home *** CD-ROM | disk | FTP | other *** search
- From: kwb@hplvec.LVLD.HP.COM (Keith Blackwell)
- Date: Tue, 1 Sep 1992 17:13:39 GMT
- Subject: Re: Unkillable processes
- Message-ID: <560036@hplvec.LVLD.HP.COM>
- Organization: Hewlett-Packard Co., Loveland, CO
- Path: sparky!uunet!gossip.pyramid.com!decwrl!sdd.hp.com!hpscdc!cupnews0.cup.hp.com!hppad.waterloo.hp.com!hppad!hpfcso!hplvec!kwb
- Newsgroups: comp.sys.hp
- References: <1992Aug26.180629.385@ctp.com>
- Lines: 29
-
- In comp.sys.hp, ignatz@wam.umd.edu (Mark J. Sienkiewicz) writes:
-
- > Anyway, you can probably think of bunches more-- just think "What could
- > make a process wait in kernel mode?"
-
- After a power glitch fried our local LAN, my s300's disk had a strange
- problem (though fsck never complained): whenever a process attempted
- to access a particular directory (in which I had been working during the
- power glitch), that process would hang in an unkillable state. Obviously,
- it was not possible to remove the directory, since removing it requires
- accessing it. Even a "stat(2)" call would cause a process to hang.
- Even after rebooting (to clean up all those hung processes), the problem
- persisted. [this happened on HP-UX 8.0, by the way]
-
- I eventually had to use fsdb(1M) to twiddle bits on the disk so as to
- "remove" that directory. I had to keep running fsdb and fsck over and
- over until I got it right. Not a fun job. And a very dangerous one at
- that.
-
- I never found anything wrong with the directory's inode, though it's
- possible that the "flags" field may have had some flag set inappropriately
- and this caused the kernel to wait erroneously. I've never heard of
- anyone else having a problem like this (and I certainly don't know enough
- about HP-UX internels to explain what *really* happened), but maybe someone
- else might run into the same problem. Using fsdb is a nightmare, but it
- might be the only choice available.
-
- ---
- Keith Blackwell (not a spokesman for HP, just a fellow HP-UX user)
-