memory.c 22.5 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
 * Show the first physical section index (number) of this memory block.
105
 */
106 107
static ssize_t phys_index_show(struct device *dev,
			       struct device_attribute *attr, char *buf)
108
{
109
	struct memory_block *mem = to_memory_block(dev);
110 111 112 113 114 115
	unsigned long phys_index;

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

116
/*
117 118 119 120
 * 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.
121
 */
122 123
static ssize_t removable_show(struct device *dev, struct device_attribute *attr,
			      char *buf)
124
{
125
	struct memory_block *mem = to_memory_block(dev);
126 127
	unsigned long pfn;
	int ret = 1, i;
128

129 130 131
	if (mem->state != MEM_ONLINE)
		goto out;

132
	for (i = 0; i < sections_per_block; i++) {
133 134
		if (!present_section_nr(mem->start_section_nr + i))
			continue;
135
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
136 137 138
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

139
out:
140 141 142
	return sprintf(buf, "%d\n", ret);
}

143 144 145
/*
 * online, offline, going offline, etc.
 */
146 147
static ssize_t state_show(struct device *dev, struct device_attribute *attr,
			  char *buf)
148
{
149
	struct memory_block *mem = to_memory_block(dev);
150 151 152 153 154 155 156
	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) {
157 158 159 160 161 162 163 164 165 166 167 168 169 170
	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;
171 172 173 174 175
	}

	return len;
}

176
int memory_notify(unsigned long val, void *v)
177
{
178
	return blocking_notifier_call_chain(&memory_chain, val, v);
179 180
}

181 182 183 184 185
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

186
/*
187 188 189
 * 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.
190
 */
191
static bool pages_correctly_probed(unsigned long start_pfn)
192
{
193 194
	unsigned long section_nr = pfn_to_section_nr(start_pfn);
	unsigned long section_nr_end = section_nr + sections_per_block;
195 196 197 198 199 200 201
	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
	 */
202
	for (; section_nr < section_nr_end; section_nr++) {
203 204 205
		if (WARN_ON_ONCE(!pfn_valid(pfn)))
			return false;

206
		if (!present_section_nr(section_nr)) {
207
			pr_warn("section %ld pfn[%lx, %lx) not present\n",
208 209 210
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (!valid_section_nr(section_nr)) {
211
			pr_warn("section %ld pfn[%lx, %lx) no valid memmap\n",
212 213 214
				section_nr, pfn, pfn + PAGES_PER_SECTION);
			return false;
		} else if (online_section_nr(section_nr)) {
215
			pr_warn("section %ld pfn[%lx, %lx) is already online\n",
216
				section_nr, pfn, pfn + PAGES_PER_SECTION);
217 218
			return false;
		}
219
		pfn += PAGES_PER_SECTION;
220 221 222 223 224
	}

	return true;
}

225 226 227 228 229
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
230 231
memory_block_action(unsigned long start_section_nr, unsigned long action,
		    int online_type)
232
{
233
	unsigned long start_pfn;
234
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
235 236
	int ret;

237
	start_pfn = section_nr_to_pfn(start_section_nr);
238

239
	switch (action) {
240
	case MEM_ONLINE:
241
		if (!pages_correctly_probed(start_pfn))
242 243 244 245 246 247 248 249 250
			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: "
251
		     "%ld\n", __func__, start_section_nr, action, action);
252
		ret = -EINVAL;
253 254 255 256 257
	}

	return ret;
}

258
static int memory_block_change_state(struct memory_block *mem,
259
		unsigned long to_state, unsigned long from_state_req)
260
{
261
	int ret = 0;
262

263 264
	if (mem->state != from_state_req)
		return -EINVAL;
265

266 267 268
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

269 270 271
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

272
	mem->state = ret ? from_state_req : to_state;
273

274 275
	return ret;
}
276

277
/* The device lock serializes operations on memory_subsys_[online|offline] */
278 279
static int memory_subsys_online(struct device *dev)
{
280
	struct memory_block *mem = to_memory_block(dev);
281
	int ret;
282

283 284
	if (mem->state == MEM_ONLINE)
		return 0;
285

286
	/*
287
	 * If we are called from state_store(), online_type will be
288 289 290 291
	 * set >= 0 Otherwise we were called from the device online
	 * attribute and need to set the online_type.
	 */
	if (mem->online_type < 0)
292
		mem->online_type = MMOP_ONLINE_KEEP;
293

294
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
295

296 297
	/* clear online_type */
	mem->online_type = -1;
298 299 300 301 302

	return ret;
}

static int memory_subsys_offline(struct device *dev)
303
{
304
	struct memory_block *mem = to_memory_block(dev);
305

306 307
	if (mem->state == MEM_OFFLINE)
		return 0;
308

309 310 311 312
	/* Can't offline block with non-present sections */
	if (mem->section_count != sections_per_block)
		return -EINVAL;

313
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
314
}
315

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

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

326
	if (sysfs_streq(buf, "online_kernel"))
327
		online_type = MMOP_ONLINE_KERNEL;
328
	else if (sysfs_streq(buf, "online_movable"))
329
		online_type = MMOP_ONLINE_MOVABLE;
330
	else if (sysfs_streq(buf, "online"))
331
		online_type = MMOP_ONLINE_KEEP;
332
	else if (sysfs_streq(buf, "offline"))
333
		online_type = MMOP_OFFLINE;
334 335 336 337
	else {
		ret = -EINVAL;
		goto err;
	}
338 339

	switch (online_type) {
340 341 342
	case MMOP_ONLINE_KERNEL:
	case MMOP_ONLINE_MOVABLE:
	case MMOP_ONLINE_KEEP:
343
		/* mem->online_type is protected by device_hotplug_lock */
344 345 346
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
347
	case MMOP_OFFLINE:
348 349 350 351
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
352 353
	}

354
err:
355
	unlock_device_hotplug();
356

357
	if (ret < 0)
358
		return ret;
359 360 361
	if (ret)
		return -EINVAL;

362 363 364 365 366 367 368 369 370 371 372 373
	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?
 */
374
static ssize_t phys_device_show(struct device *dev,
375
				struct device_attribute *attr, char *buf)
376
{
377
	struct memory_block *mem = to_memory_block(dev);
378 379 380
	return sprintf(buf, "%d\n", mem->phys_device);
}

381
#ifdef CONFIG_MEMORY_HOTREMOVE
382 383 384 385 386 387 388 389 390 391 392 393 394
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);
	}
}

395
static ssize_t valid_zones_show(struct device *dev,
396 397 398
				struct device_attribute *attr, char *buf)
{
	struct memory_block *mem = to_memory_block(dev);
399
	unsigned long start_pfn = section_nr_to_pfn(mem->start_section_nr);
400
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
401
	unsigned long valid_start_pfn, valid_end_pfn;
402
	struct zone *default_zone;
403
	int nid;
404

405 406 407 408 409
	/*
	 * 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) {
410 411 412 413 414 415 416 417
		/*
		 * 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;
418 419
		strcat(buf, page_zone(pfn_to_page(start_pfn))->name);
		goto out;
420 421
	}

422
	nid = mem->nid;
423 424
	default_zone = zone_for_pfn_range(MMOP_ONLINE_KEEP, nid, start_pfn, nr_pages);
	strcat(buf, default_zone->name);
425

426 427 428 429
	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);
430
out:
431 432 433
	strcat(buf, "\n");

	return strlen(buf);
434
}
435
static DEVICE_ATTR_RO(valid_zones);
436 437
#endif

438 439 440 441
static DEVICE_ATTR_RO(phys_index);
static DEVICE_ATTR_RW(state);
static DEVICE_ATTR_RO(phys_device);
static DEVICE_ATTR_RO(removable);
442 443

/*
444
 * Show the memory block size (shared by all memory blocks).
445
 */
446 447
static ssize_t block_size_bytes_show(struct device *dev,
				     struct device_attribute *attr, char *buf)
448
{
449
	return sprintf(buf, "%lx\n", memory_block_size_bytes());
450 451
}

452
static DEVICE_ATTR_RO(block_size_bytes);
453

454 455 456 457
/*
 * Memory auto online policy.
 */

458 459
static ssize_t auto_online_blocks_show(struct device *dev,
				       struct device_attribute *attr, char *buf)
460 461 462 463 464 465 466
{
	if (memhp_auto_online)
		return sprintf(buf, "online\n");
	else
		return sprintf(buf, "offline\n");
}

