Broadcom wireless: ifconfig eth1 up result in hard-lockup

Bug #44911 reported by Andy Brook
16
Affects Status Importance Assigned to Milestone
linux-source-2.6.15 (Ubuntu)
Invalid
Medium
Unassigned

Bug Description

Binary package hint: bcm43xx-fwcutter

Im using a HP pavillion zv5383EA, AMD64 laptop. It has a broadcom wireless chipset. It worked under Breezy (after a fasion) using ndiswrapper and wpa_supplicant. In Dapper, it has never worked. There seems to be support for the bcm43xx, and involved in that is the downloading of microcode via the bcm43xx-fwcutter script.

I can use iwconfig to list wireless device at eth1, however, executing 'ifconfig eth1 up' causes a total hardware lockup - every time. If the microcode isnt applicable to my card, surely it shouldnt have been 'loaded'? and if it is the right code, why the hard lockup?

Ive been hoping this would go away with the 150+ daily updates but it hasnt yet, including a kernel update (in windows - wireless! so cant dig out kernel image #).

If the fwcutter is working for other people, maybe I need to get ndiswrapper out again and try my luck, but I cant help think it should detect incompatible cards and warn somehow.

How does ifconfig cause a hardware lock anyway!

Revision history for this message
Gaëtan Petit (gaetanp) wrote :

i can't reproduce this under 686 kernel
maybe a 64bit problem
or a driver problem

Changed in bcm43xx-fwcutter:
status: Unconfirmed → Needs Info
Revision history for this message
Andy Brook (javahollic) wrote :

#44174 was my last report (prior to fw-cutter).

If the microcode is good, I wonder if theres something in the firmware loader which breaks under 64bit?

I'll try ndsiwrapper again - until as/when I hear about any firmware /microcode updates as Im not able to progress this much on my own.

Revision history for this message
Gaëtan Petit (gaetanp) wrote :

did you use the provided script :
/usr/share/bcm43xx-fwcutter/install_bcm43xx_firmware.sh

please try and post back here

Revision history for this message
Andy Brook (javahollic) wrote :

Ive just dist-upgraded to catch all updates as of May17:

I ran the script again, it did its stuff and added files to /lib/firmware as before. If the firmware is in place (using the script) and I do ifconfig eth1 up, I get an immediate hardware lockup. If the firmware isn't in place I get 'SIOCSIFFLAGS: No such file or directory'.

Using ndiswrapper I can load up the 64bit netbc564 driver but still cannot get eth1 up, it results in the same error as above. I wonder if this boils down to an incompatibility with bcm43xx firmware and the hardware I have '564'? different family or is it basically the same? I checked iwconfig, and it lists the device as a "Broadcom 4306" without ndiswrapper drivers registered or firmware loaded into /lib/firmware - so perhaps it is natively compatibile - so why this problem?

Revision history for this message
Gaëtan Petit (gaetanp) wrote :

well i don't really know
if it is the hardware '564' or the 64bit the source of this bug

Moreover bcm43xx seems to be buggy.
It worked yesterday but today, the network applet tell me "Deconnected" even if i de-activate then activate my eth1 interface ...

Revision history for this message
BHowell (mute-howell-ersatz) wrote :

I think this may be due to the 1GB ram bug. This driver causes a kernel panic on my machine (amd64 with 1.5 GB ram). There's talk that this 1GB+ addressing bug has been found and fixed upstream.

For the record I get the following panic:

Kernel panic - not syncing: PCI-DMA high address but no IOMMU

Revision history for this message
Andy Brook (javahollic) wrote :

Interesting, Ive got 2GB onboard, I'll pull one out to verify.

Revision history for this message
BHowell (mute-howell-ersatz) wrote :

I can confirm this. I just pulled out 1GB and the module loads and runs fine. The bcm43xx module needs to be rebuilt with a kernel with (I think) the following options:

CONFIG_HIGHMEM4G=y
CONFIG_HIGHMEM=y

I have a gateway MX7118 AMD64 notebook running todays amd64 dapper with all updates. I don't have an intel box to test on.

Revision history for this message
Cory Prowse (cory-prowse) wrote :

Just confirming that I am getting the following kernel panic:
  Kernel panic - not syncing: PCI-DMA high address but no IOMMU

This occurs shortly after I cut the firmware into the /lib/firmware directory, and also during bootup. It's a hard crash needing a hardware reset (and then booting into a LiveCD to remove the firmware files from /lib/firmware).

I have a Ferrari 4005 laptop, running AMD64 kernel with 2Gig of ram and the Broadcom 4318 card.

Revision history for this message
Elliott Minor (zengrits) wrote :

I also use a Broadcom wireless card. It worked fine with Breezy, but doesn't work with a clean install of Dapper. I have an Athlon 32 processor.

When I reinstalled Breezy, the Belkin F5D7010 worked again.

Then I used the update feature to install Dapper a second time and the card stopped working the minute the new operating system took over.

The card doesn't seem to power up under Dapper. I can't activate it.

Revision history for this message
Ben Collins (ben-collins) wrote : Re: [Bug 44911] Re: Broadcom wireless: ifconfig eth1 up result in hard-lockup

> Then I used the update feature to install Dapper a second time and the
> card stopped working the minute the new operating system took over.

When you say "the minute the new operating system took over", does that
mean before or after you rebooted into the new kernel?

Revision history for this message
Elliott Minor (zengrits) wrote : RE: [Bug 44911] Re: [Bug 44911] Re: Broadcom wireless: ifconfig eth1 upresult in

Hello Ben,

Thanks for your response. I'm a little hazy on exactly when the card stopped
working.

I used the WiFi card to download all the software that transformed the
system from Breezy to Drake. I believe the card stopped working just before
the reboot. It definitely wasn't working after the reboot.

I have another computer I want to upgrade to 6.06. I'll watch that one
closely and give you a more precise report. Hopefully I can get to that this
evening.

Fortunately, I got my Belkin card working last night using some outstanding
information from the Ubuntu Forums. That particular article was titled
(howto) General 6.06 - How to: Broadcom Wireless cards.

This method doesn't rely on ndiswrapper. It allows users to install firmware
for cards that are already recognized by the system, which mine was.

Thanks for your interest. I was about to reinstall Breezy and throw up my
hands. I'd like to help others who might be having this problem, so I'll
definitely get back to you.

Is there anything else you want me to watch for when I update the other
computer?

Best,

Elliott L. Minor

Revision history for this message
Ben Collins (ben-collins) wrote :

I should have looked closer. Your issue is that you had ndiswrapper setup to operate this card in breezy, when in dapper we have a stock driver for it (although it does need firmware that we cannot redistribute).

I'll consider this bug closed, as it's not really a bug, just an upgrade note.

Changed in linux-source-2.6.15:
status: Needs Info → Rejected
Revision history for this message
BHowell (mute-howell-ersatz) wrote :

Can you please reopen this bug? I can confirm it exists. The previous commenter seems to be talking about other issues not related to the original bug report.

To re-summarize the actual bug:

bcm43xx module causes a panic on amd64 (and possibly other) if the system has more than 1GB RAM.

If you want a panic dump I can send it to you, but I believe this was documented on the bcm43xx.berlios.de site. The module needs to be rebuilt with a kernel build configured for more than 1GB ram.

Thanks!

Revision history for this message
Ben Collins (ben-collins) wrote : Re: [Bug 44911] Re: Broadcom wireless: ifconfig eth1 up result in hard-lockup

On Wed, 2006-06-07 at 18:48 +0000, BHowell wrote:
> Can you please reopen this bug? I can confirm it exists. The previous
> commenter seems to be talking about other issues not related to the
> original bug report.
>
> To re-summarize the actual bug:
>
> bcm43xx module causes a panic on amd64 (and possibly other) if the
> system has more than 1GB RAM.
>
> If you want a panic dump I can send it to you, but I believe this was
> documented on the bcm43xx.berlios.de site. The module needs to be
> rebuilt with a kernel build configured for more than 1GB ram.

There's already another bug report for that. So just dupe this one to
it, or ignore this one and move on to the existing bug report.

Revision history for this message
Cory Prowse (cory-prowse) wrote :

Ben, I've just tried to find the other bug report which will solve this issue, but could not find it when searching.

Do you have the bug number available?

Revision history for this message
Cory Prowse (cory-prowse) wrote :

After a bit of searching I foudn this bug that seems related, however it is focussed more on the PPC kernel than the AMD64-k8 kernel.

38912 : https://launchpad.net/distros/ubuntu/+bug/38912

Revision history for this message
Elliott Minor (zengrits) wrote : RE: [Bug 44911] Re: [Bug 44911] Re: Broadcom wireless: ifconfig eth1 upresult in

Hello Ben,

I appreciate the interest you showed in my wireless problem. I found a
solution on the Umbuntu forum and learned a lot while solving it.

Specifically, I learned that Drake has the necessary drivers and recognizes
my Belkin 54G card (Broadcom), but lacks for legal reasons the firmware to
make it work. Using the excellent instructions from the forum, I was able to
install the firmware and get it working.

Elliott L. Minor

Revision history for this message
Andy Brook (javahollic) wrote :

Finally got the back off my laptop and proved wireless works with 1GB of ram (ok, it can see base stations, but configuring wpa-supplicant for WPA-TKIP is the next challenge!). Have moved firmware files out of the way to stop lockup at bootup ( /lib/firmware/bcm* ).

Looking forward to an updated module at some point.

andy

Revision history for this message
Ben Collins (ben-collins) wrote :

The 1G problem has been fixed in a recent update of dapper. 2.6.15-25.

Revision history for this message
Andy Brook (javahollic) wrote :

Great, verified, 15-25 does indeed fix the hang!

Revision history for this message
Andy Brook (javahollic) wrote :

However after ifup eth1, starting wpa_supplicant now causes a hard lockup. Ill raise a separate bug for that.

Revision history for this message
enrico (bazza-conecta) wrote :

i'm Enrico from Italy
and first of all sorry for my bad english
i have a Laptop HP Pavilion DV8306ea with wifi of Broadcom
i had installed Linux Kubuntu 6.06 with 2.6.15 kernel
To install wifi i had done all
but ndiswrapper -l tell me:
Installed ndis drivers:
bcmwl5 driver present

it doesn't tell me "Hardware present"
i think my device is not switched on and even after if i load the
module on kernel is all ok
but if i go in /var/log/syslog i find he download the driver but not
(as wroten in ndiswrapper installation doc):
something like wlan: ndiswrapper ethernet device xx:xx:xx:........
but (regard my eth1 (wifi device)) i read a repetition of a line like
this: localhost dhclient:DHCPDISCOVER on eth1 to 255.255.255.2555 port
67 interval 5...7...9...4...15 (and other numbers in the consequences
line)
what happend to my wifi? where is the problem?
can someone help me to solve my problem?
i have 1 gb ram and kernel 2.6.16-26-386
thx
enrico

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.