V480R: panic: schizo_attach: could not register interrupt controller for CDMA (17)

[Marius Strobl]

On Sat, Dec 26, 2009 at 09:48:45PM +0000, Gavin Atkinson wrote:
> 
> Hi all,
> (marius@ cc'd as the panic was introduced with r185133)
> 
> I'm seeing this panic on a V480r I'm netbooting:
> 
> panic: schizo_attach: could not register interrupt controller for CDMA (17)
> 
> I've stuck the full verbose dmesg of the panic with SCHIZO_DEBUG set at 
> http://people.freebsd.org/~gavin/v480r/boot-panic
> 
> If I replace the call to panic() with a printf() (patch I'm using is at 
> http://people.freebsd.org/~gavin/v480r/schizo.diff ) then the machine will 
> boot fully and soesn't seem to show any issues.  I've put the dmesg from a 
> successful boot at 
> http://people.freebsd.org/~gavin/v480r/boot-withpatch
> 
> Ss, I guess my question is: is the panic in the code appropriate for all 
> Schizo busses, or just the first one?  Two of the four Schizo busses do 
> not seem to list this interrupt as one available on the bus.
> 
> Output of "prtconf -pPv" here: 
> http://people.freebsd.org/~gavin/v480r/prtconf-pPv
> 

Hi,

the problem is that unlike all firmware versions I've seen so far
this one actually includes the CDMA interrupt at RID 4 for all four
host-PCI bridges. The workaround is applicable to all four Schizos
and you most likely did not seen further problems with your hack
due to your machine not having any PCI-PCI bridges. Could you
please give the following patch a try instead?
http://people.freebsd.org/~marius/schizo_cdma2.diff

Have you used the Cassini interfaces in this machine so far? On
the few E480R FreeBSD has been given a try so far (but which
obviously are at least equipped with different firmware version
if not even being another hardware revision) using the on-board
NICs triggers a seemingly unrelated FATAL RESET in tl0_dmmu_miss
which if the dump generated by the firmware can be trusted
apparently is due to the fact that the CPUs erroneously report
a TSB base of zero.

Marius