numa.c 10.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 * NUMA support, based on the x86 implementation.
 *
 * Copyright (C) 2015 Cavium Inc.
 * Author: Ganapatrao Kulkarni <gkulkarni@cavium.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

K
Kefeng Wang 已提交
20 21
#define pr_fmt(fmt) "NUMA: " fmt

22
#include <linux/acpi.h>
23 24 25 26 27
#include <linux/bootmem.h>
#include <linux/memblock.h>
#include <linux/module.h>
#include <linux/of.h>

28
#include <asm/acpi.h>
29
#include <asm/sections.h>
30

31 32 33 34 35 36 37
struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
EXPORT_SYMBOL(node_data);
nodemask_t numa_nodes_parsed __initdata;
static int cpu_to_node_map[NR_CPUS] = { [0 ... NR_CPUS-1] = NUMA_NO_NODE };

static int numa_distance_cnt;
static u8 *numa_distance;
38
static bool numa_off;
39 40 41 42 43

static __init int numa_parse_early_param(char *opt)
{
	if (!opt)
		return -EINVAL;
K
Kefeng Wang 已提交
44
	if (!strncmp(opt, "off", 3))
45
		numa_off = true;
K
Kefeng Wang 已提交
46

47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110
	return 0;
}
early_param("numa", numa_parse_early_param);

cpumask_var_t node_to_cpumask_map[MAX_NUMNODES];
EXPORT_SYMBOL(node_to_cpumask_map);

#ifdef CONFIG_DEBUG_PER_CPU_MAPS

/*
 * Returns a pointer to the bitmask of CPUs on Node 'node'.
 */
const struct cpumask *cpumask_of_node(int node)
{
	if (WARN_ON(node >= nr_node_ids))
		return cpu_none_mask;

	if (WARN_ON(node_to_cpumask_map[node] == NULL))
		return cpu_online_mask;

	return node_to_cpumask_map[node];
}
EXPORT_SYMBOL(cpumask_of_node);

#endif

static void map_cpu_to_node(unsigned int cpu, int nid)
{
	set_cpu_numa_node(cpu, nid);
	if (nid >= 0)
		cpumask_set_cpu(cpu, node_to_cpumask_map[nid]);
}

void numa_clear_node(unsigned int cpu)
{
	int nid = cpu_to_node(cpu);

	if (nid >= 0)
		cpumask_clear_cpu(cpu, node_to_cpumask_map[nid]);
	set_cpu_numa_node(cpu, NUMA_NO_NODE);
}

/*
 * Allocate node_to_cpumask_map based on number of available nodes
 * Requires node_possible_map to be valid.
 *
 * Note: cpumask_of_node() is not valid until after this is done.
 * (Use CONFIG_DEBUG_PER_CPU_MAPS to check this.)
 */
static void __init setup_node_to_cpumask_map(void)
{
	int node;

	/* setup nr_node_ids if not done yet */
	if (nr_node_ids == MAX_NUMNODES)
		setup_nr_node_ids();

	/* allocate and clear the mapping */
	for (node = 0; node < nr_node_ids; node++) {
		alloc_bootmem_cpumask_var(&node_to_cpumask_map[node]);
		cpumask_clear(node_to_cpumask_map[node]);
	}

	/* cpumask_of_node() will now work */
K
Kefeng Wang 已提交
111
	pr_debug("Node to cpumask map for %d nodes\n", nr_node_ids);
112 113 114 115 116 117 118
}

/*
 *  Set the cpu to node and mem mapping
 */
void numa_store_cpu_info(unsigned int cpu)
{
119
	map_cpu_to_node(cpu, cpu_to_node_map[cpu]);
120 121 122 123 124
}

void __init early_map_cpu_to_node(unsigned int cpu, int nid)
{
	/* fallback to node 0 */
125
	if (nid < 0 || nid >= MAX_NUMNODES || numa_off)
126 127 128
		nid = 0;

	cpu_to_node_map[cpu] = nid;
129 130 131 132 133 134 135 136

	/*
	 * We should set the numa node of cpu0 as soon as possible, because it
	 * has already been set up online before. cpu_to_node(0) will soon be
	 * called.
	 */
	if (!cpu)
		set_cpu_numa_node(cpu, nid);
137 138
}

139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189
#ifdef CONFIG_HAVE_SETUP_PER_CPU_AREA
unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
EXPORT_SYMBOL(__per_cpu_offset);

static int __init early_cpu_to_node(int cpu)
{
	return cpu_to_node_map[cpu];
}

static int __init pcpu_cpu_distance(unsigned int from, unsigned int to)
{
	return node_distance(from, to);
}

static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size,
				       size_t align)
{
	int nid = early_cpu_to_node(cpu);

	return  memblock_virt_alloc_try_nid(size, align,
			__pa(MAX_DMA_ADDRESS), MEMBLOCK_ALLOC_ACCESSIBLE, nid);
}

static void __init pcpu_fc_free(void *ptr, size_t size)
{
	memblock_free_early(__pa(ptr), size);
}

void __init setup_per_cpu_areas(void)
{
	unsigned long delta;
	unsigned int cpu;
	int rc;

	/*
	 * Always reserve area for module percpu variables.  That's
	 * what the legacy allocator did.
	 */
	rc = pcpu_embed_first_chunk(PERCPU_MODULE_RESERVE,
				    PERCPU_DYNAMIC_RESERVE, PAGE_SIZE,
				    pcpu_cpu_distance,
				    pcpu_fc_alloc, pcpu_fc_free);
	if (rc < 0)
		panic("Failed to initialize percpu areas.");

	delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
	for_each_possible_cpu(cpu)
		__per_cpu_offset[cpu] = delta + pcpu_unit_offsets[cpu];
}
#endif

190 191 192 193
/**
 * numa_add_memblk - Set node id to memblk
 * @nid: NUMA node ID of the new memblk
 * @start: Start address of the new memblk
194
 * @end:  End address of the new memblk
195 196 197 198
 *
 * RETURNS:
 * 0 on success, -errno on failure.
 */
199
int __init numa_add_memblk(int nid, u64 start, u64 end)
200 201 202
{
	int ret;

203
	ret = memblock_set_node(start, (end - start), &memblock.memory, nid);
204
	if (ret < 0) {
K
Kefeng Wang 已提交
205
		pr_err("memblock [0x%llx - 0x%llx] failed to add on node %d\n",
206
			start, (end - 1), nid);
207 208 209 210
		return ret;
	}

	node_set(nid, numa_nodes_parsed);
K
Kefeng Wang 已提交
211
	pr_info("Adding memblock [0x%llx - 0x%llx] on node %d\n",
212
			start, (end - 1), nid);
213 214 215 216 217 218 219 220 221 222 223 224 225
	return ret;
}

/**
 * Initialize NODE_DATA for a node on the local memory
 */
static void __init setup_node_data(int nid, u64 start_pfn, u64 end_pfn)
{
	const size_t nd_size = roundup(sizeof(pg_data_t), SMP_CACHE_BYTES);
	u64 nd_pa;
	void *nd;
	int tnid;

K
Kefeng Wang 已提交
226 227
	pr_info("Initmem setup node %d [mem %#010Lx-%#010Lx]\n",
		nid, start_pfn << PAGE_SHIFT, (end_pfn << PAGE_SHIFT) - 1);
228 229 230 231 232

	nd_pa = memblock_alloc_try_nid(nd_size, SMP_CACHE_BYTES, nid);
	nd = __va(nd_pa);

	/* report and initialize */
K
Kefeng Wang 已提交
233
	pr_info("NODE_DATA [mem %#010Lx-%#010Lx]\n",
234 235 236
		nd_pa, nd_pa + nd_size - 1);
	tnid = early_pfn_to_nid(nd_pa >> PAGE_SHIFT);
	if (tnid != nid)
K
Kefeng Wang 已提交
237
		pr_info("NODE_DATA(%d) on node %d\n", nid, tnid);
238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293

	node_data[nid] = nd;
	memset(NODE_DATA(nid), 0, sizeof(pg_data_t));
	NODE_DATA(nid)->node_id = nid;
	NODE_DATA(nid)->node_start_pfn = start_pfn;
	NODE_DATA(nid)->node_spanned_pages = end_pfn - start_pfn;
}

/**
 * numa_free_distance
 *
 * The current table is freed.
 */
void __init numa_free_distance(void)
{
	size_t size;

	if (!numa_distance)
		return;

	size = numa_distance_cnt * numa_distance_cnt *
		sizeof(numa_distance[0]);

	memblock_free(__pa(numa_distance), size);
	numa_distance_cnt = 0;
	numa_distance = NULL;
}

/**
 *
 * Create a new NUMA distance table.
 *
 */
static int __init numa_alloc_distance(void)
{
	size_t size;
	u64 phys;
	int i, j;

	size = nr_node_ids * nr_node_ids * sizeof(numa_distance[0]);
	phys = memblock_find_in_range(0, PFN_PHYS(max_pfn),
				      size, PAGE_SIZE);
	if (WARN_ON(!phys))
		return -ENOMEM;

	memblock_reserve(phys, size);

	numa_distance = __va(phys);
	numa_distance_cnt = nr_node_ids;

	/* fill with the default distances */
	for (i = 0; i < numa_distance_cnt; i++)
		for (j = 0; j < numa_distance_cnt; j++)
			numa_distance[i * numa_distance_cnt + j] = i == j ?
				LOCAL_DISTANCE : REMOTE_DISTANCE;

