleanup action once per drm_vblank_crtc
  drm/vblank: Add vblank works
  drm/nouveau/kms/nv50-: Unroll error cleanup in nv50_head_create()
  drm/nouveau/kms/nv140-: Don't modify depth in state during atomic
    commit
  drm/nouveau/kms/nv50-: Fix disabling dithering
  drm/nouveau/kms/nv50-: s/harm/armh/g
  drm/nouveau/kms/nv140-: Track wndw mappings in nv50_head_atom
  drm/nouveau/kms/nv50-: Expose nv50_outp_atom in disp.h
  drm/nouveau/kms/nv50-: Move hard-coded object handles into header
  drm/nouveau/kms/nvd9-: Add CRC support

 drivers/gpu/drm/drm_vblank.c                | 280 +++++++-
 drivers/gpu/drm/nouveau/dispnv04/crtc.c     |  25 +-
 drivers/gpu/drm/nouveau/dispnv50/Kbuild     |   4 +
 drivers/gpu/drm/nouveau/dispnv50/atom.h     |  21 +
 drivers/gpu/drm/nouveau/dispnv50/core.h     |   4 +
 drivers/gpu/drm/nouveau/dispnv50/core907d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/core917d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/corec37d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/corec57d.c |   3 +
 drivers/gpu/drm/nouveau/dispnv50/crc.c      | 715 ++++++++++++++++++++
 drivers/gpu/drm/nouveau/dispnv50/crc.h      | 125 ++++
 drivers/gpu/drm/nouveau/dispnv50/crc907d.c  | 139 ++++
 drivers/gpu/drm/nouveau/dispnv50/crcc37d.c  | 153 +++++
 drivers/gpu/drm/nouveau/dispnv50/disp.c     |  65 +-
 drivers/gpu/drm/nouveau/dispnv50/disp.h     |  24 +
 drivers/gpu/drm/nouveau/dispnv50/handles.h  |  16 +
 drivers/gpu/drm/nouveau/dispnv50/head.c     | 142 +++-
 drivers/gpu/drm/nouveau/dispnv50/head.h     |  13 +-
 drivers/gpu/drm/nouveau/dispnv50/head907d.c |  14 +-
 drivers/gpu/drm/nouveau/dispnv50/headc37d.c |  27 +-
 drivers/gpu/drm/nouveau/dispnv50/headc57d.c |  20 +-
 drivers/gpu/drm/nouveau/dispnv50/wndw.c     |  15 +-
 drivers/gpu/drm/nouveau/nouveau_display.c   |  60 +-
 include/drm/drm_vblank.h                    |  35 +
 include/linux/kthread.h                     |  35 +
 kernel/kthread.c                            | 133 +++-
 26 files changed, 1903 insertions(+), 174 deletions(-)
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crc.c
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crc.h
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crc907d.c
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/crcc37d.c
 create mode 100644 drivers/gpu/drm/nouveau/dispnv50/handles.h

-- 
2.25.4

On Fri, May 8, 2020 at 10:48 PM Lyude Paul <lyude at redhat.com>
wrote:
>
> Since we'll be allocating resources for kthread_create_worker() in the
> next commit (which could fail and require us to clean up the mess),
> let's simplify the cleanup process a bit by registering a
> drm_vblank_init_release() action for each drm_vblank_crtc so they're
> still cleaned up if we fail to initialize one of them.
>
> Changes since v3:
> * Use drmm_add_action_or_reset() - Daniel Vetter
>
> Cc: Daniel Vetter <daniel at ffwll.ch>
> Cc: Ville Syrj?l? <ville.syrjala at linux.intel.com>
> Cc: dri-devel at lists.freedesktop.org
> Cc: nouveau at lists.freedesktop.org
> Signed-off-by: Lyude Paul <lyude at redhat.com>
Reviewed-by: Daniel Vetter <daniel.vetter at ffwll.ch>

I think the other patches need a brain that's less in w/e mode :-)

