[Xen-devel] [PATCH] track free pages live rather than count pages in all nodes/zones

[Dan Magenheimer <dan.magenheimer@xxxxxxxxxx>]

Trying to fix a livelock condition in tmem that occurs
only when the system is totally out of memory requires
the ability to easily determine if all zones in all
nodes are empty, and this must be checked at a fairly
high frequency.  So to avoid walking all the zones in
all the nodes each time, I'd like a fast way to determine
if "free_pages" is zero.  This patch tracks the sum
of the free pages in all nodes/zones.  Since I think
the value is modified only when heap_lock is held,
it need not be atomic.

I don't know this for sure, but suspect this will be
useful in other future memory utilization code, e.g.
page sharing.

This has had limited testing, though I did drive free
memory down to zero and up and down a few times with
debug on and no asserts were triggered.

On the chance that it looks good as is or is trivially modified:

Signed-off-by: Dan Magenheimer <dan.magenheimer@xxxxxxxxxx>

Thanks,
Dan

diff -r 0fb962a5dad3 xen/common/page_alloc.c
--- a/xen/common/page_alloc.c   Mon Dec 07 14:10:27 2009 +0000
+++ b/xen/common/page_alloc.c   Mon Dec 07 18:22:44 2009 -0700
@@ -222,6 +222,7 @@ static heap_by_zone_and_order_t *_heap[M
 #define heap(node, zone, order) ((*_heap[node])[zone][order])
 
 static unsigned long *avail[MAX_NUMNODES];
+static long total_avail_pages;
 
 static DEFINE_SPINLOCK(heap_lock);
 
@@ -350,6 +351,8 @@ static struct page_info *alloc_heap_page
 
     ASSERT(avail[node][zone] >= request);
     avail[node][zone] -= request;
+    total_avail_pages -= request;
+    ASSERT(total_avail_pages >= 0);
 
     spin_unlock(&heap_lock);
 
@@ -445,6 +448,8 @@ static int reserve_offlined_page(struct 
             continue;
 
         avail[node][zone]--;
+        total_avail_pages--;
+        ASSERT(total_avail_pages >= 0);
 
         page_list_add_tail(cur_head,
                            test_bit(_PGC_broken, &cur_head->count_info) ?
@@ -497,6 +502,7 @@ static void free_heap_pages(
     spin_lock(&heap_lock);
 
     avail[node][zone] += 1 << order;
+    total_avail_pages += 1 << order;
 
     /* Merge chunks as far as possible. */
     while ( order < MAX_ORDER )
@@ -816,6 +822,7 @@ static unsigned long avail_heap_pages(
 static unsigned long avail_heap_pages(
     unsigned int zone_lo, unsigned int zone_hi, unsigned int node)
 {
+#ifndef NDEBUG
     unsigned int i, zone;
     unsigned long free_pages = 0;
 
@@ -831,7 +838,16 @@ static unsigned long avail_heap_pages(
                 free_pages += avail[i][zone];
     }
 
+    ASSERT ( free_pages == total_avail_pages );
     return free_pages;
+#else
+    return total_avail_pages;
+#endif
+}
+
+unsigned long total_free_pages(void)
+{
+    return total_avail_pages;
 }
 
 void __init end_boot_allocator(void)
diff -r 0fb962a5dad3 xen/include/xen/mm.h
--- a/xen/include/xen/mm.h      Mon Dec 07 14:10:27 2009 +0000
+++ b/xen/include/xen/mm.h      Mon Dec 07 18:22:44 2009 -0700
@@ -62,6 +62,7 @@ unsigned int online_page(unsigned long m
 unsigned int online_page(unsigned long mfn, uint32_t *status);
 int offline_page(unsigned long mfn, int broken, uint32_t *status);
 int query_page_offline(unsigned long mfn, uint32_t *status);
+unsigned long total_free_pages(void);
 
 void scrub_heap_pages(void);

Attachment: freepages.patch
Description: Binary data

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel