memory.c 20.6 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 show_mem_start_phys_index(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 show_mem_removable(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 show_mem_state(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;

209 210 211 212 213 214 215 216 217 218 219
		if (!present_section_nr(section_nr)) {
			pr_warn("section %ld pfn[%lx, %lx) not present",
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (!valid_section_nr(section_nr)) {
			pr_warn("section %ld pfn[%lx, %lx) no valid memmap",
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (online_section_nr(section_nr)) {
			pr_warn("section %ld pfn[%lx, %lx) is already online",
				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
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
231
 * Must already be protected by mem_hotplug_begin().
232 233
 */
static int
234
memory_block_action(unsigned long phys_index, unsigned long action, int online_type)
235
{
236
	unsigned long start_pfn;
237
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
238 239
	int ret;

240
	start_pfn = section_nr_to_pfn(phys_index);
241

242
	switch (action) {
243
	case MEM_ONLINE:
244
		if (!pages_correctly_probed(start_pfn))
245 246 247 248 249 250 251 252 253 254 255
			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;
256 257 258 259 260
	}

	return ret;
}

261
static int memory_block_change_state(struct memory_block *mem,
262
		unsigned long to_state, unsigned long from_state_req)
263
{
264
	int ret = 0;
265

266 267
	if (mem->state != from_state_req)
		return -EINVAL;
268

269 270 271
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

272 273 274
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

275
	mem->state = ret ? from_state_req : to_state;
276

277 278
	return ret;
}
279

280
/* The device lock serializes operations on memory_subsys_[online|offline] */
281 282
static int memory_subsys_online(struct device *dev)
{
283
	struct memory_block *mem = to_memory_block(dev);
284
	int ret;
285

286 287
	if (mem->state == MEM_ONLINE)
		return 0;
288

289 290 291 292 293 294
	/*
	 * 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)
295
		mem->online_type = MMOP_ONLINE_KEEP;
296

297
	/* Already under protection of mem_hotplug_begin() */
298
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
299

300 301
	/* clear online_type */
	mem->online_type = -1;
302 303 304 305 306

	return ret;
}

static int memory_subsys_offline(struct device *dev)
307
{
308
	struct memory_block *mem = to_memory_block(dev);
309

310 311
	if (mem->state == MEM_OFFLINE)
		return 0;
312

313 314 315 316
	/* Can't offline block with non-present sections */
	if (mem->section_count != sections_per_block)
		return -EINVAL;

317
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
318
}
319

320
static ssize_t
321 322
store_mem_state(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
323
{
324
	struct memory_block *mem = to_memory_block(dev);
325
	int ret, online_type;
326

327 328 329
	ret = lock_device_hotplug_sysfs();
	if (ret)
		return ret;
330

331
	if (sysfs_streq(buf, "online_kernel"))
332
		online_type = MMOP_ONLINE_KERNEL;
333
	else if (sysfs_streq(buf, "online_movable"))
334
		online_type = MMOP_ONLINE_MOVABLE;
335
	else if (sysfs_streq(buf, "online"))
336
		online_type = MMOP_ONLINE_KEEP;
337
	else if (sysfs_streq(buf, "offline"))
338
		online_type = MMOP_OFFLINE;
339 340 341 342
	else {
		ret = -EINVAL;
		goto err;
	}
343

344 345 346 347 348 349 350 351 352
	/*
	 * 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();

353
	switch (online_type) {
354 355 356
	case MMOP_ONLINE_KERNEL:
	case MMOP_ONLINE_MOVABLE:
	case MMOP_ONLINE_KEEP:
357 358 359
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
360
	case MMOP_OFFLINE:
361 362 363 364
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
365 366
	}

367
	mem_hotplug_done();
368
err:
369
	unlock_device_hotplug();
370

371
	if (ret < 0)
372
		return ret;
373 374 375
	if (ret)
		return -EINVAL;

376 377 378 379 380 381 382 383 384 385 386 387
	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?
 */
388 389
static ssize_t show_phys_device(struct device *dev,
				struct device_attribute *attr, char *buf)
390
{
391
	struct memory_block *mem = to_memory_block(dev);
392 393 394
	return sprintf(buf, "%d\n", mem->phys_device);
}

395
#ifdef CONFIG_MEMORY_HOTREMOVE
396 397 398 399 400 401 402 403 404 405 406 407 408
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);
	}
}

409 410 411 412
static ssize_t show_valid_zones(struct device *dev,
				struct device_attribute *attr, char *buf)
{
	struct memory_block *mem = to_memory_block(dev);
413
	unsigned long start_pfn = section_nr_to_pfn(mem->start_section_nr);
414
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
415
	unsigned long valid_start_pfn, valid_end_pfn;
416
	struct zone *default_zone;
417
	int nid;
418

419 420 421 422 423
	/*
	 * 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))
424 425
		return sprintf(buf, "none\n");

426 427
	start_pfn = valid_start_pfn;
	nr_pages = valid_end_pfn - start_pfn;
428

429 430 431 432 433 434 435
	/*
	 * 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;
436 437
	}

438
	nid = pfn_to_nid(start_pfn);
439 440
	default_zone = zone_for_pfn_range(MMOP_ONLINE_KEEP, nid, start_pfn, nr_pages);
	strcat(buf, default_zone->name);
441

442 443 444 445
	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);
446
out:
447 448 449
	strcat(buf, "\n");

	return strlen(buf);
450 451 452 453
}
static DEVICE_ATTR(valid_zones, 0444, show_valid_zones, NULL);
#endif

454 455 456 457
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);
458 459 460 461 462

/*
 * Block size attribute stuff
 */
static ssize_t
463
print_block_size(struct device *dev, struct device_attribute *attr,
464
		 char *buf)
465
{
466
	return sprintf(buf, "%lx\n", get_memory_block_size());
467 468
}

