Linux 2.6.26-rc5
[linux-2.6/openmoko-kernel/knife-kernel.git] / block / genhd.c
blob129ad939f9dd7179cab6da67497133cafee7156f
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);
193 bdi = &disk->queue->backing_dev_info;
194 bdi_register_dev(bdi, MKDEV(disk->major, disk->first_minor));
195 sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi");
198 EXPORT_SYMBOL(add_disk);
199 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
201 void unlink_gendisk(struct gendisk *disk)
203 sysfs_remove_link(&disk->dev.kobj, "bdi");
204 bdi_unregister(&disk->queue->backing_dev_info);
205 blk_unregister_queue(disk);
206 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
207 disk->minors);
211 * get_gendisk - get partitioning information for a given device
212 * @dev: device to get partitioning information for
214 * This function gets the structure containing partitioning
215 * information for the given device @dev.
217 struct gendisk *get_gendisk(dev_t devt, int *part)
219 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
220 struct device *dev = kobj_to_dev(kobj);
222 return kobj ? dev_to_disk(dev) : NULL;
226 * print a full list of all partitions - intended for places where the root
227 * filesystem can't be mounted and thus to give the victim some idea of what
228 * went wrong
230 void __init printk_all_partitions(void)
232 struct device *dev;
233 struct gendisk *sgp;
234 char buf[BDEVNAME_SIZE];
235 int n;
237 mutex_lock(&block_class_lock);
238 /* For each block device... */
239 list_for_each_entry(dev, &block_class.devices, node) {
240 if (dev->type != &disk_type)
241 continue;
242 sgp = dev_to_disk(dev);
244 * Don't show empty devices or things that have been surpressed
246 if (get_capacity(sgp) == 0 ||
247 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
248 continue;
251 * Note, unlike /proc/partitions, I am showing the numbers in
252 * hex - the same format as the root= option takes.
254 printk("%02x%02x %10llu %s",
255 sgp->major, sgp->first_minor,
256 (unsigned long long)get_capacity(sgp) >> 1,
257 disk_name(sgp, 0, buf));
258 if (sgp->driverfs_dev != NULL &&
259 sgp->driverfs_dev->driver != NULL)
260 printk(" driver: %s\n",
261 sgp->driverfs_dev->driver->name);
262 else
263 printk(" (driver?)\n");
265 /* now show the partitions */
266 for (n = 0; n < sgp->minors - 1; ++n) {
267 if (sgp->part[n] == NULL)
268 continue;
269 if (sgp->part[n]->nr_sects == 0)
270 continue;
271 printk(" %02x%02x %10llu %s\n",
272 sgp->major, n + 1 + sgp->first_minor,
273 (unsigned long long)sgp->part[n]->nr_sects >> 1,
274 disk_name(sgp, n + 1, buf));
278 mutex_unlock(&block_class_lock);
281 #ifdef CONFIG_PROC_FS
282 /* iterator */
283 static void *part_start(struct seq_file *part, loff_t *pos)
285 loff_t k = *pos;
286 struct device *dev;
288 mutex_lock(&block_class_lock);
289 list_for_each_entry(dev, &block_class.devices, node) {
290 if (dev->type != &disk_type)
291 continue;
292 if (!k--)
293 return dev_to_disk(dev);
295 return NULL;
298 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
300 struct gendisk *gp = v;
301 struct device *dev;
302 ++*pos;
303 list_for_each_entry(dev, &gp->dev.node, node) {
304 if (&dev->node == &block_class.devices)
305 return NULL;
306 if (dev->type == &disk_type)
307 return dev_to_disk(dev);
309 return NULL;
312 static void part_stop(struct seq_file *part, void *v)
314 mutex_unlock(&block_class_lock);
317 static int show_partition(struct seq_file *part, void *v)
319 struct gendisk *sgp = v;
320 int n;
321 char buf[BDEVNAME_SIZE];
323 if (&sgp->dev.node == block_class.devices.next)
324 seq_puts(part, "major minor #blocks name\n\n");
326 /* Don't show non-partitionable removeable devices or empty devices */
327 if (!get_capacity(sgp) ||
328 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
329 return 0;
330 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
331 return 0;
333 /* show the full disk and all non-0 size partitions of it */
334 seq_printf(part, "%4d %4d %10llu %s\n",
335 sgp->major, sgp->first_minor,
336 (unsigned long long)get_capacity(sgp) >> 1,
337 disk_name(sgp, 0, buf));
338 for (n = 0; n < sgp->minors - 1; n++) {
339 if (!sgp->part[n])
340 continue;
341 if (sgp->part[n]->nr_sects == 0)
342 continue;
343 seq_printf(part, "%4d %4d %10llu %s\n",
344 sgp->major, n + 1 + sgp->first_minor,
345 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
346 disk_name(sgp, n + 1, buf));
349 return 0;
352 const struct seq_operations partitions_op = {
