Introduce a simple allocator for the NUMA remap space. This space is very scarce, used for structures which are best allocated node local. Issues: o alloc_remap takes a node_id where we might expect a pgdat which was intended to allow us to allocate the pgdat's using this mechanism; which we do not yet do. Probabally should have alloc_remap_node() and alloc_remap_nid() for this purpose. Matt's for_each_..._node() patches may have broken this. Go back and check those loops. Signed-off-by: Andy Whitcroft Signed-off-by: Dave Hansen --- memhotplug-dave/arch/i386/Kconfig | 5 +++ memhotplug-dave/arch/i386/mm/discontig.c | 43 ++++++++++++++++++------------ memhotplug-dave/include/asm-i386/mmzone.h | 1 memhotplug-dave/include/linux/bootmem.h | 9 ++++++ memhotplug-dave/mm/page_alloc.c | 8 ++++- 5 files changed, 49 insertions(+), 17 deletions(-) diff -puN arch/i386/Kconfig~B1-alloc_remap-i386 arch/i386/Kconfig --- memhotplug/arch/i386/Kconfig~B1-alloc_remap-i386 2005-01-04 13:49:29.000000000 -0800 +++ memhotplug-dave/arch/i386/Kconfig 2005-01-04 13:49:29.000000000 -0800 @@ -769,6 +769,11 @@ config HAVE_ARCH_BOOTMEM_NODE depends on NUMA default y +config HAVE_ARCH_ALLOC_REMAP + bool + depends on NUMA + default y + config HIGHPTE bool "Allocate 3rd-level pagetables from highmem" depends on HIGHMEM4G || HIGHMEM64G diff -puN arch/i386/mm/discontig.c~B1-alloc_remap-i386 arch/i386/mm/discontig.c --- memhotplug/arch/i386/mm/discontig.c~B1-alloc_remap-i386 2005-01-04 13:49:29.000000000 -0800 +++ memhotplug-dave/arch/i386/mm/discontig.c 2005-01-04 13:49:29.000000000 -0800 @@ -83,6 +83,9 @@ unsigned long node_remap_offset[MAX_NUMN void *node_remap_start_vaddr[MAX_NUMNODES]; void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags); +void *node_remap_end_vaddr[MAX_NUMNODES]; +void *node_remap_alloc_vaddr[MAX_NUMNODES]; + /* * FLAT - support for basic PC memory model with discontig enabled, essentially * a single node with all available processors in it with a flat @@ -137,6 +140,20 @@ static void __init allocate_pgdat(int ni } } +void *alloc_remap(int nid, unsigned long size) +{ + void *allocation = node_remap_alloc_vaddr[nid]; + + size = ALIGN(size, L1_CACHE_BYTES); + if (!allocation || (allocation + size) >= node_remap_end_vaddr[nid]) + return 0; + + node_remap_alloc_vaddr[nid] += size; + memset(allocation, 0, size); + + return allocation; +} + void __init remap_numa_kva(void) { void *vaddr; @@ -173,8 +190,8 @@ static unsigned long calculate_numa_rema printk("Reserving %ld pages of KVA for lmem_map of node %d\n", size, nid); node_remap_size[nid] = size; - reserve_pages += size; node_remap_offset[nid] = reserve_pages; + reserve_pages += size; printk("Shrinking node %d from %ld pages to %ld pages\n", nid, node_end_pfn[nid], node_end_pfn[nid] - size); node_end_pfn[nid] -= size; @@ -257,12 +274,18 @@ unsigned long __init setup_memory(void) (ulong) pfn_to_kaddr(max_low_pfn)); for_each_online_node(nid) { node_remap_start_vaddr[nid] = pfn_to_kaddr( - (highstart_pfn + reserve_pages) - node_remap_offset[nid]); + highstart_pfn + node_remap_offset[nid]); + /* Init the node remap allocator */ + node_remap_end_vaddr[nid] = node_remap_start_vaddr[nid] + + (node_remap_size[nid] * PAGE_SIZE); + node_remap_alloc_vaddr[nid] = node_remap_start_vaddr[nid] + + ALIGN(sizeof(pg_data_t), PAGE_SIZE); + allocate_pgdat(nid); printk ("node %d will remap to vaddr %08lx - %08lx\n", nid, (ulong) node_remap_start_vaddr[nid], - (ulong) pfn_to_kaddr(highstart_pfn + reserve_pages - - node_remap_offset[nid] + node_remap_size[nid])); + (ulong) pfn_to_kaddr(highstart_pfn + + node_remap_offset[nid] + node_remap_size[nid])); } printk("High memory starts at vaddr %08lx\n", (ulong) pfn_to_kaddr(highstart_pfn)); @@ -324,18 +347,6 @@ void __init zone_sizes_init(void) } } zholes_size = get_zholes_size(nid); - /* - * We let the lmem_map for node 0 be allocated from the - * normal bootmem allocator, but other nodes come from the - * remapped KVA area - mbligh - */ - if (nid) { - unsigned long lmem_map; - lmem_map = (unsigned long)node_remap_start_vaddr[nid]; - lmem_map += sizeof(pg_data_t) + PAGE_SIZE - 1; - lmem_map &= PAGE_MASK; - NODE_DATA(nid)->node_mem_map = (struct page *)lmem_map; - } free_area_init_node(nid, NODE_DATA(nid), zones_size, start, zholes_size); } diff -puN include/asm-i386/mmzone.h~B1-alloc_remap-i386 include/asm-i386/mmzone.h --- memhotplug/include/asm-i386/mmzone.h~B1-alloc_remap-i386 2005-01-04 13:49:29.000000000 -0800 +++ memhotplug-dave/include/asm-i386/mmzone.h 2005-01-04 13:49:29.000000000 -0800 @@ -16,6 +16,7 @@ #else /* summit or generic arch */ #include #endif + #else /* !CONFIG_NUMA */ #define get_memcfg_numa get_memcfg_numa_flat #define get_zholes_size(n) (0) diff -puN include/linux/bootmem.h~B1-alloc_remap-i386 include/linux/bootmem.h --- memhotplug/include/linux/bootmem.h~B1-alloc_remap-i386 2005-01-04 13:49:29.000000000 -0800 +++ memhotplug-dave/include/linux/bootmem.h 2005-01-04 13:49:29.000000000 -0800 @@ -68,6 +68,15 @@ extern void * __init __alloc_bootmem_nod __alloc_bootmem_node((pgdat), (x), PAGE_SIZE, 0) #endif /* !CONFIG_HAVE_ARCH_BOOTMEM_NODE */ +#ifdef CONFIG_HAVE_ARCH_ALLOC_REMAP +extern void *alloc_remap(int nid, unsigned long size); +#else +static inline void *alloc_remap(int nid, unsigned long size) +{ + return NULL; +} +#endif + extern unsigned long __initdata nr_kernel_pages; extern unsigned long __initdata nr_all_pages; diff -puN mm/page_alloc.c~B1-alloc_remap-i386 mm/page_alloc.c --- memhotplug/mm/page_alloc.c~B1-alloc_remap-i386 2005-01-04 13:49:29.000000000 -0800 +++ memhotplug-dave/mm/page_alloc.c 2005-01-04 13:49:29.000000000 -0800 @@ -1766,6 +1766,7 @@ static void __init free_area_init_core(s static void __init alloc_node_mem_map(struct pglist_data *pgdat) { unsigned long size; + void *map; /* * Make sure that the architecture hasn't already allocated @@ -1775,7 +1776,12 @@ static void __init alloc_node_mem_map(st return; size = (pgdat->node_spanned_pages + 1) * sizeof(struct page); - pgdat->node_mem_map = alloc_bootmem_node(pgdat, size); + + map = alloc_remap(pgdat->node_id, size); + if (!map) + map = alloc_bootmem_node(pgdat, size); + pgdat->node_mem_map = map; + #ifndef CONFIG_DISCONTIGMEM /* * With no DISCONTIG, the global mem_map is just set as node 0's _