percpu.c 63.4 KB
Newer Older
1 2 3 4 5 6 7 8 9 10
/*
 * linux/mm/percpu.c - percpu memory allocator
 *
 * Copyright (C) 2009		SUSE Linux Products GmbH
 * Copyright (C) 2009		Tejun Heo <tj@kernel.org>
 *
 * This file is released under the GPLv2.
 *
 * This is percpu allocator which can handle both static and dynamic
 * areas.  Percpu areas are allocated in chunks in vmalloc area.  Each
11 12 13 14 15 16 17
 * chunk is consisted of boot-time determined number of units and the
 * first chunk is used for static percpu variables in the kernel image
 * (special boot time alloc/init handling necessary as these areas
 * need to be brought up before allocation services are running).
 * Unit grows as necessary and all units grow or shrink in unison.
 * When a chunk is filled up, another chunk is allocated.  ie. in
 * vmalloc area
18 19 20 21 22 23 24 25
 *
 *  c0                           c1                         c2
 *  -------------------          -------------------        ------------
 * | u0 | u1 | u2 | u3 |        | u0 | u1 | u2 | u3 |      | u0 | u1 | u
 *  -------------------  ......  -------------------  ....  ------------
 *
 * Allocation is done in offset-size areas of single unit space.  Ie,
 * an area of 512 bytes at 6k in c1 occupies 512 bytes at 6k of c1:u0,
26 27 28 29
 * c1:u1, c1:u2 and c1:u3.  On UMA, units corresponds directly to
 * cpus.  On NUMA, the mapping can be non-linear and even sparse.
 * Percpu access can be done by configuring percpu base registers
 * according to cpu to unit mapping and pcpu_unit_size.
30
 *
31 32
 * There are usually many small percpu allocations many of them being
 * as small as 4 bytes.  The allocator organizes chunks into lists
33 34 35 36 37 38 39 40 41 42 43
 * according to free size and tries to allocate from the fullest one.
 * Each chunk keeps the maximum contiguous area size hint which is
 * guaranteed to be eqaul to or larger than the maximum contiguous
 * area in the chunk.  This helps the allocator not to iterate the
 * chunk maps unnecessarily.
 *
 * Allocation state in each chunk is kept using an array of integers
 * on chunk->map.  A positive value in the map represents a free
 * region and negative allocated.  Allocation inside a chunk is done
 * by scanning this map sequentially and serving the first matching
 * entry.  This is mostly copied from the percpu_modalloc() allocator.
44 45
 * Chunks can be determined from the address using the index field
 * in the page struct. The index field contains a pointer to the chunk.
46 47 48
 *
 * To use this allocator, arch code should do the followings.
 *
49
 * - drop CONFIG_HAVE_LEGACY_PER_CPU_AREA
50 51
 *
 * - define __addr_to_pcpu_ptr() and __pcpu_ptr_to_addr() to translate
52 53
 *   regular address to percpu pointer and back if they need to be
 *   different from the default
54
 *
55 56
 * - use pcpu_setup_first_chunk() during percpu area initialization to
 *   setup the first chunk containing the kernel static percpu area
57 58 59 60
 */

#include <linux/bitmap.h>
#include <linux/bootmem.h>
61
#include <linux/err.h>
62
#include <linux/list.h>
63
#include <linux/log2.h>
64 65 66 67 68 69
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/percpu.h>
#include <linux/pfn.h>
#include <linux/slab.h>
70
#include <linux/spinlock.h>
71
#include <linux/vmalloc.h>
72
#include <linux/workqueue.h>
73 74

#include <asm/cacheflush.h>
75
#include <asm/sections.h>
76 77 78 79 80
#include <asm/tlbflush.h>

#define PCPU_SLOT_BASE_SHIFT		5	/* 1-31 shares the same slot */
#define PCPU_DFL_MAP_ALLOC		16	/* start a map with 16 ents */

81 82 83 84 85 86 87 88 89 90 91 92
/* default addr <-> pcpu_ptr mapping, override in asm/percpu.h if necessary */
#ifndef __addr_to_pcpu_ptr
#define __addr_to_pcpu_ptr(addr)					\
	(void *)((unsigned long)(addr) - (unsigned long)pcpu_base_addr	\
		 + (unsigned long)__per_cpu_start)
#endif
#ifndef __pcpu_ptr_to_addr
#define __pcpu_ptr_to_addr(ptr)						\
	(void *)((unsigned long)(ptr) + (unsigned long)pcpu_base_addr	\
		 - (unsigned long)__per_cpu_start)
#endif

93 94 95 96 97 98 99 100
struct pcpu_chunk {
	struct list_head	list;		/* linked to pcpu_slot lists */
	int			free_size;	/* free bytes in the chunk */
	int			contig_hint;	/* max contiguous size hint */
	struct vm_struct	*vm;		/* mapped vmalloc region */
	int			map_used;	/* # of map entries used */
	int			map_alloc;	/* # of map entries allocated */
	int			*map;		/* allocation map */
101
	bool			immutable;	/* no [de]population allowed */
T
Tejun Heo 已提交
102
	unsigned long		populated[];	/* populated bitmap */
103 104
};

105 106
static int pcpu_unit_pages __read_mostly;
static int pcpu_unit_size __read_mostly;
107
static int pcpu_nr_units __read_mostly;
108 109 110
static int pcpu_chunk_size __read_mostly;
static int pcpu_nr_slots __read_mostly;
static size_t pcpu_chunk_struct_size __read_mostly;
111

112 113 114 115
/* cpus with the lowest and highest unit numbers */
static unsigned int pcpu_first_unit_cpu __read_mostly;
static unsigned int pcpu_last_unit_cpu __read_mostly;

116
/* the address of the first chunk which starts with the kernel static area */
117
void *pcpu_base_addr __read_mostly;
118 119
EXPORT_SYMBOL_GPL(pcpu_base_addr);

120 121 122
/* cpu -> unit map */
const int *pcpu_unit_map __read_mostly;

123 124 125 126 127 128 129 130 131 132 133 134 135 136
/*
 * The first chunk which always exists.  Note that unlike other
 * chunks, this one can be allocated and mapped in several different
 * ways and thus often doesn't live in the vmalloc area.
 */
static struct pcpu_chunk *pcpu_first_chunk;

/*
 * Optional reserved chunk.  This chunk reserves part of the first
 * chunk and serves it for reserved allocations.  The amount of
 * reserved offset is in pcpu_reserved_chunk_limit.  When reserved
 * area doesn't exist, the following variables contain NULL and 0
 * respectively.
 */
137 138 139
static struct pcpu_chunk *pcpu_reserved_chunk;
static int pcpu_reserved_chunk_limit;

140
/*
141 142 143
 * Synchronization rules.
 *
 * There are two locks - pcpu_alloc_mutex and pcpu_lock.  The former
T
Tejun Heo 已提交
144 145 146
 * protects allocation/reclaim paths, chunks, populated bitmap and
 * vmalloc mapping.  The latter is a spinlock and protects the index
 * data structures - chunk slots, chunks and area maps in chunks.
147 148 149 150 151 152 153 154 155 156 157 158 159
 *
 * During allocation, pcpu_alloc_mutex is kept locked all the time and
 * pcpu_lock is grabbed and released as necessary.  All actual memory
 * allocations are done using GFP_KERNEL with pcpu_lock released.
 *
 * Free path accesses and alters only the index data structures, so it
 * can be safely called from atomic context.  When memory needs to be
 * returned to the system, free path schedules reclaim_work which
 * grabs both pcpu_alloc_mutex and pcpu_lock, unlinks chunks to be
 * reclaimed, release both locks and frees the chunks.  Note that it's
 * necessary to grab both locks to remove a chunk from circulation as
 * allocation path might be referencing the chunk with only
 * pcpu_alloc_mutex locked.
160
 */
161 162
static DEFINE_MUTEX(pcpu_alloc_mutex);	/* protects whole alloc and reclaim */
static DEFINE_SPINLOCK(pcpu_lock);	/* protects index data structures */
163

164
static struct list_head *pcpu_slot __read_mostly; /* chunk list slots */
165

166 167 168 169
/* reclaim work to release fully free chunks, scheduled from free path */
static void pcpu_reclaim(struct work_struct *work);
static DECLARE_WORK(pcpu_reclaim_work, pcpu_reclaim);

170
static int __pcpu_size_to_slot(int size)
171
{
T
Tejun Heo 已提交
172
	int highbit = fls(size);	/* size is in bytes */
173 174 175
	return max(highbit - PCPU_SLOT_BASE_SHIFT + 2, 1);
}

176 177 178 179 180 181 182
static int pcpu_size_to_slot(int size)
{
	if (size == pcpu_unit_size)
		return pcpu_nr_slots - 1;
	return __pcpu_size_to_slot(size);
}

183 184 185 186 187 188 189 190 191 192
static int pcpu_chunk_slot(const struct pcpu_chunk *chunk)
{
	if (chunk->free_size < sizeof(int) || chunk->contig_hint < sizeof(int))
		return 0;

	return pcpu_size_to_slot(chunk->free_size);
}

static int pcpu_page_idx(unsigned int cpu, int page_idx)
{
193
	return pcpu_unit_map[cpu] * pcpu_unit_pages + page_idx;
194 195 196 197 198 199 200 201 202
}

static unsigned long pcpu_chunk_addr(struct pcpu_chunk *chunk,
				     unsigned int cpu, int page_idx)
{
	return (unsigned long)chunk->vm->addr +
		(pcpu_page_idx(cpu, page_idx) << PAGE_SHIFT);
}

T
Tejun Heo 已提交
203 204
static struct page *pcpu_chunk_page(struct pcpu_chunk *chunk,
				    unsigned int cpu, int page_idx)
205
{
T
Tejun Heo 已提交
206 207
	/* must not be used on pre-mapped chunk */
	WARN_ON(chunk->immutable);
208

T
Tejun Heo 已提交
209
	return vmalloc_to_page((void *)pcpu_chunk_addr(chunk, cpu, page_idx));
210 211
}

212 213 214 215 216 217 218 219 220 221 222 223
/* set the pointer to a chunk in a page struct */
static void pcpu_set_page_chunk(struct page *page, struct pcpu_chunk *pcpu)
{
	page->index = (unsigned long)pcpu;
}

/* obtain pointer to a chunk from a page struct */
static struct pcpu_chunk *pcpu_get_page_chunk(struct page *page)
{
	return (struct pcpu_chunk *)page->index;
}

T
Tejun Heo 已提交
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
static void pcpu_next_unpop(struct pcpu_chunk *chunk, int *rs, int *re, int end)
{
	*rs = find_next_zero_bit(chunk->populated, end, *rs);
	*re = find_next_bit(chunk->populated, end, *rs + 1);
}

static void pcpu_next_pop(struct pcpu_chunk *chunk, int *rs, int *re, int end)
{
	*rs = find_next_bit(chunk->populated, end, *rs);
	*re = find_next_zero_bit(chunk->populated, end, *rs + 1);
}

/*
 * (Un)populated page region iterators.  Iterate over (un)populated
 * page regions betwen @start and @end in @chunk.  @rs and @re should
 * be integer variables and will be set to start and end page index of
 * the current region.
 */
#define pcpu_for_each_unpop_region(chunk, rs, re, start, end)		    \
	for ((rs) = (start), pcpu_next_unpop((chunk), &(rs), &(re), (end)); \
	     (rs) < (re);						    \
	     (rs) = (re) + 1, pcpu_next_unpop((chunk), &(rs), &(re), (end)))

