memory.c 22.1 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;
}

118 119 120 121 122
/*
 * use this as the physical section index that this memsection
 * uses.
 */

123 124
static ssize_t phys_index_show(struct device *dev,
			       struct device_attribute *attr, char *buf)
125
{
126
	struct memory_block *mem = to_memory_block(dev);
127 128 129 130 131 132
	unsigned long phys_index;

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

133 134 135
/*
 * Show whether the section of memory is likely to be hot-removable
 */
136 137
static ssize_t removable_show(struct device *dev, struct device_attribute *attr,
			      char *buf)
138
{
139
	struct memory_block *mem = to_memory_block(dev);
140 141
	unsigned long pfn;
	int ret = 1, i;
142

143 144 145
	if (mem->state != MEM_ONLINE)
		goto out;

146
	for (i = 0; i < sections_per_block; i++) {
147 148
		if (!present_section_nr(mem->start_section_nr + i))
			continue;
149
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
150 151 152
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

153
out:
154 155 156
	return sprintf(buf, "%d\n", ret);
}

157 158 159
/*
 * online, offline, going offline, etc.
 */
160 161
static ssize_t state_show(struct device *dev, struct device_attribute *attr,
			  char *buf)
162
{
163
	struct memory_block *mem = to_memory_block(dev);
164 165 166 167 168 169 170
	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) {
171 172 173 174 175 176 177 178 179 180 181 182 183 184
	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;
185 186 187 188 189
	}

	return len;
}

190
int memory_notify(unsigned long val, void *v)
191
{
192
	return blocking_notifier_call_chain(&memory_chain, val, v);
193 194
}

195 196 197 198 199
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

200
/*
201 202 203
 * 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.
204
 */
205
static bool pages_correctly_probed(unsigned long start_pfn)
206
{
207 208
	unsigned long section_nr = pfn_to_section_nr(start_pfn);
	unsigned long section_nr_end = section_nr + sections_per_block;
209 210 211 212 213 214 215
	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
	 */
216
	for (; section_nr < section_nr_end; section_nr++) {
217 218 219
		if (WARN_ON_ONCE(!pfn_valid(pfn)))
			return false;

220
		if (!present_section_nr(section_nr)) {
221
			pr_warn("section %ld pfn[%lx, %lx) not present\n",
222 223 224
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (!valid_section_nr(section_nr)) {
225
			pr_warn("section %ld pfn[%lx, %lx) no valid memmap\n",
226 227 228
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (online_section_nr(section_nr)) {
229
			pr_warn("section %ld pfn[%lx, %lx) is already online\n",
230
				section_nr, pfn, pfn + PAGES_PER_SECTION);
231 232
			return false;
		}
233
		pfn += PAGES_PER_SECTION;
234 235 236 237 238
	}

	return true;
}

239 240 241 242 243
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
244 245
memory_block_action(unsigned long start_section_nr, unsigned long action,
		    int online_type)
246
{
247
	unsigned long start_pfn;
248
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
249 250
	int ret;

251
	start_pfn = section_nr_to_pfn(start_section_nr);
252

253
	switch (action) {
254
	case MEM_ONLINE:
255
		if (!pages_correctly_probed(start_pfn))
256 257 258 259 260 261 262 263 264
			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: "
265
		     "%ld\n", __func__, start_section_nr, action, action);
266
		ret = -EINVAL;
267 268 269 270 271
	}

	return ret;
}

272
static int memory_block_change_state(struct memory_block *mem,
273
		unsigned long to_state, unsigned long from_state_req)
274
{
275
	int ret = 0;
276

277 278
	if (mem->state != from_state_req)
		return -EINVAL;
279

280 281 282
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

283 284 285
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

286
	mem->state = ret ? from_state_req : to_state;
287

288 289
	return ret;
}
290

291
/* The device lock serializes operations on memory_subsys_[online|offline] */
292 293
static int memory_subsys_online(struct device *dev)
{
294
	struct memory_block *mem = to_memory_block(dev);
295
	int ret;
296

297 298
	if (mem->state == MEM_ONLINE)
		return 0;
299

300
	/*
301
	 * If we are called from state_store(), online_type will be
302 303 304 305
	 * set >= 0 Otherwise we were called from the device online
	 * attribute and need to set the online_type.
	 */
	if (mem->online_type < 0)
306
		mem->online_type = MMOP_ONLINE_KEEP;
307

308
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
309

310 311
	/* clear online_type */
	mem->online_type = -1;
312 313 314 315 316

	return ret;
}

static int memory_subsys_offline(struct device *dev)
317
{
318
	struct memory_block *mem = to_memory_block(dev);
319

320 321
	if (mem->state == MEM_OFFLINE)
		return 0;
322

323 324 325 326
	/* Can't offline block with non-present sections */
	if (mem->section_count != sections_per_block)
		return -EINVAL;

327
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
328
}
329

330 331
static ssize_t state_store(struct device *dev, struct device_attribute *attr,
			   const char *buf, size_t count)
332
{
333
	struct memory_block *mem = to_memory_block(dev);
334
	int ret, online_type;
335

336 337 338
	ret = lock_device_hotplug_sysfs();
	if (ret)
		return ret;
339

340
	if (sysfs_streq(buf, "online_kernel"))
341
		online_type = MMOP_ONLINE_KERNEL;
342
	else if (sysfs_streq(buf, "online_movable"))
343
		online_type = MMOP_ONLINE_MOVABLE;
344
	else if (sysfs_streq(buf, "online"))
345
		online_type = MMOP_ONLINE_KEEP;
346
	else if (sysfs_streq(buf, "offline"))
347
		online_type = MMOP_OFFLINE;
348 349 350 351
	else {
		ret = -EINVAL;
		goto err;
	}
352 353

	switch (online_type) {
354 355 356
	case MMOP_ONLINE_KERNEL:
	case MMOP_ONLINE_MOVABLE:
	case MMOP_ONLINE_KEEP:
357
		/* mem->online_type is protected by device_hotplug_lock */
358 359 360
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
361
	case MMOP_OFFLINE:
362 363 364 365
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
366 367
	}

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
static ssize_t phys_device_show(struct device *dev,
389
				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
static ssize_t valid_zones_show(struct device *dev,
410 411 412
				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
	/*
	 * 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) {
424 425 426 427 428 429 430 431
		/*
		 * 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;
432 433
		strcat(buf, page_zone(pfn_to_page(start_pfn))->name);
		goto out;
434 435
	}

436
	nid = mem->nid;
437 438
	default_zone = zone_for_pfn_range(MMOP_ONLINE_KEEP, nid, start_pfn, nr_pages);
	strcat(buf, default_zone->name);
439

440 441 442 443
	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);
444
out:
445 446 447
	strcat(buf, "\n");

	return strlen(buf);
448
}
449
static DEVICE_ATTR_RO(valid_zones);
450 451
#endif

452 453 454 455
static DEVICE_ATTR_RO(phys_index);
static DEVICE_ATTR_RW(state);
static DEVICE_ATTR_RO(phys_device);
static DEVICE_ATTR_RO(removable);
456 457 458 459

/*
 * Block size attribute stuff
 */
460 461
static ssize_t block_size_bytes_show(struct device *dev,
				     struct device_attribute *attr, char *buf)
462
{
463
	return sprintf(buf, "%lx\n", get_memory_block_size());
464 465
}

466
static DEVICE_ATTR_RO(block_size_bytes);
467

468 469 470 471
/*
 * Memory auto online policy.
 */

472 473
static ssize_t auto_online_blocks_show(struct device *dev,
				       struct device_attribute *attr, char *buf)
474 475 476 477 478 479 480
{
	if (memhp_auto_online)
		return sprintf(buf, "online\n");
	else
		return sprintf(buf, "offline\n");
}

481 482 483
static ssize_t auto_online_blocks_store(struct device *dev,
					struct device_attribute *attr,
					const char *buf, size_t count)
484 485 486 487 488 489 490 491 492 493 494
{
	if (sysfs_streq(buf, "online"))
		memhp_auto_online = true;
	else if (sysfs_streq(buf, "offline"))
		memhp_auto_online = false;
	else
		return -EINVAL;

	return count;
}

495
static DEVICE_ATTR_RW(auto_online_blocks);
496

497 498 499 500 501 502 503
/*
 * 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
504 505
static ssize_t probe_store(struct device *dev, struct device_attribute *attr,
			   const char *buf, size_t count)
506 507
{
	u64 phys_addr;
508
	int nid, ret;
509
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
510

511 512 513
	ret = kstrtoull(buf, 0, &phys_addr);
	if (ret)
		return ret;
514

515 516 517
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

518 519
	ret = lock_device_hotplug_sysfs();
	if (ret)
520
		return ret;
521

522
	nid = memory_add_physaddr_to_nid(phys_addr);
523 524
	ret = __add_memory(nid, phys_addr,
			   MIN_MEMORY_BLOCK_SIZE * sections_per_block);
525

526 527
	if (ret)
		goto out;
528

529 530
	ret = count;
out:
531
	unlock_device_hotplug();
532
	return ret;
533 534
}

535
static DEVICE_ATTR_WO(probe);
536 537
#endif

538 539 540 541 542 543
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
544 545 546
static ssize_t soft_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
547 548 549 550 551
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
552
	if (kstrtoull(buf, 0, &pfn) < 0)
553 554 555 556 557 558 559 560 561
		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. */
562 563 564
static ssize_t hard_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
565 566 567 568 569
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
570
	if (kstrtoull(buf, 0, &pfn) < 0)
571 572
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
573
	ret = memory_failure(pfn, 0);
574 575 576
	return ret ? ret : count;
}

577 578
static DEVICE_ATTR_WO(soft_offline_page);
static DEVICE_ATTR_WO(hard_offline_page);
579 580
#endif

581 582 583 584 585
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
586 587 588 589
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
590

591 592 593 594
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
595
static struct memory_block *find_memory_block_by_id(unsigned long block_id,
596
						    struct memory_block *hint)
597
{
598 599
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
600

601 602 603 604
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
605
		return NULL;
606
	return to_memory_block(dev);
607 608
}

