x86: implement gbpages support in change_page_attr()
[wrt350n-kernel.git] / block / genhd.c
blobde2ebb2fab436119cf16b862c720c15be6d991c4
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 static DEFINE_MUTEX(block_class_lock);
21 #ifndef CONFIG_SYSFS_DEPRECATED
22 struct kobject *block_depr;
23 #endif
26 * Can be deleted altogether. Later.
29 static struct blk_major_name {
30 struct blk_major_name *next;
31 int major;
32 char name[16];
33 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
35 /* index in the above - for now: assume no multimajor ranges */
36 static inline int major_to_index(int major)
38 return major % BLKDEV_MAJOR_HASH_SIZE;
41 #ifdef CONFIG_PROC_FS
42 void blkdev_show(struct seq_file *f, off_t offset)
44 struct blk_major_name *dp;
46 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
47 mutex_lock(&block_class_lock);
48 for (dp = major_names[offset]; dp; dp = dp->next)
49 seq_printf(f, "%3d %s\n", dp->major, dp->name);
50 mutex_unlock(&block_class_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_class_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_class_lock);
106 return ret;
109 EXPORT_SYMBOL(register_blkdev);
111 void unregister_blkdev(unsigned int major, const char *name)
113 struct blk_major_name **n;
114 struct blk_major_name *p = NULL;
115 int index = major_to_index(major);
117 mutex_lock(&block_class_lock);
118 for (n = &major_names[index]; *n; n = &(*n)->next)
119 if ((*n)->major == major)
120 break;
121 if (!*n || strcmp((*n)->name, name)) {
122 WARN_ON(1);
123 } else {
124 p = *n;
125 *n = p->next;
127 mutex_unlock(&block_class_lock);
128 kfree(p);
131 EXPORT_SYMBOL(unregister_blkdev);
133 static struct kobj_map *bdev_map;
136 * Register device numbers dev..(dev+range-1)
137 * range must be nonzero
138 * The hash chain is sorted on range, so that subranges can override.
140 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
141 struct kobject *(*probe)(dev_t, int *, void *),
142 int (*lock)(dev_t, void *), void *data)
144 kobj_map(bdev_map, devt, range, module, probe, lock, data);
147 EXPORT_SYMBOL(blk_register_region);
149 void blk_unregister_region(dev_t devt, unsigned long range)
151 kobj_unmap(bdev_map, devt, range);
154 EXPORT_SYMBOL(blk_unregister_region);
156 static struct kobject *exact_match(dev_t devt, int *part, void *data)
158 struct gendisk *p = data;
160 return &p->dev.kobj;
163 static int exact_lock(dev_t devt, void *data)
165 struct gendisk *p = data;
167 if (!get_disk(p))
168 return -1;
169 return 0;
173 * add_disk - add partitioning information to kernel list
174 * @disk: per-device partitioning information
176 * This function registers the partitioning information in @disk
177 * with the kernel.
179 void add_disk(struct gendisk *disk)
181 disk->flags |= GENHD_FL_UP;
182 blk_register_region(MKDEV(disk->major, disk->first_minor),
183 disk->minors, NULL, exact_match, exact_lock, disk);
184 register_disk(disk);
185 blk_register_queue(disk);
188 EXPORT_SYMBOL(add_disk);
189 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
191 void unlink_gendisk(struct gendisk *disk)
193 blk_unregister_queue(disk);
194 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
195 disk->minors);
199 * get_gendisk - get partitioning information for a given device
200 * @dev: device to get partitioning information for
202 * This function gets the structure containing partitioning
203 * information for the given device @dev.
205 struct gendisk *get_gendisk(dev_t devt, int *part)
207 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
208 struct device *dev = kobj_to_dev(kobj);
210 return kobj ? dev_to_disk(dev) : NULL;
214 * print a full list of all partitions - intended for places where the root
215 * filesystem can't be mounted and thus to give the victim some idea of what
216 * went wrong
218 void __init printk_all_partitions(void)
220 struct device *dev;
221 struct gendisk *sgp;
222 char buf[BDEVNAME_SIZE];
223 int n;
225 mutex_lock(&block_class_lock);
226 /* For each block device... */
227 list_for_each_entry(dev, &block_class.devices, node) {
228 if (dev->type != &disk_type)
229 continue;
230 sgp = dev_to_disk(dev);
232 * Don't show empty devices or things that have been surpressed
234 if (get_capacity(sgp) == 0 ||
235 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
236 continue;
239 * Note, unlike /proc/partitions, I am showing the numbers in
240 * hex - the same format as the root= option takes.
