block: misc updates
[linux-2.6/mini2440.git] / block / genhd.c
blob11038fbc75edcbad32ab138419d152102d926d98
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;
186 int retval;
188 disk->flags |= GENHD_FL_UP;
189 blk_register_region(MKDEV(disk->major, disk->first_minor),
190 disk->minors, NULL, exact_match, exact_lock, disk);
191 register_disk(disk);
192 blk_register_queue(disk);
194 bdi = &disk->queue->backing_dev_info;
195 bdi_register_dev(bdi, MKDEV(disk->major, disk->first_minor));
196 retval = sysfs_create_link(&disk->dev.kobj, &bdi->dev->kobj, "bdi");
197 WARN_ON(retval);
200 EXPORT_SYMBOL(add_disk);
201 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
203 void unlink_gendisk(struct gendisk *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 * @devt: device to get partitioning information for
215 * @part: returned partition index
217 * This function gets the structure containing partitioning
218 * information for the given device @devt.
220 struct gendisk *get_gendisk(dev_t devt, int *part)
222 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
223 struct device *dev = kobj_to_dev(kobj);
225 return kobj ? dev_to_disk(dev) : NULL;
229 * print a full list of all partitions - intended for places where the root
230 * filesystem can't be mounted and thus to give the victim some idea of what
231 * went wrong
233 void __init printk_all_partitions(void)
235 struct class_dev_iter iter;
236 struct device *dev;
238 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
239 while ((dev = class_dev_iter_next(&iter))) {
240 struct gendisk *disk = dev_to_disk(dev);
241 char buf[BDEVNAME_SIZE];
242 int n;
245 * Don't show empty devices or things that have been
246 * surpressed
248 if (get_capacity(disk) == 0 ||
249 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
250 continue;
253 * Note, unlike /proc/partitions, I am showing the
254 * numbers in hex - the same format as the root=
255 * option takes.
257 printk("%02x%02x %10llu %s",
258 disk->major, disk->first_minor,
259 (unsigned long long)get_capacity(disk) >> 1,
260 disk_name(disk, 0, buf));
261 if (disk->driverfs_dev != NULL &&
262 disk->driverfs_dev->driver != NULL)
263 printk(" driver: %s\n",
264 disk->driverfs_dev->driver->name);
265 else
266 printk(" (driver?)\n");
268 /* now show the partitions */
269 for (n = 0; n < disk->minors - 1; ++n) {
270 if (disk->part[n] == NULL)
271 continue;
272 if (disk->part[n]->nr_sects == 0)
273 continue;
274 printk(" %02x%02x %10llu %s\n",
275 disk->major, n + 1 + disk->first_minor,
276 (unsigned long long)disk->part[n]->nr_sects >> 1,
277 disk_name(disk, n + 1, buf));
280 class_dev_iter_exit(&iter);
283 #ifdef CONFIG_PROC_FS
284 /* iterator */
285 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
287 loff_t skip = *pos;
288 struct class_dev_iter *iter;
289 struct device *dev;
291 iter = kmalloc(GFP_KERNEL, sizeof(*iter));
292 if (!iter)
293 return ERR_PTR(-ENOMEM);
295 seqf->private = iter;
296 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
297 do {
298 dev = class_dev_iter_next(iter);
299 if (!dev)
300 return NULL;
301 } while (skip--);
303 return dev_to_disk(dev);
306 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
308 struct device *dev;
310 (*pos)++;
311 dev = class_dev_iter_next(seqf->private);
312 if (dev)
313 return dev_to_disk(dev);
315 return NULL;
318 static void disk_seqf_stop(struct seq_file *seqf, void *v)
320 struct class_dev_iter *iter = seqf->private;
322 /* stop is called even after start failed :-( */
323 if (iter) {
324 class_dev_iter_exit(iter);
325 kfree(iter);
329 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
331 static void *p;
333 p = disk_seqf_start(seqf, pos);
334 if (!IS_ERR(p) && p)
335 seq_puts(seqf, "major minor #blocks name\n\n");
336 return p;
339 static int show_partition(struct seq_file *part, void *v)
341 struct gendisk *sgp = v;
342 int n;
343 char buf[BDEVNAME_SIZE];
346 * Print header if start told us to do. This is to preserve
347 * the original behavior of not printing header if no
348 * partition exists. This hackery will be removed later with
349 * class iteration clean up.
