Merge branch 'for-upstream/acpi-test' of git://repo.or.cz/linux-2.6/linux-acpi-2...
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / block / genhd.c
blob9074f384b0970535c9048432065ef043ff231923
1 /*
2 * gendisk handling
3 */
5 #include <linux/module.h>
6 #include <linux/fs.h>
7 #include <linux/genhd.h>
8 #include <linux/kdev_t.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>
20 #include "blk.h"
22 static DEFINE_MUTEX(block_class_lock);
23 #ifndef CONFIG_SYSFS_DEPRECATED
24 struct kobject *block_depr;
25 #endif
27 static struct device_type disk_type;
30 * Can be deleted altogether. Later.
33 static struct blk_major_name {
34 struct blk_major_name *next;
35 int major;
36 char name[16];
37 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
39 /* index in the above - for now: assume no multimajor ranges */
40 static inline int major_to_index(int major)
42 return major % BLKDEV_MAJOR_HASH_SIZE;
45 #ifdef CONFIG_PROC_FS
46 void blkdev_show(struct seq_file *f, off_t offset)
48 struct blk_major_name *dp;
50 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
51 mutex_lock(&block_class_lock);
52 for (dp = major_names[offset]; dp; dp = dp->next)
53 seq_printf(f, "%3d %s\n", dp->major, dp->name);
54 mutex_unlock(&block_class_lock);
57 #endif /* CONFIG_PROC_FS */
59 int register_blkdev(unsigned int major, const char *name)
61 struct blk_major_name **n, *p;
62 int index, ret = 0;
64 mutex_lock(&block_class_lock);
66 /* temporary */
67 if (major == 0) {
68 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
69 if (major_names[index] == NULL)
70 break;
73 if (index == 0) {
74 printk("register_blkdev: failed to get major for %s\n",
75 name);
76 ret = -EBUSY;
77 goto out;
79 major = index;
80 ret = major;
83 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
84 if (p == NULL) {
85 ret = -ENOMEM;
86 goto out;
89 p->major = major;
90 strlcpy(p->name, name, sizeof(p->name));
91 p->next = NULL;
92 index = major_to_index(major);
94 for (n = &major_names[index]; *n; n = &(*n)->next) {
95 if ((*n)->major == major)
96 break;
98 if (!*n)
99 *n = p;
100 else
101 ret = -EBUSY;
103 if (ret < 0) {
104 printk("register_blkdev: cannot get major %d for %s\n",
105 major, name);
106 kfree(p);
108 out:
109 mutex_unlock(&block_class_lock);
110 return ret;
113 EXPORT_SYMBOL(register_blkdev);
115 void unregister_blkdev(unsigned int major, const char *name)
117 struct blk_major_name **n;
118 struct blk_major_name *p = NULL;
119 int index = major_to_index(major);
121 mutex_lock(&block_class_lock);
122 for (n = &major_names[index]; *n; n = &(*n)->next)
123 if ((*n)->major == major)
124 break;
125 if (!*n || strcmp((*n)->name, name)) {
126 WARN_ON(1);
127 } else {
128 p = *n;
129 *n = p->next;
131 mutex_unlock(&block_class_lock);
132 kfree(p);
135 EXPORT_SYMBOL(unregister_blkdev);
137 static struct kobj_map *bdev_map;
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.
144 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
145 struct kobject *(*probe)(dev_t, int *, void *),
146 int (*lock)(dev_t, void *), void *data)
148 kobj_map(bdev_map, devt, range, module, probe, lock, data);
151 EXPORT_SYMBOL(blk_register_region);
153 void blk_unregister_region(dev_t devt, unsigned long range)
155 kobj_unmap(bdev_map, devt, range);
158 EXPORT_SYMBOL(blk_unregister_region);
160 static struct kobject *exact_match(dev_t devt, int *part, void *data)
162 struct gendisk *p = data;
164 return &p->dev.kobj;
167 static int exact_lock(dev_t devt, void *data)
169 struct gendisk *p = data;
171 if (!get_disk(p))
172 return -1;
173 return 0;
177 * add_disk - add partitioning information to kernel list
178 * @disk: per-device partitioning information
180 * This function registers the partitioning information in @disk
181 * with the kernel.
