memory.c 17.5 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 21
#include <linux/device.h>
#include <linux/memory.h>
#include <linux/kobject.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 28
#include <asm/uaccess.h>

29 30
static DEFINE_MUTEX(mem_sysfs_mutex);

31
#define MEMORY_CLASS_NAME	"memory"
32 33 34 35 36 37 38

static int sections_per_block;

static inline int base_memory_block_id(int section_nr)
{
	return section_nr / sections_per_block;
}
39

40
static struct bus_type memory_subsys = {
41
	.name = MEMORY_CLASS_NAME,
42
	.dev_name = MEMORY_CLASS_NAME,
43 44
};

45
static BLOCKING_NOTIFIER_HEAD(memory_chain);
46

47
int register_memory_notifier(struct notifier_block *nb)
48
{
49
        return blocking_notifier_chain_register(&memory_chain, nb);
50
}
51
EXPORT_SYMBOL(register_memory_notifier);
52

53
void unregister_memory_notifier(struct notifier_block *nb)
54
{
55
        blocking_notifier_chain_unregister(&memory_chain, nb);
56
}
57
EXPORT_SYMBOL(unregister_memory_notifier);
58

59 60 61 62 63 64 65 66 67 68 69 70 71 72
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);

73 74 75 76 77 78 79
static void memory_block_release(struct device *dev)
{
	struct memory_block *mem = container_of(dev, struct memory_block, dev);

	kfree(mem);
}

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
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;
}

100 101 102 103 104
/*
 * use this as the physical section index that this memsection
 * uses.
 */

105 106
static ssize_t show_mem_start_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
107 108
{
	struct memory_block *mem =
109
		container_of(dev, struct 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
static ssize_t show_mem_end_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
118 119
{
	struct memory_block *mem =
120
		container_of(dev, struct memory_block, dev);
121 122 123 124
	unsigned long phys_index;

	phys_index = mem->end_section_nr / sections_per_block;
	return sprintf(buf, "%08lx\n", phys_index);
125 126
}

127 128 129
/*
 * Show whether the section of memory is likely to be hot-removable
 */
130 131
static ssize_t show_mem_removable(struct device *dev,
			struct device_attribute *attr, char *buf)
132
{
133 134
	unsigned long i, pfn;
	int ret = 1;
135
	struct memory_block *mem =
136
		container_of(dev, struct memory_block, dev);
137

138
	for (i = 0; i < sections_per_block; i++) {
139
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
140 141 142
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

143 144 145
	return sprintf(buf, "%d\n", ret);
}

146 147 148
/*
 * online, offline, going offline, etc.
 */
149 150
static ssize_t show_mem_state(struct device *dev,
			struct device_attribute *attr, char *buf)
151 152
{
	struct memory_block *mem =
153
		container_of(dev, struct memory_block, dev);
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
	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;
}

180
int memory_notify(unsigned long val, void *v)
181
{
182
	return blocking_notifier_call_chain(&memory_chain, val, v);
183 184
}

185 186 187 188 189
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

190 191 192 193
/*
 * The probe routines leave the pages reserved, just as the bootmem code does.
 * Make sure they're still that way.
 */
194
static bool pages_correctly_reserved(unsigned long start_pfn)
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
{
	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;
}

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
memory_block_action(unsigned long phys_index, unsigned long action, int online_type)
231
{
232
	unsigned long start_pfn;
233
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
234
	struct page *first_page;
235 236
	int ret;

237
	first_page = pfn_to_page(phys_index << PFN_SECTION_SHIFT);
238
	start_pfn = page_to_pfn(first_page);
239

240 241
	switch (action) {
		case MEM_ONLINE:
242
			if (!pages_correctly_reserved(start_pfn))
243 244
				return -EBUSY;

245
			ret = online_pages(start_pfn, nr_pages, online_type);
246 247
			break;
		case MEM_OFFLINE:
248
			ret = offline_pages(start_pfn, nr_pages);
249 250
			break;
		default:
251 252
			WARN(1, KERN_WARNING "%s(%ld, %ld) unknown action: "
			     "%ld\n", __func__, phys_index, action, action);
253 254 255 256 257 258
			ret = -EINVAL;
	}

	return ret;
}

259
static int __memory_block_change_state(struct memory_block *mem,
260 261
		unsigned long to_state, unsigned long from_state_req,
		int online_type)
262
{
263
	int ret = 0;
264

265 266 267 268 269
	if (mem->state != from_state_req) {
		ret = -EINVAL;
		goto out;
	}

