"Charles M. Hannum" writes:
> 
>    I've just installed the i386 snapshot on a machine that was working
>    fine for some time. The kernel can't manage to find a number of
>    important devices -- including the ethernet board -- and has
>    hallucinations about others including an mcd device.
> 
> Is your ethernet board at one of the addresses specified in the config
> file?

Hard to tell. The config I was building had it in the right place, of
course, which was one of the reasons to build the new one.

> There's very little I can do about the mcd probe falsely recognizing
> devices occasionally; the protocol just sucks.

Thats cool. The real problem is the panic. It would be nice, though,
if the probe didn't print out garbage characters...

>    Thinking that the right thing to do would be to build a new kernel
>    with most devices not on the machine stripped out of the kernel (which
>    usually cures this sort of madness), I began the process with an
>    innocent make depend -- which reliably throws me into ddb with a
>    vm_fault.
> 
> I can't *possibly* determine what would cause that without at least a
> stack trace from DDB.

Unfortunately, getting the data off the machine is hard -- this is
just the list of calls as hand-copied. If this sequence tells you
anything, good -- otherwise I'll try to hand transcribe the full list.

_ffs_clusteracct at +0x221
_ffs_blkpref     at +0xc31
_ffs_blkpref     at +0xc644
_ffs_blkpref     at +0x251
_ffs_alloc       at +0x138
_ffs_balloc      at +0x2c3
_ffs_write       at +0x1f3
_vn_write        at +0xcc
_write           at +0xae
_syscall         at +0x1e1

Perry