diff options
Diffstat (limited to 'arch/x86/mm/numa_64.c')
| -rw-r--r-- | arch/x86/mm/numa_64.c | 506 |
1 files changed, 328 insertions, 178 deletions
diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c index 459913beac71..8948f47fde05 100644 --- a/arch/x86/mm/numa_64.c +++ b/arch/x86/mm/numa_64.c @@ -163,30 +163,48 @@ static void * __init early_node_mem(int nodeid, unsigned long start, unsigned long end, unsigned long size, unsigned long align) { - unsigned long mem = find_e820_area(start, end, size, align); - void *ptr; + unsigned long mem; + /* + * put it on high as possible + * something will go with NODE_DATA + */ + if (start < (MAX_DMA_PFN<<PAGE_SHIFT)) + start = MAX_DMA_PFN<<PAGE_SHIFT; + if (start < (MAX_DMA32_PFN<<PAGE_SHIFT) && + end > (MAX_DMA32_PFN<<PAGE_SHIFT)) + start = MAX_DMA32_PFN<<PAGE_SHIFT; + mem = find_e820_area(start, end, size, align); + if (mem != -1L) + return __va(mem); + + /* extend the search scope */ + end = max_pfn_mapped << PAGE_SHIFT; + if (end > (MAX_DMA32_PFN<<PAGE_SHIFT)) + start = MAX_DMA32_PFN<<PAGE_SHIFT; + else + start = MAX_DMA_PFN<<PAGE_SHIFT; + mem = find_e820_area(start, end, size, align); if (mem != -1L) return __va(mem); - ptr = __alloc_bootmem_nopanic(size, align, __pa(MAX_DMA_ADDRESS)); - if (ptr == NULL) { - printk(KERN_ERR "Cannot find %lu bytes in node %d\n", + printk(KERN_ERR "Cannot find %lu bytes in node %d\n", size, nodeid); - return NULL; - } - return ptr; + + return NULL; } /* Initialize bootmem allocator for a node */ void __init setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) { - unsigned long start_pfn, last_pfn, bootmap_pages, bootmap_size; + unsigned long start_pfn, last_pfn, nodedata_phys; const int pgdat_size = roundup(sizeof(pg_data_t), PAGE_SIZE); - unsigned long bootmap_start, nodedata_phys; - void *bootmap; int nid; +#ifndef CONFIG_NO_BOOTMEM + unsigned long bootmap_start, bootmap_pages, bootmap_size; + void *bootmap; +#endif if (!end) return; @@ -200,7 +218,7 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) start = roundup(start, ZONE_ALIGN); - printk(KERN_INFO "Bootmem setup node %d %016lx-%016lx\n", nodeid, + printk(KERN_INFO "Initmem setup node %d %016lx-%016lx\n", nodeid, start, end); start_pfn = start >> PAGE_SHIFT; @@ -211,14 +229,21 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) if (node_data[nodeid] == NULL) return; nodedata_phys = __pa(node_data[nodeid]); + reserve_early(nodedata_phys, nodedata_phys + pgdat_size, "NODE_DATA"); printk(KERN_INFO " NODE_DATA [%016lx - %016lx]\n", nodedata_phys, nodedata_phys + pgdat_size - 1); + nid = phys_to_nid(nodedata_phys); + if (nid != nodeid) + printk(KERN_INFO " NODE_DATA(%d) on node %d\n", nodeid, nid); memset(NODE_DATA(nodeid), 0, sizeof(pg_data_t)); - NODE_DATA(nodeid)->bdata = &bootmem_node_data[nodeid]; + NODE_DATA(nodeid)->node_id = nodeid; NODE_DATA(nodeid)->node_start_pfn = start_pfn; NODE_DATA(nodeid)->node_spanned_pages = last_pfn - start_pfn; +#ifndef CONFIG_NO_BOOTMEM + NODE_DATA(nodeid)->bdata = &bootmem_node_data[nodeid]; + /* * Find a place for the bootmem map * nodedata_phys could be on other nodes by alloc_bootmem, @@ -227,11 +252,7 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) * of alloc_bootmem, that could clash with reserved range */ bootmap_pages = bootmem_bootmap_pages(last_pfn - start_pfn); - nid = phys_to_nid(nodedata_phys); - if (nid == nodeid) - bootmap_start = roundup(nodedata_phys + pgdat_size, PAGE_SIZE); - else - bootmap_start = roundup(start, PAGE_SIZE); + bootmap_start = roundup(nodedata_phys + pgdat_size, PAGE_SIZE); /* * SMP_CACHE_BYTES could be enough, but init_bootmem_node like * to use that to align to PAGE_SIZE @@ -239,12 +260,13 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) bootmap = early_node_mem(nodeid, bootmap_start, end, bootmap_pages<<PAGE_SHIFT, PAGE_SIZE); if (bootmap == NULL) { - if (nodedata_phys < start || nodedata_phys >= end) - free_bootmem(nodedata_phys, pgdat_size); + free_early(nodedata_phys, nodedata_phys + pgdat_size); node_data[nodeid] = NULL; return; } bootmap_start = __pa(bootmap); + reserve_early(bootmap_start, bootmap_start+(bootmap_pages<<PAGE_SHIFT), + "BOOTMAP"); bootmap_size = init_bootmem_node(NODE_DATA(nodeid), bootmap_start >> PAGE_SHIFT, @@ -253,31 +275,12 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end) printk(KERN_INFO " bootmap [%016lx - %016lx] pages %lx\n", bootmap_start, bootmap_start + bootmap_size - 1, bootmap_pages); - - free_bootmem_with_active_regions(nodeid, end); - - /* - * convert early reserve to bootmem reserve earlier - * otherwise early_node_mem could use early reserved mem - * on previous node - */ - early_res_to_bootmem(start, end); - - /* - * in some case early_node_mem could use alloc_bootmem - * to get range on other node, don't reserve that again - */ - if (nid != nodeid) - printk(KERN_INFO " NODE_DATA(%d) on node %d\n", nodeid, nid); - else - reserve_bootmem_node(NODE_DATA(nodeid), nodedata_phys, - pgdat_size, BOOTMEM_DEFAULT); nid = phys_to_nid(bootmap_start); if (nid != nodeid) printk(KERN_INFO " bootmap(%d) on node %d\n", nodeid, nid); - else - reserve_bootmem_node(NODE_DATA(nodeid), bootmap_start, - bootmap_pages<<PAGE_SHIFT, BOOTMEM_DEFAULT); + + free_bootmem_with_active_regions(nodeid, end); +#endif node_set_online(nodeid); } @@ -306,8 +309,71 @@ void __init numa_init_array(void) #ifdef CONFIG_NUMA_EMU /* Numa emulation */ +static struct bootnode nodes[MAX_NUMNODES] __initdata; +static struct bootnode physnodes[MAX_NUMNODES] __initdata; static char *cmdline __initdata; +static int __init setup_physnodes(unsigned long start, unsigned long end, + int acpi, int k8) +{ + int nr_nodes = 0; + int ret = 0; + int i; + +#ifdef CONFIG_ACPI_NUMA + if (acpi) + nr_nodes = acpi_get_nodes(physnodes); +#endif +#ifdef CONFIG_K8_NUMA + if (k8) + nr_nodes = k8_get_nodes(physnodes); +#endif + /* + * Basic sanity checking on the physical node map: there may be errors + * if the SRAT or K8 incorrectly reported the topology or the mem= + * kernel parameter is used. + */ + for (i = 0; i < nr_nodes; i++) { + if (physnodes[i].start == physnodes[i].end) + continue; + if (physnodes[i].start > end) { + physnodes[i].end = physnodes[i].start; + continue; + } + if (physnodes[i].end < start) { + physnodes[i].start = physnodes[i].end; + continue; + } + if (physnodes[i].start < start) + physnodes[i].start = start; + if (physnodes[i].end > end) + physnodes[i].end = end; + } + + /* + * Remove all nodes that have no memory or were truncated because of the + * limited address range. + */ + for (i = 0; i < nr_nodes; i++) { + if (physnodes[i].start == physnodes[i].end) + continue; + physnodes[ret].start = physnodes[i].start; + physnodes[ret].end = physnodes[i].end; + ret++; + } + + /* + * If no physical topology was detected, a