Do not allocate memory with GFP_KERNEL when holding a spin_lock, it may
sleep. Use GFP_NOWAIT instead.
Fixes: 950fd867c635 ("staging: bcm2835-camera: Replace open-coded idr with a struct idr.")
Signed-off-by: Christophe JAILLET <[email protected]>
---
drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c b/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c
index 16af735af5c3..438d548c6e24 100644
--- a/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c
+++ b/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c
@@ -186,7 +186,7 @@ get_msg_context(struct vchiq_mmal_instance *instance)
*/
spin_lock(&instance->context_map_lock);
handle = idr_alloc(&instance->context_map, msg_context,
- 0, 0, GFP_KERNEL);
+ 0, 0, GFP_NOWAIT);
spin_unlock(&instance->context_map_lock);
if (handle < 0) {
--
2.20.1
On Mon, Jun 24, 2019 at 07:33:51AM +0200, Christophe JAILLET wrote:
> Do not allocate memory with GFP_KERNEL when holding a spin_lock, it may
> sleep. Use GFP_NOWAIT instead.
>
checking for this in the rest of the kernel with a cocci spatch
<snip>
virtual report
@nonatomic@
position p;
identifier var;
@@
spin_lock(...)
... when != spin_unlock(...)
* var = idr_alloc@p(...,GFP_KERNEL);
... when != spin_unlock(...)
spin_unlock(...);
<snip>
this seems to be the only instance of this specific problem.
> Fixes: 950fd867c635 ("staging: bcm2835-camera: Replace open-coded idr with a struct idr.")
The GFP_KERNEL actually was there befor this patch so not sure if this Fixes
ref is correct - I think the GFP_KERNEL was introduced in:
4e6bafdfb9f3 ("staging: bcm2835_camera: Use a mapping table for context field of mmal_msg_header")
> Signed-off-by: Christophe JAILLET <[email protected]>
Reviewed-by: Nicholas Mc Guire <[email protected]>
> ---
> drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c b/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c
> index 16af735af5c3..438d548c6e24 100644
> --- a/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c
> +++ b/drivers/staging/vc04_services/bcm2835-camera/mmal-vchiq.c
> @@ -186,7 +186,7 @@ get_msg_context(struct vchiq_mmal_instance *instance)
> */
> spin_lock(&instance->context_map_lock);
> handle = idr_alloc(&instance->context_map, msg_context,
> - 0, 0, GFP_KERNEL);
> + 0, 0, GFP_NOWAIT);
> spin_unlock(&instance->context_map_lock);
>
> if (handle < 0) {
> --
> 2.20.1
>
Hi Christophe,
Am 24.06.2019 um 07:33 schrieb Christophe JAILLET:
> Do not allocate memory with GFP_KERNEL when holding a spin_lock, it may
> sleep. Use GFP_NOWAIT instead.
>
> Fixes: 950fd867c635 ("staging: bcm2835-camera: Replace open-coded idr with a struct idr.")
> Signed-off-by: Christophe JAILLET <[email protected]>
there has been a fix for this, which isn't upstreamed yet. The preferred
solution is to replace the spin_lock with a mutex. Since i'm currently
working on this i would take care of this.
Sorry about this.
Stefan