[arch-general] b43: wireless issues
I'm having problems getting my wireless connection to work properly. It's a broadcom card, using the b43 driver. It connects fine on login, but once I actually use the network I'm disconnected and all attempts to re-connect fail. I've found some reports of similar behaviour online, but none seem to offer any good solutions to it :-( So I'm trying my luck in here instead. One person mentioned that things started working much better if he limited the network to 802.11b (not g), but I haven't seen any obvious way of achieving this. Any hints on what I can do to test that myself? $ lspci -vnn |grep 14e4 03:00.0 Network controller [0280]: Broadcom Corporation BCM4312 802.11b/g [14e4:4315] (rev 01) I did extract some stuff from dmesg that might be of importance, but I really couldn't tell: b43 ssb0:0: firmware: requesting b43/ucode15.fw b43 ssb0:0: firmware: requesting b43/lp0initvals15.fw b43 ssb0:0: firmware: requesting b43/lp0bsinitvals15.fw b43-phy0: Loading firmware version 478.104 (2008-07-01 00:50:23) b43-phy0 ERROR: Fatal DMA error: 0x00000800, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000 b43-phy0 ERROR: This device does not support DMA on your system. Please use PIO instead. b43-phy0: Controller RESET (DMA error) ... b43-phy0: Loading firmware version 478.104 (2008-07-01 00:50:23) b43-phy0: Controller restarted Cheers, M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe
Hi Magnus, Excerpts from Magnus Therning's message of Fri, 18 Jun 2010 00:22 +0100:
It's a broadcom card, using the b43 driver. It connects fine on login, but once I actually use the network I'm disconnected and all attempts to re-connect fail.
.... One person mentioned that things started working much better if he limited the network to 802.11b (not g), but I haven't seen any obvious way of achieving this. Any hints on what I can do to test that myself?
I used to use b43, but now I don't. I had similar situations with other wifi cards. 1. Two or more access points in the vicinity are using the same channel. If you're able to change the channel, chose one not used by others. The reasons are more or less obvious. 2. If other computers are using the same access point, and my connection is getting dropped I indeed intentionally lower the bitrate: iwconfig wlan0 bitrate 24M (or lower) - it works, but the reason is unknown, I haven't investigated it. You may want to install wavemon program to monitor your connection [0], there is a package in AUR too [1]. Cheers, Sergey [0] http://eden-feed.erg.abdn.ac.uk/wavemon/ [1] http://aur.archlinux.org/packages/wavemon/wavemon/PKGBUILD
Excerpts from Magnus Therning's message of 2010-06-18 01:22:48 +0200:
I'm having problems getting my wireless connection to work properly.
It's a broadcom card, using the b43 driver. It connects fine on login, but once I actually use the network I'm disconnected and all attempts to re-connect fail.
I've found some reports of similar behaviour online, but none seem to offer any good solutions to it :-( So I'm trying my luck in here instead.
One person mentioned that things started working much better if he limited the network to 802.11b (not g), but I haven't seen any obvious way of achieving this. Any hints on what I can do to test that myself?
$ lspci -vnn |grep 14e4 03:00.0 Network controller [0280]: Broadcom Corporation BCM4312 802.11b/g [14e4:4315] (rev 01)
I did extract some stuff from dmesg that might be of importance, but I really couldn't tell:
b43 ssb0:0: firmware: requesting b43/ucode15.fw b43 ssb0:0: firmware: requesting b43/lp0initvals15.fw b43 ssb0:0: firmware: requesting b43/lp0bsinitvals15.fw b43-phy0: Loading firmware version 478.104 (2008-07-01 00:50:23)
b43-phy0 ERROR: Fatal DMA error: 0x00000800, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000 b43-phy0 ERROR: This device does not support DMA on your system. Please use PIO instead. b43-phy0: Controller RESET (DMA error) ... b43-phy0: Loading firmware version 478.104 (2008-07-01 00:50:23) b43-phy0: Controller restarted
Cheers, M
Hi. I have the same chip and same issue. I found out that the b43 devs didn't figure out yet how to get the chip to work correctly, so unless something changed during the last weeks/months there's no real solution. Using the apparently slower PIO instead of DMA is a kernel compiletime option so far, but I read that as a workaround it's a module load option beginning with .34. It's a pity, the chip should have worked beginning with I afair .30. I wasted loads of time trying to get it to work, without luck. The broadcom driver works afair, but it annoyed me for some reasons. So for now I'm looking forward to .34 to hopefully have at least working wifi, even if slow. -- Regards, Philipp -- "Wir stehen selbst enttäuscht und sehn betroffen / Den Vorhang zu und alle Fragen offen." Bertolt Brecht, Der gute Mensch von Sezuan
On Fri, Jun 18, 2010 at 08:55, Philipp Überbacher <hollunder@lavabit.com> wrote:
Hi. I have the same chip and same issue. I found out that the b43 devs didn't figure out yet how to get the chip to work correctly, so unless something changed during the last weeks/months there's no real solution. Using the apparently slower PIO instead of DMA is a kernel compiletime option so far, but I read that as a workaround it's a module load option beginning with .34. It's a pity, the chip should have worked beginning with I afair .30. I wasted loads of time trying to get it to work, without luck. The broadcom driver works afair, but it annoyed me for some reasons. So for now I'm looking forward to .34 to hopefully have at least working wifi, even if slow.