#define pcpu_for_each_pop_region(chunk, rs, re, start, end)		    \
	for ((rs) = (start), pcpu_next_pop((chunk), &(rs), &(re), (end));   \
	     (rs) < (re);						    \
	     (rs) = (re) + 1, pcpu_next_pop((chunk), &(rs), &(re), (end)))

252
/**
253 254
 * pcpu_mem_alloc - allocate memory
 * @size: bytes to allocate
255
 *
256 257 258
 * Allocate @size bytes.  If @size is smaller than PAGE_SIZE,
 * kzalloc() is used; otherwise, vmalloc() is used.  The returned
 * memory is always zeroed.
259
 *
260 261 262
 * CONTEXT:
 * Does GFP_KERNEL allocation.
 *
263
 * RETURNS:
264
 * Pointer to the allocated area on success, NULL on failure.
265
 */
266
static void *pcpu_mem_alloc(size_t size)
267
{
268 269 270 271 272 273 274 275 276
	if (size <= PAGE_SIZE)
		return kzalloc(size, GFP_KERNEL);
	else {
		void *ptr = vmalloc(size);
		if (ptr)
			memset(ptr, 0, size);
		return ptr;
	}
}
277

278 279 280 281 282 283 284 285 286
/**
 * pcpu_mem_free - free memory
 * @ptr: memory to free
 * @size: size of the area
 *
 * Free @ptr.  @ptr should have been allocated using pcpu_mem_alloc().
 */
static void pcpu_mem_free(void *ptr, size_t size)
{
287
	if (size <= PAGE_SIZE)
288
		kfree(ptr);
289
	else
290
		vfree(ptr);
291 292 293 294 295 296 297 298 299
}

/**
 * pcpu_chunk_relocate - put chunk in the appropriate chunk slot
 * @chunk: chunk of interest
 * @oslot: the previous slot it was on
 *
 * This function is called after an allocation or free changed @chunk.
 * New slot according to the changed state is determined and @chunk is
300 301
 * moved to the slot.  Note that the reserved chunk is never put on
 * chunk slots.
302 303 304
 *
 * CONTEXT:
 * pcpu_lock.
305 306 307 308 309
 */
static void pcpu_chunk_relocate(struct pcpu_chunk *chunk, int oslot)
{
	int nslot = pcpu_chunk_slot(chunk);

310
	if (chunk != pcpu_reserved_chunk && oslot != nslot) {
311 312 313 314 315 316 317 318
		if (oslot < nslot)
			list_move(&chunk->list, &pcpu_slot[nslot]);
		else
			list_move_tail(&chunk->list, &pcpu_slot[nslot]);
	}
}

/**
319 320
 * pcpu_chunk_addr_search - determine chunk containing specified address
 * @addr: address for which the chunk needs to be determined.
321
 *
322 323 324 325 326
 * RETURNS:
 * The address of the found chunk.
 */
static struct pcpu_chunk *pcpu_chunk_addr_search(void *addr)
{
327
	void *first_start = pcpu_first_chunk->vm->addr;
328

329
	/* is it in the first chunk? */
330
	if (addr >= first_start && addr < first_start + pcpu_unit_size) {
331 332
		/* is it in the reserved area? */
		if (addr < first_start + pcpu_reserved_chunk_limit)
333
			return pcpu_reserved_chunk;
334
		return pcpu_first_chunk;
335 336
	}

337 338 339 340 341 342 343 344
	/*
	 * The address is relative to unit0 which might be unused and
	 * thus unmapped.  Offset the address to the unit space of the
	 * current processor before looking it up in the vmalloc
	 * space.  Note that any possible cpu id can be used here, so
	 * there's no need to worry about preemption or cpu hotplug.
	 */
	addr += pcpu_unit_map[smp_processor_id()] * pcpu_unit_size;
345
	return pcpu_get_page_chunk(vmalloc_to_page(addr));
346 347
}

348 349 350 351 352 353 354 355
/**
 * pcpu_extend_area_map - extend area map for allocation
 * @chunk: target chunk
 *
 * Extend area map of @chunk so that it can accomodate an allocation.
 * A single allocation can split an area into three areas, so this
 * function makes sure that @chunk->map has at least two extra slots.
 *
356 357 358 359
 * CONTEXT:
 * pcpu_alloc_mutex, pcpu_lock.  pcpu_lock is released and reacquired
 * if area map is extended.
 *
360 361 362 363 364 365 366 367 368 369 370 371 372
 * RETURNS:
 * 0 if noop, 1 if successfully extended, -errno on failure.
 */
static int pcpu_extend_area_map(struct pcpu_chunk *chunk)
{
	int new_alloc;
	int *new;
	size_t size;

	/* has enough? */
	if (chunk->map_alloc >= chunk->map_used + 2)
		return 0;

373 374
	spin_unlock_irq(&pcpu_lock);

375 376 377 378 379
	new_alloc = PCPU_DFL_MAP_ALLOC;
	while (new_alloc < chunk->map_used + 2)
		new_alloc *= 2;

	new = pcpu_mem_alloc(new_alloc * sizeof(new[0]));
380 381
	if (!new) {
		spin_lock_irq(&pcpu_lock);
382
		return -ENOMEM;
383 384 385 386 387 388 389 390 391
	}

	/*
	 * Acquire pcpu_lock and switch to new area map.  Only free
	 * could have happened inbetween, so map_used couldn't have
	 * grown.
	 */
	spin_lock_irq(&pcpu_lock);
	BUG_ON(new_alloc < chunk->map_used + 2);
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407

	size = chunk->map_alloc * sizeof(chunk->map[0]);
	memcpy(new, chunk->map, size);

	/*
	 * map_alloc < PCPU_DFL_MAP_ALLOC indicates that the chunk is
	 * one of the first chunks and still using static map.
	 */
	if (chunk->map_alloc >= PCPU_DFL_MAP_ALLOC)
		pcpu_mem_free(chunk->map, size);

	chunk->map_alloc = new_alloc;
	chunk->map = new;
	return 0;
}

408 409 410 411
/**
 * pcpu_split_block - split a map block
 * @chunk: chunk of interest
 * @i: index of map block to split
T
Tejun Heo 已提交
412 413
 * @head: head size in bytes (can be 0)
 * @tail: tail size in bytes (can be 0)
414 415 416 417 418 419 420 421 422
 *
 * Split the @i'th map block into two or three blocks.  If @head is
 * non-zero, @head bytes block is inserted before block @i moving it
 * to @i+1 and reducing its size by @head bytes.
 *
 * If @tail is non-zero, the target block, which can be @i or @i+1
 * depending on @head, is reduced by @tail bytes and @tail byte block
 * is inserted after the target block.
 *
423
 * @chunk->map must have enough free slots to accomodate the split.
424 425 426
 *
 * CONTEXT:
 * pcpu_lock.
427
 */
428 429
static void pcpu_split_block(struct pcpu_chunk *chunk, int i,
			     int head, int tail)
430 431
{
	int nr_extra = !!head + !!tail;
432

433
	BUG_ON(chunk->map_alloc < chunk->map_used + nr_extra);
434

435
	/* insert new subblocks */
436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452
	memmove(&chunk->map[i + nr_extra], &chunk->map[i],
		sizeof(chunk->map[0]) * (chunk->map_used - i));
	chunk->map_used += nr_extra;

	if (head) {
		chunk->map[i + 1] = chunk->map[i] - head;
		chunk->map[i++] = head;
	}
	if (tail) {
		chunk->map[i++] -= tail;
		chunk->map[i] = tail;
	}
}

/**
 * pcpu_alloc_area - allocate area from a pcpu_chunk
 * @chunk: chunk of interest
T
Tejun Heo 已提交
453
 * @size: wanted size in bytes
454 455 456 457 458 459
 * @align: wanted align
 *
 * Try to allocate @size bytes area aligned at @align from @chunk.
 * Note that this function only allocates the offset.  It doesn't
 * populate or map the area.
 *
460 461
 * @chunk->map must have at least two free slots.
 *
462 463 464
 * CONTEXT:
 * pcpu_lock.
 *
465
 * RETURNS:
466 467
 * Allocated offset in @chunk on success, -1 if no matching area is
 * found.
468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
 */
static int pcpu_alloc_area(struct pcpu_chunk *chunk, int size, int align)
{
	int oslot = pcpu_chunk_slot(chunk);
	int max_contig = 0;
	int i, off;

	for (i = 0, off = 0; i < chunk->map_used; off += abs(chunk->map[i++])) {
		bool is_last = i + 1 == chunk->map_used;
		int head, tail;

		/* extra for alignment requirement */
		head = ALIGN(off, align) - off;
		BUG_ON(i == 0 && head != 0);

		if (chunk->map[i] < 0)
			continue;
		if (chunk->map[i] < head + size) {
			max_contig = max(chunk->map[i], max_contig);
			continue;
		}

		/*
		 * If head is small or the previous block is free,
		 * merge'em.  Note that 'small' is defined as smaller
		 * than sizeof(int), which is very small but isn't too
		 * uncommon for percpu allocations.
		 */
		if (head && (head < sizeof(int) || chunk->map[i - 1] > 0)) {
			if (chunk->map[i - 1] > 0)
				chunk->map[i - 1] += head;
			else {
				chunk->map[i - 1] -= head;
				chunk->free_size -= head;
			}
			chunk->map[i] -= head;
			off += head;
			head = 0;
		}

		/* if tail is small, just keep it around */
		tail = chunk->map[i] - head - size;
		if (tail < sizeof(int))
			tail = 0;

		/* split if warranted */
		if (head || tail) {
515
			pcpu_split_block(chunk, i, head, tail);
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
			if (head) {
				i++;
				off += head;
				max_contig = max(chunk->map[i - 1], max_contig);
			}
			if (tail)
				max_contig = max(chunk->map[i + 1], max_contig);
		}

		/* update hint and mark allocated */
		if (is_last)
			chunk->contig_hint = max_contig; /* fully scanned */
		else
			chunk->contig_hint = max(chunk->contig_hint,
						 max_contig);

		chunk->free_size -= chunk->map[i];
		chunk->map[i] = -chunk->map[i];

		pcpu_chunk_relocate(chunk, oslot);
		return off;
	}

	chunk->contig_hint = max_contig;	/* fully scanned */
	pcpu_chunk_relocate(chunk, oslot);

542 543
	/* tell the upper layer that this chunk has no matching area */
	return -1;
544 545 546 547 548 549 550 551 552 553
}

/**
 * pcpu_free_area - free area to a pcpu_chunk
 * @chunk: chunk of interest
 * @freeme: offset of area to free
 *
 * Free area starting from @freeme to @chunk.  Note that this function
 * only modifies the allocation map.  It doesn't depopulate or unmap
 * the area.
554 555 556
 *
 * CONTEXT:
 * pcpu_lock.
557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592
 */
static void pcpu_free_area(struct pcpu_chunk *chunk, int freeme)
{
	int oslot = pcpu_chunk_slot(chunk);
	int i, off;

	for (i = 0, off = 0; i < chunk->map_used; off += abs(chunk->map[i++]))
		if (off == freeme)
			break;
	BUG_ON(off != freeme);
	BUG_ON(chunk->map[i] > 0);

	chunk->map[i] = -chunk->map[i];
	chunk->free_size += chunk->map[i];

