memory.c 16.1 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

static int sections_per_block;

static inline int base_memory_block_id(int section_nr)
{
	return section_nr / sections_per_block;
}
40
41

static struct sysdev_class memory_sysdev_class = {
42
	.name = MEMORY_CLASS_NAME,
43
44
};

45
static const char *memory_uevent_name(struct kset *kset, struct kobject *kobj)
46
47
48
49
{
	return MEMORY_CLASS_NAME;
}

50
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->start_section_nr / 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
static ssize_t show_mem_start_phys_index(struct sys_device *dev,
142
			struct sysdev_attribute *attr, char *buf)
143
144
145
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
	unsigned long phys_index;

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

static ssize_t show_mem_end_phys_index(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
	unsigned long phys_index;

	phys_index = mem->end_section_nr / sections_per_block;
	return sprintf(buf, "%08lx\n", phys_index);
161
162
}

163
164
165
/*
 * Show whether the section of memory is likely to be hot-removable
 */
166
167
static ssize_t show_mem_removable(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
168
{
169
170
	unsigned long i, pfn;
	int ret = 1;
171
172
173
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);

174
	for (i = 0; i < sections_per_block; i++) {
175
		pfn = section_nr_to_pfn(mem->start_section_nr + i);
176
177
178
		ret &= is_mem_section_removable(pfn, PAGES_PER_SECTION);
	}

179
180
181
	return sprintf(buf, "%d\n", ret);
}

182
183
184
/*
 * online, offline, going offline, etc.
 */
185
186
static ssize_t show_mem_state(struct sys_device *dev,
			struct sysdev_attribute *attr, char *buf)
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
{
	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;
}

216
int memory_notify(unsigned long val, void *v)
217
{
218
	return blocking_notifier_call_chain(&memory_chain, val, v);
219
220
}

221
222
223
224
225
int memory_isolate_notify(unsigned long val, void *v)
{
	return atomic_notifier_call_chain(&memory_isolate_chain, val, v);
}

226
227
228
229
230
/*
 * MEMORY_HOTPLUG depends on SPARSEMEM in mm/Kconfig, so it is
 * OK to have direct references to sparsemem variables in here.
 */
static int
231
memory_block_action(unsigned long phys_index, unsigned long action)
232
233
234
{
	int i;
	unsigned long start_pfn, start_paddr;
235
	unsigned long nr_pages = PAGES_PER_SECTION * sections_per_block;
236
237
238
	struct page *first_page;
	int ret;

239
	first_page = pfn_to_page(phys_index << PFN_SECTION_SHIFT);
240
241
242
243
244
245
246

	/*
	 * The probe routines leave the pages reserved, just
	 * as the bootmem code does.  Make sure they're still
	 * that way.
	 */
	if (action == MEM_ONLINE) {
247
		for (i = 0; i < nr_pages; i++) {
248
249
250
251
			if (PageReserved(first_page+i))
				continue;

			printk(KERN_WARNING "section number %ld page number %d "
252
253
				"not reserved, was it already online?\n",
				phys_index, i);
254
255
256
257
258
259
260
			return -EBUSY;
		}
	}

	switch (action) {
		case MEM_ONLINE:
			start_pfn = page_to_pfn(first_page);
261
			ret = online_pages(start_pfn, nr_pages);
262
263
264
265
			break;
		case MEM_OFFLINE:
			start_paddr = page_to_pfn(first_page) << PAGE_SHIFT;
			ret = remove_memory(start_paddr,
266
					    nr_pages << PAGE_SHIFT);
267
268
			break;
		default:
269
270
			WARN(1, KERN_WARNING "%s(%ld, %ld) unknown action: "
			     "%ld\n", __func__, phys_index, action, action);
271
272
273
274
275
276
277
278
279
			ret = -EINVAL;
	}

	return ret;
}

static int memory_block_change_state(struct memory_block *mem,
		unsigned long to_state, unsigned long from_state_req)
{
280
	int ret = 0;
281

282
	mutex_lock(&mem->state_mutex);
283
284
285
286
287
288

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

289
290
291
	if (to_state == MEM_OFFLINE)
		mem->state = MEM_GOING_OFFLINE;

292
	ret = memory_block_action(mem->start_section_nr, to_state);
293

294
	if (ret)
295
		mem->state = from_state_req;
296
	else
297
298
299
		mem->state = to_state;

out:
300
	mutex_unlock(&mem->state_mutex);
301
302
303
304
	return ret;
}

static ssize_t
305
306
store_mem_state(struct sys_device *dev,
		struct sysdev_attribute *attr, const char *buf, size_t count)
307
308
309
310
311
312
313
314
315
316
{
	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);
317

