> a SPARC 1 box with a Cycle Computers SPARC 5 upgrade board running
> 4.1.4 has panicked twice yesterday with a data access exception (see
> below). The crashes were quite spectacular with open files (including
> NIS master passwd and groups) filled with binary garbage and
> and ownership for the root directory trashed. I do have most
> patches installed and a few more (see below).
> Question: Is this a software issue or is it the hardware acting up? Is
> there anything that I could learn from the crash messages or a saved
An application chown()ed the root directory to an ordinary user and
removed read/execute permissions for the rest of the world, causing the
sytem to go down in flames.
I apologize for not reading the FAQ. Teaches me about relying on keyword
searches. Numeric crash dumps from the system logs are useless, see
Subject 6.1 of the FAQ. Thanks to Kevin.Sheehan@uniq.com.au for pointing
After enabling crash dumps I tried to re-create the situation leading to
the crash and found out that I had cause and effect reversed. I was
using Helios Ethershare Admin to add a dummy user for Mac file access.
Since he is not supposed to log in at all I gave him /bin/false as login
shell and / as home directory. Unfortunately Ethershare Admin chown()s
the home directory to the user and takes away permissions for others.
This caused the machine to crash after a while.
The funny thing is that the machine sort of boots even with incorrect
permissions for /. It just denies login to everyone but root. If on the
other hand permission are taken away while someone is logged on, the
"Ryan Lewis" <email@example.com> was on vacation.
Joel Lee <firstname.lastname@example.org> suggested a software problem.
Dieter Gobbers <email@example.com> and
Drexx Laggui CSA <firstname.lastname@example.org> suggested hardware problems.
Kevin.Sheehan@uniq.com.au pointed me to crash dumps.
Thank you for taking the time to think about my question.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:09 CDT