Skip to contentSkip to conversation

iwl3945 driver is still unusable!

DISCUSSION Thomas Bächler 4 participants 9 messages Jan 13, 2008 – Jan 20, 2008

lore
Share & tools
lore
#0 lore raw

Some history of this problem can be found here:
http://bughost.org/bugzilla/show_bug.cgi?id=1468

I want to describe my problem here again to get some more people to pay attention, as Intel doesn't seem to be able to reproduce this problem, although it is quite easy and I am not the only one who experiences it.

I will start with my setup: I own a Lenovo 3000 N100, which includes a Core 2 Duo CPU and an Intel 3945ABG Wireless LAN chip. I am working with Archlinux x86_64 and use wpa_supplicant 0.5.9 (0.5.8 and 0.5.7 in the past) combined with autowifi, a small tool that starts dhclient automatically once a wireless connection has been established by wpa_supplicant.

I am roaming with my Laptop a lot, so I have to switch between networks often. After I had some problems with the driver over half a year ago, I
figured out this test case:

- Connect to my own WPA network.
- Tell wpa_supplicant to switch to an open network.
- Switch to another open network or (if none available) back to my WPA network.

The first network I use after loading the driver always works fine. After some switching networks (always the second or at most third network I try to connect to), I don't get DHCP anymore. Sometimes I still get a DHCPNAK, but after that, there is no response to my DHCPDISCOVER messages. When apppropriate, I assign a static IP and still cannot transmit any traffic. However, I can always reconnect to the last working network (so long-term, the driver works with exactly one network).

One of the debug outputs I generated for the bug report (see above) suggests that only the first package is transmitted over the link, after that no data is transmitted at all!

This problem persists even if I
- restart wpa_supplicant
- unload the driver and the mac80211 stack completely and reload them
- do both of the above and then suspend to RAM, then resume later and try again

This issue is not limited to switching between encrypted and unencrypted networks. I am pretty sure that if I connect to my WPA network here at home, go to the university and connect to their WPA network, the issue persists (however, this is not a practical test case, as there is a 30 minute walk in between).

These are the test cases I did today (I could get similar results with older kernel versions and drivers):

Linux 2.6.23.13, iwlwifi 1.2.23 from Intel's website:
- Connect to WPA network succeeds
- Connect to an open network succeeds
- Going back to the WPA network: Associates fine, but no DHCP, with static IP no traffic can be transmitted

Linux 2.6.24-rc7-gd0c4c9d4 (latest git as of today), included iwl3945 version:
- Connect to WPA network succeeds
- Connect to an open network: Associates fine, no DHCP (same network that I could connect to in the last test).

Linux 2.6.24-rc7-gd0c4c9d4 (latest git as of today), iwlwifi 1.2.23:
- Does not compile!

All these test cases _always_ worked perfectly with the old ipw3945 driver. Unloading the iwl driver and loading the ipw driver always makes my connection work again. BUT, the ipw3945 driver has been discontinued and does not compile anymore with 2.6.24 (as iwl3945 is considered a replacement, which it is obviously not).
All this means, if I update to 2.6.24 (which I want to, due to several improvements), I have no usable driver for my wireless anymore.

What I need is a solution, meaning either a working iwl3945 version or a patch to ipw3945 to compile with 2.6.24.

Thanks in advance for any help or comments.

#1 lore raw

I have a Thinkpad T61. Though I have a feeling your problem could be your software combination. For kicks could try Ubuntu 7.10 (even as just running from a Live CD) and see what happens. I often roam a lot with my T61, but I do have an Intel 4965 wireless chipset. The Gutsy Live CD can be found here:
http://www.ubuntu.com/getubuntu/download

I have been using Gnome & Network Manger with WPA2 at home & some other places without issue (all using DHCP).

Could you try it out and let the list know of the result. You actually may have found a real bug. Though it seems like other things may be in the way. So something know nto be somewhat sane and seeing the same result whould help a lot. So try running the Gutsy Live CD and using Gnome & Network Manger and see if you see the same results switching between these networks. This issue could just be the combination of software you are using...it's worth a shot.

