Date: Fri, 10 Feb 2012 15:31:20 -0500 (EST)
From: Alan Stern <stern@rowland.harvard.edu>
To: Tejun Heo <tj@kernel.org>, Jens Axboe <axboe@kernel.dk>
cc: Linux-pm mailing list <linux-pm@vger.kernel.org>,
        Kernel development list <linux-kernel@vger.kernel.org>
Subject: Bug in disk event polling
Message-ID: <Pine.LNX.4.44L0.1202101519020.1281-100000@iolanthe.rowland.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1333
Lines: 32

Tejun:

Don't ask me why this hasn't shown up earlier...  There's a big fat bug 
in the implementation of disk event polling.

The polling is done using the system_nrt_wq work queue, which isn't
freezable.  As a result, polling continues while the system is
preparing for suspend or hibernation.

Obviously I/O to suspended devices doesn't work well.  Somewhat less 
obviously, error recovery for the failed I/O attempts can interfere 
with normal system resume.

You can see this for yourself easily enough by suspending or
hibernating while a USB flash drive is plugged in.  You don't even need
to go through the full suspend procedure; the first two stages are
enough (echo devices >/sys/power/pm_test).  Check the system log
afterward; most likely you'll find the flash drive got errors and had
to be unregistered and re-enumerated.

I have verified that changing all occurrences of system_nrt_wq in 
block/genhd.c to system_freezable_wq fixes the bug.  However this may 
not be the way you want to solve it; you may prefer to have a freezable 
non-reentrant work queue.

Alan Stern

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/