[PATCH] reduced latency in dentry and inode cache shrinking

Shrinking a huge number of dentries or inodes can hold dcache_lock or
inode_lock for a long time.  Not only does this hold off preemption -
holding those locks basically shuts down the whole VFS.

A neat fix for all such caches is to chunk the work up at the
shrink_slab() level.

I made the chunksize pretty small, for scalability reasons - avoid
holding the lock for too long so another CPU can come in, acquire it
and go off to do some work.

diff --git a/mm/vmscan.c b/mm/vmscan.c
index b88d59b482d40311998ef1698b3a92d65ff34317..af45e9a9ab4e9df9f0b4d936f1e9f66fd6556696 100644 (file)
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -155,10 +155,18 @@ static int shrink_slab(long scanned,  unsigned int gfp_mask)
                do_div(delta, pages + 1);
                shrinker->nr += delta;
                if (shrinker->nr > SHRINK_BATCH) {
-                       long nr = shrinker->nr;
+                       long nr_to_scan = shrinker->nr;
 
                        shrinker->nr = 0;
-                       (*shrinker->shrinker)(nr, gfp_mask);
+                       while (nr_to_scan) {
+                               long this_scan = nr_to_scan;
+
+                               if (this_scan > 128)
+                                       this_scan = 128;
+                               (*shrinker->shrinker)(this_scan, gfp_mask);
+                               nr_to_scan -= this_scan;
+                               cond_resched();
+                       }
                }
        }
        up(&shrinker_sem);