Puppy Linux Discussion Forum

[tempestuous]

These Linux wifi drivers are for Puppy 2.0, 2.01 and the remastered "2.02-regression".
They are not compatible with the standard Puppy 2.02

Update Aug 2006: The download location has changed slightly - http://dotpups.de/dotpups/Wifi/

A dedicated driver is one written specifically for Linux, as opposed to ndiswrapper (and the commercial Linuxant DriverLoader) which "wraps" a Windows driver.
Dedicated Linux drivers are the optimal solution, but usually involve extra effort. I have not provided a fancy wizard, you need to read the relevant README to use these drivers.

ndiswrapper has its place. It's a less complicated solution (you don't even need to know what chipset you have) and it's your only option when there's no Linux driver available.
This download site also has wifi-1.0.4-beta-2.pup (originally packaged by Bladehunter) which automates and configures ndiswrapper. It contains many popular Windows drivers ready-to-go.

So to the new drivers. With Puppy2's 2.6.16 kernel there's some bad news:
- atmelwlandriver won't compile, but there is a standard kernel driver for Atmel PCI & PCMCIA devices which should be fine.

EDIT July 20 2006, the BerliOS at76c503a driver for Atmel USB devices is now compatible.

- linux-wlan-ng (for Prism2/3) - the PCMCIA versions won't compile, but that's no problem, the HostAP driver will substitute. Fortunately the linux-wlan-ng USB driver is OK.

- Lucent-Agere Hermes1&2 (proprietary) - won't compile, but there's a standard kernel driver which should be just as good.

The good news is that there are now 4 previously unavailable wifi drivers:

Broadcom bcm43xx - I have provided the "softmac" version which is considered more developed and stable, plus the "devicescape" version which supports full (not just PSK) WPA encryption. Both versions contain additional wifi network modules, which, in the case of the softmac version are designed to overwrite Puppy2's existing wifi modules.

NOTE: Several of these driver packages here contain optimised/modified wifi network modules. Don't try to "mix-and-match" these modules. The Linux wifi network stack is undergoing change aimed at new features, but several different versions of the stack exist in the meantime.

Prism54 usb/softmac - this is the "islsm" version which supports USB adaptors with Prism54 softmac chips, and PCI/PCMCIA softmac versions.

Marvell - "mrv8k" driver for Marvell Libertas chipsets.

ipw3945 - for the Intel PRO/Wireless 3945ABG.

EDIT June 26/July 9 2006 - The "new code" rt2x00 Ralink drivers are not stable.
So the old code versions are here, too, which I have called "Ralink-legacy" and this package includes the recent rt61 and rt73 drivers.

Plus there are updated versions of drivers I earlier provided for Puppy 1.0.4-1.0.9 at http://mymirrors.homelinux.org/puppy/wireless/
- acx (TexasInstruments), adm8211, Hostap (Prism2/3), HostAP Daemon, ipw2100, ipw2200, linux-wlan-ng (Prism2/3 USB), MADWiFi (Atheros), Orinoco-with-rfmon-patch, Prism54, ralink-legacy, rt2x00 (Ralink), rtl8180-sa2400 (Realtek), zd1211 (Zydas), and wpa_supplicant.

I have also included "Aircrack-ng" at kirk's suggestion, for wifi network packet collection.

edit June 26 2006 - kirk's "Kismet" dotpup is now here, too. Kismet is an older wifi sniffer than Aircrack-ng. It's much larger, but has a simple gui, and will work with a larger range range of wifi adaptors.
The patched version of the orinoco drivers is now here, too, which adds support for Kismet.

[Lobster]

Good job

Wi-fi is the future no doubt. Which would you say is the best Puppy wifi card for people upgrading? At the moment I am hard wired through a router modem combo. Others will be very pleased with this . . .

Well done. Reports and testers needed?

[tempestuous]

I have a Prism2-based USB adaptor myself (by D-Link) with the linux-wlan-ng driver. This works rock-solid, but it's an old b-mode device.

From what I've heard, Ralink-based devices are good. I can say that I'm impressed by the "trim" nature of its driver, the rt2x00. For models containing this chip see here http://ralink.rapla.net/

Atheros chips are well regarded, but there's no driver for Atheros USB models.
PrismGT chips are well regarded, but the latest PrismGT chips are "softmac" and the driver for this is very new.

Intel PRO/Wireless chips are not well regarded.
I know someone who has a Marvell-based device which has very poor range.

[Lobster]

Many thanks - here is a wi-fi code that might be useful for city dwellers with lap dogs
http://news.com.com/2100-1033-939546.html

[BlackAdder]

Sorry to report that the rt2x00 driver does not work for me. The wireless adapter is an MSI PC54G2 which has an RT2500 chipset.
The modules 80211, rate_control and rt2500 get loaded at bootup without any intervention by me. iwconfig then shows wmaster0 and wlan0 (but not ra0). But wlan0 will not accept a WEP key and/or associate with an access point.
The same adapter works fine with ndiswrapper and the MSI Windows XP drivers.
The test was done using the Seamonkey version of Puppy 2.00.
Will try a couple of other adapters in the next day or two.

[tempestuous]

Sorry to hear. A few suggestions -

Run dmesg to see if there are any clues.

