Rob Herring <robh@kernel.org> writes:

> On Mon, Apr 1, 2019 at 10:43 AM Eric Anholt <eric@anholt.net> wrote:
>>
>> Chris Wilson <chris@chris-wilson.co.uk> writes:
>>
>> > Quoting Daniel Vetter (2019-04-01 14:06:48)
>> >> On Mon, Apr 1, 2019 at 9:47 AM Rob Herring <robh@kernel.org> wrote:
>> >> > +{
>> >> > +       int i, ret = 0;
>> >> > +       struct drm_gem_object *obj;
>> >> > +
>> >> > +       spin_lock(&filp->table_lock);
>> >> > +
>> >> > +       for (i = 0; i < count; i++) {
>> >> > +               /* Check if we currently have a reference on the object */
>> >> > +               obj = idr_find(&filp->object_idr, handle[i]);
>> >> > +               if (!obj) {
>> >> > +                       ret = -ENOENT;
>> >
>> > Unwind previous drm_gem_object_get(), the caller has no idea how many
>> > were processed before the error.
>>
>> I had the same thought, but the pattern we have is that you're loading
>> into a refcounted struct that will free the BOs when you're done,
>> anyway.
>
> The real bug here is if allocation of the array fails. The BO array
> may be NULL when the count is not. So this V3D cleanup hunk:
>
> for (i = 0; i < exec->bo_count; i++)
>   drm_gem_object_put_unlocked(&exec->bo[i]->base.base);
>   kvfree(exec->bo);
>
> Needs to be wrapped with 'if (exec->bo)'. We have a similar problem
> with fence arrays too.

Yeah, seems legit.  Not really going to write new patches when I've got
month-old critical patches I can't get acked, though. :/