ThinkPad 11a/b/g/n Wireless LAN Mini Express Adapter
|
This is a WiFi Adapter that is installed in a Mini-PCI Express slot IBM partnumber 42T0825 [1] ContentsFeatures
|
|
Identification
To determine the chipset your card uses, issue the following commands:
# update-pciids # lspci | egrep -i 'network|atheros|wireless' 03:00.0 Network controller: Atheros Communications, Inc. AR5418 802.11a/b/g/n Wireless PCI Express Adapter (rev 01)
If you get something different than above despite having the most current PCI IDs, please report it here!
Linux WiFi drivers
madwifi
For native Linux support, a development snapshot of the "madwifi" driver, which uses a non-free binary-only HAL, can be used. It does not support the draft 11n protocol, but will work in "legacy" 11g mode. Support for this chipset was initially planned for release 0.9.4, but this inclusion had to be postponed as a critical update with a bug fix for compilation with the 2.6.24 kernel was necessary before the pre-release trunk was sufficiently stable. Thus, it is still necessary to download the prerelease snapshot from trunk using subversion.
There is a howto, which describes the procedure for getting the snapshot to work.
There is an old ticket for this card at madwifi, #1001.
There is an new ticket for this card at madwifi-branch, #1243. Insert non-formatted text here
ath9k
There is now active development of a free driver ath9k that is purported to support the draft 11n protocol and is aiming for inclusion in the mainline Linux kernel. If you're impatient and want to try it out NOW, however, you'll have to Howto install the development version of atk9k.
Using the Windows Driver in Linux
If you have a weak stomach for pre-release software, you can always use "ndiswrapper" (>= 1.29) to wrap the Windows driver supplied by Lenovo. This isn't as bad as you think. It does work like a charm, but you may have problems if you're using a 64 bit kernel since it's not clear that a 64 bit windows XP driver exists (ndiswrapper currently doesn't support Vista drivers). Here's the Howto.
Problems/Bugs?
Non-Maskable Interrupt with madwifi
A number of folks have reported getting errors while using the experimental madwifi driver with the AR5418. After hours of flawless operation, the Kernel sometimes throws an NMI after which, the wifi dies. Aside from rebooting, suspending (to ram or disk) and resuming seems to be the only method to recover.
Uhhuh. NMI received for unknown reason b0 on CPU 0. You have some hardware problem, likely on the PCI bus. Dazed and confused, but trying to continue
wifi0: rx FIFO overrun; resetting wifi0: rx FIFO overrun; resetting wifi0: rx FIFO overrun; resetting
See the following threads and bug reports: [2] [3] [4] [5] [6]
A possible preventitive measure is issuing the command
# iwpriv <device> bgscan 0
each time you load the driver. To make this change "permanent", you could add this command to your distribution's ifup networking scripts as described below.
Debian
Open up /etc/network/interfaces in a text editor and find the entry for your wireless device. Which should look something like
iface wlan0 inet manual
wpa-driver madwifi
wpa-roam /etc/network/wpa_supplicant.conf
and add the line
post-up iwpriv wlan0 bgscan 0
to the end of it where of course you substitute your interface name (e.g., "ath0") for "wlan0".
AP dropout with madwifi
You may experience random disconnections from the AP when using the madwifi driver (madwifi ticket 1004). It seems to be loosely correlated with how busy the AP is (or possibly how much interference is present). This would appear to be a beacon miss problem (the AP is too busy to acknowledge the existence of your station at regular enough intervals) which can be confirmed by running
# echo 0xc80000 > /proc/sys/net/<device>/debug
and then
$ dmesg
right after you have been disconnected. If it says something about a beacon miss, you might see an improvement by upgrading to the most recent subversion of the driver. It appears that there has been significant improvement on this problem.
There is also a recently added private ioctl which controls how many beacon misses (or alternatively how many beaconless milliseconds) the card will tolerate before dropping the connection and trying to reassociate. You can confirm that your version of the driver has this feature by running
$ /sbin/iwpriv -a 2>/dev/null . Which should give output that looks something like
wlan0 get_bmiss_ms:20480 wlan0 get_bmiss:200
The numbers in the above output are rather high compared to the built in defaults, but it is probably better to have them too high than too low seeing as how reassociation is far from an instantaneous process and you probably don't need a fast dissociation reflex unless you see a significant performance improvement upon reassocation. To set these values use
# iwpriv <device> bmiss <#of misses> or
# iwpriv <device> bmiss_ms <#of ms of misses allowed>
Similar the iwpriv commands in the previous section, you should add these commands to your ifup network scripts as described below to apply them persistently.
