[PATCH v2 0/2] Fix usage counter leak by adding a general sync ops

[PATCH v2 0/2] Fix usage counter leak by adding a general sync ops

[Zhang Qilong]

In many case, we need to check return value of pm_runtime_get_sync,
but it brings a trouble to the usage counter processing. Many callers
forget to decrease the usage counter when it failed. It has been
discussed a lot[0][1]. So we add a function to deal with the usage
counter for better coding and view. Then, we replace pm_runtime_get_sync
with it in fec_main.c

Zhang Qilong (2):
  PM: runtime: Add a general runtime get sync operation to deal with
    usage counter
  net: fec: Fix reference count leak in fec series ops

 drivers/net/ethernet/freescale/fec_main.c | 12 ++++-----
 include/linux/pm_runtime.h                | 30 +++++++++++++++++++++++
 2 files changed, 35 insertions(+), 7 deletions(-)

-- 
2.25.4

[PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Zhang Qilong]

In many case, we need to check return value of pm_runtime_get_sync, but
it brings a trouble to the usage counter processing. Many callers forget
to decrease the usage counter when it failed. It has been discussed a
lot[0][1]. So we add a function to deal with the usage counter for better
coding.

[0]https://lkml.org/lkml/2020/6/14/88
[1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520095148.10995-1-dinghao.liu@zju.edu.cn/
Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
---
 include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
index 4b708f4e8eed..6549ce764400 100644
--- a/include/linux/pm_runtime.h
+++ b/include/linux/pm_runtime.h
@@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device *dev)
 	return __pm_runtime_resume(dev, RPM_GET_PUT);
 }
 
+/**
+ * pm_runtime_general_get - Bump up usage counter of a device and resume it.
+ * @dev: Target device.
+ *
+ * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
+ * of it synchronously. If __pm_runtime_resume return negative value(device is in
+ * error state), we to need decrease the usage counter before it return. If
+ * __pm_runtime_resume return positive value, it means the runtime of device has
+ * already been in active state, and we let the new wrapper return zero instead.
+ *
+ * The possible return values of this function is zero or negative value.
+ * zero:
+ *    - it means resume succeeed or runtime of device has already been active, the
+ *      runtime PM usage counter of @dev remains incremented.
+ * negative:
+ *    - it means failure and the runtime PM usage counter of @dev has been balanced.
+ */
+static inline int pm_runtime_general_get(struct device *dev)
+{
+	int ret = 0;
+
+	ret = __pm_runtime_resume(dev, RPM_GET_PUT);
+	if (ret < 0) {
+		pm_runtime_put_noidle(dev);
+		return ret;
+	}
+
+	return 0;
+}
+
 /**
  * pm_runtime_put - Drop device usage counter and queue up "idle check" if 0.
  * @dev: Target device.
-- 
2.25.4

[PATCH v2 2/2] net: fec: Fix reference count leak in fec series ops

[Zhang Qilong]

pm_runtime_get_sync() will increment pm usage at first and it will
resume the device later. If runtime of the device has error or
device is in inaccessible state(or other error state), resume
operation will fail. If we do not call put operation to decrease
the reference, it will result in reference count leak. Moreover,
this device cannot enter the idle state and always stay busy or other
non-idle state later. So we fixed it by replacing it with
pm_runtime_general_get.

Fixes: 8fff755e9f8d0 ("net: fec: Ensure clocks are enabled while using mdio bus")
Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
---
 drivers/net/ethernet/freescale/fec_main.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index d7919555250d..695720f8263f 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c
@@ -1808,7 +1808,7 @@ static int fec_enet_mdio_read(struct mii_bus *bus, int mii_id, int regnum)
 	int ret = 0, frame_start, frame_addr, frame_op;
 	bool is_c45 = !!(regnum & MII_ADDR_C45);
 
-	ret = pm_runtime_get_sync(dev);
+	ret = pm_runtime_general_get(dev);
 	if (ret < 0)
 		return ret;
 
@@ -1867,11 +1867,9 @@ static int fec_enet_mdio_write(struct mii_bus *bus, int mii_id, int regnum,
 	int ret, frame_start, frame_addr;
 	bool is_c45 = !!(regnum & MII_ADDR_C45);
 
