memory.c 15.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * drivers/base/memory.c - basic Memory class support
 *
 * 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/sysdev.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/topology.h>
17
#include <linux/capability.h>
18 19 20 21 22
#include <linux/device.h>
#include <linux/memory.h>
#include <linux/kobject.h>
#include <linux/memory_hotplug.h>
#include <linux/mm.h>
23
#include <linux/mutex.h>
24
#include <linux/stat.h>
25
#include <linux/slab.h>
26

27 28 29
#include <asm/atomic.h>
#include <asm/uaccess.h>

30 31
static DEFINE_MUTEX(mem_sysfs_mutex);

32
#define MEMORY_CLASS_NAME	"memory"
33 34 35 36 37 38 39 40
#define MIN_MEMORY_BLOCK_SIZE	(1 << SECTION_SIZE_BITS)

static int sections_per_block;

static inline int base_memory_block_id(int section_nr)
{
	return section_nr / sections_per_block;
}
41 42

static struct sysdev_class memory_sysdev_class = {
43
	.name = MEMORY_CLASS_NAME,
44 45
};

46
static const char *memory_uevent_name(struct kset *kset, struct kobject *kobj)
47 48 49 50
{
	return MEMORY_CLASS_NAME;
}

51
static int memory_uevent(struct kset *kset, struct kobject *obj, struct kobj_uevent_env *env)
52 53 54 55 56 57
{
	int retval = 0;

	return retval;
}

58
static const struct kset_uevent_ops memory_uevent_ops = {
59 60
	.name		= memory_uevent_name,
	.uevent		= memory_uevent,
61 62
};

63
static BLOCKING_NOTIFIER_HEAD(memory_chain);
64

65
int register_memory_notifier(struct notifier_block *nb)
66
{
67
        return blocking_notifier_chain_register(&memory_chain, nb);
68
}
69
EXPORT_SYMBOL(register_memory_notifier);
70

71
void unregister_memory_notifier(struct notifier_block *nb)
72
{
73
        blocking_notifier_chain_unregister(&memory_chain, nb);
74
}
75
EXPORT_SYMBOL(unregister_memory_notifier);
76

77 78 79 80 81 82 83 84 85 86 87 88 89 90
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);

91 92 93
/*
 * register_memory - Setup a sysfs device for a memory block
 */
94
static
95
int register_memory(struct memory_block *memory)
96 97 98 99
{
	int error;

	memory->sysdev.cls = &memory_sysdev_class;
100
	memory->sysdev.id = memory->phys_index / sections_per_block;
101 102 103 104 105 106

	error = sysdev_register(&memory->sysdev);
	return error;
}

static void
107
unregister_memory(struct memory_block *memory)
108 109 110
{
	BUG_ON(memory->sysdev.cls != &memory_sysdev_class);

111 112
	/* drop the ref. we got in remove_memory_block() */
	kobject_put(&memory->sysdev.kobj);
113 114 115
	sysdev_unregister(&memory->sysdev);
}

116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
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;
}

136 137 138 139 140
/*
 * use this as the physical section index that this memsection
 * uses.
 */

141 142
static ssize_t show_mem_phys_index(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
143 144 145
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
146
	return sprintf(buf, "%08lx\n", mem->phys_index / sections_per_block);
147 148
}

149 150 151
/*
 * Show whether the section of memory is likely to be hot-removable
 */
152 153
static ssize_t show_mem_removable(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
154
{
155 156
	unsigned long i, pfn;
	int ret = 1;
157 158 159
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);

160 161 162 163 164
	for (i = 0; i < sections_per_block; i++) {
		pfn = section_nr_to_pfn(mem->phys_index + i);
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

165 166 167
	return sprintf(buf, "%d\n", ret);
}

168 169 170
/*
 * online, offline, going offline, etc.
 */
171 172
static ssize_t show_mem_state(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
	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;
}

202
int memory_notify(unsigned long val, void *v)
203
{
204
	return blocking_notifier_call_chain(&memory_chain, val, v);
205 206
}

207 208 209 210 211
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

212 213 214 215 216
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
217
memory_section_action(unsigned long phys_index, unsigned long action)
218 219 220 221 222 223
{
	int i;
	unsigned long start_pfn, start_paddr;
	struct page *first_page;
	int ret;

224
	first_page = pfn_to_page(phys_index << PFN_SECTION_SHIFT);
225 226 227 228 229 230 231 232 233 234 235 236

	/*
	 * The probe routines leave the pages reserved, just
	 * as the bootmem code does.  Make sure they're still
	 * that way.
	 */
	if (action == MEM_ONLINE) {
		for (i = 0; i < PAGES_PER_SECTION; i++) {
			if (PageReserved(first_page+i))
				continue;

			printk(KERN_WARNING "section number %ld page number %d "
237 238
				"not reserved, was it already online?\n",
				phys_index, i);
239 240 241 242 243 244 245 246 247 248 249 250 251 252 253
			return -EBUSY;
		}
	}

	switch (action) {
		case MEM_ONLINE:
			start_pfn = page_to_pfn(first_page);
			ret = online_pages(start_pfn, PAGES_PER_SECTION);
			break;
		case MEM_OFFLINE:
			start_paddr = page_to_pfn(first_page) << PAGE_SHIFT;
			ret = remove_memory(start_paddr,
					    PAGES_PER_SECTION << PAGE_SHIFT);
			break;
		default:
254 255
			WARN(1, KERN_WARNING "%s(%ld, %ld) unknown action: "
			     "%ld\n", __func__, phys_index, action, action);
256 257 258 259 260 261 262 263 264
			ret = -EINVAL;
	}

	return ret;
}

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

267
	mutex_lock(&mem->state_mutex);
268 269 270 271 272 273

	if (mem->state != from_state_req) {
		ret = -EINVAL;
		goto out;
	}

274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

	for (i = 0; i < sections_per_block; i++) {
		ret = memory_section_action(mem->phys_index + i, to_state);
		if (ret)
			break;
	}

	if (ret) {
		for (i = 0; i < sections_per_block; i++)
			memory_section_action(mem->phys_index + i,
					      from_state_req);

		mem->state = from_state_req;
	} else
290 291 292
		mem->state = to_state;

out:
293
	mutex_unlock(&mem->state_mutex);
294 295 296 297
	return ret;
}

static ssize_t
298 299
store_mem_state(struct sys_device *dev,
		struct sysdev_attribute *attr, const char *buf, size_t count)
300 301 302 303 304 305 306 307 308 309
{
	struct memory_block *mem;
	int ret = -EINVAL;

	mem = container_of(dev, struct memory_block, sysdev);

	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);
310

311 312 313 314 315 316 317 318 319 320 321 322 323 324
	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?
 */
325 326
static ssize_t show_phys_device(struct sys_device *dev,
				struct sysdev_attribute *attr, char *buf)
327 328 329 330 331 332 333 334 335
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
	return sprintf(buf, "%d\n", mem->phys_device);
}

static SYSDEV_ATTR(phys_index, 0444, show_mem_phys_index, NULL);
static SYSDEV_ATTR(state, 0644, show_mem_state, store_mem_state);
static SYSDEV_ATTR(phys_device, 0444, show_phys_device, NULL);
336
static SYSDEV_ATTR(removable, 0444, show_mem_removable, NULL);
337 338 339 340 341 342 343 344 345 346

#define mem_create_simple_file(mem, attr_name)	\
	sysdev_create_file(&mem->sysdev, &attr_##attr_name)
#define mem_remove_simple_file(mem, attr_name)	\
	sysdev_remove_file(&mem->sysdev, &attr_##attr_name)

