Re: [Linaro-mm-sig] [PATCH v2] dma-buf/sw_sync: Avoid recursive lock during fence signal

2023-08-27 Thread Christian König 

Am 22.08.23 um 19:15 schrieb Rob Clark:

On Tue, Aug 22, 2023 at 6:01 AM Christian König
 wrote:

Am 18.08.23 um 16:59 schrieb Rob Clark:

From: Rob Clark 

If a signal callback releases the sw_sync fence, that will trigger a
deadlock as the timeline_fence_release recurses onto the fence->lock
(used both for signaling and the the timeline tree).

To avoid that, temporarily hold an extra reference to the signalled
fences until after we drop the lock.

(This is an alternative implementation of 
https://patchwork.kernel.org/patch/11664717/
which avoids some potential UAF issues with the original patch.)

v2: Remove now obsolete comment, use list_move_tail() and
  list_del_init()

Reported-by: Bas Nieuwenhuizen 
Fixes: d3c6dd1fb30d ("dma-buf/sw_sync: Synchronize signal vs syncpt free")
Signed-off-by: Rob Clark 

Reviewed-by: Christian König 

Thanks, any chance you could take this via drm-misc?


I've already pushed this quite a while ago.

At the moment I have problem answering because AMD has a new security 
policy which makes it impossible to push patches and access mails at the 
same time.


We are working with our IT to get this fixed, but at the moment its 
eating my time.


Sorry for the delay,
Christian.



BR,
-R


---
   drivers/dma-buf/sw_sync.c | 18 +-
   1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
index 63f0aeb66db6..f0a35277fd84 100644
--- a/drivers/dma-buf/sw_sync.c
+++ b/drivers/dma-buf/sw_sync.c
@@ -191,6 +191,7 @@ static const struct dma_fence_ops timeline_fence_ops = {
*/
   static void sync_timeline_signal(struct sync_timeline *obj, unsigned int inc)
   {
+ LIST_HEAD(signalled);
   struct sync_pt *pt, *next;

   trace_sync_timeline(obj);
@@ -203,21 +204,20 @@ static void sync_timeline_signal(struct sync_timeline 
*obj, unsigned int inc)
   if (!timeline_fence_signaled(>base))
   break;

- list_del_init(>link);
+ dma_fence_get(>base);
+
+ list_move_tail(>link, );
   rb_erase(>node, >pt_tree);

- /*
-  * A signal callback may release the last reference to this
-  * fence, causing it to be freed. That operation has to be
-  * last to avoid a use after free inside this loop, and must
-  * be after we remove the fence from the timeline in order to
-  * prevent deadlocking on timeline->lock inside
-  * timeline_fence_release().
-  */
   dma_fence_signal_locked(>base);
   }

   spin_unlock_irq(>lock);
+
+ list_for_each_entry_safe(pt, next, , link) {
+ list_del_init(>link);
+ dma_fence_put(>base);
+ }
   }

   /**

Re: [Linaro-mm-sig] [PATCH v2] dma-buf/sw_sync: Avoid recursive lock during fence signal

2023-08-22 Thread Rob Clark 
On Tue, Aug 22, 2023 at 6:01 AM Christian König
 wrote:
>
> Am 18.08.23 um 16:59 schrieb Rob Clark:
> > From: Rob Clark 
> >
> > If a signal callback releases the sw_sync fence, that will trigger a
> > deadlock as the timeline_fence_release recurses onto the fence->lock
> > (used both for signaling and the the timeline tree).
> >
> > To avoid that, temporarily hold an extra reference to the signalled
> > fences until after we drop the lock.
> >
> > (This is an alternative implementation of 
> > https://patchwork.kernel.org/patch/11664717/
> > which avoids some potential UAF issues with the original patch.)
> >
> > v2: Remove now obsolete comment, use list_move_tail() and
> >  list_del_init()
> >
> > Reported-by: Bas Nieuwenhuizen 
> > Fixes: d3c6dd1fb30d ("dma-buf/sw_sync: Synchronize signal vs syncpt free")
> > Signed-off-by: Rob Clark 
>
> Reviewed-by: Christian König 

Thanks, any chance you could take this via drm-misc?

