USB: usb_wwan: fix race between write and resume
authorxiao jin <jin.xiao@intel.com>
Mon, 26 May 2014 17:23:14 +0000 (19:23 +0200)
committerBen Hutchings <ben@decadent.org.uk>
Fri, 11 Jul 2014 12:33:40 +0000 (13:33 +0100)
commit0985cf0a59c203045a80abfaa5647285673b23ba
tree33580c90c483c0c08feace34cceb73992d8e3cc0
parent7253b828394395d96c35986a9c7c15665ba2e430
USB: usb_wwan: fix race between write and resume

commit d9e93c08d8d985e5ef89436ebc9f4aad7e31559f upstream.

We find a race between write and resume. usb_wwan_resume run play_delayed()
and spin_unlock, but intfdata->suspended still is not set to zero.
At this time usb_wwan_write is called and anchor the urb to delay
list. Then resume keep running but the delayed urb have no chance
to be commit until next resume. If the time of next resume is far
away, tty will be blocked in tty_wait_until_sent during time. The
race also can lead to writes being reordered.

This patch put play_Delayed and intfdata->suspended together in the
spinlock, it's to avoid the write race during resume.

Fixes: 383cedc3bb43 ("USB: serial: full autosuspend support for the
option driver")

Signed-off-by: xiao jin <jin.xiao@intel.com>
Signed-off-by: Zhang, Qi1 <qi1.zhang@intel.com>
Reviewed-by: David Cohen <david.a.cohen@linux.intel.com>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: there's no need to check for portdata == NULL]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
drivers/usb/serial/usb_wwan.c