[PATCH] dma-buf: fix debugfs versus rcu and fence dumping v2

[j***@redhat.com]

From: Jérôme Glisse <***@redhat.com>

The debugfs take reference on fence without dropping them. Also the
rcu section are not well balance. Fix all that ...

Changed since v1:
- moved fobj logic around to be rcu safe

Signed-off-by: Jérôme Glisse <***@redhat.com>
Cc: Christian König <***@amd.com>
Cc: Daniel Vetter <***@ffwll.ch>
Cc: Sumit Semwal <***@linaro.org>
Cc: linux-***@vger.kernel.org
Cc: dri-***@lists.freedesktop.org
Cc: linaro-mm-***@lists.linaro.org
Cc: Stéphane Marchesin <***@chromium.org>
Cc: ***@vger.kernel.org
---
drivers/dma-buf/dma-buf.c | 21 ++++++++++++++++-----
1 file changed, 16 insertions(+), 5 deletions(-)

diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c
index 13884474d158..9688d99894d6 100644
--- a/drivers/dma-buf/dma-buf.c
+++ b/drivers/dma-buf/dma-buf.c
@@ -1048,27 +1048,38 @@ static int dma_buf_debug_show(struct seq_file *s, void *unused)
while (true) {
seq = read_seqcount_begin(&robj->seq);
rcu_read_lock();
- fobj = rcu_dereference(robj->fence);
- shared_count = fobj ? fobj->shared_count : 0;
fence = rcu_dereference(robj->fence_excl);
+ fence = dma_fence_get_rcu(fence);
if (!read_seqcount_retry(&robj->seq, seq))
break;
rcu_read_unlock();
}
-
- if (fence)
+ if (fence) {
seq_printf(s, "\tExclusive fence: %s %s %ssignalled\n",
fence->ops->get_driver_name(fence),
fence->ops->get_timeline_name(fence),
dma_fence_is_signaled(fence) ? "" : "un");
- for (i = 0; i < shared_count; i++) {
+ dma_fence_put(fence);
+ }
+
+ rcu_read_lock();
+ fobj = rcu_dereference(robj->fence);
+ shared_count = fobj ? fobj->shared_count : 0;
+ for (i = 0, fence = NULL; i < shared_count; i++) {
fence = rcu_dereference(fobj->shared[i]);
if (!dma_fence_get_rcu(fence))
continue;
+ rcu_read_unlock();
+
seq_printf(s, "\tShared fence: %s %s %ssignalled\n",
fence->ops->get_driver_name(fence),
fence->ops->get_timeline_name(fence),
dma_fence_is_signaled(fence) ? "" : "un");
+ dma_fence_put(fence);
+
+ rcu_read_lock();
+ fobj = rcu_dereference(robj->fence);
+ shared_count = fobj ? fobj->shared_count : 0;
}
rcu_read_unlock();

[Chris Wilson]

Post by j***@redhat.com
The debugfs take reference on fence without dropping them. Also the
rcu section are not well balance. Fix all that ...

Wouldn't the code be a lot simpler (and a consistent snapshot) if it used
reservation_object_get_fences_rcu()?
-Chris

[Koenig, Christian]

Post by Chris Wilson

Post by j***@redhat.com
The debugfs take reference on fence without dropping them. Also the
rcu section are not well balance. Fix all that ...

Wouldn't the code be a lot simpler (and a consistent snapshot) if it used
reservation_object_get_fences_rcu()?

Yeah, thought about that as well.

Or even better move that code into reservation_object.c as
reservation_object_show_fences() or something like that.

Christian.

Post by Chris Wilson
-Chris

[Sasha Levin]

Hi,

[This is an automated email]

This commit has been processed because it contains a -stable tag.
The stable tag indicates that it's relevant for the following trees: all

The bot has tested the following trees: v4.19.7, v4.14.86, v4.9.143, v4.4.166, v3.18.128,

v4.19.7: Build OK!
v4.14.86: Build OK!
v4.9.143: Failed to apply! Possible dependencies:
5eb2c72c8acc ("dma-buf: fence debugging")

v4.4.166: Failed to apply! Possible dependencies:
5eb2c72c8acc ("dma-buf: fence debugging")

v3.18.128: Failed to apply! Possible dependencies:
5eb2c72c8acc ("dma-buf: fence debugging")


How should we proceed with this patch?

--
Thanks,
Sasha