memory.c 20.2 KB
Newer Older
1
/*
2
 * Memory subsystem support
3 4 5 6 7 8 9 10 11 12 13 14 15
 *
 * Written by Matt Tolentino <matthew.e.tolentino@intel.com>
 *            Dave Hansen <haveblue@us.ibm.com>
 *
 * This file provides the necessary infrastructure to represent
 * a SPARSEMEM-memory-model system's physical memory in /sysfs.
 * All arch-independent code that assumes MEMORY_HOTPLUG requires
 * SPARSEMEM should be contained here, or in mm/memory_hotplug.c.
 */

#include <linux/module.h>
#include <linux/init.h>
#include <linux/topology.h>
16
#include <linux/capability.h>
17 18 19 20
#include <linux/device.h>
#include <linux/memory.h>
#include <linux/memory_hotplug.h>
#include <linux/mm.h>
21
#include <linux/mutex.h>
22
#include <linux/stat.h>
23
#include <linux/slab.h>
24

Arun Sharma's avatar
Arun Sharma committed
25
#include <linux/atomic.h>
26
#include <linux/uaccess.h>
27

28 29
static DEFINE_MUTEX(mem_sysfs_mutex);

30
#define MEMORY_CLASS_NAME	"memory"
31

32 33
#define to_memory_block(dev) container_of(dev, struct memory_block, dev)

34 35 36 37 38 39
static int sections_per_block;

static inline int base_memory_block_id(int section_nr)
{
	return section_nr / sections_per_block;
}
40

41 42 43
static int memory_subsys_online(struct device *dev);
static int memory_subsys_offline(struct device *dev);

44
static struct bus_type memory_subsys = {
45
	.name = MEMORY_CLASS_NAME,
46
	.dev_name = MEMORY_CLASS_NAME,
47 48
	.online = memory_subsys_online,
	.offline = memory_subsys_offline,
49 50
};

51
static BLOCKING_NOTIFIER_HEAD(memory_chain);
52

53
int register_memory_notifier(struct notifier_block *nb)
54
{
55
	return blocking_notifier_chain_register(&memory_chain, nb);
56
}
57
EXPORT_SYMBOL(register_memory_notifier);
58

59
void unregister_memory_notifier(struct notifier_block *nb)
60
{
61
	blocking_notifier_chain_unregister(&memory_chain, nb);
62
}
63
EXPORT_SYMBOL(unregister_memory_notifier);
64

65 66 67 68 69 70 71 72 73 74 75 76 77 78
static ATOMIC_NOTIFIER_HEAD(memory_isolate_chain);

int register_memory_isolate_notifier(struct notifier_block *nb)
{
	return atomic_notifier_chain_register(&memory_isolate_chain, nb);
}
EXPORT_SYMBOL(register_memory_isolate_notifier);

void unregister_memory_isolate_notifier(struct notifier_block *nb)
{
	atomic_notifier_chain_unregister(&memory_isolate_chain, nb);
}
EXPORT_SYMBOL(unregister_memory_isolate_notifier);

79 80
static void memory_block_release(struct device *dev)
{
81
	struct memory_block *mem = to_memory_block(dev);
82 83 84 85

	kfree(mem);
}

86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
unsigned long __weak memory_block_size_bytes(void)
{
	return MIN_MEMORY_BLOCK_SIZE;
}

static unsigned long get_memory_block_size(void)
{
	unsigned long block_sz;

	block_sz = memory_block_size_bytes();

	/* Validate blk_sz is a power of 2 and not less than section size */
	if ((block_sz & (block_sz - 1)) || (block_sz < MIN_MEMORY_BLOCK_SIZE)) {
		WARN_ON(1);
		block_sz = MIN_MEMORY_BLOCK_SIZE;
	}

	return block_sz;
}

106 107 108 109 110
/*
 * use this as the physical section index that this memsection
 * uses.
 */

111 112
static ssize_t show_mem_start_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
113
{
114
	struct memory_block *mem = to_memory_block(dev);
115 116 117 118 119 120
	unsigned long phys_index;

	phys_index = mem->start_section_nr / sections_per_block;
	return sprintf(buf, "%08lx\n", phys_index);
}

121 122 123
/*
 * Show whether the section of memory is likely to be hot-removable
 */
124 125
static ssize_t show_mem_removable(struct device *dev,
			struct device_attribute *attr, char *buf)
126
{
127 128
	unsigned long i, pfn;
	int ret = 1;
129
	struct memory_block *mem = to_memory_block(dev);
130

131 132 133
	if (mem->state != MEM_ONLINE)
		goto out;

134
	for (i = 0; i < sections_per_block; i++) {
135 136
		if (!present_section_nr(mem->start_section_nr + i))
			continue;
137
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
138 139 140
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

141
out:
142 143 144
	return sprintf(buf, "%d\n", ret);
}

145 146 147
/*
 * online, offline, going offline, etc.
 */
148 149
static ssize_t show_mem_state(struct device *dev,
			struct device_attribute *attr, char *buf)
150
{
151
	struct memory_block *mem = to_memory_block(dev);
152 153 154 155 156 157 158
	ssize_t len = 0;

	/*
	 * We can probably put these states in a nice little array
	 * so that they're not open-coded
	 */
	switch (mem->state) {
159 160 161 162 163 164 165 166 167 168 169 170 171 172
	case MEM_ONLINE:
		len = sprintf(buf, "online\n");
		break;
	case MEM_OFFLINE:
		len = sprintf(buf, "offline\n");
		break;
	case MEM_GOING_OFFLINE:
		len = sprintf(buf, "going-offline\n");
		break;
	default:
		len = sprintf(buf, "ERROR-UNKNOWN-%ld\n",
				mem->state);
		WARN_ON(1);
		break;
173 174 175 176 177
	}

	return len;
}

178
int memory_notify(unsigned long val, void *v)
179
{
180
	return blocking_notifier_call_chain(&memory_chain, val, v);
181 182
}

183 184 185 186 187
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

188 189 190 191
/*
 * The probe routines leave the pages reserved, just as the bootmem code does.
 * Make sure they're still that way.
 */
192
static bool pages_correctly_reserved(unsigned long start_pfn)
193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222
{
	int i, j;
	struct page *page;
	unsigned long pfn = start_pfn;

	/*
	 * memmap between sections is not contiguous except with
	 * SPARSEMEM_VMEMMAP. We lookup the page once per section
	 * and assume memmap is contiguous within each section
	 */
	for (i = 0; i < sections_per_block; i++, pfn += PAGES_PER_SECTION) {
		if (WARN_ON_ONCE(!pfn_valid(pfn)))
			return false;
		page = pfn_to_page(pfn);

		for (j = 0; j < PAGES_PER_SECTION; j++) {
			if (PageReserved(page + j))
				continue;

			printk(KERN_WARNING "section number %ld page number %d "
				"not reserved, was it already online?\n",
				pfn_to_section_nr(pfn), j);

			return false;
		}
	}

	return true;
}

223 224 225
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
226
 * Must already be protected by mem_hotplug_begin().
227 228
 */
static int
229
memory_block_action(unsigned long phys_index, unsigned long action, int online_type)
230
{
231
	unsigned long start_pfn;
232
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
233 234
	int ret;

235
	start_pfn = section_nr_to_pfn(phys_index);
236

237
	switch (action) {
238 239 240 241 242 243 244 245 246 247 248 249 250
	case MEM_ONLINE:
		if (!pages_correctly_reserved(start_pfn))
			return -EBUSY;

		ret = online_pages(start_pfn, nr_pages, online_type);
		break;
	case MEM_OFFLINE:
		ret = offline_pages(start_pfn, nr_pages);
		break;
	default:
		WARN(1, KERN_WARNING "%s(%ld, %ld) unknown action: "
		     "%ld\n", __func__, phys_index, action, action);
		ret = -EINVAL;
251 252 253 254 255
	}

	return ret;
}

256
static int memory_block_change_state(struct memory_block *mem,
257
		unsigned long to_state, unsigned long from_state_req)
258
{
259
	int ret = 0;
260

