NFS: Don't clear PG_writeback until after we've processed unstable writes
[linux-2.6/kvm.git] / block / genhd.c
blob441432a142f231b871750caa5fed9a82c0cd750a
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 struct subsystem block_subsys;
21 static DEFINE_MUTEX(block_subsys_lock);
24 * Can be deleted altogether. Later.
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];
33 /* index in the above - for now: assume no multimajor ranges */
34 static inline int major_to_index(int major)
36 return major % BLKDEV_MAJOR_HASH_SIZE;
39 #ifdef CONFIG_PROC_FS
41 void blkdev_show(struct seq_file *f, off_t offset)
43 struct blk_major_name *dp;
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);
53 #endif /* CONFIG_PROC_FS */
55 int register_blkdev(unsigned int major, const char *name)
57 struct blk_major_name **n, *p;
58 int index, ret = 0;
60 mutex_lock(&block_subsys_lock);
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;
69 if (index == 0) {
70 printk("register_blkdev: failed to get major for %s\n",
71 name);
72 ret = -EBUSY;
73 goto out;
75 major = index;
76 ret = major;
79 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
80 if (p == NULL) {
81 ret = -ENOMEM;
82 goto out;
85 p->major = major;
86 strlcpy(p->name, name, sizeof(p->name));
87 p->next = NULL;
88 index = major_to_index(major);
90 for (n = &major_names[index]; *n; n = &(*n)->next) {
91 if ((*n)->major == major)
92 break;
94 if (!*n)
95 *n = p;
96 else
97 ret = -EBUSY;
99 if (ret < 0) {
100 printk("register_blkdev: cannot get major %d for %s\n",
101 major, name);
102 kfree(p);
104 out:
105 mutex_unlock(&block_subsys_lock);
106 return ret;
109 EXPORT_SYMBOL(register_blkdev);
111 /* todo: make void - error printk here */
112 int unregister_blkdev(unsigned int major, const char *name)
114 struct blk_major_name **n;
115 struct blk_major_name *p = NULL;
116 int index = major_to_index(major);
117 int ret = 0;
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;
129 mutex_unlock(&block_subsys_lock);
130 kfree(p);
132 return ret;
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 dev, 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, dev, range, module, probe, lock, data);
151 EXPORT_SYMBOL(blk_register_region);
153 void blk_unregister_region(dev_t dev, unsigned long range)
155 kobj_unmap(bdev_map, dev, range);
158 EXPORT_SYMBOL(blk_unregister_region);
160 static struct kobject *exact_match(dev_t dev, int *part, void *data)
162 struct gendisk *p = data;
163 return &p->kobj;
166 static int exact_lock(dev_t dev, void *data)
168 struct gendisk *p = data;
170 if (!get_disk(p))
171 return -1;
172 return 0;
176 * add_disk - add partitioning information to kernel list
177 * @disk: per-device partitioning information
179 * This function registers the partitioning information in @disk
180 * with the kernel.
182 void add_disk(struct gendisk *disk)
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);
191 EXPORT_SYMBOL(add_disk);
192 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
194 void unlink_gendisk(struct gendisk *disk)
196 blk_unregister_queue(disk);
197 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
198 disk->minors);
201 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
204 * get_gendisk - get partitioning information for a given device
205 * @dev: device to get partitioning information for
207 * This function gets the structure containing partitioning
208 * information for the given device @dev.
