Re: Install panic w/ 2003-12-09 jpsnap...

> > During install, I picked up the following panic shortly after
> > formatting and during chunk 4 of the base dist.  :-/ -sc
> > 
> > kernel: type 12 trap, code=0
> > Stopped at   _mtx_assert=0x4e:	movl	0x1c(%ebx0,%eax
> > db> trace
> > _mtx_asssert(0,1,c0877786,63e,1000) at _mtx_assert+0x4e
> > vm_page_set_invalid(c17f0c30,0,1000,554,0) at vm_page_set_invalid+0x35
> > brelse(c8010b88,0,ce893b90,c336f480,c3075c80,ce893bc4,346,c0865d8b0 at softdep_dissk_io_initiation+0xc4
> > spec_xtrategy(c3372924,c8010818,d5b,c0865d8b,c8010818) at spec_xstrategy+0x117
> > spec_specstrategy(ce893bf8,f4,c179f6d8,4,c8010818) at spec_specstrategy+0x72
> > ufs_strategy()
> > ufs_vnoperate()
> > bwrite()
> > vfs_bio_awrite()
> > fluusshbufqueues()
> > buf_daemon()
> > fork_exit()
> > fork_trampoline()
> 
> I think this is the same as kern/59792.  There seem to be some problem
> if soft-updates are enabled on the root filesystem.

Actually, I think the analysis in the PR may be wrong.  Use
atacontrol(8) and set the mode to pio4 and this problem goes away.
Something like:

atacontrol mode 0 pio4 biospio

Whether or not it's a workaround and masks the problem, or if it's the
real deal, I don't know.  It feelss like this is an ATAng problem and
not a softupdates problem.  Though by looking at the back trace, it
would sure seem like a softupdates problem.  Either ATAng is tickling
a bug in softupdates (doubtful, I have this problem without
softupdates), or, ATAng is incorrectly handling UDMA and is causing
problems for softupdates.

Anyway, now that I've hosed my machine enough times, I'm setting:

echo "hw.ata.ata_dma=0" >> /boot/loader.conf

and my HDD has worked just fine, however my CD-ROM drive doesn't like
this and refuses to be recognized as a valid source for install data
(nevermind that I booted from it!).  ACPI (everyone's favorite
scapegoat) wasn't the culprit and can be safely enabled from what I
can tell.

dmesg -a attached from a boot -v'ed kernel.  WD's diagnostics
(extended test) say this drive is a-okay.  -sc

-- 
Sean Chittenden