xen/manage: Always freeze/thaw processes when suspend/resuming
authorRoss Lagerwall <ross.lagerwall@citrix.com>
Mon, 18 Aug 2014 09:41:36 +0000 (10:41 +0100)
committerDavid Vrabel <david.vrabel@citrix.com>
Tue, 2 Sep 2014 14:36:59 +0000 (15:36 +0100)
Always freeze processes when suspending and thaw processes when resuming
to prevent a race noticeable with HVM guests.

This prevents a deadlock where the khubd kthread (which is designed to
be freezable) acquires a usb device lock and then tries to allocate
memory which requires the disk which hasn't been resumed yet.
Meanwhile, the xenwatch thread deadlocks waiting for the usb device
lock.

Freezing processes fixes this because the khubd thread is only thawed
after the xenwatch thread finishes resuming all the devices.

Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Cc: stable@vger.kernel.org
drivers/xen/manage.c

index 5f1e1f3cd18619ed2899dcc5e967ca274f690189..f8bb36f9d9cef862c6c502b5ba25b7d38476cd85 100644 (file)
@@ -103,16 +103,11 @@ static void do_suspend(void)
 
        shutting_down = SHUTDOWN_SUSPEND;
 
-#ifdef CONFIG_PREEMPT
-       /* If the kernel is preemptible, we need to freeze all the processes
-          to prevent them from being in the middle of a pagetable update
-          during suspend. */
        err = freeze_processes();
        if (err) {
                pr_err("%s: freeze failed %d\n", __func__, err);
                goto out;
        }
-#endif
 
        err = dpm_suspend_start(PMSG_FREEZE);
        if (err) {
@@ -157,10 +152,8 @@ out_resume:
        dpm_resume_end(si.cancelled ? PMSG_THAW : PMSG_RESTORE);
 
 out_thaw:
-#ifdef CONFIG_PREEMPT
        thaw_processes();
 out:
-#endif
        shutting_down = SHUTDOWN_INVALID;
 }
 #endif /* CONFIG_HIBERNATE_CALLBACKS */