On Wed, May 13, 2026 at 1:42 AM Boris Brezillon
<[email protected]> wrote:
>
> On Tue, 12 May 2026 14:55:30 -0700
> Chia-I Wu <[email protected]> wrote:
>
> > On Tue, May 12, 2026 at 4:54 AM Boris Brezillon
> > <[email protected]> wrote:
> > >
> > > Rather than assuming an interrupt is always expected for request
> > > acks, temporarily enable the relevant interrupts when the polling-wait
> > > failed. This should hopefully reduce the number of interrupts the CPU
> > > has to process.
> > >
> > > Signed-off-by: Boris Brezillon <[email protected]>
> > WIth minor comments below, Reviewed-by: Chia-I Wu <[email protected]>
> > > ---
> > > drivers/gpu/drm/panthor/panthor_fw.c | 34
> > > +++++++++++++++++++--------------
> > > drivers/gpu/drm/panthor/panthor_sched.c | 5 +++--
> > > 2 files changed, 23 insertions(+), 16 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/panthor/panthor_fw.c
> > > b/drivers/gpu/drm/panthor/panthor_fw.c
> > > index 8239a6951569..f5e0ceca4130 100644
> > > --- a/drivers/gpu/drm/panthor/panthor_fw.c
> > > +++ b/drivers/gpu/drm/panthor/panthor_fw.c
> > > @@ -1039,16 +1039,10 @@ static void panthor_fw_init_global_iface(struct
> > > panthor_device *ptdev)
> > > glb_iface->input->progress_timer = PROGRESS_TIMEOUT_CYCLES >>
> > > PROGRESS_TIMEOUT_SCALE_SHIFT;
> > > glb_iface->input->idle_timer = panthor_fw_conv_timeout(ptdev,
> > > IDLE_HYSTERESIS_US);
> > >
> > > - /* Enable interrupts we care about. */
> > > - glb_iface->input->ack_irq_mask = GLB_CFG_ALLOC_EN |
> > > - GLB_PING |
> > > - GLB_CFG_PROGRESS_TIMER |
> > > - GLB_CFG_POWEROFF_TIMER |
> > > - GLB_IDLE_EN |
> > > - GLB_IDLE;
> > > -
> > > - if (panthor_fw_has_glb_state(ptdev))
> > > - glb_iface->input->ack_irq_mask |= GLB_STATE_MASK;
> > > + /* Enable interrupts for asynchronous events that are not
> > > + * triggered by request acks.
> > > + */
> > > + glb_iface->input->ack_irq_mask = GLB_IDLE;
> > We should static_assert or & with GLB_EVT_MASK. Same for CSG and CS.
>
> Yep, good idea, I'll add a static_assert() in all places where
> ->ack_irq_mask is set.
>
> >
> > >
> > > panthor_fw_update_reqs(glb_iface, req, GLB_IDLE_EN |
> > > GLB_COUNTER_EN,
> > > GLB_IDLE_EN | GLB_COUNTER_EN);
> > > @@ -1318,8 +1312,8 @@ void panthor_fw_unplug(struct panthor_device *ptdev)
> > > * Return: 0 on success, -ETIMEDOUT otherwise.
> > > */
> > > static int panthor_fw_wait_acks(const u32 *req_ptr, const u32 *ack_ptr,
> > > - wait_queue_head_t *wq,
> > > - u32 req_mask, u32 *acked,
> > > + u32 *ack_irq_mask_ptr, spinlock_t *lock,
> > > + wait_queue_head_t *wq, u32 req_mask, u32
> > > *acked,
> > > u32 timeout_ms)
> > > {
> > > u32 ack, req = READ_ONCE(*req_ptr) & req_mask;
> > > @@ -1334,8 +1328,16 @@ static int panthor_fw_wait_acks(const u32
> > > *req_ptr, const u32 *ack_ptr,
> > > if (!ret)
> > > return 0;
> > >
> > > - if (wait_event_timeout(*wq, (READ_ONCE(*ack_ptr) & req_mask) ==
> > > req,
> > > - msecs_to_jiffies(timeout_ms)))
> > > + scoped_guard(spinlock_irqsave, lock)
> > > + *ack_irq_mask_ptr |= req_mask;
> > > +
> > > + ret = wait_event_timeout(*wq, (READ_ONCE(*ack_ptr) & req_mask) ==
> > > req,
> > > + msecs_to_jiffies(timeout_ms));
> > > +
> > > + scoped_guard(spinlock_irqsave, lock)
> > > + *ack_irq_mask_ptr &= ~req_mask;
> > We should add a comment saying that this is safe because
> > {GLB,CSG,CS}_REQ_MASK and {GLB,CSG,CS}_EVT_MASK are disjoint, and thus
> > req_mask and ack_irq_mask are disjoint.
>
> You mean the ack_irq_mask set at init time? Because
> xxx_iface->input->ack_irq_mask is moving target now.
>
> Well, if we expand on safety matters, I'd say none of this is safe
> since it relies on the caller knowing what it does and passing a valid
> req_mask. But I'll add a comment mentioning that the original
> ack_irq_mask shouldn't intersect with any of the bits that might be set
> in req_mask (that's basically the static_assert() you suggested).
The callers of panthor_fw_wait_acks validate req_mask. With a
static_assert on ack_irq_mask at init, we are sure the two masks are
disjoint upon entry. We just need a comment explaining why
save-and-restore is unnecessary.