K
Kefeng Wang 已提交
294
	pr_debug("Initialized distance table, cnt=%d\n", numa_distance_cnt);
295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314

	return 0;
}

/**
 * numa_set_distance - Set inter node NUMA distance from node to node.
 * @from: the 'from' node to set distance
 * @to: the 'to'  node to set distance
 * @distance: NUMA distance
 *
 * Set the distance from node @from to @to to @distance.
 * If distance table doesn't exist, a warning is printed.
 *
 * If @from or @to is higher than the highest known node or lower than zero
 * or @distance doesn't make sense, the call is ignored.
 *
 */
void __init numa_set_distance(int from, int to, int distance)
{
	if (!numa_distance) {
K
Kefeng Wang 已提交
315
		pr_warn_once("Warning: distance table not allocated yet\n");
316 317 318 319 320
		return;
	}

	if (from >= numa_distance_cnt || to >= numa_distance_cnt ||
			from < 0 || to < 0) {
K
Kefeng Wang 已提交
321
		pr_warn_once("Warning: node ids are out of bound, from=%d to=%d distance=%d\n",
322 323 324 325 326 327
			    from, to, distance);
		return;
	}

	if ((u8)distance != distance ||
	    (from == to && distance != LOCAL_DISTANCE)) {
K
Kefeng Wang 已提交
328
		pr_warn_once("Warning: invalid distance parameter, from=%d to=%d distance=%d\n",
329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
			     from, to, distance);
		return;
	}

	numa_distance[from * numa_distance_cnt + to] = distance;
}

/**
 * Return NUMA distance @from to @to
 */
int __node_distance(int from, int to)
{
	if (from >= numa_distance_cnt || to >= numa_distance_cnt)
		return from == to ? LOCAL_DISTANCE : REMOTE_DISTANCE;
	return numa_distance[from * numa_distance_cnt + to];
}
EXPORT_SYMBOL(__node_distance);

static int __init numa_register_nodes(void)
{
	int nid;
	struct memblock_region *mblk;

	/* Check that valid nid is set to memblks */
	for_each_memblock(memory, mblk)
		if (mblk->nid == NUMA_NO_NODE || mblk->nid >= MAX_NUMNODES) {
K
Kefeng Wang 已提交
355
			pr_warn("Warning: invalid memblk node %d [mem %#010Lx-%#010Lx]\n",
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
				mblk->nid, mblk->base,
				mblk->base + mblk->size - 1);
			return -EINVAL;
		}

	/* Finally register nodes. */
	for_each_node_mask(nid, numa_nodes_parsed) {
		unsigned long start_pfn, end_pfn;

		get_pfn_range_for_nid(nid, &start_pfn, &end_pfn);
		setup_node_data(nid, start_pfn, end_pfn);
		node_set_online(nid);
	}

	/* Setup online nodes to actual nodes*/
	node_possible_map = numa_nodes_parsed;

	return 0;
}

static int __init numa_init(int (*init_func)(void))
{
	int ret;

	nodes_clear(numa_nodes_parsed);
	nodes_clear(node_possible_map);
	nodes_clear(node_online_map);
	numa_free_distance();

	ret = numa_alloc_distance();
	if (ret < 0)
		return ret;

	ret = init_func();
	if (ret < 0)
		return ret;

393 394
	if (nodes_empty(numa_nodes_parsed)) {
		pr_info("No NUMA configuration found\n");
395
		return -EINVAL;
396
	}
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420

	ret = numa_register_nodes();
	if (ret < 0)
		return ret;

	setup_node_to_cpumask_map();

	return 0;
}

/**
 * dummy_numa_init - Fallback dummy NUMA init
 *
 * Used if there's no underlying NUMA architecture, NUMA initialization
 * fails, or NUMA is disabled on the command line.
 *
 * Must online at least one node (node 0) and add memory blocks that cover all
 * allowed memory. It is unlikely that this function fails.
 */
static int __init dummy_numa_init(void)
{
	int ret;
	struct memblock_region *mblk;

421 422
	if (numa_off)
		pr_info("NUMA disabled\n"); /* Forced off on command line. */
K
Kefeng Wang 已提交
423 424
	pr_info("Faking a node at [mem %#018Lx-%#018Lx]\n",
		0LLU, PFN_PHYS(max_pfn) - 1);
425 426

	for_each_memblock(memory, mblk) {
427
		ret = numa_add_memblk(0, mblk->base, mblk->base + mblk->size);
428 429 430 431 432 433 434
		if (!ret)
			continue;

		pr_err("NUMA init failed\n");
		return ret;
	}

435
	numa_off = true;
436 437 438 439 440 441 442 443 444 445 446 447
	return 0;
}

/**
 * arm64_numa_init - Initialize NUMA
 *
 * Try each configured NUMA initialization method until one succeeds.  The
 * last fallback is dummy single node config encomapssing whole memory.
 */
void __init arm64_numa_init(void)
{
	if (!numa_off) {
448 449 450
		if (!acpi_disabled && !numa_init(arm64_acpi_numa_init))
			return;
		if (acpi_disabled && !numa_init(of_numa_init))
451 452 453 454 455
			return;
	}

	numa_init(dummy_numa_init);
}