/*
 * Block size attribute stuff
 */
static ssize_t
347 348
print_block_size(struct sysdev_class *class, struct sysdev_class_attribute *attr,
		 char *buf)
349
{
350
	return sprintf(buf, "%lx\n", get_memory_block_size());
351 352
}

353
static SYSDEV_CLASS_ATTR(block_size_bytes, 0444, print_block_size, NULL);
354 355 356

static int block_size_init(void)
{
357
	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
358
				&attr_block_size_bytes.attr);
359 360 361 362 363 364 365 366 367 368
}

/*
 * 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
369 370
memory_probe_store(struct class *class, struct class_attribute *attr,
		   const char *buf, size_t count)
371 372
{
	u64 phys_addr;
373
	int nid;
374 375 376 377
	int ret;

	phys_addr = simple_strtoull(buf, NULL, 0);

378 379
	nid = memory_add_physaddr_to_nid(phys_addr);
	ret = add_memory(nid, phys_addr, PAGES_PER_SECTION << PAGE_SHIFT);
380 381 382 383 384 385

	if (ret)
		count = ret;

	return count;
}
386
static CLASS_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
387 388 389

static int memory_probe_init(void)
{
390
	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
391
				&class_attr_probe.attr);
392 393
}
#else
394 395 396 397
static inline int memory_probe_init(void)
{
	return 0;
}
398 399
#endif

400 401 402 403 404 405 406
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
407 408 409
store_soft_offline_page(struct class *class,
			struct class_attribute *attr,
			const char *buf, size_t count)
410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
{
	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
426 427 428
store_hard_offline_page(struct class *class,
			struct class_attribute *attr,
			const char *buf, size_t count)
429 430 431 432 433 434 435 436 437 438 439 440
{
	int ret;
	u64 pfn;
	if (!capable(CAP_SYS_ADMIN))
		return -EPERM;
	if (strict_strtoull(buf, 0, &pfn) < 0)
		return -EINVAL;
	pfn >>= PAGE_SHIFT;
	ret = __memory_failure(pfn, 0, 0);
	return ret ? ret : count;
}

441 442
static CLASS_ATTR(soft_offline_page, 0644, NULL, store_soft_offline_page);
static CLASS_ATTR(hard_offline_page, 0644, NULL, store_hard_offline_page);
443 444 445 446 447 448

static __init int memory_fail_init(void)
{
	int err;

	err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
449
				&class_attr_soft_offline_page.attr);
450 451
	if (!err)
		err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
452
				&class_attr_hard_offline_page.attr);
453 454 455 456 457 458 459 460 461
	return err;
}
#else
static inline int memory_fail_init(void)
{
	return 0;
}
#endif

462 463 464 465 466
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
467 468 469 470
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
471

472 473
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
474 475 476 477 478
{
	struct kobject *kobj;
	struct sys_device *sysdev;
	struct memory_block *mem;
	char name[sizeof(MEMORY_CLASS_NAME) + 9 + 1];
479
	int block_id = base_memory_block_id(__section_nr(section));
480

481 482
	kobj = hint ? &hint->sysdev.kobj : NULL;

483 484 485 486
	/*
	 * This only works because we know that section == sysdev->id
	 * slightly redundant with sysdev_register()
	 */
487
	sprintf(&name[0], "%s%d", MEMORY_CLASS_NAME, block_id);
488

489
	kobj = kset_find_obj_hinted(&memory_sysdev_class.kset, name, kobj);
490 491 492 493 494 495 496 497 498
	if (!kobj)
		return NULL;

	sysdev = container_of(kobj, struct sys_device, kobj);
	mem = container_of(sysdev, struct memory_block, sysdev);

	return mem;
}

499 500 501 502 503 504 505 506 507 508 509 510 511
/*
 * 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.
 *
 * This could be made generic for all sysdev classes.
 */
struct memory_block *find_memory_block(struct mem_section *section)
{
	return find_memory_block_hinted(section, NULL);
}

