kern/144755: [iwi] [panic] iwi panic when issuing /etc/rc.d/netif restart on 8-STABLE r205159

[Bernhard Schmidt]

The following reply was made to PR kern/144755; it has been noted by GNATS.

From: Bernhard Schmidt <bschmidt at techwires.net>
To: Edwin Groothuis <edwin at mavetju.org>, bug-followup at freebsd.org
Cc:  
Subject: Re: kern/144755: [iwi] [panic] iwi panic when issuing /etc/rc.d/netif 
	restart on 8-STABLE r205159
Date: Thu, 5 Aug 2010 10:08:42 +0200

 On Thu, Aug 5, 2010 at 00:31, Edwin Groothuis <edwin at mavetju.org> wrote:
 > Hello Bernhard!
 >
 > On Wed, Jul 28, 2010 at 12:11:17PM +0200, Bernhard Schmidt wrote:
 >> Can you check if wpa_supplicant gets started twice after
 >> "/etc/rc.d/netif restart"?
 >
 > According to core.txt there are two wpa_supplicant processes, but
 > I don't know if it is caused by a fork of itself or if it is caused
 > by something started twice:
 >
 > =A0 =A00 =A07084 =A0 =A0 1 =A0 0 =A056 =A00 =A05192 =A0 =A0 0 select Ds =
 =A0 =A0?? =A03686:38.00 [wpa_suppli
 > =A0 =A00 =A07085 =A0 =A0 1 =A0 0 =A057 =A00 =A05192 =A0 =A0 0 select Ds =
 =A0 =A0?? =A02617:11.00 [wpa_suppli
 >
 > Based on the PPID of 1 for both, I would say it gets started twice.
 
 Indeed.
 
 >
 > [..]
 >> I have the feeling that this is a more or less known issue, a race
 >> between devd (/etc/pccard_ether) and /etc/rc.d/netif. There is a small
 >> window (a few ms) where this can happen and our net80211 isn't capable
 >> of handling two running wpa_supplicants on the same interface. Which
 >> leads to all kind of weird issues, .e.g. panics.
 >
 > Nice! Is there a workaround or patch available for testing?
 
 A temporary workaround is to add a "sleep 1" in /etc/pccard_ether,
 something I'd rather not commit. I haven't yet found a clean solution
 for this.. any pointers welcome. :)
 
 --=20
 Bernhard