Debian
Open up /etc/network/interfaces in a text editor and find the entry for your wireless device. Which should look something like
iface wlan0 inet manual
wpa-driver madwifi
wpa-roam /etc/network/wpa_supplicant.conf
and add the line
pre-up iwpriv wlan0 bmiss 200
to the end of it where of course you substitute your interface name (e.g., "ath0") for "wlan0".
No Draft n support in Linux (madwifi and ndiswrapper)
Once you have your shiny new wireless card working, you can give yourself a pat on the back for a job well done. However, if your are a perfectionist and enough of a sucker to go for a pricey draft n router, your job is not done. Well actually maybe it is and mayhap you shouldn't go for that router either. Though the card can indeed be made to work under Linux, it does not appear that it will operate under the draft n protocol.
Tests with the DLink DIR-655, which uses the same AR5008 chipset family, proved discouraging. The first test using the windows driver natively in windows does indeed work. The connection should be listed as 300Mbit/s and should connect in the router's "802.11ng only" mode. Not having access to any linux benchmarking tools, there was only ssh which gave transfer speeds of ~3.5MByte/s. Not exactly the 10x as promised by the box even when accounting for ssh overhead, but a decent improvement to 11g speeds.
As for Linux, things get a little dicey. Using no authentication/encryption, wpa_supplicant (with -D wext) and the ndiswrapper driver, the card is only able to connect if the router is set to "802.11g only" mode. In "Mixed 802.11ng, 802.11g and 802.11b" mode, there is only a brief period of connectivity where no ip address is assigned and then the connection is dropped. In "802.11ng only" mode, no connection at all.
Though it does appear that there are some out there looking at 11n for the new ath5k open-hal branch of madwifi which has yet to support this chip at all[7], this ticket does not bode well for the current trunk madwifi driver. Indeed, using the madwifi driver (and giving wpa_supplicant either the -D madwifi or -D wext option) is more promising, but the result is still the same: no 11n connection. Using the router's "Mixed 802.11ng, 802.11g and 802.11b" mode does work, but only provides (admittedly solid) 11g speeds of 20*10^6bits/sec using the TCP_STREAM benchmark in the netperf suite. Again, however "802.11ng only" mode doesn't work. A connection seems to happen and will sit there for some time while no ip address is assigned even by invoking dhclient manually.
The kernel message upon probing the madwifi ath_pci module shows no 11n rates message with corresponding 300Mbps confirming that such capabilities are not within the scope of the driver.
$modprobe ath_pci $dmesg . . . [28696.420000] ath_hal: 0.9.30.13 (AR5210, AR5211, AR5212, AR5416, RF5111, RF5112, RF2413, RF5413, RF2133) [28696.430000] wlan: 0.8.4.2 (svn r2708) [28696.430000] ath_pci: 0.9.4.5 (svn r2708) [28696.430000] ACPI: PCI Interrupt 0000:03:00.0[A] -> GSI 17 (level, low) -> IRQ 21 [28696.430000] PCI: Setting latency timer of device 0000:03:00.0 to 64 [28696.570000] ath_pci: switching rfkill capability off [28696.580000] ath_rate_sample: 1.2 (svn r2708) [28696.580000] ath_pci: switching per-packet transmit power control off [28696.580000] wifi0: 11a rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps [28696.580000] wifi0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps [28696.580000] wifi0: 11g rates: 1Mbps 2Mbps 5.5Mbps 11Mbps 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps [28696.580000] wifi0: turboA rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps [28696.580000] wifi0: turboG rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps [28696.580000] wifi0: H/W encryption support: WEP AES AES_CCM TKIP [28696.580000] wifi0: mac 12.10 phy 8.1 radio 12.0 [28696.580000] wifi0: Use hw queue 1 for WME_AC_BE traffic [28696.580000] wifi0: Use hw queue 0 for WME_AC_BK traffic [28696.580000] wifi0: Use hw queue 2 for WME_AC_VI traffic [28696.580000] wifi0: Use hw queue 3 for WME_AC_VO traffic [28696.580000] wifi0: Use hw queue 8 for CAB traffic [28696.580000] wifi0: Use hw queue 9 for beacons [28696.580000] wifi0: Atheros 5418: mem=0xedf00000, irq=21 [28696.590000] udev: renamed network interface ath0 to wlan0
Hardware switch
Some ThinkPads have a hardware switch that must be in the on position for the radio to work, regardless of driver state:
In addition to hard-switching the wireless card, the switch also generates an acpi event on transition from hi->lo and vice versa. It is however the same event in both directions.
