Re: auich fifo underrun

[buhrow%lothlorien.nfbcal.org@localhost (Brian Buhrow)]

        Hello.  This sounds like a problem I had with a Dell D400 laptop where
the ath(4) card and the auich(4) devices shared an interrupt.  In my case,
the problem was that the acpi code tried to redirect the interrupt the BIOS
assigned to these devices to another interrupt.  The symptom was that if I
tried to use both devices at once, the interrupt controller would
apparently freak out and stop passing interrupts to everything on that
particular interrupt.  I could get things going again by stopping the use
of the audio device, and ifconfigging the ath(4) device down and up again.
        The solution, which works well, but is somewhat controversial, is
documented in kern/37538.
        I don't know that this is the same problem, but it sounds identical,
and the symptoms, as you  have described them, seem the same.  You might
try the patches in that bug report, and see if the problems persist.
Alternatively, to see if it's the same problem, try booting a kernel
without ACPI and see how things behave.  If they work well, modulo the loss
of functionality you suffer without ACPI, it's probably a similar issue to
the bug, and you should try the patches in that bug.  If they fix the
problem, then we can decide how to best address the issue, which is
apparently larger than just on one older Dell laptop.
-Brian
On Feb 12,  8:52pm, Dieter Baron wrote:
} Subject: Re: auich fifo underrun
} In article <20080212181824.GA26235%britannica.bec.de@localhost> Joerg wrote:
} : On Tue, Feb 12, 2008 at 12:10:25PM -0600, David Young wrote:
} : > On Tue, Feb 12, 2008 at 05:44:04PM +0100, Dieter Baron wrote:
} : > >   Do iwi0 and auich0 share an interrupt?  (I experience interrupt
} : > > sharing problems between auich0 and wm0 (both at interrupt 11).)
} : > 
} : > Audio devices are a bit unusual in that they auto-suspend.  Looks to me
} : > like auich does not detach its interrupt handler when it is suspended.
} : > It should.
} 
} : auich skips interrupt processing if the device is currently inactive.
} : Given that it is not supposed to get any interrupts at that time, that
} : should be good enough. It might need to clear pending interrupts though.
} 
}   I don't think that is the problem.  I hear sound for about 30
} seconds, then sound and all network traffic stops.  A ifconfig down/up
} restores networking.  If I play sound from an USB disk (uhci also uses
} interrupt 11), access to the USB disk hangs and umount fails (even on
} shutdown).  As an additional data point, I never had that problem with
} iwi instead of wm.
} 
}                                               yours,
}                                               dillo
>-- End of excerpt from Dieter Baron