On Sat, Apr 7, 2018 at 8:50 AM, Darren Hart <dvhart@xxxxxxxxxxxxx> wrote: > On Fri, Apr 06, 2018 at 10:37:29PM -0700, João Paulo Rechi Vita wrote: >> When the module is removed the led workqueue is destroyed in the remove >> callback, before the led device is unregistered from the led subsystem. >> >> This leads to a NULL pointer derefence when the led device is >> unregistered automatically later as part of the module removal cleanup. >> Bellow is the backtrace showing the problem. >> > > Thanks João Paulo, > ... > >> Unregistering the led device on the remove callback before destroying the >> workqueue avoids this problem. >> >> https://bugzilla.kernel.org/show_bug.cgi?id=196097 >> >> Reported-by: Dun Hum <bitter.taste@xxxxxxx> >> Signed-off-by: João Paulo Rechi Vita <jprvita@xxxxxxxxxxxx> >> --- >> drivers/platform/x86/asus-wireless.c | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/drivers/platform/x86/asus-wireless.c b/drivers/platform/x86/asus-wireless.c >> index 343e12547660..ecd715c82de5 100644 >> --- a/drivers/platform/x86/asus-wireless.c >> +++ b/drivers/platform/x86/asus-wireless.c >> @@ -181,6 +181,7 @@ static int asus_wireless_remove(struct acpi_device *adev) >> { >> struct asus_wireless_data *data = acpi_driver_data(adev); >> >> + devm_led_classdev_unregister(&adev->dev, &data->led); >> if (data->wq) >> destroy_workqueue(data->wq); >> return 0; > > asus_wireless_add only calls devm_led_classdev_register() iff the workqueue is > successfully created. It seems like it would make sense to move the > devm_led_classdev_unregister() call within the 'if (data->wq)' condition block. > I agree. I didn't do it this way initially as I believe the call to devm_led_classdev_unregister() would be harmless in any case, but it indeed best to avoid it if we can. > This should also cc stable. > Adding to v2. -- João Paulo Rechi Vita http://about.me/jprvita
