[PATCH] drm/ttm: Fix ttm_bo_move_memcpy() for subclassed struct ttm_resource

[PATCH] drm/ttm: Fix ttm_bo_move_memcpy() for subclassed struct ttm_resource

[Thomas Hellström]

The code was making a copy of a struct ttm_resource. However,
recently the struct ttm_resources were allowed to be subclassed and
also were allowed to be malloced, hence the driver could end up assuming
the copy we handed it was subclassed and worse, the original could have
been freed at this point.

Fix this by using the original struct ttm_resource before it is
potentially freed in ttm_bo_move_sync_cleanup()

Reported-by: Ben Skeggs <skeggsb@gmail.com>
Reported-by: Dave Airlie <airlied@gmail.com>
Cc: Christian König <christian.koenig@amd.com>
Fixes: 3bf3710e3718 ("drm/ttm: Add a generic TTM memcpy move for page-based iomem")
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
---
 drivers/gpu/drm/ttm/ttm_bo_util.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 5c20d0541cc3..c893c3db2623 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -139,7 +139,6 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	struct ttm_resource *src_mem = bo->resource;
 	struct ttm_resource_manager *src_man =
 		ttm_manager_type(bdev, src_mem->mem_type);
-	struct ttm_resource src_copy = *src_mem;
 	union {
 		struct ttm_kmap_iter_tt tt;
 		struct ttm_kmap_iter_linear_io io;
@@ -173,11 +172,10 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	if (!(clear && ttm && !(ttm->page_flags & TTM_PAGE_FLAG_ZERO_ALLOC)))
 		ttm_move_memcpy(clear, dst_mem->num_pages, dst_iter, src_iter);
 
-	src_copy = *src_mem;
+	if (!src_iter->ops->maps_tt)
+		ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, src_mem);
 	ttm_bo_move_sync_cleanup(bo, dst_mem);
 
-	if (!src_iter->ops->maps_tt)
-		ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, &src_copy);
 out_src_iter:
 	if (!dst_iter->ops->maps_tt)
 		ttm_kmap_iter_linear_io_fini(&_dst_iter.io, bdev, dst_mem);
-- 
2.31.1

Re: [PATCH] drm/ttm: Fix ttm_bo_move_memcpy() for subclassed struct ttm_resource

[Christian König]

Am 30.08.21 um 09:48 schrieb Thomas Hellström:
> The code was making a copy of a struct ttm_resource. However,
> recently the struct ttm_resources were allowed to be subclassed and
> also were allowed to be malloced, hence the driver could end up assuming
> the copy we handed it was subclassed and worse, the original could have
> been freed at this point.
>
> Fix this by using the original struct ttm_resource before it is
> potentially freed in ttm_bo_move_sync_cleanup()
>
> Reported-by: Ben Skeggs <skeggsb@gmail.com>
> Reported-by: Dave Airlie <airlied@gmail.com>
> Cc: Christian König <christian.koenig@amd.com>
> Fixes: 3bf3710e3718 ("drm/ttm: Add a generic TTM memcpy move for page-based iomem")
> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>

Reviewed-by: Christian König <christian.koenig@amd.com>

> ---
>   drivers/gpu/drm/ttm/ttm_bo_util.c | 6 ++----
>   1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
> index 5c20d0541cc3..c893c3db2623 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo_util.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
> @@ -139,7 +139,6 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
>   	struct ttm_resource *src_mem = bo->resource;
>   	struct ttm_resource_manager *src_man =
>   		ttm_manager_type(bdev, src_mem->mem_type);
> -	struct ttm_resource src_copy = *src_mem;
>   	union {
>   		struct ttm_kmap_iter_tt tt;
>   		struct ttm_kmap_iter_linear_io io;
> @@ -173,11 +172,10 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
>   	if (!(clear && ttm && !(ttm->page_flags & TTM_PAGE_FLAG_ZERO_ALLOC)))
>   		ttm_move_memcpy(clear, dst_mem->num_pages, dst_iter, src_iter);
>   
> -	src_copy = *src_mem;
> +	if (!src_iter->ops->maps_tt)
> +		ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, src_mem);
>   	ttm_bo_move_sync_cleanup(bo, dst_mem);
>   
> -	if (!src_iter->ops->maps_tt)
> -		ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, &src_copy);
>   out_src_iter:
>   	if (!dst_iter->ops->maps_tt)
>   		ttm_kmap_iter_linear_io_fini(&_dst_iter.io, bdev, dst_mem);

Re: [PATCH] drm/ttm: Fix ttm_bo_move_memcpy() for subclassed struct ttm_resource

[Ben Skeggs]

