[lkml]   [2001]   [Apr]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Patch in this message
Subject[PATCH][RFC] appling pressure to icache and dcache - simplified

Rik asked, "Can it be made simpler?"

So I went back to the basics, inserted a printk in kswapd and watched
the dentry_stat and inodes_stat numbers for a while. I observed
the following pattern. The dentry cache grows as does the number of
unused entries in it. Unless we shrink this cache objects do not seem
to be reused. At the same time the inode cache usually kept about 15%

At this point I starting to shrink the dcache. The goal being to keep
the size of the cache as observed in /proc/slabinfo reasonable without
much overhead. From experimenting, it turns out that if the shrink
call is made when there is over 50% free space the cache stays small.
Using 66% is not quite as aggressive but achieves its effect with about
half the shrink calls.

With the pressure on the dcache, I looked at the icache numbers. With
the dcache shrinking the amount of free space in the icache was much
higher. It turns out that using the same logic as above with, 80% as
the amount of free space, it works well.

Here are the results against 2.4.3-ac3


--- linux.ac3.orig/mm/vmscan.c Sat Apr 7 15:20:49 2001
+++ linux/mm/vmscan.c Sat Apr 7 12:37:27 2001
@@ -997,6 +997,21 @@
refill_inactive_scan(DEF_PRIORITY, 0);

+ /*
+ * Here we apply pressure to the dcache and icache.
+ * The nr_inodes and nr_dentry track the used part of
+ * the slab caches. When there is more than X% objs free
+ * in these lists, as reported by the nr_unused fields,
+ * there is a very good chance that shrinking will free
+ * pages from the slab caches. For the dcache 66% works,
+ * and 80% seems optimal for the icache.
+ */
+ if ((dentry_stat.nr_unused+(dentry_stat.nr_unused>>1)) > dentry_stat.nr_dentry)
+ shrink_dcache_memory(DEF_PRIORITY, GFP_KSWAPD);
+ if ((inodes_stat.nr_unused+(inodes_stat.nr_unused>>2)) > inodes_stat.nr_inodes)
+ shrink_icache_memory(DEF_PRIORITY, GFP_KSWAPD);
/* Once a second, recalculate some VM stats. */
if (time_after(jiffies, recalc + HZ)) {
recalc = jiffies;
Ed Tomlinson <>

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 13:24    [W:0.019 / U:8.384 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site