353 .start = part_start,
354 .next = part_next,
355 .stop = part_stop,
356 .show = show_partition
358 #endif
361 static struct kobject *base_probe(dev_t devt, int *part, void *data)
363 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
364 /* Make old-style 2.4 aliases work */
365 request_module("block-major-%d", MAJOR(devt));
366 return NULL;
369 static int __init genhd_device_init(void)
371 int error = class_register(&block_class);
372 if (unlikely(error))
373 return error;
374 bdev_map = kobj_map_init(base_probe, &block_class_lock);
375 blk_dev_init();
377 #ifndef CONFIG_SYSFS_DEPRECATED
378 /* create top-level block dir */
379 block_depr = kobject_create_and_add("block", NULL);
380 #endif
381 return 0;
384 subsys_initcall(genhd_device_init);
386 static ssize_t disk_range_show(struct device *dev,
387 struct device_attribute *attr, char *buf)
389 struct gendisk *disk = dev_to_disk(dev);
391 return sprintf(buf, "%d\n", disk->minors);
394 static ssize_t disk_removable_show(struct device *dev,
395 struct device_attribute *attr, char *buf)
397 struct gendisk *disk = dev_to_disk(dev);
399 return sprintf(buf, "%d\n",
400 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
403 static ssize_t disk_size_show(struct device *dev,
404 struct device_attribute *attr, char *buf)
406 struct gendisk *disk = dev_to_disk(dev);
408 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
411 static ssize_t disk_capability_show(struct device *dev,
412 struct device_attribute *attr, char *buf)
414 struct gendisk *disk = dev_to_disk(dev);
416 return sprintf(buf, "%x\n", disk->flags);
419 static ssize_t disk_stat_show(struct device *dev,
420 struct device_attribute *attr, char *buf)
422 struct gendisk *disk = dev_to_disk(dev);
424 preempt_disable();
425 disk_round_stats(disk);
426 preempt_enable();
427 return sprintf(buf,
428 "%8lu %8lu %8llu %8u "
429 "%8lu %8lu %8llu %8u "
430 "%8u %8u %8u"
431 "\n",
432 disk_stat_read(disk, ios[READ]),
433 disk_stat_read(disk, merges[READ]),
434 (unsigned long long)disk_stat_read(disk, sectors[READ]),
435 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
436 disk_stat_read(disk, ios[WRITE]),
437 disk_stat_read(disk, merges[WRITE]),
438 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
439 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
440 disk->in_flight,
441 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
442 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
445 #ifdef CONFIG_FAIL_MAKE_REQUEST
446 static ssize_t disk_fail_show(struct device *dev,
447 struct device_attribute *attr, char *buf)
449 struct gendisk *disk = dev_to_disk(dev);
451 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
454 static ssize_t disk_fail_store(struct device *dev,
455 struct device_attribute *attr,
456 const char *buf, size_t count)
458 struct gendisk *disk = dev_to_disk(dev);
459 int i;
461 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
462 if (i == 0)
463 disk->flags &= ~GENHD_FL_FAIL;
464 else
465 disk->flags |= GENHD_FL_FAIL;
468 return count;
471 #endif
473 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
474 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
475 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
476 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
477 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
478 #ifdef CONFIG_FAIL_MAKE_REQUEST
479 static struct device_attribute dev_attr_fail =
480 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
481 #endif
483 static struct attribute *disk_attrs[] = {
484 &dev_attr_range.attr,
485 &dev_attr_removable.attr,
486 &dev_attr_size.attr,
487 &dev_attr_capability.attr,
488 &dev_attr_stat.attr,
489 #ifdef CONFIG_FAIL_MAKE_REQUEST
490 &dev_attr_fail.attr,
491 #endif
492 NULL
495 static struct attribute_group disk_attr_group = {
496 .attrs = disk_attrs,
499 static struct attribute_group *disk_attr_groups[] = {
500 &disk_attr_group,
501 NULL
504 static void disk_release(struct device *dev)
506 struct gendisk *disk = dev_to_disk(dev);
508 kfree(disk->random);
509 kfree(disk->part);
510 free_disk_stats(disk);
511 kfree(disk);
513 struct class block_class = {
514 .name = "block",
517 static struct device_type disk_type = {
518 .name = "disk",
519 .groups = disk_attr_groups,
520 .release = disk_release,
524 * aggregate disk stat collector. Uses the same stats that the sysfs
525 * entries do, above, but makes them available through one seq_file.