single node is faked to cover + * the entire address space. + */ + if (!ret) { + physnodes[ret].start = start; + physnodes[ret].end = end; + ret = 1; + } + return ret; +} + /* * Setups up nid to range from addr to addr + size. If the end * boundary is greater than max_addr, then max_addr is used instead. @@ -315,11 +381,9 @@ static char *cmdline __initdata; * allocation past addr and -1 otherwise. addr is adjusted to be at * the end of the node. */ -static int __init setup_node_range(int nid, struct bootnode *nodes, u64 *addr, - u64 size, u64 max_addr) +static int __init setup_node_range(int nid, u64 *addr, u64 size, u64 max_addr) { int ret = 0; - nodes[nid].start = *addr; *addr += size; if (*addr >= max_addr) { @@ -335,167 +399,234 @@ static int __init setup_node_range(int nid, struct bootnode *nodes, u64 *addr, } /* - * Splits num_nodes nodes up equally starting at node_start. The return value - * is the number of nodes split up and addr is adjusted to be at the end of the - * last node allocated. + * Sets up nr_nodes fake nodes interleaved over physical nodes ranging from addr + * to max_addr. The return value is the number of nodes allocated. */ -static int __init split_nodes_equally(struct bootnode *nodes, u64 *addr, - u64 max_addr, int node_start, - int num_nodes) +static int __init split_nodes_interleave(u64 addr, u64 max_addr, + int nr_phys_nodes, int nr_nodes) { - unsigned int big; + nodemask_t physnode_mask = NODE_MASK_NONE; u64 size; + int big; + int ret = 0; int i; - if (num_nodes <= 0) + if (nr_nodes <= 0) return -1; - if (num_nodes > MAX_NUMNODES) - num_nodes = MAX_NUMNODES; - size = (max_addr - *addr - e820_hole_size(*addr, max_addr)) / - num_nodes; + if (nr_nodes > MAX_NUMNODES) { + pr_info("numa=fake=%d too large, reducing to %d\n", + nr_nodes, MAX_NUMNODES); + nr_nodes = MAX_NUMNODES; + } + + size = (max_addr - addr - e820_hole_size(addr, max_addr)) / nr_nodes; /* * Calculate the number of big nodes that can be allocated as a result - * of consolidating the leftovers. + * of consolidating the remainder. */ - big = ((size & ~FAKE_NODE_MIN_HASH_MASK) * num_nodes) / - FAKE_NODE_MIN_SIZE; + big = ((size & ~FAKE_NODE_MIN_HASH_MASK) * nr_nodes) / + FAKE_NODE_MIN_SIZE; - /* Round down to nearest FAKE_NODE_MIN_SIZE. */ size &= FAKE_NODE_MIN_HASH_MASK; if (!size) { - printk(KERN_ERR "Not enough memory for each node. " - "NUMA emulation disabled.\n"); + pr_err("Not enough memory for each node. " + "NUMA emulation disabled.\n"); return -1; } - for (i = node_start; i < num_nodes + node_start; i++) { - u64 end = *addr + size; + for (i = 0; i < nr_phys_nodes; i++) + if (physnodes[i].start != physnodes[i].end) + node_set(i, physnode_mask); - if (i < big) - end += FAKE_NODE_MIN_SIZE; - /* - * The final node can have the remaining system RAM. Other - * nodes receive roughly the same amount of available pages. - */ - if (i == num_nodes + node_start - 1) - end = max_addr; - else - while (end - *addr - e820_hole_size(*addr, end) < - size) { + /* + * Continue to fill physical nodes with fake nodes until there is no + * memory left on any of them. + */ + while (nodes_weight(physnode_mask)) { + for_each_node_mask(i, physnode_mask) { + u64 end = physnodes[i].start + size; + u64 dma32_end = PFN_PHYS(MAX_DMA32_PFN); + + if (ret < big) + end += FAKE_NODE_MIN_SIZE; + + /* + * Continue to add memory to this fake node if its + * non-reserved