270 271 272
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

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

275
	if (ret) {
276
		mem->state = from_state_req;
277 278
		goto out;
	}
279

280 281 282 283 284 285 286 287 288 289 290
	mem->state = to_state;
	switch (mem->state) {
	case MEM_OFFLINE:
		kobject_uevent(&mem->dev.kobj, KOBJ_OFFLINE);
		break;
	case MEM_ONLINE:
		kobject_uevent(&mem->dev.kobj, KOBJ_ONLINE);
		break;
	default:
		break;
	}
291 292 293 294
out:
	return ret;
}

295
static int memory_block_change_state(struct memory_block *mem,
296 297
		unsigned long to_state, unsigned long from_state_req,
		int online_type)
298 299 300 301
{
	int ret;

	mutex_lock(&mem->state_mutex);
302 303
	ret = __memory_block_change_state(mem, to_state, from_state_req,
					  online_type);
304 305 306 307
	mutex_unlock(&mem->state_mutex);

	return ret;
}
308
static ssize_t
309 310
store_mem_state(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
311 312 313 314
{
	struct memory_block *mem;
	int ret = -EINVAL;

315
	mem = container_of(dev, struct memory_block, dev);
316

317 318 319 320 321 322 323 324 325 326 327 328
	if (!strncmp(buf, "online_kernel", min_t(int, count, 13)))
		ret = memory_block_change_state(mem, MEM_ONLINE,
						MEM_OFFLINE, ONLINE_KERNEL);
	else if (!strncmp(buf, "online_movable", min_t(int, count, 14)))
		ret = memory_block_change_state(mem, MEM_ONLINE,
						MEM_OFFLINE, ONLINE_MOVABLE);
	else if (!strncmp(buf, "online", min_t(int, count, 6)))
		ret = memory_block_change_state(mem, MEM_ONLINE,
						MEM_OFFLINE, ONLINE_KEEP);
	else if(!strncmp(buf, "offline", min_t(int, count, 7)))
		ret = memory_block_change_state(mem, MEM_OFFLINE,
						MEM_ONLINE, -1);
329

330 331 332 333 334 335 336 337 338 339 340 341 342 343
	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?
 */
344 345
static ssize_t show_phys_device(struct device *dev,
				struct device_attribute *attr, char *buf)
346 347
{
	struct memory_block *mem =
348
		container_of(dev, struct memory_block, dev);
349 350 351
	return sprintf(buf, "%d\n", mem->phys_device);
}

352 353 354 355 356
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);
357 358 359 360 361

/*
 * Block size attribute stuff
 */
static ssize_t
362
print_block_size(struct device *dev, struct device_attribute *attr,
363
		 char *buf)
364
{
365
	return sprintf(buf, "%lx\n", get_memory_block_size());
366 367
}

368
static DEVICE_ATTR(block_size_bytes, 0444, print_block_size, NULL);
369 370 371 372 373 374 375 376 377

/*
 * 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
378
memory_probe_store(struct device *dev, struct device_attribute *attr,
379
		   const char *buf, size_t count)
380 381
{
	u64 phys_addr;
382
	int nid;
383
	int i, ret;
384
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
385 386 387

	phys_addr = simple_strtoull(buf, NULL, 0);

388 389 390
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

391 392 393 394 395
	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)
396
			goto out;
397 398 399

		phys_addr += MIN_MEMORY_BLOCK_SIZE;
	}
400

401 402 403
	ret = count;
out:
	return ret;
404 405
}

406
static DEVICE_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
407 408
#endif

409 410 411 412 413 414 415
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
416 417
store_soft_offline_page(struct device *dev,
			struct device_attribute *attr,
418
			const char *buf, size_t count)
419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (strict_strtoull(buf, 0, &pfn) < 0)
		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
435 436
store_hard_offline_page(struct device *dev,
			struct device_attribute *attr,
437
			const char *buf, size_t count)
438 439 440 441 442 443 444 445
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (strict_strtoull(buf, 0, &pfn) < 0)
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
446
	ret = memory_failure(pfn, 0, 0);
447 448 449
	return ret ? ret : count;
}

450 451
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);
452 453
#endif

454 455 456 457 458
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
459 460 461 462
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
463

464 465 466 467
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
468 469
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
470
{
471
	int block_id = base_memory_block_id(__section_nr(section));
472 473
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
474

475 476 477 478
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
479
		return NULL;
480
	return container_of(dev, struct memory_block, dev);
481 482
}

483 484 485 486 487 488
/*
 * 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.
 *
489
 * This could be made generic for all device subsystems.
490 491 492 493 494 495
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530
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)
{
	int error;