BR,
-R

>
> > ---
> >   drivers/dma-buf/sw_sync.c | 18 +-
> >   1 file changed, 9 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
> > index 63f0aeb66db6..f0a35277fd84 100644
> > --- a/drivers/dma-buf/sw_sync.c
> > +++ b/drivers/dma-buf/sw_sync.c
> > @@ -191,6 +191,7 @@ static const struct dma_fence_ops timeline_fence_ops = {
> >*/
> >   static void sync_timeline_signal(struct sync_timeline *obj, unsigned int 
> > inc)
> >   {
> > + LIST_HEAD(signalled);
> >   struct sync_pt *pt, *next;
> >
> >   trace_sync_timeline(obj);
> > @@ -203,21 +204,20 @@ static void sync_timeline_signal(struct sync_timeline 
> > *obj, unsigned int inc)
> >   if (!timeline_fence_signaled(>base))
> >   break;
> >
> > - list_del_init(>link);
> > + dma_fence_get(>base);
> > +
> > + list_move_tail(>link, );
> >   rb_erase(>node, >pt_tree);
> >
> > - /*
> > -  * A signal callback may release the last reference to this
> > -  * fence, causing it to be freed. That operation has to be
> > -  * last to avoid a use after free inside this loop, and must
> > -  * be after we remove the fence from the timeline in order to
> > -  * prevent deadlocking on timeline->lock inside
> > -  * timeline_fence_release().
> > -  */
> >   dma_fence_signal_locked(>base);
> >   }
> >
> >   spin_unlock_irq(>lock);
> > +
> > + list_for_each_entry_safe(pt, next, , link) {
> > + list_del_init(>link);
> > + dma_fence_put(>base);
> > + }
> >   }
> >
> >   /**
>

Re: [Linaro-mm-sig] [PATCH v2] dma-buf/sw_sync: Avoid recursive lock during fence signal

2023-08-22 Thread Christian König 

Am 18.08.23 um 16:59 schrieb Rob Clark:

From: Rob Clark 

If a signal callback releases the sw_sync fence, that will trigger a
deadlock as the timeline_fence_release recurses onto the fence->lock
(used both for signaling and the the timeline tree).

To avoid that, temporarily hold an extra reference to the signalled
fences until after we drop the lock.

(This is an alternative implementation of 
https://patchwork.kernel.org/patch/11664717/
which avoids some potential UAF issues with the original patch.)

v2: Remove now obsolete comment, use list_move_tail() and
 list_del_init()

Reported-by: Bas Nieuwenhuizen 
Fixes: d3c6dd1fb30d ("dma-buf/sw_sync: Synchronize signal vs syncpt free")
Signed-off-by: Rob Clark 


Reviewed-by: Christian König 


---
  drivers/dma-buf/sw_sync.c | 18 +-
  1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
index 63f0aeb66db6..f0a35277fd84 100644
--- a/drivers/dma-buf/sw_sync.c
+++ b/drivers/dma-buf/sw_sync.c
@@ -191,6 +191,7 @@ static const struct dma_fence_ops timeline_fence_ops = {
   */
  static void sync_timeline_signal(struct sync_timeline *obj, unsigned int inc)
  {
+   LIST_HEAD(signalled);
struct sync_pt *pt, *next;
  
  	trace_sync_timeline(obj);

@@ -203,21 +204,20 @@ static void sync_timeline_signal(struct sync_timeline 
*obj, unsigned int inc)
if (!timeline_fence_signaled(>base))
break;
  
-		list_del_init(>link);

+   dma_fence_get(>base);
+
+   list_move_tail(>link, );
rb_erase(>node, >pt_tree);
  
-		/*

-* A signal callback may release the last reference to this
-* fence, causing it to be freed. That operation has to be
-* last to avoid a use after free inside this loop, and must
-* be after we remove the fence from the timeline in order to
-* prevent deadlocking on timeline->lock inside
-* timeline_fence_release().
-*/
dma_fence_signal_locked(>base);
}
  
  	spin_unlock_irq(>lock);

+
+   list_for_each_entry_safe(pt, next, , link) {
+   list_del_init(>link);
+   dma_fence_put(>base);
+   }
  }
  
  /**