swap_duplicate: trying to duplicate unused page

David Wong (wongste@hhb.com.my)
Fri, 20 Feb 1998 09:59:22


I've been having a problem on one of my linux 2.0.33 servers for a few
weeks now.

The kernel gives out "swap_duplicate: trying to duplicate unused page"
error messages in the logs, followed by "Hmm.. Trying to use unallocated
swap (0000f700)", repeated ad nauseum until i 'swapoff -a; swapon -a'.
After restarting sendmail (it's an email server) and the pop daemon,
everything works fine until the next time the same error occurs.

The problem started on 2.0.30, so i changed to 2.0.33, but same thing
happened.

Before you tell me it's a hardware problem, I have changed the hardware
over the last few weeks in the following order.

Original System:
Pentium 166 MMX, Adaptec 2940, 2x 3c905 PCI, 32 MB DIMMs

Then i changed the DIMMs, but the same error occured.
so i suspected maybe linux doesn't like dimms, so i changed the motherboard
and CPU to a Pentium 166 with 32MB SIMMs, but kept the scsi card and the
hdd, but the same error occured.

so i suspected the scsi card and hdd, so i changed both, the HDD to another
Quantum Fireball 4G, the scsi to another adaptec 2940, but the same error
occurred. Then i changed the scsi card to a NCR53c815, and you guessed it,
the same error. The reason that I didn't change the HDD immediately after
the error was discovered because I ran badblocks on all the partitions
(include swap) and they checked out fine.

so all in all i've changed everything except the 3c905, because we only
have the one type, and these have been working fine on another linux box
before being installed in this system.

The error occurs at 8 am, during a period of large number of smtp
connections from ~30 hosts, made at the same minute, initiated as part of a
cron job. The error only happens when swap file > 0 before the start of
this period. if swap == 0 before 8 am, the error will NOT occur. This
behaviour has been fairly consistent.

Where can i even begin to find out what's causing this problem? More
importantly, what workarounds are possible to prevent this from happening,
before I have to replace the system with a Sun. I consider myself fairly
knowledgeable in Linux. I have 6 Linux servers running on our network, all
without problems except for this one. TIA.

David

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu