[grabben@acc.umu.se: Loading xfs module on Solaris 2.5.1 gives kernel panic]

[Assar Westerlund]

Bjorn Grabbe <grabben at acc.umu.se> writes:
> Resending the mail since I don't think it got through the first time.

Only subscribes can post directly to the list for the usual spam
reasons.  Other mails we have to forward manually and our availability
might be slightly less than 24 hours per day.

> BAD TRAP: type=9 rp=fc10a8e4 addr=14 mmu_fsr=326 rw=1
> modload: Data fault
> kernel read fault at addr=0x14, pme=0x0
> MMU sfsr=326: Invalid Address on supv data fetch at level 3
> pte addr = 0xf6bedd00, level = 3
> pid=8194, pc=0xf0081368, sp=0xfc10a930, psr=0x119000c1, context=71
> g1-g7: 0, 20000000, ffffffff, 0, 0, 1, f6b6e6e0
> Begin traceback... sp = fc10a930
> Called from f0080f64, fp=fc10a9b8, args=14 f5b31ac4 3 f6c5f000 f6c5f000 0
> Called from f0083240, fp=fc10aa18, args=f63c4220 f63c4258 f6bd13e0 ae 9 f6bd1408
> Called from f01048b4, fp=fc10aa78, args=f64348c0 258 f63c4220 f0272a80 f63c4228 
> 0
> Called from f01043cc, fp=fc10aad8, args=f64348c0 0 0 0 0 f028cbf8
> Called from f0103714, fp=fc10ab38, args=f64348c0 0 fc10ab9c 0 17 f01036ec
> Called from f00923e0, fp=fc10aba0, args=fc10ac90 fc10ac00 f0000000 98 980 f6bfa4
> 90
> Called from f006d5c8, fp=fc10ac08, args=0 fc10ac64 effff7a0 66730044 f027b2c4 f6
> ba1c00
> Called from 10924, fp=effffba0, args=2 effffc04 effffc10 0 0 effffd04
> End traceback...
> panic: Data fault
> syncing file systems... done
>  6104 static and sysmap kernel pages
>   161 dynamic kernel data pages
>   434 kernel-pageable pages
>     0 segkmap kernel pages
>     0 segvn kernel pages
>   177 current user process pages
>  6876 total pages (6876 chunks)
> 
> The machine is a Sparc10 with 4 CY605 cpus.
> SunOS Release 5.5.1 Version Generic_103640-31 [UNIX(R) System V Release 4.0]

The SMP-code is not really that well tested, but I haven't heard of
anyone having similar problems.  Do you get this crash when loading
the kernel module or when doing something else?

Can you also make sure your solaris does a savecore and then run adb
on the crash dump?  You should have a directory called
/var/crash/`hostname` and make sure savecore is not commented out in
/etc/rc*.d/*

And then try something like:

adb -k unix.0 vmcore.0
$c

Thanks.

/assar