i have a laptop inspiron 5100 with broadcom b44 ethernet and bcm4301 wireless lan cards onboard. using kernel 2.6.23 everything is ok, on kernel 2.6.24 and higher (new 2.0 driver with new ssb) the system hangs on boot in udevsettle. I booted into bb and found out that the a modprobe b44 hangs the system. Following error is in dmesg: b44: Invalid MAC address found in EEPROM b44 ssb0:1: Problem fetching invariants of chip, aborting. b44: probe of ssb0:1 failed with error -22 In 3 of 4 times the (cold) boot process locks hard, but sometimes i get a working system, but the error is still in the logs. I googled, some say ubuntu ibex is fixed, but i could not find out how. Will report on that later. Reproducible: Always Steps to Reproduce: 1. turn on the computer Actual Results: system boots and hangs while "waiting for udev...", fan gets loud .-( Expected Results: boot into nice kde startup screen. gentoo-sources-2.6.23 works (old version 1.1 of drivers) gentoo-sources-2.6.24 and later does not work (version 2.0 of driver) I tried some vanilla sources and gentoo sources up to gentoo-sources 2.6.27-r5 no problem with udev, hal, etc (I ermerged latest stable ones)
So, does this mean the problem disappears when upgrading to gentoo-sources 2.6.27-r5 ? Which is quite possible because of http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.27.y.git;a=commitdiff;h=f225763a7d6c92c4932dbd528437997078496fcc
I confirm that vanilla-sources-2.6.27.9 dos not show crashes but I get another problem: The new driver sais I should toglle my rfkill button in order to get a working interface. BUT I HAVE NO RFKILL BUTTON...
(In reply to comment #2) > I confirm that vanilla-sources-2.6.27.9 dos not show crashes but I get another > problem: The new driver sais I should toglle my rfkill button in order to > get a working interface. > BUT I HAVE NO RFKILL BUTTON... Can you please attach dmesg output from kernel 2.6.27.9? Does yourr ethernet interface work? Is it possible to bring the wireless interface up by using software, i. e. iwconfig?
I hacked b43legacy to ignore the hardware button reading and always return on in rfkill.c. After that, I was able to scan my network and assosicate with my ap. But whatever I did, the interface did not receive any packets, neither with dhclient or with manual ip setup. And yes, b44 works correctly dmesg output follows /ulf
Created attachment 175886 [details] dmesg output of session this is a hacked b43legacy to ignore the hardware rfkill hardware button setting!! can load and unload module without hangs and crashes multiple times can turn on card can scan network can associate with ap using wep encryption can not get dhcp lease can set up ip manual using ifconfig try ping i see tx packets but no rx packets - destination unreachable checked ip adress,broadcast etc. multiple times but they are correct!!
Anyway, the original problem (hang on boot) seems to be gone with latest 2.6.27 kernel. So I'll close this bug. Please open a new/separate bug for the other problem, e. g. b43legacy driver assumes there is a hardware radio switch although there is none. For the strategy will be to push this bug upstream (kernel.org), please use vanilla sources 2.6.27.10 to produce the problem and (for now) do not hack the driver. :-) Attach your kernel config, dmesg and lspci output to the bug. Also state on the (new) bug which has been the latest kernel version that worked for you. Finally you should check, if the problem persist with latest 2.6.28-rc sources, because you'll be asked to do so, when taking the bug upstream. :-) Thanks
closing this bug is ok, but the b43legacy driver still does not work for me As I started hacking and its holiday time, i subscribed to b43xx-dev and will try to fix this with the developers directly. I will send an update here if it works again. thanks Ulf
For anyone else who finds the rfkill bug interesting and wants to follow the updates on the bcm43xx-dev mailing list, I'm giving you the links for the appropriate threads: https://lists.berlios.de/pipermail/bcm43xx-dev/2008-December/008424.html https://lists.berlios.de/pipermail/bcm43xx-dev/2008-December/008425.html
(In reply to comment #7) > closing this bug is ok, but the b43legacy driver still does not work for me > > As I started hacking and its holiday time, > i subscribed to b43xx-dev and will try to fix this with the developers > directly. > I will send an update here if it works again. Thanks. Though feel free to file the bug right away. After all it's obviously a kernel regression.