>The 'real' shrinking function, check_dcache_memory()
>is being called from fs/namei.c in the functions
>__namei() (called from open(), etc...) and open_namei(),
>so the dcache will only shrink on open()s of files.
This is what I thought developing the patch.
>The added call from kswapd is, IMHO, a good thing.
>This is mainly because small-mem machines don't always do
>enough open()s to recover from a bad case of find...
Yes. check_dcache_memory() could be also very more aggressive against the
dcache memory. On the 8m laptop of my friend I removed the check of
(page_cache_size > (num_physpages >> 1)), the goal limits and I given more
weight to dentry_stat.want_pages and things was going better (at least my
friend could return to use 2.1 with performance like to 2.0).
Andrea[s] Arcangeli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu