Re: SMP with LOCKDEBUG crashes

[Michael Lorenz <macallan%netbsd.org@localhost>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Jul 9, 2008, at 11:59, John D. Baker wrote:

> I reported a problem like this some time ago in
>
>   http://mail-index.netbsd.org/port-sparc/2007/06/24/0000.html
>
> and subsequent messages.
>
> I haven't had time to investigate this for some time.  I'll be able to
> set that machine up in about a week and have another go at it for both
> NetBSD-4 and -current.

That's something completely different. You're using 4.0, I use - 
current. Mine crashes before attaching CPUs, yours makes it to  
multiuser. Also, I get a similar crash without LOCKDEBUG.

Anyway, I'm pretty sure my SS20 crashes in percpu_cpu_enlarge() -  
there's
        CPU_INFO_ITERATOR cii;
        struct cpu_info *ci;
... and I vaguely remember problems with CPU_INFO_ITERATOR in the  
past. I guess we don't setup the (struct cpuinfo)s early enough on  
sparc.
With LOCKDEBUG we crash in lockdebug_alloc() - no CPU_INFO_ITERATOR  
there but it does use curcpu() which probably doesn't do The Right  
Thing that early in kernel startup with SMP support.

Does that ring any bells with anyone?

have fun
Michael

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBSHa9T8pnzkX8Yg2nAQJYJAgArWerOKW/YmTa1KppKpRMK6M61NM/uikA
Vpd3lC8ZGkyh0obZzPoQpN3AM/D3zp1TcnJDccZ8ctLqXSWDqN8HL842FNkyRFpb
siWv2vD9gOII3Mpg1aIfcvMTMkR+brl2fVStCgj6Sp6DS51eT92R67py9OAEpn6k
g1ETu8ChZpaczFC5yiFerCCwWIyhRZCcEjtYbvc1Zai8Q3agh1qzfAn9aiT8Z1LR
FBi25PMNRQxtO0EDXreJPbllE8ctNMX7jg6RKCz2WNuGXfBsr96+TrzGsqKGynow
xMQFvgIEMB1IQyE5PERv5bAkb60WgWlOY1/bMYXWhFVXXeKgomfDIA==
=gLW+
-----END PGP SIGNATURE-----