The handling of the "no PAG" corner case is a bit of a wart on AFS
(IMHO).  But I think we can actually take advantage of it to make exim
work smoothly.

Michael, currently the get-token script runs *before* exim4 setuid()'s
itself to a non-root user.  Can we change that to execute *after* exim
has done a setuid() to the delivery user's userid?  I think that would
make things work, although not as cleanly as I would like.

Here's why it should work: since exim is spawned at boot time and is
not invoked via pagsh, it has no PAG.  So, provided we acquire tokens
*after* switching to the user's userid, those tokens will belong to
all non-PAGged processes with that userid on that machine.

Here's why it is not as clean as I would like: my major concern at
this point is that hcoop doesn't use runsv(8), so if some admin does
an "/etc/init.d/exim4 restart" and that admin's bash has a PAG, exim
will inherit the PAG.  That could be bad.  I don't know if there's a
way to have a process eliminate its PAG (and thereby move to the "no
PAG" case).

  - a

-- 
PGP/GPG: 5C9F F366 C9CF 2145 E770  B1B8 EFB1 462D A146 C380


_______________________________________________
HCoop-SysAdmin mailing list
[email protected]
http://hcoop.net/cgi-bin/mailman/listinfo/hcoop-sysadmin

Reply via email to