xref: /linux/block/genhd.c (revision ccea15f45eb0ab12d658f88b5d4be005cb2bb1a7) 
1 /*
2  *  gendisk handling
3  */
4 
5 #include <linux/config.h>
6 #include <linux/module.h>
7 #include <linux/fs.h>
8 #include <linux/genhd.h>
9 #include <linux/kernel.h>
10 #include <linux/blkdev.h>
11 #include <linux/init.h>
12 #include <linux/spinlock.h>
13 #include <linux/seq_file.h>
14 #include <linux/slab.h>
15 #include <linux/kmod.h>
16 #include <linux/kobj_map.h>
17 #include <linux/buffer_head.h>
18 #include <linux/mutex.h>
19 
20 static struct subsystem block_subsys;
21 
22 static DEFINE_MUTEX(block_subsys_lock);
23 
24 /*
25  * Can be deleted altogether. Later.
26  *
27  */
28 static struct blk_major_name {
29 	struct blk_major_name *next;
30 	int major;
31 	char name[16];
32 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
33 
34 /* index in the above - for now: assume no multimajor ranges */
35 static inline int major_to_index(int major)
36 {
37 	return major % BLKDEV_MAJOR_HASH_SIZE;
38 }
39 
40 #ifdef CONFIG_PROC_FS
41 
42 void blkdev_show(struct seq_file *f, off_t offset)
43 {
44 	struct blk_major_name *dp;
45 
46 	if (offset < BLKDEV_MAJOR_HASH_SIZE) {
47 		mutex_lock(&block_subsys_lock);
48 		for (dp = major_names[offset]; dp; dp = dp->next)
49 			seq_printf(f, "%3d %s\n", dp->major, dp->name);
50 		mutex_unlock(&block_subsys_lock);
51 	}
52 }
53 
54 #endif /* CONFIG_PROC_FS */
55 
56 int register_blkdev(unsigned int major, const char *name)
57 {
58 	struct blk_major_name **n, *p;
59 	int index, ret = 0;
60 
61 	mutex_lock(&block_subsys_lock);
62 
63 	/* temporary */
64 	if (major == 0) {
65 		for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
66 			if (major_names[index] == NULL)
67 				break;
68 		}
69 
70 		if (index == 0) {
71 			printk("register_blkdev: failed to get major for %s\n",
72 			       name);
73 			ret = -EBUSY;
74 			goto out;
75 		}
76 		major = index;
77 		ret = major;
78 	}
79 
80 	p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
81 	if (p == NULL) {
82 		ret = -ENOMEM;
83 		goto out;
84 	}
85 
86 	p->major = major;
87 	strlcpy(p->name, name, sizeof(p->name));
88 	p->next = NULL;
89 	index = major_to_index(major);
90 
91 	for (n = &major_names[index]; *n; n = &(*n)->next) {
92 		if ((*n)->major == major)
93 			break;
94 	}
95 	if (!*n)
96 		*n = p;
97 	else
98 		ret = -EBUSY;
99 
100 	if (ret < 0) {
101 		printk("register_blkdev: cannot get major %d for %s\n",
102 		       major, name);
103 		kfree(p);
104 	}
105 out:
106 	mutex_unlock(&block_subsys_lock);
107 	return ret;
108 }
109 
110 EXPORT_SYMBOL(register_blkdev);
111 
112 /* todo: make void - error printk here */
113 int unregister_blkdev(unsigned int major, const char *name)
114 {
115 	struct blk_major_name **n;
116 	struct blk_major_name *p = NULL;
117 	int index = major_to_index(major);
118 	int ret = 0;
119 
120 	mutex_lock(&block_subsys_lock);
121 	for (n = &major_names[index]; *n; n = &(*n)->next)
122 		if ((*n)->major == major)
123 			break;
124 	if (!*n || strcmp((*n)->name, name))
125 		ret = -EINVAL;
126 	else {
127 		p = *n;
128 		*n = p->next;
129 	}
130 	mutex_unlock(&block_subsys_lock);
131 	kfree(p);
132 
133 	return ret;
134 }
135 
136 EXPORT_SYMBOL(unregister_blkdev);
137 
138 static struct kobj_map *bdev_map;
139 
140 /*
141  * Register device numbers dev..(dev+range-1)
142  * range must be nonzero
143  * The hash chain is sorted on range, so that subranges can override.