On Mon, 30 Aug 2021 at 17:48, Thomas Hellström
<thomas.hellstrom@linux.intel.com> wrote:
>
> The code was making a copy of a struct ttm_resource. However,
> recently the struct ttm_resources were allowed to be subclassed and
> also were allowed to be malloced, hence the driver could end up assuming
> the copy we handed it was subclassed and worse, the original could have
> been freed at this point.
>
> Fix this by using the original struct ttm_resource before it is
> potentially freed in ttm_bo_move_sync_cleanup()
>
> Reported-by: Ben Skeggs <skeggsb@gmail.com>
> Reported-by: Dave Airlie <airlied@gmail.com>
> Cc: Christian König <christian.koenig@amd.com>
> Fixes: 3bf3710e3718 ("drm/ttm: Add a generic TTM memcpy move for page-based iomem")
> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
That's basically identical to what I came up with locally, so:

Reviewed-by: Ben Skeggs <bskeggs@redhat.com>

> ---
>  drivers/gpu/drm/ttm/ttm_bo_util.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
> index 5c20d0541cc3..c893c3db2623 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo_util.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
> @@ -139,7 +139,6 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
>         struct ttm_resource *src_mem = bo->resource;
>         struct ttm_resource_manager *src_man =
>                 ttm_manager_type(bdev, src_mem->mem_type);
> -       struct ttm_resource src_copy = *src_mem;
>         union {
>                 struct ttm_kmap_iter_tt tt;
>                 struct ttm_kmap_iter_linear_io io;
> @@ -173,11 +172,10 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
>         if (!(clear && ttm && !(ttm->page_flags & TTM_PAGE_FLAG_ZERO_ALLOC)))
>                 ttm_move_memcpy(clear, dst_mem->num_pages, dst_iter, src_iter);
>
> -       src_copy = *src_mem;
> +       if (!src_iter->ops->maps_tt)
> +               ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, src_mem);
>         ttm_bo_move_sync_cleanup(bo, dst_mem);
>
> -       if (!src_iter->ops->maps_tt)
> -               ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, &src_copy);
>  out_src_iter:
>         if (!dst_iter->ops->maps_tt)
>                 ttm_kmap_iter_linear_io_fini(&_dst_iter.io, bdev, dst_mem);
> --
> 2.31.1
>

Re: [PATCH] drm/ttm: Fix ttm_bo_move_memcpy() for subclassed struct ttm_resource

[Daniel Vetter]

On Mon, Aug 30, 2021 at 09:48:35AM +0200, Thomas Hellström wrote:
> The code was making a copy of a struct ttm_resource. However,
> recently the struct ttm_resources were allowed to be subclassed and
> also were allowed to be malloced, hence the driver could end up assuming
> the copy we handed it was subclassed and worse, the original could have
> been freed at this point.
> 
> Fix this by using the original struct ttm_resource before it is
> potentially freed in ttm_bo_move_sync_cleanup()
> 
> Reported-by: Ben Skeggs <skeggsb@gmail.com>
> Reported-by: Dave Airlie <airlied@gmail.com>
> Cc: Christian König <christian.koenig@amd.com>
> Fixes: 3bf3710e3718 ("drm/ttm: Add a generic TTM memcpy move for page-based iomem")
> Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>

This doesn't apply cleanly to drm-misc-next-fixes, do we need a backmerge
or something? Can you pls coordinate with Maarten and then also poke for a
pull request so this isn't stuck?

Also since 5.14 is released this needs cc: stable.
-Daniel

> ---
>  drivers/gpu/drm/ttm/ttm_bo_util.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
> index 5c20d0541cc3..c893c3db2623 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo_util.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
> @@ -139,7 +139,6 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
>  	struct ttm_resource *src_mem = bo->resource;
>  	struct ttm_resource_manager *src_man =
>  		ttm_manager_type(bdev, src_mem->mem_type);
> -	struct ttm_resource src_copy = *src_mem;
>  	union {
>  		struct ttm_kmap_iter_tt tt;
>  		struct ttm_kmap_iter_linear_io io;
> @@ -173,11 +172,10 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
>  	if (!(clear && ttm && !(ttm->page_flags & TTM_PAGE_FLAG_ZERO_ALLOC)))
>  		ttm_move_memcpy(clear, dst_mem->num_pages, dst_iter, src_iter);
>  
> -	src_copy = *src_mem;
> +	if (!src_iter->ops->maps_tt)
> +		ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, src_mem);
>  	ttm_bo_move_sync_cleanup(bo, dst_mem);
>  
> -	if (!src_iter->ops->maps_tt)
> -		ttm_kmap_iter_linear_io_fini(&_src_iter.io, bdev, &src_copy);
>  out_src_iter:
>  	if (!dst_iter->ops->maps_tt)
>  		ttm_kmap_iter_linear_io_fini(&_dst_iter.io, bdev, dst_mem);
> -- 
> 2.31.1
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch