Path: utzoo!utgpu!watserv1!watmath!att!dptg!mtune!jcm
From: jcm@mtune.ATT.COM (John McMillan)
Newsgroups: comp.sys.att
Subject: Re: Fixdisk problems?
Summary: O... THAT...
Keywords: cursed
Message-ID: <309@mtune.ATT.COM>
Date: 12 Feb 90 15:02:24 GMT
References: <111@spirit.UUCP> <301@mtune.ATT.COM> <1896@neoucom.UUCP>
Distribution: na
Organization: AT&T BL Middletown/Lincroft NJ USA
Lines: 49

In article <1896@neoucom.UUCP> wtm@neoucom.UUCP (Bill Mayhew) writes:
:
>This is what the panic message looked like.  The panic scribbled
>over the metermaid and some of the other stuff on the screen, so it
>was a bit tough to read.  Seems somewhat elusive because I tried to
>duplicate the conditions again and have not been able to get the
>crash.
>
>type = 0x02, pid = 17144, pc = 0x6C09, rps = 0x2000, .... 0x4BB5C
				     ^-SPURIOUS: probably on screen at time 
				     of panic.  REAL value was 0x6C0
						====================
>GSR = 8D00, BSR0 = 7C07, BSR1 = 2400 PHYSPF = 0
>D0 = ff, D1 = 6030, D2 = 301, D3 = 5
>D4 = 52, D5 = 400, D6 C000, D7 = 400
>A0 = 30300, A1 = 72400, A2 = 4BB5C, A3 = 70E08
>A4 = 70884, A5 = 413B8, A6 = 70820, userA7=2FF138/kernA7=707C4
>KI-RAM@6BC:000422D8 51C8FFFC 4E75227C 004000E0
>KS-RAM@707C4:
> 000101C6 000302FC 00072400 00000400 000270EC 00000000 00000034 08000007
> 0B560000 13312002 00051052 000001F4 00400602 00000000 00000001 FFFB7C30
:
>Anybody have any ideas?

	A serious-but-unexplained problem of the UNIX-pc has been
	a long-occuring PANIC in 'bcopy'.  For almost everyone it
	is a great rarity.  I don't know that I've ever had one in
	five years, but it's probably been there since day-0.

	The data you present places the panic in the kernel's 'bcopy'
	routine where it's paging in a TEXT page from a program.
	Comparing A0 (from_ptr) and A1 (to_ptr) to their sources
	in the Stack (KS-RAM[1,2]), I'd _GUESS_ the Memory-READ
	action had occurred (*from_ptr++) but the WRITE panic'd
	the kernel for unknown reasons.

	The problem's of enough concern that there's a debugging aid
	hidden in the BCOPY code: the HPTE for the FROM page is tucked
	in a register (D1): 6030.  This value is reasonable.

	In simple fact: I've never seen a 'bcopy' panic where EVERYTHING
	wasn't reasonable !-(  It is probably time to add a test to
	the panic software that identifies the In-BCOPY state and
	dumps the to_ptr's HPTE as well.  [Yes, Karl: you may be
	squeeking again, soon !-]

Everyone bored enough?

john mcmillan -- att!mtune!jcm -- muttering for self, not THEM