memory.c 21.9 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
static int sections_per_block;

37
static inline unsigned long base_memory_block_id(unsigned long section_nr)
38 39 40
{
	return section_nr / sections_per_block;
}
41

42
static inline unsigned long pfn_to_block_id(unsigned long pfn)
43 44 45 46
{
	return base_memory_block_id(pfn_to_section_nr(pfn));
}

47 48 49 50 51
static inline unsigned long phys_to_block_id(unsigned long phys)
{
	return pfn_to_block_id(PFN_DOWN(phys));
}

52 53 54
static int memory_subsys_online(struct device *dev);
static int memory_subsys_offline(struct device *dev);

55
static struct bus_type memory_subsys = {
56
	.name = MEMORY_CLASS_NAME,
57
	.dev_name = MEMORY_CLASS_NAME,
58 59
	.online = memory_subsys_online,
	.offline = memory_subsys_offline,
60 61
};

62
static BLOCKING_NOTIFIER_HEAD(memory_chain);
63

64
int register_memory_notifier(struct notifier_block *nb)
65
{
66
	return blocking_notifier_chain_register(&memory_chain, nb);
67
}
68
EXPORT_SYMBOL(register_memory_notifier);
69

70
void unregister_memory_notifier(struct notifier_block *nb)
71
{
72
	blocking_notifier_chain_unregister(&memory_chain, nb);
73
}
74
EXPORT_SYMBOL(unregister_memory_notifier);
75

76 77 78 79 80 81 82 83 84 85 86 87 88 89
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);

90 91
static void memory_block_release(struct device *dev)
{
92
	struct memory_block *mem = to_memory_block(dev);
93 94 95 96

	kfree(mem);
}

97 98 99 100
unsigned long __weak memory_block_size_bytes(void)
{
	return MIN_MEMORY_BLOCK_SIZE;
}
101
EXPORT_SYMBOL_GPL(memory_block_size_bytes);
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117

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;
}

118
/*
119
 * Show the first physical section index (number) of this memory block.
120
 */
121 122
static ssize_t phys_index_show(struct device *dev,
			       struct device_attribute *attr, char *buf)
123
{
124
	struct memory_block *mem = to_memory_block(dev);
125 126 127 128 129 130
	unsigned long phys_index;

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

131
/*
132 133 134 135
 * Show whether the memory block is likely to be offlineable (or is already
 * offline). Once offline, the memory block could be removed. The return
 * value does, however, not indicate that there is a way to remove the
 * memory block.
136
 */
137 138
static ssize_t removable_show(struct device *dev, struct device_attribute *attr,
			      char *buf)
139
{
140
	struct memory_block *mem = to_memory_block(dev);
141 142
	unsigned long pfn;
	int ret = 1, i;
143

144 145 146
	if (mem->state != MEM_ONLINE)
		goto out;

147
	for (i = 0; i < sections_per_block; i++) {
148 149
		if (!present_section_nr(mem->start_section_nr + i))
			continue;
150
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
151 152 153
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

154
out:
155 156 157
	return sprintf(buf, "%d\n", ret);
}

158 159 160
/*
 * online, offline, going offline, etc.
 */
161 162
static ssize_t state_show(struct device *dev, struct device_attribute *attr,
			  char *buf)
163
{
164
	struct memory_block *mem = to_memory_block(dev);
165 166 167 168 169 170 171
	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) {
172 173 174 175 176 177 178 179 180 181 182 183 184 185
	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;
186 187 188 189 190
	}

	return len;
}

191
int memory_notify(unsigned long val, void *v)
192
{
193
	return blocking_notifier_call_chain(&memory_chain, val, v);
194 195
}

196 197 198 199 200
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

201
/*
202 203 204
 * 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.
205
 */
206
static bool pages_correctly_probed(unsigned long start_pfn)
207
{
208 209
	unsigned long section_nr = pfn_to_section_nr(start_pfn);
	unsigned long section_nr_end = section_nr + sections_per_block;
210 211 212 213 214 215 216
	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
	 */
217
	for (; section_nr < section_nr_end; section_nr++) {
218 219 220
		if (WARN_ON_ONCE(!pfn_valid(pfn)))
			return false;

221
		if (!present_section_nr(section_nr)) {
222
			pr_warn("section %ld pfn[%lx, %lx) not present\n",
223 224 225
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (!valid_section_nr(section_nr)) {
226
			pr_warn("section %ld pfn[%lx, %lx) no valid memmap\n",
227 228 229
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (online_section_nr(section_nr)) {
230
			pr_warn("section %ld pfn[%lx, %lx) is already online\n",
231
				section_nr, pfn, pfn + PAGES_PER_SECTION);
232 233
			return false;
		}
234
		pfn += PAGES_PER_SECTION;
235 236 237 238 239
	}

	return true;
}

240 241 242 243 244
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
245 246
memory_block_action(unsigned long start_section_nr, unsigned long action,
		    int online_type)
247
{
248
	unsigned long start_pfn;
249
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
250 251
	int ret;

252
	start_pfn = section_nr_to_pfn(start_section_nr);
253

254
	switch (action) {
255
	case MEM_ONLINE:
256
		if (!pages_correctly_probed(start_pfn))
257 258 259 260 261 262 263 264 265
			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: "
266
		     "%ld\n", __func__, start_section_nr, action, action);
267
		ret = -EINVAL;
268 269 270 271 272
	}

	return ret;
}

273
static int memory_block_change_state(struct memory_block *mem,
274
		unsigned long to_state, unsigned long from_state_req)
275
{
276
	int ret = 0;
277

278 279
	if (mem->state != from_state_req)
		return -EINVAL;
280

281 282 283
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

284 285 286
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

287
	mem->state = ret ? from_state_req : to_state;
288

289 290
	return ret;
}
291

292
/* The device lock serializes operations on memory_subsys_[online|offline] */
293 294
static int memory_subsys_online(struct device *dev)
{
295
	struct memory_block *mem = to_memory_block(dev);
296
	int ret;
297

298 299
	if (mem->state == MEM_ONLINE)
		return 0;
300

301
	/*
302
	 * If we are called from state_store(), online_type will be
303 304 305 306
	 * set >= 0 Otherwise we were called from the device online
	 * attribute and need to set the online_type.
	 */
	if (mem->online_type < 0)
307
		mem->online_type = MMOP_ONLINE_KEEP;
308

309
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
310

311 312
	/* clear online_type */
	mem->online_type = -1;
313 314 315 316 317

	return ret;
}

static int memory_subsys_offline(struct device *dev)
318
{
319
	struct memory_block *mem = to_memory_block(dev);
320

321 322
	if (mem->state == MEM_OFFLINE)
		return 0;
323

324 325 326 327
	/* Can't offline block with non-present sections */
	if (mem->section_count != sections_per_block)
		return -EINVAL;

328
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
329
}
330

331 332
static ssize_t state_store(struct device *dev, struct device_attribute *attr,
			   const char *buf, size_t count)
333
{
334
	struct memory_block *mem = to_memory_block(dev);
335
	int ret, online_type;
336

337 338 339
	ret = lock_device_hotplug_sysfs();
	if (ret)
		return ret;
340

341
	if (sysfs_streq(buf, "online_kernel"))
342
		online_type = MMOP_ONLINE_KERNEL;
343
	else if (sysfs_streq(buf, "online_movable"))
344
		online_type = MMOP_ONLINE_MOVABLE;
345
	else if (sysfs_streq(buf, "online"))
346
		online_type = MMOP_ONLINE_KEEP;
347
	else if (sysfs_streq(buf, "offline"))
348
		online_type = MMOP_OFFLINE;
349 350 351 352
	else {
		ret = -EINVAL;
		goto err;
	}
353 354

	switch (online_type) {
355 356 357
	case MMOP_ONLINE_KERNEL:
	case MMOP_ONLINE_MOVABLE:
	case MMOP_ONLINE_KEEP:
358
		/* mem->online_type is protected by device_hotplug_lock */
359 360 361
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
362
	case MMOP_OFFLINE:
363 364 365 366
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
367 368
	}

369
err:
370
	unlock_device_hotplug();
371

372
	if (ret < 0)
373
		return ret;
374 375 376
	if (ret)
		return -EINVAL;

377 378 379 380 381 382 383 384 385 386 387 388
	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?
 */
389
static ssize_t phys_device_show(struct device *dev,
390
				struct device_attribute *attr, char *buf)
391
{
392
	struct memory_block *mem = to_memory_block(dev);
393 394 395
	return sprintf(buf, "%d\n", mem->phys_device);
}

396
#ifdef CONFIG_MEMORY_HOTREMOVE
397 398 399 400 401 402 403 404 405 406 407 408 409
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);
	}
}

410
static ssize_t valid_zones_show(struct device *dev,
411 412 413
				struct device_attribute *attr, char *buf)
{
	struct memory_block *mem = to_memory_block(dev);
414
	unsigned long start_pfn = section_nr_to_pfn(mem->start_section_nr);
415
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
416
	unsigned long valid_start_pfn, valid_end_pfn;
417
	struct zone *default_zone;
418
	int nid;
419

420 421 422 423 424
	/*
	 * 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) {
425 426 427 428 429 430 431 432
		/*
		 * 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;
433 434
		strcat(buf, page_zone(pfn_to_page(start_pfn))->name);
		goto out;
435 436
	}

437
	nid = mem->nid;
438 439
	default_zone = zone_for_pfn_range(MMOP_ONLINE_KEEP, nid, start_pfn, nr_pages);
	strcat(buf, default_zone->name);
440

441 442 443 444
	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);
445
out:
446 447 448
	strcat(buf, "\n");

	return strlen(buf);
449
}
450
static DEVICE_ATTR_RO(valid_zones);
451 452
#endif

453 454 455 456
static DEVICE_ATTR_RO(phys_index);
static DEVICE_ATTR_RW(state);
static DEVICE_ATTR_RO(phys_device);
static DEVICE_ATTR_RO(removable);
457 458

/*
459
 * Show the memory block size (shared by all memory blocks).
460
 */
461 462
static ssize_t block_size_bytes_show(struct device *dev,
				     struct device_attribute *attr, char *buf)
463
{
464
	return sprintf(buf, "%lx\n", get_memory_block_size());
465 466
}

467
static DEVICE_ATTR_RO(block_size_bytes);
468

469 470 471 472
/*
 * Memory auto online policy.
 */

473 474
static ssize_t auto_online_blocks_show(struct device *dev,
				       struct device_attribute *attr, char *buf)
475 476 477 478 479 480 481
{
	if (memhp_auto_online)
		return sprintf(buf, "online\n");
	else
		return sprintf(buf, "offline\n");
}

482 483 484
static ssize_t auto_online_blocks_store(struct device *dev,
					struct device_attribute *attr,
					const char *buf, size_t count)
485 486 487 488 489 490 491 492 493 494 495
{
	if (sysfs_streq(buf, "online"))
		memhp_auto_online = true;
	else if (sysfs_streq(buf, "offline"))
		memhp_auto_online = false;
	else
		return -EINVAL;

	return count;
}

496
static DEVICE_ATTR_RW(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
505 506
static ssize_t probe_store(struct device *dev, struct device_attribute *attr,
			   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
	ret = lock_device_hotplug_sysfs();
	if (ret)
521
		return ret;
522

523
	nid = memory_add_physaddr_to_nid(phys_addr);
524 525
	ret = __add_memory(nid, phys_addr,
			   MIN_MEMORY_BLOCK_SIZE * sections_per_block);
526

527 528
	if (ret)
		goto out;
529

530 531
	ret = count;
out:
532
	unlock_device_hotplug();
533
	return ret;
534 535
}

536
static DEVICE_ATTR_WO(probe);
537 538
#endif

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

/* Soft offline a page */
545 546 547
static ssize_t soft_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
548 549 550 551 552
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
553
	if (kstrtoull(buf, 0, &pfn) < 0)
554 555 556 557 558 559 560 561 562
		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. */
563 564 565
static ssize_t hard_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
566 567 568 569 570
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
571
	if (kstrtoull(buf, 0, &pfn) < 0)
572 573
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
574
	ret = memory_failure(pfn, 0);
575 576 577
	return ret ? ret : count;
}

578 579
static DEVICE_ATTR_WO(soft_offline_page);
static DEVICE_ATTR_WO(hard_offline_page);
580 581
#endif

582 583 584 585 586
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
587 588 589 590
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
591

592 593
/* A reference for the returned memory block device is acquired. */
static struct memory_block *find_memory_block_by_id(unsigned long block_id)
594
{
595
	struct device *dev;
596

597 598
	dev = subsys_find_device_by_id(&memory_subsys, block_id, NULL);
	return dev ? to_memory_block(dev) : NULL;
599 600
}

601 602 603 604 605 606
/*
 * 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.
 *
607
 * This could be made generic for all device subsystems.
608 609 610
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
611 612 613
	unsigned long block_id = base_memory_block_id(__section_nr(section));

	return find_memory_block_by_id(block_id);
614 615
}

616 617 618 619 620
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
621 622 623
#ifdef CONFIG_MEMORY_HOTREMOVE
	&dev_attr_valid_zones.attr,
#endif
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
	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)
{
642 643
	int ret;

644 645 646 647
	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;
648
	memory->dev.offline = memory->state == MEM_OFFLINE;
649

650 651 652 653 654
	ret = device_register(&memory->dev);
	if (ret)
		put_device(&memory->dev);

	return ret;
655 656
}

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

664
	mem = find_memory_block_by_id(block_id);
665 666 667 668
	if (mem) {
		put_device(&mem->dev);
		return -EEXIST;
	}
669
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
670 671 672
	if (!mem)
		return -ENOMEM;

673
	mem->start_section_nr = block_id * sections_per_block;
674
	mem->end_section_nr = mem->start_section_nr + sections_per_block - 1;
675
	mem->state = state;
676
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
677
	mem->phys_device = arch_get_memory_phys_device(start_pfn);
678
	mem->nid = NUMA_NO_NODE;
679

680 681 682 683 684 685
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

686
static int add_memory_block(unsigned long base_section_nr)
687
{
688
	int ret, section_count = 0;
689
	struct memory_block *mem;
690
	unsigned long nr;
691

692 693 694
	for (nr = base_section_nr; nr < base_section_nr + sections_per_block;
	     nr++)
		if (present_section_nr(nr))
695
			section_count++;
696

697 698
	if (section_count == 0)
		return 0;
699 700
	ret = init_memory_block(&mem, base_memory_block_id(base_section_nr),
				MEM_ONLINE);
701 702 703 704
	if (ret)
		return ret;
	mem->section_count = section_count;
	return 0;
705 706
}

707 708 709 710 711 712 713 714 715 716
static void unregister_memory(struct memory_block *memory)
{
	if (WARN_ON_ONCE(memory->dev.bus != &memory_subsys))
		return;

	/* drop the ref. we got via find_memory_block() */
	put_device(&memory->dev);
	device_unregister(&memory->dev);
}

717
/*
718 719 720
 * Create memory block devices for the given memory area. Start and size
 * have to be aligned to memory block granularity. Memory block devices
 * will be initialized as offline.
721
 */
722
int create_memory_block_devices(unsigned long start, unsigned long size)
723
{
724 725
	const unsigned long start_block_id = pfn_to_block_id(PFN_DOWN(start));
	unsigned long end_block_id = pfn_to_block_id(PFN_DOWN(start + size));
726
	struct memory_block *mem;
727 728
	unsigned long block_id;
	int ret = 0;
729

730 731 732
	if (WARN_ON_ONCE(!IS_ALIGNED(start, memory_block_size_bytes()) ||
			 !IS_ALIGNED(size, memory_block_size_bytes())))
		return -EINVAL;
733

