From: Jarek Poplawski <redacted>
Date: Wed, 11 Jun 2008 07:08:47 +0000

On 11-06-2008 07:40, David Miller wrote:
...

quoted

Ok, I did an audit of all the drivers under drivers/net that invoke
flush_scheduled_work().  Here is my audit analysis and the patch I
came up with.  The only deadlock'y case I couldn't fix right now is
the cassini driver, see below for details and the patch:

8139too: Calls flush_scheduled_work() in device remove method, OK.

I have some doubt here: this rtl8139_thred() work function seems to
rearm if netif_running(). Is it guaranteed a dev is down while
running this flush_scheduled_work()?

Indeed, perhaps it should do the flush_scheduled_work()
after the unregister_netdev().