144  */
145 void blk_register_region(dev_t dev, unsigned long range, struct module *module,
146 			 struct kobject *(*probe)(dev_t, int *, void *),
147 			 int (*lock)(dev_t, void *), void *data)
148 {
149 	kobj_map(bdev_map, dev, range, module, probe, lock, data);
150 }
151 
152 EXPORT_SYMBOL(blk_register_region);
153 
154 void blk_unregister_region(dev_t dev, unsigned long range)
155 {
156 	kobj_unmap(bdev_map, dev, range);
157 }
158 
159 EXPORT_SYMBOL(blk_unregister_region);
160 
161 static struct kobject *exact_match(dev_t dev, int *part, void *data)
162 {
163 	struct gendisk *p = data;
164 	return &p->kobj;
165 }
166 
167 static int exact_lock(dev_t dev, void *data)
168 {
169 	struct gendisk *p = data;
170 
171 	if (!get_disk(p))
172 		return -1;
173 	return 0;
174 }
175 
176 /**
177  * add_disk - add partitioning information to kernel list
178  * @disk: per-device partitioning information
179  *
180  * This function registers the partitioning information in @disk
181  * with the kernel.
182  */
183 void add_disk(struct gendisk *disk)
184 {
185 	disk->flags |= GENHD_FL_UP;
186 	blk_register_region(MKDEV(disk->major, disk->first_minor),
187 			    disk->minors, NULL, exact_match, exact_lock, disk);
188 	register_disk(disk);
189 	blk_register_queue(disk);
190 }
191 
192 EXPORT_SYMBOL(add_disk);
193 EXPORT_SYMBOL(del_gendisk);	/* in partitions/check.c */
194 
195 void unlink_gendisk(struct gendisk *disk)
196 {
197 	blk_unregister_queue(disk);
198 	blk_unregister_region(MKDEV(disk->major, disk->first_minor),
199 			      disk->minors);
200 }
201 
202 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
203 
204 /**
205  * get_gendisk - get partitioning information for a given device
206  * @dev: device to get partitioning information for
207  *
208  * This function gets the structure containing partitioning
209  * information for the given device @dev.
210  */
211 struct gendisk *get_gendisk(dev_t dev, int *part)
212 {
213 	struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
214 	return  kobj ? to_disk(kobj) : NULL;
215 }
216 
217 #ifdef CONFIG_PROC_FS
218 /* iterator */
219 static void *part_start(struct seq_file *part, loff_t *pos)
220 {
221 	struct list_head *p;
222 	loff_t l = *pos;
223 
224 	mutex_lock(&block_subsys_lock);
225 	list_for_each(p, &block_subsys.kset.list)
226 		if (!l--)
227 			return list_entry(p, struct gendisk, kobj.entry);
228 	return NULL;
229 }
230 
231 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
232 {
233 	struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
234 	++*pos;
235 	return p==&block_subsys.kset.list ? NULL :
236 		list_entry(p, struct gendisk, kobj.entry);
237 }
238 
239 static void part_stop(struct seq_file *part, void *v)
240 {
241 	mutex_unlock(&block_subsys_lock);
242 }
243 
244 static int show_partition(struct seq_file *part, void *v)
245 {
246 	struct gendisk *sgp = v;
247 	int n;
248 	char buf[BDEVNAME_SIZE];
249 
250 	if (&sgp->kobj.entry == block_subsys.kset.list.next)
251 		seq_puts(part, "major minor  #blocks  name\n\n");
252 
253 	/* Don't show non-partitionable removeable devices or empty devices */
254 	if (!get_capacity(sgp) ||
255 			(sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
256 		return 0;
257 	if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
258 		return 0;
259 
260 	/* show the full disk and all non-0 size partitions of it */
261 	seq_printf(part, "%4d  %4d %10llu %s\n",
262 		sgp->major, sgp->first_minor,
263 		(unsigned long long)get_capacity(sgp) >> 1,
264 		disk_name(sgp, 0, buf));
265 	for (n = 0; n < sgp->minors - 1; n++) {
266 		if (!sgp->part[n])
267 			continue;
268 		if (sgp->part[n]->nr_sects == 0)
269 			continue;
270 		seq_printf(part, "%4d  %4d %10llu %s\n",
271 			sgp->major, n + 1 + sgp->first_minor,
272 			(unsigned long long)sgp->part[n]->nr_sects >> 1 ,
