[PATCH] usbmon: fix tiny race exposed by the fastboot patches

[PATCH] usbmon: fix tiny race exposed by the fastboot patches

[Arjan van de Ven]

From: Arjan van de Ven <arjan@linux.intel.com>
Subject: [PATCH] usbmon: fix tiny race exposed by the fastboot patches

usbmon registers the notifier chain, takes the bus lock
and then goes to scan the existing devices for hooking up.

unfortunately, if usb_mon gets initialized while USB bus discovery is going on,
it's possible that usbmon gets a notifier on one cpu (which runs without USB locks),
and the scan is going on and also finds the new bus, resulting in a double sysfs
registration, which then produces a WARNING.

Pete Zaitcev did the bug diagnostics on this one

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
---
 drivers/usb/mon/mon_main.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/usb/mon/mon_main.c b/drivers/usb/mon/mon_main.c
index 442d807..5e0ab42 100644
--- a/drivers/usb/mon/mon_main.c
+++ b/drivers/usb/mon/mon_main.c
@@ -361,12 +361,12 @@ static int __init mon_init(void)
 	}
 	// MOD_INC_USE_COUNT(which_module?);
 
-	usb_register_notify(&mon_nb);
 
 	mutex_lock(&usb_bus_list_lock);
 	list_for_each_entry (ubus, &usb_bus_list, bus_list) {
 		mon_bus_init(ubus);
 	}
+	usb_register_notify(&mon_nb);
 	mutex_unlock(&usb_bus_list_lock);
 	return 0;
 
-- 
1.5.5.1



-- 
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org

Re: [PATCH] usbmon: fix tiny race exposed by the fastboot patches

[Pete Zaitcev]

On Sat, 13 Sep 2008 12:48:58 -0700, Arjan van de Ven <arjan@infradead.org> wrote:

> +++ b/drivers/usb/mon/mon_main.c
> @@ -361,12 +361,12 @@ static int __init mon_init(void)
>  	}
>  	// MOD_INC_USE_COUNT(which_module?);
>  
> -	usb_register_notify(&mon_nb);
>  
>  	mutex_lock(&usb_bus_list_lock);
>  	list_for_each_entry (ubus, &usb_bus_list, bus_list) {
>  		mon_bus_init(ubus);
>  	}
> +	usb_register_notify(&mon_nb);
>  	mutex_unlock(&usb_bus_list_lock);

Looks like the above does not add new deadlocks and does not
miss notifications, so ACK.

-- Pete