[PATCH] dma-buf: Move BUG_ON from _add_shared_fence to _add_shared_inplace
Chris Wilson
chris at chris-wilson.co.uk
Wed Jun 27 11:50:42 UTC 2018
Quoting Michel Dänzer (2018-06-26 15:31:47)
> From: Michel Dänzer <michel.daenzer at amd.com>
>
> Fixes the BUG_ON spuriously triggering under the following
> circumstances:
>
> * ttm_eu_reserve_buffers processes a list containing multiple BOs using
> the same reservation object, so it calls
> reservation_object_reserve_shared with that reservation object once
> for each such BO.
> * In reservation_object_reserve_shared, old->shared_count ==
> old->shared_max - 1, so obj->staged is freed in preparation of an
> in-place update.
> * ttm_eu_fence_buffer_objects calls reservation_object_add_shared_fence
> once for each of the BOs above, always with the same fence.
> * The first call adds the fence in the remaining free slot, after which
> old->shared_count == old->shared_max.
>
> In the next call to reservation_object_add_shared_fence, the BUG_ON
> triggers. However, nothing bad would happen in
> reservation_object_add_shared_inplace, since the fence is already in the
> reservation object.
>
> Prevent this by moving the BUG_ON to where an overflow would actually
> happen (e.g. if a buggy caller didn't call
> reservation_object_reserve_shared before).
>
> Cc: stable at vger.kernel.org
> Signed-off-by: Michel Dänzer <michel.daenzer at amd.com>
I've convinced myself (or rather have not found a valid argument
against) that being able to call reserve_shared + add_shared multiple
times for the same fence is an intended part of reservation_object API
I'd double check with Christian though.
Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
> drivers/dma-buf/reservation.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/dma-buf/reservation.c b/drivers/dma-buf/reservation.c
> index 314eb1071cce..532545b9488e 100644
> --- a/drivers/dma-buf/reservation.c
> +++ b/drivers/dma-buf/reservation.c
> @@ -141,6 +141,7 @@ reservation_object_add_shared_inplace(struct reservation_object *obj,
> if (signaled) {
> RCU_INIT_POINTER(fobj->shared[signaled_idx], fence);
> } else {
> + BUG_ON(fobj->shared_count >= fobj->shared_max);
Personally I would just let kasan detect this and throw away the BUG_ON
or at least move it behind some DMABUF_BUG_ON().
-Chris
More information about the dri-devel
mailing list