-	ret = pm_runtime_get_sync(dev);
+	ret = pm_runtime_general_get(dev);
 	if (ret < 0)
 		return ret;
-	else
-		ret = 0;
 
 	if (is_c45) {
 		frame_start = FEC_MMFR_ST_C45;
@@ -2275,7 +2273,7 @@ static void fec_enet_get_regs(struct net_device *ndev,
 	u32 i, off;
 	int ret;
 
-	ret = pm_runtime_get_sync(dev);
+	ret = pm_runtime_general_get(dev);
 	if (ret < 0)
 		return;
 
@@ -2976,7 +2974,7 @@ fec_enet_open(struct net_device *ndev)
 	int ret;
 	bool reset_again;
 
-	ret = pm_runtime_get_sync(&fep->pdev->dev);
+	ret = pm_runtime_general_get(&fep->pdev->dev);
 	if (ret < 0)
 		return ret;
 
@@ -3770,7 +3768,7 @@ fec_drv_remove(struct platform_device *pdev)
 	struct device_node *np = pdev->dev.of_node;
 	int ret;
 
-	ret = pm_runtime_get_sync(&pdev->dev);
+	ret = pm_runtime_general_get(&pdev->dev);
 	if (ret < 0)
 		return ret;
 
-- 
2.25.4

Re: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Rafael J. Wysocki]

On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong <zhangqilong3@huawei.com> wrote:
>
> In many case, we need to check return value of pm_runtime_get_sync, but
> it brings a trouble to the usage counter processing. Many callers forget
> to decrease the usage counter when it failed. It has been discussed a
> lot[0][1]. So we add a function to deal with the usage counter for better
> coding.
>
> [0]https://lkml.org/lkml/2020/6/14/88
> [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520095148.10995-1-dinghao.liu@zju.edu.cn/
> Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> ---
>  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
>  1 file changed, 30 insertions(+)
>
> diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> index 4b708f4e8eed..6549ce764400 100644
> --- a/include/linux/pm_runtime.h
> +++ b/include/linux/pm_runtime.h
> @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device *dev)
>         return __pm_runtime_resume(dev, RPM_GET_PUT);
>  }
>
> +/**
> + * pm_runtime_general_get - Bump up usage counter of a device and resume it.
> + * @dev: Target device.
> + *
> + * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
> + * of it synchronously. If __pm_runtime_resume return negative value(device is in
> + * error state), we to need decrease the usage counter before it return. If
> + * __pm_runtime_resume return positive value, it means the runtime of device has
> + * already been in active state, and we let the new wrapper return zero instead.
> + *
> + * The possible return values of this function is zero or negative value.
> + * zero:
> + *    - it means resume succeeed or runtime of device has already been active, the
> + *      runtime PM usage counter of @dev remains incremented.
> + * negative:
> + *    - it means failure and the runtime PM usage counter of @dev has been balanced.

The kerneldoc above is kind of noisy and it is hard to figure out what
the helper really does from it.

You could basically say something like "Resume @dev synchronously and
if that is successful, increment its runtime PM usage counter.  Return
0 if the runtime PM usage counter of @dev has been incremented or a
negative error code otherwise."

> + */
> +static inline int pm_runtime_general_get(struct device *dev)

What about pm_runtime_resume_and_get()?

> +{
> +       int ret = 0;

This extra initialization is not necessary.

You can initialize ret to the __pm_runtime_resume() return value right away.

> +
> +       ret = __pm_runtime_resume(dev, RPM_GET_PUT);
> +       if (ret < 0) {
> +               pm_runtime_put_noidle(dev);
> +               return ret;
> +       }
> +
> +       return 0;
> +}
> +
>  /**
>   * pm_runtime_put - Drop device usage counter and queue up "idle check" if 0.
>   * @dev: Target device.
> --

Re: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Ulf Hansson]

On Mon, 9 Nov 2020 at 16:20, Rafael J. Wysocki <rafael@kernel.org> wrote:
>
> On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong <zhangqilong3@huawei.com> wrote:
> >
> > In many case, we need to check return value of pm_runtime_get_sync, but
> > it brings a trouble to the usage counter processing. Many callers forget
> > to decrease the usage counter when it failed. It has been discussed a
> > lot[0][1]. So we add a function to deal with the usage counter for better
> > coding.
> >
> > [0]https://lkml.org/lkml/2020/6/14/88
> > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520095148.10995-1-dinghao.liu@zju.edu.cn/
> > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > ---
> >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> >  1 file changed, 30 insertions(+)
> >
> > diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> > index 4b708f4e8eed..6549ce764400 100644
> > --- a/include/linux/pm_runtime.h
> > +++ b/include/linux/pm_runtime.h
> > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device *dev)
> >         return __pm_runtime_resume(dev, RPM_GET_PUT);
> >  }
> >
> > +/**
> > + * pm_runtime_general_get - Bump up usage counter of a device and resume it.
> > + * @dev: Target device.
> > + *
> > + * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
> > + * of it synchronously. If __pm_runtime_resume return negative value(device is in
> > + * error state), we to need decrease the usage counter before it return. If
> > + * __pm_runtime_resume return positive value, it means the runtime of device has
> > + * already been in active state, and we let the new wrapper return zero instead.
> > + *
> > + * The possible return values of this function is zero or negative value.
> > + * zero:
> > + *    - it means resume succeeed or runtime of device has already been active, the
> > + *      runtime PM usage counter of @dev remains incremented.
> > + * negative:
> > + *    - it means failure and the runtime PM usage counter of @dev has been balanced.
>
> The kerneldoc above is kind of noisy and it is hard to figure out what
> the helper really does from it.
>
> You could basically say something like "Resume @dev synchronously and
> if that is successful, increment its runtime PM usage counter.  Return
> 0 if the runtime PM usage counter of @dev has been incremented or a
> negative error code otherwise."
>
> > + */
> > +static inline int pm_runtime_general_get(struct device *dev)
>
> What about pm_runtime_resume_and_get()?

We already have pm_runtime_get_if_active() - so perhaps
pm_runtime_get_if_suspended() could be an option as well?

>
> > +{
> > +       int ret = 0;
>
> This extra initialization is not necessary.
>
> You can initialize ret to the __pm_runtime_resume() return value right away.
>
> > +
> > +       ret = __pm_runtime_resume(dev, RPM_GET_PUT);
> > +       if (ret < 0) {
> > +               pm_runtime_put_noidle(dev);
> > +               return ret;
> > +       }
> > +
> > +       return 0;
> > +}
> > +
> >  /**
> >   * pm_runtime_put - Drop device usage counter and queue up "idle check" if 0.
> >   * @dev: Target device.
> > --

Kind regards
Uffe

答复: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[zhangqilong]

> operation to deal with usage counter
> 
> On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong <zhangqilong3@huawei.com>
> wrote:
> >
> > In many case, we need to check return value of pm_runtime_get_sync,
> > but it brings a trouble to the usage counter processing. Many callers
> > forget to decrease the usage counter when it failed. It has been
> > discussed a lot[0][1]. So we add a function to deal with the usage
> > counter for better coding.
> >
> > [0]https://lkml.org/lkml/2020/6/14/88
> > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/202005200951
> > 48.10995-1-dinghao.liu@zju.edu.cn/
> > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > ---
> >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> >  1 file changed, 30 insertions(+)
> >
> > diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> > index 4b708f4e8eed..6549ce764400 100644
> > --- a/include/linux/pm_runtime.h
> > +++ b/include/linux/pm_runtime.h
> > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device
> *dev)
> >         return __pm_runtime_resume(dev, RPM_GET_PUT);  }
> >
> > +/**
> > + * pm_runtime_general_get - Bump up usage counter of a device and
> resume it.
> > + * @dev: Target device.
> > + *
> > + * Increase runtime PM usage counter of @dev first, and carry out
> > +runtime-resume
> > + * of it synchronously. If __pm_runtime_resume return negative
> > +value(device is in
> > + * error state), we to need decrease the usage counter before it
> > +return. If
> > + * __pm_runtime_resume return positive value, it means the runtime of
> > +device has
> > + * already been in active state, and we let the new wrapper return zero
> instead.
> > + *
> > + * The possible return values of this function is zero or negative value.
> > + * zero:
> > + *    - it means resume succeeed or runtime of device has already been
> active, the
> > + *      runtime PM usage counter of @dev remains incremented.
> > + * negative:
> > + *    - it means failure and the runtime PM usage counter of @dev has
> been balanced.
> 
> The kerneldoc above is kind of noisy and it is hard to figure out what the helper
> really does from it.
> 
> You could basically say something like "Resume @dev synchronously and if that
> is successful, increment its runtime PM usage counter.  Return
> 0 if the runtime PM usage counter of @dev has been incremented or a negative
> error code otherwise."
> 

How about the following description.
/**
390  * pm_runtime_general_get - Bump up usage counter of a device and resume it.
391  * @dev: Target device.
392  *
393  * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
394  * of it synchronously. If __pm_runtime_resume return negative value(device is in
395  * error state), we to need decrease the usage counter before it return. If
396  * __pm_runtime_resume return positive value, it means the runtime of device has
397  * already been in active state, and we let the new wrapper return zero instead.
398  *
399  * Resume @dev synchronously and if that is successful, and increment its runtime
400  * PM usage counter if it turn out to equal to 0. The runtime PM usage counter of
401  * @dev has been incremented or a negative error code otherwise.
402  */

Thanks,
Zhang

> > + */
> > +static inline int pm_runtime_general_get(struct device *dev)
> 
> What about pm_runtime_resume_and_get()?
> 

I think it's OK.

> > +{
> > +       int ret = 0;
> 
> This extra initialization is not necessary.
> 
> You can initialize ret to the __pm_runtime_resume() return value right away.
> 

OK, good idea.

> > +
> > +       ret = __pm_runtime_resume(dev, RPM_GET_PUT);
> > +       if (ret < 0) {
> > +               pm_runtime_put_noidle(dev);
> > +               return ret;
> > +       }
> > +
> > +       return 0;
> > +}
> > +
> >  /**
> >   * pm_runtime_put - Drop device usage counter and queue up "idle check"
> if 0.
> >   * @dev: Target device.
> > --

Re: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Rafael J. Wysocki]

On Mon, Nov 9, 2020 at 4:50 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
>
> On Mon, 9 Nov 2020 at 16:20, Rafael J. Wysocki <rafael@kernel.org> wrote:
> >
> > On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong <zhangqilong3@huawei.com> wrote:
> > >
> > > In many case, we need to check return value of pm_runtime_get_sync, but
> > > it brings a trouble to the usage counter processing. Many callers forget
> > > to decrease the usage counter when it failed. It has been discussed a
> > > lot[0][1]. So we add a function to deal with the usage counter for better
> > > coding.
> > >
> > > [0]https://lkml.org/lkml/2020/6/14/88
> > > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520095148.10995-1-dinghao.liu@zju.edu.cn/
> > > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > > ---
> > >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> > >  1 file changed, 30 insertions(+)
> > >
> > > diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> > > index 4b708f4e8eed..6549ce764400 100644
> > > --- a/include/linux/pm_runtime.h
> > > +++ b/include/linux/pm_runtime.h
> > > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device *dev)
> > >         return __pm_runtime_resume(dev, RPM_GET_PUT);
> > >  }
> > >
> > > +/**
> > > + * pm_runtime_general_get - Bump up usage counter of a device and resume it.
> > > + * @dev: Target device.
> > > + *
> > > + * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
> > > + * of it synchronously. If __pm_runtime_resume return negative value(device is in
> > > + * error state), we to need decrease the usage counter before it return. If
> > > + * __pm_runtime_resume return positive value, it means the runtime of device has
> > > + * already been in active state, and we let the new wrapper return zero instead.
> > > + *
> > > + * The possible return values of this function is zero or negative value.
> > > + * zero:
> > > + *    - it means resume succeeed or runtime of device has already been active, the
> > > + *      runtime PM usage counter of @dev remains incremented.
> > > + * negative:
> > > + *    - it means failure and the runtime PM usage counter of @dev has been balanced.
> >
> > The kerneldoc above is kind of noisy and it is hard to figure out what
> > the helper really does from it.
> >
> > You could basically say something like "Resume @dev synchronously and
> > if that is successful, increment its runtime PM usage counter.  Return
> > 0 if the runtime PM usage counter of @dev has been incremented or a
> > negative error code otherwise."
> >
> > > + */
> > > +static inline int pm_runtime_general_get(struct device *dev)
> >
> > What about pm_runtime_resume_and_get()?
>
> We already have pm_runtime_get_if_active() - so perhaps
> pm_runtime_get_if_suspended() could be an option as well?

It doesn't work this way, though.

The "get" happens even if the device has not been suspended.

Re: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Rafael J. Wysocki]

On Mon, Nov 9, 2020 at 4:50 PM zhangqilong <zhangqilong3@huawei.com> wrote:
>
> > operation to deal with usage counter
> >
> > On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong <zhangqilong3@huawei.com>
> > wrote:
> > >
> > > In many case, we need to check return value of pm_runtime_get_sync,
> > > but it brings a trouble to the usage counter processing. Many callers
> > > forget to decrease the usage counter when it failed. It has been
> > > discussed a lot[0][1]. So we add a function to deal with the usage
> > > counter for better coding.
> > >
> > > [0]https://lkml.org/lkml/2020/6/14/88
> > > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/202005200951
> > > 48.10995-1-dinghao.liu@zju.edu.cn/
> > > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > > ---
> > >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> > >  1 file changed, 30 insertions(+)
> > >
> > > diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> > > index 4b708f4e8eed..6549ce764400 100644
> > > --- a/include/linux/pm_runtime.h
> > > +++ b/include/linux/pm_runtime.h
> > > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device
> > *dev)
> > >         return __pm_runtime_resume(dev, RPM_GET_PUT);  }
> > >
> > > +/**
> > > + * pm_runtime_general_get - Bump up usage counter of a device and
> > resume it.
> > > + * @dev: Target device.
> > > + *
> > > + * Increase runtime PM usage counter of @dev first, and carry out
> > > +runtime-resume
> > > + * of it synchronously. If __pm_runtime_resume return negative
> > > +value(device is in
> > > + * error state), we to need decrease the usage counter before it
> > > +return. If
> > > + * __pm_runtime_resume return positive value, it means the runtime of
> > > +device has
> > > + * already been in active state, and we let the new wrapper return zero
> > instead.
> > > + *
> > > + * The possible return values of this function is zero or negative value.
> > > + * zero:
> > > + *    - it means resume succeeed or runtime of device has already been
> > active, the
> > > + *      runtime PM usage counter of @dev remains incremented.
> > > + * negative:
> > > + *    - it means failure and the runtime PM usage counter of @dev has
> > been balanced.
> >
> > The kerneldoc above is kind of noisy and it is hard to figure out what the helper
> > really does from it.
> >
> > You could basically say something like "Resume @dev synchronously and if that
> > is successful, increment its runtime PM usage counter.  Return
> > 0 if the runtime PM usage counter of @dev has been incremented or a negative
> > error code otherwise."
> >
>
> How about the following description.
> /**
> 390  * pm_runtime_general_get - Bump up usage counter of a device and resume it.
> 391  * @dev: Target device.
> 392  *
> 393  * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
> 394  * of it synchronously. If __pm_runtime_resume return negative value(device is in
> 395  * error state), we to need decrease the usage counter before it return. If
> 396  * __pm_runtime_resume return positive value, it means the runtime of device has
> 397  * already been in active state, and we let the new wrapper return zero instead.
> 398  *

If you add the paragraph below, the one above becomes redundant IMV.

> 399  * Resume @dev synchronously and if that is successful, and increment its runtime

"Resume @dev synchronously and if that is successful, increment its runtime"

(drop the extra "and").

> 400  * PM usage counter if it turn out to equal to 0. The runtime PM usage counter of

The "if it turn out to equal to 0" phrase is redundant (and the
grammar in it is incorrect).

> 401  * @dev has been incremented or a negative error code otherwise.
> 402  */

Why don't you use what I said verbatim?

Re: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Ulf Hansson]

On Mon, 9 Nov 2020 at 16:54, Rafael J. Wysocki <rafael@kernel.org> wrote:
>
> On Mon, Nov 9, 2020 at 4:50 PM Ulf Hansson <ulf.hansson@linaro.org> wrote:
> >
> > On Mon, 9 Nov 2020 at 16:20, Rafael J. Wysocki <rafael@kernel.org> wrote:
> > >
> > > On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong <zhangqilong3@huawei.com> wrote:
> > > >
> > > > In many case, we need to check return value of pm_runtime_get_sync, but
> > > > it brings a trouble to the usage counter processing. Many callers forget
> > > > to decrease the usage counter when it failed. It has been discussed a
> > > > lot[0][1]. So we add a function to deal with the usage counter for better
> > > > coding.
> > > >
> > > > [0]https://lkml.org/lkml/2020/6/14/88
> > > > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520095148.10995-1-dinghao.liu@zju.edu.cn/
> > > > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > > > ---
> > > >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> > > >  1 file changed, 30 insertions(+)
> > > >
> > > > diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> > > > index 4b708f4e8eed..6549ce764400 100644
> > > > --- a/include/linux/pm_runtime.h
> > > > +++ b/include/linux/pm_runtime.h
> > > > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct device *dev)
> > > >         return __pm_runtime_resume(dev, RPM_GET_PUT);
> > > >  }
> > > >
> > > > +/**
> > > > + * pm_runtime_general_get - Bump up usage counter of a device and resume it.
> > > > + * @dev: Target device.
> > > > + *
> > > > + * Increase runtime PM usage counter of @dev first, and carry out runtime-resume
> > > > + * of it synchronously. If __pm_runtime_resume return negative value(device is in
> > > > + * error state), we to need decrease the usage counter before it return. If
> > > > + * __pm_runtime_resume return positive value, it means the runtime of device has
> > > > + * already been in active state, and we let the new wrapper return zero instead.
> > > > + *
> > > > + * The possible return values of this function is zero or negative value.
> > > > + * zero:
> > > > + *    - it means resume succeeed or runtime of device has already been active, the
> > > > + *      runtime PM usage counter of @dev remains incremented.
> > > > + * negative:
> > > > + *    - it means failure and the runtime PM usage counter of @dev has been balanced.
> > >
> > > The kerneldoc above is kind of noisy and it is hard to figure out what
> > > the helper really does from it.
> > >
> > > You could basically say something like "Resume @dev synchronously and
> > > if that is successful, increment its runtime PM usage counter.  Return
> > > 0 if the runtime PM usage counter of @dev has been incremented or a
> > > negative error code otherwise."
> > >
> > > > + */
> > > > +static inline int pm_runtime_general_get(struct device *dev)
> > >
> > > What about pm_runtime_resume_and_get()?
> >
> > We already have pm_runtime_get_if_active() - so perhaps
> > pm_runtime_get_if_suspended() could be an option as well?
>
> It doesn't work this way, though.
>
> The "get" happens even if the device has not been suspended.

Yes, that's right - so pm_runtime_resume_and_get() is probably the
best we can pick then.

Kind regards
Uffe

答复: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[zhangqilong]

Hi

> 
> On Mon, Nov 9, 2020 at 4:50 PM zhangqilong <zhangqilong3@huawei.com>
> wrote:
> >
> > > operation to deal with usage counter
> > >
> > > On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong
> > > <zhangqilong3@huawei.com>
> > > wrote:
> > > >
> > > > In many case, we need to check return value of
> > > > pm_runtime_get_sync, but it brings a trouble to the usage counter
> > > > processing. Many callers forget to decrease the usage counter when
> > > > it failed. It has been discussed a lot[0][1]. So we add a function
> > > > to deal with the usage counter for better coding.
> > > >
> > > > [0]https://lkml.org/lkml/2020/6/14/88
> > > > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520
> > > > 0951 48.10995-1-dinghao.liu@zju.edu.cn/
> > > > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > > > ---
> > > >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> > > >  1 file changed, 30 insertions(+)
> > > >
> > > > diff --git a/include/linux/pm_runtime.h
> > > > b/include/linux/pm_runtime.h index 4b708f4e8eed..6549ce764400
> > > > 100644
> > > > --- a/include/linux/pm_runtime.h
> > > > +++ b/include/linux/pm_runtime.h
> > > > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct
> > > > device
> > > *dev)
> > > >         return __pm_runtime_resume(dev, RPM_GET_PUT);  }
> > > >
> > > > +/**
> > > > + * pm_runtime_general_get - Bump up usage counter of a device and
> > > resume it.
> > > > + * @dev: Target device.
> > > > + *
> > > > + * Increase runtime PM usage counter of @dev first, and carry out
> > > > +runtime-resume
> > > > + * of it synchronously. If __pm_runtime_resume return negative
> > > > +value(device is in
> > > > + * error state), we to need decrease the usage counter before it
> > > > +return. If
> > > > + * __pm_runtime_resume return positive value, it means the
> > > > +runtime of device has
> > > > + * already been in active state, and we let the new wrapper
> > > > +return zero
> > > instead.
> > > > + *
> > > > + * The possible return values of this function is zero or negative value.
> > > > + * zero:
> > > > + *    - it means resume succeeed or runtime of device has already been
> > > active, the
> > > > + *      runtime PM usage counter of @dev remains incremented.
> > > > + * negative:
> > > > + *    - it means failure and the runtime PM usage counter of @dev has
> > > been balanced.
> > >
> > > The kerneldoc above is kind of noisy and it is hard to figure out
> > > what the helper really does from it.
> > >
> > > You could basically say something like "Resume @dev synchronously
> > > and if that is successful, increment its runtime PM usage counter.
> > > Return
> > > 0 if the runtime PM usage counter of @dev has been incremented or a
> > > negative error code otherwise."
> > >
> >
> > How about the following description.
> > /**
> > 390  * pm_runtime_general_get - Bump up usage counter of a device and
> resume it.
> > 391  * @dev: Target device.
> > 392  *
> > 393  * Increase runtime PM usage counter of @dev first, and carry out
> > runtime-resume
> > 394  * of it synchronously. If __pm_runtime_resume return negative
> > value(device is in
> > 395  * error state), we to need decrease the usage counter before it
> > return. If
> > 396  * __pm_runtime_resume return positive value, it means the runtime
> > of device has
> > 397  * already been in active state, and we let the new wrapper return zero
> instead.
> > 398  *
> 
> If you add the paragraph below, the one above becomes redundant IMV.
> 
> > 399  * Resume @dev synchronously and if that is successful, and
> > increment its runtime
> 
> "Resume @dev synchronously and if that is successful, increment its runtime"
> 
> (drop the extra "and").
> 
> > 400  * PM usage counter if it turn out to equal to 0. The runtime PM
> > usage counter of
> 
> The "if it turn out to equal to 0" phrase is redundant (and the grammar in it is
> incorrect).
> 
> > 401  * @dev has been incremented or a negative error code otherwise.
> > 402  */
> 
> Why don't you use what I said verbatim?

I had misunderstand just now, sorry for that. The description is as follows:
389 /**
390  * pm_runtime_resume_and_get - Bump up usage counter of a device and resume it.
391  * @dev: Target device.
392  *
393  * Resume @dev synchronously if that is successful, increment its runtime PM
394  * usage counter.  Return 0 if the runtime PM usage counter of @dev has been
395  * incremented or a negative error code otherwise.
396  */

Do you think it's OK?

Thanks,
Zhang

Re: [PATCH v2 1/2] PM: runtime: Add a general runtime get sync operation to deal with usage counter

[Rafael J. Wysocki]

On Mon, Nov 9, 2020 at 5:15 PM zhangqilong <zhangqilong3@huawei.com> wrote:
>
> Hi
>
> >
> > On Mon, Nov 9, 2020 at 4:50 PM zhangqilong <zhangqilong3@huawei.com>
> > wrote:
> > >
> > > > operation to deal with usage counter
> > > >
> > > > On Mon, Nov 9, 2020 at 4:00 PM Zhang Qilong
> > > > <zhangqilong3@huawei.com>
> > > > wrote:
> > > > >
> > > > > In many case, we need to check return value of
> > > > > pm_runtime_get_sync, but it brings a trouble to the usage counter
> > > > > processing. Many callers forget to decrease the usage counter when
> > > > > it failed. It has been discussed a lot[0][1]. So we add a function
> > > > > to deal with the usage counter for better coding.
> > > > >
> > > > > [0]https://lkml.org/lkml/2020/6/14/88
> > > > > [1]https://patchwork.ozlabs.org/project/linux-tegra/patch/20200520
> > > > > 0951 48.10995-1-dinghao.liu@zju.edu.cn/
> > > > > Signed-off-by: Zhang Qilong <zhangqilong3@huawei.com>
> > > > > ---
> > > > >  include/linux/pm_runtime.h | 30 ++++++++++++++++++++++++++++++
> > > > >  1 file changed, 30 insertions(+)
> > > > >
> > > > > diff --git a/include/linux/pm_runtime.h
> > > > > b/include/linux/pm_runtime.h index 4b708f4e8eed..6549ce764400
> > > > > 100644
> > > > > --- a/include/linux/pm_runtime.h
> > > > > +++ b/include/linux/pm_runtime.h
> > > > > @@ -386,6 +386,36 @@ static inline int pm_runtime_get_sync(struct
> > > > > device
> > > > *dev)
> > > > >         return __pm_runtime_resume(dev, RPM_GET_PUT);  }
> > > > >
> > > > > +/**
> > > > > + * pm_runtime_general_get - Bump up usage counter of a device and
> > > > resume it.
> > > > > + * @dev: Target device.
> > > > > + *
> > > > > + * Increase runtime PM usage counter of @dev first, and carry out
> > > > > +runtime-resume
> > > > > + * of it synchronously. If __pm_runtime_resume return negative
> > > > > +value(device is in
> > > > > + * error state), we to need decrease the usage counter before it
> > > > > +return. If
> > > > > + * __pm_runtime_resume return positive value, it means the
> > > > > +runtime of device has
> > > > > + * already been in active state, and we let the new wrapper
> > > > > +return zero
> > > > instead.
> > > > > + *
> > > > > + * The possible return values of this function is zero or negative value.
> > > > > + * zero:
> > > > > + *    - it means resume succeeed or runtime of device has already been
> > > > active, the
> > > > > + *      runtime PM usage counter of @dev remains incremented.
> > > > > + * negative:
> > > > > + *    - it means failure and the runtime PM usage counter of @dev has
> > > > been balanced.
> > > >
> > > > The kerneldoc above is kind of noisy and it is hard to figure out
> > > > what the helper really does from it.
> > > >
> > > > You could basically say something like "Resume @dev synchronously
> > > > and if that is successful, increment its runtime PM usage counter.
> > > > Return
> > > > 0 if the runtime PM usage counter of @dev has been incremented or a
> > > > negative error code otherwise."
> > > >
> > >
> > > How about the following description.
> > > /**
> > > 390  * pm_runtime_general_get - Bump up usage counter of a device and
> > resume it.
> > > 391  * @dev: Target device.
> > > 392  *
> > > 393  * Increase runtime PM usage counter of @dev first, and carry out
> > > runtime-resume
> > > 394  * of it synchronously. If __pm_runtime_resume return negative
> > > value(device is in
> > > 395  * error state), we to need decrease the usage counter before it
> > > return. If
> > > 396  * __pm_runtime_resume return positive value, it means the runtime
> > > of device has
> > > 397  * already been in active state, and we let the new wrapper return zero
> > instead.
> > > 398  *
> >
> > If you add the paragraph below, the one above becomes redundant IMV.
> >
> > > 399  * Resume @dev synchronously and if that is successful, and
> > > increment its runtime
> >
> > "Resume @dev synchronously and if that is successful, increment its runtime"
> >
> > (drop the extra "and").
> >
> > > 400  * PM usage counter if it turn out to equal to 0. The runtime PM
> > > usage counter of
> >
> > The "if it turn out to equal to 0" phrase is redundant (and the grammar in it is
> > incorrect).
> >
> > > 401  * @dev has been incremented or a negative error code otherwise.
> > > 402  */
> >
> > Why don't you use what I said verbatim?
>
> I had misunderstand just now, sorry for that. The description is as follows:
> 389 /**
> 390  * pm_runtime_resume_and_get - Bump up usage counter of a device and resume it.
> 391  * @dev: Target device.
> 392  *
> 393  * Resume @dev synchronously if that is successful, increment its runtime PM

"Resume @dev synchronously and if that is successful, increment its runtime PM"

(missing "and").

> 394  * usage counter.  Return 0 if the runtime PM usage counter of @dev has been
> 395  * incremented or a negative error code otherwise.
> 396  */
>
> Do you think it's OK?

Apart from the above typo, yes it is.

Thanks!