Date: Mon, 17 Jan 2011 21:34:32 +0100 From: Bernhard Schmidt <bschmidt@freebsd.org> To: freebsd-net@freebsd.org Subject: Re: devd and 802.11 Message-ID: <201101172134.32367.bschmidt@freebsd.org> In-Reply-To: <201101052151.46883.bschmidt@freebsd.org> References: <201101052151.46883.bschmidt@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi, no objections so far. If non arise till this weekend, I going to commit it. On Wednesday 05 January 2011 21:51:46 you wrote: > Hi, > > While working on the 'netif restart' issue (kern/153594) I noticed that > there are to entries in devd.conf with shady usefulness. > > detach 0 { > media-type "802.11"; > action "/etc/pccard_ether $device-name stop"; > }; > attach 0 { > media-type "802.11"; > action "/etc/pccard_ether $device-name start"; > }; > > I assume that the general use-case for devd is to automatically configure > wireless devices as soon as those are plugged in (PCMCIA, ..). Those are > already handled by > > notify 0 { > match "system" "IFNET"; > match "type" "ATTACH"; > action "/etc/pccard_ether $subsystem start"; > }; > > this will invoke netif which then will create and configure the wlan0 > device. While wlan0 is created, devd is poked again and this time the > other rule kicks in and starts netif again. This effectively results in 2 > 'netif start wlan0' calls. Due to the racy nature of this all tries to > ensure wpa_supplicant has exclusive access to the wlan device fail, > wpa_supplicant gets started twice and each instance is fighting for > ownership. > > That devd entries a quite old (pre-VAP), the media-type might have changed > and this was just overlooked, so, I did some tests to figure out if those > are really necessary. After removing the 2 entries, I tried 3 things: > - plug in PCMCIA card while system is up > - remove PCMCIA device while system is up > - kldunload if_drv; kldload if_drv > in both cases, the wireless device was configure according to the settings > in rc.conf. Also, cleaning up after kldunload or removal of the device > worked as expected. > > If I'm not missing an use-case here (pointers?) and there are no complains, > the attached diff is worth being committed. > > -- > Bernhard > -- Bernhard
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201101172134.32367.bschmidt>