Mailing List Archive

On Thu, 2004-10-14 at 19:56 -0600, Gabriel M. Beddingfield wrote:
> I have an Intel Pro/Wireless 2100 card in a new laptop. The card works fine
> in Knoppix and Windows. I can't get it to work in gentoo.
>
> I followed the install instructions from Intel, compiled it into the kernel
> (not a module). Put the firmware in /usr/lib/hotplug/firmware... all per
> Intel instruction.

Is there any particular reason you're not building it as a module? Why
not try the ebuild first?

--
Andrew Ross <aross@westnet.com.au>


--
gentoo-user@gentoo.org mailing list

Andrew Ross wrote:

> Is there any particular reason you're not building it as a module? Why
> not try the ebuild first?

Reasons:
1. When installing gentoo, the ebuild failed. Compiler errors. I just
moved on to something else. (Was tired after hours of menuconfig, fstab
editing, choosing cron daemons...)
2. Intel says I can.
3. I still needed a kernel compile to enable 802.11g tools, encryption, and
firmware loading.
4. I plan to always have the module loaded... so why not compile it in?
5. I presume that there are performance advantages to not having it as a
module.

:-)

--
Gabriel M. Beddingfield


--
gentoo-user@gentoo.org mailing list

On Thu, 2004-10-14 at 22:49 -0500, Gabriel M. Beddingfield wrote:
> Andrew Ross wrote:
>
> > Is there any particular reason you're not building it as a module? Why
> > not try the ebuild first?
>
> Reasons:
> 1. When installing gentoo, the ebuild failed. Compiler errors. I just
> moved on to something else. (Was tired after hours of menuconfig, fstab
> editing, choosing cron daemons...)

Well, I'm glad to hear you at least tried the ebuild first, even if you
did forget to mention it :-)

With regard to the compile errors, please check bugs.gentoo.org and if
you problem is not already reported, file a bug yourself. If you like,
cc me in the bug report.

> 4. I plan to always have the module loaded... so why not compile it in?

It makes tracking down bugs difficult, because you can't load/unload the
code and pass it different parameters. If you ask for help, pretty much
everyone is going to ask you to unload the modules, clear the kernel
message log (dmesg -c), then load the modules with a set debug level
(modprobe ipw2100 debug=some_val) and post the output of dmesg. Of
course, you won't be able to do that, so they probably won't help you
(or even be _able_ to help you).

It also means you must repatch and recompile your kernel when you want
to use a newer version of ipw2100.

> 5. I presume that there are performance advantages to not having it as a
> module.

I highly doubt that, unless you are talking about some theoretical
performance increase that would only appear on dedicated benchmarks and
only when your kernel is built without _any_ modules and without even
the ability to load modules (which, admittedly, offers an increase in
security, but that's not what we're talking about).

I might have got a little carried away there, but I'm almost certain
that there would be no throughput penalties or increase in resource
consumption from a module.

Cheers

--
Andrew Ross <aross@westnet.com.au>


--
gentoo-user@gentoo.org mailing list

Andrew Ross wrote:

> Well, I'm glad to hear you at least tried the ebuild first, even if you
> did forget to mention it :-)

Well... I can't mention *EVERYTHING* -- then nobody would reply. ;-)

BTW, Thank YOU for your replies!

> With regard to the compile errors, please check bugs.gentoo.org and if
> you problem is not already reported, file a bug yourself. If you like,
> cc me in the bug report.

Before you posted, I took your suggestion and compiled the kernel without
ipw2100... making sure to add support for encryption and the like. Then
'emerge ipw2100'. No compile errors... seemed OK. But no little antenna
light. Not even after 'modprobe ipw2100'.

> everyone is going to ask you to unload the modules, clear the kernel
> message log (dmesg -c), then load the modules with a set debug level
> (modprobe ipw2100 debug=some_val) and post the output of dmesg. Of

# dmesg -c
...lots of output...
# modprobe ipw2100 debug=5
# dmesg
ieee80211_crypt: registered algorithm 'NULL'
ipw2100: Intel(R) PRO/Wireless 2100 Network Driver, 0.54
ipw2100: Copyright(c) 2003-2004 Intel Corporation
ACPI: PCI interrupt 0000:02:02.0[A] -> GSI 11 (level, low) -> IRQ 11
ipw2100: 0000:02:02.0: Detected at mem: 0xD0200000-0xD0200FFF -> e079a000,
irq: 11
eth1: Using hotplug firmware load.
eth1: Firmware 'ipw2100-1.2.fw' not available or load failed.
eth1: ipw2100_get_firmware failed: -2
eth1: Failed to power on the adapter.
eth1: Failed to start the firmware.
#

Looks basically the same. :-(

> It also means you must repatch and recompile your kernel when you want
> to use a newer version of ipw2100.