512 513
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
514
{
515
	struct memory_block *mem;
516
	unsigned long start_pfn;
517
	int scn_nr;
518 519
	int ret = 0;

520
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
521 522 523
	if (!mem)
		return -ENOMEM;

524 525
	scn_nr = __section_nr(section);
	mem->phys_index = base_memory_block_id(scn_nr) * sections_per_block;
526
	mem->state = state;
527
	mem->section_count++;
528 529 530 531
	mutex_init(&mem->state_mutex);
	start_pfn = section_nr_to_pfn(mem->phys_index);
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

532
	ret = register_memory(mem);
533 534 535 536 537 538 539 540
	if (!ret)
		ret = mem_create_simple_file(mem, phys_index);
	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);
541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560

	*memory = mem;
	return ret;
}

static int add_memory_section(int nid, struct mem_section *section,
			unsigned long state, enum mem_add_context context)
{
	struct memory_block *mem;
	int ret = 0;

	mutex_lock(&mem_sysfs_mutex);

	mem = find_memory_block(section);
	if (mem) {
		mem->section_count++;
		kobject_put(&mem->sysdev.kobj);
	} else
		ret = init_memory_block(&mem, section, state);

561
	if (!ret) {
562 563
		if (context == HOTPLUG &&
		    mem->section_count == sections_per_block)
564 565 566
			ret = register_mem_sect_under_node(mem, nid);
	}

567
	mutex_unlock(&mem_sysfs_mutex);
568 569 570
	return ret;
}

571 572 573 574 575
int remove_memory_block(unsigned long node_id, struct mem_section *section,
		int phys_device)
{
	struct memory_block *mem;

576
	mutex_lock(&mem_sysfs_mutex);
577
	mem = find_memory_block(section);
578 579 580 581 582 583 584 585

	mem->section_count--;
	if (mem->section_count == 0) {
		unregister_mem_sect_under_nodes(mem);
		mem_remove_simple_file(mem, phys_index);
		mem_remove_simple_file(mem, state);
		mem_remove_simple_file(mem, phys_device);
		mem_remove_simple_file(mem, removable);
586 587 588 589
		unregister_memory(mem);
		kfree(mem);
	} else
		kobject_put(&mem->sysdev.kobj);
590

591
	mutex_unlock(&mem_sysfs_mutex);
592 593 594 595 596 597 598
	return 0;
}

/*
 * need an interface for the VM to add new memory regions,
 * but without onlining it.
 */
599
int register_new_memory(int nid, struct mem_section *section)
600
{
601
	return add_memory_section(nid, section, MEM_OFFLINE, HOTPLUG);
602 603 604 605
}

int unregister_memory_section(struct mem_section *section)
{
606
	if (!present_section(section))
607 608 609 610 611 612 613 614 615 616 617 618
		return -EINVAL;

	return remove_memory_block(0, section, 0);
}

/*
 * Initialize the sysfs support for memory devices...
 */
int __init memory_dev_init(void)
{
	unsigned int i;
	int ret;
619
	int err;
620
	unsigned long block_sz;
621

622
	memory_sysdev_class.kset.uevent_ops = &memory_uevent_ops;
623
	ret = sysdev_class_register(&memory_sysdev_class);
624 625
	if (ret)
		goto out;
626

627 628 629
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

630 631 632 633 634
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
	for (i = 0; i < NR_MEM_SECTIONS; i++) {
635
		if (!present_section_nr(i))
636
			continue;
637 638
		err = add_memory_section(0, __nr_to_section(i), MEM_ONLINE,
					 BOOT);
639 640
		if (!ret)
			ret = err;
641 642
	}

643
	err = memory_probe_init();
644 645 646
	if (!ret)
		ret = err;
	err = memory_fail_init();
647 648 649 650 651 652 653
	if (!ret)
		ret = err;
	err = block_size_init();
	if (!ret)
		ret = err;
out:
	if (ret)
654
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
655 656
	return ret;
}