diff mbox

[v2,2/2] dma-buf/sw_sync: force signal all unsignaled fences on dying timeline

Message ID 20170907190246.16425-2-gustavo@padovan.org (mailing list archive)
State New, archived
Headers show

Commit Message

Gustavo Padovan Sept. 7, 2017, 7:02 p.m. UTC 
From: Dominik Behr <dbehr@chromium.org>

To avoid hanging userspace components that might have been waiting on the
active fences of the destroyed timeline we need to signal with error all
remaining fences on such timeline.

This restore the default behaviour of the Android sw_sync framework, which
Android still relies on. It was broken on the dma fence conversion a few
years ago and never fixed.

v2: Do not bother with cleanup do the list (Chris Wilson)

Signed-off-by: Dominik Behr <dbehr@chromium.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
---
 drivers/dma-buf/sw_sync.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

Comments

Chris Wilson Sept. 7, 2017, 7:11 p.m. UTC | #1 
Quoting Gustavo Padovan (2017-09-07 20:02:46)
> From: Dominik Behr <dbehr@chromium.org>
> 
> To avoid hanging userspace components that might have been waiting on the
> active fences of the destroyed timeline we need to signal with error all
> remaining fences on such timeline.
> 
> This restore the default behaviour of the Android sw_sync framework, which
> Android still relies on. It was broken on the dma fence conversion a few
> years ago and never fixed.
> 
> v2: Do not bother with cleanup do the list (Chris Wilson)
> 
> Signed-off-by: Dominik Behr <dbehr@chromium.org>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

> ---
>  drivers/dma-buf/sw_sync.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
> index 38cc7389a6c1..f183eef074fd 100644
> --- a/drivers/dma-buf/sw_sync.c
> +++ b/drivers/dma-buf/sw_sync.c
> @@ -321,9 +321,19 @@ static int sw_sync_debugfs_open(struct inode *inode, struct file *file)
>  static int sw_sync_debugfs_release(struct inode *inode, struct file *file)
>  {
>         struct sync_timeline *obj = file->private_data;
> +       struct sync_pt *pt, *next;
>  
>         smp_wmb();
>  
> +       spin_lock_irq(&obj->lock);

Given the spinlock, that uncommented barrier (what is it paired with?)
above is redundant.
-Chris
Gustavo Padovan Sept. 8, 2017, 1:03 p.m. UTC | #2 
Hi Chris,

2017-09-07 Chris Wilson <chris@chris-wilson.co.uk>:

> Quoting Gustavo Padovan (2017-09-07 20:02:46)
> > From: Dominik Behr <dbehr@chromium.org>
> > 
> > To avoid hanging userspace components that might have been waiting on the
> > active fences of the destroyed timeline we need to signal with error all
> > remaining fences on such timeline.
> > 
> > This restore the default behaviour of the Android sw_sync framework, which
> > Android still relies on. It was broken on the dma fence conversion a few
> > years ago and never fixed.
> > 
> > v2: Do not bother with cleanup do the list (Chris Wilson)
> > 
> > Signed-off-by: Dominik Behr <dbehr@chromium.org>
> > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
> 
> > ---
> >  drivers/dma-buf/sw_sync.c | 10 ++++++++++
> >  1 file changed, 10 insertions(+)
> > 
> > diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
> > index 38cc7389a6c1..f183eef074fd 100644
> > --- a/drivers/dma-buf/sw_sync.c
> > +++ b/drivers/dma-buf/sw_sync.c
> > @@ -321,9 +321,19 @@ static int sw_sync_debugfs_open(struct inode *inode, struct file *file)
> >  static int sw_sync_debugfs_release(struct inode *inode, struct file *file)
> >  {
> >         struct sync_timeline *obj = file->private_data;
> > +       struct sync_pt *pt, *next;
> >  
> >         smp_wmb();
> >  
> > +       spin_lock_irq(&obj->lock);
> 
> Given the spinlock, that uncommented barrier (what is it paired with?)
> above is redundant.

Okay, I'll remove the barrier and push that patch, I assume your r-b
will comtemplate that as well?

Gustavo
Chris Wilson Sept. 8, 2017, 1:07 p.m. UTC | #3 
Quoting Gustavo Padovan (2017-09-08 14:03:00)
> Hi Chris,
> 
> 2017-09-07 Chris Wilson <chris@chris-wilson.co.uk>:
> 
> > Quoting Gustavo Padovan (2017-09-07 20:02:46)
> > > From: Dominik Behr <dbehr@chromium.org>
> > > 
> > > To avoid hanging userspace components that might have been waiting on the
> > > active fences of the destroyed timeline we need to signal with error all
> > > remaining fences on such timeline.
> > > 
> > > This restore the default behaviour of the Android sw_sync framework, which
> > > Android still relies on. It was broken on the dma fence conversion a few
> > > years ago and never fixed.
> > > 
> > > v2: Do not bother with cleanup do the list (Chris Wilson)
> > > 
> > > Signed-off-by: Dominik Behr <dbehr@chromium.org>
> > > Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
> > Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
> > 
> > > ---
> > >  drivers/dma-buf/sw_sync.c | 10 ++++++++++
> > >  1 file changed, 10 insertions(+)
> > > 
> > > diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
> > > index 38cc7389a6c1..f183eef074fd 100644
> > > --- a/drivers/dma-buf/sw_sync.c
> > > +++ b/drivers/dma-buf/sw_sync.c
> > > @@ -321,9 +321,19 @@ static int sw_sync_debugfs_open(struct inode *inode, struct file *file)
> > >  static int sw_sync_debugfs_release(struct inode *inode, struct file *file)
> > >  {
> > >         struct sync_timeline *obj = file->private_data;
> > > +       struct sync_pt *pt, *next;
> > >  
> > >         smp_wmb();
> > >  
> > > +       spin_lock_irq(&obj->lock);
> > 
> > Given the spinlock, that uncommented barrier (what is it paired with?)
> > above is redundant.
> 
> Okay, I'll remove the barrier and push that patch, I assume your r-b
> will comtemplate that as well?

No worries, either as one patch or two, slap my r-b on it.
-Chris
diff mbox

Patch

diff --git a/drivers/dma-buf/sw_sync.c b/drivers/dma-buf/sw_sync.c
index 38cc7389a6c1..f183eef074fd 100644
--- a/drivers/dma-buf/sw_sync.c
+++ b/drivers/dma-buf/sw_sync.c
@@ -321,9 +321,19 @@  static int sw_sync_debugfs_open(struct inode *inode, struct file *file)
 static int sw_sync_debugfs_release(struct inode *inode, struct file *file)
 {
 	struct sync_timeline *obj = file->private_data;
+	struct sync_pt *pt, *next;
 
 	smp_wmb();
 
+	spin_lock_irq(&obj->lock);
+
+	list_for_each_entry_safe(pt, next, &obj->pt_list, link) {
+		dma_fence_set_error(&pt->base, -ENOENT);
+		dma_fence_signal_locked(&pt->base);
+	}
+
+	spin_unlock_irq(&obj->lock);
+
 	sync_timeline_put(obj);
 	return 0;
 }