Bug#737109: [pkg-wpa-devel] Bug#737109: hostapd: Bridged interface dropped from bridge

2014-01-30 Thread Stefan Lippers-Hollmann
Hi

On Thursday 30 January 2014, Mark Hindley wrote:
[...]
 I am using hostapd in a bridged wlan/eth setup. The wifi card is 
 
  00:08.0 Ethernet controller: Atheros Communications Inc. AR5212/AR5213 
 Wireless Network Adapter (rev 01)
[...]
 When using hostapd/stable, clients using the wlan are sometimes suddenly 
 unable to communicate through the bridge and wlan0 is no longer present 
 in the output of brctl show br0. 
 
 At the same time syslog shows:
 
 /var/log/syslog.2.gz:Jan 19 11:31:07 titan kernel: device wlan0.sta1 entered 
 promiscuous mode
 /var/log/syslog.2.gz:Jan 19 11:31:07 titan kernel: br0: port 3(wlan0.sta1) 
 entering forwarding state
 /var/log/syslog.2.gz:Jan 19 11:31:07 titan kernel: br0: port 3(wlan0.sta1) 
 entering forwarding state
 /var/log/syslog.2.gz:Jan 19 11:31:11 titan ntpd[3468]: Listen normally on 19 
 wlan0.sta1 fe80::20f:3dff:feaa:96f0 UDP 123
 /var/log/syslog.2.gz:Jan 19 11:31:18 titan kernel: wlan0.sta1: no IPv6 
 routers present
 /var/log/syslog.2.gz:Jan 19 11:31:22 titan kernel: br0: port 3(wlan0.sta1) 
 entering forwarding state
 /var/log/syslog.2.gz:Jan 19 11:31:22 titan kernel: device wlan0.sta1 left 
 promiscuous mode
 /var/log/syslog.2.gz:Jan 19 11:31:22 titan kernel: br0: port 3(wlan0.sta1) 
 entering disabled state
 /var/log/syslog.2.gz:Jan 19 11:31:24 titan ntpd[3468]: Deleting interface #19 
 wlan0.sta1, fe80::20f:3dff:feaa:96f0#123, interface stats: received=0, 
 sent=0, dropped=0, active_time=13 secs
 
 Wireless traffic across the bridge can be restored by adding wlan0 back 
 to the bridge with brctl addif br0 wlan0
 
 There is a similar ticket in the OpenWRT lists at 
 https://dev.openwrt.org/ticket/9257
 
 Their fix is https://dev.openwrt.org/changeset/26724
 
 I have rebuilt hostapd with an updated version of the same patch and it 
 also seems to fix the problem for me. Perhaps you would consider including it?
 
 My patch below. The only change I made to the OpenWRT version was to 
 reflect the move of drv-ioctl_sock to drv-global-ioctl_sock and to 
 refresh the line numbers.
[...]

Thanks a lot for investigating this so well and providing a patch, 
which seems to have gotten decent testing and looks to be pretty 
straight forward. However I'm concerned that this particular patch
appears to be around three years old, without having been merged into
hostapd upstream, despite the patch author usually being quite active
in upstream development[1] of these wireless needs...

Given that the old bugtracker at w1.fi no longer exists, I can't 
confirm at the moment if this patch had been submitted upstream and/ or
if it has been rejected for any reasons, which makes me a bit reluctant
to apply it to Debian. So far I haven't come to a conclusion yet and
while this patch might not be part of the very next wpa upload, I'll 
keep it in mind.

Regards
Stefan Lippers-Hollmann

[1] I'm aware that OpenWrt is probably the only party actively working
on 4addr support


signature.asc
Description: This is a digitally signed message part.


Bug#737109: [pkg-wpa-devel] Bug#737109: hostapd: Bridged interface dropped from bridge

2014-01-30 Thread Mark Hindley
On Thu, Jan 30, 2014 at 12:42:07PM +, Stefan Lippers-Hollmann wrote:
[...]
 Thanks a lot for investigating this so well and providing a patch, 
 which seems to have gotten decent testing and looks to be pretty 
 straight forward. However I'm concerned that this particular patch
 appears to be around three years old, without having been merged into
 hostapd upstream, despite the patch author usually being quite active
 in upstream development[1] of these wireless needs...
 
 Given that the old bugtracker at w1.fi no longer exists, I can't 
 confirm at the moment if this patch had been submitted upstream and/ or
 if it has been rejected for any reasons, which makes me a bit reluctant
 to apply it to Debian. So far I haven't come to a conclusion yet and
 while this patch might not be part of the very next wpa upload, I'll 
 keep it in mind.

Thanks. I understand your reluctance. I am not that familiar with 
hostapd, but wanted to share a fix for my issue ;-) 

Having a quick look through the current hostapd GIT 
(http://hostap.epitest.fi/cgit/hostap/tree/src/drivers/driver_nl80211.c), 
I think the patch *functionality* is incorporated:

HUNK 1 appears around line 9376.

The changes to wpa_driver_nl80211_if_remove() are no longer directly 
necessary because the order of the logic has been changed and most of 
the function is skipped for non WPA_IF_AP_BSS by the test

 if (type != WPA_IF_AP_BSS)
return 0; 

at line 9780.

The GIT implementation of wpa_driver_nl80211_if_remove() may well be 
cleaner and more preferable. Perhaps that would be a better fix?

I hope that helps.

Mark


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org