Arch Linux

Description:

Even when things are working properly, netcfg often reports the following:

Failed to connect to wpa_supplicant - wpa_ctrl_open: No such file or directory

Actually, it is the wpa_check() function in /usr/lib/network/wireless.subr. netcfg does:

1) start wpa_supplicant -wB, which returns immediately (before the daemon has opened its control socket)
2) wpa_check() does: wpa_cli status | grep wpa_state

The wpa_cli call often happens before wpa_supplicant has created the ctrl_interface socket, so wpa_cli fails to connect. This error message is confusing for users who are trying to debug connection problems (and distracting even once the netcfg setup is working).

It can be argued that upstream wpa_supplicant daemon should not return until the control socket is available. See upstream bug http://w1.fi/bugz/show_bug.cgi?id=283

netcfg2 can work around this by just suppressing this error output. Here's a patch:

--- /tmp/wireless.subr.orig 2008-10-21 00:02:13.000000000 -0600
+++ /usr/lib/network/wireless.subr 2008-10-20 23:23:56.000000000 -0600
@@ -9,7 +9,7 @@
[[ -z "$TIMEOUT" ]] && TIMEOUT=15
let timeout=0
while [[ $timeout -ne $TIMEOUT ]]; do
- eval `wpa_cli status|grep wpa_state`
+ eval `wpa_cli status 2> /dev/null | grep wpa_state`
[[ "$wpa_state" = "COMPLETED" ]] && return 0
sleep 1
let timeout++

Steps to reproduce:

Using core/netcfg 2.1.2-1 and core/wpa_supplicant 0.5.10-1

$ cat /etc/network.d/wlan-god
CONNECTION="wireless"
INTERFACE=wlan0
SCAN="yes"
SECURITY="wpa-config"
WPA_CONF="/etc/wpa_supplicant.conf"
#WPA_OPTS="-Dwext -B -iwlan0"
ESSID="god"
IP="dhcp"
DHCP_TIMEOUT=60
TIMEOUT=20

$ cat /etc/wpa_supplicant.conf
ctrl_interface=/var/run/wpa_supplicant
ctrl_interface_group=wheel

network={
ssid="god"
psk=[REMOVED]
}

$ sudo sh -x /usr/bin/netcfg2 wlan-god
...
+ wpa_supplicant -wB -P/var/run/wpa_supplicant_wlan0.pid -iwlan0 -c /etc/wpa_supplicant.conf -Dwext
+ sleep 1
+ [[ ! -f /var/run/wpa_supplicant_wlan0.pid ]]
+ wpa_check wlan0 20
+ INTERFACE=wlan0
+ TIMEOUT=20
+ [[ -z 20 ]]
+ let timeout=0
+ [[ 0 -ne 20 ]]
++ wpa_cli status
++ grep wpa_state
Failed to connect to wpa_supplicant - wpa_ctrl_open: No such file or directory
+ eval
+ [[ '' = \C\O\M\P\L\E\T\E\D ]]
+ sleep 1
+ let timeout++
+ [[ 1 -ne 20 ]]
++ wpa_cli status
++ grep wpa_state
Failed to connect to wpa_supplicant - wpa_ctrl_open: No such file or directory
+ eval
+ [[ '' = \C\O\M\P\L\E\T\E\D ]]
+ sleep 1
+ let timeout++
+ [[ 2 -ne 20 ]]
++ wpa_cli status
++ grep wpa_state
Failed to connect to wpa_supplicant - wpa_ctrl_open: No such file or directory
+ eval
+ [[ '' = \C\O\M\P\L\E\T\E\D ]]
+ sleep 1
+ let timeout++
+ [[ 3 -ne 20 ]]
++ wpa_cli status
++ grep wpa_state
+ eval wpa_state=COMPLETED
++ wpa_state=COMPLETED
+ [[ COMPLETED = \C\O\M\P\L\E\T\E\D ]]
+ return 0
...


I'm calling this "Medium" severity because it is quite misleading regarding why netcfg is failing. In my case, I needed to increase DHCP_TIMEOUT significantly, but it took a long time to determine this because I was trying to find out what was wrong with wpa_supplicant.