2020-09-17 22:45:54

by Stephen Boyd

Subject: [PATCH] drm/msm/dp: Sleep properly in dp_hpd_handler kthread

We shouldn't be waiting for an event here with a timeout of 100ms when
we're not in the 'timeout' arm of the if condition. Instead we should be
sleeping in the interruptible state (S) until something happens and we
need to wakeup. Right now this kthread is running almost all the time
because it sleeps for 100ms, wakes up, sees there's nothing to do, and
then starts the process all over again. Looking at top it shows up in
the D state (uninterruptible) because it uses wait_event_timeout(). FIx
this up.

Cc: Tanmay Shah <[email protected]>
Cc: Kuogee Hsieh <[email protected]>
Reported-by: Douglas Anderson <[email protected]>
Fixes: 8ede2ecc3e5e ("drm/msm/dp: Add DP compliance tests on Snapdragon Chipsets")
Signed-off-by: Stephen Boyd <[email protected]>
---

Based on msm-next-dp of https://gitlab.freedesktop.org/drm/msm.git

drivers/gpu/drm/msm/dp/dp_display.c | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/msm/dp/dp_display.c b/drivers/gpu/drm/msm/dp/dp_display.c
index 05a97e097edf..e175aa3fd3a9 100644
--- a/drivers/gpu/drm/msm/dp/dp_display.c
+++ b/drivers/gpu/drm/msm/dp/dp_display.c
@@ -970,9 +970,8 @@ static int hpd_event_thread(void *data)
(dp_priv->event_pndx == dp_priv->event_gndx),
EVENT_TIMEOUT);
} else {
- wait_event_timeout(dp_priv->event_q,
- (dp_priv->event_pndx != dp_priv->event_gndx),
- EVENT_TIMEOUT);
+ wait_event_interruptible(dp_priv->event_q,
+ (dp_priv->event_pndx != dp_priv->event_gndx));
}
spin_lock_irqsave(&dp_priv->event_lock, flag);
todo = &dp_priv->event_list[dp_priv->event_gndx];

base-commit: 937f941ca06f2f3ab64baebf31be2c16d57ae7b8
--
Sent by a computer, using git, on the internet

2020-09-17 23:24:26

by Kuogee Hsieh

Subject: Re: [PATCH] drm/msm/dp: Sleep properly in dp_hpd_handler kthread

On 2020-09-17 15:44, Stephen Boyd wrote:
> We shouldn't be waiting for an event here with a timeout of 100ms when
> we're not in the 'timeout' arm of the if condition. Instead we should
> be
> sleeping in the interruptible state (S) until something happens and we
> need to wakeup. Right now this kthread is running almost all the time
> because it sleeps for 100ms, wakes up, sees there's nothing to do, and
> then starts the process all over again. Looking at top it shows up in
> the D state (uninterruptible) because it uses wait_event_timeout(). FIx
> this up.
>
> Cc: Tanmay Shah <[email protected]>
> Cc: Kuogee Hsieh <[email protected]>
> Reported-by: Douglas Anderson <[email protected]>
> Fixes: 8ede2ecc3e5e ("drm/msm/dp: Add DP compliance tests on
> Snapdragon Chipsets")
> Signed-off-by: Stephen Boyd <[email protected]>
Reviewed-by: Kuogee Hsieh <[email protected]>
> ---
>
> Based on msm-next-dp of https://gitlab.freedesktop.org/drm/msm.git
>
> drivers/gpu/drm/msm/dp/dp_display.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/msm/dp/dp_display.c
> b/drivers/gpu/drm/msm/dp/dp_display.c
> index 05a97e097edf..e175aa3fd3a9 100644
> --- a/drivers/gpu/drm/msm/dp/dp_display.c
> +++ b/drivers/gpu/drm/msm/dp/dp_display.c
> @@ -970,9 +970,8 @@ static int hpd_event_thread(void *data)
> (dp_priv->event_pndx == dp_priv->event_gndx),
> EVENT_TIMEOUT);
> } else {
> - wait_event_timeout(dp_priv->event_q,
> - (dp_priv->event_pndx != dp_priv->event_gndx),
> - EVENT_TIMEOUT);
> + wait_event_interruptible(dp_priv->event_q,
> + (dp_priv->event_pndx != dp_priv->event_gndx));
> }
> spin_lock_irqsave(&dp_priv->event_lock, flag);
> todo = &dp_priv->event_list[dp_priv->event_gndx];
>
> base-commit: 937f941ca06f2f3ab64baebf31be2c16d57ae7b8