261 262
	if (mem->state != from_state_req)
		return -EINVAL;
263

264 265 266
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

267 268 269
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

270
	mem->state = ret ? from_state_req : to_state;
271

272 273
	return ret;
}
274

275
/* The device lock serializes operations on memory_subsys_[online|offline] */
276 277
static int memory_subsys_online(struct device *dev)
{
278
	struct memory_block *mem = to_memory_block(dev);
279
	int ret;
280

281 282
	if (mem->state == MEM_ONLINE)
		return 0;
283

284 285 286 287 288 289
	/*
	 * If we are called from store_mem_state(), online_type will be
	 * set >= 0 Otherwise we were called from the device online
	 * attribute and need to set the online_type.
	 */
	if (mem->online_type < 0)
290
		mem->online_type = MMOP_ONLINE_KEEP;
291

292
	/* Already under protection of mem_hotplug_begin() */
293
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
294

295 296
	/* clear online_type */
	mem->online_type = -1;
297 298 299 300 301

	return ret;
}

static int memory_subsys_offline(struct device *dev)
302
{
303
	struct memory_block *mem = to_memory_block(dev);
304

305 306
	if (mem->state == MEM_OFFLINE)
		return 0;
307

308 309 310 311
	/* Can't offline block with non-present sections */
	if (mem->section_count != sections_per_block)
		return -EINVAL;

312
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
313
}
314

315
static ssize_t
316 317
store_mem_state(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
318
{
319
	struct memory_block *mem = to_memory_block(dev);
320
	int ret, online_type;
321

322 323 324
	ret = lock_device_hotplug_sysfs();
	if (ret)
		return ret;
325

326
	if (sysfs_streq(buf, "online_kernel"))
327
		online_type = MMOP_ONLINE_KERNEL;
328
	else if (sysfs_streq(buf, "online_movable"))
329
		online_type = MMOP_ONLINE_MOVABLE;
330
	else if (sysfs_streq(buf, "online"))
331
		online_type = MMOP_ONLINE_KEEP;
332
	else if (sysfs_streq(buf, "offline"))
333
		online_type = MMOP_OFFLINE;
334 335 336 337
	else {
		ret = -EINVAL;
		goto err;
	}
338

339 340 341 342 343 344 345 346 347
	/*
	 * Memory hotplug needs to hold mem_hotplug_begin() for probe to find
	 * the correct memory block to online before doing device_online(dev),
	 * which will take dev->mutex.  Take the lock early to prevent an
	 * inversion, memory_subsys_online() callbacks will be implemented by
	 * assuming it's already protected.
	 */
	mem_hotplug_begin();

348
	switch (online_type) {
349 350 351
	case MMOP_ONLINE_KERNEL:
	case MMOP_ONLINE_MOVABLE:
	case MMOP_ONLINE_KEEP:
352 353 354
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
355
	case MMOP_OFFLINE:
356 357 358 359
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
360 361
	}

362
	mem_hotplug_done();
363
err:
364
	unlock_device_hotplug();
365

366
	if (ret < 0)
367
		return ret;
368 369 370
	if (ret)
		return -EINVAL;

371 372 373 374 375 376 377 378 379 380 381 382
	return count;
}

/*
 * phys_device is a bad name for this.  What I really want
 * is a way to differentiate between memory ranges that
 * are part of physical devices that constitute
 * a complete removable unit or fru.
 * i.e. do these ranges belong to the same physical device,
 * s.t. if I offline all of these sections I can then
 * remove the physical device?
 */
383 384
static ssize_t show_phys_device(struct device *dev,
				struct device_attribute *attr, char *buf)
385
{
386
	struct memory_block *mem = to_memory_block(dev);
387 388 389
	return sprintf(buf, "%d\n", mem->phys_device);
}

390
#ifdef CONFIG_MEMORY_HOTREMOVE
391 392 393 394 395 396 397 398 399 400 401 402 403
static void print_allowed_zone(char *buf, int nid, unsigned long start_pfn,
		unsigned long nr_pages, int online_type,
		struct zone *default_zone)
{
	struct zone *zone;

	zone = zone_for_pfn_range(online_type, nid, start_pfn, nr_pages);
	if (zone != default_zone) {
		strcat(buf, " ");
		strcat(buf, zone->name);
	}
}

