x86, NUMA: Trim numa meminfo with max_pfn in a separate loop

During testing 32bit numa unifying code from tj, found one system with
more than 64g fails to use numa.  It turns out we do not trim numa
meminfo correctly against max_pfn in case start address of a node is
higher than 64GiB.  Bug fix made it to tip tree.

This patch moves the checking and trimming to a separate loop.  So we
don't need to compare low/high in following merge loops.  It makes the
code more readable.

Also it makes the node merge printouts less strange.  On a 512GiB numa
system with 32bit,

before:
> NUMA: Node 0 [0,a0000) + [100000,80000000) -> [0,80000000)
> NUMA: Node 0 [0,80000000) + [100000000,1080000000) -> [0,1000000000)

after:
> NUMA: Node 0 [0,a0000) + [100000,80000000) -> [0,80000000)
> NUMA: Node 0 [0,80000000) + [100000000,1000000000) -> [0,1000000000)

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
[Updated patch description and comment slightly.]
Signed-off-by: Tejun Heo <tj@kernel.org>

diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
index 9a0ed31..f5510d8 100644 (file)
--- a/arch/x86/mm/numa.c
+++ b/arch/x86/mm/numa.c
@@ -270,6 +270,7 @@ int __init numa_cleanup_meminfo(struct numa_meminfo *mi)
        const u64 high = PFN_PHYS(max_pfn);
        int i, j, k;
 
+       /* first, trim all entries */
        for (i = 0; i < mi->nr_blks; i++) {
                struct numa_memblk *bi = &mi->blk[i];
 
@@ -278,10 +279,13 @@ int __init numa_cleanup_meminfo(struct numa_meminfo *mi)
                bi->end = min(bi->end, high);
 
                /* and there's no empty block */
-               if (bi->start >= bi->end) {
+               if (bi->start >= bi->end)
                        numa_remove_memblk_from(i--, mi);
-                       continue;
-               }
+       }
+
+       /* merge neighboring / overlapping entries */
+       for (i = 0; i < mi->nr_blks; i++) {
+               struct numa_memblk *bi = &mi->blk[i];
 
                for (j = i + 1; j < mi->nr_blks; j++) {
                        struct numa_memblk *bj = &mi->blk[j];
@@ -311,8 +315,8 @@ int __init numa_cleanup_meminfo(struct numa_meminfo *mi)
                         */
                        if (bi->nid != bj->nid)
                                continue;
-                       start = max(min(bi->start, bj->start), low);
-                       end = min(max(bi->end, bj->end), high);
+                       start = min(bi->start, bj->start);
+                       end = max(bi->end, bj->end);
                        for (k = 0; k < mi->nr_blks; k++) {
                                struct numa_memblk *bk = &mi->blk[k];
 
@@ -332,6 +336,7 @@ int __init numa_cleanup_meminfo(struct numa_meminfo *mi)
                }
        }
 
+       /* clear unused ones */
        for (i = mi->nr_blks; i < ARRAY_SIZE(mi->blk); i++) {
                mi->blk[i].start = mi->blk[i].end = 0;
                mi->blk[i].nid = NUMA_NO_NODE;