Cheers, Daniel
> ---
>  drivers/gpu/drm/drm_vblank.c | 23 ++++++++++-------------
>  1 file changed, 10 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
> index 758bf74e1cab..a4a9013584e2 100644
> --- a/drivers/gpu/drm/drm_vblank.c
> +++ b/drivers/gpu/drm/drm_vblank.c
> @@ -491,16 +491,12 @@ static void vblank_disable_fn(struct timer_list *t)
>
>  static void drm_vblank_init_release(struct drm_device *dev, void *ptr)
>  {
> -       unsigned int pipe;
> -
> -       for (pipe = 0; pipe < dev->num_crtcs; pipe++) {
> -               struct drm_vblank_crtc *vblank = &dev->vblank[pipe];
> +       struct drm_vblank_crtc *vblank = ptr;
>
> -               WARN_ON(READ_ONCE(vblank->enabled) &&
> -                       drm_core_check_feature(dev, DRIVER_MODESET));
> +       WARN_ON(READ_ONCE(vblank->enabled) &&
> +               drm_core_check_feature(dev, DRIVER_MODESET));
>
> -               del_timer_sync(&vblank->disable_timer);
> -       }
> +       del_timer_sync(&vblank->disable_timer);
>  }
>
>  /**
> @@ -510,7 +506,7 @@ static void drm_vblank_init_release(struct drm_device
*dev, void *ptr)
>   *
>   * This function initializes vblank support for @num_crtcs display
pipelines.
>   * Cleanup is handled automatically through a cleanup function added with
> - * drmm_add_action().
> + * drmm_add_action_or_reset().
>   *
>   * Returns:
>   * Zero on success or a negative error code on failure.
> @@ -529,10 +525,6 @@ int drm_vblank_init(struct drm_device *dev, unsigned
int num_crtcs)
>
>         dev->num_crtcs = num_crtcs;
>
> -       ret = drmm_add_action(dev, drm_vblank_init_release, NULL);
> -       if (ret)
> -               return ret;
> -
>         for (i = 0; i < num_crtcs; i++) {
>                 struct drm_vblank_crtc *vblank = &dev->vblank[i];
>
> @@ -541,6 +533,11 @@ int drm_vblank_init(struct drm_device *dev, unsigned
int num_crtcs)
>                 init_waitqueue_head(&vblank->queue);
>                 timer_setup(&vblank->disable_timer,
vblank_disable_fn, 0);
>                 seqlock_init(&vblank->seqlock);
> +
> +               ret = drmm_add_action_or_reset(dev,
drm_vblank_init_release,
> +                                              vblank);
> +               if (ret)
> +                       return ret;
>         }
>
>         DRM_INFO("Supports vblank timestamp caching Rev 2
(21.10.2013).\n");
> --
> 2.25.4
>

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

Hello,

On Fri, May 08, 2020 at 04:46:51PM -0400, Lyude Paul
wrote:
> +bool kthread_queue_flush_work(struct kthread_work *work,
> +			      struct kthread_flush_work *fwork);
> +void __kthread_flush_work_fn(struct kthread_work *work);
As an exposed interface, this doesn't seem great. What the user wants to say
is "wait for the current instance of this guy" and the interface is
asking
them to queue an extra work item whose queueing return state should be
checked and depending on that result wait on its internal completion.

I'm skeptical this is a good idea in general given that unless you define
"this instance" at the time of queueing the work item which is being
waited-upon, there's no way to guarantee that the instance you're
queueing
the flush work item on is the instance you want unless the queuer is holding
external synchronization which prevents the instance from running. That's a
really confusing semantics to expose in the interface.

What the above means is that the ordering that you want is only defined
through your own locking and that maybe suggests that the sequencing should
be implemented on that side too. It may be a bit more code but a sequence
counter + wait queue might be the better solution here.

Thanks.

-- 
tejun

On Fri, May 08, 2020 at 04:46:52PM -0400, Lyude Paul
wrote:
> Add some simple wrappers around incrementing/decrementing
> kthread_work.cancelling under lock, along with checking whether queuing
> is currently allowed on a given kthread_work, which we'll use want to
> implement work cancelling with DRM's vblank work helpers.
Am I correct in assuming that what you want is "cancel this and block
further queueing until the state is cleared"? I agree that'd be
something
really useful to have. That said, There are a few areas that I think can be
improved upon:

* It'd be better if we separate this state into its own thing rather than
  mixing with canceling state which has potential to make things really
  confusing. It doesn't have to be a separate field unless you want disable
  depth for work item disable (and I don't think you do). It can just be a
  high bit in the same field but I think the two states should be separate
  one way or the other.

* I'm always a bit skeptical about state querying interfaces which
aren't
  synchronized to anything. They're useful in many cases but also prone to
  being misused. If you absoultely have to have them, can you please add
  explicit comment explaining the lack of synchronization around it - ie.
  unless you're the one setting and clearing the flag and queueing the task,
  it isn't synchronized against anything.

* In the same vein, I'm not too sure about stand-alone block interface.
  Unless I'm the sole queuer or there are further locking around queueing,
  what good does setting blocking do? There's no way to guarantee that the
  flag is seen by someone else trying to queue it and trying to flush the
  work item after queueing doesn't help either. The only way to make that
  interface meaningful is doing it together with cancel - so, you say
"block
  further queueing and cancel / flush whatever is in flight or queued",
  which actually gives you a useful invariant.

* A simliar argument can be made about unblock too although that's an a lot
  more relaxed situation in that unblocking and queueing oneself always
  works and that the user might not care which future instance of queueing
  will start succeeding.

Thanks.

