Re: [RFC PATCH 2/5] driver core: introduce global device ADD/DEL notifier

[Andy Green <andy.green@linaro.org>]

On 02/12/12 23:01, the mail apparently from Ming Lei included:
> The global device ADD/DEL notifier is introduced so that
> some platform code can bind some device resource to the
> device. When this platform code runs, there is no any bus
> information about the device, so have to resort to the
> global notifier.
>
> Cc: Andy Green <andy.green@linaro.org>
> Cc: Roger Quadros <rogerq@ti.com>
> Cc: Alan Stern <stern@rowland.harvard.edu>
> Cc: Felipe Balbi <balbi@ti.com>
> Signed-off-by: Ming Lei <tom.leiming@gmail.com>
> ---
>   drivers/base/core.c    |   21 +++++++++++++++++++++
>   include/linux/device.h |   13 +++++++++++++
>   2 files changed, 34 insertions(+)
>
> diff --git a/drivers/base/core.c b/drivers/base/core.c
> index a235085..37f11ff 100644
> --- a/drivers/base/core.c
> +++ b/drivers/base/core.c
> @@ -43,6 +43,9 @@ static __init int sysfs_deprecated_setup(char *arg)
>   early_param("sysfs.deprecated", sysfs_deprecated_setup);
>   #endif
>
> +/* global device nofifier */
> +struct blocking_notifier_head dev_notifier;
> +
>   int (*platform_notify)(struct device *dev) = NULL;
>   int (*platform_notify_remove)(struct device *dev) = NULL;
>   static struct kobject *dev_kobj;
> @@ -1041,6 +1044,8 @@ int device_add(struct device *dev)
>   	if (platform_notify)
>   		platform_notify(dev);
>
> +	blocking_notifier_call_chain(&dev_notifier, DEV_NOTIFY_ADD_DEVICE, dev);
> +
>   	error = device_create_file(dev, &uevent_attr);
>   	if (error)
>   		goto attrError;
> @@ -1228,6 +1233,7 @@ void device_del(struct device *dev)
>   	device_pm_remove(dev);
>   	driver_deferred_probe_del(dev);
>
> +	blocking_notifier_call_chain(&dev_notifier, DEV_NOTIFY_DEL_DEVICE, dev);
>   	/* Notify the platform of the removal, in case they
>   	 * need to do anything...
>   	 */
> @@ -1376,6 +1382,8 @@ struct device *device_find_child(struct device *parent, void *data,
>
>   int __init devices_init(void)
>   {
> +	BLOCKING_INIT_NOTIFIER_HEAD(&dev_notifier);
> +
>   	devices_kset = kset_create_and_add("devices", &device_uevent_ops, NULL);
>   	if (!devices_kset)
>   		return -ENOMEM;
> @@ -1995,6 +2003,19 @@ int dev_printk(const char *level, const struct device *dev,
>   }
>   EXPORT_SYMBOL(dev_printk);
>
> +/* The notifier should be avoided as far as possible */
> +int dev_register_notifier(struct notifier_block *nb)
> +{
> +	return blocking_notifier_chain_register(&dev_notifier, nb);
> +}
> +EXPORT_SYMBOL_GPL(dev_register_notifier);
> +
> +int dev_unregister_notifier(struct notifier_block *nb)
> +{
> +	return blocking_notifier_chain_unregister(&dev_notifier, nb);
> +}
> +EXPORT_SYMBOL_GPL(dev_unregister_notifier);
> +
>   #define define_dev_printk_level(func, kern_level)		\
>   int func(const struct device *dev, const char *fmt, ...)	\
>   {								\
> diff --git a/include/linux/device.h b/include/linux/device.h
> index 43dcda9..aeb54f6 100644
> --- a/include/linux/device.h
> +++ b/include/linux/device.h
> @@ -179,6 +179,19 @@ extern int bus_unregister_notifier(struct bus_type *bus,
>   #define BUS_NOTIFY_UNBOUND_DRIVER	0x00000006 /* driver is unbound
>   						      from the device */
>
> +/* All 2 notifers below get called with the target struct device *
> + * as an argument. Note that those functions are likely to be called
> + * with the device lock held in the core, so be careful.
> + */
> +#define DEV_NOTIFY_ADD_DEVICE          0x00000001 /* device added */
> +#define DEV_NOTIFY_DEL_DEVICE          0x00000002 /* device removed */
> +extern int dev_register_notifier(struct notifier_block *nb);
> +extern int dev_unregister_notifier(struct notifier_block *nb);
> +
> +extern struct kset *bus_get_kset(struct bus_type *bus);
> +extern struct klist *bus_get_device_klist(struct bus_type *bus);
> +
> +
>   extern struct kset *bus_get_kset(struct bus_type *bus);
>   extern struct klist *bus_get_device_klist(struct bus_type *bus);

Device de/registraton time is not necessarily a good choice for the 
notification point.  At boot time, platform_devices may be being 
registered with no sign of driver and anything getting enabled in the 
notifier without further filtering (at each notifier...) will then 
always be enabled the whole session whether in use or not.

probe() and remove() are more interesting because at that point the 
driver is present and we're definitely instantiating the thing or 
finished with its instantiation, and it's valid for non-usb devices too.

In the one case you're servicing in the series, usb hub port, the device 
is very unusual in that it has no driver.  However if you're going to 
add a global notifier in generic device code, you should have it do the 
right thing for normal device case so other people can use it... how I 
solved this for hub port was to simply normalize hub port by introducing 
a stub driver for it, so you get a probe / remove like anything else.

-Andy

-- 
Andy Green | TI Landing Team Leader
Linaro.org │ Open source software for ARM SoCs | Follow Linaro
http://facebook.com/pages/Linaro/155974581091106  - 
http://twitter.com/#!/linaroorg - http://linaro.org/linaro-blog
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html