contigmalloc for low address in SI_SUB_KMEM

Matthew Dillon dillon at apollo.backplane.com
Wed Sep 17 10:30:19 PDT 2003


:Yes, your fix is working fine here, as I replied to you on kernel@ list.
:
:The message you replied was originally sent on Septemper 6th,
:but have stayed on myrealbox server until now. Unfortunately at that time,
:myrealbox.com mail server seemed to be under heavy attack from virus-
:..
 
    Exellent.

    This reminds me of perpetual issues we had at BEST Internet with mail
    delivery.  In the days before dirpref, dirhash, and good directory
    caching in the VM system something would go wrong and the boxes would
    get behind and start down a slippery performance slope which they would
    take 24 hours or more to recover from.

    I finally solved the problem by creating multiple queues (this was long
    before sendmail had a multi-queue capability built in).  Each time
    sendmail ran a queue file unsuccessfully a script would move the queue
    file into the next higher queue.  Each successive directory had fewer
    queue runners and longer retry periods.  That plus we round-robined
    incoming MX to three boxes and separated outgoing mail to run through
    three other boxes.

:...
:I thought it was dropped. I'm not familiar with how mail messages are
:dealt with inside myrealbox, but sometimes they seems to deliver those
:postponed messages several days later.

    Yah.  The problem with this particular virus is that it uses legitimate
    envelope addresses so the traffic is compounded by mail systems 'returning'
    delivery failures as well as virus scanners 'returning' delivery failures.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>





More information about the Bugs mailing list