2021-11-04 22:27:03

by Rob Clark

[permalink] [raw]
Subject: [PATCH] drm/msm/devfreq: Fix OPP refcnt leak

From: Rob Clark <[email protected]>

Reported-by: Douglas Anderson <[email protected]>
Fixes: 9bc95570175a ("drm/msm: Devfreq tuning")
Signed-off-by: Rob Clark <[email protected]>
---
drivers/gpu/drm/msm/msm_gpu_devfreq.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/msm/msm_gpu_devfreq.c b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
index d32b729b4616..9bf8600b6eea 100644
--- a/drivers/gpu/drm/msm/msm_gpu_devfreq.c
+++ b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
@@ -20,8 +20,6 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
struct msm_gpu *gpu = dev_to_gpu(dev);
struct dev_pm_opp *opp;

- opp = devfreq_recommended_opp(dev, freq, flags);
-
/*
* If the GPU is idle, devfreq is not aware, so just ignore
* it's requests
@@ -31,6 +29,8 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
return 0;
}

+ opp = devfreq_recommended_opp(dev, freq, flags);
+
if (IS_ERR(opp))
return PTR_ERR(opp);

--
2.31.1


2021-11-05 01:06:46

by Doug Anderson

[permalink] [raw]
Subject: Re: [PATCH] drm/msm/devfreq: Fix OPP refcnt leak

Hi,

On Thu, Nov 4, 2021 at 3:23 PM Rob Clark <[email protected]> wrote:
>
> From: Rob Clark <[email protected]>
>
> Reported-by: Douglas Anderson <[email protected]>
> Fixes: 9bc95570175a ("drm/msm: Devfreq tuning")
> Signed-off-by: Rob Clark <[email protected]>
> ---
> drivers/gpu/drm/msm/msm_gpu_devfreq.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)

Reviewed-by: Douglas Anderson <[email protected]>

2021-11-05 06:17:28

by Steev Klimaszewski

[permalink] [raw]
Subject: Re: [PATCH] drm/msm/devfreq: Fix OPP refcnt leak


On 11/4/21 5:28 PM, Rob Clark wrote:
> From: Rob Clark <[email protected]>
>
> Reported-by: Douglas Anderson <[email protected]>
> Fixes: 9bc95570175a ("drm/msm: Devfreq tuning")
> Signed-off-by: Rob Clark <[email protected]>
> ---
> drivers/gpu/drm/msm/msm_gpu_devfreq.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/msm/msm_gpu_devfreq.c b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
> index d32b729b4616..9bf8600b6eea 100644
> --- a/drivers/gpu/drm/msm/msm_gpu_devfreq.c
> +++ b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
> @@ -20,8 +20,6 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
> struct msm_gpu *gpu = dev_to_gpu(dev);
> struct dev_pm_opp *opp;
>
> - opp = devfreq_recommended_opp(dev, freq, flags);
> -
> /*
> * If the GPU is idle, devfreq is not aware, so just ignore
> * it's requests
> @@ -31,6 +29,8 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
> return 0;
> }
>
> + opp = devfreq_recommended_opp(dev, freq, flags);
> +
> if (IS_ERR(opp))
> return PTR_ERR(opp);
>

Testing this here on the Lenovo Yoga C630, and I'm starting to see in my
dmesg output

[   36.337061] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[   36.388122] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[   36.810941] wcd934x-codec wcd934x-codec.3.auto: Port Closed RX port
1, value 4
[   36.811914] wcd934x-codec wcd934x-codec.3.auto: Port Closed RX port
2, value 4
[  198.794946] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  198.845698] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  502.285421] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  502.339427] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  503.361469] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  503.412757] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  503.871480] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  503.922712] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  503.974474] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  504.025501] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  505.923563] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  505.974513] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  510.313052] usb 3-1.3: USB disconnect, device number 4
[  519.677148] usb 3-1.3: new high-speed USB device number 5 using xhci-hcd
[  519.793394] usb 3-1.3: New USB device found, idVendor=5986,
idProduct=2115, bcdDevice=54.20
[  519.793441] usb 3-1.3: New USB device strings: Mfr=1, Product=2,
SerialNumber=0
[  519.793472] usb 3-1.3: Product: Integrated Camera
[  519.793495] usb 3-1.3: Manufacturer: SunplusIT Inc
[  519.861020] usb 3-1.3: Found UVC 1.00 device Integrated Camera
(5986:2115)
[  519.892879] input: Integrated Camera: Integrated C as
/devices/platform/soc@0/a8f8800.usb/a800000.dwc3/xhci-hcd.1.auto/usb3/3-1/3-1.3/3-1.3:1.0/input/input27
[  520.283839] devfreq 5000000.gpu: Couldn't update frequency transition
information.
[  520.335854] devfreq 5000000.gpu: Couldn't update frequency transition
information.


Is this intended?

2021-11-05 16:44:28

by Doug Anderson

[permalink] [raw]
Subject: Re: [PATCH] drm/msm/devfreq: Fix OPP refcnt leak

Hi,

On Thu, Nov 4, 2021 at 9:32 PM Steev Klimaszewski <[email protected]> wrote:
>
>
> On 11/4/21 5:28 PM, Rob Clark wrote:
> > From: Rob Clark <[email protected]>
> >
> > Reported-by: Douglas Anderson <[email protected]>
> > Fixes: 9bc95570175a ("drm/msm: Devfreq tuning")
> > Signed-off-by: Rob Clark <[email protected]>
> > ---
> > drivers/gpu/drm/msm/msm_gpu_devfreq.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/msm/msm_gpu_devfreq.c b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
> > index d32b729b4616..9bf8600b6eea 100644
> > --- a/drivers/gpu/drm/msm/msm_gpu_devfreq.c
> > +++ b/drivers/gpu/drm/msm/msm_gpu_devfreq.c
> > @@ -20,8 +20,6 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
> > struct msm_gpu *gpu = dev_to_gpu(dev);
> > struct dev_pm_opp *opp;
> >
> > - opp = devfreq_recommended_opp(dev, freq, flags);
> > -
> > /*
> > * If the GPU is idle, devfreq is not aware, so just ignore
> > * it's requests
> > @@ -31,6 +29,8 @@ static int msm_devfreq_target(struct device *dev, unsigned long *freq,
> > return 0;
> > }
> >
> > + opp = devfreq_recommended_opp(dev, freq, flags);
> > +
> > if (IS_ERR(opp))
> > return PTR_ERR(opp);
> >
>
> Testing this here on the Lenovo Yoga C630, and I'm starting to see in my
> dmesg output
>
> [ 36.337061] devfreq 5000000.gpu: Couldn't update frequency transition
> information.
> [ 36.388122] devfreq 5000000.gpu: Couldn't update frequency transition
> information.

Ah, I think this makes sense. We're now storing a frequency which
might not match an actual "opp" and I suppose that we must return it
in some cases.

I guess a simple fix is to still call devfreq_recommended_opp() in the
idle case but just call dev_pm_opp_put() to fix the leak.

-Doug