memory is less than the per-node size. + */ + while (end - physnodes[i].start - + e820_hole_size(physnodes[i].start, end) < size) { end += FAKE_NODE_MIN_SIZE; - if (end > max_addr) { - end = max_addr; + if (end > physnodes[i].end) { + end = physnodes[i].end; break; } } - if (setup_node_range(i, nodes, addr, end - *addr, max_addr) < 0) - break; + + /* + * If there won't be at least FAKE_NODE_MIN_SIZE of + * non-reserved memory in ZONE_DMA32 for the next node, + * this one must extend to the boundary. + */ + if (end < dma32_end && dma32_end - end - + e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE) + end = dma32_end; + + /* + * If there won't be enough non-reserved memory for the + * next node, this one must extend to the end of the + * physical node. + */ + if (physnodes[i].end - end - + e820_hole_size(end, physnodes[i].end) < size) + end = physnodes[i].end; + + /* + * Avoid allocating more nodes than requested, which can + * happen as a result of rounding down each node's size + * to FAKE_NODE_MIN_SIZE. + */ + if (nodes_weight(physnode_mask) + ret >= nr_nodes) + end = physnodes[i].end; + + if (setup_node_range(ret++, &physnodes[i].start, + end - physnodes[i].start, + physnodes[i].end) < 0) + node_clear(i, physnode_mask); + } } - return i - node_start + 1; + return ret; } /* - * Splits the remaining system RAM into chunks of size. The remaining memory is - * always assigned to a final node and can be asymmetric. Returns the number of - * nodes split. + * Returns the end address of a node so that there is at least `size' amount of + * non-reserved memory or `max_addr' is reached. */ -static int __init split_nodes_by_size(struct bootnode *nodes, u64 *addr, - u64 max_addr, int node_start, u64 size) +static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size) { - int i = node_start; - size = (size << 20) & FAKE_NODE_MIN_HASH_MASK; - while (!setup_node_range(i++, nodes, addr, size, max_addr)) - ; - return i - node_start; + u64 end = start + size; + + while (end - start - e820_hole_size(start, end) < size) { + end += FAKE_NODE_MIN_SIZE; + if (end > max_addr) { + end = max_addr; + break; + } + } + return end; } /* - * Sets up the system RAM area from start_pfn to last_pfn according to the - * numa=fake command-line option. + * Sets up fake nodes of `size' interleaved over physical nodes ranging from + * `addr' to `max_addr'. The return value is the number of nodes allocated. */ -static struct bootnode nodes[MAX_NUMNODES] __initdata; - -static int __init numa_emulation(unsigned long start_pfn, unsigned long last_pfn) +static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size) { - u64 size, addr = start_pfn << PAGE_SHIFT; - u64 max_addr = last_pfn << PAGE_SHIFT; - int num_nodes = 0, num = 0, coeff_flag, coeff = -1, i; + nodemask_t physnode_mask = NODE_MASK_NONE; + u64 min_size; + int ret = 0; + int i; - memset(&nodes, 0, sizeof(nodes)); + if (!size) + return -1; /* - * If the numa=fake command-line is just a single number N, split the - * system RAM into N fake nodes. + * The limit on emulated nodes is MAX_NUMNODES, so the size per node is + * increased accordingly if the requested size is too small. This + * creates a uniform distribution of node sizes across the entire + * machine (but not necessarily over physical nodes). */ - if (!strchr(cmdline, '*') && !strchr(cmdline, ',')) { - long n = simple_strtol(cmdline, NULL, 0); - - num_nodes = split_nodes_equally(nodes, &addr, max_addr, 0, n); - if (num_nodes < 0) - return num_nodes; - goto out; + min_size = (max_addr - addr - e820_hole_size(addr, max_addr)) / + MAX_NUMNODES; + min_size = max(min_size, FAKE_NODE_MIN_SIZE); + if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size) + min_size = (min_size + FAKE_NODE_MIN_SIZE) & + FAKE_NODE_MIN_HASH_MASK; + if (size < min_size) { + pr_err("Fake node size %LuMB too small, increasing to %LuMB\n", + size >> 20, min_size >> 20); + size = min_size; } + size &= FAKE_NODE_MIN_HASH_MASK; - /* Parse the command line. */ - for (coeff_flag = 0; ; cmdline++) { - if (*cmdline && isdigit(*cmdline)) { - num = num * 10 + *cmdline - '0'; - continue; - } - if (*cmdline == '*') { - if (num > 0) - coeff = num; - coeff_flag = 1; - } - if (!*cmdline || *cmdline == ',') { - if (!coeff_flag) - coeff = 1; + for (i = 0; i < MAX_NUMNODES; i++) + if (physnodes[i].start != physnodes[i].end) + node_set(i, physnode_mask); + /* + * Fill physical nodes with fake nodes of size until there is no memory + * left on any of them. + */ + while (nodes_weight(physnode_mask)) { + for_each_node_mask(i, physnode_mask) { + u64 dma32_end = MAX_DMA32_PFN << PAGE_SHIFT; + u64 end; + + end = find_end_of_node(physnodes[i].start, + physnodes[i].end, size); /* - * Round down to the nearest FAKE_NODE_MIN_SIZE. - * Command-line coefficients are in megabytes. + * If there won't be at least FAKE_NODE_MIN_SIZE of + * non-reserved memory in ZONE_DMA32 for the next node, + * this one must extend to the boundary. */ - size = ((u64)num << 20) & FAKE_NODE_MIN_HASH_MASK; - if (size) - for (i = 0; i < coeff; i++, num_nodes++) - if (setup_node_range(num_nodes, nodes, - &addr, size, max_addr) < 0) - goto done; - if (!*cmdline) - break; - coeff_flag = 0; - coeff = -1; + if (end < dma32_end && dma32_end - end - + e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE) + end = dma32_end; + + /* + * If there won't be enough non-reserved memory for the + * next node, this one must extend to the end of the + * physical node. + */ + if (physnodes[i].end - end - + e820_hole_size(end, physnodes[i].end) < size) + end = physnodes[i].end; + + /* + * Setup the fake node that will be allocated as bootmem + * later. If setup_node_range() returns non-zero, there + * is no more memory available on this physical node. + */ + if (setup_node_range(ret++, &physnodes[i].start, + end - physnodes[i].start, + physnodes[i].end) < 0) + node_clear(i, physnode_mask); } - num = 0; } -done: - if (!num_nodes) - return -1; - /* Fill remainder of system RAM, if appropriate. */ - if (addr < max_addr) { - if (coeff_flag && coeff < 0) { - /* Split remaining nodes into num-sized chunks */ - num_nodes += split_nodes_by_size(nodes, &addr, max_addr, - num_nodes, num); - goto out; - } - switch (*(cmdline - 1)) { - case '*': - /* Split remaining nodes into coeff chunks */ - if (coeff <= 0) - break; - num_nodes += split_nodes_equally(nodes, &addr, max_addr, - num_nodes, coeff); - break; - case ',': - /* Do not allocate remaining system RAM */ - break; - default: - /* Give one final node */ - setup_node_range(num_nodes, nodes, &addr, - max_addr - addr, max_addr); - num_nodes++; - } + return ret; +} + +/* + * Sets up the system RAM area from start_pfn to last_pfn according to the + * numa=fake command-line option. + */ +static int __init numa_emulation(unsigned