Re: [PATCH v3] driver core: Move device_links_purge() after bus_remove_device()

Re: [PATCH v3] driver core: Move device_links_purge() after bus_remove_device()

[Rafael J. Wysocki]

On Friday, October 20, 2017 3:23:16 AM CEST Jeffy Chen wrote:
> The current ordering of code in device_del() triggers a WARN_ON()
> in device_links_purge(), because of an unexpected link status.
> 
> The device_links_unbind_consumers() call in device_release_driver()
> has to take place before device_links_purge() for the status of all
> links to be correct, so move the device_links_purge() call in
> device_del() after the invocation of bus_remove_device() which calls
> device_release_driver().
> 
> Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

OK, thanks.

Greg, will you take this?

> ---
> 
> Changes in v3:
> Remove wrong signed-off-by.
> 
> Changes in v2:
> Update commit message provided by Rafael.
> 
>  drivers/base/core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/base/core.c b/drivers/base/core.c
> index 12ebd055724c..2e683cdf4a08 100644
> --- a/drivers/base/core.c
> +++ b/drivers/base/core.c
> @@ -1958,7 +1958,6 @@ void device_del(struct device *dev)
>  		blocking_notifier_call_chain(&dev->bus->p->bus_notifier,
>  					     BUS_NOTIFY_DEL_DEVICE, dev);
>  
> -	device_links_purge(dev);
>  	dpm_sysfs_remove(dev);
>  	if (parent)
>  		klist_del(&dev->p->knode_parent);
> @@ -1986,6 +1985,7 @@ void device_del(struct device *dev)
>  	device_pm_remove(dev);
>  	driver_deferred_probe_del(dev);
>  	device_remove_properties(dev);
> +	device_links_purge(dev);
>  
>  	/* Notify the platform of the removal, in case they
>  	 * need to do anything...
> 

[PATCH v3] driver core: Move device_links_purge() after bus_remove_device()

[Jeffy Chen]

The current ordering of code in device_del() triggers a WARN_ON()
in device_links_purge(), because of an unexpected link status.

The device_links_unbind_consumers() call in device_release_driver()
has to take place before device_links_purge() for the status of all
links to be correct, so move the device_links_purge() call in
device_del() after the invocation of bus_remove_device() which calls
device_release_driver().

Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
---

Changes in v3:
Remove wrong signed-off-by.

Changes in v2:
Update commit message provided by Rafael.

 drivers/base/core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/base/core.c b/drivers/base/core.c
index 12ebd055724c..2e683cdf4a08 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -1958,7 +1958,6 @@ void device_del(struct device *dev)
 		blocking_notifier_call_chain(&dev->bus->p->bus_notifier,
 					     BUS_NOTIFY_DEL_DEVICE, dev);
 
-	device_links_purge(dev);
 	dpm_sysfs_remove(dev);
 	if (parent)
 		klist_del(&dev->p->knode_parent);
@@ -1986,6 +1985,7 @@ void device_del(struct device *dev)
 	device_pm_remove(dev);
 	driver_deferred_probe_del(dev);
 	device_remove_properties(dev);
+	device_links_purge(dev);
 
 	/* Notify the platform of the removal, in case they
 	 * need to do anything...
-- 
2.11.0

Re: [PATCH v3] driver core: Move device_links_purge() after bus_remove_device()

[Greg KH]

On Fri, Oct 20, 2017 at 03:22:07AM +0200, Rafael J. Wysocki wrote:
> On Friday, October 20, 2017 3:23:16 AM CEST Jeffy Chen wrote:
> > The current ordering of code in device_del() triggers a WARN_ON()
> > in device_links_purge(), because of an unexpected link status.
> > 
> > The device_links_unbind_consumers() call in device_release_driver()
> > has to take place before device_links_purge() for the status of all
> > links to be correct, so move the device_links_purge() call in
> > device_del() after the invocation of bus_remove_device() which calls
> > device_release_driver().
> > 
> > Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
> > Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> 
> OK, thanks.
> 
> Greg, will you take this?

Yes I will.  But I don't understand if this is a bugfix that needs to go
to all stable kernels to resolve a problem that has always been there,
or it is ok for 4.15-rc1, as it's just a "do things in the correct
order" type of change.

Any hints?

thanks,

greg k-h

Re: [PATCH v3] driver core: Move device_links_purge() after bus_remove_device()

[Rafael J. Wysocki]

On Friday, October 20, 2017 8:22:01 AM CEST Greg KH wrote:
> On Fri, Oct 20, 2017 at 03:22:07AM +0200, Rafael J. Wysocki wrote:
> > On Friday, October 20, 2017 3:23:16 AM CEST Jeffy Chen wrote:
> > > The current ordering of code in device_del() triggers a WARN_ON()
> > > in device_links_purge(), because of an unexpected link status.
> > > 
> > > The device_links_unbind_consumers() call in device_release_driver()
> > > has to take place before device_links_purge() for the status of all
> > > links to be correct, so move the device_links_purge() call in
> > > device_del() after the invocation of bus_remove_device() which calls
> > > device_release_driver().
> > > 
> > > Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
> > > Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > 
> > OK, thanks.
> > 
> > Greg, will you take this?
> 
> Yes I will.  But I don't understand if this is a bugfix that needs to go
> to all stable kernels to resolve a problem that has always been there,
> or it is ok for 4.15-rc1, as it's just a "do things in the correct
> order" type of change.
> 
> Any hints?

Technically the problem has always been there, but it is only visible if
device links are used in a specific way and I'm not sure whether or not
there is any code in the tree where that's the case.

That said AFAICS it should be safe to add

Fixes: 9ed9895370ae (driver core: Functional dependencies tracking support)

to it and handle it accordingly.

Thanks,
Rafael

Re: [PATCH v3] driver core: Move device_links_purge() after bus_remove_device()

[jeffy]

Hi Rafael,

On 10/20/2017 07:17 PM, Rafael J. Wysocki wrote:
> Technically the problem has always been there, but it is only visible if
> device links are used in a specific way and I'm not sure whether or not
> there is any code in the tree where that's the case.
>
> That said AFAICS it should be safe to add
>
> Fixes: 9ed9895370ae (driver core: Functional dependencies tracking support)

right, forgot to do so, thanks for noticing :)
>
> to it and handle it accordingly.
>
> Thanks,
> Rafael