-- 
tejun

Minor nitpick.? s/queuing/queueing/ and s/queuable/queueable/.

On 2020-05-08 15:46, Lyude Paul wrote:
> Add some simple wrappers around incrementing/decrementing
> kthread_work.cancelling under lock, along with checking whether queuing
> is currently allowed on a given kthread_work, which we'll use want to
> implement work cancelling with DRM's vblank work helpers.
>
> Cc: Daniel Vetter <daniel at ffwll.ch>
> Cc: Tejun Heo <tj at kernel.org>
> Cc: Ville Syrj?l? <ville.syrjala at linux.intel.com>
> Cc: dri-devel at lists.freedesktop.org
> Cc: nouveau at lists.freedesktop.org
> Signed-off-by: Lyude Paul <lyude at redhat.com>
> ---
>   include/linux/kthread.h | 19 +++++++++++++++++
>   kernel/kthread.c        | 46 +++++++++++++++++++++++++++++++++++++++++
>   2 files changed, 65 insertions(+)
>
> diff --git a/include/linux/kthread.h b/include/linux/kthread.h
> index 0006540ce7f9..c6fee200fced 100644
> --- a/include/linux/kthread.h
> +++ b/include/linux/kthread.h
> @@ -211,9 +211,28 @@ void kthread_flush_worker(struct kthread_worker
*worker);
>   
>   bool kthread_cancel_work_sync(struct kthread_work *work);
>   bool kthread_cancel_delayed_work_sync(struct kthread_delayed_work *work);
> +void kthread_block_work_queuing(struct kthread_worker *worker,
> +				struct kthread_work *work);
> +void kthread_unblock_work_queuing(struct kthread_worker *worker,
> +				  struct kthread_work *work);
>   
>   void kthread_destroy_worker(struct kthread_worker *worker);
>   
> +/**
> + * kthread_work_queuable - whether or not a kthread work can be queued
> + * @work: The kthread work to check
> + *
> + * Checks whether or not queuing @work is currently blocked from queuing,
> + * either by kthread_cancel_work_sync() and friends or
> + * kthread_block_work_queuing().
> + *
> + * Returns: whether or not the @work may be queued.
> + */
> +static inline bool kthread_work_queuable(struct kthread_work *work)
> +{
> +	return READ_ONCE(work->canceling) == 0;
> +}
> +
>   struct cgroup_subsys_state;
>   
>   #ifdef CONFIG_BLK_CGROUP
> diff --git a/kernel/kthread.c b/kernel/kthread.c
> index c1f8ec9d5836..f8a5c5a87cc6 100644
> --- a/kernel/kthread.c
> +++ b/kernel/kthread.c
> @@ -1187,6 +1187,52 @@ bool kthread_cancel_delayed_work_sync(struct
kthread_delayed_work *dwork)
>   }
>   EXPORT_SYMBOL_GPL(kthread_cancel_delayed_work_sync);
>   
> +/**
> + * kthread_block_work_queuing - prevent a kthread_work from being queued
> + * without actually cancelling it
> + * @worker: kthread worker to use
> + * @work: work to block queuing on
> + *
> + * Prevents @work from being queued using kthread_queue_work() and
friends,
> + * but doesn't attempt to cancel any previous queuing. The caller must
unblock
> + * queuing later by calling kthread_unblock_work_queuing(). This call can
be
> + * called multiple times.
> + *
> + * See also: kthread_work_queuable()
> + */
> +void kthread_block_work_queuing(struct kthread_worker *worker,
> +				struct kthread_work *work)
> +{
> +	unsigned long flags;
> +
> +	raw_spin_lock_irqsave(&worker->lock, flags);
> +	work->canceling++;
> +	raw_spin_unlock_irqrestore(&worker->lock, flags);
> +}
> +EXPORT_SYMBOL_GPL(kthread_block_work_queuing);
> +
> +/**
> + * kthread_unblock_work_queuing - unblock queuing on a kthread_work
> + * @worker: kthread worker to use
> + * @work: work to unblock queuing on
> + *
> + * Removes a request to prevent @work from being queued with
> + * kthread_queue_work() and friends, so that it may potentially be queued
> + * again.
> + *
> + * See also: kthread_work_queuable()
> + */
> +void kthread_unblock_work_queuing(struct kthread_worker *worker,
> +				  struct kthread_work *work)
> +{
> +	unsigned long flags;
> +
> +	raw_spin_lock_irqsave(&worker->lock, flags);
> +	WARN_ON_ONCE(--work->canceling < 0);
> +	raw_spin_unlock_irqrestore(&worker->lock, flags);
> +}
> +EXPORT_SYMBOL_GPL(kthread_unblock_work_queuing);
> +
>   /**
>    * kthread_flush_worker - flush all current works on a kthread_worker
>    * @worker: worker to flush