[PATCH] dma-buf: fix unintended pin/unpin warnings
Daniel Vetter
daniel at ffwll.ch
Thu May 20 14:13:38 UTC 2021
On Thu, May 20, 2021 at 01:37:47PM +0200, Christian König wrote:
> Am 17.05.21 um 16:57 schrieb Daniel Vetter:
> > On Mon, May 17, 2021 at 10:09:13AM -0400, Alex Deucher wrote:
> > > On Mon, May 17, 2021 at 7:57 AM Christian König
> > > <ckoenig.leichtzumerken at gmail.com> wrote:
> > > > DMA-buf internal users call the pin/unpin functions without having a
> > > > dynamic attachment. Avoid the warning and backtrace in the logs.
> > > >
> > > > Signed-off-by: Christian König <christian.koenig at amd.com>
> > > > Bugs: https://gitlab.freedesktop.org/drm/intel/-/issues/3481
> > > > Fixes: c545781e1c55 ("dma-buf: doc polish for pin/unpin")
> > > > CC: stable at kernel.org
> > > Reviewed-by: Alex Deucher <alexander.deucher at amd.com>
> > Hm this means we're losing the dma_resv_assert_held check, do we have that
> > in amdgpu callbacks to make sure we're not accidentally breaking stuff
> > later on?
>
> Mhm, well this is just for calling the pin/unpin internally from the DMA-buf
> framework itself.
It's both, but currently I think no users for it (since it would be
pinning for display).
Maybe we could do __ internal wrappers to keep the lockdep checks?
> Need to double check, but I think all those cases either have a
> dma_resv_assert_held() or are locking the reservation themselves before
> calling the function.
Yeah right now it's still all good, but that changes easily :-)
> But yeah, rather good point.
Do you plan to type something that I can ack?
-Daniel
>
> Christian.
>
> >
> > But yeah lgtm, Reviewed-by: Daniel Vetter <daniel.vetter at ffwll.ch>
> >
> > > > ---
> > > > drivers/dma-buf/dma-buf.c | 10 +++++-----
> > > > 1 file changed, 5 insertions(+), 5 deletions(-)
> > > >
> > > > diff --git a/drivers/dma-buf/dma-buf.c b/drivers/dma-buf/dma-buf.c
> > > > index f264b70c383e..eadd1eaa2fb5 100644
> > > > --- a/drivers/dma-buf/dma-buf.c
> > > > +++ b/drivers/dma-buf/dma-buf.c
> > > > @@ -760,7 +760,7 @@ dma_buf_dynamic_attach(struct dma_buf *dmabuf, struct device *dev,
> > > >
> > > > if (dma_buf_is_dynamic(attach->dmabuf)) {
> > > > dma_resv_lock(attach->dmabuf->resv, NULL);
> > > > - ret = dma_buf_pin(attach);
> > > > + ret = dmabuf->ops->pin(attach);
> > > > if (ret)
> > > > goto err_unlock;
> > > > }
> > > > @@ -786,7 +786,7 @@ dma_buf_dynamic_attach(struct dma_buf *dmabuf, struct device *dev,
> > > >
> > > > err_unpin:
> > > > if (dma_buf_is_dynamic(attach->dmabuf))
> > > > - dma_buf_unpin(attach);
> > > > + dmabuf->ops->unpin(attach);
> > > >
> > > > err_unlock:
> > > > if (dma_buf_is_dynamic(attach->dmabuf))
> > > > @@ -843,7 +843,7 @@ void dma_buf_detach(struct dma_buf *dmabuf, struct dma_buf_attachment *attach)
> > > > __unmap_dma_buf(attach, attach->sgt, attach->dir);
> > > >
> > > > if (dma_buf_is_dynamic(attach->dmabuf)) {
> > > > - dma_buf_unpin(attach);
> > > > + dmabuf->ops->unpin(attach);
> > > > dma_resv_unlock(attach->dmabuf->resv);
> > > > }
> > > > }
> > > > @@ -956,7 +956,7 @@ struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach,
> > > > if (dma_buf_is_dynamic(attach->dmabuf)) {
> > > > dma_resv_assert_held(attach->dmabuf->resv);
> > > > if (!IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY)) {
> > > > - r = dma_buf_pin(attach);
> > > > + r = attach->dmabuf->ops->pin(attach);
> > > > if (r)
> > > > return ERR_PTR(r);
> > > > }
> > > > @@ -968,7 +968,7 @@ struct sg_table *dma_buf_map_attachment(struct dma_buf_attachment *attach,
> > > >
> > > > if (IS_ERR(sg_table) && dma_buf_is_dynamic(attach->dmabuf) &&
> > > > !IS_ENABLED(CONFIG_DMABUF_MOVE_NOTIFY))
> > > > - dma_buf_unpin(attach);
> > > > + attach->dmabuf->ops->unpin(attach);
> > > >
> > > > if (!IS_ERR(sg_table) && attach->dmabuf->ops->cache_sgt_mapping) {
> > > > attach->sgt = sg_table;
> > > > --
> > > > 2.25.1
> > > >
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
More information about the dri-devel
mailing list