long start_pfn, + unsigned long last_pfn, int acpi, int k8) +{ + u64 addr = start_pfn << PAGE_SHIFT; + u64 max_addr = last_pfn << PAGE_SHIFT; + int num_phys_nodes; + int num_nodes; + int i; + + num_phys_nodes = setup_physnodes(addr, max_addr, acpi, k8); + /* + * If the numa=fake command-line contains a 'M' or 'G', it represents + * the fixed node size. Otherwise, if it is just a single number N, + * split the system RAM into N fake nodes. + */ + if (strchr(cmdline, 'M') || strchr(cmdline, 'G')) { + u64 size; + + size = memparse(cmdline, &cmdline); + num_nodes = split_nodes_size_interleave(addr, max_addr, size); + } else { + unsigned long n; + + n = simple_strtoul(cmdline, NULL, 0); + num_nodes = split_nodes_interleave(addr, max_addr, num_phys_nodes, n); } -out: + + if (num_nodes < 0) + return num_nodes; memnode_shift = compute_hash_shift(nodes, num_nodes, NULL); if (memnode_shift < 0) { memnode_shift = 0; @@ -505,14 +636,10 @@ out: } /* - * We need to vacate all active ranges that may have been registered by - * SRAT and set acpi_numa to -1 so that srat_disabled() always returns - * true. NUMA emulation has succeeded so we will not scan ACPI nodes. + * We need to vacate all active ranges that may have been registered for + * the e820 memory map. */ remove_all_active_ranges(); -#ifdef CONFIG_ACPI_NUMA - acpi_numa = -1; -#endif for_each_node_mask(i, node_possible_map) { e820_register_active_regions(i, nodes[i].start >> PAGE_SHIFT, nodes[i].end >> PAGE_SHIFT); @@ -524,7 +651,8 @@ out: } #endif /* CONFIG_NUMA_EMU */ -void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn) +void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn, + int acpi, int k8) { int i; @@ -532,23 +660,22 @@ void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn) nodes_clear(node_online_map); #ifdef CONFIG_NUMA_EMU - if (cmdline && !numa_emulation(start_pfn, last_pfn)) + if (cmdline && !numa_emulation(start_pfn, last_pfn, acpi, k8)) return; nodes_clear(node_possible_map); nodes_clear(node_online_map); #endif #ifdef CONFIG_ACPI_NUMA - if (!numa_off && !acpi_scan_nodes(start_pfn << PAGE_SHIFT, - last_pfn << PAGE_SHIFT)) + if (!numa_off && acpi && !acpi_scan_nodes(start_pfn << PAGE_SHIFT, + last_pfn << PAGE_SHIFT)) return; nodes_clear(node_possible_map); nodes_clear(node_online_map); #endif #ifdef CONFIG_K8_NUMA - if (!numa_off && !k8_scan_nodes(start_pfn<<PAGE_SHIFT, - last_pfn<<PAGE_SHIFT)) + if (!numa_off && k8 && !k8_scan_nodes()) return; nodes_clear(node_possible_map); nodes_clear(node_online_map); @@ -579,6 +706,10 @@ unsigned long __init numa_free_all_bootmem(void) for_each_online_node(i) pages += free_all_bootmem_node(NODE_DATA(i)); +#ifdef CONFIG_NO_BOOTMEM + pages += free_all_memory_core_early(MAX_NUMNODES); +#endif + return pages; } @@ -601,6 +732,25 @@ static __init int numa_setup(char *opt) early_param("numa", numa_setup); #ifdef CONFIG_NUMA + +static __init int find_near_online_node(int node) +{ + int n, val; + int min_val = INT_MAX; + int best_node = -1; + + for_each_online_node(n) { + val = node_distance(node, n); + + if (val < min_val) { + min_val = val; + best_node = n; + } + } + + return best_node; +} + /* * Setup early cpu_to_node. * @@ -632,7 +782,7 @@ void __init init_cpu_to_node(void) if (node == NUMA_NO_NODE) continue; if (!node_online(node)) - continue; + node = find_near_online_node(node); numa_set_node(cpu, node); } } |