Re: scheduler problems with linux 2.0.11/redhat 3.0.3

[Matti Aarnio <mea@nic.funet.fi>]

> Hi
> 
> I installed zmailer-2.99.36 on my home machine but the scheduler seems
> to be broken:
> 
> fred:/var/log/mail# ls -l /var/log/mail/scheduler
> -rw-rw-r--   1 root     root     165788672 Aug 12 03:00 /var/log/mail/scheduler
> 
> scheduler log contains this:
> 
> bind:mailq socket: Address already in use

	There are two scheduler's running at the system, or for
	some reason you have a lingering (but dying?) socket
	left around blocking a bind() for passive listener socket
	from being created for the  mailq  service.

	OR you are running the scheduler as non-root ?
	(It is possible, but not recommended..)

> Synchronous startup completed, messages: 0 (0 skipped) recipients: 0
> ***********************************************************************
> bind:mailq socket: Address already in use
> **** recursed into mux()! ***
> **** recursed into mux()! ***
> 
> .. and then 165MB of the recursed into mux() messages. 
> This after ca 30mins of uptime. The only message I sent was a testmsg
> to a local user.
> 
> Any ideas?

	I have seen it occasionally, and it should not happen, but as
	it does, I guess I must look at it again.

	That means, among others, that  mux()  exits uncleanly without
	clearing the "in_select" flag..

	.. searching the "return" statements ...  does reveal one return
	without clearing the set in_select variable.  Oops.

	OTOH it does happen only when the  mailq  socket has not been opened.

	Fix:   Move  "in_select = 1;" in the  mux() of  scheduler/transport.c
	right front of select() statement.  (So it is after a test of
	"if (maxf < 0)". )

> -Andi
> --
> |andi@mlm.extern.lrz-muenchen.de     Nonsense is better than no sense at all.

	/Matti Aarnio <mea@nic.funet.fi>