467 468 469
static ssize_t auto_online_blocks_store(struct device *dev,
					struct device_attribute *attr,
					const char *buf, size_t count)
470 471 472 473 474 475 476 477 478 479 480
{
	if (sysfs_streq(buf, "online"))
		memhp_auto_online = true;
	else if (sysfs_streq(buf, "offline"))
		memhp_auto_online = false;
	else
		return -EINVAL;

	return count;
}

481
static DEVICE_ATTR_RW(auto_online_blocks);
482

483 484 485 486 487 488 489
/*
 * 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
490 491
static ssize_t probe_store(struct device *dev, struct device_attribute *attr,
			   const char *buf, size_t count)
492 493
{
	u64 phys_addr;
494
	int nid, ret;
495
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
496

497 498 499
	ret = kstrtoull(buf, 0, &phys_addr);
	if (ret)
		return ret;
500

501 502 503
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

504 505
	ret = lock_device_hotplug_sysfs();
	if (ret)
506
		return ret;
507

508
	nid = memory_add_physaddr_to_nid(phys_addr);
509 510
	ret = __add_memory(nid, phys_addr,
			   MIN_MEMORY_BLOCK_SIZE * sections_per_block);
511

512 513
	if (ret)
		goto out;
514

515 516
	ret = count;
out:
517
	unlock_device_hotplug();
518
	return ret;
519 520
}

521
static DEVICE_ATTR_WO(probe);
522 523
#endif

524 525 526 527 528 529
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
530 531 532
static ssize_t soft_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       const char *buf, size_t count)
533 534 535 536 537
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
538
	if (kstrtoull(buf, 0, &pfn) < 0)
539 540
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
541
	ret = soft_offline_page(pfn, 0);
542 543 544 545
	return ret == 0 ? count : ret;
}

/* Forcibly offline a page, including killing processes. */
546 547 548
static ssize_t hard_offline_page_store(struct device *dev,
				       struct device_attribute *attr,
				       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
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
557
	ret = memory_failure(pfn, 0);
558 559 560
	return ret ? ret : count;
}

561 562
static DEVICE_ATTR_WO(soft_offline_page);
static DEVICE_ATTR_WO(hard_offline_page);
563 564
#endif

565 566 567 568 569
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
570 571 572 573
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
574

575 576
/* A reference for the returned memory block device is acquired. */
static struct memory_block *find_memory_block_by_id(unsigned long block_id)
577
{
578
	struct device *dev;
579

580 581
	dev = subsys_find_device_by_id(&memory_subsys, block_id, NULL);
	return dev ? to_memory_block(dev) : NULL;
582 583
}

584 585 586 587 588 589
/*
 * 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.
 *
590
 * This could be made generic for all device subsystems.
591 592 593
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
594 595 596
	unsigned long block_id = base_memory_block_id(__section_nr(section));

	return find_memory_block_by_id(block_id);
597 598
}

599 600 601 602 603
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
604 605 606
#ifdef CONFIG_MEMORY_HOTREMOVE
	&dev_attr_valid_zones.attr,
#endif
607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624
	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)
{
625 626
	int ret;

627 628 629 630
	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;
631
	memory->dev.offline = memory->state == MEM_OFFLINE;
632

633 634 635 636 637
	ret = device_register(&memory->dev);
	if (ret)
		put_device(&memory->dev);

	return ret;
638 639
}

640 641
static int init_memory_block(struct memory_block **memory,
			     unsigned long block_id, unsigned long state)
642
{
643
	struct memory_block *mem;
644 645 646
	unsigned long start_pfn;
	int ret = 0;

647
	mem = find_memory_block_by_id(block_id);
648 649 650 651
	if (mem) {
		put_device(&mem->dev);
		return -EEXIST;
	}
652
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
653 654 655
	if (!mem)
		return -ENOMEM;

656
	mem->start_section_nr = block_id * sections_per_block;
657
	mem->state = state;
658
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
659
	mem->phys_device = arch_get_memory_phys_device(start_pfn);
660
	mem->nid = NUMA_NO_NODE;
661

662 663 664 665 666 667
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

668
static int add_memory_block(unsigned long base_section_nr)
669
{
670
	int ret, section_count = 0;
671
	struct memory_block *mem;
672
	unsigned long nr;
673