527 * The output looks suspiciously like /proc/partitions with a bunch of
528 * extra fields.
531 static void *diskstats_start(struct seq_file *part, loff_t *pos)
533 loff_t k = *pos;
534 struct device *dev;
536 mutex_lock(&block_class_lock);
537 list_for_each_entry(dev, &block_class.devices, node) {
538 if (dev->type != &disk_type)
539 continue;
540 if (!k--)
541 return dev_to_disk(dev);
543 return NULL;
546 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
548 struct gendisk *gp = v;
549 struct device *dev;
551 ++*pos;
552 list_for_each_entry(dev, &gp->dev.node, node) {
553 if (&dev->node == &block_class.devices)
554 return NULL;
555 if (dev->type == &disk_type)
556 return dev_to_disk(dev);
558 return NULL;
561 static void diskstats_stop(struct seq_file *part, void *v)
563 mutex_unlock(&block_class_lock);
566 static int diskstats_show(struct seq_file *s, void *v)
568 struct gendisk *gp = v;
569 char buf[BDEVNAME_SIZE];
570 int n = 0;
573 if (&gp->dev.kobj.entry == block_class.devices.next)
574 seq_puts(s, "major minor name"
575 " rio rmerge rsect ruse wio wmerge "
576 "wsect wuse running use aveq"
577 "\n\n");
580 preempt_disable();
581 disk_round_stats(gp);
582 preempt_enable();
583 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
584 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
585 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
586 (unsigned long long)disk_stat_read(gp, sectors[0]),
587 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
588 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
589 (unsigned long long)disk_stat_read(gp, sectors[1]),
590 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
591 gp->in_flight,
592 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
593 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
595 /* now show all non-0 size partitions of it */
596 for (n = 0; n < gp->minors - 1; n++) {
597 struct hd_struct *hd = gp->part[n];
599 if (!hd || !hd->nr_sects)
600 continue;
602 preempt_disable();
603 part_round_stats(hd);
604 preempt_enable();
605 seq_printf(s, "%4d %4d %s %lu %lu %llu "
606 "%u %lu %lu %llu %u %u %u %u\n",
607 gp->major, n + gp->first_minor + 1,
608 disk_name(gp, n + 1, buf),
609 part_stat_read(hd, ios[0]),
610 part_stat_read(hd, merges[0]),
611 (unsigned long long)part_stat_read(hd, sectors[0]),
612 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
613 part_stat_read(hd, ios[1]),
614 part_stat_read(hd, merges[1]),
615 (unsigned long long)part_stat_read(hd, sectors[1]),
616 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
617 hd->in_flight,
618 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
619 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
623 return 0;
626 const struct seq_operations diskstats_op = {
627 .start = diskstats_start,
628 .next = diskstats_next,
629 .stop = diskstats_stop,
630 .show = diskstats_show
633 static void media_change_notify_thread(struct work_struct *work)
635 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
636 char event[] = "MEDIA_CHANGE=1";
637 char *envp[] = { event, NULL };
640 * set enviroment vars to indicate which event this is for
641 * so that user space will know to go check the media status.