404 405 406 407
static ssize_t show_valid_zones(struct device *dev,
				struct device_attribute *attr, char *buf)
{
	struct memory_block *mem = to_memory_block(dev);
408
	unsigned long start_pfn = section_nr_to_pfn(mem->start_section_nr);
409
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
410
	unsigned long valid_start_pfn, valid_end_pfn;
411
	struct zone *default_zone;
412
	int nid;
413

414 415 416 417 418
	/*
	 * The block contains more than one zone can not be offlined.
	 * This can happen e.g. for ZONE_DMA and ZONE_DMA32
	 */
	if (!test_pages_in_a_zone(start_pfn, start_pfn + nr_pages, &valid_start_pfn, &valid_end_pfn))
419 420
		return sprintf(buf, "none\n");

421 422
	start_pfn = valid_start_pfn;
	nr_pages = valid_end_pfn - start_pfn;
423

424 425 426 427 428 429 430
	/*
	 * Check the existing zone. Make sure that we do that only on the
	 * online nodes otherwise the page_zone is not reliable
	 */
	if (mem->state == MEM_ONLINE) {
		strcat(buf, page_zone(pfn_to_page(start_pfn))->name);
		goto out;
431 432
	}

433
	nid = pfn_to_nid(start_pfn);
434 435
	default_zone = zone_for_pfn_range(MMOP_ONLINE_KEEP, nid, start_pfn, nr_pages);
	strcat(buf, default_zone->name);
436

437 438 439 440
	print_allowed_zone(buf, nid, start_pfn, nr_pages, MMOP_ONLINE_KERNEL,
			default_zone);
	print_allowed_zone(buf, nid, start_pfn, nr_pages, MMOP_ONLINE_MOVABLE,
			default_zone);
441
out:
442 443 444
	strcat(buf, "\n");

	return strlen(buf);
445 446 447 448
}
static DEVICE_ATTR(valid_zones, 0444, show_valid_zones, NULL);
#endif

449 450 451 452
static DEVICE_ATTR(phys_index, 0444, show_mem_start_phys_index, NULL);
static DEVICE_ATTR(state, 0644, show_mem_state, store_mem_state);
static DEVICE_ATTR(phys_device, 0444, show_phys_device, NULL);
static DEVICE_ATTR(removable, 0444, show_mem_removable, NULL);
453 454 455 456 457

/*
 * Block size attribute stuff
 */
static ssize_t
458
print_block_size(struct device *dev, struct device_attribute *attr,
459
		 char *buf)
460
{
461
	return sprintf(buf, "%lx\n", get_memory_block_size());
462 463
}

464
static DEVICE_ATTR(block_size_bytes, 0444, print_block_size, NULL);
465

466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
/*
 * Memory auto online policy.
 */

static ssize_t
show_auto_online_blocks(struct device *dev, struct device_attribute *attr,
			char *buf)
{
	if (memhp_auto_online)
		return sprintf(buf, "online\n");
	else
		return sprintf(buf, "offline\n");
}

static ssize_t
store_auto_online_blocks(struct device *dev, struct device_attribute *attr,
			 const char *buf, size_t count)
{
	if (sysfs_streq(buf, "online"))
		memhp_auto_online = true;
	else if (sysfs_streq(buf, "offline"))
		memhp_auto_online = false;
	else
		return -EINVAL;

	return count;
}

static DEVICE_ATTR(auto_online_blocks, 0644, show_auto_online_blocks,
		   store_auto_online_blocks);

497 498 499 500 501 502 503 504
/*
 * Some architectures will have custom drivers to do this, and
 * will not need to do it from userspace.  The fake hot-add code
 * as well as ppc64 will do all of their discovery in userspace
 * and will require this interface.
 */
#ifdef CONFIG_ARCH_MEMORY_PROBE
static ssize_t
505
memory_probe_store(struct device *dev, struct device_attribute *attr,
506
		   const char *buf, size_t count)
507 508
{
	u64 phys_addr;
509
	int nid, ret;
510
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
511