Surprisingly, on this new laptop that's not so painful. :)

>> 5. I presume that there are performance advantages to not having it as a
>> module.
>
> I highly doubt that, unless you are talking about some theoretical
[snip]

Fair enough.

more info...

# iwconfig
eth0 no wireless extensions.

lo no wireless extensions.

#

Note: the wireless should be eth1, right?

--
Gabriel M. Beddingfield


--
gentoo-user@gentoo.org mailing list

On Fri, 2004-10-15 at 00:40 -0500, Gabriel M. Beddingfield wrote:

> Before you posted, I took your suggestion and compiled the kernel without
> ipw2100... making sure to add support for encryption and the like. Then
> 'emerge ipw2100'. No compile errors... seemed OK. But no little antenna
> light. Not even after 'modprobe ipw2100'.

The LED isn't controlled by ipw2100, but rather in a laptop-model
specific manner eg. http://sourceforge.net/projects/acpi4asus/

> > everyone is going to ask you to unload the modules, clear the kernel
> > message log (dmesg -c), then load the modules with a set debug level
> > (modprobe ipw2100 debug=some_val) and post the output of dmesg. Of
>
> # dmesg -c
> ...lots of output...

"dmesg -c > /dev/null" :-)

> # modprobe ipw2100 debug=5
> # dmesg
> ieee80211_crypt: registered algorithm 'NULL'
> ipw2100: Intel(R) PRO/Wireless 2100 Network Driver, 0.54
> ipw2100: Copyright(c) 2003-2004 Intel Corporation
> ACPI: PCI interrupt 0000:02:02.0[A] -> GSI 11 (level, low) -> IRQ 11
> ipw2100: 0000:02:02.0: Detected at mem: 0xD0200000-0xD0200FFF -> e079a000,
> irq: 11
> eth1: Using hotplug firmware load.
> eth1: Firmware 'ipw2100-1.2.fw' not available or load failed.
> eth1: ipw2100_get_firmware failed: -2
> eth1: Failed to power on the adapter.
> eth1: Failed to start the firmware.
> #
>
> Looks basically the same. :-(

For starters, the latest version of ipw2100 is 0.56, not 0.54. The
latest ebuild is 0.55-r2, so try that first.

What version of hotplug are you using? What is the output of
"cat /proc/sys/kernel/hotplug" ?

> more info...
>
> # iwconfig
> eth0 no wireless extensions.
>
> lo no wireless extensions.
>
> Note: the wireless should be eth1, right?

Yep.

--
Andrew Ross <aross@westnet.com.au>


--
gentoo-user@gentoo.org mailing list

I'm currently using ipw2100 v0.52 without a problem in my HP nx7000. So
I may have some hints for you:

1. The ipw2100 firmware will be stored in /usr/lib/hotplug/firmware
and consists of ipw2100-1.2.fw, ipw2100-1.2-p.fw and ipw2100-1.2-i.fw

2. If these files are present verify that your kernel is correctly
configured for WLAN e.g.:

CONFIG_KMOD=Y
CONFIG_CRYPTO_ARC4=Y
CONFIG_NET_RADIO=Y
CONFIG_FW_LOADER=Y (!!!)

These are kernel 2.6.8.1 settings BTW.

3. emerge wireless-tools

4. Power your WLAN device!!! Most notebooks (like mine) have an button
to turn on/off the radio device. Normally the radio is shut off by default.

5. Here's a little script that I use to start my WLAN

Assumptions:
a.) ipw2100 is loaded
b.) eth0 is a built-in NIC
c.) eth1 is your WLAN device
d.) we're using WEP
e.) the access point is acting as a DHCP server
f.) xxx is the ESSID
g.) yyy is the channel
h.) zzz is the WEP keyphrase

#!/bin/bash
/etc/init.d/net.eth0 stop
ifconfig eth1 down
iwconfig eth1 mode monitor
#iwconfig eth1 sens -90
iwconfig eth1 essid xxx
iwconfig eth1 channel yyy
iwconfig eth1 key restricted s:zzz
iwconfig eth1 mode auto
ifconfig eth1 up
dhcpcd eth1
ifconfig eth1

As I said: that worked for me.

