[PATCH v2 06/18] dma-buf: Use sequence counter with associated wound/wait mutex
Daniel Vetter
daniel at ffwll.ch
Mon Jun 8 14:32:21 UTC 2020
On Mon, Jun 08, 2020 at 02:57:17AM +0200, Ahmed S. Darwish wrote:
> A sequence counter write side critical section must be protected by some
> form of locking to serialize writers. If the serialization primitive is
> not disabling preemption implicitly, preemption has to be explicitly
> disabled before entering the sequence counter write side critical
> section.
>
> The dma-buf reservation subsystem uses plain sequence counters to manage
> updates to reservations. Writer serialization is accomplished through a
> wound/wait mutex.
>
> Acquiring a wound/wait mutex does not disable preemption, so this needs
> to be done manually before and after the write side critical section.
>
> Use the newly-added seqcount_ww_mutex_t instead:
>
> - It associates the ww_mutex with the sequence count, which enables
> lockdep to validate that the write side critical section is properly
> serialized.
>
> - It removes the need to explicitly add preempt_disable/enable()
> around the write side critical section because the write_begin/end()
> functions for this new data type automatically do this.
>
> If lockdep is disabled this ww_mutex lock association is compiled out
> and has neither storage size nor runtime overhead.
>
> Signed-off-by: Ahmed S. Darwish <a.darwish at linutronix.de>
I'm not seeing the patch that adds the seqcount ww_mutex glue and not
quite motivated enough to grab it from lore, so someone else needs to
check the details. Just
Acked-by: Daniel Vetter <daniel.vetter at ffwll.ch>
for merging through whatever tree/branch makes sense from me.
-Daniel
> ---
> drivers/dma-buf/dma-resv.c | 8 +-------
> drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 2 --
> include/linux/dma-resv.h | 2 +-
> 3 files changed, 2 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/dma-buf/dma-resv.c b/drivers/dma-buf/dma-resv.c
> index 590ce7ad60a0..3aba2b2bfc48 100644
> --- a/drivers/dma-buf/dma-resv.c
> +++ b/drivers/dma-buf/dma-resv.c
> @@ -128,7 +128,7 @@ subsys_initcall(dma_resv_lockdep);
> void dma_resv_init(struct dma_resv *obj)
> {
> ww_mutex_init(&obj->lock, &reservation_ww_class);
> - seqcount_init(&obj->seq);
> + seqcount_ww_mutex_init(&obj->seq, &obj->lock);
>
> RCU_INIT_POINTER(obj->fence, NULL);
> RCU_INIT_POINTER(obj->fence_excl, NULL);
> @@ -259,7 +259,6 @@ void dma_resv_add_shared_fence(struct dma_resv *obj, struct dma_fence *fence)
> fobj = dma_resv_get_list(obj);
> count = fobj->shared_count;
>
> - preempt_disable();
> write_seqcount_begin(&obj->seq);
>
> for (i = 0; i < count; ++i) {
> @@ -281,7 +280,6 @@ void dma_resv_add_shared_fence(struct dma_resv *obj, struct dma_fence *fence)
> smp_store_mb(fobj->shared_count, count);
>
> write_seqcount_end(&obj->seq);
> - preempt_enable();
> dma_fence_put(old);
> }
> EXPORT_SYMBOL(dma_resv_add_shared_fence);
> @@ -308,14 +306,12 @@ void dma_resv_add_excl_fence(struct dma_resv *obj, struct dma_fence *fence)
> if (fence)
> dma_fence_get(fence);
>
> - preempt_disable();
> write_seqcount_begin(&obj->seq);
> /* write_seqcount_begin provides the necessary memory barrier */
> RCU_INIT_POINTER(obj->fence_excl, fence);
> if (old)
> old->shared_count = 0;
> write_seqcount_end(&obj->seq);
> - preempt_enable();
>
> /* inplace update, no shared fences */
> while (i--)
> @@ -393,13 +389,11 @@ int dma_resv_copy_fences(struct dma_resv *dst, struct dma_resv *src)
> src_list = dma_resv_get_list(dst);
> old = dma_resv_get_excl(dst);
>
> - preempt_disable();
> write_seqcount_begin(&dst->seq);
> /* write_seqcount_begin provides the necessary memory barrier */
> RCU_INIT_POINTER(dst->fence_excl, new);
> RCU_INIT_POINTER(dst->fence, dst_list);
> write_seqcount_end(&dst->seq);
> - preempt_enable();
>
> dma_resv_list_free(src_list);
> dma_fence_put(old);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> index 6a5b91d23fd9..c71c0bb6ce26 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> @@ -258,11 +258,9 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo,
> new->shared_count = k;
>
> /* Install the new fence list, seqcount provides the barriers */
> - preempt_disable();
> write_seqcount_begin(&resv->seq);
> RCU_INIT_POINTER(resv->fence, new);
> write_seqcount_end(&resv->seq);
> - preempt_enable();
>
> /* Drop the references to the removed fences or move them to ef_list */
> for (i = j, k = 0; i < old->shared_count; ++i) {
> diff --git a/include/linux/dma-resv.h b/include/linux/dma-resv.h
> index a6538ae7d93f..d44a77e8a7e3 100644
> --- a/include/linux/dma-resv.h
> +++ b/include/linux/dma-resv.h
> @@ -69,7 +69,7 @@ struct dma_resv_list {
> */
> struct dma_resv {
> struct ww_mutex lock;
> - seqcount_t seq;
> + seqcount_ww_mutex_t seq;
>
> struct dma_fence __rcu *fence_excl;
> struct dma_resv_list __rcu *fence;
> --
> 2.20.1
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
More information about the amd-gfx
mailing list