> > that's what I did now. I've gnuplotted a few graphs from the collected
> > information which can be found here: http://www.foobar.de/memgraphs/.
>
> Fascinating. There's a deep dip in the dentry_cache - I wonder where all
> that memory for that particular line has disappeared to.
IIRC, when memory pressures start to become acute, the dcache gets suddenly
pruned, which accounts for the drop of ~300 in dentry_cache. Note that
pruning the dcache will free memory elsewhere as well, particularly inodes,
which aren't allocated using the slab allocator.
> I would have
> expected the other lines to show a corresponding dip going upwards. So I
> would suspect a leak in the dentry_cache code in certain circumstances.
I'm don't think so, but I'm not an expert at this.
The thing that __really__ worries me is that skbuff_head_cache is >100,000.
It's currently *3* on my system, with a max of 125.
Joerg: what's running on this system? Do you see any odd network behaviour?
> Also, the steadily gowing-upwards curves for the rest is disturbing. I
> don't see any sign of memory being freed-up over time. There should be a
> point where it stabilises and starts re-using buffers instead of grabbing
> more and more memory as it requires.
*nods* Deep weirdness seems to be occurring. I'd be happier if one of the
more experienced mm people poked their heads in here.
S.
-- "Water? Never touch the stuff! Fish fuck in it." -- W. C. Fields- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/