643 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
644 put_device(gd->driverfs_dev);
647 #if 0
648 void genhd_media_change_notify(struct gendisk *disk)
650 get_device(disk->driverfs_dev);
651 schedule_work(&disk->async_notify);
653 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
654 #endif /* 0 */
656 dev_t blk_lookup_devt(const char *name, int part)
658 struct device *dev;
659 dev_t devt = MKDEV(0, 0);
661 mutex_lock(&block_class_lock);
662 list_for_each_entry(dev, &block_class.devices, node) {
663 if (strcmp(dev->bus_id, name) == 0) {
664 struct gendisk *disk = dev_to_disk(dev);
666 if (part < disk->minors)
667 devt = MKDEV(MAJOR(dev->devt),
668 MINOR(dev->devt) + part);
669 break;
672 mutex_unlock(&block_class_lock);
674 return devt;
676 EXPORT_SYMBOL(blk_lookup_devt);
678 struct gendisk *alloc_disk(int minors)
680 return alloc_disk_node(minors, -1);
683 struct gendisk *alloc_disk_node(int minors, int node_id)
685 struct gendisk *disk;
687 disk = kmalloc_node(sizeof(struct gendisk),
688 GFP_KERNEL | __GFP_ZERO, node_id);
689 if (disk) {
690 if (!init_disk_stats(disk)) {
691 kfree(disk);
692 return NULL;
694 if (minors > 1) {
695 int size = (minors - 1) * sizeof(struct hd_struct *);
696 disk->part = kmalloc_node(size,
697 GFP_KERNEL | __GFP_ZERO, node_id);
698 if (!disk->part) {
699 free_disk_stats(disk);
700 kfree(disk);
701 return NULL;
704 disk->minors = minors;
705 rand_initialize_disk(disk);
706 disk->dev.class = &block_class;
707 disk->dev.type = &disk_type;
708 device_initialize(&disk->dev);
709 INIT_WORK(&disk->async_notify,
710 media_change_notify_thread);
712 return disk;
715 EXPORT_SYMBOL(alloc_disk);
716 EXPORT_SYMBOL(alloc_disk_node);
718 struct kobject *get_disk(struct gendisk *disk)
720 struct module *owner;
721 struct kobject *kobj;
723 if (!disk->fops)
724 return NULL;
725 owner = disk->fops->owner;
726 if (owner && !try_module_get(owner))
727 return NULL;
728 kobj = kobject_get(&disk->dev.kobj);
729 if (kobj == NULL) {
730 module_put(owner);
731 return NULL;
733 return kobj;
737 EXPORT_SYMBOL(get_disk);
739 void put_disk(struct gendisk *disk)
741 if (disk)
742 kobject_put(&disk->dev.kobj);
745 EXPORT_SYMBOL(put_disk);
747 void set_device_ro(struct block_device *bdev, int flag)
749 if (bdev->bd_contains != bdev)
750 bdev->bd_part->policy = flag;
751 else
752 bdev->bd_disk->policy = flag;
755 EXPORT_SYMBOL(set_device_ro);
757 void set_disk_ro(struct gendisk *disk, int flag)
759 int i;
760 disk->policy = flag;
761 for (i = 0; i < disk->minors - 1; i++)
762 if (disk->part[i]) disk->part[i]->policy = flag;
765 EXPORT_SYMBOL(set_disk_ro);
767 int bdev_read_only(struct block_device *bdev)
769 if (!bdev)
770 return 0;
771 else if (bdev->bd_contains != bdev)
772 return bdev->bd_part->policy;
773 else
774 return bdev->bd_disk->policy;
777 EXPORT_SYMBOL(bdev_read_only);
779 int invalidate_partition(struct gendisk *disk, int index)
781 int res = 0;
782 struct block_device *bdev = bdget_disk(disk, index);
783 if (bdev) {
784 fsync_bdev(bdev);
785 res = __invalidate_device(bdev);
786 bdput(bdev);
788 return res;
791 EXPORT_SYMBOL(invalidate_partition);