This got me confused for a bit while looking over this code: I had been
planning on adding some blocking function calls into this function, but
seeing the irqsave/irqrestore variants of spin_(un)lock() didn't make it
very clear whether or not that would actually be safe.
So I went ahead and reviewed every single driver in the kernel that uses
this function, and they all fall into three categories:
* Driver probe code
* ->atomic_disable() callbacks
* Legacy modesetting callbacks
All of these will be guaranteed to have IRQs enabled, which means it's
perfectly safe to block here. Just to make things a little less
confusing to others in the future, let's switch over to
spin_lock_irq()/spin_unlock_irq() to make that fact a little more
obvious.
Signed-off-by: Lyude Paul <[email protected]>
Cc: Daniel Vetter <[email protected]>
Cc: Ville Syrjälä <[email protected]>
---
drivers/gpu/drm/drm_vblank.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
index ce5c1e1d29963..e895f5331fdb4 100644
--- a/drivers/gpu/drm/drm_vblank.c
+++ b/drivers/gpu/drm/drm_vblank.c
@@ -1283,13 +1283,12 @@ void drm_crtc_vblank_off(struct drm_crtc *crtc)
struct drm_pending_vblank_event *e, *t;
ktime_t now;
- unsigned long irqflags;
u64 seq;
if (drm_WARN_ON(dev, pipe >= dev->num_crtcs))
return;
- spin_lock_irqsave(&dev->event_lock, irqflags);
+ spin_lock_irq(&dev->event_lock);
spin_lock(&dev->vbl_lock);
drm_dbg_vbl(dev, "crtc %d, vblank enabled %d, inmodeset %d\n",
@@ -1325,7 +1324,7 @@ void drm_crtc_vblank_off(struct drm_crtc *crtc)
drm_vblank_put(dev, pipe);
send_vblank_event(dev, e, seq, now);
}
- spin_unlock_irqrestore(&dev->event_lock, irqflags);
+ spin_unlock_irq(&dev->event_lock);
/* Will be reset by the modeset helpers when re-enabling the crtc by
* calling drm_calc_timestamping_constants(). */
--
2.26.2
On Wed, Jun 24, 2020 at 07:03:09PM -0400, Lyude Paul wrote:
> This got me confused for a bit while looking over this code: I had been
> planning on adding some blocking function calls into this function, but
> seeing the irqsave/irqrestore variants of spin_(un)lock() didn't make it
> very clear whether or not that would actually be safe.
>
> So I went ahead and reviewed every single driver in the kernel that uses
> this function, and they all fall into three categories:
>
> * Driver probe code
> * ->atomic_disable() callbacks
> * Legacy modesetting callbacks
>
> All of these will be guaranteed to have IRQs enabled, which means it's
> perfectly safe to block here. Just to make things a little less
> confusing to others in the future, let's switch over to
> spin_lock_irq()/spin_unlock_irq() to make that fact a little more
> obvious.
>
> Signed-off-by: Lyude Paul <[email protected]>
> Cc: Daniel Vetter <[email protected]>
> Cc: Ville Syrj?l? <[email protected]>
I think the patch is correct, but now we're having a bit a inconsistency,
since all other functions where the same applies still use _irqsave. I
looked through the file and I think drm_vblank_get, drm_crtc_vblank_reset,
drm_crtc_vblank_on and drm_legacy_vblank_post_modeset,
drm_queue_vblank_event and drm_crtc_queue_sequence_ioctl are all candiates
for the same cleanup.
Maybe follow up patches for less confusion?
On this:
Reviewed-by: Daniel Vetter <[email protected]>
> ---
> drivers/gpu/drm/drm_vblank.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_vblank.c b/drivers/gpu/drm/drm_vblank.c
> index ce5c1e1d29963..e895f5331fdb4 100644
> --- a/drivers/gpu/drm/drm_vblank.c
> +++ b/drivers/gpu/drm/drm_vblank.c
> @@ -1283,13 +1283,12 @@ void drm_crtc_vblank_off(struct drm_crtc *crtc)
> struct drm_pending_vblank_event *e, *t;
>
> ktime_t now;
> - unsigned long irqflags;
> u64 seq;
>
> if (drm_WARN_ON(dev, pipe >= dev->num_crtcs))
> return;
>
> - spin_lock_irqsave(&dev->event_lock, irqflags);
> + spin_lock_irq(&dev->event_lock);
>
> spin_lock(&dev->vbl_lock);
> drm_dbg_vbl(dev, "crtc %d, vblank enabled %d, inmodeset %d\n",
> @@ -1325,7 +1324,7 @@ void drm_crtc_vblank_off(struct drm_crtc *crtc)
> drm_vblank_put(dev, pipe);
> send_vblank_event(dev, e, seq, now);
> }
> - spin_unlock_irqrestore(&dev->event_lock, irqflags);
> + spin_unlock_irq(&dev->event_lock);
>
> /* Will be reset by the modeset helpers when re-enabling the crtc by
> * calling drm_calc_timestamping_constants(). */
> --
> 2.26.2
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch