On Tue, 11 Aug 2026 16:33:18 +0200
Osama Abdelkader <[email protected]> wrote:
> CS interrupts are reported through firmware-provided bits in the CSG
> interface. sched_process_csg_irq_locked() iterates over those bits and
> passes each CS ID to cs_slot_process_irq_locked().
>
> panthor_fw_get_cs_iface() only bounds the CS ID against MAX_CS_PER_CSG,
> while the firmware can expose fewer CS slots. If an IRQ bit is reported
> for a slot beyond sched->cs_slot_count, the driver can fetch a
> zero-initialized interface entry and dereference cs_iface->input.
>
> Ignore IRQs for slots that were not exposed by the firmware. The CSG IRQ
> acknowledgment is already updated from cs_irq_req before processing the
> individual CS IRQs, so ignored bits are still acknowledged.
>
> Fixes: de8548813824 ("drm/panthor: Add the scheduler logical block")
> Cc: [email protected]
> Signed-off-by: Osama Abdelkader <[email protected]>
> ---
> drivers/gpu/drm/panthor/panthor_sched.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/gpu/drm/panthor/panthor_sched.c
> b/drivers/gpu/drm/panthor/panthor_sched.c
> index 64749750e6ee..1e230936e99c 100644
> --- a/drivers/gpu/drm/panthor/panthor_sched.c
> +++ b/drivers/gpu/drm/panthor/panthor_sched.c
> @@ -1705,6 +1705,9 @@ static bool cs_slot_process_irq_locked(struct
> panthor_device *ptdev,
>
> lockdep_assert_held(&ptdev->scheduler->lock);
>
> + if (cs_id >= ptdev->scheduler->cs_slot_count)
if (drm_WARN_ON_ONCE(&ptdev->base, cs_id >=
ptdev->scheduler->cs_slot_count))
Since this is certainly not supposed to happen. Looks good otherwise.
Reviewed-by: Boris Brezillon <[email protected]>
> + return false;
> +
> cs_iface = panthor_fw_get_cs_iface(ptdev, csg_id, cs_id);
> req = cs_iface->input->req;
> ack = cs_iface->output->ack;