242 printk("%02x%02x %10llu %s",
243 sgp->major, sgp->first_minor,
244 (unsigned long long)get_capacity(sgp) >> 1,
245 disk_name(sgp, 0, buf));
246 if (sgp->driverfs_dev != NULL &&
247 sgp->driverfs_dev->driver != NULL)
248 printk(" driver: %s\n",
249 sgp->driverfs_dev->driver->name);
250 else
251 printk(" (driver?)\n");
253 /* now show the partitions */
254 for (n = 0; n < sgp->minors - 1; ++n) {
255 if (sgp->part[n] == NULL)
256 continue;
257 if (sgp->part[n]->nr_sects == 0)
258 continue;
259 printk(" %02x%02x %10llu %s\n",
260 sgp->major, n + 1 + sgp->first_minor,
261 (unsigned long long)sgp->part[n]->nr_sects >> 1,
262 disk_name(sgp, n + 1, buf));
266 mutex_unlock(&block_class_lock);
269 #ifdef CONFIG_PROC_FS
270 /* iterator */
271 static void *part_start(struct seq_file *part, loff_t *pos)
273 loff_t k = *pos;
274 struct device *dev;
276 mutex_lock(&block_class_lock);
277 list_for_each_entry(dev, &block_class.devices, node) {
278 if (dev->type != &disk_type)
279 continue;
280 if (!k--)
281 return dev_to_disk(dev);
283 return NULL;
286 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
288 struct gendisk *gp = v;
289 struct device *dev;
290 ++*pos;
291 list_for_each_entry(dev, &gp->dev.node, node) {
292 if (&dev->node == &block_class.devices)
293 return NULL;
294 if (dev->type == &disk_type)
295 return dev_to_disk(dev);
297 return NULL;
300 static void part_stop(struct seq_file *part, void *v)
302 mutex_unlock(&block_class_lock);
305 static int show_partition(struct seq_file *part, void *v)
307 struct gendisk *sgp = v;
308 int n;
309 char buf[BDEVNAME_SIZE];
311 if (&sgp->dev.node == block_class.devices.next)
312 seq_puts(part, "major minor #blocks name\n\n");
314 /* Don't show non-partitionable removeable devices or empty devices */
315 if (!get_capacity(sgp) ||
316 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
317 return 0;
318 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
319 return 0;
321 /* show the full disk and all non-0 size partitions of it */
322 seq_printf(part, "%4d %4d %10llu %s\n",
323 sgp->major, sgp->first_minor,
324 (unsigned long long)get_capacity(sgp) >> 1,
325 disk_name(sgp, 0, buf));
326 for (n = 0; n < sgp->minors - 1; n++) {
327 if (!sgp->part[n])
328 continue;
329 if (sgp->part[n]->nr_sects == 0)
330 continue;
331 seq_printf(part, "%4d %4d %10llu %s\n",
332 sgp->major, n + 1 + sgp->first_minor,
333 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
334 disk_name(sgp, n + 1, buf));
337 return 0;
340 const struct seq_operations partitions_op = {
341 .start = part_start,
342 .next = part_next,
343 .stop = part_stop,
344 .show = show_partition
346 #endif
349 extern int blk_dev_init(void);
351 static struct kobject *base_probe(dev_t devt, int *part, void *data)
353 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
354 /* Make old-style 2.4 aliases work */
355 request_module("block-major-%d", MAJOR(devt));
356 return NULL;
359 static int __init genhd_device_init(void)
361 class_register(&block_class);
362 bdev_map = kobj_map_init(base_probe, &block_class_lock);
363 blk_dev_init();
365 #ifndef CONFIG_SYSFS_DEPRECATED
366 /* create top-level block dir */
367 block_depr = kobject_create_and_add("block", NULL);
368 #endif
369 return 0;
372 subsys_initcall(genhd_device_init);
374 static ssize_t disk_range_show(struct device *dev,
375 struct device_attribute *attr, char *buf)
377 struct gendisk *disk = dev_to_disk(dev);
379 return sprintf(buf, "%d\n", disk->minors);
382 static ssize_t disk_removable_show(struct device *dev,
383 struct device_attribute *attr, char *buf)
385 struct gendisk *disk = dev_to_disk(dev);
387 return sprintf(buf, "%d\n",
388 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
391 static ssize_t disk_size_show(struct device *dev,
392 struct device_attribute *attr, char *buf)
394 struct gendisk *disk = dev_to_disk(dev);
396 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
399 static ssize_t disk_capability_show(struct device *dev,
400 struct device_attribute *attr, char *buf)
402 struct gendisk *disk = dev_to_disk(dev);
404 return sprintf(buf, "%x\n", disk->flags);
407 static ssize_t disk_stat_show(struct device *dev,
408 struct device_attribute *attr, char *buf)
410 struct gendisk *disk = dev_to_disk(dev);
412 preempt_disable();
413 disk_round_stats(disk);
414 preempt_enable();
415 return sprintf(buf,
416 "%8lu %8lu %8llu %8u "
417 "%8lu %8lu %8llu %8u "
418 "%8u %8u %8u"
419 "\n",
420 disk_stat_read(disk, ios[READ]),
421 disk_stat_read(disk, merges[READ]),
422 (unsigned long long)disk_stat_read(disk, sectors[READ]),
423 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
424 disk_stat_read(disk, ios[WRITE]),
425 disk_stat_read(disk, merges[WRITE]),
426 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
