memory.c 19.8 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2
/*
3
 * Memory subsystem support
4 5 6 7 8 9 10 11 12 13 14 15 16
 *
 * 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>
17
#include <linux/capability.h>
18 19 20 21
#include <linux/device.h>
#include <linux/memory.h>
#include <linux/memory_hotplug.h>
#include <linux/mm.h>
22
#include <linux/mutex.h>
23
#include <linux/stat.h>
24
#include <linux/slab.h>
25

Arun Sharma's avatar
Arun Sharma committed
26
#include <linux/atomic.h>
27
#include <linux/uaccess.h>
28

29 30
static DEFINE_MUTEX(mem_sysfs_mutex);

31
#define MEMORY_CLASS_NAME	"memory"
32

33 34
#define to_memory_block(dev) container_of(dev, struct memory_block, dev)

35 36 37 38 39 40
static int sections_per_block;

static inline int base_memory_block_id(int section_nr)
{
	return section_nr / sections_per_block;
}
41

42 43 44
static int memory_subsys_online(struct device *dev);
static int memory_subsys_offline(struct device *dev);

45
static struct bus_type memory_subsys = {
46
	.name = MEMORY_CLASS_NAME,
47
	.dev_name = MEMORY_CLASS_NAME,
48 49
	.online = memory_subsys_online,
	.offline = memory_subsys_offline,
50 51
};

52
static BLOCKING_NOTIFIER_HEAD(memory_chain);
53

54
int register_memory_notifier(struct notifier_block *nb)
55
{
56
	return blocking_notifier_chain_register(&memory_chain, nb);
57
}
58
EXPORT_SYMBOL(register_memory_notifier);
59

60
void unregister_memory_notifier(struct notifier_block *nb)
61
{
62
	blocking_notifier_chain_unregister(&memory_chain, nb);
63
}
64
EXPORT_SYMBOL(unregister_memory_notifier);
65

66 67 68 69 70 71 72 73 74 75 76 77 78 79
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);

80 81
static void memory_block_release(struct device *dev)
{
82
	struct memory_block *mem = to_memory_block(dev);
83 84 85 86

	kfree(mem);
}

87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106
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;
}

107 108 109 110 111
/*
 * use this as the physical section index that this memsection
 * uses.
 */

112 113
static ssize_t phys_index_show(struct device *dev,
			       struct device_attribute *attr, char *buf)
114
{
115
	struct memory_block *mem = to_memory_block(dev);
116 117 118 119 120 121
	unsigned long phys_index;

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

122 123 124
/*
 * Show whether the section of memory is likely to be hot-removable
 */
125 126
static ssize_t removable_show(struct device *dev, struct device_attribute *attr,
			      char *buf)
127
{
128 129
	unsigned long i, pfn;
	int ret = 1;
130
	struct memory_block *mem = to_memory_block(dev);
131

132 133 134
	if (mem->state != MEM_ONLINE)
		goto out;

135
	for (i = 0; i < sections_per_block; i++) {
136 137
		if (!present_section_nr(mem->start_section_nr + i))
			continue;
138
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
139 140 141
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

142
out:
143 144 145
	return sprintf(buf, "%d\n", ret);
}

146 147 148
/*
 * online, offline, going offline, etc.
 */
149 150
static ssize_t state_show(struct device *dev, struct device_attribute *attr,
			  char *buf)
151
{
152
	struct memory_block *mem = to_memory_block(dev);
153 154 155 156 157 158 159
	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) {
160 161 162 163 164 165 166 167 168 169 170 171 172 173
	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;
174 175 176 177 178
	}

	return len;
}

179
int memory_notify(unsigned long val, void *v)
180
{
181
	return blocking_notifier_call_chain(&memory_chain, val, v);
182 183
}

184 185 186 187 188
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

189
/*
190 191 192
 * The probe routines leave the pages uninitialized, just as the bootmem code
 * does. Make sure we do not access them, but instead use only information from
 * within sections.
193
 */
194
static bool pages_correctly_probed(unsigned long start_pfn)
195
{
196 197
	unsigned long section_nr = pfn_to_section_nr(start_pfn);
	unsigned long section_nr_end = section_nr + sections_per_block;
198 199 200 201 202 203 204
	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
	 */
205
	for (; section_nr < section_nr_end; section_nr++) {
206 207 208
		if (WARN_ON_ONCE(!pfn_valid(pfn)))
			return false;

209
		if (!present_section_nr(section_nr)) {
210
			pr_warn("section %ld pfn[%lx, %lx) not present\n",
211 212 213
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (!valid_section_nr(section_nr)) {
214
			pr_warn("section %ld pfn[%lx, %lx) no valid memmap\n",
215 216 217
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (online_section_nr(section_nr)) {
218
			pr_warn("section %ld pfn[%lx, %lx) is already online\n",
219
				section_nr, pfn, pfn + PAGES_PER_SECTION);
220 221
			return false;
		}
222
		pfn += PAGES_PER_SECTION;
223 224 225 226 227
	}

	return true;
}

228 229 230 231 232
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
233
memory_block_action(unsigned long phys_index, unsigned long action, int online_type)
234
{
235
	unsigned long start_pfn;
236
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
237 238
	int ret;

239
	start_pfn = section_nr_to_pfn(phys_index);
240

241
	switch (action) {
242
	case MEM_ONLINE:
243
		if (!pages_correctly_probed(start_pfn))
244 245 246 247 248 249 250 251 252 253 254
			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;
255 256 257 258 259
	}

	return ret;
}

260
static int memory_block_change_state(struct memory_block *mem,
261
		unsigned long to_state, unsigned long from_state_req)
262
{
263
	int ret = 0;
264

265 266
	if (mem->state != from_state_req)
		return -EINVAL;
267

268 269 270
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

271 272 273
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

274
	mem->state = ret ? from_state_req : to_state;
275

276 277
	return ret;
}
278

279
/* The device lock serializes operations on memory_subsys_[online|offline] */
280 281
static int memory_subsys_online(struct device *dev)
{
282
	struct memory_block *mem = to_memory_block(dev);
283
	int ret;
284

285 286
	if (mem->state == MEM_ONLINE)
		return 0;
287

288
	/*
289
	 * If we are called from state_store(), online_type will be
290 291 292 293
	 * set >= 0 Otherwise we were called from the device online
	 * attribute and need to set the online_type.
	 */
	if (mem->online_type < 0)
294
		mem->online_type = MMOP_ONLINE_KEEP;
295

296
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
297

298 299
	/* clear online_type */
	mem->online_type = -1;
300 301 302 303 304

	return ret;
}

static int memory_subsys_offline(struct device *dev)
305
{
306
	struct memory_block *mem = to_memory_block(dev);
307

308 309
	if (mem->state == MEM_OFFLINE)
		return 0;
310

311 312 313 314
	/* Can't offline block with non-present sections */
	if (mem->section_count != sections_per_block)
		return -EINVAL;

315
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
316
}
317

318 319
static ssize_t state_store(struct device *dev, struct device_attribute *attr,
			   const char *buf, size_t count)
320
{
321
	struct memory_block *mem = to_memory_block(dev);
322
	int ret, online_type;
323

324 325 326
	ret = lock_device_hotplug_sysfs();
	if (ret)
		return ret;
327

328
	if (sysfs_streq(buf, "online_kernel"))
329
		online_type = MMOP_ONLINE_KERNEL;
330
	else if (sysfs_streq(buf, "online_movable"))
331
		online_type = MMOP_ONLINE_MOVABLE;
332
	else if (sysfs_streq(buf, "online"))
333
		online_type = MMOP_ONLINE_KEEP;
334
	else if (sysfs_streq(buf, "offline"))
335
		online_type = MMOP_OFFLINE;
336 337 338 339
	else {
		ret = -EINVAL;
		goto err;
	}
340 341

	switch (online_type) {
342 343 344
	case MMOP_ONLINE_KERNEL:
	case MMOP_ONLINE_MOVABLE:
	case MMOP_ONLINE_KEEP:
345
		/* mem->online_type is protected by device_hotplug_lock */
346 347 348
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
349
	case MMOP_OFFLINE:
350 351 352 353
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
354 355
	}

356
err:
357
	unlock_device_hotplug();
358

359
	if (ret < 0)
360
		return ret;
361 362 363
	if (ret)
		return -EINVAL;

364 365 366 367 368 369 370 371 372 373 374 375
	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?
 */
376
static ssize_t phys_device_show(struct device *dev,
377
				struct device_attribute *attr, char *buf)
378
{
379
	struct memory_block *mem = to_memory_block(dev);
380 381 382
	return sprintf(buf, "%d\n", mem->phys_device);
}

383
#ifdef CONFIG_MEMORY_HOTREMOVE
384 385 386 387 388 389 390 391 392 393 394 395 396
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);
	}
}

397
static ssize_t valid_zones_show(struct device *dev,
398 399 400
				struct device_attribute *attr, char *buf)
{
	struct memory_block *mem = to_memory_block(dev);
401
	unsigned long start_pfn = section_nr_to_pfn(mem->start_section_nr);
402
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
403
	unsigned long valid_start_pfn, valid_end_pfn;
404
	struct zone *default_zone;
405
	int nid;
406

407 408 409 410 411
	/*
	 * 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) {
412 413 414 415 416 417 418 419
		/*
		 * 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))
			return sprintf(buf, "none\n");
		start_pfn = valid_start_pfn;
420 421
		strcat(buf, page_zone(pfn_to_page(start_pfn))->name);
		goto out;
422 423
	}

424
	nid = mem->nid;
425 426
	default_zone = zone_for_pfn_range(MMOP_ONLINE_KEEP, nid, start_pfn, nr_pages);
	strcat(buf, default_zone->name);
427

428 429 430 431
	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);
432
out:
433 434 435
	strcat(buf, "\n");

	return strlen(buf);
436
}
437
static DEVICE_ATTR_RO(valid_zones);
438 439
#endif

440 441 442 443
static DEVICE_ATTR_RO(phys_index);
static DEVICE_ATTR_RW(state);
static DEVICE_ATTR_RO(phys_device);
static DEVICE_ATTR_RO(removable);
444 445 446 447

/*
 * Block size attribute stuff
 */
448 449
static ssize_t block_size_bytes_show(struct device *dev,
				     struct device_attribute *attr, char *buf)
450
{
451
	return sprintf(buf, "%lx\n", get_memory_block_size());
452 453
}

454
static DEVICE_ATTR_RO(block_size_bytes);
455

456 457 458 459
/*
 * Memory auto online policy.
 */

460 461
static ssize_t auto_online_blocks_show(struct device *dev,
				       struct device_attribute *attr, char *buf)
462 463 464 465 466 467 468
{
	if (memhp_auto_online)
		return sprintf(buf, "online\n");
	else
		return sprintf(buf, "offline\n");
}

469 470 471
static ssize_t auto_online_blocks_store(struct device *dev,
					struct device_attribute *attr,
					const char *buf, size_t count)
472 473 474 475 476 477 478 479 480 481 482
{
	if (sysfs_streq(buf, "online"))
		memhp_auto_online = true;
	else if (sysfs_streq(buf, "offline"))
		memhp_auto_online = false;
	else
		return -EINVAL;

	return count;
}

483
static DEVICE_ATTR_RW(auto_online_blocks);
484

485 486 487 488 489 490 491
/*
 * 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
492 493
static ssize_t probe_store(struct device *dev, struct device_attribute *attr,
			   const char *buf, size_t count)
494 495
{
	u64 phys_addr;
496
	int nid, ret;
497
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
498

499 500 501
	ret = kstrtoull(buf, 0, &phys_addr);
	if (ret)
		return ret;
502

503 504 505
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

506 507 508 509
	ret = lock_device_hotplug_sysfs();
	if (ret)
		goto out;

510
	nid = memory_add_physaddr_to_nid(phys_addr);
511 512
	ret = __add_memory(nid, phys_addr,
			   MIN_MEMORY_BLOCK_SIZE * sections_per_block);
513

514 515
	if (ret)
		goto out;
516

517 518
	ret = count;
out:
519
	unlock_device_hotplug();
520
	return ret;
521 522
}

523
static DEVICE_ATTR_WO(probe);
524 525
#endif

526 527 528 529 530 531
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
532 533 534
static ssize_t soft_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
535 536 537 538 539
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
540
	if (kstrtoull(buf, 0, &pfn) < 0)
541 542 543 544 545 546 547 548 549
		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. */
550 551 552
static ssize_t hard_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
553 554 555 556 557
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
558
	if (kstrtoull(buf, 0, &pfn) < 0)
559 560
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
561
	ret = memory_failure(pfn, 0);
562 563 564
	return ret ? ret : count;
}

565 566
static DEVICE_ATTR_WO(soft_offline_page);
static DEVICE_ATTR_WO(hard_offline_page);
567 568
#endif

