2022-06-10 18:00:33

by Rob Clark

Subject: [PATCH] drm/msm/gem: Drop early returns in close/purge vma

From: Rob Clark <[email protected]>

Keep the warn, but drop the early return. If we do manage to hit this
sort of issue, skipping the cleanup just makes things worse (dangling
drm_mm_nodes when the msm_gem_vma is freed, etc). Whereas the worst
that happens if we tear down a mapping the GPU is accessing is that we
get GPU iova faults, but otherwise the world keeps spinning.

Signed-off-by: Rob Clark <[email protected]>
---
drivers/gpu/drm/msm/msm_gem_vma.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
index 3c1dc9241831..c471aebcdbab 100644
--- a/drivers/gpu/drm/msm/msm_gem_vma.c
+++ b/drivers/gpu/drm/msm/msm_gem_vma.c
@@ -62,8 +62,7 @@ void msm_gem_purge_vma(struct msm_gem_address_space *aspace,
unsigned size = vma->node.size;

/* Print a message if we try to purge a vma in use */
- if (GEM_WARN_ON(msm_gem_vma_inuse(vma)))
- return;
+ GEM_WARN_ON(msm_gem_vma_inuse(vma));

/* Don't do anything if the memory isn't mapped */
if (!vma->mapped)
@@ -128,8 +127,7 @@ msm_gem_map_vma(struct msm_gem_address_space *aspace,
void msm_gem_close_vma(struct msm_gem_address_space *aspace,
struct msm_gem_vma *vma)
{
- if (GEM_WARN_ON(msm_gem_vma_inuse(vma) || vma->mapped))
- return;
+ GEM_WARN_ON(msm_gem_vma_inuse(vma) || vma->mapped);

spin_lock(&aspace->lock);
if (vma->iova)
--
2.36.1

2022-06-12 04:02:00

by Steev Klimaszewski

Subject: Re: [PATCH] drm/msm/gem: Drop early returns in close/purge vma

Hi Rob,

On 6/10/22 12:20 PM, Rob Clark wrote:
> From: Rob Clark <[email protected]>
>
> Keep the warn, but drop the early return. If we do manage to hit this
> sort of issue, skipping the cleanup just makes things worse (dangling
> drm_mm_nodes when the msm_gem_vma is freed, etc). Whereas the worst
> that happens if we tear down a mapping the GPU is accessing is that we
> get GPU iova faults, but otherwise the world keeps spinning.
>
> Signed-off-by: Rob Clark <[email protected]>
> ---
> drivers/gpu/drm/msm/msm_gem_vma.c | 6 ++----
> 1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
> index 3c1dc9241831..c471aebcdbab 100644
> --- a/drivers/gpu/drm/msm/msm_gem_vma.c
> +++ b/drivers/gpu/drm/msm/msm_gem_vma.c
> @@ -62,8 +62,7 @@ void msm_gem_purge_vma(struct msm_gem_address_space *aspace,
> unsigned size = vma->node.size;
>
> /* Print a message if we try to purge a vma in use */
> - if (GEM_WARN_ON(msm_gem_vma_inuse(vma)))
> - return;
> + GEM_WARN_ON(msm_gem_vma_inuse(vma));
>
> /* Don't do anything if the memory isn't mapped */
> if (!vma->mapped)
> @@ -128,8 +127,7 @@ msm_gem_map_vma(struct msm_gem_address_space *aspace,
> void msm_gem_close_vma(struct msm_gem_address_space *aspace,
> struct msm_gem_vma *vma)
> {
> - if (GEM_WARN_ON(msm_gem_vma_inuse(vma) || vma->mapped))
> - return;
> + GEM_WARN_ON(msm_gem_vma_inuse(vma) || vma->mapped);
>
> spin_lock(&aspace->lock);
> if (vma->iova)

I've seen the splat on the Lenovo Yoga C630 here, and have tested this
patch, and as described, the splat still happens, but the system is
still able to be used.

Tested-by: Steev Klimaszewski <[email protected]>

2022-06-15 15:10:51

by Rob Clark

Subject: Re: [PATCH] drm/msm/gem: Drop early returns in close/purge vma

On Sat, Jun 11, 2022 at 11:16 AM Steev Klimaszewski <[email protected]> wrote:
>
> Hi Rob,
>
> On 6/10/22 12:20 PM, Rob Clark wrote:
> > From: Rob Clark <[email protected]>
> >
> > Keep the warn, but drop the early return. If we do manage to hit this
> > sort of issue, skipping the cleanup just makes things worse (dangling
> > drm_mm_nodes when the msm_gem_vma is freed, etc). Whereas the worst
> > that happens if we tear down a mapping the GPU is accessing is that we
> > get GPU iova faults, but otherwise the world keeps spinning.
> >

forgot this initially:

Reported-by: Steev Klimaszewski <[email protected]>

> > Signed-off-by: Rob Clark <[email protected]>
> > ---
> > drivers/gpu/drm/msm/msm_gem_vma.c | 6 ++----
> > 1 file changed, 2 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/msm_gem_vma.c b/drivers/gpu/drm/msm/msm_gem_vma.c
> > index 3c1dc9241831..c471aebcdbab 100644
> > --- a/drivers/gpu/drm/msm/msm_gem_vma.c
> > +++ b/drivers/gpu/drm/msm/msm_gem_vma.c
> > @@ -62,8 +62,7 @@ void msm_gem_purge_vma(struct msm_gem_address_space *aspace,
> > unsigned size = vma->node.size;
> >
> > /* Print a message if we try to purge a vma in use */
> > - if (GEM_WARN_ON(msm_gem_vma_inuse(vma)))
> > - return;
> > + GEM_WARN_ON(msm_gem_vma_inuse(vma));
> >
> > /* Don't do anything if the memory isn't mapped */
> > if (!vma->mapped)
> > @@ -128,8 +127,7 @@ msm_gem_map_vma(struct msm_gem_address_space *aspace,
> > void msm_gem_close_vma(struct msm_gem_address_space *aspace,
> > struct msm_gem_vma *vma)
> > {
> > - if (GEM_WARN_ON(msm_gem_vma_inuse(vma) || vma->mapped))
> > - return;
> > + GEM_WARN_ON(msm_gem_vma_inuse(vma) || vma->mapped);
> >
> > spin_lock(&aspace->lock);
> > if (vma->iova)
>
> I've seen the splat on the Lenovo Yoga C630 here, and have tested this
> patch, and as described, the splat still happens, but the system is
> still able to be used.
>
> Tested-by: Steev Klimaszewski <[email protected]>
>