427 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
428 disk->in_flight,
429 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
430 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
433 #ifdef CONFIG_FAIL_MAKE_REQUEST
434 static ssize_t disk_fail_show(struct device *dev,
435 struct device_attribute *attr, char *buf)
437 struct gendisk *disk = dev_to_disk(dev);
439 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
442 static ssize_t disk_fail_store(struct device *dev,
443 struct device_attribute *attr,
444 const char *buf, size_t count)
446 struct gendisk *disk = dev_to_disk(dev);
447 int i;
449 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
450 if (i == 0)
451 disk->flags &= ~GENHD_FL_FAIL;
452 else
453 disk->flags |= GENHD_FL_FAIL;
456 return count;
459 #endif
461 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
462 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
463 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
464 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
465 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
466 #ifdef CONFIG_FAIL_MAKE_REQUEST
467 static struct device_attribute dev_attr_fail =
468 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
469 #endif
471 static struct attribute *disk_attrs[] = {
472 &dev_attr_range.attr,
473 &dev_attr_removable.attr,
474 &dev_attr_size.attr,
475 &dev_attr_capability.attr,
476 &dev_attr_stat.attr,
477 #ifdef CONFIG_FAIL_MAKE_REQUEST
478 &dev_attr_fail.attr,
479 #endif
480 NULL
483 static struct attribute_group disk_attr_group = {
484 .attrs = disk_attrs,
487 static struct attribute_group *disk_attr_groups[] = {
488 &disk_attr_group,
489 NULL
492 static void disk_release(struct device *dev)
494 struct gendisk *disk = dev_to_disk(dev);
496 kfree(disk->random);
497 kfree(disk->part);
498 free_disk_stats(disk);
499 kfree(disk);
501 struct class block_class = {
502 .name = "block",
505 struct device_type disk_type = {
506 .name = "disk",
507 .groups = disk_attr_groups,
508 .release = disk_release,
512 * aggregate disk stat collector. Uses the same stats that the sysfs
513 * entries do, above, but makes them available through one seq_file.
515 * The output looks suspiciously like /proc/partitions with a bunch of
516 * extra fields.
519 static void *diskstats_start(struct seq_file *part, loff_t *pos)
521 loff_t k = *pos;
522 struct device *dev;
524 mutex_lock(&block_class_lock);
525 list_for_each_entry(dev, &block_class.devices, node) {
526 if (dev->type != &disk_type)
527 continue;
528 if (!k--)
529 return dev_to_disk(dev);
531 return NULL;
534 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
536 struct gendisk *gp = v;
537 struct device *dev;
539 ++*pos;
540 list_for_each_entry(dev, &gp->dev.node, node) {
541 if (&dev->node == &block_class.devices)
542 return NULL;
543 if (dev->type == &disk_type)
544 return dev_to_disk(dev);
546 return NULL;
549 static void diskstats_stop(struct seq_file *part, void *v)
551 mutex_unlock(&block_class_lock);
554 static int diskstats_show(struct seq_file *s, void *v)
556 struct gendisk *gp = v;
557 char buf[BDEVNAME_SIZE];
558 int n = 0;
561 if (&gp->dev.kobj.entry == block_class.devices.next)
562 seq_puts(s, "major minor name"
563 " rio rmerge rsect ruse wio wmerge "
564 "wsect wuse running use aveq"
565 "\n\n");
568 preempt_disable();
569 disk_round_stats(gp);
570 preempt_enable();
571 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
572 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
573 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
574 (unsigned long long)disk_stat_read(gp, sectors[0]),
575 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
576 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
577 (unsigned long long)disk_stat_read(gp, sectors[1]),
578 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
579 gp->in_flight,
580 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
581 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
583 /* now show all non-0 size partitions of it */
584 for (n = 0; n < gp->minors - 1; n++) {
585 struct hd_struct *hd = gp->part[n];
587 if (hd && hd->nr_sects)
588 seq_printf(s, "%4d %4d %s %u %u %u %u\n",
589 gp->major, n + gp->first_minor + 1,
590 disk_name(gp, n + 1, buf),
591 hd->ios[0], hd->sectors[0],
592 hd->ios[1], hd->sectors[1]);
595 return 0;
598 const struct seq_operations diskstats_op = {
599 .start = diskstats_start,
600 .next = diskstats_next,
601 .stop = diskstats_stop,
602 .show = diskstats_show
605 static void media_change_notify_thread(struct work_struct *work)
607 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
608 char event[] = "MEDIA_CHANGE=1";
609 char *envp[] = { event, NULL };
612 * set enviroment vars to indicate which event this is for
613 * so that user space will know to go check the media status.
