[PATCH] driver core: unbind consumers before locking parent on device unbind

[PATCH] driver core: unbind consumers before locking parent on device unbind

[Lucas Stach]

On forced unbind the parent lock is taken, as the device might be a USB
device. The device to be unbound might have active consumers, which
must be unbound before calling device_release_driver(), otherwise this
function might deadlock when trying to take the parent lock during
consumer unbind.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
---
 drivers/base/bus.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/base/bus.c b/drivers/base/bus.c
index 6470eb8088f4..23df171a6a64 100644
--- a/drivers/base/bus.c
+++ b/drivers/base/bus.c
@@ -186,6 +186,9 @@ static ssize_t unbind_store(struct device_driver *drv, const char *buf,
 
 	dev = bus_find_device_by_name(bus, NULL, buf);
 	if (dev && dev->driver == drv) {
+		while (device_links_busy(dev)) {
+			device_links_unbind_consumers(dev);
+		}
 		if (dev->parent)	/* Needed for USB */
 			device_lock(dev->parent);
 		device_release_driver(dev);
-- 
2.11.0

Re: [PATCH] driver core: unbind consumers before locking parent on device unbind

[Greg Kroah-Hartman]

On Fri, Feb 17, 2017 at 07:36:29PM +0100, Lucas Stach wrote:
> On forced unbind the parent lock is taken, as the device might be a USB
> device. The device to be unbound might have active consumers, which
> must be unbound before calling device_release_driver(), otherwise this
> function might deadlock when trying to take the parent lock during
> consumer unbind.

Really?  What type of device does this fail on today?

thanks,

greg k-h

Re: [PATCH] driver core: unbind consumers before locking parent on device unbind

[Lucas Stach]

Am Freitag, den 17.02.2017, 10:42 -0800 schrieb Greg Kroah-Hartman:
> On Fri, Feb 17, 2017 at 07:36:29PM +0100, Lucas Stach wrote:
> > On forced unbind the parent lock is taken, as the device might be a USB
> > device. The device to be unbound might have active consumers, which
> > must be unbound before calling device_release_driver(), otherwise this
> > function might deadlock when trying to take the parent lock during
> > consumer unbind.
> 
> Really?  What type of device does this fail on today?

Regular devices on a MMIO bus. __device_release_driver() will take the
parent lock unconditionally when it unbinds consumer devices. So if the
forced unbound device is on the same bus as the consumer we will
deadlock right there.

Regards,
Lucas