	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;

	error = device_register(&memory->dev);
	return error;
}

531 532
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
533
{
534
	struct memory_block *mem;
535
	unsigned long start_pfn;
536
	int scn_nr;
537 538
	int ret = 0;

539
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
540 541 542
	if (!mem)
		return -ENOMEM;

543
	scn_nr = __section_nr(section);
544 545 546
	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;
547
	mem->state = state;
548
	mem->section_count++;
549
	mutex_init(&mem->state_mutex);
550
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
551 552
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

553 554 555 556 557 558 559
	ret = register_memory(mem);

	*memory = mem;
	return ret;
}

static int add_memory_section(int nid, struct mem_section *section,
560
			struct memory_block **mem_p,
561 562
			unsigned long state, enum mem_add_context context)
{
563 564
	struct memory_block *mem = NULL;
	int scn_nr = __section_nr(section);
565 566 567 568
	int ret = 0;

	mutex_lock(&mem_sysfs_mutex);

569 570 571 572 573 574 575 576 577 578 579
	if (context == BOOT) {
		/* same memory block ? */
		if (mem_p && *mem_p)
			if (scn_nr >= (*mem_p)->start_section_nr &&
			    scn_nr <= (*mem_p)->end_section_nr) {
				mem = *mem_p;
				kobject_get(&mem->dev.kobj);
			}
	} else
		mem = find_memory_block(section);

580 581
	if (mem) {
		mem->section_count++;
582
		kobject_put(&mem->dev.kobj);
583
	} else {
584
		ret = init_memory_block(&mem, section, state);
585 586 587 588 589
		/* store memory_block pointer for next loop */
		if (!ret && context == BOOT)
			if (mem_p)
				*mem_p = mem;
	}
590

591
	if (!ret) {
592 593
		if (context == HOTPLUG &&
		    mem->section_count == sections_per_block)
594 595 596
			ret = register_mem_sect_under_node(mem, nid);
	}

597
	mutex_unlock(&mem_sysfs_mutex);
598 599 600
	return ret;
}

601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
/*
 * 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)
{
	return add_memory_section(nid, section, NULL, MEM_OFFLINE, HOTPLUG);
}

#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() */
	kobject_put(&memory->dev.kobj);
	device_unregister(&memory->dev);
}

static int remove_memory_block(unsigned long node_id,
			       struct mem_section *section, int phys_device)
623 624 625
{
	struct memory_block *mem;

626
	mutex_lock(&mem_sysfs_mutex);
627
	mem = find_memory_block(section);
628
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
629 630

	mem->section_count--;
631
	if (mem->section_count == 0)
632
		unregister_memory(mem);
633
	else
634
		kobject_put(&mem->dev.kobj);
635

636
	mutex_unlock(&mem_sysfs_mutex);
637 638 639 640 641
	return 0;
}

int unregister_memory_section(struct mem_section *section)
{
642
	if (!present_section(section))
643 644 645 646
		return -EINVAL;

	return remove_memory_block(0, section, 0);
}
647
#endif /* CONFIG_MEMORY_HOTREMOVE */
648

649 650 651 652 653 654 655 656 657
/*
 * offline one memory block. If the memory block has been offlined, do nothing.
 */
int offline_memory_block(struct memory_block *mem)
{
	int ret = 0;

	mutex_lock(&mem->state_mutex);
	if (mem->state != MEM_OFFLINE)
658
		ret = __memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE, -1);
659 660 661 662 663
	mutex_unlock(&mem->state_mutex);

	return ret;
}

664 665 666 667 668 669
/* return true if the memory block is offlined, otherwise, return false */
bool is_memblock_offlined(struct memory_block *mem)
{
	return mem->state == MEM_OFFLINE;
}

670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
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,
};

693 694 695 696 697 698 699
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
700
	int err;
701
	unsigned long block_sz;
702
	struct memory_block *mem = NULL;
703

704
	ret = subsys_system_register(&memory_subsys, memory_root_attr_groups);
705 706
	if (ret)
		goto out;
707

708 709 710
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

711 712 713 714 715
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
	for (i = 0; i < NR_MEM_SECTIONS; i++) {
716
		if (!present_section_nr(i))
717
			continue;
718 719 720 721
		/* don't need to reuse memory_block if only one per block */
		err = add_memory_section(0, __nr_to_section(i),
				 (sections_per_block == 1) ? NULL : &mem,
					 MEM_ONLINE,
722
					 BOOT);
723 724
		if (!ret)
			ret = err;
725 726
	}

727 728
out:
	if (ret)
729
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
730 731
	return ret;
}