The patch titled x86,nobootmem: make alloc_bootmem_node fall back to other node when 32bit numa is used has been added to the -mm tree. Its filename is x86nobootmem-make-alloc_bootmem_node-fall-back-to-other-node-when-32bit-numa-is-used.patch Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's *** Remember to use Documentation/SubmitChecklist when testing your code *** See http://userweb.kernel.org/~akpm/stuff/added-to-mm.txt to find out what to do about this The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/ ------------------------------------------------------ Subject: x86,nobootmem: make alloc_bootmem_node fall back to other node when 32bit numa is used From: Yinghai Lu <yinghai@xxxxxxxxxx> Borislav Petkov reported his 32bit numa system has problem: [ 0.000000] Reserving total of 4c00 pages for numa KVA remap [ 0.000000] kva_start_pfn ~ 32800 max_low_pfn ~ 375fe [ 0.000000] max_pfn = 238000 [ 0.000000] 8202MB HIGHMEM available. [ 0.000000] 885MB LOWMEM available. [ 0.000000] mapped low ram: 0 - 375fe000 [ 0.000000] low ram: 0 - 375fe000 [ 0.000000] alloc (nid=8 100000 - 7ee00000) (1000000 - ffffffff) 1000 1000 => 34e7000 [ 0.000000] alloc (nid=8 100000 - 7ee00000) (1000000 - ffffffff) 200 40 => 34c9d80 [ 0.000000] alloc (nid=0 100000 - 7ee00000) (1000000 - ffffffffffffffff) 180 40 => 34e6140 [ 0.000000] alloc (nid=1 80000000 - c7e60000) (1000000 - ffffffffffffffff) 240 40 => 80000000 [ 0.000000] BUG: unable to handle kernel paging request at 40000000 [ 0.000000] IP: [<c2c8cff1>] __alloc_memory_core_early+0x147/0x1d6 [ 0.000000] *pdpt = 0000000000000000 *pde = f000ff53f000ff00 ... [ 0.000000] Call Trace: [ 0.000000] [<c2c8b4f8>] ? __alloc_bootmem_node+0x216/0x22f [ 0.000000] [<c2c90c9b>] ? sparse_early_usemaps_alloc_node+0x5a/0x10b [ 0.000000] [<c2c9149e>] ? sparse_init+0x1dc/0x499 [ 0.000000] [<c2c79118>] ? paging_init+0x168/0x1df [ 0.000000] [<c2c780ff>] ? native_pagetable_setup_start+0xef/0x1bb looks like it allocate much high address for bootmem. Try to cut limit with get_max_mapped() Reported-by: Borislav Petkov <borislav.petkov@xxxxxxx> Tested-by: Conny Seidel <conny.seidel@xxxxxxx> Signed-off-by: Yinghai Lu <yinghai@xxxxxxxxxx> Cc: <stable@xxxxxxxxxx> [2.6.34.x] Cc: Ingo Molnar <mingo@xxxxxxx> Cc: "H. Peter Anvin" <hpa@xxxxxxxxx> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx> Cc: Johannes Weiner <hannes@xxxxxxxxxxx> Cc: Lee Schermerhorn <lee.schermerhorn@xxxxxx> Cc: Mel Gorman <mel@xxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/bootmem.c | 24 ++++++++++++++++++++---- mm/page_alloc.c | 3 +++ 2 files changed, 23 insertions(+), 4 deletions(-) diff -puN mm/bootmem.c~x86nobootmem-make-alloc_bootmem_node-fall-back-to-other-node-when-32bit-numa-is-used mm/bootmem.c --- a/mm/bootmem.c~x86nobootmem-make-alloc_bootmem_node-fall-back-to-other-node-when-32bit-numa-is-used +++ a/mm/bootmem.c @@ -833,15 +833,24 @@ static void * __init ___alloc_bootmem_no void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, unsigned long align, unsigned long goal) { + void *ptr; + if (WARN_ON_ONCE(slab_is_available())) return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); #ifdef CONFIG_NO_BOOTMEM - return __alloc_memory_core_early(pgdat->node_id, size, align, + ptr = __alloc_memory_core_early(pgdat->node_id, size, align, + goal, -1ULL); + if (ptr) + return ptr; + + ptr = __alloc_memory_core_early(MAX_NUMNODES, size, align, goal, -1ULL); #else - return ___alloc_bootmem_node(pgdat->bdata, size, align, goal, 0); + ptr = ___alloc_bootmem_node(pgdat->bdata, size, align, goal, 0); #endif + + return ptr; } void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size, @@ -977,14 +986,21 @@ void * __init __alloc_bootmem_low(unsign void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size, unsigned long align, unsigned long goal) { + void *ptr; + if (WARN_ON_ONCE(slab_is_available())) return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); #ifdef CONFIG_NO_BOOTMEM - return __alloc_memory_core_early(pgdat->node_id, size, align, + ptr = __alloc_memory_core_early(pgdat->node_id, size, align, + goal, ARCH_LOW_ADDRESS_LIMIT); + if (ptr) + return ptr; + ptr = __alloc_memory_core_early(MAX_NUMNODES, size, align, goal, ARCH_LOW_ADDRESS_LIMIT); #else - return ___alloc_bootmem_node(pgdat->bdata, size, align, + ptr = ___alloc_bootmem_node(pgdat->bdata, size, align, goal, ARCH_LOW_ADDRESS_LIMIT); #endif + return ptr; } diff -puN mm/page_alloc.c~x86nobootmem-make-alloc_bootmem_node-fall-back-to-other-node-when-32bit-numa-is-used mm/page_alloc.c --- a/mm/page_alloc.c~x86nobootmem-make-alloc_bootmem_node-fall-back-to-other-node-when-32bit-numa-is-used +++ a/mm/page_alloc.c @@ -3634,6 +3634,9 @@ void * __init __alloc_memory_core_early( int i; void *ptr; + if (limit > get_max_mapped()) + limit = get_max_mapped(); + /* need to go over early_node_map to find out good range for node */ for_each_active_range_index_in_nid(i, nid) { u64 addr; _ Patches currently in -mm which might be from yinghai@xxxxxxxxxx are x86-numa-fix-boot-without-ram-on-node0-again.patch x86-numa-fix-boot-without-ram-on-node0-again-fix.patch linux-next.patch x86nobootmem-make-alloc_bootmem_node-fall-back-to-other-node-when-32bit-numa-is-used.patch kernel-range-remove-unused-definition-of-array_size.patch -- To unsubscribe from this list: send the line "unsubscribe mm-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html