On Sun, Sep 28, 2025 at 1:49 AM Pavan Bobba <[email protected]> wrote:
>
> Replace the open-coded polling with schedule() in vmw_fallback_wait()
> by schedule_hrtimeout(). The old code wakes up at jiffy granularity and
> leads to unnecessary CPU wakeups during fence waits.
>
> schedule_hrtimeout() provides high-resolution sleep with finer control,
> reducing CPU utilization without affecting fence correctness. For the
> non-interruptible case, use schedule_timeout_uninterruptible().
>
> Signed-off-by: Pavan Bobba <[email protected]>
> ---
>  drivers/gpu/drm/vmwgfx/vmwgfx_irq.c | 16 ++++++----------
>  1 file changed, 6 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c 
> b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
> index 05773eb394d3..64045b0efafc 100644
> --- a/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
> +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_irq.c
> @@ -202,16 +202,12 @@ int vmw_fallback_wait(struct vmw_private *dev_priv,
>                 if (lazy)
>                         schedule_timeout(1);
>                 else if ((++count & 0x0F) == 0) {
> -                       /**
> -                        * FIXME: Use schedule_hr_timeout here for
> -                        * newer kernels and lower CPU utilization.
> -                        */
> -
> -                       __set_current_state(TASK_RUNNING);
> -                       schedule();
> -                       __set_current_state((interruptible) ?
> -                                           TASK_INTERRUPTIBLE :
> -                                           TASK_UNINTERRUPTIBLE);
> +                       ktime_t delta = ktime_set(0, NSEC_PER_MSEC);
> +
> +                       if (interruptible)
> +                               schedule_hrtimeout(&delta, HRTIMER_MODE_REL);
> +                       else
> +                               schedule_timeout_uninterruptible(delta);
>                 }
>                 if (interruptible && signal_pending(current)) {
>                         ret = -ERESTARTSYS;
> --

I don't remember exactly the schedule family of functions but isn't
schedule_hrtimeout leaving the task in a running state? In general it
looks like with the patch the task's current state doesn't match what
was expected, plus I'm not sure if I quite get why the uninterruptible
non-lazy case is being replaced with a lazy wait of NSEC_PER_MSEC's.
It'd be great if you could explain a little bit better what you're
doing here because the commit message is missing an explanation for
either of those.

z

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to