Quoted 4 lines · On Jan 13, 2008 3:33 AM, Thomas B=E4chler <[email protected]> wrote:

Some history of this problem can be found here:
http://bughost.org/bugzilla/show_bug.cgi?id=3D1468

I want to describe my problem here again to get some more people to p=

ay

Quoted 1 line

attention, as Intel doesn't seem to be able to reproduce this problem=

,

Quoted 1 line

although it is quite easy and I am not the only one who experiences i=

t.

Quoted 2 lines

I will start with my setup: I own a Lenovo 3000 N100, which includes =

a

Quoted 1 line

Core 2 Duo CPU and an Intel 3945ABG Wireless LAN chip. I am working w=

ith

Quoted 6 lines

Archlinux x86_64 and use wpa_supplicant 0.5.9 (0.5.8 and 0.5.7 in the past) combined with autowifi, a small tool that starts dhclient automatically once a wireless connection has been established by wpa_supplicant.

I am roaming with my Laptop a lot, so I have to switch between networ=

ks

Quoted 1 line

often. After I had some problems with the driver over half a year ago=

, I

Quoted 5 lines

figured out this test case:

- Connect to my own WPA network.
- Tell wpa_supplicant to switch to an open network.
- Switch to another open network or (if none available) back to my WP=

A

Quoted 7 lines

network.

The first network I use after loading the driver always works fine. After some switching networks (always the second or at most third network I try to connect to), I don't get DHCP anymore. Sometimes I still get a DHCPNAK, but after that, there is no response to my DHCPDISCOVER messages. When apppropriate, I assign a static IP and st=

ill

Quoted 1 line

cannot transmit any traffic. However, I can always reconnect to the l=

ast

Quoted 1 line

working network (so long-term, the driver works with exactly one netw=

ork).

Quoted 3 lines

One of the debug outputs I generated for the bug report (see above) suggests that only the first package is transmitted over the link, af=

ter

Quoted 9 lines

that no data is transmitted at all!

This problem persists even if I
- restart wpa_supplicant
- unload the driver and the mac80211 stack completely and reload them
- do both of the above and then suspend to RAM, then resume later and try again

This issue is not limited to switching between encrypted and unencryp=

ted

Quoted 1 line

networks. I am pretty sure that if I connect to my WPA network here a=

t

Quoted 1 line

home, go to the university and connect to their WPA network, the issu=

e

Quoted 1 line

persists (however, this is not a practical test case, as there is a 3=

0

Quoted 3 lines

minute walk in between).

These are the test cases I did today (I could get similar results wit=

h

Quoted 14 lines

older kernel versions and drivers):

Linux 2.6.23.13, iwlwifi 1.2.23 from Intel's website:
- Connect to WPA network succeeds
- Connect to an open network succeeds
- Going back to the WPA network: Associates fine, but no DHCP, with static IP no traffic can be transmitted

Linux 2.6.24-rc7-gd0c4c9d4 (latest git as of today), included iwl3945 version:
- Connect to WPA network succeeds
- Connect to an open network: Associates fine, no DHCP (same network that I could connect to in the last test).

Quoted 5 lines

Linux 2.6.24-rc7-gd0c4c9d4 (latest git as of today), iwlwifi 1.2.23:
- Does not compile!

All these test cases _always_ worked perfectly with the old ipw3945 driver. Unloading the iwl driver and loading the ipw driver always ma=

kes

Quoted 1 line

my connection work again. BUT, the ipw3945 driver has been discontinu=

ed

Quoted 3 lines

and does not compile anymore with 2.6.24 (as iwl3945 is considered a replacement, which it is obviously not).
All this means, if I update to 2.6.24 (which I want to, due to severa=

l

Quoted 3 lines

improvements), I have no usable driver for my wireless anymore.

What I need is a solution, meaning either a working iwl3945 version o=

r a

Quoted 6 lines

patch to ipw3945 to compile with 2.6.24.

Thanks in advance for any help or comments.

---------------------------------------------------------------------=

----

Quoted 4 lines

Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/mark=

etplace

Quoted 5 lines

_______________________________________________
Ipw3945-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ipw3945-devel

-
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
In reply to Thomas Bächler
Quoted 1 line · Thomas B=E4chler schrieb:

What I need is a solution, meaning either a working iwl3945 version o=

r a

Quoted 1 line

patch to ipw3945 to compile with 2.6.24.

