memory.c 17.4 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
/*
 * register_memory - Setup a sysfs device for a memory block
 */
83
static
84
int register_memory(struct memory_block *memory)
85 86 87
{
	int error;

88 89
	memory->dev.bus = &memory_subsys;
	memory->dev.id = memory->start_section_nr / sections_per_block;
90
	memory->dev.release = memory_block_release;
91

92
	error = device_register(&memory->dev);
93 94 95 96
	return error;
}

static void
97
unregister_memory(struct memory_block *memory)
98
{
99
	BUG_ON(memory->dev.bus != &memory_subsys);
100

101
	/* drop the ref. we got in remove_memory_block() */
102 103
	kobject_put(&memory->dev.kobj);
	device_unregister(&memory->dev);
104 105
}

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
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;
}

126 127 128 129 130
/*
 * use this as the physical section index that this memsection
 * uses.
 */

131 132
static ssize_t show_mem_start_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
133 134
{
	struct memory_block *mem =
135
		container_of(dev, struct memory_block, dev);
136 137 138 139 140 141
	unsigned long phys_index;

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

142 143
static ssize_t show_mem_end_phys_index(struct device *dev,
			struct device_attribute *attr, char *buf)
144 145
{
	struct memory_block *mem =
146
		container_of(dev, struct memory_block, dev);
147 148 149 150
	unsigned long phys_index;

	phys_index = mem->end_section_nr / sections_per_block;
	return sprintf(buf, "%08lx\n", phys_index);
151 152
}

153 154 155
/*
 * Show whether the section of memory is likely to be hot-removable
 */
156 157
static ssize_t show_mem_removable(struct device *dev,
			struct device_attribute *attr, char *buf)
158
{
159 160
	unsigned long i, pfn;
	int ret = 1;
161
	struct memory_block *mem =
162
		container_of(dev, struct memory_block, dev);
163

164
	for (i = 0; i < sections_per_block; i++) {
165
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
166 167 168
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

169 170 171
	return sprintf(buf, "%d\n", ret);
}

172 173 174
/*
 * online, offline, going offline, etc.
 */
175 176
static ssize_t show_mem_state(struct device *dev,
			struct device_attribute *attr, char *buf)
177 178
{
	struct memory_block *mem =
179
		container_of(dev, struct memory_block, dev);
180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
	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;
}

206
int memory_notify(unsigned long val, void *v)
207
{
208
	return blocking_notifier_call_chain(&memory_chain, val, v);
209 210
}

211 212 213 214 215
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
/*
 * The probe routines leave the pages reserved, just as the bootmem code does.
 * Make sure they're still that way.
 */
static bool pages_correctly_reserved(unsigned long start_pfn,
					unsigned long nr_pages)
{
	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;
}

252 253 254 255 256
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
257
memory_block_action(unsigned long phys_index, unsigned long action)
258
{
259
	unsigned long start_pfn;
260
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
261
	struct page *first_page;
262 263
	int ret;

264
	first_page = pfn_to_page(phys_index << PFN_SECTION_SHIFT);
265
	start_pfn = page_to_pfn(first_page);
266

267 268
	switch (action) {
		case MEM_ONLINE:
269 270 271
			if (!pages_correctly_reserved(start_pfn, nr_pages))
				return -EBUSY;

272
			ret = online_pages(start_pfn, nr_pages);
273 274
			break;
		case MEM_OFFLINE:
275
			ret = offline_pages(start_pfn, nr_pages);
276 277
			break;
		default:
278 279
			WARN(1, KERN_WARNING "%s(%ld, %ld) unknown action: "
			     "%ld\n", __func__, phys_index, action, action);
280 281 282 283 284 285
			ret = -EINVAL;
	}

	return ret;
}

286
static int __memory_block_change_state(struct memory_block *mem,
287 288
		unsigned long to_state, unsigned long from_state_req)
{
289
	int ret = 0;
290

291 292 293 294 295
	if (mem->state != from_state_req) {
		ret = -EINVAL;
		goto out;
	}

296 297 298
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

299
	ret = memory_block_action(mem->start_section_nr, to_state);
300

301
	if (ret) {
302
		mem->state = from_state_req;
303 304
		goto out;
	}
305

306 307 308 309 310 311 312 313 314 315 316
	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;
	}
317 318 319 320
out:
	return ret;
}

321 322 323 324 325 326 327 328 329 330 331
static int memory_block_change_state(struct memory_block *mem,
		unsigned long to_state, unsigned long from_state_req)
{
	int ret;

	mutex_lock(&mem->state_mutex);
	ret = __memory_block_change_state(mem, to_state, from_state_req);
	mutex_unlock(&mem->state_mutex);

	return ret;
}
332
static ssize_t
333 334
store_mem_state(struct device *dev,
		struct device_attribute *attr, const char *buf, size_t count)
335 336 337 338
{
	struct memory_block *mem;
	int ret = -EINVAL;

