Hi All,

Cc: more people + linux-input@vger.kernel.org

On 01/07/2016 03:49 AM, kernel test robot wrote:

FYI, we noticed the below changes on

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
commit 013c074f8642d8e815ad670601f8e27155a74b57 ("PM / sleep: prohibit devices probing during suspend/hibernation")

<4>[   33.021805] serio serio0: device_attach() failed for isa0060/serio0 (i8042 KBD port), error: -517
<4>[   33.021813] serio serio1: device_attach() failed for isa0060/serio1 (i8042 AUX port), error: -517
<6>[   33.021906] PM: resume of devices complete after 171.688 msecs

To reproduce:

         git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
         cd lkp-tests
         bin/lkp install job.yaml  # job file is attached in this email
         bin/lkp run     job.yaml

I've done some investigation of the above log output:
1) Small introduction first: commit 013c074f8642 was created, because it was identified
that it's unsafe to bind/probe drivers during suspend resume, mainly because this will
cause changing of pointers on PM data structures on the fly for devices which already
tracked by Suspend core.

2) above log prints can be generated by the below call chain (I've checked the code only):

drivers/input/serio/serio.c
serio_resume()
 - serio_queue_event(serio, NULL, SERIO_RECONNECT_PORT)
   - queue_work(system_long_wq, &serio_event_work);
 ....
 - serio_handle_event()
   - serio_reconnect_port()
	- serio_reconnect_driver() --- return <0
		
	- serio_disconnect_port()
	  - device_release_driver() --- unbind
 	- serio_find_driver()
	   error = device_attach(&serio->dev);
	   if (error < 0)
		dev_warn(&serio->dev,
			 "device_attach() failed for %s (%s), error: %d\n",
			 serio->phys, serio->name, error);

  device_attach() will return -EPROBE_DEFER(-517) because the device probing is not allowed yet.

As per my understanding (and it's not clear from issue description), above situation
should not break functionality of the system - just device's probe will be delayed till the
end of dpm_complete(). So, I propose consider the option - do not print above warnings if 
return code from device_attach() is -EPROBE_DEFER.

diff --git a/drivers/input/serio/serio.c b/drivers/input/serio/serio.c
index 8f82897..1ca7f55 100644
--- a/drivers/input/serio/serio.c
+++ b/drivers/input/serio/serio.c

@@ -134,7 +134,7 @@ static void serio_find_driver(struct serio *serio)
        int error;
 
        error = device_attach(&serio->dev);
-       if (error < 0)
+       if (error < 0 && error != -EPROBE_DEFER)
                dev_warn(&serio->dev,
                         "device_attach() failed for %s (%s), error: %d\n",
                         serio->phys, serio->name, error);

[1] https://lkml.org/lkml/2015/9/11/554
-- 
regards,
-grygorii