	/* merge with previous? */
	if (i > 0 && chunk->map[i - 1] >= 0) {
		chunk->map[i - 1] += chunk->map[i];
		chunk->map_used--;
		memmove(&chunk->map[i], &chunk->map[i + 1],
			(chunk->map_used - i) * sizeof(chunk->map[0]));
		i--;
	}
	/* merge with next? */
	if (i + 1 < chunk->map_used && chunk->map[i + 1] >= 0) {
		chunk->map[i] += chunk->map[i + 1];
		chunk->map_used--;
		memmove(&chunk->map[i + 1], &chunk->map[i + 2],
			(chunk->map_used - (i + 1)) * sizeof(chunk->map[0]));
	}

	chunk->contig_hint = max(chunk->map[i], chunk->contig_hint);
	pcpu_chunk_relocate(chunk, oslot);
}

/**
T
Tejun Heo 已提交
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616
 * pcpu_get_pages_and_bitmap - get temp pages array and bitmap
 * @chunk: chunk of interest
 * @bitmapp: output parameter for bitmap
 * @may_alloc: may allocate the array
 *
 * Returns pointer to array of pointers to struct page and bitmap,
 * both of which can be indexed with pcpu_page_idx().  The returned
 * array is cleared to zero and *@bitmapp is copied from
 * @chunk->populated.  Note that there is only one array and bitmap
 * and access exclusion is the caller's responsibility.
 *
 * CONTEXT:
 * pcpu_alloc_mutex and does GFP_KERNEL allocation if @may_alloc.
 * Otherwise, don't care.
 *
 * RETURNS:
 * Pointer to temp pages array on success, NULL on failure.
 */
static struct page **pcpu_get_pages_and_bitmap(struct pcpu_chunk *chunk,
					       unsigned long **bitmapp,
					       bool may_alloc)
{
	static struct page **pages;
	static unsigned long *bitmap;
617
	size_t pages_size = pcpu_nr_units * pcpu_unit_pages * sizeof(pages[0]);
T
Tejun Heo 已提交
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714
	size_t bitmap_size = BITS_TO_LONGS(pcpu_unit_pages) *
			     sizeof(unsigned long);

	if (!pages || !bitmap) {
		if (may_alloc && !pages)
			pages = pcpu_mem_alloc(pages_size);
		if (may_alloc && !bitmap)
			bitmap = pcpu_mem_alloc(bitmap_size);
		if (!pages || !bitmap)
			return NULL;
	}

	memset(pages, 0, pages_size);
	bitmap_copy(bitmap, chunk->populated, pcpu_unit_pages);

	*bitmapp = bitmap;
	return pages;
}

/**
 * pcpu_free_pages - free pages which were allocated for @chunk
 * @chunk: chunk pages were allocated for
 * @pages: array of pages to be freed, indexed by pcpu_page_idx()
 * @populated: populated bitmap
 * @page_start: page index of the first page to be freed
 * @page_end: page index of the last page to be freed + 1
 *
 * Free pages [@page_start and @page_end) in @pages for all units.
 * The pages were allocated for @chunk.
 */
static void pcpu_free_pages(struct pcpu_chunk *chunk,
			    struct page **pages, unsigned long *populated,
			    int page_start, int page_end)
{
	unsigned int cpu;
	int i;

	for_each_possible_cpu(cpu) {
		for (i = page_start; i < page_end; i++) {
			struct page *page = pages[pcpu_page_idx(cpu, i)];

			if (page)
				__free_page(page);
		}
	}
}

/**
 * pcpu_alloc_pages - allocates pages for @chunk
 * @chunk: target chunk
 * @pages: array to put the allocated pages into, indexed by pcpu_page_idx()
 * @populated: populated bitmap
 * @page_start: page index of the first page to be allocated
 * @page_end: page index of the last page to be allocated + 1
 *
 * Allocate pages [@page_start,@page_end) into @pages for all units.
 * The allocation is for @chunk.  Percpu core doesn't care about the
 * content of @pages and will pass it verbatim to pcpu_map_pages().
 */
static int pcpu_alloc_pages(struct pcpu_chunk *chunk,
			    struct page **pages, unsigned long *populated,
			    int page_start, int page_end)
{
	const gfp_t gfp = GFP_KERNEL | __GFP_HIGHMEM | __GFP_COLD;
	unsigned int cpu;
	int i;

	for_each_possible_cpu(cpu) {
		for (i = page_start; i < page_end; i++) {
			struct page **pagep = &pages[pcpu_page_idx(cpu, i)];

			*pagep = alloc_pages_node(cpu_to_node(cpu), gfp, 0);
			if (!*pagep) {
				pcpu_free_pages(chunk, pages, populated,
						page_start, page_end);
				return -ENOMEM;
			}
		}
	}
	return 0;
}

/**
 * pcpu_pre_unmap_flush - flush cache prior to unmapping
 * @chunk: chunk the regions to be flushed belongs to
 * @page_start: page index of the first page to be flushed
 * @page_end: page index of the last page to be flushed + 1
 *
 * Pages in [@page_start,@page_end) of @chunk are about to be
 * unmapped.  Flush cache.  As each flushing trial can be very
 * expensive, issue flush on the whole region at once rather than
 * doing it for each cpu.  This could be an overkill but is more
 * scalable.
 */
static void pcpu_pre_unmap_flush(struct pcpu_chunk *chunk,
				 int page_start, int page_end)
{
715 716 717
	flush_cache_vunmap(
		pcpu_chunk_addr(chunk, pcpu_first_unit_cpu, page_start),
		pcpu_chunk_addr(chunk, pcpu_last_unit_cpu, page_end));
T
Tejun Heo 已提交
718 719 720 721 722 723 724 725 726
}

static void __pcpu_unmap_pages(unsigned long addr, int nr_pages)
{
	unmap_kernel_range_noflush(addr, nr_pages << PAGE_SHIFT);
}

/**
 * pcpu_unmap_pages - unmap pages out of a pcpu_chunk
727
 * @chunk: chunk of interest
T
Tejun Heo 已提交
728 729
 * @pages: pages array which can be used to pass information to free
 * @populated: populated bitmap
730 731 732 733
 * @page_start: page index of the first page to unmap
 * @page_end: page index of the last page to unmap + 1
 *
 * For each cpu, unmap pages [@page_start,@page_end) out of @chunk.
T
Tejun Heo 已提交
734 735 736 737
 * Corresponding elements in @pages were cleared by the caller and can
 * be used to carry information to pcpu_free_pages() which will be
 * called after all unmaps are finished.  The caller should call
 * proper pre/post flush functions.
738
 */
T
Tejun Heo 已提交
739 740 741
static void pcpu_unmap_pages(struct pcpu_chunk *chunk,
			     struct page **pages, unsigned long *populated,
			     int page_start, int page_end)
742 743
{
	unsigned int cpu;
T
Tejun Heo 已提交
744
	int i;
745

T
Tejun Heo 已提交
746 747 748
	for_each_possible_cpu(cpu) {
		for (i = page_start; i < page_end; i++) {
			struct page *page;
749

T
Tejun Heo 已提交
750 751 752 753 754 755 756
			page = pcpu_chunk_page(chunk, cpu, i);
			WARN_ON(!page);
			pages[pcpu_page_idx(cpu, i)] = page;
		}
		__pcpu_unmap_pages(pcpu_chunk_addr(chunk, cpu, page_start),
				   page_end - page_start);
	}
757

T
Tejun Heo 已提交
758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777
	for (i = page_start; i < page_end; i++)
		__clear_bit(i, populated);
}

/**
 * pcpu_post_unmap_tlb_flush - flush TLB after unmapping
 * @chunk: pcpu_chunk the regions to be flushed belong to
 * @page_start: page index of the first page to be flushed
 * @page_end: page index of the last page to be flushed + 1
 *
 * Pages [@page_start,@page_end) of @chunk have been unmapped.  Flush
 * TLB for the regions.  This can be skipped if the area is to be
 * returned to vmalloc as vmalloc will handle TLB flushing lazily.
 *
 * As with pcpu_pre_unmap_flush(), TLB flushing also is done at once
 * for the whole region.
 */
static void pcpu_post_unmap_tlb_flush(struct pcpu_chunk *chunk,
				      int page_start, int page_end)
{
778 779 780
	flush_tlb_kernel_range(
		pcpu_chunk_addr(chunk, pcpu_first_unit_cpu, page_start),
		pcpu_chunk_addr(chunk, pcpu_last_unit_cpu, page_end));
781 782
}

783 784 785 786 787 788 789 790
static int __pcpu_map_pages(unsigned long addr, struct page **pages,
			    int nr_pages)
{
	return map_kernel_range_noflush(addr, nr_pages << PAGE_SHIFT,
					PAGE_KERNEL, pages);
}

/**
T
Tejun Heo 已提交
791
 * pcpu_map_pages - map pages into a pcpu_chunk
792
 * @chunk: chunk of interest
T
Tejun Heo 已提交
793 794
 * @pages: pages array containing pages to be mapped
 * @populated: populated bitmap
795 796 797
 * @page_start: page index of the first page to map
 * @page_end: page index of the last page to map + 1
 *
T
Tejun Heo 已提交
798 799 800 801 802 803 804
 * For each cpu, map pages [@page_start,@page_end) into @chunk.  The
 * caller is responsible for calling pcpu_post_map_flush() after all
 * mappings are complete.
 *
 * This function is responsible for setting corresponding bits in
 * @chunk->populated bitmap and whatever is necessary for reverse
 * lookup (addr -> chunk).
805
 */
T
Tejun Heo 已提交
806 807 808
static int pcpu_map_pages(struct pcpu_chunk *chunk,
			  struct page **pages, unsigned long *populated,
			  int page_start, int page_end)
809
{
T
Tejun Heo 已提交
810 811
	unsigned int cpu, tcpu;
	int i, err;
812 813 814

	for_each_possible_cpu(cpu) {
		err = __pcpu_map_pages(pcpu_chunk_addr(chunk, cpu, page_start),
T
Tejun Heo 已提交
815
				       &pages[pcpu_page_idx(cpu, page_start)],
816 817
				       page_end - page_start);
		if (err < 0)
T
Tejun Heo 已提交
818
			goto err;
819 820
	}

T
Tejun Heo 已提交
821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
	/* mapping successful, link chunk and mark populated */
	for (i = page_start; i < page_end; i++) {
		for_each_possible_cpu(cpu)
			pcpu_set_page_chunk(pages[pcpu_page_idx(cpu, i)],
					    chunk);
		__set_bit(i, populated);
	}

	return 0;

err:
	for_each_possible_cpu(tcpu) {
		if (tcpu == cpu)
			break;
		__pcpu_unmap_pages(pcpu_chunk_addr(chunk, tcpu, page_start),
				   page_end - page_start);
	}
	return err;
}

/**
 * pcpu_post_map_flush - flush cache after mapping
 * @chunk: pcpu_chunk the regions to be flushed belong to
 * @page_start: page index of the first page to be flushed
 * @page_end: page index of the last page to be flushed + 1
 *
 * Pages [@page_start,@page_end) of @chunk have been mapped.  Flush
 * cache.
 *
 * As with pcpu_pre_unmap_flush(), TLB flushing also is done at once
 * for the whole region.
 */
static void pcpu_post_map_flush(struct pcpu_chunk *chunk,
				int page_start, int page_end)
{
856 857 858
	flush_cache_vmap(
		pcpu_chunk_addr(chunk, pcpu_first_unit_cpu, page_start),
		pcpu_chunk_addr(chunk, pcpu_last_unit_cpu, page_end));
859 860
}

861 862 863 864
/**
 * pcpu_depopulate_chunk - depopulate and unmap an area of a pcpu_chunk
 * @chunk: chunk to depopulate
 * @off: offset to the area to depopulate
T
Tejun Heo 已提交
865
 * @size: size of the area to depopulate in bytes
866 867 868 869 870
 * @flush: whether to flush cache and tlb or not
 *
 * For each cpu, depopulate and unmap pages [@page_start,@page_end)
 * from @chunk.  If @flush is true, vcache is flushed before unmapping
 * and tlb after.
871 872 873
 *
 * CONTEXT:
 * pcpu_alloc_mutex.
874
 */
T
Tejun Heo 已提交
875
static void pcpu_depopulate_chunk(struct pcpu_chunk *chunk, int off, int size)
876 877 878
{
	int page_start = PFN_DOWN(off);
	int page_end = PFN_UP(off + size);
T
Tejun Heo 已提交
879 880 881 882 883 884 885 886 887 888
	struct page **pages;
	unsigned long *populated;
	int rs, re;

	/* quick path, check whether it's empty already */
	pcpu_for_each_unpop_region(chunk, rs, re, page_start, page_end) {
		if (rs == page_start && re == page_end)
			return;
		break;
	}
889

T
Tejun Heo 已提交
890 891
	/* immutable chunks can't be depopulated */
	WARN_ON(chunk->immutable);
892

T
Tejun Heo 已提交
893 894 895 896 897 898 899
	/*
	 * If control reaches here, there must have been at least one
	 * successful population attempt so the temp pages array must
	 * be available now.
	 */
	pages = pcpu_get_pages_and_bitmap(chunk, &populated, false);
	BUG_ON(!pages);
900

T
Tejun Heo 已提交
901 902
	/* unmap and free */
	pcpu_pre_unmap_flush(chunk, page_start, page_end);
903

T
Tejun Heo 已提交
904 905
	pcpu_for_each_pop_region(chunk, rs, re, page_start, page_end)
		pcpu_unmap_pages(chunk, pages, populated, rs, re);
906

T
Tejun Heo 已提交
907 908 909 910
	/* no need to flush tlb, vmalloc will handle it lazily */

	pcpu_for_each_pop_region(chunk, rs, re, page_start, page_end)
		pcpu_free_pages(chunk, pages, populated, rs, re);
911

T
Tejun Heo 已提交
912 913
	/* commit new bitmap */
	bitmap_copy(chunk->populated, populated, pcpu_unit_pages);
914 915 916 917 918 919
}

/**
 * pcpu_populate_chunk - populate and map an area of a pcpu_chunk
 * @chunk: chunk of interest
 * @off: offset to the area to populate
T
Tejun Heo 已提交
920
 * @size: size of the area to populate in bytes
921 922 923
 *
 * For each cpu, populate and map pages [@page_start,@page_end) into
 * @chunk.  The area is cleared on return.
924 925 926
 *
 * CONTEXT:
 * pcpu_alloc_mutex, does GFP_KERNEL allocation.
927 928 929 930 931
 */
static int pcpu_populate_chunk(struct pcpu_chunk *chunk, int off, int size)
{
	int page_start = PFN_DOWN(off);
	int page_end = PFN_UP(off + size);
T
Tejun Heo 已提交
932 933 934
	int free_end = page_start, unmap_end = page_start;
	struct page **pages;
	unsigned long *populated;
935
	unsigned int cpu;
T
Tejun Heo 已提交
936
	int rs, re, rc;
937

T
Tejun Heo 已提交
938 939 940 941 942 943
	/* quick path, check whether all pages are already there */
	pcpu_for_each_pop_region(chunk, rs, re, page_start, page_end) {
		if (rs == page_start && re == page_end)
			goto clear;
		break;
	}
944

T
Tejun Heo 已提交
945 946
	/* need to allocate and map pages, this chunk can't be immutable */
	WARN_ON(chunk->immutable);
947

T
Tejun Heo 已提交
948 949 950
	pages = pcpu_get_pages_and_bitmap(chunk, &populated, true);
	if (!pages)
		return -ENOMEM;
951

T
Tejun Heo 已提交
952 953 954 955 956 957
	/* alloc and map */
	pcpu_for_each_unpop_region(chunk, rs, re, page_start, page_end) {
		rc = pcpu_alloc_pages(chunk, pages, populated, rs, re);
		if (rc)
			goto err_free;
		free_end = re;
958 959
	}

T
Tejun Heo 已提交
960 961 962 963 964 965 966
	pcpu_for_each_unpop_region(chunk, rs, re, page_start, page_end) {
		rc = pcpu_map_pages(chunk, pages, populated, rs, re);
		if (rc)
			goto err_unmap;
		unmap_end = re;
	}
	pcpu_post_map_flush(chunk, page_start, page_end);
967

T
Tejun Heo 已提交
968 969 970
	/* commit new bitmap */
	bitmap_copy(chunk->populated, populated, pcpu_unit_pages);
clear:
971
	for_each_possible_cpu(cpu)
972
		memset((void *)pcpu_chunk_addr(chunk, cpu, 0) + off, 0, size);
973
	return 0;
T
Tejun Heo 已提交
974 975 976 977 978 979 980 981 982 983

err_unmap:
	pcpu_pre_unmap_flush(chunk, page_start, unmap_end);
	pcpu_for_each_unpop_region(chunk, rs, re, page_start, unmap_end)
		pcpu_unmap_pages(chunk, pages, populated, rs, re);
	pcpu_post_unmap_tlb_flush(chunk, page_start, unmap_end);
err_free:
	pcpu_for_each_unpop_region(chunk, rs, re, page_start, free_end)
		pcpu_free_pages(chunk, pages, populated, rs, re);
	return rc;
984 985 986 987 988 989 990 991
}

static void free_pcpu_chunk(struct pcpu_chunk *chunk)
{
	if (!chunk)
		return;
	if (chunk->vm)
		free_vm_area(chunk->vm);
992
	pcpu_mem_free(chunk->map, chunk->map_alloc * sizeof(chunk->map[0]));
993 994 995 996 997 998 999 1000 1001 1002 1003
	kfree(chunk);
}

static struct pcpu_chunk *alloc_pcpu_chunk(void)
{
	struct pcpu_chunk *chunk;

	chunk = kzalloc(pcpu_chunk_struct_size, GFP_KERNEL);
	if (!chunk)
		return NULL;

1004
	chunk->map = pcpu_mem_alloc(PCPU_DFL_MAP_ALLOC * sizeof(chunk->map[0]));
1005 1006 1007
	chunk->map_alloc = PCPU_DFL_MAP_ALLOC;
	chunk->map[chunk->map_used++] = pcpu_unit_size;

1008
	chunk->vm = get_vm_area(pcpu_chunk_size, VM_ALLOC);
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
	if (!chunk->vm) {
		free_pcpu_chunk(chunk);
		return NULL;
	}

	INIT_LIST_HEAD(&chunk->list);
	chunk->free_size = pcpu_unit_size;
	chunk->contig_hint = pcpu_unit_size;

	return chunk;
}

/**
1022
 * pcpu_alloc - the percpu allocator
T
Tejun Heo 已提交
1023
 * @size: size of area to allocate in bytes
1024
 * @align: alignment of area (max PAGE_SIZE)
1025
 * @reserved: allocate from the reserved chunk if available
1026
 *
1027 1028 1029 1030
 * Allocate percpu area of @size bytes aligned at @align.
 *
 * CONTEXT:
 * Does GFP_KERNEL allocation.
1031 1032 1033 1034
 *
 * RETURNS:
 * Percpu pointer to the allocated area on success, NULL on failure.
 */
1035
static void *pcpu_alloc(size_t size, size_t align, bool reserved)
1036 1037 1038 1039
{
	struct pcpu_chunk *chunk;
	int slot, off;

1040
	if (unlikely(!size || size > PCPU_MIN_UNIT_SIZE || align > PAGE_SIZE)) {
1041 1042 1043 1044 1045
		WARN(true, "illegal size (%zu) or align (%zu) for "
		     "percpu allocation\n", size, align);
		return NULL;
	}

1046 1047
	mutex_lock(&pcpu_alloc_mutex);
	spin_lock_irq(&pcpu_lock);
1048

1049 1050 1051
	/* serve reserved allocations from the reserved chunk if available */
	if (reserved && pcpu_reserved_chunk) {
		chunk = pcpu_reserved_chunk;
1052 1053
		if (size > chunk->contig_hint ||
		    pcpu_extend_area_map(chunk) < 0)
1054
			goto fail_unlock;
1055 1056 1057
		off = pcpu_alloc_area(chunk, size, align);
		if (off >= 0)
			goto area_found;
1058
		goto fail_unlock;
1059 1060
	}

1061
restart:
1062
	/* search through normal chunks */
1063 1064 1065 1066
	for (slot = pcpu_size_to_slot(size); slot < pcpu_nr_slots; slot++) {
		list_for_each_entry(chunk, &pcpu_slot[slot], list) {
			if (size > chunk->contig_hint)
				continue;
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076

			switch (pcpu_extend_area_map(chunk)) {
			case 0:
				break;
			case 1:
				goto restart;	/* pcpu_lock dropped, restart */
			default:
				goto fail_unlock;
			}

1077 1078 1079 1080 1081 1082 1083
			off = pcpu_alloc_area(chunk, size, align);
			if (off >= 0)
				goto area_found;
		}
	}

	/* hmmm... no space left, create a new chunk */
1084 1085
	spin_unlock_irq(&pcpu_lock);

1086 1087
	chunk = alloc_pcpu_chunk();
	if (!chunk)
1088 1089 1090
		goto fail_unlock_mutex;

	spin_lock_irq(&pcpu_lock);
1091
	pcpu_chunk_relocate(chunk, -1);
1092
	goto restart;
1093 1094

area_found:
1095 1096
	spin_unlock_irq(&pcpu_lock);

1097 1098
	/* populate, map and clear the area */
	if (pcpu_populate_chunk(chunk, off, size)) {
1099
		spin_lock_irq(&pcpu_lock);
1100
		pcpu_free_area(chunk, off);
1101
		goto fail_unlock;
1102 1103
	}

1104 1105
	mutex_unlock(&pcpu_alloc_mutex);

1106
	/* return address relative to unit0 */
1107 1108 1109 1110 1111 1112 1113
	return __addr_to_pcpu_ptr(chunk->vm->addr + off);

fail_unlock:
	spin_unlock_irq(&pcpu_lock);
fail_unlock_mutex:
	mutex_unlock(&pcpu_alloc_mutex);
	return NULL;
1114
}
1115 1116 1117 1118 1119 1120 1121 1122 1123

/**
 * __alloc_percpu - allocate dynamic percpu area
 * @size: size of area to allocate in bytes
 * @align: alignment of area (max PAGE_SIZE)
 *
 * Allocate percpu area of @size bytes aligned at @align.  Might
 * sleep.  Might trigger writeouts.
 *
1124 1125 1126
 * CONTEXT:
 * Does GFP_KERNEL allocation.
 *
1127 1128 1129 1130 1131 1132 1133
 * RETURNS:
 * Percpu pointer to the allocated area on success, NULL on failure.
 */
void *__alloc_percpu(size_t size, size_t align)
{
	return pcpu_alloc(size, align, false);
}
1134 1135
EXPORT_SYMBOL_GPL(__alloc_percpu);

1136 1137 1138 1139 1140 1141 1142 1143 1144
/**
 * __alloc_reserved_percpu - allocate reserved percpu area
 * @size: size of area to allocate in bytes
 * @align: alignment of area (max PAGE_SIZE)
 *
 * Allocate percpu area of @size bytes aligned at @align from reserved
 * percpu area if arch has set it up; otherwise, allocation is served
 * from the same dynamic area.  Might sleep.  Might trigger writeouts.
 *
1145 1146 1147
 * CONTEXT:
 * Does GFP_KERNEL allocation.
 *
1148 1149 1150 1151 1152 1153 1154 1155
 * RETURNS:
 * Percpu pointer to the allocated area on success, NULL on failure.
 */
void *__alloc_reserved_percpu(size_t size, size_t align)
{
	return pcpu_alloc(size, align, true);
}

1156 1157 1158 1159 1160
/**
 * pcpu_reclaim - reclaim fully free chunks, workqueue function
 * @work: unused
 *
 * Reclaim all fully free chunks except for the first one.
1161 1162 1163
 *
 * CONTEXT:
 * workqueue context.
1164 1165
 */
static void pcpu_reclaim(struct work_struct *work)
1166
{
1167 1168 1169 1170
	LIST_HEAD(todo);
	struct list_head *head = &pcpu_slot[pcpu_nr_slots - 1];
	struct pcpu_chunk *chunk, *next;

1171 1172
	mutex_lock(&pcpu_alloc_mutex);
	spin_lock_irq(&pcpu_lock);
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183

	list_for_each_entry_safe(chunk, next, head, list) {
		WARN_ON(chunk->immutable);

		/* spare the first one */
		if (chunk == list_first_entry(head, struct pcpu_chunk, list))
			continue;

		list_move(&chunk->list, &todo);
	}

1184
	spin_unlock_irq(&pcpu_lock);
1185 1186

	list_for_each_entry_safe(chunk, next, &todo, list) {
T
Tejun Heo 已提交
1187
		pcpu_depopulate_chunk(chunk, 0, pcpu_unit_size);
1188 1189
		free_pcpu_chunk(chunk);
	}
T
Tejun Heo 已提交
1190 1191

	mutex_unlock(&pcpu_alloc_mutex);
1192 1193 1194 1195 1196 1197
}

/**
 * free_percpu - free percpu area
 * @ptr: pointer to area to free
 *
1198 1199 1200 1201
 * Free percpu area @ptr.
 *
 * CONTEXT:
 * Can be called from atomic context.
1202 1203 1204 1205 1206
 */
void free_percpu(void *ptr)
{
	void *addr = __pcpu_ptr_to_addr(ptr);
	struct pcpu_chunk *chunk;
1207
	unsigned long flags;
1208 1209 1210 1211 1212
	int off;

	if (!ptr)
		return;

1213
	spin_lock_irqsave(&pcpu_lock, flags);
1214 1215 1216 1217 1218 1219

	chunk = pcpu_chunk_addr_search(addr);
	off = addr - chunk->vm->addr;

	pcpu_free_area(chunk, off);

1220
	/* if there are more than one fully free chunks, wake up grim reaper */
1221 1222 1223
	if (chunk->free_size == pcpu_unit_size) {
		struct pcpu_chunk *pos;

1224
		list_for_each_entry(pos, &pcpu_slot[pcpu_nr_slots - 1], list)
1225
			if (pos != chunk) {
1226
				schedule_work(&pcpu_reclaim_work);
1227 1228 1229 1230
				break;
			}
	}

1231
	spin_unlock_irqrestore(&pcpu_lock, flags);
1232 1233 1234
}
EXPORT_SYMBOL_GPL(free_percpu);

1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
static inline size_t pcpu_calc_fc_sizes(size_t static_size,
					size_t reserved_size,
					ssize_t *dyn_sizep)
{
	size_t size_sum;

	size_sum = PFN_ALIGN(static_size + reserved_size +
			     (*dyn_sizep >= 0 ? *dyn_sizep : 0));
	if (*dyn_sizep != 0)
		*dyn_sizep = size_sum - static_size - reserved_size;

	return size_sum;
}

/**
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
 * pcpu_alloc_alloc_info - allocate percpu allocation info
 * @nr_groups: the number of groups
 * @nr_units: the number of units
 *
 * Allocate ai which is large enough for @nr_groups groups containing
 * @nr_units units.  The returned ai's groups[0].cpu_map points to the
 * cpu_map array which is long enough for @nr_units and filled with
 * NR_CPUS.  It's the caller's responsibility to initialize cpu_map
 * pointer of other groups.
 *
 * RETURNS:
 * Pointer to the allocated pcpu_alloc_info on success, NULL on
 * failure.
 */
struct pcpu_alloc_info * __init pcpu_alloc_alloc_info(int nr_groups,
						      int nr_units)
{
	struct pcpu_alloc_info *ai;
	size_t base_size, ai_size;
	void *ptr;
	int unit;

	base_size = ALIGN(sizeof(*ai) + nr_groups * sizeof(ai->groups[0]),
			  __alignof__(ai->groups[0].cpu_map[0]));
	ai_size = base_size + nr_units * sizeof(ai->groups[0].cpu_map[0]);

	ptr = alloc_bootmem_nopanic(PFN_ALIGN(ai_size));
	if (!ptr)
		return NULL;
	ai = ptr;
	ptr += base_size;

	ai->groups[0].cpu_map = ptr;

	for (unit = 0; unit < nr_units; unit++)
		ai->groups[0].cpu_map[unit] = NR_CPUS;

	ai->nr_groups = nr_groups;
	ai->__ai_size = PFN_ALIGN(ai_size);

	return ai;
}

/**
 * pcpu_free_alloc_info - free percpu allocation info
 * @ai: pcpu_alloc_info to free
 *
 * Free @ai which was allocated by pcpu_alloc_alloc_info().
 */
void __init pcpu_free_alloc_info(struct pcpu_alloc_info *ai)
{
	free_bootmem(__pa(ai), ai->__ai_size);
}

/**
 * pcpu_build_alloc_info - build alloc_info considering distances between CPUs
1306
 * @reserved_size: the size of reserved percpu area in bytes
1307 1308 1309
 * @dyn_size: free size for dynamic allocation in bytes, -1 for auto
 * @atom_size: allocation atom size
 * @cpu_distance_fn: callback to determine distance between cpus, optional
1310
 *
1311 1312 1313
 * This function determines grouping of units, their mappings to cpus
 * and other parameters considering needed percpu size, allocation
 * atom size and distances between CPUs.
1314
 *
1315 1316 1317 1318 1319
 * Groups are always mutliples of atom size and CPUs which are of
 * LOCAL_DISTANCE both ways are grouped together and share space for
 * units in the same group.  The returned configuration is guaranteed
 * to have CPUs on different nodes on different groups and >=75% usage
 * of allocated virtual address space.
1320 1321
 *
 * RETURNS:
1322 1323
 * On success, pointer to the new allocation_info is returned.  On
 * failure, ERR_PTR value is returned.
1324
 */
1325 1326 1327 1328
struct pcpu_alloc_info * __init pcpu_build_alloc_info(
				size_t reserved_size, ssize_t dyn_size,
				size_t atom_size,
				pcpu_fc_cpu_distance_fn_t cpu_distance_fn)
1329 1330 1331 1332
{
	static int group_map[NR_CPUS] __initdata;
	static int group_cnt[NR_CPUS] __initdata;
	const size_t static_size = __per_cpu_end - __per_cpu_start;
1333
	int group_cnt_max = 0, nr_groups = 1, nr_units = 0;
1334 1335
	size_t size_sum, min_unit_size, alloc_size;
	int upa, max_upa, uninitialized_var(best_upa);	/* units_per_alloc */
1336
	int last_allocs, group, unit;
1337
	unsigned int cpu, tcpu;
1338 1339
	struct pcpu_alloc_info *ai;
	unsigned int *cpu_map;
1340 1341 1342

	/*
	 * Determine min_unit_size, alloc_size and max_upa such that
1343
	 * alloc_size is multiple of atom_size and is the smallest
1344 1345 1346
	 * which can accomodate 4k aligned segments which are equal to
	 * or larger than min_unit_size.
	 */
1347
	size_sum = pcpu_calc_fc_sizes(static_size, reserved_size, &dyn_size);
1348 1349
	min_unit_size = max_t(size_t, size_sum, PCPU_MIN_UNIT_SIZE);

1350
	alloc_size = roundup(min_unit_size, atom_size);
1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362
	upa = alloc_size / min_unit_size;
	while (alloc_size % upa || ((alloc_size / upa) & ~PAGE_MASK))
		upa--;
	max_upa = upa;

	/* group cpus according to their proximity */
	for_each_possible_cpu(cpu) {
		group = 0;
	next_group:
		for_each_possible_cpu(tcpu) {
			if (cpu == tcpu)
				break;
1363
			if (group_map[tcpu] == group && cpu_distance_fn &&
1364 1365 1366
			    (cpu_distance_fn(cpu, tcpu) > LOCAL_DISTANCE ||
			     cpu_distance_fn(tcpu, cpu) > LOCAL_DISTANCE)) {
				group++;
1367
				nr_groups = max(nr_groups, group + 1);
1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
				goto next_group;
			}
		}
		group_map[cpu] = group;
		group_cnt[group]++;
		group_cnt_max = max(group_cnt_max, group_cnt[group]);
	}

	/*
	 * Expand unit size until address space usage goes over 75%
	 * and then as much as possible without using more address
	 * space.
	 */
	last_allocs = INT_MAX;
	for (upa = max_upa; upa; upa--) {
		int allocs = 0, wasted = 0;

		if (alloc_size % upa || ((alloc_size / upa) & ~PAGE_MASK))
			continue;

1388
		for (group = 0; group < nr_groups; group++) {
1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407
			int this_allocs = DIV_ROUND_UP(group_cnt[group], upa);
			allocs += this_allocs;
			wasted += this_allocs * upa - group_cnt[group];
		}

		/*
		 * Don't accept if wastage is over 25%.  The
		 * greater-than comparison ensures upa==1 always
		 * passes the following check.
		 */
		if (wasted > num_possible_cpus() / 3)
			continue;

		/* and then don't consume more memory */
		if (allocs > last_allocs)
			break;
		last_allocs = allocs;
		best_upa = upa;
	}
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439
	upa = best_upa;

