From: Ricardo Galli (gallir_at_uib.es)
Date: 2002-08-28 14:42:13 UTC
On Wednesday 28 August 2002 16:19, Jouni Malinen shaped the electrons to say:
> On Wed, Aug 28, 2002 at 01:14:31AM +0200, Ricardo Galli wrote:
> > Confirmed, after hours of ping-flooding from end to end without a glitch,
> > the problem seems to reside in a race condition in the interface syscalls
> > (wavemon and iwconfig kill the machine if run at the same time of a ping
> > flood).
>
> Yes, indeed. Set RID function did not survive CmdCompl changes without
> problems. There was a horrible bug in it that caused deadlocks when any
> other command (e.g., transmit for TX packet) was pending. Set RID called
> hfa384x_cmd() (that can sleep after the CmdCompl changes) while holding
> a spinlock and with interrupts disabled.. This is now fixed in CVS.
Just tested, and it still kills the machine (just with few iwconfig executions), although I don't know if you mean you fixed both, or just the "unrelated" second one.
--
ricardo
A paperless office has about as much a chance as a paperless bathroom