Have fun!

Andrew Ross schrieb:
> On Fri, 2004-10-15 at 00:40 -0500, Gabriel M. Beddingfield wrote:
>
>
>>Before you posted, I took your suggestion and compiled the kernel without
>>ipw2100... making sure to add support for encryption and the like. Then
>>'emerge ipw2100'. No compile errors... seemed OK. But no little antenna
>>light. Not even after 'modprobe ipw2100'.
>
>
> The LED isn't controlled by ipw2100, but rather in a laptop-model
> specific manner eg. http://sourceforge.net/projects/acpi4asus/
>
>
>>>everyone is going to ask you to unload the modules, clear the kernel
>>>message log (dmesg -c), then load the modules with a set debug level
>>>(modprobe ipw2100 debug=some_val) and post the output of dmesg. Of
>>
>># dmesg -c
>>...lots of output...
>
>
> "dmesg -c > /dev/null" :-)
>
>
>># modprobe ipw2100 debug=5
>># dmesg
>>ieee80211_crypt: registered algorithm 'NULL'
>>ipw2100: Intel(R) PRO/Wireless 2100 Network Driver, 0.54
>>ipw2100: Copyright(c) 2003-2004 Intel Corporation
>>ACPI: PCI interrupt 0000:02:02.0[A] -> GSI 11 (level, low) -> IRQ 11
>>ipw2100: 0000:02:02.0: Detected at mem: 0xD0200000-0xD0200FFF -> e079a000,
>>irq: 11
>>eth1: Using hotplug firmware load.
>>eth1: Firmware 'ipw2100-1.2.fw' not available or load failed.
>>eth1: ipw2100_get_firmware failed: -2
>>eth1: Failed to power on the adapter.
>>eth1: Failed to start the firmware.
>>#
>>
>>Looks basically the same. :-(
>
>
> For starters, the latest version of ipw2100 is 0.56, not 0.54. The
> latest ebuild is 0.55-r2, so try that first.
>
> What version of hotplug are you using? What is the output of
> "cat /proc/sys/kernel/hotplug" ?
>
>
>>more info...
>>
>># iwconfig
>>eth0 no wireless extensions.
>>
>>lo no wireless extensions.
>>
>>Note: the wireless should be eth1, right?
>
>
> Yep.
>

--
Mit freundlichen Grüßen

Heinz Sporn

SPORN it-freelancing

Mobile: ++43 (0)699 / 127 827 07
Email: heinz.sporn@sporn-it.com
heinz.sporn@utanet.at
Snail: Steyrer Str. 20
A-4540 Bad Hall
Austria / Europe

--
gentoo-user@gentoo.org mailing list

Andrew Ross wrote:

> The LED isn't controlled by ipw2100, but rather in a laptop-model
> specific manner eg. http://sourceforge.net/projects/acpi4asus/

It wouldn't surprise me if it was tied to the hardware... that appears to be
how it's done on the Thinkpads in general. Regardless... all I know is
that if the light is on -- wireless can work. If it's off -- then there is
a problem with the card. IIUC, the ipw2100 driver "turns on" the card's
hardware and loads firmware into it.

> "dmesg -c > /dev/null" :-)

:-)

> For starters, the latest version of ipw2100 is 0.56, not 0.54. The
> latest ebuild is 0.55-r2, so try that first.
>
> What version of hotplug are you using? What is the output of
> "cat /proc/sys/kernel/hotplug" ?

.54 was what emerge loaded. Perhaps .55 is masked? I'll play with it
later.

In the mean-time... the problem is resolved. See other post.

Thanks for all your help, Andrew! You're a gem!

--
Gabriel M. Beddingfield


--
gentoo-user@gentoo.org mailing list

Heinz Sporn wrote:

> 1. The ipw2100 firmware will be stored in /usr/lib/hotplug/firmware
> and consists of ipw2100-1.2.fw, ipw2100-1.2-p.fw and ipw2100-1.2-i.fw

Heinz: Thanks! That's where the firmware was loaded. It reminded me to go
and change FIRMWARE_DIR back to /usr/lib/hotplug/firmware. When compiled
in kernel, I tried different settings of this variable and moved firmware
to different locations to see if I could stumble on an answer.

This corrects the problem!! Thanks a bunch!

Also, thanks for the script in your post. That is my next step!

> Have fun!

Believe it or not... I AM!

--
Gabriel M. Beddingfield


--
gentoo-user@gentoo.org mailing list