Re: [PATCH v8 4/8] OF: platform: Add OF notifier handler

[Pantelis Antoniou <pantelis.antoniou-OWPKS81ov/FWk0Htik3J/w@public.gmane.org>]

Hi Grant,

> On Nov 14, 2014, at 01:29 , Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org> wrote:
> 
> On Tue, 28 Oct 2014 22:36:01 +0200
> , Pantelis Antoniou <pantelis.antoniou-OWPKS81ov/FWk0Htik3J/w@public.gmane.org>
> wrote:
>> Add OF notifier handler needed for creating/destroying platform devices
>> according to dynamic runtime changes in the DT live tree.
>> 
>> Signed-off-by: Pantelis Antoniou <pantelis.antoniou-OWPKS81ov/FWk0Htik3J/w@public.gmane.org>
> 
> Hi Pantelis,
> 
> Some comments below. Feel free to send me fixup patches instead of
> respinning.
> 

Sure, I’ll get around to send updated patches over the weekend.

> g.
> 
>> ---
>> drivers/base/platform.c     | 18 +++++++++--
>> drivers/of/platform.c       | 78 +++++++++++++++++++++++++++++++++++++++++++++
>> include/linux/of_platform.h | 10 ++++++
>> 3 files changed, 103 insertions(+), 3 deletions(-)
>> 
>> diff --git a/drivers/base/platform.c b/drivers/base/platform.c
>> index b2afc29..282bfec 100644
>> --- a/drivers/base/platform.c
>> +++ b/drivers/base/platform.c
>> @@ -1002,10 +1002,22 @@ int __init platform_bus_init(void)
>> 
>> 	error = device_register(&platform_bus);
>> 	if (error)
>> -		return error;
>> -	error =  bus_register(&platform_bus_type);
>> +		goto err_out;
>> +
>> +	error = bus_register(&platform_bus_type);
>> +	if (error)
>> +		goto err_unreg_dev;
>> +
>> +	error = of_platform_register_reconfig_notifier();
>> 	if (error)
>> -		device_unregister(&platform_bus);
>> +		goto err_unreg_bus;
> 
> We really don't want to fail out here. If the notifier registration fails, it
> doesn't make sense to cause the entire boot to fail.
> 
> Instead of refactoring the function, just add the call to
> of_platform_register_reconfig_notifier(), and WARN_ON() failure without
> bailing.
> 
> (Actually, you don't need to send me a fixup for this; I've gone ahead
> and fixed it up in my tree)
> 

I see. Thanks.

I’m curious under which condition the of_platform_register_reconfig_notifier()
would fail. Only under extreme low memory conditions (and very early in the
boot-sequence). I doubt we’ll get very far - the boot-sequence will croak shortly
after. 

>> +
>> +	return 0;
>> +err_unreg_bus:
>> +	bus_unregister(&platform_bus_type);
>> +err_unreg_dev:
>> +	device_unregister(&platform_bus);
>> +err_out:
>> 	return error;
>> }
>> 
>> diff --git a/drivers/of/platform.c b/drivers/of/platform.c
>> index 3b64d0b..aa8db92 100644
>> --- a/drivers/of/platform.c
>> +++ b/drivers/of/platform.c
>> @@ -546,4 +546,82 @@ void of_platform_depopulate(struct device *parent)
>> }
>> EXPORT_SYMBOL_GPL(of_platform_depopulate);
>> 
>> +#ifdef CONFIG_OF_DYNAMIC
>> +
>> +static struct notifier_block platform_of_notifier;
>> +
>> +static int of_platform_notify(struct notifier_block *nb,
>> +				unsigned long action, void *arg)
>> +{
>> +	struct platform_device *pdev_parent, *pdev;
>> +	struct device_node *dn;
>> +	int state;
>> +	bool children_left;
>> +
>> +	state = of_reconfig_get_state_change(action, arg);
>> +
>> +	/* no change? */
>> +	if (state == -1)
>> +		return NOTIFY_OK;
>> +
>> +	switch (action) {
>> +	case OF_RECONFIG_ATTACH_NODE:
>> +	case OF_RECONFIG_DETACH_NODE:
>> +		dn = arg;
>> +		break;
>> +	case OF_RECONFIG_ADD_PROPERTY:
>> +	case OF_RECONFIG_REMOVE_PROPERTY:
>> +	case OF_RECONFIG_UPDATE_PROPERTY:
>> +		dn = ((struct of_prop_reconfig *)arg)->dn;
>> +		break;
>> +	default:
>> +		return NOTIFY_OK;
>> +	}
>> +
>> +	if (state) {
>> +
>> +		/* verify that the parent is a bus */
>> +		if (!of_match_node(of_default_bus_match_table, dn->parent))
>> +			return NOTIFY_OK;	/* not for us */
> 
> This doesn't work reliably. Not all callers of of_platform_populate use
> the of_default_bus_match_table. The code needs to actively track the
> nodes that were used to create child devices. We've got a flag in the
> device node now that you can use for that; OF_POPULATED_BUS.
> 

I see. That’s a relatively new flag no?

>> +
>> +		/* pdev_parent may be NULL when no bus platform device */
>> +		pdev_parent = of_find_device_by_node(dn->parent);
>> +		pdev = of_platform_device_create(dn, NULL,
>> +				pdev_parent ? &pdev_parent->dev : NULL);
>> +		of_dev_put(pdev_parent);
>> +
>> +		if (pdev == NULL) {
>> +			pr_err("%s: failed to create for '%s'\n",
>> +					__func__, dn->full_name);
>> +			/* of_platform_device_create tosses the error code */
>> +			return notifier_from_errno(-EINVAL);
>> +		}
>> +
>> +	} else {
>> +
>> +		/* find our device by node */
>> +		pdev = of_find_device_by_node(dn);
>> +		if (pdev == NULL)
>> +			return NOTIFY_OK;	/* no? not meant for us */
>> +
>> +		/* unregister takes one ref away */
>> +		of_platform_device_destroy(&pdev->dev, &children_left);
>> +
>> +		/* and put the reference of the find */
>> +		of_dev_put(pdev);
>> +
>> +	}
>> +
>> +	return NOTIFY_OK;
>> +}
>> +
>> +int of_platform_register_reconfig_notifier(void)
>> +{
>> +	platform_of_notifier.notifier_call = of_platform_notify;
>> +	return of_reconfig_notifier_register(&platform_of_notifier);
>> +}
>> +EXPORT_SYMBOL_GPL(of_platform_register_reconfig_notifier);
>> +
>> +#endif
>> +
>> #endif /* CONFIG_OF_ADDRESS */
>> diff --git a/include/linux/of_platform.h b/include/linux/of_platform.h
>> index c2b0627..01fe5d6 100644
>> --- a/include/linux/of_platform.h
>> +++ b/include/linux/of_platform.h
>> @@ -84,4 +84,14 @@ static inline int of_platform_populate(struct device_node *root,
>> static inline void of_platform_depopulate(struct device *parent) { }
>> #endif
>> 
>> +#ifdef CONFIG_OF_DYNAMIC
>> +extern int of_platform_register_reconfig_notifier(void);
>> +#else
>> +static inline int of_platform_register_reconfig_notifier(void)
>> +{
>> +	return 0;
>> +}
>> +#endif
>> +
>> +
>> #endif	/* _LINUX_OF_PLATFORM_H */
>> -- 
>> 1.7.12