339
	mem = container_of(dev, struct memory_block, dev);
340 341 342 343 344

	if (!strncmp(buf, "online", min((int)count, 6)))
		ret = memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE);
	else if(!strncmp(buf, "offline", min((int)count, 7)))
		ret = memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
345

346 347 348 349 350 351 352 353 354 355 356 357 358 359
	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?
 */
360 361
static ssize_t show_phys_device(struct device *dev,
				struct device_attribute *attr, char *buf)
362 363
{
	struct memory_block *mem =
364
		container_of(dev, struct memory_block, dev);
365 366 367
	return sprintf(buf, "%d\n", mem->phys_device);
}

368 369 370 371 372
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);
373 374

#define mem_create_simple_file(mem, attr_name)	\
375
	device_create_file(&mem->dev, &dev_attr_##attr_name)
376
#define mem_remove_simple_file(mem, attr_name)	\
377
	device_remove_file(&mem->dev, &dev_attr_##attr_name)
378 379 380 381 382

/*
 * Block size attribute stuff
 */
static ssize_t
383
print_block_size(struct device *dev, struct device_attribute *attr,
384
		 char *buf)
385
{
386
	return sprintf(buf, "%lx\n", get_memory_block_size());
387 388
}

389
static DEVICE_ATTR(block_size_bytes, 0444, print_block_size, NULL);
390 391 392

static int block_size_init(void)
{
393 394
	return device_create_file(memory_subsys.dev_root,
				  &dev_attr_block_size_bytes);
395 396 397 398 399 400 401 402 403 404
}

/*
 * 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
405
memory_probe_store(struct device *dev, struct device_attribute *attr,
406
		   const char *buf, size_t count)
407 408
{
	u64 phys_addr;
409
	int nid;
410
	int i, ret;
411
	unsigned long pages_per_block = PAGES_PER_SECTION * sections_per_block;
412 413 414

	phys_addr = simple_strtoull(buf, NULL, 0);

415 416 417
	if (phys_addr & ((pages_per_block << PAGE_SHIFT) - 1))
		return -EINVAL;

418 419 420 421 422
	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)
423
			goto out;
424 425 426

		phys_addr += MIN_MEMORY_BLOCK_SIZE;
	}
427

428 429 430
	ret = count;
out:
	return ret;
431
}
432
static DEVICE_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
433 434 435

static int memory_probe_init(void)
{
436
	return device_create_file(memory_subsys.dev_root, &dev_attr_probe);
437 438
}
#else
439 440 441 442
static inline int memory_probe_init(void)
{
	return 0;
}
443 444
#endif

445 446 447 448 449 450 451
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
452 453
store_soft_offline_page(struct device *dev,
			struct device_attribute *attr,
454
			const char *buf, size_t count)
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470
{
	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
471 472
store_hard_offline_page(struct device *dev,
			struct device_attribute *attr,
473
			const char *buf, size_t count)
474 475 476 477 478 479 480 481
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (strict_strtoull(buf, 0, &pfn) < 0)
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
482
	ret = memory_failure(pfn, 0, 0);
483 484 485
	return ret ? ret : count;
}

486 487
static DEVICE_ATTR(soft_offline_page, 0644, NULL, store_soft_offline_page);
static DEVICE_ATTR(hard_offline_page, 0644, NULL, store_hard_offline_page);
488 489 490 491 492

static __init int memory_fail_init(void)
{
	int err;

493 494
	err = device_create_file(memory_subsys.dev_root,
				&dev_attr_soft_offline_page);
495
	if (!err)
496 497
		err = device_create_file(memory_subsys.dev_root,
				&dev_attr_hard_offline_page);
498 499 500 501 502 503 504 505 506
	return err;
}
#else
static inline int memory_fail_init(void)
{
	return 0;
}
#endif

507 508 509 510 511
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
512 513 514 515
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
516

517 518 519 520
/*
 * A reference for the returned object is held and the reference for the
 * hinted object is released.
 */
521 522
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
523
{
524
	int block_id = base_memory_block_id(__section_nr(section));
525 526
	struct device *hintdev = hint ? &hint->dev : NULL;
	struct device *dev;
527

528 529 530 531
	dev = subsys_find_device_by_id(&memory_subsys, block_id, hintdev);
	if (hint)
		put_device(&hint->dev);
	if (!dev)
532
		return NULL;
533
	return container_of(dev, struct memory_block, dev);
534 535
}

