[26836] in Hotline Meeting
bombardier fixed
daemon@ATHENA.MIT.EDU (Sam Hartman)
Thu Feb 9 17:33:45 1995
Date: Thu, 9 Feb 1995 17:33:34 -0500
From: Sam Hartman <hartmans@MIT.EDU>
To: rs6kswat@MIT.EDU
Cc: shalabi@MIT.EDU, hotline@MIT.EDU
I fixed bombardier today. Apparently, either before or after
the update, it ran low on space in /usr. I assume the update
completed successfully, but am not sure. We need to address this
space problem.
Anyway, somehow, CellServDB and SetuidCells got mangled;
SetuidCells was basically empty, and CellServDB was not being parsed
correctly by afsd. So, on reboot, it did a synctree of /srvd, which
promptly removed the set-uid bits from all the binaries in
/bin/athena, including attach. This is bad. On another reboot,
CellServDB got so mangled that AFS was unable to mount, so you got a
connection timed out to any access to AFS. In general, this made the
machine difficult to use.
I believe this identifies the following areas to work on:
1) We need to resolve the space determining issue in updates so it
tends to work reliably.
2) We need to fix config_afs so it notices if it is out of disk
space.
3) A workstation should not run synctree against a srvd pulled from a
cell that it doesn't trust for set-uid.
]
--Sam
