>That'd be great. Because I can't imagine, why the
>'bare-LFs' thing should only affect qmails on Solaris 7 -
>and why it should trigger this undeterministic. If bare
>LFs would be the reason, it should trigger on the first
Well in the case I was looking at, the machines in question
are used by tens of thousands of users every day - so
some of these may have broken clients. Alternatively, mail
is coming in for those same users from broken servers. As
I write, the systems are still happy (no recurrence of the
000's of procs), so it really looks like bare LF was the
issue. Since it's now 'Fixed', I'm unlikely to be back with
that system for a while....
Contrary to your assertion that it only affects Solaris 7, I did
see that other systems (Linux) were affected in other messages
in the archive (go search...) - so I suspect all platforms could
have this issue if 'Stoned' enough by broken mailers.
I suppose it's worth mentioning that the issue was sporadic,
so that the apparent DoS would last for maybe an hour... which
suggests dialup users to me, with broken clients, rather than
Unfortunately at present dialup "Outbound" mail is handled by the
same machines as Internet "Inbound" mail, meaning that a
"DoS" like this from either source degrades both services. This
will be changed to having separate "Inbound" and "Outbound"
machines - this reduces the impact of this problem.
I think for the systems concerned, bare-LF mailers must be
pretty rare, but once a couple started appearing, it spelt trouble.
From: Toens Bueker[SMTP:[EMAIL PROTECTED]]
Sent: 29 July 2000 23:36
To: [EMAIL PROTECTED]
Subject: Re: qmail-1.03 on Solaris is broken
Andrew Richards <[EMAIL PROTECTED]> wrote:
> The SMTP service may issue a QUIT, and immediately try again,
> resulting in a potential loop."
> The actual qmail-smtpd error message re bare LFs is
> 451 See http://pobox.com/~djb/docs/smtplf.html
> which would trigger the above fault if Microsoft's software does
> indeed send bare LFs - contributors suggest it does.
> Anyway, part of my reason for posting was to speculate on why
> a mailserver might get a flood of SMTP connections.
Now, I'm testing qmails behaviour under these conditions,
'cause I need to relay a quite reasonable amount of mail
through it a few times a week. This is no spam, though.
> The above bare LF issue is obviously one, as are
> smtpstone and a DoS. In my case, fixing the bare LF
> problem fixed the many-procs problem, by fixing the
> thing that was triggering it, but there may still be
> something that is 'broken' in Solaris 2.7. If I'm
> feeling brave, and happen to be working with that system
> again, I'll try smtpstone-ing it...
That'd be great. Because I can't imagine, why the
'bare-LFs' thing should only affect qmails on Solaris 7 -
and why it should trigger this undeterministic. If bare
LFs would be the reason, it should trigger on the first
Linux. The dot in /.