674 675 676
	for (nr = base_section_nr; nr < base_section_nr + sections_per_block;
	     nr++)
		if (present_section_nr(nr))
677
			section_count++;
678

679 680
	if (section_count == 0)
		return 0;
681 682
	ret = init_memory_block(&mem, base_memory_block_id(base_section_nr),
				MEM_ONLINE);
683 684 685 686
	if (ret)
		return ret;
	mem->section_count = section_count;
	return 0;
687 688
}

689 690 691 692 693 694 695 696 697 698
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);
}

699
/*
700 701 702
 * 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.
703
 */
704
int create_memory_block_devices(unsigned long start, unsigned long size)
705
{
706 707
	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));
708
	struct memory_block *mem;
709 710
	unsigned long block_id;
	int ret = 0;
711

712 713 714
	if (WARN_ON_ONCE(!IS_ALIGNED(start, memory_block_size_bytes()) ||
			 !IS_ALIGNED(size, memory_block_size_bytes())))
		return -EINVAL;
715

716 717
	mutex_lock(&mem_sysfs_mutex);
	for (block_id = start_block_id; block_id != end_block_id; block_id++) {
718
		ret = init_memory_block(&mem, block_id, MEM_OFFLINE);
719
		if (ret)
720 721 722 723 724 725 726
			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++) {
727
			mem = find_memory_block_by_id(block_id);
728 729 730
			mem->section_count = 0;
			unregister_memory(mem);
		}
731 732
	}
	mutex_unlock(&mem_sysfs_mutex);
733
	return ret;
734 735
}

736 737 738 739 740 741
/*
 * 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)
742
{
743 744
	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));
745
	struct memory_block *mem;
746
	unsigned long block_id;
747

748 749
	if (WARN_ON_ONCE(!IS_ALIGNED(start, memory_block_size_bytes()) ||
			 !IS_ALIGNED(size, memory_block_size_bytes())))
750 751
		return;

752
	mutex_lock(&mem_sysfs_mutex);
753
	for (block_id = start_block_id; block_id != end_block_id; block_id++) {
754
		mem = find_memory_block_by_id(block_id);
755 756 757 758
		if (WARN_ON_ONCE(!mem))
			continue;
		mem->section_count = 0;
		unregister_memory_block_under_nodes(mem);
759
		unregister_memory(mem);
760
	}
761
	mutex_unlock(&mem_sysfs_mutex);
762 763
}

764 765 766 767 768 769
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

770 771 772 773 774 775 776 777 778 779 780
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,
781
	&dev_attr_auto_online_blocks.attr,
782 783 784 785 786 787 788 789 790 791 792 793
	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,
};

794 795 796
/*
 * Initialize the sysfs support for memory devices...
 */
797
void __init memory_dev_init(void)
798 799
{
	int ret;
800
	int err;
801
	unsigned long block_sz, nr;
802

803 804 805 806 807 808
	/* Validate the configured memory block size */
	block_sz = memory_block_size_bytes();
	if (!is_power_of_2(block_sz) || block_sz < MIN_MEMORY_BLOCK_SIZE)
		panic("Memory block size not suitable: 0x%lx\n", block_sz);
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

809
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
810 811
	if (ret)
		goto out;
812 813 814 815 816

	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
817
	mutex_lock(&mem_sysfs_mutex);
818 819 820
	for (nr = 0; nr <= __highest_present_section_nr;
	     nr += sections_per_block) {
		err = add_memory_block(nr);
821 822
		if (!ret)
			ret = err;
823
	}
824
	mutex_unlock(&mem_sysfs_mutex);
825

826 827
out:
	if (ret)
828
		panic("%s() failed: %d\n", __func__, ret);
829
}
830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854

/**
 * 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;

855 856 857
	if (!size)
		return 0;

858
	for (block_id = start_block_id; block_id <= end_block_id; block_id++) {
859
		mem = find_memory_block_by_id(block_id);
860 861 862 863 864 865 866 867 868 869
		if (!mem)
			continue;

		ret = func(mem, arg);
		put_device(&mem->dev);
		if (ret)
			break;
	}
	return ret;
}
870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905

struct for_each_memory_block_cb_data {
	walk_memory_blocks_func_t func;
	void *arg;
};

static int for_each_memory_block_cb(struct device *dev, void