On Tue, 18 Apr 2000, Jeff V. Merkey wrote:
>
> Also, minor correction. It was dput() that was oinking in 2.4 and not
> d_lookup(). This is probably worse since dput() may trigger inode
> updates in underlying file systems. It took a lot longer to copy the
> files under 2.3.99 than 2.2.15.
Oh, but that changes the picture completely. There is a lot of things that
could change the dput() frequency - it's called in dcache trimming and any
increase of memory pressure would lead to more dput() calls. And amount of
time spent in dput() depends on the contents of dcache, so...
Something bad happened with VM balance lately. I suspect that it's a
different issue and what you see in dput() is consequence of VM problems.
If that's what changed between -pre3 and -pre5 - we'ld better start
looking at the code on memory-pressure pathes changed in that period.
However, there was a bunch of changes in lookup mechanism during that
time. I don't think that VM problems are related to that, but...
Could you try to profile 2.3.51, 2.3.99-pre3 and 2.3.99-pre5 and compare
the results? That would probably help to distinguish between the sources
of that shit...
-
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/
This archive was generated by hypermail 2b29 : Sun Apr 23 2000 - 21:00:14 EST