351 if (part->private) {
352 seq_puts(part, "major minor #blocks name\n\n");
353 part->private = NULL;
356 /* Don't show non-partitionable removeable devices or empty devices */
357 if (!get_capacity(sgp) ||
358 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
359 return 0;
360 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
361 return 0;
363 /* show the full disk and all non-0 size partitions of it */
364 seq_printf(part, "%4d %4d %10llu %s\n",
365 sgp->major, sgp->first_minor,
366 (unsigned long long)get_capacity(sgp) >> 1,
367 disk_name(sgp, 0, buf));
368 for (n = 0; n < sgp->minors - 1; n++) {
369 if (!sgp->part[n])
370 continue;
371 if (sgp->part[n]->nr_sects == 0)
372 continue;
373 seq_printf(part, "%4d %4d %10llu %s\n",
374 sgp->major, n + 1 + sgp->first_minor,
375 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
376 disk_name(sgp, n + 1, buf));
379 return 0;
382 const struct seq_operations partitions_op = {
383 .start = show_partition_start,
384 .next = disk_seqf_next,
385 .stop = disk_seqf_stop,
386 .show = show_partition
388 #endif
391 static struct kobject *base_probe(dev_t devt, int *part, void *data)
393 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
394 /* Make old-style 2.4 aliases work */
395 request_module("block-major-%d", MAJOR(devt));
396 return NULL;
399 static int __init genhd_device_init(void)
401 int error;
403 block_class.dev_kobj = sysfs_dev_block_kobj;
404 error = class_register(&block_class);
405 if (unlikely(error))
406 return error;
407 bdev_map = kobj_map_init(base_probe, &block_class_lock);
408 blk_dev_init();
410 #ifndef CONFIG_SYSFS_DEPRECATED
411 /* create top-level block dir */
412 block_depr = kobject_create_and_add("block", NULL);
413 #endif
414 return 0;
417 subsys_initcall(genhd_device_init);
419 static ssize_t disk_range_show(struct device *dev,
420 struct device_attribute *attr, char *buf)
422 struct gendisk *disk = dev_to_disk(dev);
424 return sprintf(buf, "%d\n", disk->minors);
427 static ssize_t disk_removable_show(struct device *dev,
428 struct device_attribute *attr, char *buf)
430 struct gendisk *disk = dev_to_disk(dev);
432 return sprintf(buf, "%d\n",
433 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
436 static ssize_t disk_ro_show(struct device *dev,
437 struct device_attribute *attr, char *buf)
439 struct gendisk *disk = dev_to_disk(dev);
441 return sprintf(buf, "%d\n", disk->policy ? 1 : 0);
444 static ssize_t disk_size_show(struct device *dev,
445 struct device_attribute *attr, char *buf)
447 struct gendisk *disk = dev_to_disk(dev);
449 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
452 static ssize_t disk_capability_show(struct device *dev,
453 struct device_attribute *attr, char *buf)
455 struct gendisk *disk = dev_to_disk(dev);
457 return sprintf(buf, "%x\n", disk->flags);
460 static ssize_t disk_stat_show(struct device *dev,
461 struct device_attribute *attr, char *buf)
463 struct gendisk *disk = dev_to_disk(dev);
465 preempt_disable();
466 disk_round_stats(disk);
467 preempt_enable();
468 return sprintf(buf,
469 "%8lu %8lu %8llu %8u "
470 "%8lu %8lu %8llu %8u "
471 "%8u %8u %8u"
472 "\n",
473 disk_stat_read(disk, ios[READ]),
474 disk_stat_read(disk, merges[READ]),
475 (unsigned long long)disk_stat_read(disk, sectors[READ]),
476 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
477 disk_stat_read(disk, ios[WRITE]),
478 disk_stat_read(disk, merges[WRITE]),
479 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
480 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
481 disk->in_flight,
482 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
483 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
486 #ifdef CONFIG_FAIL_MAKE_REQUEST
487 static ssize_t disk_fail_show(struct device *dev,
488 struct device_attribute *attr, char *buf)
490 struct gendisk *disk = dev_to_disk(dev);
492 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
495 static ssize_t disk_fail_store(struct device *dev,
496 struct device_attribute *attr,
