Memory abuse with NFS Root filesystem?

Memory abuse with NFS Root filesystem?

[Skip Gaede]

Folks,

I have a Mac client that's running 2.4.19-pre10, XFree86 
(4.2.0) -query + Mozilla 1.0 and losing lots of size-2048 
buffers. The rate loss is approximately 1000 buffers/hour, 
no reaping of the cache is ever performed, and after 8 
hours or so, the system is unusable. I'm seeing this loss 
of memory after I boot with an initrd, get an IP address 
with dhclient and do a pivot_root onto the read-only NFS 
file system.

I can also boot the the client to a bash prompt on the
 local hard drive and then run XFree86 -query + Mozilla as
 before. Under these conditions, both with and without the
 local filesystem mounted RO, the size-2048 buffer
 utilization is nominal.

Local boot: slabinfo after 30 minutes:
size-2048      5   50  2048    3    25  1 

NFS Root: slabinfo after 15 & 60 minutes
size-2048   187 228  2048  94  114  1  (15 min)
size-2048   918 928  2048 460 464  1  (60 min)

Anyone ever seen or heard of this before? Any thoughts on
how to isolate this to a root cause?

Thanks,
Skip

-------------------------------------------------------

Re: Memory abuse with NFS Root filesystem?

[Trond Myklebust]

>>>>> " " == Skip Gaede <sgaede@attbi.com> writes:

     > Local boot: slabinfo after 30 minutes: size-2048 5 50 2048 3 25
     > 1

     > NFS Root: slabinfo after 15 & 60 minutes size-2048 187 228 2048
     > 94 114 1 (15 min) size-2048 918 928 2048 460 464 1 (60 min)

     > Anyone ever seen or heard of this before? Any thoughts on how
     > to isolate this to a root cause?

Have you tried combining the 'local boot' test with NFS?

AFAICS there are no memory leaks on NFS with x86 systems, but there
might be something wierd going on with other architectures.

Cheers,
   Trond