512 513 514
	ret = kstrtoull(buf, 0, &phys_addr);
	if (ret)
		return ret;
515

516 517 518
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

519 520 521
	nid = memory_add_physaddr_to_nid(phys_addr);
	ret = add_memory(nid, phys_addr,
			 MIN_MEMORY_BLOCK_SIZE * sections_per_block);
522

523 524
	if (ret)
		goto out;
525

526 527 528
	ret = count;
out:
	return ret;
529 530
}

531
static DEVICE_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
532 533
#endif

534 535 536 537 538 539 540
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
541 542
store_soft_offline_page(struct device *dev,
			struct device_attribute *attr,
543
			const char *buf, size_t count)
544 545 546 547 548
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
549
	if (kstrtoull(buf, 0, &pfn) < 0)
550 551 552 553 554 555 556 557 558 559
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
	if (!pfn_valid(pfn))
		return -ENXIO;
	ret = soft_offline_page(pfn_to_page(pfn), 0);
	return ret == 0 ? count : ret;
}

/* Forcibly offline a page, including killing processes. */
static ssize_t
560 561
store_hard_offline_page(struct device *dev,
			struct device_attribute *attr,
562
			const char *buf, size_t count)
563 564 565 566 567
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
568
	if (kstrtoull(buf, 0, &pfn) < 0)
569 570
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
571
	ret = memory_failure(pfn, 0, 0);
572 573 574
	return ret ? ret : count;
}

575 576
static DEVICE_ATTR(soft_offline_page, S_IWUSR, NULL, store_soft_offline_page);
static DEVICE_ATTR(hard_offline_page, S_IWUSR, NULL, store_hard_offline_page);
577 578
#endif

579 580 581 582 583
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
584 585 586 587
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
588

589 590 591 592
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
593 594
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
595
{
596
	int block_id = base_memory_block_id(__section_nr(section));
597 598
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
599

600 601 602 603
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
604
		return NULL;
605
	return to_memory_block(dev);
606 607
}

608 609 610 611 612 613
/*
 * For now, we have a linear search to go find the appropriate
 * memory_block corresponding to a particular phys_index. If
 * this gets to be a real problem, we can always use a radix
 * tree or something here.
 *
614
 * This could be made generic for all device subsystems.
615 616 617 618 619 620
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

621 622 623 624 625
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
626 627 628
#ifdef CONFIG_MEMORY_HOTREMOVE
	&dev_attr_valid_zones.attr,
#endif
629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650
	NULL
};

static struct attribute_group memory_memblk_attr_group = {
	.attrs = memory_memblk_attrs,
};

static const struct attribute_group *memory_memblk_attr_groups[] = {
	&memory_memblk_attr_group,
	NULL,
};

/*
 * register_memory - Setup a sysfs device for a memory block
 */
static
int register_memory(struct memory_block *memory)
{
	memory->dev.bus = &memory_subsys;
	memory->dev.id = memory->start_section_nr / sections_per_block;
	memory->dev.release = memory_block_release;
	memory->dev.groups = memory_memblk_attr_groups;
651
	memory->dev.offline = memory->state == MEM_OFFLINE;
652

653
	return device_register(&memory->dev);
654 655
}

656 657
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
658
{
659
	struct memory_block *mem;
660
	unsigned long start_pfn;
661
	int scn_nr;
662 663
	int ret = 0;

664
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
665 666 667
	if (!mem)
		return -ENOMEM;

668
	scn_nr = __section_nr(section);
669 670 671
	mem->start_section_nr =
			base_memory_block_id(scn_nr) * sections_per_block;
	mem->end_section_nr = mem->start_section_nr + sections_per_block - 1;
672
	mem->state = state;
673
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
674 675
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

676 677 678 679 680 681
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

682
static int add_memory_block(int base_section_nr)
683
{
684 685
	struct memory_block *mem;
	int i, ret, section_count = 0, section_nr;
686

687 688 689 690 691 692 693 694
	for (i = base_section_nr;
	     (i < base_section_nr + sections_per_block) && i < NR_MEM_SECTIONS;
	     i++) {
		if (!present_section_nr(i))
			continue;
		if (section_count == 0)
			section_nr = i;
		section_count++;
695 696
	}

