Re: b43 dma error

[Michael Buesch <mb@bu3sch.de>]

On Tuesday 08 September 2009 20:06:49 Gábor Stefanik wrote:
> On Tue, Sep 8, 2009 at 7:50 PM, John Daiker<daikerjohn@gmail.com> wrote:
> > On 09/08/2009 07:54 AM, Larry Finger wrote:
> >>
> >> Michael Buesch wrote:
> >>>
> >>> On Tuesday 08 September 2009 15:47:32 Dave Young wrote:
> >>>>
> >>>> I tested wireless-testing b43 driver, but got "Fatal DMA error"
> >>>> then the controller keep restarting...
> >>>>
> >>>> Please tell what I can provide or test, Thanks.
> >>>
> >>> Is this a regression? If so, please bisect.
> >>
> >> It is something specific to his system as I don't see anything like
> >> this. In addition, there are some users on the openSUSE forums that
> >> have implemented the latest compat-wireless and switched away from
> >> Broadcom wl to b43 on their LP PHY devices. So far, no complaints from
> >> them.
> >>
> >> Please reboot so that we see the ssb output as well. Use the command
> >>
> >> dmesg | egrep "ssb|b43"
> >>
> >> That way we will be able to see exactly what kind of device you have
> >> and what revisions are in it. AFAIK, the testing to date has been
> >> limited to Rev 1 PHYs and Rev 2 radios.
> >>
> >> Larry
> >>
> >
> > I can confirm the same issue.  I have a HP Mini 1116NR with a Broadcom 4312.
> >  Looks to be a PHY 1, Radio 2:
> >
> > --snip--
> > [  456.165296] b43-phy0 debug: Found PHY: Analog 6, Type 5, Revision 1
> > [  456.165364] b43-phy0 debug: Found Radio: Manuf 0x17F, Version 0x2062,
> > Revision 2
> > --snip--
> >
> > I've attached the output of 'dmesg | egrep "ssb|b43"', my kernel config, and
> > lspci -vv and lspci -nn
> >
> > Note: With the dmesg output, I had unloaded the b43 module previous
> > 'modprobe -r b43' and then loaded it again with debug output: 'modprobe b43
> > verbose=3'
> >
> > John Daiker
> >
> 
> Again, please test with v478 or v5xx firmware.
> 

This won't help. The DMA engine reports a descriptor error. That means the device
was unable to fetch the control data via DMA. Most of the time this indicates
a bug/failure/quirk in the platform DMA implementation.

-- 
Greetings, Michael.