569 570 571 572 573
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
574 575 576 577
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
578

579 580 581 582
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
583 584
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
585
{
586
	int block_id = base_memory_block_id(__section_nr(section));
587 588
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
589

590 591 592 593
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
594
		return NULL;
595
	return to_memory_block(dev);
596 597
}

598 599 600 601 602 603
/*
 * 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.
 *
604
 * This could be made generic for all device subsystems.
605 606 607 608 609 610
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

611 612 613 614 615
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
616 617 618
#ifdef CONFIG_MEMORY_HOTREMOVE
	&dev_attr_valid_zones.attr,
#endif
619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
	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)
{
637 638
	int ret;

639 640 641 642
	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;
643
	memory->dev.offline = memory->state == MEM_OFFLINE;
644

645 646 647 648 649
	ret = device_register(&memory->dev);
	if (ret)
		put_device(&memory->dev);

	return ret;
650 651
}

652 653
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
654
{
655
	struct memory_block *mem;
656
	unsigned long start_pfn;
657
	int scn_nr;
658 659
	int ret = 0;

660
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
661 662 663
	if (!mem)
		return -ENOMEM;

664
	scn_nr = __section_nr(section);
665 666 667
	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;
668
	mem->state = state;
669
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
670 671
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

672 673 674 675 676 677
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

678
static int add_memory_block(int base_section_nr)
679
{
680 681
	struct memory_block *mem;
	int i, ret, section_count = 0, section_nr;
682

683
	for (i = base_section_nr;
684
	     i < base_section_nr + sections_per_block;
685 686 687 688 689 690
	     i++) {
		if (!present_section_nr(i))
			continue;
		if (section_count == 0)
			section_nr = i;
		section_count++;
691 692
	}

693 694 695 696 697 698 699
	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;
700 701
}

702 703 704 705
/*
 * need an interface for the VM to add new memory regions,
 * but without onlining it.
 */
706
int hotplug_memory_register(int nid, struct mem_section *section)
707
{
708 709
	int ret = 0;
	struct memory_block *mem;
710 711 712

	mutex_lock(&mem_sysfs_mutex);

713 714 715 716 717 718 719 720
	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;
721
		mem->section_count++;
722 723 724 725
	}

out:
	mutex_unlock(&mem_sysfs_mutex);
726
	return ret;
727 728 729 730 731 732 733 734
}

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

735
	/* drop the ref. we got in remove_memory_section() */
736
	put_device(&memory->dev);
737 738 739
	device_unregister(&memory->dev);
}

740
static int remove_memory_section(unsigned long node_id,
741
			       struct mem_section *section, int phys_device)
742 743 744
{
	struct memory_block *mem;

745
	mutex_lock(&mem_sysfs_mutex);
746 747 748 749 750

	/*
	 * Some users of the memory hotplug do not want/need memblock to
	 * track all sections. Skip over those.
	 */
751
	mem = find_memory_block(section);
752 753 754
	if (!mem)
		goto out_unlock;

755
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
756 757

	mem->section_count--;
758
	if (mem->section_count == 0)
759
		unregister_memory(mem);
760
	else
761
		put_device(&mem->dev);
762

763
out_unlock:
764
	mutex_unlock(&mem_sysfs_mutex);
765 766 767 768 769
	return 0;
}

int unregister_memory_section(struct mem_section *section)
{
770
	if (!present_section(section))
771 772
		return -EINVAL;

773
	return remove_memory_section(0, section, 0);
774
}
775
#endif /* CONFIG_MEMORY_HOTREMOVE */
776

777 778 779 780 781 782
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

783 784 785 786 787 788 789 790 791 792 793
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,
794
	&dev_attr_auto_online_blocks.attr,
795 796 797 798 799 800 801 802 803 804 805 806
	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,
};

807 808 809 810 811 812 813
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
814
	int err;
815
	unsigned long block_sz;
816

817
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
818 819
	if (ret)
		goto out;
820

821 822 823
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

824 825 826 827
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
828
	mutex_lock(&mem_sysfs_mutex);
829 830
	for (i = 0; i <= __highest_present_section_nr;
		i += sections_per_block) {
831
		err = add_memory_block(i);
832 833
		if (!ret)
			ret = err;
834
	}
835
	mutex_unlock(&mem_sysfs_mutex);
836

837 838
out:
	if (ret)
839
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
840 841
	return ret;
}