	/* allocate and fill alloc_info */
	for (group = 0; group < nr_groups; group++)
		nr_units += roundup(group_cnt[group], upa);

	ai = pcpu_alloc_alloc_info(nr_groups, nr_units);
	if (!ai)
		return ERR_PTR(-ENOMEM);
	cpu_map = ai->groups[0].cpu_map;

	for (group = 0; group < nr_groups; group++) {
		ai->groups[group].cpu_map = cpu_map;
		cpu_map += roundup(group_cnt[group], upa);
	}

	ai->static_size = static_size;
	ai->reserved_size = reserved_size;
	ai->dyn_size = dyn_size;
	ai->unit_size = alloc_size / upa;
	ai->atom_size = atom_size;
	ai->alloc_size = alloc_size;

	for (group = 0, unit = 0; group_cnt[group]; group++) {
		struct pcpu_group_info *gi = &ai->groups[group];

		/*
		 * Initialize base_offset as if all groups are located
		 * back-to-back.  The caller should update this to
		 * reflect actual allocation.
		 */
		gi->base_offset = unit * ai->unit_size;
1440 1441 1442

		for_each_possible_cpu(cpu)
			if (group_map[cpu] == group)
1443 1444 1445
				gi->cpu_map[gi->nr_units++] = cpu;
		gi->nr_units = roundup(gi->nr_units, upa);
		unit += gi->nr_units;
1446
	}
1447
	BUG_ON(unit != nr_units);
1448

1449
	return ai;
1450 1451
}

1452 1453 1454 1455 1456 1457 1458 1459 1460
/**
 * pcpu_dump_alloc_info - print out information about pcpu_alloc_info
 * @lvl: loglevel
 * @ai: allocation info to dump
 *
 * Print out information about @ai using loglevel @lvl.
 */
static void pcpu_dump_alloc_info(const char *lvl,
				 const struct pcpu_alloc_info *ai)
1461
{
1462
	int group_width = 1, cpu_width = 1, width;
1463
	char empty_str[] = "--------";
1464 1465 1466 1467 1468 1469 1470
	int alloc = 0, alloc_end = 0;
	int group, v;
	int upa, apl;	/* units per alloc, allocs per line */

	v = ai->nr_groups;
	while (v /= 10)
		group_width++;
1471

1472
	v = num_possible_cpus();
1473
	while (v /= 10)
1474 1475
		cpu_width++;
	empty_str[min_t(int, cpu_width, sizeof(empty_str) - 1)] = '\0';
1476

1477 1478 1479
	upa = ai->alloc_size / ai->unit_size;
	width = upa * (cpu_width + 1) + group_width + 3;
	apl = rounddown_pow_of_two(max(60 / width, 1));
1480

1481 1482 1483
	printk("%spcpu-alloc: s%zu r%zu d%zu u%zu alloc=%zu*%zu",
	       lvl, ai->static_size, ai->reserved_size, ai->dyn_size,
	       ai->unit_size, ai->alloc_size / ai->atom_size, ai->atom_size);
1484

1485 1486 1487 1488 1489 1490 1491 1492
	for (group = 0; group < ai->nr_groups; group++) {
		const struct pcpu_group_info *gi = &ai->groups[group];
		int unit = 0, unit_end = 0;

		BUG_ON(gi->nr_units % upa);
		for (alloc_end += gi->nr_units / upa;
		     alloc < alloc_end; alloc++) {
			if (!(alloc % apl)) {
1493
				printk("\n");
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503
				printk("%spcpu-alloc: ", lvl);
			}
			printk("[%0*d] ", group_width, group);

			for (unit_end += upa; unit < unit_end; unit++)
				if (gi->cpu_map[unit] != NR_CPUS)
					printk("%0*d ", cpu_width,
					       gi->cpu_map[unit]);
				else
					printk("%s ", empty_str);
1504 1505 1506 1507 1508
		}
	}
	printk("\n");
}

1509
/**
1510
 * pcpu_setup_first_chunk - initialize the first percpu chunk
1511
 * @ai: pcpu_alloc_info describing how to percpu area is shaped
1512
 * @base_addr: mapped address
1513 1514 1515
 *
 * Initialize the first percpu chunk which contains the kernel static
 * perpcu area.  This function is to be called from arch percpu area
1516
 * setup path.
1517
 *
1518 1519 1520 1521 1522 1523
 * @ai contains all information necessary to initialize the first
 * chunk and prime the dynamic percpu allocator.
 *
 * @ai->static_size is the size of static percpu area.
 *
 * @ai->reserved_size, if non-zero, specifies the amount of bytes to
1524 1525 1526 1527 1528 1529 1530
 * reserve after the static area in the first chunk.  This reserves
 * the first chunk such that it's available only through reserved
 * percpu allocation.  This is primarily used to serve module percpu
 * static areas on architectures where the addressing model has
 * limited offset range for symbol relocations to guarantee module
 * percpu symbols fall inside the relocatable range.
 *
1531 1532 1533
 * @ai->dyn_size determines the number of bytes available for dynamic
 * allocation in the first chunk.  The area between @ai->static_size +
 * @ai->reserved_size + @ai->dyn_size and @ai->unit_size is unused.
1534
 *
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550
 * @ai->unit_size specifies unit size and must be aligned to PAGE_SIZE
 * and equal to or larger than @ai->static_size + @ai->reserved_size +
 * @ai->dyn_size.
 *
 * @ai->atom_size is the allocation atom size and used as alignment
 * for vm areas.
 *
 * @ai->alloc_size is the allocation size and always multiple of
 * @ai->atom_size.  This is larger than @ai->atom_size if
 * @ai->unit_size is larger than @ai->atom_size.
 *
 * @ai->nr_groups and @ai->groups describe virtual memory layout of
 * percpu areas.  Units which should be colocated are put into the
 * same group.  Dynamic VM areas will be allocated according to these
 * groupings.  If @ai->nr_groups is zero, a single group containing
 * all units is assumed.
1551
 *
1552 1553
 * The caller should have mapped the first chunk at @base_addr and
 * copied static data to each unit.
1554
 *
1555 1556 1557 1558 1559 1560 1561
 * If the first chunk ends up with both reserved and dynamic areas, it
 * is served by two chunks - one to serve the core static and reserved
 * areas and the other for the dynamic area.  They share the same vm
 * and page map but uses different area allocation map to stay away
 * from each other.  The latter chunk is circulated in the chunk slots
 * and available for dynamic allocation like any other chunks.
 *
1562 1563 1564 1565
 * RETURNS:
 * The determined pcpu_unit_size which can be used to initialize
 * percpu access.
 */
1566 1567
size_t __init pcpu_setup_first_chunk(const struct pcpu_alloc_info *ai,
				     void *base_addr)
1568
{
1569
	static struct vm_struct first_vm;
1570
	static int smap[2], dmap[2];
1571 1572
	size_t dyn_size = ai->dyn_size;
	size_t size_sum = ai->static_size + ai->reserved_size + dyn_size;
1573
	struct pcpu_chunk *schunk, *dchunk = NULL;
1574 1575 1576
	unsigned int cpu;
	int *unit_map;
	int group, unit, i;
1577

1578
	/* sanity checks */
1579 1580
	BUILD_BUG_ON(ARRAY_SIZE(smap) >= PCPU_DFL_MAP_ALLOC ||
		     ARRAY_SIZE(dmap) >= PCPU_DFL_MAP_ALLOC);
1581 1582
	BUG_ON(ai->nr_groups <= 0);
	BUG_ON(!ai->static_size);
1583
	BUG_ON(!base_addr);
1584 1585 1586 1587 1588
	BUG_ON(ai->unit_size < size_sum);
	BUG_ON(ai->unit_size & ~PAGE_MASK);
	BUG_ON(ai->unit_size < PCPU_MIN_UNIT_SIZE);

	pcpu_dump_alloc_info(KERN_DEBUG, ai);
1589

1590
	/* determine number of units and verify and initialize pcpu_unit_map */
1591
	unit_map = alloc_bootmem(nr_cpu_ids * sizeof(unit_map[0]));
1592

1593 1594 1595
	for (cpu = 0; cpu < nr_cpu_ids; cpu++)
		unit_map[cpu] = NR_CPUS;
	pcpu_first_unit_cpu = NR_CPUS;
1596

1597 1598
	for (group = 0, unit = 0; group < ai->nr_groups; group++, unit += i) {
		const struct pcpu_group_info *gi = &ai->groups[group];
1599

1600 1601 1602 1603
		for (i = 0; i < gi->nr_units; i++) {
			cpu = gi->cpu_map[i];
			if (cpu == NR_CPUS)
				continue;
1604

1605 1606 1607 1608 1609 1610 1611
			BUG_ON(cpu > nr_cpu_ids || !cpu_possible(cpu));
			BUG_ON(unit_map[cpu] != NR_CPUS);

			unit_map[cpu] = unit + i;
			if (pcpu_first_unit_cpu == NR_CPUS)
				pcpu_first_unit_cpu = cpu;
		}
1612
	}
1613 1614 1615 1616 1617 1618 1619
	pcpu_last_unit_cpu = cpu;
	pcpu_nr_units = unit;

	for_each_possible_cpu(cpu)
		BUG_ON(unit_map[cpu] == NR_CPUS);

	pcpu_unit_map = unit_map;
1620 1621

	/* determine basic parameters */
1622
	pcpu_unit_pages = ai->unit_size >> PAGE_SHIFT;
1623
	pcpu_unit_size = pcpu_unit_pages << PAGE_SHIFT;
1624
	pcpu_chunk_size = pcpu_nr_units * pcpu_unit_size;
T
Tejun Heo 已提交
1625 1626
	pcpu_chunk_struct_size = sizeof(struct pcpu_chunk) +
		BITS_TO_LONGS(pcpu_unit_pages) * sizeof(unsigned long);
1627

1628 1629 1630 1631
	first_vm.flags = VM_ALLOC;
	first_vm.size = pcpu_chunk_size;
	first_vm.addr = base_addr;

1632 1633 1634 1635 1636
	/*
	 * Allocate chunk slots.  The additional last slot is for
	 * empty chunks.
	 */
	pcpu_nr_slots = __pcpu_size_to_slot(pcpu_unit_size) + 2;
1637 1638 1639 1640
	pcpu_slot = alloc_bootmem(pcpu_nr_slots * sizeof(pcpu_slot[0]));
	for (i = 0; i < pcpu_nr_slots; i++)
		INIT_LIST_HEAD(&pcpu_slot[i]);

1641 1642 1643 1644 1645 1646 1647
	/*
	 * Initialize static chunk.  If reserved_size is zero, the
	 * static chunk covers static area + dynamic allocation area
	 * in the first chunk.  If reserved_size is not zero, it
	 * covers static area + reserved area (mostly used for module
	 * static percpu allocation).
	 */
1648 1649 1650
	schunk = alloc_bootmem(pcpu_chunk_struct_size);
	INIT_LIST_HEAD(&schunk->list);
	schunk->vm = &first_vm;
1651 1652
	schunk->map = smap;
	schunk->map_alloc = ARRAY_SIZE(smap);
1653
	schunk->immutable = true;
T
Tejun Heo 已提交
1654
	bitmap_fill(schunk->populated, pcpu_unit_pages);
1655

1656 1657
	if (ai->reserved_size) {
		schunk->free_size = ai->reserved_size;
1658
		pcpu_reserved_chunk = schunk;
1659
		pcpu_reserved_chunk_limit = ai->static_size + ai->reserved_size;
1660 1661 1662 1663
	} else {
		schunk->free_size = dyn_size;
		dyn_size = 0;			/* dynamic area covered */
	}
1664
	schunk->contig_hint = schunk->free_size;
1665