734 735
	mutex_lock(&mem_sysfs_mutex);
	for (block_id = start_block_id; block_id != end_block_id; block_id++) {
736
		ret = init_memory_block(&mem, block_id, MEM_OFFLINE);
737
		if (ret)
738 739 740 741 742 743 744
			break;
		mem->section_count = sections_per_block;
	}
	if (ret) {
		end_block_id = block_id;
		for (block_id = start_block_id; block_id != end_block_id;
		     block_id++) {
745
			mem = find_memory_block_by_id(block_id);
746 747 748
			mem->section_count = 0;
			unregister_memory(mem);
		}
749 750
	}
	mutex_unlock(&mem_sysfs_mutex);
751
	return ret;
752 753
}

754 755 756 757 758 759
/*
 * Remove memory block devices for the given memory area. Start and size
 * have to be aligned to memory block granularity. Memory block devices
 * have to be offline.
 */
void remove_memory_block_devices(unsigned long start, unsigned long size)
760
{
761 762
	const unsigned long start_block_id = pfn_to_block_id(PFN_DOWN(start));
	const unsigned long end_block_id = pfn_to_block_id(PFN_DOWN(start + size));
763
	struct memory_block *mem;
764
	unsigned long block_id;
765

766 767
	if (WARN_ON_ONCE(!IS_ALIGNED(start, memory_block_size_bytes()) ||
			 !IS_ALIGNED(size, memory_block_size_bytes())))
768 769
		return;

770
	mutex_lock(&mem_sysfs_mutex);
771
	for (block_id = start_block_id; block_id != end_block_id; block_id++) {
772
		mem = find_memory_block_by_id(block_id);
773 774 775 776
		if (WARN_ON_ONCE(!mem))
			continue;
		mem->section_count = 0;
		unregister_memory_block_under_nodes(mem);
777
		unregister_memory(mem);
778
	}
779
	mutex_unlock(&mem_sysfs_mutex);
780 781
}

782 783 784 785 786 787
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

788 789 790 791 792 793 794 795 796 797 798
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,
799
	&dev_attr_auto_online_blocks.attr,
800 801 802 803 804 805 806 807 808 809 810 811
	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,
};

812 813 814 815 816 817
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	int ret;
818
	int err;
819
	unsigned long block_sz, nr;
820

821
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
822 823
	if (ret)
		goto out;
824

825 826 827
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

828 829 830 831
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
832
	mutex_lock(&mem_sysfs_mutex);
833 834 835
	for (nr = 0; nr <= __highest_present_section_nr;
	     nr += sections_per_block) {
		err = add_memory_block(nr);
836 837
		if (!ret)
			ret = err;
838
	}
839
	mutex_unlock(&mem_sysfs_mutex);
840

841 842
out:
	if (ret)
843
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
844 845
	return ret;
}
846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870

/**
 * walk_memory_blocks - walk through all present memory blocks overlapped
 *			by the range [start, start + size)
 *
 * @start: start address of the memory range
 * @size: size of the memory range
 * @arg: argument passed to func
 * @func: callback for each memory section walked
 *
 * This function walks through all present memory blocks overlapped by the
 * range [start, start + size), calling func on each memory block.
 *
 * In case func() returns an error, walking is aborted and the error is
 * returned.
 */
int walk_memory_blocks(unsigned long start, unsigned long size,
		       void *arg, walk_memory_blocks_func_t func)
{
	const unsigned long start_block_id = phys_to_block_id(start);
	const unsigned long end_block_id = phys_to_block_id(start + size - 1);
	struct memory_block *mem;
	unsigned long block_id;
	int ret = 0;

871 872 873
	if (!size)
		return 0;

874
	for (block_id = start_block_id; block_id <= end_block_id; block_id++) {
875
		mem = find_memory_block_by_id(block_id);
876 877 878 879 880 881 882 883 884 885
		if (!mem)
			continue;

		ret = func(mem, arg);
		put_device(&mem->dev);
		if (ret)
			break;
	}
	return ret;
}