210 struct gendisk *get_gendisk(dev_t dev, int *part)
212 struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
213 return kobj ? to_disk(kobj) : NULL;
216 #ifdef CONFIG_PROC_FS
217 /* iterator */
218 static void *part_start(struct seq_file *part, loff_t *pos)
220 struct list_head *p;
221 loff_t l = *pos;
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;
230 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
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);
238 static void part_stop(struct seq_file *part, void *v)
240 mutex_unlock(&block_subsys_lock);
243 static int show_partition(struct seq_file *part, void *v)
245 struct gendisk *sgp = v;
246 int n;
247 char buf[BDEVNAME_SIZE];
249 if (&sgp->kobj.entry == block_subsys.kset.list.next)
250 seq_puts(part, "major minor #blocks name\n\n");
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;
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));
275 return 0;
278 struct seq_operations partitions_op = {
279 .start =part_start,
280 .next = part_next,
281 .stop = part_stop,
282 .show = show_partition
284 #endif
287 extern int blk_dev_init(void);
289 static struct kobject *base_probe(dev_t dev, int *part, void *data)
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;
297 static int __init genhd_device_init(void)
299 int err;
301 bdev_map = kobj_map_init(base_probe, &block_subsys_lock);
302 blk_dev_init();
303 err = subsystem_register(&block_subsys);
304 if (err < 0)
305 printk(KERN_WARNING "%s: subsystem_register error: %d\n",
306 __FUNCTION__, err);
307 return err;
310 subsys_initcall(genhd_device_init);
315 * kobject & sysfs bindings for block devices
317 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
318 char *page)
320 struct gendisk *disk = to_disk(kobj);
321 struct disk_attribute *disk_attr =
322 container_of(attr,struct disk_attribute,attr);
323 ssize_t ret = -EIO;
325 if (disk_attr->show)
326 ret = disk_attr->show(disk,page);
327 return ret;
330 static ssize_t disk_attr_store(struct kobject * kobj, struct attribute * attr,
331 const char *page, size_t count)
333 struct gendisk *disk = to_disk(kobj);
334 struct disk_attribute *disk_attr =
335 container_of(attr,struct disk_attribute,attr);
336 ssize_t ret = 0;
338 if (disk_attr->store)
339 ret = disk_attr->store(disk, page, count);
340 return ret;
343 static struct sysfs_ops disk_sysfs_ops = {
344 .show = &disk_attr_show,
345 .store = &disk_attr_store,
348 static ssize_t disk_uevent_store(struct gendisk * disk,
349 const char *buf, size_t count)
351 kobject_uevent(&disk->kobj, KOBJ_ADD);
352 return count;
354 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
356 dev_t base = MKDEV(disk->major, disk->first_minor);
357 return print_dev_t(page, base);
359 static ssize_t disk_range_read(struct gendisk * disk, char *page)
361 return sprintf(page, "%d\n", disk->minors);
363 static ssize_t disk_removable_read(struct gendisk * disk, char *page)
365 return sprintf(page, "%d\n",
366 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
369 static ssize_t disk_size_read(struct gendisk * disk, char *page)
371 return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
374 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
376 preempt_disable();
377 disk_round_stats(disk);
378 preempt_enable();
379 return sprintf(page,
380 "%8lu %8lu %8llu %8u "
381 "%8lu %8lu %8llu %8u "
382 "%8u %8u %8u"
383 "\n",
384 disk_stat_read(disk, ios[READ]),
385 disk_stat_read(disk, merges[READ]),
386 (unsigned long long)disk_stat_read(disk, sectors[READ]),
387 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
388 disk_stat_read(disk, ios[WRITE]),
389 disk_stat_read(disk, merges[WRITE]),
390 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
391 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
392 disk->in_flight,
393 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
394 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
396 static struct disk_attribute disk_attr_uevent = {
397 .attr = {.name = "uevent", .mode = S_IWUSR },
398 .store = disk_uevent_store
400 static struct disk_attribute disk_attr_dev = {
401 .attr = {.name = "dev", .mode = S_IRUGO },
402 .show = disk_dev_read
