Yeah. I set up a test network and have two machines up to about 10 million
interrupts on their ethernet cards, and no discernable leak. Furthermore,
the machine I have been watching had memory come *back* after a few days of
seeming to drain away. I'm in the process of looking into this further,
but it appears that we're seeing the NFS server gobbling up memory. I
appear to have jumped on this too early -- we had a machine running a
previous prerelease die running out of memory, and saw another report of
a machine running out of memory, and jumped to a conclusion without proper
investigation. <blush>
All but one of the reports have involved NFS. The test machines I've got
going are not doing NFS, and they aren't showing any problem.
>Have the people seeing the leak patched up to pre9 or did you use the big
>pre9 tar.gz Linus uploaded? I've been using pre9 from the big tar. Are
>you all running with at least one IDE disk? My systems have SCSI only
>(some NCR-810, some Buslogic 946, some 948).
We aren't using IDE disks, but it probably doesn't matter what we are
using, considering that we probably aren't seeing this problem after
all.
We ARE using glibc and we ARE using the latest nfs server from Olaf.
Preliminary tests show that it does not appear to be happening with
libc 5 and Olaf's latest nfs server.
michaelkjohnson
"Magazines all too frequently lead to books and should be regarded by the
prudent as the heavy petting of literature." -- Fran Lebowitz