[PATCH RFC 0/1] mm, page_alloc: fix incorrect zone_statistics data

From: Jia He
Date: Mon Dec 12 2016 - 00:59:39 EST



In commit b9f00e147f27 ("mm, page_alloc: reduce branches in
zone_statistics"), it reconstructed the code to reduce the branch miss rate.
Compared with the original logic, it assumed if !(flag & __GFP_OTHER_NODE)
z->node would not be equal to preferred_zone->node. That seems to be
incorrect.

Here is what I catch, dumpstack() is triggered when z->node ==
preferred_zone->node and z->node != numa_node_id()

z=5,prefer=5,local=4, flag&OTHER_NODE=0
[c000000cdcaef440] [c0000000002e88cc] cache_grow_begin+0xcc/0x500
[c000000cdcaef6f0] [c0000000002ecb44] do_tune_cpucache+0x64/0x100
[c000000cdcaef750] [c0000000002ecc7c] enable_cpucache+0x9c/0x180
[c000000cdcaef7d0] [c0000000002ed01c] __kmem_cache_create+0x1ec/00x2c0
[c000000cdcaef820] [c000000000291c98] create_cache+0xb8/0x240
[c000000cdcaef890] [c000000000291fa8] kmem_cache_create+0x188/0x2290
[c000000cdcaef950] [d000000011dc5c70] ext4_mb_init+0x3c0/0x5e0 [eext4]
[c000000cdcaef9f0] [d000000011daaedc] ext4_fill_super+0x266c/0x33390 [ext4]
[c000000cdcaefb30] [c000000000328b8c] mount_bdev+0x22c/0x260
[c000000cdcaefbd0] [d000000011da1fa8] ext4_mount+0x48/0x60 [ext4]
[c000000cdcaefc10] [c00000000032a11c] mount_fs+0x8c/0x230
[c000000cdcaefcb0] [c000000000351f98] vfs_kern_mount+0x78/0x180
[c000000cdcaefd00] [c000000000356d68] do_mount+0x258/0xea0
[c000000cdcaefde0] [c000000000357da0] SyS_mount+0xa0/0x110
[c000000cdcaefe30] [c00000000000bd84] system_call+0x38/0xe0

Before this patch, the numa_miss and numa_foreign looked very odd:
linux:~ # numastat
node0 node1 node2 node3 node4 node5 node6
numa_hit 42216 0 0 0 96755 0 0
numa_miss 1 718 711 726 860 712 719
numa_foreign 1 718 711 726 860 712 719
interleave_hit 631 638 632 641 621 633 636
local_node 42216 0 0 0 96755 0 0
other_node 0 0 0 0 0 0 0

After this patch
linux:~ # numastat
node0 node1 node2 node3 node4 node5 node6
numa_hit 177891 718 711 726 60302 712 719
numa_miss 0 196944 237222 253424 0 36265 0
numa_foreign 723855 0 0 0 0 0 0
interleave_hit 631 638 632 641 621 633 636
local_node 177891 0 0 0 59444 0 0
other_node 0 718 711 726 858 712 719
Jia He (1):
mm, page_alloc: fix incorrect zone_statistics data

mm/page_alloc.c | 3 +++
1 file changed, 3 insertions(+)

--
2.5.5