Use struct_size() instead of hand writing it.
This is less verbose and more informative.
Signed-off-by: Su Hui <[email protected]>
---
drivers/gpu/drm/nouveau/nvif/object.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/nouveau/nvif/object.c b/drivers/gpu/drm/nouveau/nvif/object.c
index 4d1aaee8fe15..4bd693aa4ee0 100644
--- a/drivers/gpu/drm/nouveau/nvif/object.c
+++ b/drivers/gpu/drm/nouveau/nvif/object.c
@@ -65,7 +65,7 @@ nvif_object_sclass_get(struct nvif_object *object, struct nvif_sclass **psclass)
u32 size;
while (1) {
- size = sizeof(*args) + cnt * sizeof(args->sclass.oclass[0]);
+ size = struct_size(args, sclass.oclass, cnt);
if (!(args = kmalloc(size, GFP_KERNEL)))
return -ENOMEM;
args->ioctl.version = 0;
--
2.30.2
On Wed, May 31, 2023 at 12:38:26PM +0800, Su Hui wrote:
> Use struct_size() instead of hand writing it.
> This is less verbose and more informative.
>
> Signed-off-by: Su Hui <[email protected]>
> ---
> drivers/gpu/drm/nouveau/nvif/object.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/nouveau/nvif/object.c b/drivers/gpu/drm/nouveau/nvif/object.c
> index 4d1aaee8fe15..4bd693aa4ee0 100644
> --- a/drivers/gpu/drm/nouveau/nvif/object.c
> +++ b/drivers/gpu/drm/nouveau/nvif/object.c
> @@ -65,7 +65,7 @@ nvif_object_sclass_get(struct nvif_object *object, struct nvif_sclass **psclass)
> u32 size;
>
> while (1) {
> - size = sizeof(*args) + cnt * sizeof(args->sclass.oclass[0]);
> + size = struct_size(args, sclass.oclass, cnt);
This is from the original code, but now that you are using the
struct_size() macro static checkers will complain about it. (Maybe they
don't yet?). size is a u32. Never save struct_size() returns to
anything except unsigned long or size_t. (ssize_t is also fine, I
suppose). Otherwise, you do not benefit from the integer overflow
checking.
> if (!(args = kmalloc(size, GFP_KERNEL)))
> return -ENOMEM;
> args->ioctl.version = 0;
regards,
dan carpenter
On 2023/5/31 16:31, Dan Carpenter wrote:
> On Wed, May 31, 2023 at 12:38:26PM +0800, Su Hui wrote:
>> Use struct_size() instead of hand writing it.
>> This is less verbose and more informative.
>>
>> Signed-off-by: Su Hui <[email protected]>
>> ---
>> drivers/gpu/drm/nouveau/nvif/object.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/nouveau/nvif/object.c b/drivers/gpu/drm/nouveau/nvif/object.c
>> index 4d1aaee8fe15..4bd693aa4ee0 100644
>> --- a/drivers/gpu/drm/nouveau/nvif/object.c
>> +++ b/drivers/gpu/drm/nouveau/nvif/object.c
>> @@ -65,7 +65,7 @@ nvif_object_sclass_get(struct nvif_object *object, struct nvif_sclass **psclass)
>> u32 size;
>>
>> while (1) {
>> - size = sizeof(*args) + cnt * sizeof(args->sclass.oclass[0]);
>> + size = struct_size(args, sclass.oclass, cnt);
> This is from the original code, but now that you are using the
> struct_size() macro static checkers will complain about it. (Maybe they
> don't yet?). size is a u32. Never save struct_size() returns to
> anything except unsigned long or size_t. (ssize_t is also fine, I
> suppose). Otherwise, you do not benefit from the integer overflow
> checking.
Sorry, I don't notice the issue caused by type size.
You are right, this patch is wrong because of the type mismatch.
Thanks for your reply!
Su Hui
>> if (!(args = kmalloc(size, GFP_KERNEL)))
>> return -ENOMEM;
>> args->ioctl.version = 0;
> regards,
> dan carpenter