5.2-RELEASE panic in turnstile_wait

[Erik Moe]

How long has this been an issue?  Was this a problem in 5.1-RELEASE as well,
because I had a few mysterious panics in 5.1 where it look like there was
memory corruption.

Erik

> -----Original Message-----
> From: Don Lewis [mailto:truckman at FreeBSD.org]
> Sent: Thursday, January 29, 2004 1:15 AM
> To: emoe at cox.net
> Cc: jhb at FreeBSD.org; freebsd-current at FreeBSD.org
> Subject: Re: 5.2-RELEASE panic in turnstile_wait
>
>
> On 29 Jan, Erik Moe wrote:
> > My loader.conf looks like this...
> >
> > hw.pci.allow_unsupported_io_range="1"
> > hw.ata.atapi_dma="1"
> >
> > hw.snd.pcm0.vchans=4
> > hw.snd.maxautovchans=4
>
> There is a potential buffer overflow in the vchan code that can bzero()
> unrelated objects in the kernel heap, like the mutex that you found, and
> cause hard to track down system panics.  The buffer overflow can be
> triggered by certain combinations of sound hardware and software.
>
> The 5.2-RELEASE errata list contains the following note:
>
> (9 Jan 2004) The use of multiple vchans (virtual audio channels with
> dynamic mixing in software) in the pcm(4) driver has been known to cause
> some instability.
>
>
> I finally came up with a fix for this problem that worked and checked it
> into the -CURRENT source within the last 24 hours.  I'd recommend either
> disabling vchans or upgrading to -CURRENT, though you might be able to
> retrofit the contents of src/sys/dev/sound/pcm/ and rebuild your kernel.
>
>