Subject: Crashes during SCSI mounts, was: Jaz disks and swap
To: None <ADAMGOOD@delphi.com>
From: Hauke Fath <hauke@Espresso.Rhein-Neckar.DE>
List: port-mac68k
Date: 08/07/1997 21:23:43
At 16:34 Uhr +0200 07.08.1997, ADAMGOOD@delphi.com wrote:

>Well, I decided to try moving my whole system to a Jaz disk.  I partitioned
>it with APS.2.7.3 and gave it 150MB Mac partition, 100MB A/UX Root,
>36MB Swap, and the rest A/UX Usr.
>
>For about a second everything seemed OK.  Then the thing paniced and gave
>me this string:
>
>pmap_alloc_pv: pgi_nfree inconsistent
>
>And dumped me into the debugger.  A stacktrace shows the panic was from:
>
>_pmap_alloc_pv()
>_pmap_enter()
>_vmfault()
>_trap()
>
>I haven't done extensive work with the VM system, but it looks to me like
>it page faulted and couldn't find a page somewhere.
>_addrerr()

You mean, like this (from my private crash log)?

[root@wurzel] ~ # fsck -p /dev/rsd1g
sd1: could not mode sense (4/5); using fictitious geometry
Aug  4 08:04:04 wurzel /netbsd: sd1: could not mode sense (4/5); using
fictitiou
s geometry
sd1: could not mode sense (4/5); using fictitious geometry
panic: pmap_alloc_pv: pgi_nfree inconsistent
Stopped at      _Debugger+0x6:  unlk    a6
db> t
_Debugger(2004,82ad70,bce2c,bcd29,140ad8) + 6
_panic(bcd29,140ad8,d7c1c,82ada4,bda26) + 40
_pmap_alloc_pv(79f9000,1,79f9000,30000,ffffffff) + d6
_pmap_enter(d7c1c,79f9000,5c9000,3,1) + 208
_vmapbuf(7636f0,2000,30000,2000,6bfa280) + b4
_physio(b2f76,0,d0e,100000,b330c) + 192
_sdread(d0e,82af24,0) + 1c
_spec_read(82aed4,82aee8,38410,82aed4,110) + a6
_ufsspec_read(82aed4) + 28
_vn_read(6c72680,82af24,6bfa280) + b8
_sys_read(6c75000,82af88,82af80) + a0
_syscall(3) + 10e
_trap0() + e
db>

-- Not much hope here. Ever since I used NetBSD with an MO (for more than
two years now) I have seen this kind of crash, especially during 'first
access' to media. Some kernels used to crash so often I was unable to even
fsck(8) the disk. And sometimes this even happens with "real" hard disks.

The Jaz/Zip and the MO drive are both removable and they both don't support
the SCSI mode page 4 which otherwise holds faked drive geometry data. Maybe
the kernel has some problems there.

>I have no idea what the problem is with this.  Can anyone give me a hand
>please?  Has anyone used a Jaz disk alone for NetBSD so far?  Did anyone
>have this problem?

You might have more luck with a "real" hard disk...

	hauke



--
"It's never straight up and down"     (DEVO)