404 static struct disk_attribute disk_attr_range = {
405 .attr = {.name = "range", .mode = S_IRUGO },
406 .show = disk_range_read
408 static struct disk_attribute disk_attr_removable = {
409 .attr = {.name = "removable", .mode = S_IRUGO },
410 .show = disk_removable_read
412 static struct disk_attribute disk_attr_size = {
413 .attr = {.name = "size", .mode = S_IRUGO },
414 .show = disk_size_read
416 static struct disk_attribute disk_attr_stat = {
417 .attr = {.name = "stat", .mode = S_IRUGO },
418 .show = disk_stats_read
421 #ifdef CONFIG_FAIL_MAKE_REQUEST
423 static ssize_t disk_fail_store(struct gendisk * disk,
424 const char *buf, size_t count)
426 int i;
428 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
429 if (i == 0)
430 disk->flags &= ~GENHD_FL_FAIL;
431 else
432 disk->flags |= GENHD_FL_FAIL;
435 return count;
437 static ssize_t disk_fail_read(struct gendisk * disk, char *page)
439 return sprintf(page, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
441 static struct disk_attribute disk_attr_fail = {
442 .attr = {.name = "make-it-fail", .mode = S_IRUGO | S_IWUSR },
443 .store = disk_fail_store,
444 .show = disk_fail_read
447 #endif
449 static struct attribute * default_attrs[] = {
450 &disk_attr_uevent.attr,
451 &disk_attr_dev.attr,
452 &disk_attr_range.attr,
453 &disk_attr_removable.attr,
454 &disk_attr_size.attr,
455 &disk_attr_stat.attr,
456 #ifdef CONFIG_FAIL_MAKE_REQUEST
457 &disk_attr_fail.attr,
458 #endif
459 NULL,
462 static void disk_release(struct kobject * kobj)
464 struct gendisk *disk = to_disk(kobj);
465 kfree(disk->random);
466 kfree(disk->part);
467 free_disk_stats(disk);
468 kfree(disk);
471 static struct kobj_type ktype_block = {
472 .release = disk_release,
473 .sysfs_ops = &disk_sysfs_ops,
474 .default_attrs = default_attrs,
477 extern struct kobj_type ktype_part;
479 static int block_uevent_filter(struct kset *kset, struct kobject *kobj)
481 struct kobj_type *ktype = get_ktype(kobj);
483 return ((ktype == &ktype_block) || (ktype == &ktype_part));
486 static int block_uevent(struct kset *kset, struct kobject *kobj, char **envp,
487 int num_envp, char *buffer, int buffer_size)
489 struct kobj_type *ktype = get_ktype(kobj);
490 struct device *physdev;
491 struct gendisk *disk;
492 struct hd_struct *part;
493 int length = 0;
494 int i = 0;
496 if (ktype == &ktype_block) {
497 disk = container_of(kobj, struct gendisk, kobj);
498 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
499 &length, "MINOR=%u", disk->first_minor);
500 } else if (ktype == &ktype_part) {
501 disk = container_of(kobj->parent, struct gendisk, kobj);
502 part = container_of(kobj, struct hd_struct, kobj);
503 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
504 &length, "MINOR=%u",
505 disk->first_minor + part->partno);
506 } else
507 return 0;
509 add_uevent_var(envp, num_envp, &i, buffer, buffer_size, &length,
510 "MAJOR=%u", disk->major);
512 /* add physical device, backing this device */
513 physdev = disk->driverfs_dev;
514 if (physdev) {
515 char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
517 add_uevent_var(envp, num_envp, &i, buffer, buffer_size,
518 &length, "PHYSDEVPATH=%s", path);
519 kfree(path);
521 if (physdev->bus)
522 add_uevent_var(envp, num_envp, &i,
523 buffer, buffer_size, &length,
524 "PHYSDEVBUS=%s",
525 physdev->bus->name);
527 if (physdev->driver)
528 add_uevent_var(envp, num_envp, &i,
529 buffer, buffer_size, &length,
530 "PHYSDEVDRIVER=%s",
531 physdev->driver->name);
534 /* terminate, set to next free slot, shrink available space */
535 envp[i] = NULL;
536 envp = &envp[i];
537 num_envp -= i;
538 buffer = &buffer[length];
539 buffer_size -= length;
541 return 0;
544 static struct kset_uevent_ops block_uevent_ops = {
545 .filter = block_uevent_filter,
546 .uevent = block_uevent,
549 decl_subsys(block, &ktype_block, &block_uevent_ops);
552 * aggregate disk stat collector. Uses the same stats that the sysfs
553 * entries do, above, but makes them available through one seq_file.
554 * Watching a few disks may be efficient through sysfs, but watching
555 * all of them will be more efficient through this interface.
557 * The output looks suspiciously like /proc/partitions with a bunch of
558 * extra fields.
561 /* iterator */
562 static void *diskstats_start(struct seq_file *part, loff_t *pos)
564 loff_t k = *pos;
565 struct list_head *p;
567 mutex_lock(&block_subsys_lock);
568 list_for_each(p, &block_subsys.kset.list)
569 if (!k--)
570 return list_entry(p, struct gendisk, kobj.entry);
571 return NULL;
574 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
576 struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
577 ++*pos;
578 return p==&block_subsys.kset.list ? NULL :
579 list_entry(p, struct gendisk, kobj.entry);
582 static void diskstats_stop(struct seq_file *part, void *v)
584 mutex_unlock(&block_subsys_lock);
587 static int diskstats_show(struct seq_file *s, void *v)
589 struct gendisk *gp = v;
590 char buf[BDEVNAME_SIZE];
591 int n = 0;
594 if (&sgp->kobj.entry == block_subsys.kset.list.next)
595 seq_puts(s, "major minor name"
596 " rio rmerge rsect ruse wio wmerge "
597 "wsect wuse running use aveq"
598 "\n\n");
601 preempt_disable();
602 disk_round_stats(gp);
603 preempt_enable();
604 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
605 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
606 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
607 (unsigned long long)disk_stat_read(gp, sectors[0]),
608 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
609 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
610 (unsigned long long)disk_stat_read(gp, sectors[1]),
611 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
612 gp->in_flight,
613 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
614 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
616 /* now show all non-0 size partitions of it */
617 for (n = 0; n < gp->minors - 1; n++) {
618 struct hd_struct *hd = gp->part[n];
620 if (hd && hd->nr_sects)
621 seq_printf(s, "%4d %4d %s %u %u %u %u\n",
622 gp->major, n + gp->first_minor + 1,
623 disk_name(gp, n + 1, buf),
624 hd->ios[0], hd->sectors[0],
625 hd->ios[1], hd->sectors[1]);
628 return 0;
631 struct seq_operations diskstats_op = {
632 .start = diskstats_start,
633 .next = diskstats_next,
634 .stop = diskstats_stop,
635 .show = diskstats_show
638 struct gendisk *alloc_disk(int minors)
640 return alloc_disk_node(minors, -1);
643 struct gendisk *alloc_disk_node(int minors, int node_id)
645 struct gendisk *disk;
647 disk = kmalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
648 if (disk) {
649 memset(disk, 0, sizeof(struct gendisk));
650 if (!init_disk_stats(disk)) {
651 kfree(disk);
652 return NULL;
654 if (minors > 1) {
655 int size = (minors - 1) * sizeof(struct hd_struct *);
656 disk->part = kmalloc_node(size, GFP_KERNEL, node_id);
657 if (!disk->part) {
658 kfree(disk);
659 return NULL;
661 memset(disk->part, 0, size);
663 disk->minors = minors;
664 kobj_set_kset_s(disk,block_subsys);
665 kobject_init(&disk->kobj);
666 rand_initialize_disk(disk);
668 return disk;
671 EXPORT_SYMBOL(alloc_disk);
672 EXPORT_SYMBOL(alloc_disk_node);
674 struct kobject *get_disk(struct gendisk *disk)
676 struct module *owner;
677 struct kobject *kobj;
679 if (!disk->fops)
680 return NULL;
681 owner = disk->fops->owner;
682 if (owner && !try_module_get(owner))
683 return NULL;
684 kobj = kobject_get(&disk->kobj);
685 if (kobj == NULL) {
686 module_put(owner);
687 return NULL;
689 return kobj;
693 EXPORT_SYMBOL(get_disk);
695 void put_disk(struct gendisk *disk)
697 if (disk)
698 kobject_put(&disk->kobj);
701 EXPORT_SYMBOL(put_disk);
703 void set_device_ro(struct block_device *bdev, int flag)
705 if (bdev->bd_contains != bdev)
706 bdev->bd_part->policy = flag;
707 else
708 bdev->bd_disk->policy = flag;
711 EXPORT_SYMBOL(set_device_ro);
713 void set_disk_ro(struct gendisk *disk, int flag)
715 int i;
716 disk->policy = flag;
717 for (i = 0; i < disk->minors - 1; i++)
718 if (disk->part[i]) disk->part[i]->policy = flag;
721 EXPORT_SYMBOL(set_disk_ro);
723 int bdev_read_only(struct block_device *bdev)
725 if (!bdev)
726 return 0;
727 else if (bdev->bd_contains != bdev)
728 return bdev->bd_part->policy;
729 else
730 return bdev->bd_disk->policy;
733 EXPORT_SYMBOL(bdev_read_only);
735 int invalidate_partition(struct gendisk *disk, int index)
737 int res = 0;
738 struct block_device *bdev = bdget_disk(disk, index);
739 if (bdev) {
740 fsync_bdev(bdev);
741 res = __invalidate_device(bdev);
742 bdput(bdev);
744 return res;
747 EXPORT_SYMBOL(invalidate_partition);