[PATCH V1] accel/amdxdna: Use vma->vm_mm consistently in amdxdna_hmm_register()

[PATCH V1] accel/amdxdna: Use vma->vm_mm consistently in amdxdna_hmm_register()

[Lizhi Hou]

amdxdna_hmm_register() keys the umap list lookup and the interval
notifier insertion on current->mm, while amdxdna_hmm_unregister() and
amdxdna_mark_mapp_invalid() match the same list entries against
vma->vm_mm. The two are identical for the ->mmap() callback, so this is
not a functional change, but the inconsistency makes the code harder to
follow.

Use vma->vm_mm in both places. It is the mm whose mmap_lock is held
across ->mmap(), which is what mmu_interval_notifier_insert_locked()
asserts on, and it is the mm the teardown paths look the mapping up by.

Assisted-by: LLM
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
---
 drivers/accel/amdxdna/amdxdna_gem.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/accel/amdxdna/amdxdna_gem.c b/drivers/accel/amdxdna/amdxdna_gem.c
index 09e48c794e1b..92125bc753c7 100644
--- a/drivers/accel/amdxdna/amdxdna_gem.c
+++ b/drivers/accel/amdxdna/amdxdna_gem.c
@@ -443,7 +443,7 @@ static int amdxdna_hmm_register(struct amdxdna_gem_obj *abo,
 
 	down_read(&xdna->notifier_lock);
 	list_for_each_entry(mapp, &abo->mem.umap_list, node) {
-		if (compare_range(mapp, current->mm, addr, addr + len)) {
+		if (compare_range(mapp, vma->vm_mm, addr, addr + len)) {
 			up_read(&xdna->notifier_lock);
 			return 0;
 		}
@@ -477,7 +477,7 @@ static int amdxdna_hmm_register(struct amdxdna_gem_obj *abo,
 	kref_init(&mapp->refcnt);
 
 	ret = mmu_interval_notifier_insert_locked(&mapp->notifier,
-						  current->mm,
+						  vma->vm_mm,
 						  addr,
 						  len,
 						  &amdxdna_hmm_ops);

base-commit: d2a8cee91bfa8616f713c7c8f539a81e8e3554ea
-- 
2.34.1

Re: [PATCH V1] accel/amdxdna: Use vma->vm_mm consistently in amdxdna_hmm_register()

[Max Zhen]

On 11/6/2026 Fri 12:35, Lizhi Hou wrote:
> amdxdna_hmm_register() keys the umap list lookup and the interval
> notifier insertion on current->mm, while amdxdna_hmm_unregister() and
> amdxdna_mark_mapp_invalid() match the same list entries against
> vma->vm_mm. The two are identical for the ->mmap() callback, so this is
> not a functional change, but the inconsistency makes the code harder to
> follow.
> 
> Use vma->vm_mm in both places. It is the mm whose mmap_lock is held
> across ->mmap(), which is what mmu_interval_notifier_insert_locked()
> asserts on, and it is the mm the teardown paths look the mapping up by.
> 
> Assisted-by: LLM
> Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Reviewed-by: Max Zhen <max.zhen@amd.com>
> ---
>   drivers/accel/amdxdna/amdxdna_gem.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/accel/amdxdna/amdxdna_gem.c b/drivers/accel/amdxdna/amdxdna_gem.c
> index 09e48c794e1b..92125bc753c7 100644
> --- a/drivers/accel/amdxdna/amdxdna_gem.c
> +++ b/drivers/accel/amdxdna/amdxdna_gem.c
> @@ -443,7 +443,7 @@ static int amdxdna_hmm_register(struct amdxdna_gem_obj *abo,
>   
>   	down_read(&xdna->notifier_lock);
>   	list_for_each_entry(mapp, &abo->mem.umap_list, node) {
> -		if (compare_range(mapp, current->mm, addr, addr + len)) {
> +		if (compare_range(mapp, vma->vm_mm, addr, addr + len)) {
>   			up_read(&xdna->notifier_lock);
>   			return 0;
>   		}
> @@ -477,7 +477,7 @@ static int amdxdna_hmm_register(struct amdxdna_gem_obj *abo,
>   	kref_init(&mapp->refcnt);
>   
>   	ret = mmu_interval_notifier_insert_locked(&mapp->notifier,
> -						  current->mm,
> +						  vma->vm_mm,
>   						  addr,
>   						  len,
>   						  &amdxdna_hmm_ops);
> 
> base-commit: d2a8cee91bfa8616f713c7c8f539a81e8e3554ea