536 537 538 539 540 541
/*
 * 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.
 *
542
 * This could be made generic for all device subsystems.
543 544 545 546 547 548
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

549 550
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
551
{
552
	struct memory_block *mem;
553
	unsigned long start_pfn;
554
	int scn_nr;
555 556
	int ret = 0;

557
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
558 559 560
	if (!mem)
		return -ENOMEM;

561
	scn_nr = __section_nr(section);
562 563 564
	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;
565
	mem->state = state;
566
	mem->section_count++;
567
	mutex_init(&mem->state_mutex);
568
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
569 570
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

571
	ret = register_memory(mem);
572 573
	if (!ret)
		ret = mem_create_simple_file(mem, phys_index);
574 575
	if (!ret)
		ret = mem_create_simple_file(mem, end_phys_index);
576 577 578 579 580 581
	if (!ret)
		ret = mem_create_simple_file(mem, state);
	if (!ret)
		ret = mem_create_simple_file(mem, phys_device);
	if (!ret)
		ret = mem_create_simple_file(mem, removable);
582 583 584 585 586 587

	*memory = mem;
	return ret;
}

static int add_memory_section(int nid, struct mem_section *section,
588
			struct memory_block **mem_p,
589 590
			unsigned long state, enum mem_add_context context)
{
591 592
	struct memory_block *mem = NULL;
	int scn_nr = __section_nr(section);
593 594 595 596
	int ret = 0;

	mutex_lock(&mem_sysfs_mutex);

597 598 599 600 601 602 603 604 605 606 607
	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);

608 609
	if (mem) {
		mem->section_count++;
610
		kobject_put(&mem->dev.kobj);
611
	} else {
612
		ret = init_memory_block(&mem, section, state);
613 614 615 616 617
		/* store memory_block pointer for next loop */
		if (!ret && context == BOOT)
			if (mem_p)
				*mem_p = mem;
	}
618

619
	if (!ret) {
620 621
		if (context == HOTPLUG &&
		    mem->section_count == sections_per_block)
622 623 624
			ret = register_mem_sect_under_node(mem, nid);
	}

625
	mutex_unlock(&mem_sysfs_mutex);
626 627 628
	return ret;
}

629 630 631 632 633
int remove_memory_block(unsigned long node_id, struct mem_section *section,
		int phys_device)
{
	struct memory_block *mem;

634
	mutex_lock(&mem_sysfs_mutex);
635
	mem = find_memory_block(section);
636
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
637 638 639 640

	mem->section_count--;
	if (mem->section_count == 0) {
		mem_remove_simple_file(mem, phys_index);
641
		mem_remove_simple_file(mem, end_phys_index);
642 643 644
		mem_remove_simple_file(mem, state);
		mem_remove_simple_file(mem, phys_device);
		mem_remove_simple_file(mem, removable);
645 646
		unregister_memory(mem);
	} else
647
		kobject_put(&mem->dev.kobj);
648

649
	mutex_unlock(&mem_sysfs_mutex);
650 651 652 653 654 655 656
	return 0;
}

/*
 * need an interface for the VM to add new memory regions,
 * but without onlining it.
 */
657
int register_new_memory(int nid, struct mem_section *section)
658
{
659
	return add_memory_section(nid, section, NULL, MEM_OFFLINE, HOTPLUG);
660 661 662 663
}

int unregister_memory_section(struct mem_section *section)
{
664
	if (!present_section(section))
665 666 667 668 669
		return -EINVAL;

	return remove_memory_block(0, section, 0);
}

670 671 672 673 674 675 676 677 678 679 680 681 682 683 684
/*
 * 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)
		ret = __memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE);
	mutex_unlock(&mem->state_mutex);

	return ret;
}

685 686 687 688 689 690 691
/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
692
	int err;
693
	unsigned long block_sz;
694
	struct memory_block *mem = NULL;
695

696
	ret = subsys_system_register(&memory_subsys, NULL);
697 698
	if (ret)
		goto out;
699

700 701 702
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

703 704 705 706 707
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
	for (i = 0; i < NR_MEM_SECTIONS; i++) {
708
		if (!present_section_nr(i))
709
			continue;
710 711 712 713
		/* 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,
714
					 BOOT);
715 716
		if (!ret)
			ret = err;
717 718
	}

719
	err = memory_probe_init();
720 721 722
	if (!ret)
		ret = err;
	err = memory_fail_init();
723 724 725 726 727 728 729
	if (!ret)
		ret = err;
	err = block_size_init();
	if (!ret)
		ret = err;
out:
	if (ret)
730
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
731 732
	return ret;
}