697 698 699 700 701 702 703
	if (section_count == 0)
		return 0;
	ret = init_memory_block(&mem, __nr_to_section(section_nr), MEM_ONLINE);
	if (ret)
		return ret;
	mem->section_count = section_count;
	return 0;
704 705
}

706 707 708 709 710 711
/*
 * need an interface for the VM to add new memory regions,
 * but without onlining it.
 */
int register_new_memory(int nid, struct mem_section *section)
{
712 713
	int ret = 0;
	struct memory_block *mem;
714 715 716

	mutex_lock(&mem_sysfs_mutex);

717 718 719 720 721 722 723 724
	mem = find_memory_block(section);
	if (mem) {
		mem->section_count++;
		put_device(&mem->dev);
	} else {
		ret = init_memory_block(&mem, section, MEM_OFFLINE);
		if (ret)
			goto out;
725
		mem->section_count++;
726 727 728 729 730 731
	}

	if (mem->section_count == sections_per_block)
		ret = register_mem_sect_under_node(mem, nid);
out:
	mutex_unlock(&mem_sysfs_mutex);
732
	return ret;
733 734 735 736 737 738 739 740 741
}

#ifdef CONFIG_MEMORY_HOTREMOVE
static void
unregister_memory(struct memory_block *memory)
{
	BUG_ON(memory->dev.bus != &memory_subsys);

	/* drop the ref. we got in remove_memory_block() */
742
	put_device(&memory->dev);
743 744 745
	device_unregister(&memory->dev);
}

746
static int remove_memory_section(unsigned long node_id,
747
			       struct mem_section *section, int phys_device)
748 749 750
{
	struct memory_block *mem;

751
	mutex_lock(&mem_sysfs_mutex);
752 753 754 755 756

	/*
	 * Some users of the memory hotplug do not want/need memblock to
	 * track all sections. Skip over those.
	 */
757
	mem = find_memory_block(section);
758 759 760
	if (!mem)
		goto out_unlock;

761
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
762 763

	mem->section_count--;
764
	if (mem->section_count == 0)
765
		unregister_memory(mem);
766
	else
767
		put_device(&mem->dev);
768

769
out_unlock:
770
	mutex_unlock(&mem_sysfs_mutex);
771 772 773 774 775
	return 0;
}

int unregister_memory_section(struct mem_section *section)
{
776
	if (!present_section(section))
777 778
		return -EINVAL;

779
	return remove_memory_section(0, section, 0);
780
}
781
#endif /* CONFIG_MEMORY_HOTREMOVE */
782

783 784 785 786 787 788
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

789 790 791 792 793 794 795 796 797 798 799
static struct attribute *memory_root_attrs[] = {
#ifdef CONFIG_ARCH_MEMORY_PROBE
	&dev_attr_probe.attr,
#endif

#ifdef CONFIG_MEMORY_FAILURE
	&dev_attr_soft_offline_page.attr,
	&dev_attr_hard_offline_page.attr,
#endif

	&dev_attr_block_size_bytes.attr,
800
	&dev_attr_auto_online_blocks.attr,
801 802 803 804 805 806 807 808 809 810 811 812
	NULL
};

static struct attribute_group memory_root_attr_group = {
	.attrs = memory_root_attrs,
};

static const struct attribute_group *memory_root_attr_groups[] = {
	&memory_root_attr_group,
	NULL,
};

813 814 815 816 817 818 819
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
820
	int err;
821
	unsigned long block_sz;
822

823
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
824 825
	if (ret)
		goto out;
826

827 828 829
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

830 831 832 833
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
834
	mutex_lock(&mem_sysfs_mutex);
835
	for (i = 0; i < NR_MEM_SECTIONS; i += sections_per_block) {
836 837 838 839
		/* Don't iterate over sections we know are !present: */
		if (i > __highest_present_section_nr)
			break;

840
		err = add_memory_block(i);
841 842
		if (!ret)
			ret = err;
843
	}
844
	mutex_unlock(&mem_sysfs_mutex);
845

846 847
out:
	if (ret)
848
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
849 850
	return ret;
}