273 			disk_name(sgp, n + 1, buf));
274 	}
275 
276 	return 0;
277 }
278 
279 struct seq_operations partitions_op = {
280 	.start =part_start,
281 	.next =	part_next,
282 	.stop =	part_stop,
283 	.show =	show_partition
284 };
285 #endif
286 
287 
288 extern int blk_dev_init(void);
289 
290 static struct kobject *base_probe(dev_t dev, int *part, void *data)
291 {
292 	if (request_module("block-major-%d-%d", MAJOR(dev), MINOR(dev)) > 0)
293 		/* Make old-style 2.4 aliases work */
294 		request_module("block-major-%d", MAJOR(dev));
295 	return NULL;
296 }
297 
298 static int __init genhd_device_init(void)
299 {
300 	bdev_map = kobj_map_init(base_probe, &block_subsys_lock);
301 	blk_dev_init();
302 	subsystem_register(&block_subsys);
303 	return 0;
304 }
305 
306 subsys_initcall(genhd_device_init);
307 
308 
309 
310 /*
311  * kobject & sysfs bindings for block devices
312  */
313 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
314 			      char *page)
315 {
316 	struct gendisk *disk = to_disk(kobj);
317 	struct disk_attribute *disk_attr =
318 		container_of(attr,struct disk_attribute,attr);
319 	ssize_t ret = -EIO;
320 
321 	if (disk_attr->show)
322 		ret = disk_attr->show(disk,page);
323 	return ret;
324 }
325 
326 static ssize_t disk_attr_store(struct kobject * kobj, struct attribute * attr,
327 			       const char *page, size_t count)
328 {
329 	struct gendisk *disk = to_disk(kobj);
330 	struct disk_attribute *disk_attr =
331 		container_of(attr,struct disk_attribute,attr);
332 	ssize_t ret = 0;
333 
334 	if (disk_attr->store)
335 		ret = disk_attr->store(disk, page, count);
336 	return ret;
337 }
338 
339 static struct sysfs_ops disk_sysfs_ops = {
340 	.show	= &disk_attr_show,
341 	.store	= &disk_attr_store,
342 };
343 
344 static ssize_t disk_uevent_store(struct gendisk * disk,
345 				 const char *buf, size_t count)
346 {
347 	kobject_uevent(&disk->kobj, KOBJ_ADD);
348 	return count;
349 }
350 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
351 {
352 	dev_t base = MKDEV(disk->major, disk->first_minor);
353 	return print_dev_t(page, base);
354 }
355 static ssize_t disk_range_read(struct gendisk * disk, char *page)
356 {
357 	return sprintf(page, "%d\n", disk->minors);
358 }
359 static ssize_t disk_removable_read(struct gendisk * disk, char *page)
360 {
361 	return sprintf(page, "%d\n",
362 		       (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
363 
364 }
365 static ssize_t disk_size_read(struct gendisk * disk, char *page)
366 {
367 	return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
368 }
369 
370 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
371 {
372 	preempt_disable();
373 	disk_round_stats(disk);
374 	preempt_enable();
375 	return sprintf(page,
376 		"%8lu %8lu %8llu %8u "
377 		"%8lu %8lu %8llu %8u "
378 		"%8u %8u %8u"
379 		"\n",
380 		disk_stat_read(disk, ios[READ]),
381 		disk_stat_read(disk, merges[READ]),
382 		(unsigned long long)disk_stat_read(disk, sectors[READ]),
383 		jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
384 		disk_stat_read(disk, ios[WRITE]),
385 		disk_stat_read(disk, merges[WRITE]),
386 		(unsigned long long)disk_stat_read(disk, sectors[WRITE]),
387 		jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
388 		disk->in_flight,
389 		jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
390 		jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
391 }
392 static struct disk_attribute disk_attr_uevent = {
393 	.attr = {.name = "uevent", .mode = S_IWUSR },
394 	.store	= disk_uevent_store
395 };
396 static struct disk_attribute disk_attr_dev = {
397 	.attr = {.name = "dev", .mode = S_IRUGO },
398 	.show	= disk_dev_read
399 };
400 static struct disk_attribute disk_attr_range = {
401 	.attr = {.name = "range", .mode = S_IRUGO },
402 	.show	= disk_range_read
403 };
404 static struct disk_attribute disk_attr_removable = {
405 	.attr = {.name = "removable", .mode = S_IRUGO },
406 	.show	= disk_removable_read
407 };
408 static struct disk_attribute disk_attr_size = {
409 	.attr = {.name = "size", .mode = S_IRUGO },
410 	.show	= disk_size_read
411 };
412 static struct disk_attribute disk_attr_stat = {
413 	.attr = {.name = "stat", .mode = S_IRUGO },
414 	.show	= disk_stats_read
415 };
416 
417 static struct attribute * default_attrs[] = {
418 	&disk_attr_uevent.attr,
419 	&disk_attr_dev.attr,
420 	&disk_attr_range.attr,
421 	&disk_attr_removable.attr,
422 	&disk_attr_size.attr,
423 	&disk_attr_stat.attr,
424 	NULL,
425 };
426 
427 static void disk_release(struct kobject * kobj)
428 {
429 	struct gendisk *disk = to_disk(kobj);
430 	kfree(disk->random);
431 	kfree(disk->part);
432 	free_disk_stats(disk);
433 	kfree(disk);
434 }
435 
436 static struct kobj_type ktype_block = {
437 	.release	= disk_release,
438 	.sysfs_ops	= &disk_sysfs_ops,
439 	.default_attrs	= default_attrs,
440 };
441 
442 extern struct kobj_type ktype_part;
443 
444 static int block_uevent_filter(struct kset *kset, struct kobject *kobj)
445 {
446 	struct kobj_type *ktype = get_ktype(kobj);
447 
448 	return ((ktype == &ktype_block) || (ktype == &ktype_part));
449 }
450 
451 static int block_uevent(struct kset *kset, struct kobject *kobj, char **envp,
452 			 int num_envp, char *buffer, int buffer_size)
453 {
454 	struct kobj_type *ktype = get_ktype(kobj);
455 	struct device *physdev;
456 	struct gendisk *disk;
457 	struct hd_struct *part;
458 	int length = 0;
459 	int i = 0;
460 
461 	if (ktype == &ktype_block) {
462 		disk = container_of(kobj, struct gendisk, kobj);
463 		add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
464 			       &length, "MINOR=%u", disk->first_minor);
465 	} else if (ktype == &ktype_part) {
466 		disk = container_of(kobj->parent, struct gendisk, kobj);
467 		part = container_of(kobj, struct hd_struct, kobj);
468 		add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
469 			       &length, "MINOR=%u",
470 			       disk->first_minor + part->partno);
471 	} else
472 		return 0;
473 
474 	add_uevent_var(envp, num_envp, &i, buffer, buffer_size, &length,
475 		       "MAJOR=%u", disk->major);
476 
477 	/* add physical device, backing this device  */
478 	physdev = disk->driverfs_dev;
479 	if (physdev) {
480 		char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
481 
482 		add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
483 			       &length, "PHYSDEVPATH=%s", path);
484 		kfree(path);
485 
486 		if (physdev->bus)
487 			add_uevent_var(envp, num_envp, &i,
488 				       buffer, buffer_size, &length,
489 				       "PHYSDEVBUS=%s",
490 				       physdev->bus->name);
491 
492 		if (physdev->driver)
493 			add_uevent_var(envp, num_envp, &i,
494 				       buffer, buffer_size, &length,
495 				       "PHYSDEVDRIVER=%s",
496 				       physdev->driver->name);
497 	}
498 
499 	/* terminate, set to next free slot, shrink available space */
500 	envp[i] = NULL;
501 	envp = &envp[i];
502 	num_envp -= i;
503 	buffer = &buffer[length];
504 	buffer_size -= length;
505 
506 	return 0;
507 }
508 
509 static struct kset_uevent_ops block_uevent_ops = {
510 	.filter		= block_uevent_filter,
511 	.uevent		= block_uevent,
512 };
513 
514 /* declare block_subsys. */
515 static decl_subsys(block, &ktype_block, &block_uevent_ops);
516 
517 
518 /*
519  * aggregate disk stat collector.  Uses the same stats that the sysfs
520  * entries do, above, but makes them available through one seq_file.
521  * Watching a few disks may be efficient through sysfs, but watching
522  * all of them will be more efficient through this interface.
523  *
524  * The output looks suspiciously like /proc/partitions with a bunch of
525  * extra fields.
526  */
527 
528 /* iterator */
529 static void *diskstats_start(struct seq_file *part, loff_t *pos)
530 {
531 	loff_t k = *pos;
532 	struct list_head *p;
533 
534 	mutex_lock(&block_subsys_lock);
535 	list_for_each(p, &block_subsys.kset.list)
536 		if (!k--)
537 			return list_entry(p, struct gendisk, kobj.entry);
538 	return NULL;
539 }
540 
541 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
542 {
543 	struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
544 	++*pos;
545 	return p==&block_subsys.kset.list ? NULL :
546 		list_entry(p, struct gendisk, kobj.entry);
547 }
548 
549 static void diskstats_stop(struct seq_file *part, void *v)
550 {
551 	mutex_unlock(&block_subsys_lock);
552 }
553 
554 static int diskstats_show(struct seq_file *s, void *v)
555 {
556 	struct gendisk *gp = v;
557 	char buf[BDEVNAME_SIZE];
558 	int n = 0;
559 
560 	/*
561 	if (&sgp->kobj.entry == block_subsys.kset.list.next)
562 		seq_puts(s,	"major minor name"
563 				"     rio rmerge rsect ruse wio wmerge "
564 				"wsect wuse running use aveq"
565 				"\n\n");
566 	*/
567 
568 	preempt_disable();
569 	disk_round_stats(gp);
570 	preempt_enable();
571 	seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
572 		gp->major, n + gp->first_minor, disk_name(gp, n, buf),
573 		disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
574 		(unsigned long long)disk_stat_read(gp, sectors[0]),
575 		jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
576 		disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
577 		(unsigned long long)disk_stat_read(gp, sectors[1]),
578 		jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
579 		gp->in_flight,
580 		jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
581 		jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
582 
583 	/* now show all non-0 size partitions of it */
584 	for (n = 0; n < gp->minors - 1; n++) {
585 		struct hd_struct *hd = gp->part[n];
586 
587 		if (hd && hd->nr_sects)
588 			seq_printf(s, "%4d %4d %s %u %u %u %u\n",
589 				gp->major, n + gp->first_minor + 1,
590 				disk_name(gp, n + 1, buf),
591 				hd->ios[0], hd->sectors[0],
592 				hd->ios[1], hd->sectors[1]);
593 	}
594 
595 	return 0;
596 }
597 
598 struct seq_operations diskstats_op = {
599 	.start	= diskstats_start,
600 	.next	= diskstats_next,
601 	.stop	= diskstats_stop,
602 	.show	= diskstats_show
603 };
604 
605 struct gendisk *alloc_disk(int minors)
606 {
607 	return alloc_disk_node(minors, -1);
608 }
609 
610 struct gendisk *alloc_disk_node(int minors, int node_id)
611 {
612 	struct gendisk *disk;
613 
614 	disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
615 	if (disk) {
616 		memset(disk, 0, sizeof(struct gendisk));
617 		if (!init_disk_stats(disk)) {
618 			kfree(disk);
619 			return NULL;
620 		}
621 		if (minors > 1) {
622 			int size = (minors - 1) * sizeof(struct hd_struct *);
623 			disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
624 			if (!disk->part) {
625 				kfree(disk);
626 				return NULL;
627 			}
628 			memset(disk->part, 0, size);
629 		}
630 		disk->minors = minors;
631 		kobj_set_kset_s(disk,block_subsys);
632 		kobject_init(&disk->kobj);
633 		rand_initialize_disk(disk);
634 	}
635 	return disk;
636 }
637 
638 EXPORT_SYMBOL(alloc_disk);
639 EXPORT_SYMBOL(alloc_disk_node);
640 
641 struct kobject *get_disk(struct gendisk *disk)
642 {
643 	struct module *owner;
644 	struct kobject *kobj;
645 
646 	if (!disk->fops)
647 		return NULL;
648 	owner = disk->fops->owner;
649 	if (owner && !try_module_get(owner))
650 		return NULL;
651 	kobj = kobject_get(&disk->kobj);
652 	if (kobj == NULL) {
653 		module_put(owner);
654 		return NULL;
655 	}
656 	return kobj;
657 
658 }
659 
660 EXPORT_SYMBOL(get_disk);
661 
662 void put_disk(struct gendisk *disk)
663 {
664 	if (disk)
665 		kobject_put(&disk->kobj);
666 }
667 
668 EXPORT_SYMBOL(put_disk);
669 
670 void set_device_ro(struct block_device *bdev, int flag)
671 {
672 	if (bdev->bd_contains != bdev)
673 		bdev->bd_part->policy = flag;
674 	else
675 		bdev->bd_disk->policy = flag;
676 }
677 
678 EXPORT_SYMBOL(set_device_ro);
679 
680 void set_disk_ro(struct gendisk *disk, int flag)
681 {
682 	int i;
683 	disk->policy = flag;
684 	for (i = 0; i < disk->minors - 1; i++)
685 		if (disk->part[i]) disk->part[i]->policy = flag;
686 }
687 
688 EXPORT_SYMBOL(set_disk_ro);
689 
690 int bdev_read_only(struct block_device *bdev)
691 {
692 	if (!bdev)
693 		return 0;
694 	else if (bdev->bd_contains != bdev)
695 		return bdev->bd_part->policy;
696 	else
697 		return bdev->bd_disk->policy;
698 }
699 
700 EXPORT_SYMBOL(bdev_read_only);
701 
702 int invalidate_partition(struct gendisk *disk, int index)
703 {
704 	int res = 0;
705 	struct block_device *bdev = bdget_disk(disk, index);
706 	if (bdev) {
707 		fsync_bdev(bdev);
708 		res = __invalidate_device(bdev);
709 		bdput(bdev);
710 	}
711 	return res;
712 }
713 
714 EXPORT_SYMBOL(invalidate_partition);
715