kladit@xxxxxxxxxxx (Klaus Dittrich) wrote:
Can you narrow the onset of the problem down to any particular kernel>snapshot?
Did it and here is the answer.
kernel-2.6.7 and bk's up to 2.6.7-bk7 survived a du -s,
kernels starting with 2.6.7-bk8 did not.
Dammit, -bk7 to -bk8 is a 1.8M diff. Relevant changes include the switch
to the rcu callbacks (make them take an rcu_head* rather than a void*) and
the introduction of /proc/sys/vm/vfs_cache_pressure.
So the immediate question is: please check the contents of your
vfs_cache_pressure tunable. It should be 100. A setting of zero would
cause this behaviour.
Compiler gcc-3.4.1
It would be useful to try a different compiler version.
There's _something_ different in your setup. If we can work out what this
factor is, it will lead us to the bug.