memory.c 17.8 KB
Newer Older
1
/*
2
 * Memory subsystem support
3 4 5 6 7 8 9 10 11 12 13 14 15
 *
 * 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>
16
#include <linux/capability.h>
17 18 19 20
#include <linux/device.h>
#include <linux/memory.h>
#include <linux/memory_hotplug.h>
#include <linux/mm.h>
21
#include <linux/mutex.h>
22
#include <linux/stat.h>
23
#include <linux/slab.h>
24

Arun Sharma's avatar
Arun Sharma committed
25
#include <linux/atomic.h>
26 27
#include <asm/uaccess.h>

28 29
static DEFINE_MUTEX(mem_sysfs_mutex);

30
#define MEMORY_CLASS_NAME	"memory"
31 32 33 34 35 36 37

static int sections_per_block;

static inline int base_memory_block_id(int section_nr)
{
	return section_nr / sections_per_block;
}
38

39 40 41
static int memory_subsys_online(struct device *dev);
static int memory_subsys_offline(struct device *dev);

42
static struct bus_type memory_subsys = {
43
	.name = MEMORY_CLASS_NAME,
44
	.dev_name = MEMORY_CLASS_NAME,
45 46
	.online = memory_subsys_online,
	.offline = memory_subsys_offline,
47 48
};

49
static BLOCKING_NOTIFIER_HEAD(memory_chain);
50

51
int register_memory_notifier(struct notifier_block *nb)
52
{
53
        return blocking_notifier_chain_register(&memory_chain, nb);
54
}
55
EXPORT_SYMBOL(register_memory_notifier);
56

57
void unregister_memory_notifier(struct notifier_block *nb)
58
{
59
        blocking_notifier_chain_unregister(&memory_chain, nb);
60
}
61
EXPORT_SYMBOL(unregister_memory_notifier);
62

63 64 65 66 67 68 69 70 71 72 73 74 75 76
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);

77 78 79 80 81 82 83
static void memory_block_release(struct device *dev)
{
	struct memory_block *mem = container_of(dev, struct memory_block, dev);

	kfree(mem);
}

84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
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;
}

104 105 106 107 108
/*
 * use this as the physical section index that this memsection
 * uses.
 */

109 110
static ssize_t show_mem_start_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
111 112
{
	struct memory_block *mem =
113
		container_of(dev, struct memory_block, dev);
114 115 116 117 118 119
	unsigned long phys_index;

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

120 121
static ssize_t show_mem_end_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
122 123
{
	struct memory_block *mem =
124
		container_of(dev, struct memory_block, dev);
125 126 127 128
	unsigned long phys_index;

	phys_index = mem->end_section_nr / sections_per_block;
	return sprintf(buf, "%08lx\n", phys_index);
129 130
}

131 132 133
/*
 * Show whether the section of memory is likely to be hot-removable
 */
134 135
static ssize_t show_mem_removable(struct device *dev,
			struct device_attribute *attr, char *buf)
136
{
137 138
	unsigned long i, pfn;
	int ret = 1;
139
	struct memory_block *mem =
140
		container_of(dev, struct memory_block, dev);
141

142
	for (i = 0; i < sections_per_block; i++) {
143
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
144 145 146
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

147 148 149
	return sprintf(buf, "%d\n", ret);
}

150 151 152
/*
 * online, offline, going offline, etc.
 */
153 154
static ssize_t show_mem_state(struct device *dev,
			struct device_attribute *attr, char *buf)
155 156
{
	struct memory_block *mem =
157
		container_of(dev, struct memory_block, dev);
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
	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) {
		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;
	}

	return len;
}

184
int memory_notify(unsigned long val, void *v)
185
{
186
	return blocking_notifier_call_chain(&memory_chain, val, v);
187 188
}

189 190 191 192 193
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

194 195 196 197
/*
 * The probe routines leave the pages reserved, just as the bootmem code does.
 * Make sure they're still that way.
 */
198
static bool pages_correctly_reserved(unsigned long start_pfn)
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228
{
	int i, j;
	struct page *page;
	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
	 */
	for (i = 0; i < sections_per_block; i++, pfn += PAGES_PER_SECTION) {
		if (WARN_ON_ONCE(!pfn_valid(pfn)))
			return false;
		page = pfn_to_page(pfn);

		for (j = 0; j < PAGES_PER_SECTION; j++) {
			if (PageReserved(page + j))
				continue;

			printk(KERN_WARNING "section number %ld page number %d "
				"not reserved, was it already online?\n",
				pfn_to_section_nr(pfn), j);

			return false;
		}
	}

	return true;
}

229 230 231 232 233
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
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
	struct page *first_page;
239 240
	int ret;

241
	first_page = pfn_to_page(phys_index << PFN_SECTION_SHIFT);
242
	start_pfn = page_to_pfn(first_page);
243

