percpu: speed alloc_pcpu_area() up

If we know that first N areas are all in use, we can obviously skip
them when searching for a free one.  And that kind of hint is very
easy to maintain.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Tejun Heo <tj@kernel.org>

diff --git a/mm/percpu.c b/mm/percpu.c
index 49dfccf..c7206d0 100644 (file)
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -106,6 +106,7 @@ struct pcpu_chunk {
        int                     map_alloc;      /* # of map entries allocated */
        int                     *map;           /* allocation map */
        void                    *data;          /* chunk data */
+       int                     first_free;     /* no free below this */
        bool                    immutable;      /* no [de]population allowed */
        unsigned long           populated[];    /* populated bitmap */
 };
@@ -441,9 +442,10 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, int size, int align)
        int oslot = pcpu_chunk_slot(chunk);
        int max_contig = 0;
        int i, off;
+       bool seen_free = false;
        int *p;
 
-       for (i = 0, p = chunk->map; i < chunk->map_used; i++, p++) {
+       for (i = chunk->first_free, p = chunk->map + i; i < chunk->map_used; i++, p++) {
                int head, tail;
                int this_size;
 
@@ -456,6 +458,10 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, int size, int align)
 
                this_size = (p[1] & ~1) - off;
                if (this_size < head + size) {
+                       if (!seen_free) {
+                               chunk->first_free = i;
+                               seen_free = true;
+                       }
                        max_contig = max(this_size, max_contig);
                        continue;
                }
@@ -491,6 +497,10 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, int size, int align)
                        chunk->map_used += nr_extra;
 
                        if (head) {
+                               if (!seen_free) {
+                                       chunk->first_free = i;
+                                       seen_free = true;
+                               }
                                *++p = off += head;
                                ++i;
                                max_contig = max(head, max_contig);
@@ -501,6 +511,9 @@ static int pcpu_alloc_area(struct pcpu_chunk *chunk, int size, int align)
                        }
                }
 
+               if (!seen_free)
+                       chunk->first_free = i + 1;
+
                /* update hint and mark allocated */
                if (i + 1 == chunk->map_used)
                        chunk->contig_hint = max_contig; /* fully scanned */
@@ -558,6 +571,9 @@ static void pcpu_free_area(struct pcpu_chunk *chunk, int freeme)
        }
        BUG_ON(off != freeme);
 
+       if (i < chunk->first_free)
+               chunk->first_free = i;
+
        p = chunk->map + i;
        *p = off &= ~1;
        chunk->free_size += (p[1] & ~1) - off;