xref: /linux/block/genhd.c (revision f3d9478b2ce468c3115b02ecae7e975990697f15) 
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 struct subsystem block_subsys;
21 static DEFINE_MUTEX(block_subsys_lock);
22 
23 /*
24  * Can be deleted altogether. Later.
25  *
26  */
27 static struct blk_major_name {
28 	struct blk_major_name *next;
29 	int major;
30 	char name[16];
31 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
32 
33 /* index in the above - for now: assume no multimajor ranges */
34 static inline int major_to_index(int major)
35 {
36 	return major % BLKDEV_MAJOR_HASH_SIZE;
37 }
38 
39 #ifdef CONFIG_PROC_FS
40 
41 void blkdev_show(struct seq_file *f, off_t offset)
42 {
43 	struct blk_major_name *dp;
44 
45 	if (offset < BLKDEV_MAJOR_HASH_SIZE) {
46 		mutex_lock(&block_subsys_lock);
47 		for (dp = major_names[offset]; dp; dp = dp->next)
48 			seq_printf(f, "%3d %s\n", dp->major, dp->name);
49 		mutex_unlock(&block_subsys_lock);
50 	}
51 }
52 
53 #endif /* CONFIG_PROC_FS */
54 
55 int register_blkdev(unsigned int major, const char *name)
56 {
57 	struct blk_major_name **n, *p;
58 	int index, ret = 0;
59 
60 	mutex_lock(&block_subsys_lock);
61 
62 	/* temporary */
63 	if (major == 0) {
64 		for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
65 			if (major_names[index] == NULL)
66 				break;
67 		}
68 
69 		if (index == 0) {
70 			printk("register_blkdev: failed to get major for %s\n",
71 			       name);
72 			ret = -EBUSY;
73 			goto out;
74 		}
75 		major = index;
76 		ret = major;
77 	}
78 
79 	p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
80 	if (p == NULL) {
81 		ret = -ENOMEM;
82 		goto out;
83 	}
84 
85 	p->major = major;
86 	strlcpy(p->name, name, sizeof(p->name));
87 	p->next = NULL;
88 	index = major_to_index(major);
89 
90 	for (n = &major_names[index]; *n; n = &(*n)->next) {
91 		if ((*n)->major == major)
92 			break;
93 	}
94 	if (!*n)
95 		*n = p;
96 	else
97 		ret = -EBUSY;
98 
99 	if (ret < 0) {
100 		printk("register_blkdev: cannot get major %d for %s\n",
101 		       major, name);
102 		kfree(p);
103 	}
104 out:
105 	mutex_unlock(&block_subsys_lock);
106 	return ret;
107 }
108 
109 EXPORT_SYMBOL(register_blkdev);
110 
111 /* todo: make void - error printk here */
112 int unregister_blkdev(unsigned int major, const char *name)
113 {
114 	struct blk_major_name **n;
115 	struct blk_major_name *p = NULL;
116 	int index = major_to_index(major);
117 	int ret = 0;
118 
119 	mutex_lock(&block_subsys_lock);
120 	for (n = &major_names[index]; *n; n = &(*n)->next)
121 		if ((*n)->major == major)
122 			break;
123 	if (!*n || strcmp((*n)->name, name))
124 		ret = -EINVAL;
125 	else {
126 		p = *n;
127 		*n = p->next;
128 	}
129 	mutex_unlock(&block_subsys_lock);
130 	kfree(p);
131 
132 	return ret;
133 }
134 
135 EXPORT_SYMBOL(unregister_blkdev);
136 
137 static struct kobj_map *bdev_map;
138 
139 /*
140  * Register device numbers dev..(dev+range-1)
141  * range must be nonzero
142  * The hash chain is sorted on range, so that subranges can override.
143  */
144 void blk_register_region(dev_t dev, unsigned long range, struct module *module,
145 			 struct kobject *(*probe)(dev_t, int *, void *),
146 			 int (*lock)(dev_t, void *), void *data)
147 {
148 	kobj_map(bdev_map, dev, range, module, probe, lock, data);
149 }
150 
151 EXPORT_SYMBOL(blk_register_region);
152 
153 void blk_unregister_region(dev_t dev, unsigned long range)
154 {
155 	kobj_unmap(bdev_map, dev, range);
156 }
157 
158 EXPORT_SYMBOL(blk_unregister_region);
159 
160 static struct kobject *exact_match(dev_t dev, int *part, void *data)
161 {
162 	struct gendisk *p = data;
163 	return &p->kobj;
164 }
165 
166 static int exact_lock(dev_t dev, void *data)
167 {
168 	struct gendisk *p = data;
169 
170 	if (!get_disk(p))
171 		return -1;
172 	return 0;
173 }
174 
175 /**
176  * add_disk - add partitioning information to kernel list
177  * @disk: per-device partitioning information
178  *
179  * This function registers the partitioning information in @disk
180  * with the kernel.
181  */
182 void add_disk(struct gendisk *disk)
183 {
184 	disk->flags |= GENHD_FL_UP;
185 	blk_register_region(MKDEV(disk->major, disk->first_minor),
186 			    disk->minors, NULL, exact_match, exact_lock, disk);
187 	register_disk(disk);
188 	blk_register_queue(disk);
189 }
190 
191 EXPORT_SYMBOL(add_disk);
192 EXPORT_SYMBOL(del_gendisk);	/* in partitions/check.c */
193 
194 void unlink_gendisk(struct gendisk *disk)
195 {
196 	blk_unregister_queue(disk);
197 	blk_unregister_region(MKDEV(disk->major, disk->first_minor),
198 			      disk->minors);
199 }
200 
201 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
202 
203 /**
204  * get_gendisk - get partitioning information for a given device
205  * @dev: device to get partitioning information for
206  *
207  * This function gets the structure containing partitioning
208  * information for the given device @dev.
209  */
210 struct gendisk *get_gendisk(dev_t dev, int *part)
211 {
212 	struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
213 	return  kobj ? to_disk(kobj) : NULL;
214 }
215 
216 #ifdef CONFIG_PROC_FS
217 /* iterator */
218 static void *part_start(struct seq_file *part, loff_t *pos)
219 {
220 	struct list_head *p;
221 	loff_t l = *pos;
222 
223 	mutex_lock(&block_subsys_lock);
224 	list_for_each(p, &block_subsys.kset.list)
225 		if (!l--)
226 			return list_entry(p, struct gendisk, kobj.entry);
227 	return NULL;
228 }
229 
230 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
231 {
232 	struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
233 	++*pos;
234 	return p==&block_subsys.kset.list ? NULL :
235 		list_entry(p, struct gendisk, kobj.entry);
236 }
237 
238 static void part_stop(struct seq_file *part, void *v)
239 {
240 	mutex_unlock(&block_subsys_lock);
241 }
242 
243 static int show_partition(struct seq_file *part, void *v)
244 {
245 	struct gendisk *sgp = v;
246 	int n;
247 	char buf[BDEVNAME_SIZE];
248 
249 	if (&sgp->kobj.entry == block_subsys.kset.list.next)
250 		seq_puts(part, "major minor  #blocks  name\n\n");
251 
252 	/* Don't show non-partitionable removeable devices or empty devices */
253 	if (!get_capacity(sgp) ||
254 			(sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
255 		return 0;
256 	if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
257 		return 0;
258 
259 	/* show the full disk and all non-0 size partitions of it */
260 	seq_printf(part, "%4d  %4d %10llu %s\n",
261 		sgp->major, sgp->first_minor,
262 		(unsigned long long)get_capacity(sgp) >> 1,
263 		disk_name(sgp, 0, buf));
264 	for (n = 0; n < sgp->minors - 1; n++) {
265 		if (!sgp->part[n])
266 			continue;
267 		if (sgp->part[n]->nr_sects == 0)
268 			continue;
269 		seq_printf(part, "%4d  %4d %10llu %s\n",
270 			sgp->major, n + 1 + sgp->first_minor,
271 			(unsigned long long)sgp->part[n]->nr_sects >> 1 ,
272 			disk_name(sgp, n + 1, buf));
273 	}
274 
275 	return 0;
276 }
277 
278 struct seq_operations partitions_op = {
279 	.start =part_start,
280 	.next =	part_next,
281 	.stop =	part_stop,
282 	.show =	show_partition
283 };
284 #endif
285 
286 
287 extern int blk_dev_init(void);
288 
289 static struct kobject *base_probe(dev_t dev, int *part, void *data)
290 {
291 	if (request_module("block-major-%d-%d", MAJOR(dev), MINOR(dev)) > 0)
292 		/* Make old-style 2.4 aliases work */
293 		request_module("block-major-%d", MAJOR(dev));
294 	return NULL;
295 }
296 
297 static int __init genhd_device_init(void)
298 {
299 	bdev_map = kobj_map_init(base_probe, &block_subsys_lock);
300 	blk_dev_init();
301 	subsystem_register(&block_subsys);
302 	return 0;
303 }
304 
305 subsys_initcall(genhd_device_init);
306 
307 
308 
309 /*
310  * kobject & sysfs bindings for block devices
311  */
312 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
313 			      char *page)
314 {
315 	struct gendisk *disk = to_disk(kobj);
316 	struct disk_attribute *disk_attr =
317 		container_of(attr,struct disk_attribute,attr);
318 	ssize_t ret = -EIO;
319 
320 	if (disk_attr->show)
321 		ret = disk_attr->show(disk,page);
322 	return ret;
323 }
324 
325 static ssize_t disk_attr_store(struct kobject * kobj, struct attribute * attr,
326 			       const char *page, size_t count)
327 {
328 	struct gendisk *disk = to_disk(kobj);
329 	struct disk_attribute *disk_attr =
330 		container_of(attr,struct disk_attribute,attr);
331 	ssize_t ret = 0;
332 
333 	if (disk_attr->store)
334 		ret = disk_attr->store(disk, page, count);
335 	return ret;
336 }
337 
338 static struct sysfs_ops disk_sysfs_ops = {
339 	.show	= &disk_attr_show,
340 	.store	= &disk_attr_store,
341 };
342 
343 static ssize_t disk_uevent_store(struct gendisk * disk,
344 				 const char *buf, size_t count)
345 {
346 	kobject_uevent(&disk->kobj, KOBJ_ADD);
347 	return count;
348 }
349 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
350 {
351 	dev_t base = MKDEV(disk->major, disk->first_minor);
352 	return print_dev_t(page, base);
353 }
354 static ssize_t disk_range_read(struct gendisk * disk, char *page)
355 {
356 	return sprintf(page, "%d\n", disk->minors);
357 }
358 static ssize_t disk_removable_read(struct gendisk * disk, char *page)
359 {
360 	return sprintf(page, "%d\n",
361 		       (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
362 
363 }
364 static ssize_t disk_size_read(struct gendisk * disk, char *page)
365 {
366 	return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
367 }
368 
369 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
370 {
371 	preempt_disable();
372 	disk_round_stats(disk);
373 	preempt_enable();
374 	return sprintf(page,
375 		"%8lu %8lu %8llu %8u "
376 		"%8lu %8lu %8llu %8u "
377 		"%8u %8u %8u"
378 		"\n",
379 		disk_stat_read(disk, ios[READ]),
380 		disk_stat_read(disk, merges[READ]),
381 		(unsigned long long)disk_stat_read(disk, sectors[READ]),
382 		jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
383 		disk_stat_read(disk, ios[WRITE]),
384 		disk_stat_read(disk, merges[WRITE]),
385 		(unsigned long long)disk_stat_read(disk, sectors[WRITE]),
386 		jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
387 		disk->in_flight,
388 		jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
389 		jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
390 }
391 static struct disk_attribute disk_attr_uevent = {
392 	.attr = {.name = "uevent", .mode = S_IWUSR },
393 	.store	= disk_uevent_store
394 };
395 static struct disk_attribute disk_attr_dev = {
396 	.attr = {.name = "dev", .mode = S_IRUGO },
397 	.show	= disk_dev_read
398 };
399 static struct disk_attribute disk_attr_range = {
400 	.attr = {.name = "range", .mode = S_IRUGO },
401 	.show	= disk_range_read
402 };
403 static struct disk_attribute disk_attr_removable = {
404 	.attr = {.name = "removable", .mode = S_IRUGO },
405 	.show	= disk_removable_read
406 };
407 static struct disk_attribute disk_attr_size = {
408 	.attr = {.name = "size", .mode = S_IRUGO },
409 	.show	= disk_size_read
410 };
411 static struct disk_attribute disk_attr_stat = {
412 	.attr = {.name = "stat", .mode = S_IRUGO },
413 	.show	= disk_stats_read
414 };
415 
416 static struct attribute * default_attrs[] = {
417 	&disk_attr_uevent.attr,
418 	&disk_attr_dev.attr,
419 	&disk_attr_range.attr,
420 	&disk_attr_removable.attr,
421 	&disk_attr_size.attr,
422 	&disk_attr_stat.attr,
423 	NULL,
424 };
425 
426 static void disk_release(struct kobject * kobj)
427 {
428 	struct gendisk *disk = to_disk(kobj);
429 	kfree(disk->random);
430 	kfree(disk->part);
431 	free_disk_stats(disk);
432 	kfree(disk);
433 }
434 
435 static struct kobj_type ktype_block = {
436 	.release	= disk_release,
437 	.sysfs_ops	= &disk_sysfs_ops,
438 	.default_attrs	= default_attrs,
439 };
440 
441 extern struct kobj_type ktype_part;
442 
443 static int block_uevent_filter(struct kset *kset, struct kobject *kobj)
444 {
445 	struct kobj_type *ktype = get_ktype(kobj);
446 
447 	return ((ktype == &ktype_block) || (ktype == &ktype_part));
448 }
449 
450 static int block_uevent(struct kset *kset, struct kobject *kobj, char **envp,
451 			 int num_envp, char *buffer, int buffer_size)
452 {
453 	struct kobj_type *ktype = get_ktype(kobj);
454 	struct device *physdev;
455 	struct gendisk *disk;
456 	struct hd_struct *part;
457 	int length = 0;
458 	int i = 0;
459 
460 	if (ktype == &ktype_block) {
461 		disk = container_of(kobj, struct gendisk, kobj);
462 		add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
463 			       &length, "MINOR=%u", disk->first_minor);
464 	} else if (ktype == &ktype_part) {
465 		disk = container_of(kobj->parent, struct gendisk, kobj);
466 		part = container_of(kobj, struct hd_struct, kobj);
467 		add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
468 			       &length, "MINOR=%u",
469 			       disk->first_minor + part->partno);
470 	} else
471 		return 0;
472 
473 	add_uevent_var(envp, num_envp, &i, buffer, buffer_size, &length,
474 		       "MAJOR=%u", disk->major);
475 
476 	/* add physical device, backing this device  */
477 	physdev = disk->driverfs_dev;
478 	if (physdev) {
479 		char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
480 
481 		add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
482 			       &length, "PHYSDEVPATH=%s", path);
483 		kfree(path);
484 
485 		if (physdev->bus)
486 			add_uevent_var(envp, num_envp, &i,
487 				       buffer, buffer_size, &length,
488 				       "PHYSDEVBUS=%s",
489 				       physdev->bus->name);
490 
491 		if (physdev->driver)
492 			add_uevent_var(envp, num_envp, &i,
493 				       buffer, buffer_size, &length,
494 				       "PHYSDEVDRIVER=%s",
495 				       physdev->driver->name);
496 	}
497 
498 	/* terminate, set to next free slot, shrink available space */
499 	envp[i] = NULL;
500 	envp = &envp[i];
501 	num_envp -= i;
502 	buffer = &buffer[length];
503 	buffer_size -= length;
504 
505 	return 0;
506 }
507 
508 static struct kset_uevent_ops block_uevent_ops = {
509 	.filter		= block_uevent_filter,
510 	.uevent		= block_uevent,
511 };
512 
513 decl_subsys(block, &ktype_block, &block_uevent_ops);
514 
515 /*
516  * aggregate disk stat collector.  Uses the same stats that the sysfs
517  * entries do, above, but makes them available through one seq_file.
518  * Watching a few disks may be efficient through sysfs, but watching
519  * all of them will be more efficient through this interface.
520  *
521  * The output looks suspiciously like /proc/partitions with a bunch of
522  * extra fields.
523  */
524 
525 /* iterator */
526 static void *diskstats_start(struct seq_file *part, loff_t *pos)
527 {
528 	loff_t k = *pos;
529 	struct list_head *p;
530 
531 	mutex_lock(&block_subsys_lock);
532 	list_for_each(p, &block_subsys.kset.list)
533 		if (!k--)
534 			return list_entry(p, struct gendisk, kobj.entry);
535 	return NULL;
536 }
537 
538 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
539 {
540 	struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
541 	++*pos;
542 	return p==&block_subsys.kset.list ? NULL :
543 		list_entry(p, struct gendisk, kobj.entry);
544 }
545 
546 static void diskstats_stop(struct seq_file *part, void *v)
547 {
548 	mutex_unlock(&block_subsys_lock);
549 }
550 
551 static int diskstats_show(struct seq_file *s, void *v)
552 {
553 	struct gendisk *gp = v;
554 	char buf[BDEVNAME_SIZE];
555 	int n = 0;
556 
557 	/*
558 	if (&sgp->kobj.entry == block_subsys.kset.list.next)
559 		seq_puts(s,	"major minor name"
560 				"     rio rmerge rsect ruse wio wmerge "
561 				"wsect wuse running use aveq"
562 				"\n\n");
563 	*/
564 
565 	preempt_disable();
566 	disk_round_stats(gp);
567 	preempt_enable();
568 	seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
569 		gp->major, n + gp->first_minor, disk_name(gp, n, buf),
570 		disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
571 		(unsigned long long)disk_stat_read(gp, sectors[0]),
572 		jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
573 		disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
574 		(unsigned long long)disk_stat_read(gp, sectors[1]),
575 		jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
576 		gp->in_flight,
577 		jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
578 		jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
579 
580 	/* now show all non-0 size partitions of it */
581 	for (n = 0; n < gp->minors - 1; n++) {
582 		struct hd_struct *hd = gp->part[n];
583 
584 		if (hd && hd->nr_sects)
585 			seq_printf(s, "%4d %4d %s %u %u %u %u\n",
586 				gp->major, n + gp->first_minor + 1,
587 				disk_name(gp, n + 1, buf),
588 				hd->ios[0], hd->sectors[0],
589 				hd->ios[1], hd->sectors[1]);
590 	}
591 
592 	return 0;
593 }
594 
595 struct seq_operations diskstats_op = {
596 	.start	= diskstats_start,
597 	.next	= diskstats_next,
598 	.stop	= diskstats_stop,
599 	.show	= diskstats_show
600 };
601 
602 struct gendisk *alloc_disk(int minors)
603 {
604 	return alloc_disk_node(minors, -1);
605 }
606 
607 struct gendisk *alloc_disk_node(int minors, int node_id)
608 {
609 	struct gendisk *disk;
610 
611 	disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
612 	if (disk) {
613 		memset(disk, 0, sizeof(struct gendisk));
614 		if (!init_disk_stats(disk)) {
615 			kfree(disk);
616 			return NULL;
617 		}
618 		if (minors > 1) {
619 			int size = (minors - 1) * sizeof(struct hd_struct *);
620 			disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
621 			if (!disk->part) {
622 				kfree(disk);
623 				return NULL;
624 			}
625 			memset(disk->part, 0, size);
626 		}
627 		disk->minors = minors;
628 		kobj_set_kset_s(disk,block_subsys);
629 		kobject_init(&disk->kobj);
630 		rand_initialize_disk(disk);
631 	}
632 	return disk;
633 }
634 
635 EXPORT_SYMBOL(alloc_disk);
636 EXPORT_SYMBOL(alloc_disk_node);
637 
638 struct kobject *get_disk(struct gendisk *disk)
639 {
640 	struct module *owner;
641 	struct kobject *kobj;
642 
643 	if (!disk->fops)
644 		return NULL;
645 	owner = disk->fops->owner;
646 	if (owner && !try_module_get(owner))
647 		return NULL;
648 	kobj = kobject_get(&disk->kobj);
649 	if (kobj == NULL) {
650 		module_put(owner);
651 		return NULL;
652 	}
653 	return kobj;
654 
655 }
656 
657 EXPORT_SYMBOL(get_disk);
658 
659 void put_disk(struct gendisk *disk)
660 {
661 	if (disk)
662 		kobject_put(&disk->kobj);
663 }
664 
665 EXPORT_SYMBOL(put_disk);
666 
667 void set_device_ro(struct block_device *bdev, int flag)
668 {
669 	if (bdev->bd_contains != bdev)
670 		bdev->bd_part->policy = flag;
671 	else
672 		bdev->bd_disk->policy = flag;
673 }
674 
675 EXPORT_SYMBOL(set_device_ro);
676 
677 void set_disk_ro(struct gendisk *disk, int flag)
678 {
679 	int i;
680 	disk->policy = flag;
681 	for (i = 0; i < disk->minors - 1; i++)
682 		if (disk->part[i]) disk->part[i]->policy = flag;
683 }
684 
685 EXPORT_SYMBOL(set_disk_ro);
686 
687 int bdev_read_only(struct block_device *bdev)
688 {
689 	if (!bdev)
690 		return 0;
691 	else if (bdev->bd_contains != bdev)
692 		return bdev->bd_part->policy;
693 	else
694 		return bdev->bd_disk->policy;
695 }
696 
697 EXPORT_SYMBOL(bdev_read_only);
698 
699 int invalidate_partition(struct gendisk *disk, int index)
700 {
701 	int res = 0;
702 	struct block_device *bdev = bdget_disk(disk, index);
703 	if (bdev) {
704 		fsync_bdev(bdev);
705 		res = __invalidate_device(bdev);
706 		bdput(bdev);
707 	}
708 	return res;
709 }
710 
711 EXPORT_SYMBOL(invalidate_partition);
712