615 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
616 put_device(gd->driverfs_dev);
619 void genhd_media_change_notify(struct gendisk *disk)
621 get_device(disk->driverfs_dev);
622 schedule_work(&disk->async_notify);
624 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
626 dev_t blk_lookup_devt(const char *name)
628 struct device *dev;
629 dev_t devt = MKDEV(0, 0);
631 mutex_lock(&block_class_lock);
632 list_for_each_entry(dev, &block_class.devices, node) {
633 if (strcmp(dev->bus_id, name) == 0) {
634 devt = dev->devt;
635 break;
638 mutex_unlock(&block_class_lock);
640 return devt;
643 EXPORT_SYMBOL(blk_lookup_devt);
645 struct gendisk *alloc_disk(int minors)
647 return alloc_disk_node(minors, -1);
650 struct gendisk *alloc_disk_node(int minors, int node_id)
652 struct gendisk *disk;
654 disk = kmalloc_node(sizeof(struct gendisk),
655 GFP_KERNEL | __GFP_ZERO, node_id);
656 if (disk) {
657 if (!init_disk_stats(disk)) {
658 kfree(disk);
659 return NULL;
661 if (minors > 1) {
662 int size = (minors - 1) * sizeof(struct hd_struct *);
663 disk->part = kmalloc_node(size,
664 GFP_KERNEL | __GFP_ZERO, node_id);
665 if (!disk->part) {
666 free_disk_stats(disk);
667 kfree(disk);
668 return NULL;
671 disk->minors = minors;
672 rand_initialize_disk(disk);
673 disk->dev.class = &block_class;
674 disk->dev.type = &disk_type;
675 device_initialize(&disk->dev);
676 INIT_WORK(&disk->async_notify,
677 media_change_notify_thread);
679 return disk;
682 EXPORT_SYMBOL(alloc_disk);
683 EXPORT_SYMBOL(alloc_disk_node);
685 struct kobject *get_disk(struct gendisk *disk)
687 struct module *owner;
688 struct kobject *kobj;
690 if (!disk->fops)
691 return NULL;
692 owner = disk->fops->owner;
693 if (owner && !try_module_get(owner))
694 return NULL;
695 kobj = kobject_get(&disk->dev.kobj);
696 if (kobj == NULL) {
697 module_put(owner);
698 return NULL;
700 return kobj;
704 EXPORT_SYMBOL(get_disk);
706 void put_disk(struct gendisk *disk)
708 if (disk)
709 kobject_put(&disk->dev.kobj);
712 EXPORT_SYMBOL(put_disk);
714 void set_device_ro(struct block_device *bdev, int flag)
716 if (bdev->bd_contains != bdev)
717 bdev->bd_part->policy = flag;
718 else
719 bdev->bd_disk->policy = flag;
722 EXPORT_SYMBOL(set_device_ro);
724 void set_disk_ro(struct gendisk *disk, int flag)
726 int i;
727 disk->policy = flag;
728 for (i = 0; i < disk->minors - 1; i++)
729 if (disk->part[i]) disk->part[i]->policy = flag;
732 EXPORT_SYMBOL(set_disk_ro);
734 int bdev_read_only(struct block_device *bdev)
736 if (!bdev)
737 return 0;
738 else if (bdev->bd_contains != bdev)
739 return bdev->bd_part->policy;
740 else
741 return bdev->bd_disk->policy;
744 EXPORT_SYMBOL(bdev_read_only);
746 int invalidate_partition(struct gendisk *disk, int index)
748 int res = 0;
749 struct block_device *bdev = bdget_disk(disk, index);
750 if (bdev) {
751 fsync_bdev(bdev);
752 res = __invalidate_device(bdev);
753 bdput(bdev);
755 return res;
758 EXPORT_SYMBOL(invalidate_partition);