[lvc-project] [PATCH] drm/amdgpu: fix NULL dereference in gfx_v9_0_kcq() and kiq_init_queue()

[Alex Deucher]

On Wed, Jun 4, 2025 at 3:30 PM SDL <sdl at nppct.ru> wrote:
>
>
> > On Sat, May 24, 2025 at 2:14 AM Alexey Nepomnyashih <sdl at nppct.ru> wrote:
> >> A potential NULL pointer dereference may occur when accessing
> >> tmp_mqd->cp_hqd_pq_control without verifying that tmp_mqd is non-NULL.
> >> This may happen if mqd_backup[mqd_idx] is unexpectedly NULL.
> >>
> >> Although a NULL check for mqd_backup[mqd_idx] existed previously, it was
> >> moved to a position after the dereference in a recent commit, which
> >> renders it ineffective.
> > I don't think it's possible for mqd_backup to be NULL at this point.
> > We would have failed earlier in init if the mqd backup allocation
> > failed.
> >
> > Alex
> In scenarios such as GPU reset or power management resume, there is no
> strict
> guarantee that amdgpu_gfx_mqd_sw_init() (via ->sw_init()) is invoked before
> gfx_v9_0_kiq_init_queue(). As a result, mqd_backup[] may remain
> uninitialized,
> and dereferencing it without a NULL check can lead to a crash.
>
> Most other uses of mqd_backup[] in the driver explicitly check for NULL,
> indicating that uninitialized entries are an expected condition and
> should be handled
> accordingly.

sw_init() is only called once at driver load time.  everything is
allocated at that point.  If that fails, the driver would not have
loaded in the first place.  I don't think it's possible for it to be
NULL.

Alex

>
> Alexey