469
static DEVICE_ATTR(block_size_bytes, 0444, print_block_size, NULL);
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
/*
 * 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);

502 503 504 505 506 507 508 509
/*
 * 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
510
memory_probe_store(struct device *dev, struct device_attribute *attr,
511
		   const char *buf, size_t count)
512 513
{
	u64 phys_addr;
514
	int nid, ret;
515
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
516

517 518 519
	ret = kstrtoull(buf, 0, &phys_addr);
	if (ret)
		return ret;
520

521 522 523
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

524 525 526
	nid = memory_add_physaddr_to_nid(phys_addr);
	ret = add_memory(nid, phys_addr,
			 MIN_MEMORY_BLOCK_SIZE * sections_per_block);
527

528 529
	if (ret)
		goto out;
530

531 532 533
	ret = count;
out:
	return ret;
534 535
}

536
static DEVICE_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
537 538
#endif

539 540 541 542 543 544 545
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
546 547
store_soft_offline_page(struct device *dev,
			struct device_attribute *attr,
548
			const char *buf, size_t count)
549 550 551 552 553
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
554
	if (kstrtoull(buf, 0, &pfn) < 0)
555 556 557 558 559 560 561 562 563 564
		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
565 566
store_hard_offline_page(struct device *dev,
			struct device_attribute *attr,
567
			const char *buf, size_t count)
568 569 570 571 572
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
573
	if (kstrtoull(buf, 0, &pfn) < 0)
574 575
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
576
	ret = memory_failure(pfn, 0);
577 578 579
	return ret ? ret : count;
}

580 581
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);
582 583
#endif

584 585 586 587 588
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
589 590 591 592
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
593

594 595 596 597
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
598 599
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
600
{
601
	int block_id = base_memory_block_id(__section_nr(section));
602 603
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
604

605 606 607 608
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
609
		return NULL;
610
	return to_memory_block(dev);
611 612
}

613 614 615 616 617 618
/*
 * 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.
 *
619
 * This could be made generic for all device subsystems.
620 621 622 623 624 625
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

626 627 628 629 630
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
631 632 633
#ifdef CONFIG_MEMORY_HOTREMOVE
	&dev_attr_valid_zones.attr,
#endif
634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
	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;
656
	memory->dev.offline = memory->state == MEM_OFFLINE;
657

658
	return device_register(&memory->dev);
659 660
}

661 662
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
663
{
664
	struct memory_block *mem;
665
	unsigned long start_pfn;
666
	int scn_nr;
667 668
	int ret = 0;

669
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
670 671 672
	if (!mem)
		return -ENOMEM;

673
	scn_nr = __section_nr(section);
674 675 676
	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;
677
	mem->state = state;
678
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
679 680
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

681 682 683 684 685 686
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

687
static int add_memory_block(int base_section_nr)
688
{
689 690
	struct memory_block *mem;
	int i, ret, section_count = 0, section_nr;
691

692 693 694 695 696 697 698 699
	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++;
700 701
	}

702 703 704 705 706 707 708
	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;
709 710
}

711 712 713 714 715 716
/*
 * 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)
{
717 718
	int ret = 0;
	struct memory_block *mem;
719 720 721

	mutex_lock(&mem_sysfs_mutex);

722 723 724 725 726 727 728 729
	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;
730
		mem->section_count++;
731 732 733 734 735 736
	}

	if (mem->section_count == sections_per_block)
		ret = register_mem_sect_under_node(mem, nid);
out:
	mutex_unlock(&mem_sysfs_mutex);
737
	return ret;
738 739 740 741 742 743 744 745 746
}

#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() */
747
	put_device(&memory->dev);
748 749 750
	device_unregister(&memory->dev);
}

751
static int remove_memory_section(unsigned long node_id,
752
			       struct mem_section *section, int phys_device)
753 754 755
{
	struct memory_block *mem;

756
	mutex_lock(&mem_sysfs_mutex);
757 758 759 760 761

	/*
	 * Some users of the memory hotplug do not want/need memblock to
	 * track all sections. Skip over those.
	 */
762
	mem = find_memory_block(section);
763 764 765
	if (!mem)
		goto out_unlock;

766
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
767 768

	mem->section_count--;
769
	if (mem->section_count == 0)
770
		unregister_memory(mem);
771
	else
772
		put_device(&mem->dev);
773

774
out_unlock:
775
	mutex_unlock(&mem_sysfs_mutex);
776 777 778 779 780
	return 0;
}

int unregister_memory_section(struct mem_section *section)
{
781
	if (!present_section(section))
782 783
		return -EINVAL;

784
	return remove_memory_section(0, section, 0);
785
}
786
#endif /* CONFIG_MEMORY_HOTREMOVE */
787

788 789 790 791 792 793
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

794 795 796 797 798 799 800 801 802 803 804
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,
805
	&dev_attr_auto_online_blocks.attr,
806 807 808 809 810 811 812 813 814 815 816 817
	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,
};

818 819 820 821 822 823 824
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
825
	int err;
826
	unsigned long block_sz;
827

828
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
829 830
	if (ret)
		goto out;
831

832 833 834
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

835 836 837 838
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
839
	mutex_lock(&mem_sysfs_mutex);
840
	for (i = 0; i < NR_MEM_SECTIONS; i += sections_per_block) {
841 842 843 844
		/* Don't iterate over sections we know are !present: */
		if (i > __highest_present_section_nr)
			break;

845
		err = add_memory_block(i);
846 847
		if (!ret)
			ret = err;
848
	}
849
	mutex_unlock(&mem_sysfs_mutex);
850

851 852
out:
	if (ret)
853
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
854 855
	return ret;
}