projects / thirdparty / kernel / stable-queue.git / blame
| Commit | Line | Data |
|---|---|---|
| 37554d48 SL |
1 | From 051db87772d9867fc2406338717675f5e12d25c8 Mon Sep 17 00:00:00 2001 |
| 2 | From: Linxu Fang <fanglinxu@huawei.com> | |
| 3 | Date: Mon, 13 May 2019 17:19:17 -0700 | |
| 4 | Subject: mem-hotplug: fix node spanned pages when we have a node with only | |
| 5 | ZONE_MOVABLE | |
| 6 | ||
| 7 | [ Upstream commit 299c83dce9ea3a79bb4b5511d2cb996b6b8e5111 ] | |
| 8 | ||
| 9 | 342332e6a925 ("mm/page_alloc.c: introduce kernelcore=mirror option") and | |
| 10 | later patches rewrote the calculation of node spanned pages. | |
| 11 | ||
| 12 | e506b99696a2 ("mem-hotplug: fix node spanned pages when we have a movable | |
| 13 | node"), but the current code still has problems, | |
| 14 | ||
| 15 | When we have a node with only zone_movable and the node id is not zero, | |
| 16 | the size of node spanned pages is double added. | |
| 17 | ||
| 18 | That's because we have an empty normal zone, and zone_start_pfn or | |
| 19 | zone_end_pfn is not between arch_zone_lowest_possible_pfn and | |
| 20 | arch_zone_highest_possible_pfn, so we need to use clamp to constrain the | |
| 21 | range just like the commit <96e907d13602> (bootmem: Reimplement | |
| 22 | __absent_pages_in_range() using for_each_mem_pfn_range()). | |
| 23 | ||
| 24 | e.g. | |
| 25 | Zone ranges: | |
| 26 | DMA [mem 0x0000000000001000-0x0000000000ffffff] | |
| 27 | DMA32 [mem 0x0000000001000000-0x00000000ffffffff] | |
| 28 | Normal [mem 0x0000000100000000-0x000000023fffffff] | |
| 29 | Movable zone start for each node | |
| 30 | Node 0: 0x0000000100000000 | |
| 31 | Node 1: 0x0000000140000000 | |
| 32 | Early memory node ranges | |
| 33 | node 0: [mem 0x0000000000001000-0x000000000009efff] | |
| 34 | node 0: [mem 0x0000000000100000-0x00000000bffdffff] | |
| 35 | node 0: [mem 0x0000000100000000-0x000000013fffffff] | |
| 36 | node 1: [mem 0x0000000140000000-0x000000023fffffff] | |
| 37 | ||
| 38 | node 0 DMA spanned:0xfff present:0xf9e absent:0x61 | |
| 39 | node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020 | |
| 40 | node 0 Normal spanned:0 present:0 absent:0 | |
| 41 | node 0 Movable spanned:0x40000 present:0x40000 absent:0 | |
| 42 | On node 0 totalpages(node_present_pages): 1048446 | |
| 43 | node_spanned_pages:1310719 | |
| 44 | node 1 DMA spanned:0 present:0 absent:0 | |
| 45 | node 1 DMA32 spanned:0 present:0 absent:0 | |
| 46 | node 1 Normal spanned:0x100000 present:0x100000 absent:0 | |
| 47 | node 1 Movable spanned:0x100000 present:0x100000 absent:0 | |
| 48 | On node 1 totalpages(node_present_pages): 2097152 | |
| 49 | node_spanned_pages:2097152 | |
| 50 | Memory: 6967796K/12582392K available (16388K kernel code, 3686K rwdata, | |
| 51 | 4468K rodata, 2160K init, 10444K bss, 5614596K reserved, 0K | |
| 52 | cma-reserved) | |
| 53 | ||
| 54 | It shows that the current memory of node 1 is double added. | |