318
319
320
321
322
323
324
325
326
327
328
329
330
331
	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?
 */
332
333
static ssize_t show_phys_device(struct sys_device *dev,
				struct sysdev_attribute *attr, char *buf)
334
335
336
337
338
339
{
	struct memory_block *mem =
		container_of(dev, struct memory_block, sysdev);
	return sprintf(buf, "%d\n", mem->phys_device);
}

340
341
static SYSDEV_ATTR(phys_index, 0444, show_mem_start_phys_index, NULL);
static SYSDEV_ATTR(end_phys_index, 0444, show_mem_end_phys_index, NULL);
342
343
static SYSDEV_ATTR(state, 0644, show_mem_state, store_mem_state);
static SYSDEV_ATTR(phys_device, 0444, show_phys_device, NULL);
344
static SYSDEV_ATTR(removable, 0444, show_mem_removable, NULL);
345
346
347
348
349
350
351
352
353
354

#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
355
356
print_block_size(struct sysdev_class *class, struct sysdev_class_attribute *attr,
		 char *buf)
357
{
358
	return sprintf(buf, "%lx\n", get_memory_block_size());
359
360
}

361
static SYSDEV_CLASS_ATTR(block_size_bytes, 0444, print_block_size, NULL);
362
363
364

static int block_size_init(void)
{
365
	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
366
				&attr_block_size_bytes.attr);
367
368
369
370
371
372
373
374
375
376
}

/*
 * 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
377
378
memory_probe_store(struct class *class, struct class_attribute *attr,
		   const char *buf, size_t count)
379
380
{
	u64 phys_addr;
381
	int nid;
382
	int i, ret;
383
384
385

	phys_addr = simple_strtoull(buf, NULL, 0);

386
387
388
389
390
	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)
391
			goto out;
392
393
394

		phys_addr += MIN_MEMORY_BLOCK_SIZE;
	}
395

396
397
398
	ret = count;
out:
	return ret;
399
}
400
static CLASS_ATTR(probe, S_IWUSR, NULL, memory_probe_store);
401
402
403

static int memory_probe_init(void)
{
404
	return sysfs_create_file(&memory_sysdev_class.kset.kobj,
405
				&class_attr_probe.attr);
406
407
}
#else
408
409
410
411
static inline int memory_probe_init(void)
{
	return 0;
}
412
413
#endif

414
415
416
417
418
419
420
#ifdef CONFIG_MEMORY_FAILURE
/*
 * Support for offlining pages of memory
 */

/* Soft offline a page */
static ssize_t
421
422
423
store_soft_offline_page(struct class *class,
			struct class_attribute *attr,
			const char *buf, size_t count)
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
{
	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
440
441
442
store_hard_offline_page(struct class *class,
			struct class_attribute *attr,
			const char *buf, size_t count)
443
444
445
446
447
448
449
450
451
452
453
454
{
	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;
}

455
456
static CLASS_ATTR(soft_offline_page, 0644, NULL, store_soft_offline_page);
static CLASS_ATTR(hard_offline_page, 0644, NULL, store_hard_offline_page);
457
458
459
460
461
462

static __init int memory_fail_init(void)
{
	int err;

	err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
463
				&class_attr_soft_offline_page.attr);
464
465
	if (!err)
		err = sysfs_create_file(&memory_sysdev_class.kset.kobj,
466
				&class_attr_hard_offline_page.attr);
467
468
469
470
471
472
473
474
475
	return err;
}
#else
static inline int memory_fail_init(void)
{
	return 0;
}
#endif

476
477
478
479
480
/*
 * Note that phys_device is optional.  It is here to allow for
 * differentiation between which *physical* devices each
 * section belongs to...
 */
481
482
483
484
int __weak arch_get_memory_phys_device(unsigned long start_pfn)
{
	return 0;
}
485

486
487
struct memory_block *find_memory_block_hinted(struct mem_section *section,
					      struct memory_block *hint)
488
489
490
491
492
{
	struct kobject *kobj;
	struct sys_device *sysdev;
	struct memory_block *mem;
	char name[sizeof(MEMORY_CLASS_NAME) + 9 + 1];
493
	int block_id = base_memory_block_id(__section_nr(section));
494

495
496
	kobj = hint ? &hint->sysdev.kobj : NULL;

497
498
499
500
	/*
	 * This only works because we know that section == sysdev->id
	 * slightly redundant with sysdev_register()
	 */
501
	sprintf(&name[0], "%s%d", MEMORY_CLASS_NAME, block_id);
502

503
	kobj = kset_find_obj_hinted(&memory_sysdev_class.kset, name, kobj);
504
505
506
507
508
509
510
511
512
	if (!kobj)
		return NULL;

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

	return mem;
}

513
514
515
516
517
518
519
520
521
522
523
524
525
/*
 * 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);
}

526
527
static int init_memory_block(struct memory_block **memory,
			     struct mem_section *section, unsigned long state)
528
{
529
	struct memory_block *mem;
530
	unsigned long start_pfn;
531
	int scn_nr;
532
533
	int ret = 0;

534
	mem = kzalloc(sizeof(*mem), GFP_KERNEL);
535
536
537
	if (!mem)
		return -ENOMEM;

538
	scn_nr = __section_nr(section);
539
540
541
	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;
542
	mem->state = state;
543
	mem->section_count++;
544
	mutex_init(&mem->state_mutex);
545
	start_pfn = section_nr_to_pfn(mem->start_section_nr);
546
547
	mem->phys_device = arch_get_memory_phys_device(start_pfn);

548
	ret = register_memory(mem);
549
550
	if (!ret)
		ret = mem_create_simple_file(mem, phys_index);
551
552
	if (!ret)
		ret = mem_create_simple_file(mem, end_phys_index);
553
554
555
556
557
558
	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);
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578

	*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);

579
	if (!ret) {
580
581
		if (context == HOTPLUG &&
		    mem->section_count == sections_per_block)
582
583
584
			ret = register_mem_sect_under_node(mem, nid);
	}

585
	mutex_unlock(&mem_sysfs_mutex);
586
587
588
	return ret;
}

589
590
591
592
593
int remove_memory_block(unsigned long node_id, struct mem_section *section,
		int phys_device)
{
	struct memory_block *mem;

594
	mutex_lock(&mem_sysfs_mutex);
595
	mem = find_memory_block(section);
596
	unregister_mem_sect_under_nodes(mem, __section_nr(section));
597
598
599
600

	mem->section_count--;
	if (mem->section_count == 0) {
		mem_remove_simple_file(mem, phys_index);
601
		mem_remove_simple_file(mem, end_phys_index);
602
603
604
		mem_remove_simple_file(mem, state);
		mem_remove_simple_file(mem, phys_device);
		mem_remove_simple_file(mem, removable);
605
606
607
608
		unregister_memory(mem);
		kfree(mem);
	} else
		kobject_put(&mem->sysdev.kobj);
609

610
	mutex_unlock(&mem_sysfs_mutex);
611
612
613
614
615
616
617
	return 0;
}

/*
 * need an interface for the VM to add new memory regions,
 * but without onlining it.
 */
618
int register_new_memory(int nid, struct mem_section *section)
619
{
620
	return add_memory_section(nid, section, MEM_OFFLINE, HOTPLUG);
621
622
623
624
}

int unregister_memory_section(struct mem_section *section)
{
625
	if (!present_section(section))
626
627
628
629
630
631
632
633
634
635
636
637
		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;
638
	int err;
639
	unsigned long block_sz;
640

641
	memory_sysdev_class.kset.uevent_ops = &memory_uevent_ops;
642
	ret = sysdev_class_register(&memory_sysdev_class);
643
644
	if (ret)
		goto out;
645

646
647
648
	block_sz = get_memory_block_size();
	sections_per_block = block_sz / MIN_MEMORY_BLOCK_SIZE;

649
650
651
652
653
	/*
	 * Create entries for memory sections that were found
	 * during boot and have been initialized
	 */
	for (i = 0; i < NR_MEM_SECTIONS; i++) {
654
		if (!present_section_nr(i))
655
			continue;
656
657
		err = add_memory_section(0, __nr_to_section(i), MEM_ONLINE,
					 BOOT);
658
659
		if (!ret)
			ret = err;
660
661
	}

662
	err = memory_probe_init();
663
664
665
	if (!ret)
		ret = err;
	err = memory_fail_init();
666
667
668
669
670
671
672
	if (!ret)
		ret = err;
	err = block_size_init();
	if (!ret)
		ret = err;
out:
	if (ret)
673
		printk(KERN_ERR "%s() failed: %d\n", __func__, ret);
674
675
	return ret;
}