244 245
	switch (action) {
		case MEM_ONLINE:
246
			if (!pages_correctly_reserved(start_pfn))
247 248
				return -EBUSY;

249
			ret = online_pages(start_pfn, nr_pages, online_type);
250 251
			break;
		case MEM_OFFLINE:
252
			ret = offline_pages(start_pfn, nr_pages);
253 254
			break;
		default:
255 256
			WARN(1, KERN_WARNING "%s(%ld, %ld) unknown action: "
			     "%ld\n", __func__, phys_index, action, action);
257 258 259 260 261 262
			ret = -EINVAL;
	}

	return ret;
}

263 264
static int memory_block_change_state(struct memory_block *mem,
		unsigned long to_state, unsigned long from_state_req)
265
{
266
	int ret = 0;
267

268 269
	if (mem->state != from_state_req)
		return -EINVAL;
270

271 272 273
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

274 275 276
	ret = memory_block_action(mem->start_section_nr, to_state,
				mem->online_type);

277
	mem->state = ret ? from_state_req : to_state;
278

279 280
	return ret;
}
281

282
/* The device lock serializes operations on memory_subsys_[online|offline] */
283 284 285 286
static int memory_subsys_online(struct device *dev)
{
	struct memory_block *mem = container_of(dev, struct memory_block, dev);
	int ret;
287

288 289 290 291 292 293 294 295 296 297
	if (mem->state == MEM_ONLINE)
		return 0;

	/*
	 * 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)
		mem->online_type = ONLINE_KEEP;
298

299 300 301 302
	ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);

	/* clear online_type */
	mem->online_type = -1;
303 304 305 306 307 308 309 310

	return ret;
}

static int memory_subsys_offline(struct device *dev)
{
	struct memory_block *mem = container_of(dev, struct memory_block, dev);

311 312
	if (mem->state == MEM_OFFLINE)
		return 0;
313

314
	return memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
315 316
}

317
static ssize_t
318 319
store_mem_state(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
320 321
{
	struct memory_block *mem;
322
	int ret, online_type;
323

324
	mem = container_of(dev, struct memory_block, dev);
325

326 327
	lock_device_hotplug();

328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
	if (!strncmp(buf, "online_kernel", min_t(int, count, 13)))
		online_type = ONLINE_KERNEL;
	else if (!strncmp(buf, "online_movable", min_t(int, count, 14)))
		online_type = ONLINE_MOVABLE;
	else if (!strncmp(buf, "online", min_t(int, count, 6)))
		online_type = ONLINE_KEEP;
	else if (!strncmp(buf, "offline", min_t(int, count, 7)))
		online_type = -1;
	else
		return -EINVAL;

	switch (online_type) {
	case ONLINE_KERNEL:
	case ONLINE_MOVABLE:
	case ONLINE_KEEP:
		/*
		 * mem->online_type is not protected so there can be a
		 * race here.  However, when racing online, the first
		 * will succeed and the second will just return as the
		 * block will already be online.  The online type
		 * could be either one, but that is expected.
		 */
		mem->online_type = online_type;
		ret = device_online(&mem->dev);
		break;
	case -1:
		ret = device_offline(&mem->dev);
		break;
	default:
		ret = -EINVAL; /* should never happen */
358 359 360
	}

	unlock_device_hotplug();
361

362 363 364 365 366 367 368 369 370 371 372 373 374 375
	if (ret)
		return ret;
	return count;
}

/*
 * phys_device is a bad name for this.  What I really want
 * is a way to differentiate between memory ranges that
 * are part of physical devices that constitute
 * a complete removable unit or fru.
 * i.e. do these ranges belong to the same physical device,
 * s.t. if I offline all of these sections I can then
 * remove the physical device?
 */
376 377
static ssize_t show_phys_device(struct device *dev,
				struct device_attribute *attr, char *buf)
378 379
{
	struct memory_block *mem =
380
		container_of(dev, struct memory_block, dev);
381 382 383
	return sprintf(buf, "%d\n", mem->phys_device);
}

384 385 386 387 388
static DEVICE_ATTR(phys_index, 0444, show_mem_start_phys_index, NULL);
static DEVICE_ATTR(end_phys_index, 0444, show_mem_end_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);
389 390 391 392 393

/*
 * Block size attribute stuff
 */
static ssize_t
394
print_block_size(struct device *dev, struct device_attribute *attr,
395
		 char *buf)
396
{
397
	return sprintf(buf, "%lx\n", get_memory_block_size());
398 399
}

400
static DEVICE_ATTR(block_size_bytes, 0444, print_block_size, NULL);
401 402 403 404 405 406 407 408 409

/*
 * 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
410
memory_probe_store(struct device *dev, struct device_attribute *attr,
411
		   const char *buf, size_t count)
412 413
{
	u64 phys_addr;
414
	int nid;
415
	int i, ret;
416
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
417 418 419

	phys_addr = simple_strtoull(buf, NULL, 0);