| 55 | After this patch, the problem is fixed. | |
| 56 | ||
| 57 | node 0 DMA spanned:0xfff present:0xf9e absent:0x61 | |
| 58 | node 0 DMA32 spanned:0xff000 present:0xbefe0 absent:0x40020 | |
| 59 | node 0 Normal spanned:0 present:0 absent:0 | |
| 60 | node 0 Movable spanned:0x40000 present:0x40000 absent:0 | |
| 61 | On node 0 totalpages(node_present_pages): 1048446 | |
| 62 | node_spanned_pages:1310719 | |
| 63 | node 1 DMA spanned:0 present:0 absent:0 | |
| 64 | node 1 DMA32 spanned:0 present:0 absent:0 | |
| 65 | node 1 Normal spanned:0 present:0 absent:0 | |
| 66 | node 1 Movable spanned:0x100000 present:0x100000 absent:0 | |
| 67 | On node 1 totalpages(node_present_pages): 1048576 | |
| 68 | node_spanned_pages:1048576 | |
| 69 | memory: 6967796K/8388088K available (16388K kernel code, 3686K rwdata, | |
| 70 | 4468K rodata, 2160K init, 10444K bss, 1420292K reserved, 0K | |
| 71 | cma-reserved) | |
| 72 | ||
| 73 | Link: http://lkml.kernel.org/r/1554178276-10372-1-git-send-email-fanglinxu@huawei.com | |
| 74 | Signed-off-by: Linxu Fang <fanglinxu@huawei.com> | |
| 75 | Cc: Taku Izumi <izumi.taku@jp.fujitsu.com> | |
| 76 | Cc: Xishi Qiu <qiuxishi@huawei.com> | |
| 77 | Cc: Michal Hocko <mhocko@suse.com> | |
| 78 | Cc: Vlastimil Babka <vbabka@suse.cz> | |
| 79 | Cc: Pavel Tatashin <pavel.tatashin@microsoft.com> | |
| 80 | Cc: Oscar Salvador <osalvador@suse.de> | |
| 81 | Signed-off-by: Andrew Morton <akpm@linux-foundation.org> | |
| 82 | Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> | |
| 83 | Signed-off-by: Sasha Levin <sashal@kernel.org> | |
| 84 | --- | |
| 85 | mm/page_alloc.c | 6 ++++-- | |
| 86 | 1 file changed, 4 insertions(+), 2 deletions(-) | |
| 87 | ||
| 88 | diff --git a/mm/page_alloc.c b/mm/page_alloc.c | |
| 89 | index 8e6932a140b8..2d04bd2e1ced 100644 | |
| 90 | --- a/mm/page_alloc.c | |
| 91 | +++ b/mm/page_alloc.c | |
| 92 | @@ -5937,13 +5937,15 @@ static unsigned long __meminit zone_spanned_pages_in_node(int nid, | |
| 93 | unsigned long *zone_end_pfn, | |
| 94 | unsigned long *ignored) | |
| 95 | { | |
| 96 | + unsigned long zone_low = arch_zone_lowest_possible_pfn[zone_type]; | |
| 97 | + unsigned long zone_high = arch_zone_highest_possible_pfn[zone_type]; | |
| 98 | /* When hotadd a new node from cpu_up(), the node should be empty */ | |
| 99 | if (!node_start_pfn && !node_end_pfn) | |
| 100 | return 0; | |
| 101 | ||
| 102 | /* Get the start and end of the zone */ | |
| 103 | - *zone_start_pfn = arch_zone_lowest_possible_pfn[zone_type]; | |
| 104 | - *zone_end_pfn = arch_zone_highest_possible_pfn[zone_type]; | |
| 105 | + *zone_start_pfn = clamp(node_start_pfn, zone_low, zone_high); | |
| 106 | + *zone_end_pfn = clamp(node_end_pfn, zone_low, zone_high); | |
| 107 | adjust_zone_range_for_zone_movable(nid, zone_type, | |
| 108 | node_start_pfn, node_end_pfn, | |
| 109 | zone_start_pfn, zone_end_pfn); | |
| 110 | -- | |
| 111 | 2.20.1 | |
| 112 |