Hi,
On 22/02/2022 14:46, Yongzhi Liu wrote:
[why]
i915_gem_object_put_pages_phys() frees pages and standard
pattern is to allow caller to not care if it's
NULL or not. This will reduce burden on
the callers to perform this check.
I don't think we want this in general, and definitely not just for phys
objects and leaving others not handle NULL.
See comment at the call site for these vfuncs:
/*
* XXX Temporary hijinx to avoid updating all backends to handle
* NULL pages. In the future, when we have more asynchronous
* get_pages backends we should be better able to handle the
* cancellation of the async task in a more uniform manner.
*/
if (!IS_ERR_OR_NULL(pages))
obj->ops->put_pages(obj, pages);
Maarten owns this area these days so he can comment as well.
Regards,
Tvrtko
[how]
Fix it by adding Null check.
Signed-off-by: Yongzhi Liu <lyz_cs@xxxxxxxxxx>
---
drivers/gpu/drm/i915/gem/i915_gem_phys.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_phys.c b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
index ca6faff..09c3dcb 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_phys.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_phys.c
@@ -95,8 +95,13 @@ void
i915_gem_object_put_pages_phys(struct drm_i915_gem_object *obj,
struct sg_table *pages)
{
- dma_addr_t dma = sg_dma_address(pages->sgl);
- void *vaddr = sg_page(pages->sgl);
+ dma_addr_t dma;
+ void *vaddr;
+
+ if (!pages)
+ return;
+ dma = sg_dma_address(pages->sgl);
+ vaddr = sg_page(pages->sgl);
__i915_gem_object_release_shmem(obj, pages, false);