I was hoping for someone from Intel or someone with knowledge of the code to comment on both these issues:

- Can ipw3945 be modified to compile with 2.6.24?
- Can the newest iwl3945 from the website be modified to compile with 2.6.24?

-
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw
Quoted 8 lines · On , Thomas B=E4chler wrote:
Quoted 2 lines · Thomas B=E4chler schrieb:

What I need is a solution, meaning either a working iwl3945 version or a patch to ipw3945 to compile with 2.6.24.

=20
I was hoping for someone from Intel or someone with knowledge of the code to comment on both these issues:
=20
- Can ipw3945 be modified to compile with 2.6.24?

ipw3945 is deprecated - we focus our effort on iwlwifi at this time.

Quoted 2 lines

- Can the newest iwl3945 from the website be modified to compile with 2.6.24?

iwlwifi is included in 2.6.24 so you can use it from there. If you are = interested in the latest iwlwifi development (more recent than what is = in 2.6.24) then you can work from John Linville's wireless-2.6 (#everyt= hing) repository.

Reinette
-
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#4 lore raw
Quoted 11 lines · Reinette
Quoted 7 lines · Thomas Bächler
Quoted 2 lines · Thomas Bächler

What I need is a solution, meaning either a working iwl3945 version or a patch to ipw3945 to compile with 2.6.24.

I was hoping for someone from Intel or someone with knowledge of the code to comment on both these issues:

- Can ipw3945 be modified to compile with 2.6.24?

ipw3945 is deprecated - we focus our effort on iwlwifi at this time.

Which doesn't help anyone if it does not work. ipw3945 may be deprecated in your eyes, but for me it is the only way to make my wireless NIC work reliably. From my point of view ipw3945 is stable and iwl3945 is _alpha_ software.

Quoted 4 lines
Quoted 2 lines

- Can the newest iwl3945 from the website be modified to compile with 2.6.24?

iwlwifi is included in 2.6.24 so you can use it from there. If you are interested in the latest iwlwifi development (more recent than what is in 2.6.24) then you can work from John Linville's wireless-2.6 (#everything) repository.

It is, but if you read my E-Mail, you will notice that the driver included in 2.6.24 is unusable, thus my request to use a newer driver with a _stable_ kernel. The wireless-dev is based on the latest linux-2.6 git, which is not always stable.

#5 lore raw
Quoted 8 lines · On Friday, January 18, 2008 12:42 PM, Thomas B=E4chler wrote:
Quoted 1 line

ipw3945 is deprecated - we focus our effort on iwlwifi at this time.

=20
Which doesn't help anyone if it does not work. ipw3945 may be deprecated in your eyes, but for me it is the only way to make my wireless NIC work
reliably. From my point of view ipw3945 is stable and iwl3945 is _alpha_
software.

Please understand that we need to focus limited resources. ipw3945 is d= eprecated and we focus on iwlwifi, which supports both 3945 and 4965. U= nfortunately we have been unable to reproduce your problem.=20

Quoted 4 lines

It is, but if you read my E-Mail, you will notice that the driver included in 2.6.24 is unusable, thus my request to use a newer driver with a _stable_ kernel. The wireless-dev is based on the latest linux-2.6 git, which is not always stable.

=46or this purpose I suggest that you consider the compat-wireless proj= ect. See http://www.kernel.org/pub/scm/linux/kernel/git/mcgrof/compat-w=
ireless-2.6.git and http://linuxwireless.org/download/compat-wireless-2=
=2E6/.

Reinette

-
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#6 lore raw

If the ipw3945 is no longer going to be maintained. You guys should consider removing it from the mainline kernel. This is in favor of iwl3945 (which is in the kernel also).

On Jan 18, 2008 3:09 PM, Chatre, Reinette <[email protected]> w= rote:

Quoted 3 lines · On Friday, January 18, 2008 12:42 PM, Thomas B=E4chler wrote:
Quoted 2 lines
Quoted 2 lines

ipw3945 is deprecated - we focus our effort on iwlwifi at this tim=

e.

Quoted 9 lines
Quoted 7 lines

Which doesn't help anyone if it does not work. ipw3945 may be deprecated in your eyes, but for me it is the only way to make my wireless NIC work
reliably. From my point of view ipw3945 is stable and iwl3945 is _alpha_
software.