1666
	schunk->map[schunk->map_used++] = -ai->static_size;
1667 1668 1669
	if (schunk->free_size)
		schunk->map[schunk->map_used++] = schunk->free_size;

1670 1671
	/* init dynamic chunk if necessary */
	if (dyn_size) {
T
Tejun Heo 已提交
1672
		dchunk = alloc_bootmem(pcpu_chunk_struct_size);
1673 1674 1675 1676
		INIT_LIST_HEAD(&dchunk->list);
		dchunk->vm = &first_vm;
		dchunk->map = dmap;
		dchunk->map_alloc = ARRAY_SIZE(dmap);
1677
		dchunk->immutable = true;
T
Tejun Heo 已提交
1678
		bitmap_fill(dchunk->populated, pcpu_unit_pages);
1679 1680 1681 1682 1683 1684

		dchunk->contig_hint = dchunk->free_size = dyn_size;
		dchunk->map[dchunk->map_used++] = -pcpu_reserved_chunk_limit;
		dchunk->map[dchunk->map_used++] = dchunk->free_size;
	}

1685
	/* link the first chunk in */
1686 1687
	pcpu_first_chunk = dchunk ?: schunk;
	pcpu_chunk_relocate(pcpu_first_chunk, -1);
1688 1689

	/* we're done */
1690
	pcpu_base_addr = schunk->vm->addr;
1691 1692
	return pcpu_unit_size;
}
1693

1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
const char *pcpu_fc_names[PCPU_FC_NR] __initdata = {
	[PCPU_FC_AUTO]	= "auto",
	[PCPU_FC_EMBED]	= "embed",
	[PCPU_FC_PAGE]	= "page",
	[PCPU_FC_LPAGE]	= "lpage",
};

enum pcpu_fc pcpu_chosen_fc __initdata = PCPU_FC_AUTO;

static int __init percpu_alloc_setup(char *str)
{
	if (0)
		/* nada */;
#ifdef CONFIG_NEED_PER_CPU_EMBED_FIRST_CHUNK
	else if (!strcmp(str, "embed"))
		pcpu_chosen_fc = PCPU_FC_EMBED;
#endif
#ifdef CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK
	else if (!strcmp(str, "page"))
		pcpu_chosen_fc = PCPU_FC_PAGE;
#endif
#ifdef CONFIG_NEED_PER_CPU_LPAGE_FIRST_CHUNK
	else if (!strcmp(str, "lpage"))
		pcpu_chosen_fc = PCPU_FC_LPAGE;
#endif
	else
		pr_warning("PERCPU: unknown allocator %s specified\n", str);

	return 0;
}
early_param("percpu_alloc", percpu_alloc_setup);

1726 1727
#if defined(CONFIG_NEED_PER_CPU_EMBED_FIRST_CHUNK) || \
	!defined(CONFIG_HAVE_SETUP_PER_CPU_AREA)
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
/**
 * pcpu_embed_first_chunk - embed the first percpu chunk into bootmem
 * @reserved_size: the size of reserved percpu area in bytes
 * @dyn_size: free size for dynamic allocation in bytes, -1 for auto
 *
 * This is a helper to ease setting up embedded first percpu chunk and
 * can be called where pcpu_setup_first_chunk() is expected.
 *
 * If this function is used to setup the first chunk, it is allocated
 * as a contiguous area using bootmem allocator and used as-is without
 * being mapped into vmalloc area.  This enables the first chunk to
 * piggy back on the linear physical mapping which often uses larger
 * page size.
 *
 * When @dyn_size is positive, dynamic area might be larger than
1743 1744 1745
 * specified to fill page alignment.  When @dyn_size is auto,
 * @dyn_size is just big enough to fill page alignment after static
 * and reserved areas.
1746 1747 1748 1749 1750 1751 1752 1753
 *
 * If the needed size is smaller than the minimum or specified unit
 * size, the leftover is returned to the bootmem allocator.
 *
 * RETURNS:
 * The determined pcpu_unit_size which can be used to initialize
 * percpu access on success, -errno on failure.
 */
1754
ssize_t __init pcpu_embed_first_chunk(size_t reserved_size, ssize_t dyn_size)
1755
{
1756 1757
	struct pcpu_alloc_info *ai;
	size_t size_sum, chunk_size;
T
Tejun Heo 已提交
1758
	void *base;
1759 1760
	int unit;
	ssize_t ret;
1761

1762 1763 1764 1765 1766
	ai = pcpu_build_alloc_info(reserved_size, dyn_size, PAGE_SIZE, NULL);
	if (IS_ERR(ai))
		return PTR_ERR(ai);
	BUG_ON(ai->nr_groups != 1);
	BUG_ON(ai->groups[0].nr_units != num_possible_cpus());
1767

1768 1769
	size_sum = ai->static_size + ai->reserved_size + ai->dyn_size;
	chunk_size = ai->unit_size * num_possible_cpus();
1770

T
Tejun Heo 已提交
1771 1772 1773
	base = __alloc_bootmem_nopanic(chunk_size, PAGE_SIZE,
				       __pa(MAX_DMA_ADDRESS));
	if (!base) {
1774 1775
		pr_warning("PERCPU: failed to allocate %zu bytes for "
			   "embedding\n", chunk_size);
1776 1777
		ret = -ENOMEM;
		goto out_free_ai;
1778
	}
1779 1780

	/* return the leftover and copy */
1781 1782 1783 1784 1785
	for (unit = 0; unit < num_possible_cpus(); unit++) {
		void *ptr = base + unit * ai->unit_size;

		free_bootmem(__pa(ptr + size_sum), ai->unit_size - size_sum);
		memcpy(ptr, __per_cpu_load, ai->static_size);
1786 1787 1788
	}

	/* we're ready, commit */
T
Tejun Heo 已提交
1789
	pr_info("PERCPU: Embedded %zu pages/cpu @%p s%zu r%zu d%zu u%zu\n",
1790 1791
		PFN_DOWN(size_sum), base, ai->static_size, ai->reserved_size,
		ai->dyn_size, ai->unit_size);
1792

1793 1794 1795 1796
	ret = pcpu_setup_first_chunk(ai, base);
out_free_ai:
	pcpu_free_alloc_info(ai);
	return ret;
1797
}
1798 1799
#endif /* CONFIG_NEED_PER_CPU_EMBED_FIRST_CHUNK ||
	  !CONFIG_HAVE_SETUP_PER_CPU_AREA */
1800

1801
#ifdef CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK
1802
/**
1803
 * pcpu_page_first_chunk - map the first chunk using PAGE_SIZE pages
1804 1805 1806 1807 1808
 * @reserved_size: the size of reserved percpu area in bytes
 * @alloc_fn: function to allocate percpu page, always called with PAGE_SIZE
 * @free_fn: funtion to free percpu page, always called with PAGE_SIZE
 * @populate_pte_fn: function to populate pte
 *
1809 1810
 * This is a helper to ease setting up page-remapped first percpu
 * chunk and can be called where pcpu_setup_first_chunk() is expected.
1811 1812 1813 1814 1815 1816 1817 1818
 *
 * This is the basic allocator.  Static percpu area is allocated
 * page-by-page into vmalloc area.
 *
 * RETURNS:
 * The determined pcpu_unit_size which can be used to initialize
 * percpu access on success, -errno on failure.
 */
1819
ssize_t __init pcpu_page_first_chunk(size_t reserved_size,
1820 1821 1822
				     pcpu_fc_alloc_fn_t alloc_fn,
				     pcpu_fc_free_fn_t free_fn,
				     pcpu_fc_populate_pte_fn_t populate_pte_fn)
1823
{
1824
	static struct vm_struct vm;
1825
	struct pcpu_alloc_info *ai;
1826
	char psize_str[16];
T
Tejun Heo 已提交
1827
	int unit_pages;
1828
	size_t pages_size;
T
Tejun Heo 已提交
1829
	struct page **pages;
1830
	int unit, i, j;
1831 1832
	ssize_t ret;

1833 1834
	snprintf(psize_str, sizeof(psize_str), "%luK", PAGE_SIZE >> 10);

1835 1836 1837 1838 1839 1840 1841
	ai = pcpu_build_alloc_info(reserved_size, -1, PAGE_SIZE, NULL);
	if (IS_ERR(ai))
		return PTR_ERR(ai);
	BUG_ON(ai->nr_groups != 1);
	BUG_ON(ai->groups[0].nr_units != num_possible_cpus());

	unit_pages = ai->unit_size >> PAGE_SHIFT;
1842 1843

	/* unaligned allocations can't be freed, round up to page size */
1844 1845
	pages_size = PFN_ALIGN(unit_pages * num_possible_cpus() *
			       sizeof(pages[0]));
T
Tejun Heo 已提交
1846
	pages = alloc_bootmem(pages_size);
1847

1848
	/* allocate pages */
1849
	j = 0;
1850
	for (unit = 0; unit < num_possible_cpus(); unit++)
T
Tejun Heo 已提交
1851
		for (i = 0; i < unit_pages; i++) {
1852
			unsigned int cpu = ai->groups[0].cpu_map[unit];
1853 1854
			void *ptr;

1855
			ptr = alloc_fn(cpu, PAGE_SIZE, PAGE_SIZE);
1856
			if (!ptr) {
1857 1858
				pr_warning("PERCPU: failed to allocate %s page "
					   "for cpu%u\n", psize_str, cpu);
1859 1860
				goto enomem;
			}
T
Tejun Heo 已提交
1861
			pages[j++] = virt_to_page(ptr);
1862 1863
		}

1864 1865
	/* allocate vm area, map the pages and copy static data */
	vm.flags = VM_ALLOC;
1866
	vm.size = num_possible_cpus() * ai->unit_size;
1867 1868
	vm_area_register_early(&vm, PAGE_SIZE);

1869
	for (unit = 0; unit < num_possible_cpus(); unit++) {
1870
		unsigned long unit_addr =
1871
			(unsigned long)vm.addr + unit * ai->unit_size;
1872

T
Tejun Heo 已提交
1873
		for (i = 0; i < unit_pages; i++)
1874 1875 1876
			populate_pte_fn(unit_addr + (i << PAGE_SHIFT));

		/* pte already populated, the following shouldn't fail */
1877
		ret = __pcpu_map_pages(unit_addr, &pages[unit * unit_pages],
T
Tejun Heo 已提交
1878
				       unit_pages);
1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890
		if (ret < 0)
			panic("failed to map percpu area, err=%zd\n", ret);

		/*
		 * FIXME: Archs with virtual cache should flush local
		 * cache for the linear mapping here - something
		 * equivalent to flush_cache_vmap() on the local cpu.
		 * flush_cache_vmap() can't be used as most supporting
		 * data structures are not set up yet.
		 */

		/* copy static data */
1891
		memcpy((void *)unit_addr, __per_cpu_load, ai->static_size);
1892 1893
	}

1894
	/* we're ready, commit */
1895
	pr_info("PERCPU: %d %s pages/cpu @%p s%zu r%zu d%zu\n",
1896 1897
		unit_pages, psize_str, vm.addr, ai->static_size,
		ai->reserved_size, ai->dyn_size);
1898

