BCM4312 LP-PHY DMA issues

[Justin Azoff <JAzoff@albany.edu>]

On Tue, Nov 22, 2011 at 05:13:13PM +0100, Dominique Martinet wrote:
> Hi,
> 
> Is the dmesg output the one without pio=1?
> It says:
> [   28.236344] b43-phy0: Loading firmware version 478.104 (2008-07-01
> 00:50:23)
> [   33.772580] b43-phy0 ERROR: Fatal DMA error: 0x00000400, 0x00000000,
> 0x00000000, 0x00000000, 0x00000000, 0x00000000
> [   33.778855] b43-phy0 ERROR: This device does not support DMA on your
> system. It will now be switched to PIO.       
> [   33.784685] b43-phy0: Controller RESET (DMA error) ...
>                                       
> so I would suppose it fell back to pio=1 and missed something during the
> initialization (assuming it works when you modprobe it with pio=1
> directly)

That is correct:

 * with pio=1 it works
 * without pio=0 you get the above message, but then it is never able to connect.

> That being said, it is supposed to work on linux >= 3.0, so I'd try with
> a more recent firmware just to see if it helps.
> (cf. http://linuxwireless.org/en/users/Drivers/b43?highlight=%28b43%29#List_of_firmware
> ) 

Ah, I didn't even think to upgrade the firmware.. the debian package
that installs the firmware grabs this url:
http://downloads.openwrt.org/sources/broadcom-wl-4.178.10.4.tar.bz2

which appears to be a few versions behind. I'm installed
broadcom-wl-5.10.56.27.3_mipsel.tar.bz2 and am testing kernels now.

3.1.0-1-686-pae with the new firmware seems to be working :-)

It went from

2011-11-22 12:24:23 check=DL ok=True download_time=26.44 timeout=False min_speed=544 avg_speed=774 max_speed=888

to

2011-11-22 12:42:16 check=DL ok=True download_time=11.75 timeout=False min_speed=1376 avg_speed=1743 max_speed=1976

and CPU usage is only 15% now.

> (it's not debugging the right part, and I won't be of much use since
> I've never had a b43-compatible device, but you can have more
> information about the deauth if you run "iw event" to get a few more
> messages, or maybe wpa_supplicant debug output, since it says "local
> choice" I would suppose wpa_supplicant isn't happy with something.)

If I had to guess I think it is just hosed and timing out somehow... not
important now :-)

-- 
-- Justin Azoff
-- Network Security & Performance Analyst