497 const char *buf, size_t count)
499 struct gendisk *disk = dev_to_disk(dev);
500 int i;
502 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
503 if (i == 0)
504 disk->flags &= ~GENHD_FL_FAIL;
505 else
506 disk->flags |= GENHD_FL_FAIL;
509 return count;
512 #endif
514 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
515 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
516 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
517 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
518 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
519 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
520 #ifdef CONFIG_FAIL_MAKE_REQUEST
521 static struct device_attribute dev_attr_fail =
522 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
523 #endif
525 static struct attribute *disk_attrs[] = {
526 &dev_attr_range.attr,
527 &dev_attr_removable.attr,
528 &dev_attr_ro.attr,
529 &dev_attr_size.attr,
530 &dev_attr_capability.attr,
531 &dev_attr_stat.attr,
532 #ifdef CONFIG_FAIL_MAKE_REQUEST
533 &dev_attr_fail.attr,
534 #endif
535 NULL
538 static struct attribute_group disk_attr_group = {
539 .attrs = disk_attrs,
542 static struct attribute_group *disk_attr_groups[] = {
543 &disk_attr_group,
544 NULL
547 static void disk_release(struct device *dev)
549 struct gendisk *disk = dev_to_disk(dev);
551 kfree(disk->random);
552 kfree(disk->part);
553 free_disk_stats(disk);
554 kfree(disk);
556 struct class block_class = {
557 .name = "block",
560 static struct device_type disk_type = {
561 .name = "disk",
562 .groups = disk_attr_groups,
563 .release = disk_release,
566 #ifdef CONFIG_PROC_FS
567 static int diskstats_show(struct seq_file *s, void *v)
569 struct gendisk *gp = v;
570 char buf[BDEVNAME_SIZE];
571 int n;
574 if (&gp->dev.kobj.entry == block_class.devices.next)
575 seq_puts(s, "major minor name"
576 " rio rmerge rsect ruse wio wmerge "
577 "wsect wuse running use aveq"
578 "\n\n");
581 preempt_disable();
582 disk_round_stats(gp);
583 preempt_enable();
584 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
585 gp->major, gp->first_minor, disk_name(gp, 0, buf),
586 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
587 (unsigned long long)disk_stat_read(gp, sectors[0]),
588 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
589 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
590 (unsigned long long)disk_stat_read(gp, sectors[1]),
591 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
592 gp->in_flight,
593 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
594 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
596 /* now show all non-0 size partitions of it */
597 for (n = 0; n < gp->minors - 1; n++) {
598 struct hd_struct *hd = gp->part[n];
600 if (!hd || !hd->nr_sects)
601 continue;
603 preempt_disable();
604 part_round_stats(hd);
605 preempt_enable();
606 seq_printf(s, "%4d %4d %s %lu %lu %llu "
607 "%u %lu %lu %llu %u %u %u %u\n",
608 gp->major, n + gp->first_minor + 1,
609 disk_name(gp, n + 1, buf),
610 part_stat_read(hd, ios[0]),
611 part_stat_read(hd, merges[0]),
612 (unsigned long long)part_stat_read(hd, sectors[0]),
613 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
614 part_stat_read(hd, ios[1]),
615 part_stat_read(hd, merges[1]),
616 (unsigned long long)part_stat_read(hd, sectors[1]),
617 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
618 hd->in_flight,
619 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
620 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
624 return 0;
627 const struct seq_operations diskstats_op = {
628 .start = disk_seqf_start,
629 .next = disk_seqf_next,
630 .stop = disk_seqf_stop,
631 .show = diskstats_show
633 #endif /* CONFIG_PROC_FS */
635 static void media_change_notify_thread(struct work_struct *work)
637 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
638 char event[] = "MEDIA_CHANGE=1";
639 char *envp[] = { event, NULL };
642 * set enviroment vars to indicate which event this is for
643 * so that user space will know to go check the media status.
