Subject: Re: 32bit kernels busted
To: matthew green <mrg@eterna.com.au>
From: Andrey Petrov <petrov@netbsd.org>
List: port-sparc64
Date: 02/06/2004 15:30:36
Hi Matt,

On Sat, Feb 07, 2004 at 12:42:33AM +1100, matthew green wrote:
> 
> 
> 32 bit kernels seem busted.  my U10/440's latest kernel crashed coming
> upto multiuser, and the U5/300 has had the same problems with a GENERIC32.
> 
> the U5 looks like this, when it's trying to create an mfs /dev:
> 
> warning: no /dev/console
> inidata fault: pc=12ddd10 addr=0
> kernel trap 30: data access exception
> Stopped in pid 1.1 (init) att    netbsd:ubc_release+0x1d0:       lduh            [
> %g3 + 0x28], %g1
> db> : Creating mfs /dev (370 blocks, 896 inodes)
> db> t
> ffs_write(0, 6ea2184, 6e3bbb8, 6ea2180, 1, 6) at netbsd:ffs_write+0x32c
> VOP_WRITE(6e671b8, 6e3bcb8, 25, 1e25e80, ffffffff, 805) at netbsd:VOP_WRITE+0x30
> 
> vn_write(0, 6e94060, 6e3bcb8, 1e25e80, 1, c5) at netbsd:vn_write+0xcc
> dofilewrite(16, 0, 6e94038, 40058000, 563a, 563a) at netbsd:dofilewrite+0x78
> sys_write(9, 6e3bdd8, 6e3bdd0, 81c061ac, 6e3bf10, 200) at netbsd:sys_write+0x60
> syscall(6e3bed0, 4, 0, 3, 400de5b8, 6e3bdd8) at netbsd:syscall+0xdc
> ?(0, 40058000, 563a, 0, 0, 0) at 0x1008cb8
> db>
> 
> 

The same happened for me on U2, populating /dev with devices let the kernel
reach multiuser, I'll try to stress it. I also have the whole set of older sys
trees, I'm going to check older kernels.

And ffs_write(0) doesn't look right.

Regrads,
	Andrey