🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

Thread View: comp.sys.sgi.bugs
3 messages
3 total messages Started by Andy Poling Fri, 22 Jan 1993 00:21
Unbelievable bug that didn't happen
#82
Author: Andy Poling
Date: Fri, 22 Jan 1993 00:21
133 lines
4519 bytes
I can't believe what I just witnessed.  I've been chasing down a bug where
some process (I don't know which one - I'll get to why) is scribbling pieces
of /etc/utmp into /etc/mtab.  This I can deal with -- I just look at the
modification time on mtab and check out our process-accounting file for
which processes were running at the time, right?  Wrong.

While I was tracking this bug down, it happened -- I may have even made it
happen - I dunno.  The problem is this: even though the mtab file had been
scribbled in since I had last fixed it, the modification time of the file
had not changed!!!

I've called this in to the TAC, but I figure an article is worth a thousand
words over the phone as far as demonstrating the problem (which has occurred
three times while I was writing this article).  Plus maybe it'll ring a bell
with someone else and we can nail it down faster...

Below is included an undoctored transcript of me witnessing it.  I should
add that /etc/mtab is rally a symlink to a symlink to /loc/mtab (don't ask
why).  "age" is a simple 50 line C program that stat()s a file and prints
out it's ages with second precision.

Here's the sequence of events... 
	1) I do a "mount" and find the mtab screwed
	2) I "age" it: mod time = Jan 21 18:26:56 1993
	3) I zero it, and mount -f it back to life
	4) I do a "mount" to verify it is OK
	5) I "age" it: mod time = Jan 21 18:33:47 1993
	6) I wait awhile for the scribbling to occur
	7) I do a "mount" and voila: screwed again
	8) I "age" it: mod time = Jan 21 18:33:47 1993 (STILL!!!!!)

# mount
# age -ams /loc/mtab
/loc/mtab: access: 12 secs ago - Thu Jan 21 18:33:22 1993
modify: 398 secs ago - Thu Jan 21 18:26:56 1993
change: 398 secs ago - Thu Jan 21 18:26:56 1993
# > /loc/mtab; mount -fv /; mount -fav
/dev/root mounted on /
/dev/dsk/lv6 mounted on /src
/debug mounted on /debug
/dev/dsk/lv0 mounted on /diskless
/dev/dsk/lv3 mounted on /users
/dev/dsk/lv4 mounted on /tmp
/dev/dsk/lv1 mounted on /usr
/dev/dsk/lv5 mounted on /usr/spool/news
/dev/dsk/lv2 mounted on /usr/local
# mount
/dev/root on / type efs (rw,raw=/dev/rroot)
/dev/dsk/lv6 on /src type efs (rw,noquota,raw=/dev/rdsk/lv6)
/debug on /debug type dbg (rw,noquota)
/dev/dsk/lv0 on /diskless type efs (rw,noquota,raw=/dev/rdsk/lv0)
/dev/dsk/lv3 on /users type efs (rw,quota,raw=/dev/rdsk/lv3)
/dev/dsk/lv4 on /tmp type efs (rw,noquota,raw=/dev/rdsk/lv4)
/dev/dsk/lv1 on /usr type efs (rw,noquota,raw=/dev/rdsk/lv1)
/dev/dsk/lv5 on /usr/spool/news type efs (rw,noquota,raw=/dev/rdsk/lv5)
/dev/dsk/lv2 on /usr/local type efs (rw,noquota,raw=/dev/rdsk/lv2)
# age -ams /loc/mtab
/loc/mtab: access: 2 secs ago - Thu Jan 21 18:33:49 1993
modify: 4 secs ago - Thu Jan 21 18:33:47 1993
change: 4 secs ago - Thu Jan 21 18:33:47 1993
# mount
# age -ams /loc/mtab
/loc/mtab: access: 2 secs ago - Thu Jan 21 18:35:26 1993
modify: 101 secs ago - Thu Jan 21 18:33:47 1993
change: 101 secs ago - Thu Jan 21 18:33:47 1993