609 610 611
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
{
612
	unsigned long block_id = base_memory_block_id(__section_nr(section));
613 614 615 616

	return find_memory_block_by_id(block_id, hint);
}

617 618 619 620 621 622
/*
 * 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.
 *
623
 * This could be made generic for all device subsystems.
624 625 626 627 628 629
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

630 631 632 633 634
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
635 636 637
#ifdef CONFIG_MEMORY_HOTREMOVE
	&dev_attr_valid_zones.attr,
#endif
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)
{
656 657
	int ret;

658 659 660 661
	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;
662
	memory->dev.offline = memory->state == MEM_OFFLINE;
663

664 665 666 667 668
	ret = device_register(&memory->dev);
	if (ret)
		put_device(&memory->dev);

	return ret;
669 670
}

671 672
static int init_memory_block(struct memory_block **memory,
			     unsigned long block_id, unsigned long state)
673
{
674
	struct memory_block *mem;
675 676 677
	unsigned long start_pfn;
	int ret = 0;

678 679 680 681 682
	mem = find_memory_block_by_id(block_id, NULL);
	if (mem) {
		put_device(&mem->dev);
		return -EEXIST;
	}
683
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
684 685 686
	if (!mem)
		return -ENOMEM;

687
	mem->start_section_nr = block_id * sections_per_block;
688
	mem->end_section_nr = mem->start_section_nr + sections_per_block - 1;
689
	mem->state = state;
690
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
691 692
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

693 694 695 696 697 698
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

699
static int add_memory_block(unsigned long base_section_nr)
700
{
701
	int ret, section_count = 0;
702
	struct memory_block *mem;
703
	unsigned long nr;
704

705 706 707
	for (nr = base_section_nr; nr < base_section_nr + sections_per_block;
	     nr++)
		if (present_section_nr(nr))
708
			section_count++;
709

710 711
	if (section_count == 0)
		return 0;
712 713
	ret = init_memory_block(&mem, base_memory_block_id(base_section_nr),
				MEM_ONLINE);
714 715 716 717
	if (ret)
		return ret;
	mem->section_count = section_count;
	return 0;
718 719
}

720 721 722 723 724 725 726 727 728 729
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);
}

730
/*
731 732 733
 * 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.
734
 */
735
int create_memory_block_devices(unsigned long start, unsigned long size)
736
{
737 738
	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));
739
	struct memory_block *mem;
740 741
	unsigned long block_id;
	int ret = 0;
742

743 744 745
	if (WARN_ON_ONCE(!IS_ALIGNED(start, memory_block_size_bytes()) ||
			 !IS_ALIGNED(size, memory_block_size_bytes())))
		return -EINVAL;
746

747 748
	mutex_lock(&mem_sysfs_mutex);
	for (block_id = start_block_id; block_id != end_block_id; block_id++) {
749
		ret = init_memory_block(&mem, block_id, MEM_OFFLINE);
750
		if (ret)
751 752 753 754 755 756 757 758 759 760 761
			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++) {
			mem = find_memory_block_by_id(block_id, NULL);
			mem->section_count = 0;
			unregister_memory(mem);
		}
762 763
	}
	mutex_unlock(&mem_sysfs_mutex);
764
	return ret;
765 766
}

767 768 769 770 771 772
/*
 * 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)
773
{
774 775
	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));
776
	struct memory_block *mem;
777
	unsigned long block_id;
778

779 780
	if (WARN_ON_ONCE(!IS_ALIGNED(start, memory_block_size_bytes()) ||
			 !IS_ALIGNED(size, memory_block_size_bytes())))
781 782
		return;

783
	mutex_lock(&mem_sysfs_mutex);
784 785 786 787 788 789
	for (block_id = start_block_id; block_id != end_block_id; block_id++) {
		mem = find_memory_block_by_id(block_id, NULL);
		if (WARN_ON_ONCE(!mem))
			continue;
		mem->section_count = 0;
		unregister_memory_block_under_nodes(mem);
790
		unregister_memory(mem);
791
	}
792
	mutex_unlock(&mem_sysfs_mutex);
793 794
}

795 796 797 798 799 800
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

801 802 803 804 805 806 807 808 809 810 811
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,
812
	&dev_attr_auto_online_blocks.attr,
813 814 815 816 817 818 819 820 821 822 823 824
	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,
};

825 826 827 828 829 830
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	int ret;
831
	int err;
832
	unsigned long block_sz, nr;
833

834
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
835 836
	if (ret)
		goto out;
837

838 839 840
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

841 842 843 844
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
845
	mutex_lock(&mem_sysfs_mutex);
846 847 848
	for (nr = 0; nr <= __highest_present_section_nr;
	     nr += sections_per_block) {
		err = add_memory_block(nr);
849 850
		if (!ret)
			ret = err;
851
	}
852
	mutex_unlock(&mem_sysfs_mutex);
853

854 855
out:
	if (ret)
856
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
857 858
	return ret;
}
859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895

/**
 * 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;

	for (block_id = start_block_id; block_id <= end_block_id; block_id++) {
		mem = find_memory_block_by_id(block_id, NULL);
		if (!mem)
			continue;

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