Patch "dma-buf: add dma_fence_timestamp helper" has been added to the 6.5-stable tree

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



This is a note to let you know that I've just added the patch titled

    dma-buf: add dma_fence_timestamp helper

to the 6.5-stable tree which can be found at:
    http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
     dma-buf-add-dma_fence_timestamp-helper.patch
and it can be found in the queue-6.5 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@xxxxxxxxxxxxxxx> know about it.


>From b83ce9cb4a465b8f9a3fa45561b721a9551f60e3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Christian=20K=C3=B6nig?= <christian.koenig@xxxxxxx>
Date: Fri, 8 Sep 2023 10:27:23 +0200
Subject: dma-buf: add dma_fence_timestamp helper
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

From: Christian König <christian.koenig@xxxxxxx>

commit b83ce9cb4a465b8f9a3fa45561b721a9551f60e3 upstream.

When a fence signals there is a very small race window where the timestamp
isn't updated yet. sync_file solves this by busy waiting for the
timestamp to appear, but on other ocassions didn't handled this
correctly.

Provide a dma_fence_timestamp() helper function for this and use it in
all appropriate cases.

Another alternative would be to grab the spinlock when that happens.

v2 by teddy: add a wait parameter to wait for the timestamp to show up, in case
   the accurate timestamp is needed and/or the timestamp is not based on
   ktime (e.g. hw timestamp)
v3 chk: drop the parameter again for unified handling

Signed-off-by: Yunxiang Li <Yunxiang.Li@xxxxxxx>
Signed-off-by: Christian König <christian.koenig@xxxxxxx>
Fixes: 1774baa64f93 ("drm/scheduler: Change scheduled fence track v2")
Reviewed-by: Alex Deucher <alexander.deucher@xxxxxxx>
CC: stable@xxxxxxxxxxxxxxx
Link: https://patchwork.freedesktop.org/patch/msgid/20230929104725.2358-1-christian.koenig@xxxxxxx
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
---
 drivers/dma-buf/dma-fence-unwrap.c     |   13 ++++---------
 drivers/dma-buf/sync_file.c            |    9 +++------
 drivers/gpu/drm/scheduler/sched_main.c |    2 +-
 include/linux/dma-fence.h              |   19 +++++++++++++++++++
 4 files changed, 27 insertions(+), 16 deletions(-)

--- a/drivers/dma-buf/dma-fence-unwrap.c
+++ b/drivers/dma-buf/dma-fence-unwrap.c
@@ -76,16 +76,11 @@ struct dma_fence *__dma_fence_unwrap_mer
 		dma_fence_unwrap_for_each(tmp, &iter[i], fences[i]) {
 			if (!dma_fence_is_signaled(tmp)) {
 				++count;
-			} else if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT,
-					    &tmp->flags)) {
-				if (ktime_after(tmp->timestamp, timestamp))
-					timestamp = tmp->timestamp;
 			} else {
-				/*
-				 * Use the current time if the fence is
-				 * currently signaling.
-				 */
-				timestamp = ktime_get();
+				ktime_t t = dma_fence_timestamp(tmp);
+
+				if (ktime_after(t, timestamp))
+					timestamp = t;
 			}
 		}
 	}
--- a/drivers/dma-buf/sync_file.c
+++ b/drivers/dma-buf/sync_file.c
@@ -268,13 +268,10 @@ static int sync_fill_fence_info(struct d
 		sizeof(info->driver_name));
 
 	info->status = dma_fence_get_status(fence);
-	while (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags) &&
-	       !test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags))
-		cpu_relax();
 	info->timestamp_ns =
-		test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags) ?
-		ktime_to_ns(fence->timestamp) :
-		ktime_set(0, 0);
+		dma_fence_is_signaled(fence) ?
+			ktime_to_ns(dma_fence_timestamp(fence)) :
+			ktime_set(0, 0);
 
 	return info->status;
 }
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -929,7 +929,7 @@ drm_sched_get_cleanup_job(struct drm_gpu
 
 		if (next) {
 			next->s_fence->scheduled.timestamp =
-				job->s_fence->finished.timestamp;
+				dma_fence_timestamp(&job->s_fence->finished);
 			/* start TO timer for next job */
 			drm_sched_start_timeout(sched);
 		}
--- a/include/linux/dma-fence.h
+++ b/include/linux/dma-fence.h
@@ -568,6 +568,25 @@ static inline void dma_fence_set_error(s
 	fence->error = error;
 }
 
+/**
+ * dma_fence_timestamp - helper to get the completion timestamp of a fence
+ * @fence: fence to get the timestamp from.
+ *
+ * After a fence is signaled the timestamp is updated with the signaling time,
+ * but setting the timestamp can race with tasks waiting for the signaling. This
+ * helper busy waits for the correct timestamp to appear.
+ */
+static inline ktime_t dma_fence_timestamp(struct dma_fence *fence)
+{
+	if (WARN_ON(!test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)))
+		return ktime_get();
+
+	while (!test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags))
+		cpu_relax();
+
+	return fence->timestamp;
+}
+
 signed long dma_fence_wait_timeout(struct dma_fence *,
 				   bool intr, signed long timeout);
 signed long dma_fence_wait_any_timeout(struct dma_fence **fences,


Patches currently in stable-queue which might be from christian.koenig@xxxxxxx are

queue-6.5/drm-amdgpu-add-missing-null-check.patch
queue-6.5/dma-buf-add-dma_fence_timestamp-helper.patch



[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux