Patchwork [1/3] ACPI: add module autoloading support for ACPI enumerated devices

login
register
mail settings
Submitter Zhang Rui
Date Oct. 14, 2013, 9:10 a.m.
Message ID <1381741834-7334-1-git-send-email-rui.zhang@intel.com>
Download mbox | patch
Permalink /patch/283156/
State Deferred
Headers show

Comments

Zhang Rui - Oct. 14, 2013, 9:10 a.m.
An ACPI enumerated device may have its compatible id strings.

To support the compatible ACPI ids (acpi_device->pnp.ids),
we introduced acpi_driver_match_device() to match
the driver->acpi_match_table and acpi_device->pnp.ids.

For those drivers, MODULE_DEVICE_TABLE(acpi, xxx) is used to
exports the driver module alias in the format of
"acpi:device_compatible_ids".
But in the mean time, the current code does not export the
ACPI compatible strings as part of the module_alias for the
ACPI enumerated devices, which will break the module autoloading.

Take the following piece of code for example,
static const struct acpi_device_id xxx_acpi_match[] = {
        { "INTABCD", 0 },
        { }
};
MODULE_DEVICE_TABLE(acpi, xxx_acpi_match);

If this piece of code is used in a platform driver for
an ACPI enumerated platform device, the platform driver module_alias
is "acpi:INTABCD", but the uevent attribute of its platform device node
is "platform:INTABCD:00" (PREFIX:pdev->name).
If this piece of code is used in an i2c driver for an ACPI enumerated
i2c device, the i2c driver module_alias is "acpi:INTABCD", but
the uevent of its i2c device node is "i2c:INTABCD:00" (PREFIX:client->name).

The reason why the module autoloading is not broken for now is that
the uevent file of the ACPI device node is "acpi:INTABCD".
Thus it is the ACPI device node creation that loads the platform/i2c driver.

So this is a problem that will affect us the day when the ACPI bus
is removed from device model.

This patch introduces a new function to exporting ACPI ids as the
module_alias, for the ACPI enumerate devices.

For any drivers using MODULE_DEVICE_TALBE(acpi, blabla), the uevent file
of its associated device must be fixed by invoking this function.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
---
 drivers/acpi/scan.c  |   24 ++++++++++++++++++++++++
 include/linux/acpi.h |    8 ++++++++
 2 files changed, 32 insertions(+)
Mika Westerberg - Oct. 14, 2013, 10:14 a.m.
On Mon, Oct 14, 2013 at 05:10:32PM +0800, Zhang Rui wrote:
> An ACPI enumerated device may have its compatible id strings.
> 
> To support the compatible ACPI ids (acpi_device->pnp.ids),
> we introduced acpi_driver_match_device() to match
> the driver->acpi_match_table and acpi_device->pnp.ids.
> 
> For those drivers, MODULE_DEVICE_TABLE(acpi, xxx) is used to
> exports the driver module alias in the format of
> "acpi:device_compatible_ids".
> But in the mean time, the current code does not export the
> ACPI compatible strings as part of the module_alias for the
> ACPI enumerated devices, which will break the module autoloading.
> 
> Take the following piece of code for example,
> static const struct acpi_device_id xxx_acpi_match[] = {
>         { "INTABCD", 0 },
>         { }
> };
> MODULE_DEVICE_TABLE(acpi, xxx_acpi_match);
> 
> If this piece of code is used in a platform driver for
> an ACPI enumerated platform device, the platform driver module_alias
> is "acpi:INTABCD", but the uevent attribute of its platform device node
> is "platform:INTABCD:00" (PREFIX:pdev->name).
> If this piece of code is used in an i2c driver for an ACPI enumerated
> i2c device, the i2c driver module_alias is "acpi:INTABCD", but
> the uevent of its i2c device node is "i2c:INTABCD:00" (PREFIX:client->name).
> 
> The reason why the module autoloading is not broken for now is that
> the uevent file of the ACPI device node is "acpi:INTABCD".
> Thus it is the ACPI device node creation that loads the platform/i2c driver.
> 
> So this is a problem that will affect us the day when the ACPI bus
> is removed from device model.
> 
> This patch introduces a new function to exporting ACPI ids as the
> module_alias, for the ACPI enumerate devices.
> 
> For any drivers using MODULE_DEVICE_TALBE(acpi, blabla), the uevent file
                                        ^ TABLE?

> of its associated device must be fixed by invoking this function.
> 
> Signed-off-by: Zhang Rui <rui.zhang@intel.com>

Makes sense. There is one comment below but other than that, you can add
my,

Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>

For the three patches.

I wonder if we should do the same for SPI bus as well?

> ---
>  drivers/acpi/scan.c  |   24 ++++++++++++++++++++++++
>  include/linux/acpi.h |    8 ++++++++
>  2 files changed, 32 insertions(+)
> 
> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> index 407ad13..db6f879 100644
> --- a/drivers/acpi/scan.c
> +++ b/drivers/acpi/scan.c
> @@ -111,6 +111,30 @@ static int create_modalias(struct acpi_device *acpi_dev, char *modalias,
>  	return len;
>  }
>  
> +int acpi_device_uevent_modalias(struct device *dev, struct kobj_uevent_env *env)
> +{
> +	struct acpi_device *acpi_dev;
> +	int result;
> +	int len;
> +
> +	result = acpi_bus_get_device(ACPI_HANDLE(dev), &acpi_dev);
> +	if (result)
> +		return result;
> +
> +	if (list_empty(&acpi_dev->pnp.ids))
> +		return 0;
> +
> +	if (add_uevent_var(env, "MODALIAS="))
> +		return -ENOMEM;
> +	len = create_modalias(acpi_dev, &env->buf[env->buflen - 1],
> +				sizeof(env->buf) - env->buflen);
> +	if (len >= (sizeof(env->buf) - env->buflen))
> +		return -ENOMEM;
> +	env->buflen += len;
> +	return 0;
> +}
> +EXPORT_SYMBOL(acpi_device_uevent_modalias);

_GPL?
--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Zhang Rui - Oct. 14, 2013, 11:16 a.m.
On Mon, 2013-10-14 at 13:14 +0300, Mika Westerberg wrote:
> On Mon, Oct 14, 2013 at 05:10:32PM +0800, Zhang Rui wrote:
> > An ACPI enumerated device may have its compatible id strings.
> > 
> > To support the compatible ACPI ids (acpi_device->pnp.ids),
> > we introduced acpi_driver_match_device() to match
> > the driver->acpi_match_table and acpi_device->pnp.ids.
> > 
> > For those drivers, MODULE_DEVICE_TABLE(acpi, xxx) is used to
> > exports the driver module alias in the format of
> > "acpi:device_compatible_ids".
> > But in the mean time, the current code does not export the
> > ACPI compatible strings as part of the module_alias for the
> > ACPI enumerated devices, which will break the module autoloading.
> > 
> > Take the following piece of code for example,
> > static const struct acpi_device_id xxx_acpi_match[] = {
> >         { "INTABCD", 0 },
> >         { }
> > };
> > MODULE_DEVICE_TABLE(acpi, xxx_acpi_match);
> > 
> > If this piece of code is used in a platform driver for
> > an ACPI enumerated platform device, the platform driver module_alias
> > is "acpi:INTABCD", but the uevent attribute of its platform device node
> > is "platform:INTABCD:00" (PREFIX:pdev->name).
> > If this piece of code is used in an i2c driver for an ACPI enumerated
> > i2c device, the i2c driver module_alias is "acpi:INTABCD", but
> > the uevent of its i2c device node is "i2c:INTABCD:00" (PREFIX:client->name).
> > 
> > The reason why the module autoloading is not broken for now is that
> > the uevent file of the ACPI device node is "acpi:INTABCD".
> > Thus it is the ACPI device node creation that loads the platform/i2c driver.
> > 
> > So this is a problem that will affect us the day when the ACPI bus
> > is removed from device model.
> > 
> > This patch introduces a new function to exporting ACPI ids as the
> > module_alias, for the ACPI enumerate devices.
> > 
> > For any drivers using MODULE_DEVICE_TALBE(acpi, blabla), the uevent file
>                                         ^ TABLE?
> 
> > of its associated device must be fixed by invoking this function.
> > 
> > Signed-off-by: Zhang Rui <rui.zhang@intel.com>
> 
> Makes sense. There is one comment below but other than that, you can add
> my,
> 
> Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
> 
Thanks for reviewing.