1899
	ret = pcpu_setup_first_chunk(ai, vm.addr);
1900 1901 1902 1903
	goto out_free_ar;

enomem:
	while (--j >= 0)
T
Tejun Heo 已提交
1904
		free_fn(page_address(pages[j]), PAGE_SIZE);
1905 1906
	ret = -ENOMEM;
out_free_ar:
T
Tejun Heo 已提交
1907
	free_bootmem(__pa(pages), pages_size);
1908
	pcpu_free_alloc_info(ai);
1909 1910
	return ret;
}
1911
#endif /* CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK */
1912

1913
#ifdef CONFIG_NEED_PER_CPU_LPAGE_FIRST_CHUNK
1914 1915
struct pcpul_ent {
	void		*ptr;
1916
	void		*map_addr;
1917 1918 1919
};

static size_t pcpul_size;
1920 1921
static size_t pcpul_lpage_size;
static int pcpul_nr_lpages;
1922
static struct pcpul_ent *pcpul_map;
1923

1924
static bool __init pcpul_unit_to_cpu(int unit, const struct pcpu_alloc_info *ai,
1925 1926
				     unsigned int *cpup)
{
1927
	int group, cunit;
1928

1929 1930 1931 1932
	for (group = 0, cunit = 0; group < ai->nr_groups; group++) {
		const struct pcpu_group_info *gi = &ai->groups[group];

		if (unit < cunit + gi->nr_units) {
1933
			if (cpup)
1934
				*cpup = gi->cpu_map[unit - cunit];
1935 1936
			return true;
		}
1937 1938
		cunit += gi->nr_units;
	}
1939 1940 1941 1942

	return false;
}

1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956
static int __init pcpul_cpu_to_unit(int cpu, const struct pcpu_alloc_info *ai)
{
	int group, unit, i;

	for (group = 0, unit = 0; group < ai->nr_groups; group++, unit += i) {
		const struct pcpu_group_info *gi = &ai->groups[group];

		for (i = 0; i < gi->nr_units; i++)
			if (gi->cpu_map[i] == cpu)
				return unit + i;
	}
	BUG();
}

1957 1958
/**
 * pcpu_lpage_first_chunk - remap the first percpu chunk using large page
1959
 * @ai: pcpu_alloc_info
1960 1961 1962 1963
 * @alloc_fn: function to allocate percpu lpage, always called with lpage_size
 * @free_fn: function to free percpu memory, @size <= lpage_size
 * @map_fn: function to map percpu lpage, always called with lpage_size
 *
1964
 * This allocator uses large page to build and map the first chunk.
1965 1966 1967
 * Unlike other helpers, the caller should provide fully initialized
 * @ai.  This can be done using pcpu_build_alloc_info().  This two
 * stage initialization is to allow arch code to evaluate the
1968 1969 1970 1971 1972 1973 1974 1975 1976 1977
 * parameters before committing to it.
 *
 * Large pages are allocated as directed by @unit_map and other
 * parameters and mapped to vmalloc space.  Unused holes are returned
 * to the page allocator.  Note that these holes end up being actively
 * mapped twice - once to the physical mapping and to the vmalloc area
 * for the first percpu chunk.  Depending on architecture, this might
 * cause problem when changing page attributes of the returned area.
 * These double mapped areas can be detected using
 * pcpu_lpage_remapped().
1978 1979 1980 1981 1982
 *
 * RETURNS:
 * The determined pcpu_unit_size which can be used to initialize
 * percpu access on success, -errno on failure.
 */
1983
ssize_t __init pcpu_lpage_first_chunk(const struct pcpu_alloc_info *ai,
1984 1985 1986 1987
				      pcpu_fc_alloc_fn_t alloc_fn,
				      pcpu_fc_free_fn_t free_fn,
				      pcpu_fc_map_fn_t map_fn)
{
1988
	static struct vm_struct vm;
1989 1990
	const size_t lpage_size = ai->atom_size;
	size_t chunk_size, map_size;
1991 1992
	unsigned int cpu;
	ssize_t ret;
1993
	int i, j, unit, nr_units;
1994

1995 1996 1997
	nr_units = 0;
	for (i = 0; i < ai->nr_groups; i++)
		nr_units += ai->groups[i].nr_units;
1998

1999
	chunk_size = ai->unit_size * nr_units;
2000 2001
	BUG_ON(chunk_size % lpage_size);

2002
	pcpul_size = ai->static_size + ai->reserved_size + ai->dyn_size;
2003 2004
	pcpul_lpage_size = lpage_size;
	pcpul_nr_lpages = chunk_size / lpage_size;
2005 2006

	/* allocate pointer array and alloc large pages */
2007
	map_size = pcpul_nr_lpages * sizeof(pcpul_map[0]);
2008 2009
	pcpul_map = alloc_bootmem(map_size);

2010 2011 2012
	/* allocate all pages */
	for (i = 0; i < pcpul_nr_lpages; i++) {
		size_t offset = i * lpage_size;
2013 2014
		int first_unit = offset / ai->unit_size;
		int last_unit = (offset + lpage_size - 1) / ai->unit_size;
2015 2016
		void *ptr;

2017 2018
		/* find out which cpu is mapped to this unit */
		for (unit = first_unit; unit <= last_unit; unit++)
2019
			if (pcpul_unit_to_cpu(unit, ai, &cpu))
2020 2021 2022
				goto found;
		continue;
	found:
2023
		ptr = alloc_fn(cpu, lpage_size, lpage_size);
2024 2025 2026 2027 2028 2029
		if (!ptr) {
			pr_warning("PERCPU: failed to allocate large page "
				   "for cpu%u\n", cpu);
			goto enomem;
		}

2030 2031
		pcpul_map[i].ptr = ptr;
	}
2032

2033 2034
	/* return unused holes */
	for (unit = 0; unit < nr_units; unit++) {
2035 2036
		size_t start = unit * ai->unit_size;
		size_t end = start + ai->unit_size;
2037 2038 2039
		size_t off, next;

		/* don't free used part of occupied unit */
2040
		if (pcpul_unit_to_cpu(unit, ai, NULL))
2041 2042 2043 2044 2045 2046 2047 2048 2049
			start += pcpul_size;

		/* unit can span more than one page, punch the holes */
		for (off = start; off < end; off = next) {
			void *ptr = pcpul_map[off / lpage_size].ptr;
			next = min(roundup(off + 1, lpage_size), end);
			if (ptr)
				free_fn(ptr + off % lpage_size, next - off);
		}
2050 2051
	}

2052 2053 2054
	/* allocate address, map and copy */
	vm.flags = VM_ALLOC;
	vm.size = chunk_size;
2055
	vm_area_register_early(&vm, ai->unit_size);
2056 2057 2058 2059 2060 2061 2062

	for (i = 0; i < pcpul_nr_lpages; i++) {
		if (!pcpul_map[i].ptr)
			continue;
		pcpul_map[i].map_addr = vm.addr + i * lpage_size;
		map_fn(pcpul_map[i].ptr, lpage_size, pcpul_map[i].map_addr);
	}
2063 2064

	for_each_possible_cpu(cpu)
2065 2066
		memcpy(vm.addr + pcpul_cpu_to_unit(cpu, ai) * ai->unit_size,
		       __per_cpu_load, ai->static_size);
2067 2068

	/* we're ready, commit */
T
Tejun Heo 已提交
2069
	pr_info("PERCPU: large pages @%p s%zu r%zu d%zu u%zu\n",
2070 2071
		vm.addr, ai->static_size, ai->reserved_size, ai->dyn_size,
		ai->unit_size);
2072

2073
	ret = pcpu_setup_first_chunk(ai, vm.addr);
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095

	/*
	 * Sort pcpul_map array for pcpu_lpage_remapped().  Unmapped
	 * lpages are pushed to the end and trimmed.
	 */
	for (i = 0; i < pcpul_nr_lpages - 1; i++)
		for (j = i + 1; j < pcpul_nr_lpages; j++) {
			struct pcpul_ent tmp;

			if (!pcpul_map[j].ptr)
				continue;
			if (pcpul_map[i].ptr &&
			    pcpul_map[i].ptr < pcpul_map[j].ptr)
				continue;

			tmp = pcpul_map[i];
			pcpul_map[i] = pcpul_map[j];
			pcpul_map[j] = tmp;
		}

	while (pcpul_nr_lpages && !pcpul_map[pcpul_nr_lpages - 1].ptr)
		pcpul_nr_lpages--;
2096 2097 2098 2099

	return ret;

enomem:
2100 2101 2102
	for (i = 0; i < pcpul_nr_lpages; i++)
		if (pcpul_map[i].ptr)
			free_fn(pcpul_map[i].ptr, lpage_size);
2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124
	free_bootmem(__pa(pcpul_map), map_size);
	return -ENOMEM;
}

/**
 * pcpu_lpage_remapped - determine whether a kaddr is in pcpul recycled area
 * @kaddr: the kernel address in question
 *
 * Determine whether @kaddr falls in the pcpul recycled area.  This is
 * used by pageattr to detect VM aliases and break up the pcpu large
 * page mapping such that the same physical page is not mapped under
 * different attributes.
 *
 * The recycled area is always at the tail of a partially used large
 * page.
 *
 * RETURNS:
 * Address of corresponding remapped pcpu address if match is found;
 * otherwise, NULL.
 */
void *pcpu_lpage_remapped(void *kaddr)
{
2125 2126 2127 2128
	unsigned long lpage_mask = pcpul_lpage_size - 1;
	void *lpage_addr = (void *)((unsigned long)kaddr & ~lpage_mask);
	unsigned long offset = (unsigned long)kaddr & lpage_mask;
	int left = 0, right = pcpul_nr_lpages - 1;
2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142
	int pos;

	/* pcpul in use at all? */
	if (!pcpul_map)
		return NULL;

	/* okay, perform binary search */
	while (left <= right) {
		pos = (left + right) / 2;

		if (pcpul_map[pos].ptr < lpage_addr)
			left = pos + 1;
		else if (pcpul_map[pos].ptr > lpage_addr)
			right = pos - 1;
2143 2144
		else
			return pcpul_map[pos].map_addr + offset;
2145 2146 2147 2148
	}

	return NULL;
}
2149
#endif /* CONFIG_NEED_PER_CPU_LPAGE_FIRST_CHUNK */
2150

2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176
/*
 * Generic percpu area setup.
 *
 * The embedding helper is used because its behavior closely resembles
 * the original non-dynamic generic percpu area setup.  This is
 * important because many archs have addressing restrictions and might
 * fail if the percpu area is located far away from the previous
 * location.  As an added bonus, in non-NUMA cases, embedding is
 * generally a good idea TLB-wise because percpu area can piggy back
 * on the physical linear memory mapping which uses large page
 * mappings on applicable archs.
 */
#ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
EXPORT_SYMBOL(__per_cpu_offset);

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

	/*
	 * Always reserve area for module percpu variables.  That's
	 * what the legacy allocator did.
	 */
2177
	unit_size = pcpu_embed_first_chunk(PERCPU_MODULE_RESERVE,
2178
					   PERCPU_DYNAMIC_RESERVE);
2179 2180 2181 2182 2183 2184 2185 2186
	if (unit_size < 0)
		panic("Failed to initialized percpu areas.");

	delta = (unsigned long)pcpu_base_addr - (unsigned long)__per_cpu_start;
	for_each_possible_cpu(cpu)
		__per_cpu_offset[cpu] = delta + cpu * unit_size;
}
#endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */