At 22:59 Uhr 12.06.1996, Scott Reynolds wrote:
>On Thu, 13 Jun 1996, Hauke Fath wrote:
>
>[With ncrscsi:]
>
>> >sd2: could not mode sense (4); using fictitious geometry
>> >panic: pmap_alloc_pv: pgi_nfree inconsistent
>> >Stopped at      _Debugger+0x6:  unlk    a6
>> >db> t
>> >_Debugger(16c28,87895,99dd64,1,99dd68) + 6
>> >_panic(87895,12ac58,99dd9c,8814e,7329000) + 34
>> >_pmap_init(7329000,1,7329000,33000,ffffffff) + 484
>> >_pmap_enter(9ec64,7329000,9d9000,3,1) + 156
>> >_vmapbuf(95b830,2000,33000,2000,7cd380) + bc
>> >_physio(7b850,0,d16,100000,7bb9e) + 18c
>> >_sdread(d16,99df28,0) + 1c
>
>Looks like the same crash.

It is. It is reproducible: Insert a ffs MO | fsck -f /dev/rsd2g | BANG.
I can mount it, but I cannot fsck it.

>You might try updating your sources and see if the -current code helps you
>at all.

I haven't seen a working kernel with ncrscsi driver since 15 May.
A kernel with sbc driver from sources sup'ed last night (12 Jun,
wipux2.wifo.uni-mannheim.de) shows exactly the same behaviour.

>A suggestion:  try changing the definition of splimp() in psl.h to use
>spl4() instead of spl2() and see if the crash goes away... patch appended.

I'll try that. (2 1/2 hrs...)

>--scott




	hauke

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