420 421 422
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

423 424 425 426 427
	for (i = 0; i < sections_per_block; i++) {
		nid = memory_add_physaddr_to_nid(phys_addr);
		ret = add_memory(nid, phys_addr,
				 PAGES_PER_SECTION << PAGE_SHIFT);
		if (ret)
428
			goto out;
429 430 431

		phys_addr += MIN_MEMORY_BLOCK_SIZE;
	}
432

433 434 435
	ret = count;
out:
	return ret;
436 437
}

438
static DEVICE_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
439 440
#endif

441 442 443 444 445 446 447
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
448 449
store_soft_offline_page(struct device *dev,
			struct device_attribute *attr,
450
			const char *buf, size_t count)
451 452 453 454 455
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
456
	if (kstrtoull(buf, 0, &pfn) < 0)
457 458 459 460 461 462 463 464 465 466
		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
467 468
store_hard_offline_page(struct device *dev,
			struct device_attribute *attr,
469
			const char *buf, size_t count)
470 471 472 473 474
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
475
	if (kstrtoull(buf, 0, &pfn) < 0)
476 477
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
478
	ret = memory_failure(pfn, 0, 0);
479 480 481
	return ret ? ret : count;
}

482 483
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);
484 485
#endif

486 487 488 489 490
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
491 492 493 494
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
495

496 497 498 499
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
500 501
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
502
{
503
	int block_id = base_memory_block_id(__section_nr(section));
504 505
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
506

507 508 509 510
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
511
		return NULL;
512
	return container_of(dev, struct memory_block, dev);
513 514
}

515 516 517 518 519 520
/*
 * 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.
 *
521
 * This could be made generic for all device subsystems.
522 523 524 525 526 527
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555
static struct attribute *memory_memblk_attrs[] = {
	&dev_attr_phys_index.attr,
	&dev_attr_end_phys_index.attr,
	&dev_attr_state.attr,
	&dev_attr_phys_device.attr,
	&dev_attr_removable.attr,
	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;
556
	memory->dev.offline = memory->state == MEM_OFFLINE;
557

558
	return device_register(&memory->dev);
559 560
}

561 562
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
563
{
564
	struct memory_block *mem;
565
	unsigned long start_pfn;
566
	int scn_nr;
567 568
	int ret = 0;

569
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
570 571 572
	if (!mem)
		return -ENOMEM;

573
	scn_nr = __section_nr(section);
574 575 576
	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;
577
	mem->state = state;
578
	mem->section_count++;
579
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
580 581
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

582 583 584 585 586 587
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

588
static int add_memory_block(int base_section_nr)
589
{
590 591
	struct memory_block *mem;
	int i, ret, section_count = 0, section_nr;
592

593 594 595 596 597 598 599 600
	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++;
601 602
	}

603 604 605 606 607 608 609
	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;
610 611
}

612

613 614 615 616 617 618
/*
 * 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)
{
619 620
	int ret = 0;
	struct memory_block *mem;
621 622 623

	mutex_lock(&mem_sysfs_mutex);

624 625 626 627 628 629 630 631 632 633 634 635 636 637
	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;
	}

	if (mem->section_count == sections_per_block)
		ret = register_mem_sect_under_node(mem, nid);
out:
	mutex_unlock(&mem_sysfs_mutex);
638
	return ret;
639 640 641 642 643 644 645 646 647
}

#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() */
648
	put_device(&memory->dev);
649 650 651 652 653
	device_unregister(&memory->dev);
}

static int remove_memory_block(unsigned long node_id,
			       struct mem_section *section, int phys_device)
654 655 656
{
	struct memory_block *mem;

657
	mutex_lock(&mem_sysfs_mutex);
658
	mem = find_memory_block(section);
659
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
660 661

	mem->section_count--;
662
	if (mem->section_count == 0)
663
		unregister_memory(mem);
664
	else
665
		put_device(&mem->dev);
666

667
	mutex_unlock(&mem_sysfs_mutex);
668 669 670 671 672
	return 0;
}

int unregister_memory_section(struct mem_section *section)
{
673
	if (!present_section(section))
674 675 676 677
		return -EINVAL;

	return remove_memory_block(0, section, 0);
}
678
#endif /* CONFIG_MEMORY_HOTREMOVE */
679

680 681 682 683 684 685
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708
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,
	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,
};

709 710 711 712 713 714 715
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
716
	int err;
717
	unsigned long block_sz;
718

719
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
720 721
	if (ret)
		goto out;
722

723 724 725
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

726 727 728 729
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
730
	mutex_lock(&mem_sysfs_mutex);
731 732
	for (i = 0; i < NR_MEM_SECTIONS; i += sections_per_block) {
		err = add_memory_block(i);
733 734
		if (!ret)
			ret = err;
735
	}
736
	mutex_unlock(&mem_sysfs_mutex);
737

738 739
out:
	if (ret)
740
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
741 742
	return ret;
}