Please understand that we need to focus limited resources. ipw3945 is=

deprecated and we focus on iwlwifi, which supports both 3945 and 4965.= Unfortunately we have been unable to reproduce your problem.

Quoted 3 lines
Quoted 3 lines

It is, but if you read my E-Mail, you will notice that the driver included in 2.6.24 is unusable, thus my request to use a newer driv=

er

Quoted 4 lines
Quoted 2 lines

with a _stable_ kernel. The wireless-dev is based on the latest linux-2.6 git, which is not always stable.

For this purpose I suggest that you consider the compat-wireless proj=

ect. See http://www.kernel.org/pub/scm/linux/kernel/git/mcgrof/compat-w=
ireless-2.6.git and http://linuxwireless.org/download/compat-wireless-2=
=2E6/.

Quoted 6 lines

Reinette

---------------------------------------------------------------------=

----

Quoted 8 lines

This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Ipw3945-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/ipw3945-devel

-
To unsubscribe from this list: send the line "unsubscribe linux-wireles= s" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#7 lore raw
Quoted 4 lines · Jerone Young

If the ipw3945 is no longer going to be maintained. You guys should consider removing it from the mainline kernel. This is in favor of iwl3945 (which is in the kernel also).

ipw3945 was never in the mainline kernel.
Might be your distro shipping it as an addon.

#8 lore raw
In reply to Chatre, Reinette
Quoted 8 lines · Reinette
Quoted 6 lines

Which doesn't help anyone if it does not work. ipw3945 may be deprecated in your eyes, but for me it is the only way to make my wireless NIC work
reliably. From my point of view ipw3945 is stable and iwl3945 is _alpha_
software.

Please understand that we need to focus limited resources. ipw3945 is deprecated and we focus on iwlwifi, which supports both 3945 and 4965. Unfortunately we have been unable to reproduce your problem.

Your unability to reproduce the bug is weird, as I can reproduce this bug every time I try iwl3945, with every version I ever tried and with every kernel I ever tried. Are there different revisions of the 3945 chip which behave differently and maybe you are trying another one?

[x86_64][11:24:49][thomas@artin ~]$ (lspci; lspci -n)|grep ^03\:00 03:00.0 Network controller: Intel Corporation PRO/Wireless 3945ABG Network Connection (rev 02)
03:00.0 0280: 8086:4227 (rev 02)

Jerone suggested that it may be something in my software (although a Debian user posted in the bug report that he is also affected, and his "works for me" comment is based on iwl4965, not 3945). Which components of my system are inolved here?

- wpa_supplicant: tried versions 0.5.7 to 0.5.9, all the same. Associating always works (only data transmission after that fails), so this seems unlikely.
- openssl (which wpa_supplicant depends on): unlikely, as it is only involved in key negotiation - once the connection has been established, it does not do anything related to the actual networking.
- glibc: I think I have had more than 5 different versions of glibc installed since I first saw this bug, so a problem here seems unlikely.
- gcc: It is used to compile linux and the iwlwifi driver, I am currently using version 4.2.2, I had some 4.1 releases before, all of them with the same results. May compiling with a different version of gcc make the bug go away? Which version are you using in your tests?
- linux: tried versions from 2.6.21 (with the Intel mac80211 patch) to 2.6.24, vanilla versions as well as the Archlinux kernel (which includes some patches), all with the same results.
- the udev firmware loader: if the firmware is not loaded properly, it may lead to bugs. However, it is simply a script that gives the kernel the firmware data into sysfs, so if there is a problem here, it is in linux.
- The firmware files: I have these firmware files installed in /lib/firmware: (with md5sums)
26d10014b09439dc5a8573c2a6f85b0a /lib/firmware/iwlwifi-3945-1.ucode 3afbb112ade3ede3d02e03892ff8a201 /lib/firmware/iwlwifi-3945.ucode I guess iwlwifi uses the first one now.

So what other problems could there be on my end? When you test your driver, do you even try if actual data is transmitted after the association process completed?

I may have time to test this with the Ubuntu Live CD soon (as he suggested), but I don't think it will make a difference.

I am eager to solve this problem, but I don't know where to look anymore.