Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Medium] drm/vkms: composer_users underflow in vkms_set_crc_source()
  on error
--

commit 22fc9fb8d10e80e408b96eee292c039a0d2a1b22
Author: Oarora Etimis <[email protected]>

drm/vkms: count the composer users

This commit replaces the boolean composer_enabled flag with a reference
count so that concurrent CRC capture and writeback jobs do not disable
the composer while the other is still active.

> diff --git a/drivers/gpu/drm/vkms/vkms_composer.c 
> b/drivers/gpu/drm/vkms/vkms_composer.c
> index 0fc915a954bad..96c40eeeb61ba 100644
> --- a/drivers/gpu/drm/vkms/vkms_composer.c
> +++ b/drivers/gpu/drm/vkms/vkms_composer.c
> @@ -711,19 +711,33 @@ int vkms_verify_crc_source(struct drm_crtc *crtc, const 
> char *src_name,
>       return 0;
>  }
>  
> -void vkms_set_composer(struct vkms_output *out, bool enabled)
> +bool vkms_composer_get(struct vkms_output *out)
>  {
> -     bool old_enabled;
> +     unsigned int users;
>  
> -     if (enabled)
> -             drm_crtc_vblank_get(&out->crtc);
> +     if (drm_crtc_vblank_get(&out->crtc))
> +             return false;
> +
> +     spin_lock_irq(&out->lock);
> +     users = out->composer_users++;
> +     spin_unlock_irq(&out->lock);
> +
> +     /* All users share a single vblank reference. */
> +     if (users)
> +             drm_crtc_vblank_put(&out->crtc);
> +
> +     return true;
> +}
> +
> +void vkms_composer_put(struct vkms_output *out)
> +{
> +     unsigned int users;
>  
>       spin_lock_irq(&out->lock);
> -     old_enabled = out->composer_enabled;
> -     out->composer_enabled = enabled;
> +     users = --out->composer_users;
>       spin_unlock_irq(&out->lock);
>  
> -     if (old_enabled)
> +     if (!users)
>               drm_crtc_vblank_put(&out->crtc);
>  }
>  
> @@ -735,7 +749,10 @@ int vkms_set_crc_source(struct drm_crtc *crtc, const 
> char *src_name)
>  
>       ret = vkms_crc_parse_source(src_name, &enabled);
>  
> -     vkms_set_composer(out, enabled);
> +     if (enabled)
> +             vkms_composer_get(out);
> +     else
> +             vkms_composer_put(out);
>  
>       return ret;
>  }

[Severity: Medium]
Can this underflow out->composer_users when vkms_composer_get() or
vkms_crc_parse_source() fails?

crtc_crc_open() drops crtc->mutex after checking crtc->state->active and
before calling set_crc_source().  If the CRTC is disabled in that window,
drm_crtc_vblank_get() fails and vkms_composer_get() returns false without
incrementing out->composer_users.  Because vkms_set_crc_source() ignores
that failure and returns 0, crtc_crc_open() succeeds and the matching
crtc_crc_release() later calls vkms_set_crc_source(crtc, NULL) ->
vkms_composer_put(), decrementing out->composer_users without a prior
increment.  This either wraps out->composer_users to UINT_MAX (leaving the
composer permanently active and causing later vkms_composer_get() calls to
drop their vblank reference immediately) or prematurely drops a concurrent
writeback job's composer and vblank reference.

Also, if vkms_crc_parse_source() returns -EINVAL, enabled is false and
vkms_set_crc_source() still calls vkms_composer_put().

-- 
Sashiko AI review ยท 
https://sashiko.dev/#/patchset/[email protected]?part=1

Reply via email to