Bring down the interface (obviously "wlan0" has replaced "ra0"), unload the modules, and start again.
Turn off encryption at the access point while testing.
And use a static IP address.

ifconfig wlan0 down
rmmod rt2500pci # there is no module called "rt2500"
rmmod rate_control
rmmod 80211

modprobe 80211
modprobe rate_control
modprobe rt2500pci
ifconfig wlan0 up
iwconfig wlan0 essid MY_ESSID key none mode managed
## modify /etc/resolv.conf to include your nameservers
ifconfig wlan0 192.168.0.xx broadcast 192.168.0.255 netmask 255.255.255.0
route add default gw 192.168.0.1 wlan0 # or whatever your router's IP is

Also, do you have a "hardware button" on your PCI card? The source README says this -

"Hardware button
When the pci card contains a hardware button for the Wireless device, and the pci modules have been build with hardware button capabilities enabled (Which is the default build option, and can't be disabled), the card will send out the ACPI event ACPI_BUTTON_NOTIFY_STATUS for ACPI device <modulename> with class <modulename>_button.
When 1 is reported the button has been enabled.
When 0 is reported the button has been disabled."

If all else fails, try pressing the button!

[cplater]

Just want to report some good results. I have a Devon IT thin client that originally came with Puppy 1.0.4. Just upgraded to 2.0 on a larger 512 MB flash card. In the single slot I have a Dlink 510 PCI wireless card that has the atheros chip set. Downloaded the new mad-wifi package and installed it. Everything is running smoothly now and is talking WEP with a Zyxel mimo router. Excellent work. Bravo!

[raffy]

Thank you very much, tempestuous, for looking after wi-fi in Puppy.

Maybe we can put up later a list of cards that are easily used in Puppy2. Am glad Marvell has a few cards ready, as it is the brand that will be used in OLPC because of its innovations. (Eureka: it is the Marvell 88W8388 build).

Devon IT is quite enthusiastic about using Puppy2 in their thin clients, so it is good that cplater has already tested one card in it.

[BlackAdder]

tempestuous,
Have been through the steps that you outlined; without success I am afraid.
There were no errors showing in dmesg, but ACPI had apparently played with the interrupt when the modules loaded, so tried with acpi=off. No change to behaviour.
Of course you are correct that there is no rt2500 module in the new driver set - old Puppy 1xx habits die hard.
Cannot get to the point where ifconfig will work.
Absolutely no sign of a hardware button on the card.
This message coming to you courtesy of Puppy 109CE and the Puppy 1xx rt2500 module.
Perhaps someone else with an rt2500-based card could test also.

[BlackAdder]

Tested the laptop with a Spirit 802.11b wifi card from Dick Smith - has the rtl8180 chipset. Was remaindered by DSE about a year ago. Really cheap, like me.
Here's what worked for me:

[tempestuous]

Thanks for testing BlackAdder.
I just browsed the rt2x00 forums, and it seems that the new ralink drivers are still very much at beta stage ... although I thought the rt2500pci should be OK.
So I have compiled the old code version, now attached. It contains the "RaConfig2500" setup utility, found in /root/my-applications/bin/
It would be great if you could try this. You will need the qt3 Pupget.

Edit June 27 2006 - removed rt2500.pup.
Use ralink-legacy.pup from http://www.dotpups.de/dotpups/Wifi/wireless-drivers/

I have noted your report that the dhcpcd pid file still needs to be removed. I will amend all the README's shortly. There is sure to be more changes needed.

[tempestuous]

Darn, I have just read about quite a few more problems with the rt2x00 (new code) drivers.
So I will shortly provide the old code versions, which I will call "ralink-legacy". These, too, are sourced from rt2x00.serialmonkey.com. They're enhancements of the official Ralink drivers.

In the meantime, anyone trying the "new code" rt61pci or rt73usb drivers, you should first run "modprobe firmware_class" before any of the other modules listed in the README.

... and it would be good if someone could tell me if the RaConfig2500 utility is worth keeping in the package.

[pakt]

I finally had a chance to test two cards that have rt2500 chipsets: ASUS WL-107G and Gigabyte GN-WMKG. Neither worked with rt2x00-20060508.pup, even when I ran "modprobe firmware_class" first.

Fortunately rt2500.pup worked. I used these commands:

modprobe rt2500
iwconfig ra0 essid <my AP>
ifconfig ra0 up <-- was needed
dhcpcd -t 30 -h puppypc -d ra0

An interesting observation: with the Windows driver, the Tx-Power was 20dBm. With rt2500.pup, the Tx-Power was -3dBm. 'Link Quality' reflects those settings. Is there any way to set the 'Tx-Power' level, or should I use the Windows driver to get best range?

With rt2500.pup:

Bit Rate:11 Mb/s Tx-Power:-3 dBm
Link Quality=88/100 Signal level=-48 dBm Noise level:-193 dBm

With ndiswrapper:

Bit Rate:11 Mb/s Tx-Power:20 dBm Sensitivity=-109 dBm
Link Quality:100/100 Signal level:-37 dBm Noise level:-256 dBm

These tests where done using the 'standard' Puppy2 kernel. I'll test with my speedstep kernel as soon as I can and report back.

Oh yes, I didn't have qt installed so I couldn't test the RaConfig2500 utility. I try to do that too.

Thanks, Tempestuous, for making the drivers available.

Paul

[tempestuous]

Most wifi device settings are accessed with iwconfig or iwpriv. I searched for "Tx-Power" on the rt2x00 forum and found this http://rt2x00.serialmonkey.com/phpBB2/viewtopic.php?t=1225&start=0
As you can see, the subject is rather complex. From what I can tell, it used to be the case that "iwconfig ra0 txpower 0" would set the highest Tx-Power that the device is capable of. But it seems that the developers have changed that regime as of May 15 2006. My rt2500 compilation was from the June 9 2006 CVS snapshot, so should have the new system.
So you will just have to experiment with "iwconfig ra0 txpower xxx"

And apart from range, Linux wifi drivers are supposed to achieve better throughput. Try downloading the same file from a reliable server using the different drivers and see if one is faster than the other.

[pakt]

Thanks for the info, tempestuous.