I did find that there is a PIO-related option that I can put in modprobe.conf, but that doesn't seem to be address the problem I have (I also read that the b43 driver revert back to PIO automatically if there are DMA problems). Next step I'll try one of the daily snapshots via the compat-wireless package. If that doesn't work I'll be very irritated indeed, enough so to volunteer my services to the b43 devs for any testing they need to have done. /M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe
Excerpts from Magnus Therning's message of 2010-06-18 16:49:11 +0200:
On Fri, Jun 18, 2010 at 08:55, Philipp Überbacher <hollunder@lavabit.com> wrote:
Hi. I have the same chip and same issue. I found out that the b43 devs didn't figure out yet how to get the chip to work correctly, so unless something changed during the last weeks/months there's no real solution. Using the apparently slower PIO instead of DMA is a kernel compiletime option so far, but I read that as a workaround it's a module load option beginning with .34. It's a pity, the chip should have worked beginning with I afair .30. I wasted loads of time trying to get it to work, without luck. The broadcom driver works afair, but it annoyed me for some reasons. So for now I'm looking forward to .34 to hopefully have at least working wifi, even if slow.
I did find that there is a PIO-related option that I can put in modprobe.conf, but that doesn't seem to be address the problem I have (I also read that the b43 driver revert back to PIO automatically if there are DMA problems). Next step I'll try one of the daily snapshots via the compat-wireless package. If that doesn't work I'll be very irritated indeed, enough so to volunteer my services to the b43 devs for any testing they need to have done.
/M
I only read that stuff on their mailinglist and in some kernel log/mail that they're gonna change it to a module load time option in kernel .34. I don't know anything for sure. I don't think it reverts to PIO automatically, and I don't think it will. If you don't necessarily need the b43 driver then the broadcom thing in AUR should work without too much hassle. One of the annoying things with it is that you need to compile it against your kernel, so if you use multiple kernels you're in for fun. Another nuisance is that the wifi identifies itself as eth0 and wired as eth1. b43 is imho better in that respect, eth0 for ethernet and wlan0 for wireless just makes more sense. I also had the broadcom driver sometimes stall my machine for a couple of seconds, usually after boot, which I consider bad behavior. No idea about compat-wireless. -- Regards, Philipp -- "Wir stehen selbst enttäuscht und sehn betroffen / Den Vorhang zu und alle Fragen offen." Bertolt Brecht, Der gute Mensch von Sezuan
On 18/06/10 16:09, Philipp Überbacher wrote: [...]
I only read that stuff on their mailinglist and in some kernel log/mail that they're gonna change it to a module load time option in kernel .34. I don't know anything for sure. I don't think it reverts to PIO automatically, and I don't think it will.
I think it does. Since I do get connected long enough to start using the network, even when the DMA-related error message pops up in dmesg.
If you don't necessarily need the b43 driver then the broadcom thing in AUR should work without too much hassle. One of the annoying things with it is that you need to compile it against your kernel, so if you use multiple kernels you're in for fun. Another nuisance is that the wifi identifies itself as eth0 and wired as eth1. b43 is imho better in that respect, eth0 for ethernet and wlan0 for wireless just makes more sense. I also had the broadcom driver sometimes stall my machine for a couple of seconds, usually after boot, which I consider bad behavior.
Yes, I just installed broadcom-wl from AUR, working fine so far. I'll stick to it and move on to the next item of configuration on my laptop: suspend. So far I always get a blank screen on resume, something I know has been discussed on this list in the passed; I'm off to the archives. Cheers, M -- Magnus Therning (OpenPGP: 0xAB4DFBA4) magnus＠therning．org Jabber: magnus＠therning．org http://therning.org/magnus identi.ca|twitter: magthe
Excerpts from Magnus Therning's message of 2010-06-20 00:35:36 +0200:
On 18/06/10 16:09, Philipp Überbacher wrote: [...]
I only read that stuff on their mailinglist and in some kernel log/mail that they're gonna change it to a module load time option in kernel .34. I don't know anything for sure. I don't think it reverts to PIO automatically, and I don't think it will.
I think it does. Since I do get connected long enough to start using the network, even when the DMA-related error message pops up in dmesg.
If you don't necessarily need the b43 driver then the broadcom thing in AUR should work without too much hassle. One of the annoying things with it is that you need to compile it against your kernel, so if you use multiple kernels you're in for fun. Another nuisance is that the wifi identifies itself as eth0 and wired as eth1. b43 is imho better in that respect, eth0 for ethernet and wlan0 for wireless just makes more sense. I also had the broadcom driver sometimes stall my machine for a couple of seconds, usually after boot, which I consider bad behavior.
Yes, I just installed broadcom-wl from AUR, working fine so far. I'll stick to it and move on to the next item of configuration on my laptop: suspend.
So far I always get a blank screen on resume, something I know has been discussed on this list in the passed; I'm off to the archives.
Cheers, M
Hey again. I just upgraded to kernel 2.6.34. I still get: b43-phy0 ERROR: Fatal DMA error: 0x00000400, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000 b43-phy0 ERROR: This device does not support DMA on your system. Please use PIO instead. b43-phy0: Controller RESET (DMA error) ... but now it does fall back to PIO automatically. I'm now online with the wireless for a couple of minutes and it seems to work well so far. I had one disconnect though, and don't know why. It happened at the instant I started IRC and the mail client. Anyway, I'd call it progress. Thought I let you know. -- Regards, Philipp -- "Wir stehen selbst enttäuscht und sehn betroffen / Den Vorhang zu und alle Fragen offen." Bertolt Brecht, Der gute Mensch von Sezuan
participants (3)
-
Magnus Therning
-
Philipp Überbacher
-
Sergey Manucharian