PM: runtime: Resume the device earlier in __device_release_driver()
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>
Thu, 22 Oct 2020 15:38:22 +0000 (17:38 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 10 Nov 2020 09:29:07 +0000 (10:29 +0100)
commit50e050fb05ceaabb68064a11da44af52b6c423cd
treec1be20fb22af93f90595bb75b6dd6f38f0b01ae4
parent31dff3ab42b599a9d4ccb3798c99eaea9ad261a5
PM: runtime: Resume the device earlier in __device_release_driver()

commit 9226c504e364158a17a68ff1fe9d67d266922f50 upstream.

Since the device is resumed from runtime-suspend in
__device_release_driver() anyway, it is better to do that before
looking for busy managed device links from it to consumers, because
if there are any, device_links_unbind_consumers() will be called
and it will cause the consumer devices' drivers to unbind, so the
consumer devices will be runtime-resumed.  In turn, resuming each
consumer device will cause the supplier to be resumed and when the
runtime PM references from the given consumer to it are dropped, it
may be suspended.  Then, the runtime-resume of the next consumer
will cause the supplier to resume again and so on.

Update the code accordingly.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Fixes: 9ed9895370ae ("driver core: Functional dependencies tracking support")
Cc: All applicable <stable@vger.kernel.org> # All applicable
Tested-by: Xiang Chen <chenxiang66@hisilicon.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/base/dd.c