Re: mdelay causes BUG, please use udelay

[Jeff Garzik <jgarzik@mandrakesoft.com>]

Feldman, Scott wrote:
>>-    msec_delay(10);
>>+    usec_delay(10000);
> 
> 
> Jeff, 10000 seems on the border of what's OK.  If it's acceptable, then
> let's go for that.  Otherwise, we're going to have to chain several
> mod_timer callbacks together to do a controller reset.


That definitely wants fixing.  Since I like doing resets and similar 
slow-paths in process context -- sleep for as long as you want -- I 
would say kick over to a function called via schedule_task()

Just make sure other parts of the driver that may be called 
asynchronously, such as ethtool ioctls, are disabled.  Remember that 
tx_timeout holds the dev->xmit_lock as well, so spending a long time in 
there is a bad idea in general.

I would probably call netif_carrier_off() first thing in tx_timeout, too.

	Jeff