645 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
646 put_device(gd->driverfs_dev);
649 #if 0
650 void genhd_media_change_notify(struct gendisk *disk)
652 get_device(disk->driverfs_dev);
653 schedule_work(&disk->async_notify);
655 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
656 #endif /* 0 */
658 dev_t blk_lookup_devt(const char *name, int part)
660 dev_t devt = MKDEV(0, 0);
661 struct class_dev_iter iter;
662 struct device *dev;
664 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
665 while ((dev = class_dev_iter_next(&iter))) {
666 struct gendisk *disk = dev_to_disk(dev);
668 if (!strcmp(dev->bus_id, name) && part < disk->minors) {
669 devt = MKDEV(MAJOR(dev->devt),
670 MINOR(dev->devt) + part);
671 break;
674 class_dev_iter_exit(&iter);
675 return devt;
677 EXPORT_SYMBOL(blk_lookup_devt);
679 struct gendisk *alloc_disk(int minors)
681 return alloc_disk_node(minors, -1);
684 struct gendisk *alloc_disk_node(int minors, int node_id)
686 struct gendisk *disk;
688 disk = kmalloc_node(sizeof(struct gendisk),
689 GFP_KERNEL | __GFP_ZERO, node_id);
690 if (disk) {
691 if (!init_disk_stats(disk)) {
692 kfree(disk);
693 return NULL;
695 if (minors > 1) {
696 int size = (minors - 1) * sizeof(struct hd_struct *);
697 disk->part = kmalloc_node(size,
698 GFP_KERNEL | __GFP_ZERO, node_id);
699 if (!disk->part) {
700 free_disk_stats(disk);
701 kfree(disk);
702 return NULL;
705 disk->minors = minors;
706 rand_initialize_disk(disk);
707 disk->dev.class = &block_class;
708 disk->dev.type = &disk_type;
709 device_initialize(&disk->dev);
710 INIT_WORK(&disk->async_notify,
711 media_change_notify_thread);
713 return disk;
716 EXPORT_SYMBOL(alloc_disk);
717 EXPORT_SYMBOL(alloc_disk_node);
719 struct kobject *get_disk(struct gendisk *disk)
721 struct module *owner;
722 struct kobject *kobj;
724 if (!disk->fops)
725 return NULL;
726 owner = disk->fops->owner;
727 if (owner && !try_module_get(owner))
728 return NULL;
729 kobj = kobject_get(&disk->dev.kobj);
730 if (kobj == NULL) {
731 module_put(owner);
732 return NULL;
734 return kobj;
738 EXPORT_SYMBOL(get_disk);
740 void put_disk(struct gendisk *disk)
742 if (disk)
743 kobject_put(&disk->dev.kobj);
746 EXPORT_SYMBOL(put_disk);
748 void set_device_ro(struct block_device *bdev, int flag)
750 if (bdev->bd_contains != bdev)
751 bdev->bd_part->policy = flag;
752 else
753 bdev->bd_disk->policy = flag;
756 EXPORT_SYMBOL(set_device_ro);
758 void set_disk_ro(struct gendisk *disk, int flag)
760 int i;
761 disk->policy = flag;
762 for (i = 0; i < disk->minors - 1; i++)
763 if (disk->part[i]) disk->part[i]->policy = flag;
766 EXPORT_SYMBOL(set_disk_ro);
768 int bdev_read_only(struct block_device *bdev)
770 if (!bdev)
771 return 0;
772 else if (bdev->bd_contains != bdev)
773 return bdev->bd_part->policy;
774 else
775 return bdev->bd_disk->policy;
778 EXPORT_SYMBOL(bdev_read_only);
780 int invalidate_partition(struct gendisk *disk, int index)
782 int res = 0;
783 struct block_device *bdev = bdget_disk(disk, index);
784 if (bdev) {
785 fsync_bdev(bdev);
786 res = __invalidate_device(bdev);
787 bdput(bdev);
789 return res;
792 EXPORT_SYMBOL(invalidate_partition);