> For the three patches.
> 
> I wonder if we should do the same for SPI bus as well?
> 
yes, you are right.
Actually, I should check for buses that supports
acpi_driver_match_device() and fix them all.
Will add it in next version.

thanks,
rui
> > ---
> >  drivers/acpi/scan.c  |   24 ++++++++++++++++++++++++
> >  include/linux/acpi.h |    8 ++++++++
> >  2 files changed, 32 insertions(+)
> > 
> > diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> > index 407ad13..db6f879 100644
> > --- a/drivers/acpi/scan.c
> > +++ b/drivers/acpi/scan.c
> > @@ -111,6 +111,30 @@ static int create_modalias(struct acpi_device *acpi_dev, char *modalias,
> >  	return len;
> >  }
> >  
> > +int acpi_device_uevent_modalias(struct device *dev, struct kobj_uevent_env *env)
> > +{
> > +	struct acpi_device *acpi_dev;
> > +	int result;
> > +	int len;
> > +
> > +	result = acpi_bus_get_device(ACPI_HANDLE(dev), &acpi_dev);
> > +	if (result)
> > +		return result;
> > +
> > +	if (list_empty(&acpi_dev->pnp.ids))
> > +		return 0;
> > +
> > +	if (add_uevent_var(env, "MODALIAS="))
> > +		return -ENOMEM;
> > +	len = create_modalias(acpi_dev, &env->buf[env->buflen - 1],
> > +				sizeof(env->buf) - env->buflen);
> > +	if (len >= (sizeof(env->buf) - env->buflen))
> > +		return -ENOMEM;
> > +	env->buflen += len;
> > +	return 0;
> > +}
> > +EXPORT_SYMBOL(acpi_device_uevent_modalias);
> 
> _GPL?


--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jarkko Nikula - Oct. 14, 2013, 12:44 p.m.
Hi

On 10/14/2013 02:16 PM, Zhang Rui wrote:
> On Mon, 2013-10-14 at 13:14 +0300, Mika Westerberg wrote:
>> For the three patches.
>>
>> I wonder if we should do the same for SPI bus as well?
>>
> yes, you are right.
> Actually, I should check for buses that supports
> acpi_driver_match_device() and fix them all.
> Will add it in next version.
>
You could have my pre- "Tested-by: Jarkko Nikula 
<jarkko.nikula@linux.intel.com>".

With your set I see bunch of failing modprobe calls from udev for 
i2c:INTABCD:xy and platform:INTDCBA:xy modaliases gone.
Aaron Lu - Oct. 21, 2013, 3:20 a.m.
On 10/14/2013 05:10 PM, Zhang Rui wrote:
> An ACPI enumerated device may have its compatible id strings.
> 
> To support the compatible ACPI ids (acpi_device->pnp.ids),
> we introduced acpi_driver_match_device() to match
> the driver->acpi_match_table and acpi_device->pnp.ids.
> 
> For those drivers, MODULE_DEVICE_TABLE(acpi, xxx) is used to
> exports the driver module alias in the format of
> "acpi:device_compatible_ids".
> But in the mean time, the current code does not export the
> ACPI compatible strings as part of the module_alias for the
> ACPI enumerated devices, which will break the module autoloading.
> 
> Take the following piece of code for example,
> static const struct acpi_device_id xxx_acpi_match[] = {
>         { "INTABCD", 0 },
>         { }
> };
> MODULE_DEVICE_TABLE(acpi, xxx_acpi_match);
> 
> If this piece of code is used in a platform driver for
> an ACPI enumerated platform device, the platform driver module_alias
> is "acpi:INTABCD", but the uevent attribute of its platform device node
> is "platform:INTABCD:00" (PREFIX:pdev->name).
> If this piece of code is used in an i2c driver for an ACPI enumerated
> i2c device, the i2c driver module_alias is "acpi:INTABCD", but
> the uevent of its i2c device node is "i2c:INTABCD:00" (PREFIX:client->name).
> 
> The reason why the module autoloading is not broken for now is that
> the uevent file of the ACPI device node is "acpi:INTABCD".
> Thus it is the ACPI device node creation that loads the platform/i2c driver.
> 
> So this is a problem that will affect us the day when the ACPI bus
> is removed from device model.
> 
> This patch introduces a new function to exporting ACPI ids as the
> module_alias, for the ACPI enumerate devices.
> 
> For any drivers using MODULE_DEVICE_TALBE(acpi, blabla), the uevent file
> of its associated device must be fixed by invoking this function.
> 
> Signed-off-by: Zhang Rui <rui.zhang@intel.com>
> ---
>  drivers/acpi/scan.c  |   24 ++++++++++++++++++++++++
>  include/linux/acpi.h |    8 ++++++++
>  2 files changed, 32 insertions(+)
> 
> diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
> index 407ad13..db6f879 100644
> --- a/drivers/acpi/scan.c
> +++ b/drivers/acpi/scan.c
> @@ -111,6 +111,30 @@ static int create_modalias(struct acpi_device *acpi_dev, char *modalias,
>  	return len;
>  }
>  
> +int acpi_device_uevent_modalias(struct device *dev, struct kobj_uevent_env *env)
> +{
> +	struct acpi_device *acpi_dev;
> +	int result;
> +	int len;
> +
> +	result = acpi_bus_get_device(ACPI_HANDLE(dev), &acpi_dev);
> +	if (result)
> +		return result;
> +
> +	if (list_empty(&acpi_dev->pnp.ids))
> +		return 0;
> +
> +	if (add_uevent_var(env, "MODALIAS="))
> +		return -ENOMEM;
> +	len = create_modalias(acpi_dev, &env->buf[env->buflen - 1],
> +				sizeof(env->buf) - env->buflen);
> +	if (len >= (sizeof(env->buf) - env->buflen))
> +		return -ENOMEM;
> +	env->buflen += len;
> +	return 0;
> +}
> +EXPORT_SYMBOL(acpi_device_uevent_modalias);