183 void add_disk(struct gendisk *disk)
185 struct backing_dev_info *bdi;
187 disk->flags |= GENHD_FL_UP;
188 blk_register_region(MKDEV(disk->major, disk->first_minor),
189 disk->minors, NULL, exact_match, exact_lock, disk);
190 register_disk(disk);
191 blk_register_queue(disk);
192 blk_register_filter(disk);
194 bdi = &disk->queue->backing_dev_info;
195 bdi_register_dev(bdi, MKDEV(disk->major, disk->first_minor));
196 sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi");
199 EXPORT_SYMBOL(add_disk);
200 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
202 void unlink_gendisk(struct gendisk *disk)
204 blk_unregister_filter(disk);
205 sysfs_remove_link(&disk->dev.kobj, "bdi");
206 bdi_unregister(&disk->queue->backing_dev_info);
207 blk_unregister_queue(disk);
208 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
209 disk->minors);
213 * get_gendisk - get partitioning information for a given device
214 * @dev: device to get partitioning information for
216 * This function gets the structure containing partitioning
217 * information for the given device @dev.
219 struct gendisk *get_gendisk(dev_t devt, int *part)
221 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
222 struct device *dev = kobj_to_dev(kobj);
224 return kobj ? dev_to_disk(dev) : NULL;
228 * print a full list of all partitions - intended for places where the root
229 * filesystem can't be mounted and thus to give the victim some idea of what
230 * went wrong
232 void __init printk_all_partitions(void)
234 struct device *dev;
235 struct gendisk *sgp;
236 char buf[BDEVNAME_SIZE];
237 int n;
239 mutex_lock(&block_class_lock);
240 /* For each block device... */
241 list_for_each_entry(dev, &block_class.devices, node) {
242 if (dev->type != &disk_type)
243 continue;
244 sgp = dev_to_disk(dev);
246 * Don't show empty devices or things that have been surpressed
248 if (get_capacity(sgp) == 0 ||
249 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
250 continue;
253 * Note, unlike /proc/partitions, I am showing the numbers in
254 * hex - the same format as the root= option takes.
256 printk("%02x%02x %10llu %s",
257 sgp->major, sgp->first_minor,
258 (unsigned long long)get_capacity(sgp) >> 1,
259 disk_name(sgp, 0, buf));
260 if (sgp->driverfs_dev != NULL &&
261 sgp->driverfs_dev->driver != NULL)
262 printk(" driver: %s\n",
263 sgp->driverfs_dev->driver->name);
264 else
265 printk(" (driver?)\n");
267 /* now show the partitions */
268 for (n = 0; n < sgp->minors - 1; ++n) {
269 if (sgp->part[n] == NULL)
270 continue;
271 if (sgp->part[n]->nr_sects == 0)
272 continue;
273 printk(" %02x%02x %10llu %s\n",
274 sgp->major, n + 1 + sgp->first_minor,
275 (unsigned long long)sgp->part[n]->nr_sects >> 1,
276 disk_name(sgp, n + 1, buf));
280 mutex_unlock(&block_class_lock);
283 #ifdef CONFIG_PROC_FS
284 /* iterator */
285 static void *part_start(struct seq_file *part, loff_t *pos)
287 loff_t k = *pos;
288 struct device *dev;
290 mutex_lock(&block_class_lock);
291 list_for_each_entry(dev, &block_class.devices, node) {
292 if (dev->type != &disk_type)
293 continue;
294 if (!k--)
295 return dev_to_disk(dev);
297 return NULL;
300 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
302 struct gendisk *gp = v;
303 struct device *dev;
304 ++*pos;
305 list_for_each_entry(dev, &gp->dev.node, node) {
306 if (&dev->node == &block_class.devices)
307 return NULL;
308 if (dev->type == &disk_type)
309 return dev_to_disk(dev);
311 return NULL;
314 static void part_stop(struct seq_file *part, void *v)
316 mutex_unlock(&block_class_lock);
319 static int show_partition(struct seq_file *part, void *v)
321 struct gendisk *sgp = v;
322 int n;
323 char buf[BDEVNAME_SIZE];
325 if (&sgp->dev.node == block_class.devices.next)
326 seq_puts(part, "major minor #blocks name\n\n");
328 /* Don't show non-partitionable removeable devices or empty devices */
329 if (!get_capacity(sgp) ||
330 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
331 return 0;
332 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
333 return 0;
335 /* show the full disk and all non-0 size partitions of it */
336 seq_printf(part, "%4d %4d %10llu %s\n",
337 sgp->major, sgp->first_minor,
338 (unsigned long long)get_capacity(sgp) >> 1,
339 disk_name(sgp, 0, buf));
340 for (n = 0; n < sgp->minors - 1; n++) {
341 if (!sgp->part[n])
342 continue;
343 if (sgp->part[n]->nr_sects == 0)
344 continue;
345 seq_printf(part, "%4d %4d %10llu %s\n",
346 sgp->major, n + 1 + sgp->first_minor,
347 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
348 disk_name(sgp, n + 1, buf));
351 return 0;
354 const struct seq_operations partitions_op = {
355 .start = part_start,
356 .next = part_next,
357 .stop = part_stop,
358 .show = show_partition
360 #endif
363 static struct kobject *base_probe(dev_t devt, int *part, void *data)
365 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
366 /* Make old-style 2.4 aliases work */
367 request_module("block-major-%d", MAJOR(devt));
368 return NULL;
371 static int __init genhd_device_init(void)
373 int error = class_register(&block_class);
374 if (unlikely(error))
375 return error;
376 bdev_map = kobj_map_init(base_probe, &block_class_lock);
377 blk_dev_init();
379 #ifndef CONFIG_SYSFS_DEPRECATED
380 /* create top-level block dir */
381 block_depr = kobject_create_and_add("block", NULL);
382 #endif
383 return 0;
386 subsys_initcall(genhd_device_init);
388 static ssize_t disk_range_show(struct device *dev,
389 struct device_attribute *attr, char *buf)
391 struct gendisk *disk = dev_to_disk(dev);
393 return sprintf(buf, "%d\n", disk->minors);
396 static ssize_t disk_removable_show(struct device *dev,
397 struct device_attribute *attr, char *buf)
399 struct gendisk *disk = dev_to_disk(dev);
401 return sprintf(buf, "%d\n",
402 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
405 static ssize_t disk_ro_show(struct device *dev,
406 struct device_attribute *attr, char *buf)
408 struct gendisk *disk = dev_to_disk(dev);
410 return sprintf(buf, "%d\n", disk->policy ? 1 : 0);
413 static ssize_t disk_size_show(struct device *dev,
414 struct device_attribute *attr, char *buf)
416 struct gendisk *disk = dev_to_disk(dev);
418 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
421 static ssize_t disk_capability_show(struct device *dev,
422 struct device_attribute *attr, char *buf)
424 struct gendisk *disk = dev_to_disk(dev);
426 return sprintf(buf, "%x\n", disk->flags);
429 static ssize_t disk_stat_show(struct device *dev,
430 struct device_attribute *attr, char *buf)
432 struct gendisk *disk = dev_to_disk(dev);
434 preempt_disable();
435 disk_round_stats(disk);
436 preempt_enable();
437 return sprintf(buf,
438 "%8lu %8lu %8llu %8u "
439 "%8lu %8lu %8llu %8u "
440 "%8u %8u %8u"
441 "\n",
442 disk_stat_read(disk, ios[READ]),
443 disk_stat_read(disk, merges[READ]),
444 (unsigned long long)disk_stat_read(disk, sectors[READ]),
445 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
446 disk_stat_read(disk, ios[WRITE]),
447 disk_stat_read(disk, merges[WRITE]),
448 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
449 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
450 disk->in_flight,
451 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
452 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
455 #ifdef CONFIG_FAIL_MAKE_REQUEST
456 static ssize_t disk_fail_show(struct device *dev,
457 struct device_attribute *attr, char *buf)
459 struct gendisk *disk = dev_to_disk(dev);
461 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
464 static ssize_t disk_fail_store(struct device *dev,
465 struct device_attribute *attr,
466 const char *buf, size_t count)
468 struct gendisk *disk = dev_to_disk(dev);
469 int i;
471 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
472 if (i == 0)
473 disk->flags &= ~GENHD_FL_FAIL;
474 else
475 disk->flags |= GENHD_FL_FAIL;
478 return count;
481 #endif
483 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
484 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
485 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
486 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
487 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
488 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
489 #ifdef CONFIG_FAIL_MAKE_REQUEST
490 static struct device_attribute dev_attr_fail =
491 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
492 #endif
494 static struct attribute *disk_attrs[] = {
495 &dev_attr_range.attr,
496 &dev_attr_removable.attr,
497 &dev_attr_ro.attr,
498 &dev_attr_size.attr,
499 &dev_attr_capability.attr,
500 &dev_attr_stat.attr,
501 #ifdef CONFIG_FAIL_MAKE_REQUEST
502 &dev_attr_fail.attr,
503 #endif
504 NULL
507 static struct attribute_group disk_attr_group = {
508 .attrs = disk_attrs,
511 static struct attribute_group *disk_attr_groups[] = {
512 &disk_attr_group,
513 NULL
516 static void disk_release(struct device *dev)
518 struct gendisk *disk = dev_to_disk(dev);
520 kfree(disk->random);
521 kfree(disk->part);
522 free_disk_stats(disk);
523 kfree(disk);
525 struct class block_class = {
526 .name = "block",
529 static struct device_type disk_type = {
530 .name = "disk",
531 .groups = disk_attr_groups,
532 .release = disk_release,
536 * aggregate disk stat collector. Uses the same stats that the sysfs
537 * entries do, above, but makes them available through one seq_file.
539 * The output looks suspiciously like /proc/partitions with a bunch of
540 * extra fields.
543 static void *diskstats_start(struct seq_file *part, loff_t *pos)
545 loff_t k = *pos;
546 struct device *dev;
548 mutex_lock(&block_class_lock);
549 list_for_each_entry(dev, &block_class.devices, node) {
550 if (dev->type != &disk_type)
551 continue;
552 if (!k--)
553 return dev_to_disk(dev);
555 return NULL;
558 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
560 struct gendisk *gp = v;
561 struct device *dev;
563 ++*pos;
564 list_for_each_entry(dev, &gp->dev.node, node) {
565 if (&dev->node == &block_class.devices)
566 return NULL;
567 if (dev->type == &disk_type)
568 return dev_to_disk(dev);
570 return NULL;
573 static void diskstats_stop(struct seq_file *part, void *v)
575 mutex_unlock(&block_class_lock);
578 static int diskstats_show(struct seq_file *s, void *v)
580 struct gendisk *gp = v;
581 char buf[BDEVNAME_SIZE];
582 int n = 0;
585 if (&gp->dev.kobj.entry == block_class.devices.next)
586 seq_puts(s, "major minor name"
587 " rio rmerge rsect ruse wio wmerge "
588 "wsect wuse running use aveq"
589 "\n\n");
592 preempt_disable();
593 disk_round_stats(gp);
594 preempt_enable();
595 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
596 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
597 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
598 (unsigned long long)disk_stat_read(gp, sectors[0]),
599 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
600 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
601 (unsigned long long)disk_stat_read(gp, sectors[1]),
602 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
603 gp->in_flight,
604 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
605 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
607 /* now show all non-0 size partitions of it */
608 for (n = 0; n < gp->minors - 1; n++) {
609 struct hd_struct *hd = gp->part[n];
611 if (!hd || !hd->nr_sects)
612 continue;
614 preempt_disable();
615 part_round_stats(hd);
616 preempt_enable();
617 seq_printf(s, "%4d %4d %s %lu %lu %llu "
618 "%u %lu %lu %llu %u %u %u %u\n",
619 gp->major, n + gp->first_minor + 1,
620 disk_name(gp, n + 1, buf),
621 part_stat_read(hd, ios[0]),
622 part_stat_read(hd, merges[0]),
623 (unsigned long long)part_stat_read(hd, sectors[0]),
624 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
625 part_stat_read(hd, ios[1]),
626 part_stat_read(hd, merges[1]),
627 (unsigned long long)part_stat_read(hd, sectors[1]),
628 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
629 hd->in_flight,
630 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
631 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
635 return 0;
638 const struct seq_operations diskstats_op = {
639 .start = diskstats_start,
640 .next = diskstats_next,
641 .stop = diskstats_stop,
642 .show = diskstats_show
645 static void media_change_notify_thread(struct work_struct *work)
647 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
648 char event[] = "MEDIA_CHANGE=1";
649 char *envp[] = { event, NULL };
652 * set enviroment vars to indicate which event this is for
653 * so that user space will know to go check the media status.
655 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
656 put_device(gd->driverfs_dev);
659 #if 0
660 void genhd_media_change_notify(struct gendisk *disk)
662 get_device(disk->driverfs_dev);
663 schedule_work(&disk->async_notify);
665 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
666 #endif /* 0 */
668 dev_t blk_lookup_devt(const char *name, int part)
670 struct device *dev;
671 dev_t devt = MKDEV(0, 0);
673 mutex_lock(&block_class_lock);
674 list_for_each_entry(dev, &block_class.devices, node) {
675 if (dev->type != &disk_type)
676 continue;
677 if (strcmp(dev->bus_id, name) == 0) {
678 struct gendisk *disk = dev_to_disk(dev);
680 if (part < disk->minors)
681 devt = MKDEV(MAJOR(dev->devt),
682 MINOR(dev->devt) + part);
683 break;
686 mutex_unlock(&block_class_lock);
688 return devt;
690 EXPORT_SYMBOL(blk_lookup_devt);
692 struct gendisk *alloc_disk(int minors)
694 return alloc_disk_node(minors, -1);
697 struct gendisk *alloc_disk_node(int minors, int node_id)
699 struct gendisk *disk;
701 disk = kmalloc_node(sizeof(struct gendisk),
702 GFP_KERNEL | __GFP_ZERO, node_id);
703 if (disk) {
704 if (!init_disk_stats(disk)) {
705 kfree(disk);
706 return NULL;
708 if (minors > 1) {
709 int size = (minors - 1) * sizeof(struct hd_struct *);
710 disk->part = kmalloc_node(size,
711 GFP_KERNEL | __GFP_ZERO, node_id);
712 if (!disk->part) {
713 free_disk_stats(disk);
714 kfree(disk);
715 return NULL;
718 disk->minors = minors;
719 rand_initialize_disk(disk);
720 disk->dev.class = &block_class;
721 disk->dev.type = &disk_type;
722 device_initialize(&disk->dev);
723 INIT_WORK(&disk->async_notify,
724 media_change_notify_thread);
726 return disk;
729 EXPORT_SYMBOL(alloc_disk);
730 EXPORT_SYMBOL(alloc_disk_node);
732 struct kobject *get_disk(struct gendisk *disk)
734 struct module *owner;
735 struct kobject *kobj;
737 if (!disk->fops)
738 return NULL;
739 owner = disk->fops->owner;
740 if (owner && !try_module_get(owner))
741 return NULL;
742 kobj = kobject_get(&disk->dev.kobj);
743 if (kobj == NULL) {
744 module_put(owner);
745 return NULL;
747 return kobj;
751 EXPORT_SYMBOL(get_disk);
753 void put_disk(struct gendisk *disk)
755 if (disk)
756 kobject_put(&disk->dev.kobj);
759 EXPORT_SYMBOL(put_disk);
761 void set_device_ro(struct block_device *bdev, int flag)
763 if (bdev->bd_contains != bdev)
764 bdev->bd_part->policy = flag;
765 else
766 bdev->bd_disk->policy = flag;
769 EXPORT_SYMBOL(set_device_ro);
771 void set_disk_ro(struct gendisk *disk, int flag)
773 int i;
774 disk->policy = flag;
775 for (i = 0; i < disk->minors - 1; i++)
776 if (disk->part[i]) disk->part[i]->policy = flag;
779 EXPORT_SYMBOL(set_disk_ro);
781 int bdev_read_only(struct block_device *bdev)
783 if (!bdev)
784 return 0;
785 else if (bdev->bd_contains != bdev)
786 return bdev->bd_part->policy;
787 else
788 return bdev->bd_disk->policy;
791 EXPORT_SYMBOL(bdev_read_only);
793 int invalidate_partition(struct gendisk *disk, int index)
795 int res = 0;
796 struct block_device *bdev = bdget_disk(disk, index);
797 if (bdev) {
798 fsync_bdev(bdev);
799 res = __invalidate_device(bdev);
800 bdput(bdev);
802 return res;
805 EXPORT_SYMBOL(invalidate_partition);