Re: [PATCH] drm/i915/gvt: Fix NULL pointer dereference in slot switch
From: Tuo Li
Date: Tue Nov 17 2026 - 06:48:08 EST
On Tue, Nov 17, 2026 at 9:01 AM Zhenyu Wang <zhenyuw.linux@xxxxxxxxx> wrote:
>
> On Thu, Nov 12, 2026 at 11:58:22PM +0800, Tuo Li wrote:
> > The running slot is updated to the pending slot before updating the running
> > context. However, the condition checks the old running context instead of
> > the newly assigned running slot.
> >
> > When no pending slot exists, running_slot becomes NULL. If the previous
> > running_context is non-NULL, the code attempts to access
> > running_slot->ctx[0], resulting in a NULL pointer dereference.
> >
> > To fix this possible NULL pointer dereference, check running_slot instead
> > to ensure that running_context is updated only when a valid slot is
> > available.
>
> Well, just AI generated static check but without any runtime analysis,
> again...and can't even show kernel oops of this, as it simply can't happen..
>
> >
> > Fixes: 8453d674ae7e ("drm/i915/gvt: vGPU execlist virtualization")
> > Signed-off-by: Tuo Li <islituo@xxxxxxxxx>
> > ---
> > drivers/gpu/drm/i915/gvt/execlist.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/i915/gvt/execlist.c b/drivers/gpu/drm/i915/gvt/execlist.c
> > index 29147a9f162e..763c310f2aaf 100644
> > --- a/drivers/gpu/drm/i915/gvt/execlist.c
> > +++ b/drivers/gpu/drm/i915/gvt/execlist.c
> > @@ -73,7 +73,7 @@ static void switch_virtual_execlist_slot(struct intel_vgpu_execlist *execlist)
> >
> > execlist->running_slot = execlist->pending_slot;
> > execlist->pending_slot = NULL;
> > - execlist->running_context = execlist->running_context ?
> > + execlist->running_context = execlist->running_slot ?
> > &execlist->running_slot->ctx[0] : NULL;
> >
>
> Current code calls this switch in case whenever running_slot is
> available, so it won't have null ptr reference. But code looks a
> little weird, as current code tried to assign running_context in other
> place e.g when new request happened, or even it might seem to miss
> when enter execlist idle state...but it is still valid in switch
> call. So I think this is just harmless. If you'd like for further
> cleanup, that needs testing result for sure.
Thanks for the clarification. I will drop this patch. Sorry for the noise.
Sincerely,
Tuo Li