Looks like this function shares a lot with the existing
acpi_device_uevent, so perhaps worth making some resue of the exising
code? e.g. seperate the common code into a function and then have the
two functions call it.

-Aaron

> +
>  static ssize_t
>  acpi_device_modalias_show(struct device *dev, struct device_attribute *attr, char *buf) {
>  	struct acpi_device *acpi_dev = to_acpi_device(dev);
> diff --git a/include/linux/acpi.h b/include/linux/acpi.h
> index a5db4ae..117fa26 100644
> --- a/include/linux/acpi.h
> +++ b/include/linux/acpi.h
> @@ -401,6 +401,8 @@ static inline bool acpi_driver_match_device(struct device *dev,
>  	return !!acpi_match_device(drv->acpi_match_table, dev);
>  }
>  
> +int acpi_device_uevent_modalias(struct device *, struct kobj_uevent_env *);
> +
>  #define ACPI_PTR(_ptr)	(_ptr)
>  
>  #else	/* !CONFIG_ACPI */
> @@ -471,6 +473,12 @@ static inline bool acpi_driver_match_device(struct device *dev,
>  	return false;
>  }
>  
> +static inline int acpi_device_uevent_modalias(struct device *dev,
> +				struct kobj_uevent_env *env)
> +{
> +	return -ENODEV;
> +}
> +
>  #define ACPI_PTR(_ptr)	(NULL)
>  
>  #endif	/* !CONFIG_ACPI */
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index 407ad13..db6f879 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -111,6 +111,30 @@  static int create_modalias(struct acpi_device *acpi_dev, char *modalias,
 	return len;
 }
 
+int acpi_device_uevent_modalias(struct device *dev, struct kobj_uevent_env *env)
+{
+	struct acpi_device *acpi_dev;
+	int result;
+	int len;
+
+	result = acpi_bus_get_device(ACPI_HANDLE(dev), &acpi_dev);
+	if (result)
+		return result;
+
+	if (list_empty(&acpi_dev->pnp.ids))
+		return 0;
+
+	if (add_uevent_var(env, "MODALIAS="))
+		return -ENOMEM;
+	len = create_modalias(acpi_dev, &env->buf[env->buflen - 1],
+				sizeof(env->buf) - env->buflen);
+	if (len >= (sizeof(env->buf) - env->buflen))
+		return -ENOMEM;
+	env->buflen += len;
+	return 0;
+}
+EXPORT_SYMBOL(acpi_device_uevent_modalias);
+
 static ssize_t
 acpi_device_modalias_show(struct device *dev, struct device_attribute *attr, char *buf) {
 	struct acpi_device *acpi_dev = to_acpi_device(dev);
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index a5db4ae..117fa26 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -401,6 +401,8 @@  static inline bool acpi_driver_match_device(struct device *dev,
 	return !!acpi_match_device(drv->acpi_match_table, dev);
 }
 
+int acpi_device_uevent_modalias(struct device *, struct kobj_uevent_env *);
+
 #define ACPI_PTR(_ptr)	(_ptr)
 
 #else	/* !CONFIG_ACPI */
@@ -471,6 +473,12 @@  static inline bool acpi_driver_match_device(struct device *dev,
 	return false;
 }
 
+static inline int acpi_device_uevent_modalias(struct device *dev,
+				struct kobj_uevent_env *env)
+{
+	return -ENODEV;
+}
+
 #define ACPI_PTR(_ptr)	(NULL)
 
 #endif	/* !CONFIG_ACPI */