In case anybody's intersted, here is the sort of scribbling that I'm talking
about:

# strings /loc/mtab.hosed
ttyq10
elmo
ttyq11
tockgae_q12
ttyq12
yooyoungq13
ttyq13
+_("yam_ls
ttyq14
kroot
ttyq15
oherb
ttyq16
telnet
ttyq17
+_0gsheppardq18
ttyq18
hosemug
ttyq19
telnet
ttyq20
+_/'gwenlynsq21
ttyq21
telnet
ttyq22
latha
ttyq23
+_!Ocheckersq24
ttyq24
+_+]telnet
ttyq25
stephen
ttyq26
+_+atelnet
ttyq27
+_*Htelnet
ttyq28
telnet
ttyq29
telnet
ttyq30
telnet
ttyq31
+_*5rlogin
ttyq32
deat
ttyq33
telnet
ttyq34
telnet
ttyq35
kroot
ttyq36
isinger
ttyq37

-Andy

Andy Poling                              Internet: andy@jhunix.hcf.jhu.edu
UNIX Systems Programmer                  Bitnet: ANDY@JHUNIX
Homewood Academic Computing              Voice: (410)516-8096    
Johns Hopkins University                 UUCP: uunet!mimsy!aplcen!jhunix!andy
Re: Unbelievable bug that didn't happen
#84
Author: Paul Jackson
Date: Fri, 22 Jan 1993 02:16
17 lines
1199 bytes
In article <1jneleINN2tb@jhunix2.hcf.jhu.edu>, andy@jhunix2.hcf.jhu.edu (Andy Poling) writes:
|> While I was tracking this bug down, it happened -- I may have even made it
|> happen - I dunno.  The problem is this: even though the mtab file had been
|> scribbled in since I had last fixed it, the modification time of the file
|> had not changed!!!

Looks to me like your mtab and /etc/wtmp files are cohabitating.  That
is, they are both using a particular block of the file system.

I'd suggest you go into single user mode and fsck /.  Expect some
things such as DUP blocks.

-- 

				I won't rest till it's the best ...
				Software Production Engineer
				Paul Jackson (pj@sgi.com), x1373
Re: Unbelievable bug that didn't happen
#94
Author: Andy Poling
Date: Tue, 26 Jan 1993 20:42
27 lines
1905 bytes
In article <1jnlejINNp5m@fido.asd.sgi.com> pj@sgi.com writes:
>In article <1jneleINN2tb@jhunix2.hcf.jhu.edu>, andy@jhunix2.hcf.jhu.edu (Andy Poling) writes:
>|> While I was tracking this bug down, it happened -- I may have even made it
>|> happen - I dunno.  The problem is this: even though the mtab file had been
>|> scribbled in since I had last fixed it, the modification time of the file
>|> had not changed!!!
>
>Looks to me like your mtab and /etc/wtmp files are cohabitating.  That
>is, they are both using a particular block of the file system.
>
>I'd suggest you go into single user mode and fsck /.  Expect some
>things such as DUP blocks.

Paul takes the prize.  Each of the copies of the scribbled mtab that I had
saved for posterity showed up as BAD/DUP in phase 1 of the fsck.

Of course, how they got that way in the first place is a mystery since I
kept creating new mtab files which kept having this happen to them.

Thanks Paul!

-Andy

Andy Poling                              Internet: andy@jhunix.hcf.jhu.edu
UNIX Systems Programmer                  Bitnet: ANDY@JHUNIX
Homewood Academic Computing              Voice: (410)516-8096    
Johns Hopkins University                 UUCP: uunet!mimsy!aplcen!jhunix!andy
Thread Navigation

This is a paginated view of messages in the thread with full content displayed inline.

Messages are displayed in chronological order, with the original post highlighted in green.

Use pagination controls to navigate through all messages in large threads.

Back to All Threads