x86: use apic_*_around instead of apic_write in x86_64
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / block / genhd.c
blob00da5219ee37c02a066d45e6ceefa65d9606d5e6
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 disk->flags |= GENHD_FL_UP;
186 blk_register_region(MKDEV(disk->major, disk->first_minor),
187 disk->minors, NULL, exact_match, exact_lock, disk);
188 register_disk(disk);
189 blk_register_queue(disk);
192 EXPORT_SYMBOL(add_disk);
193 EXPORT_SYMBOL(del_gendisk); /* in partitions/check.c */
195 void unlink_gendisk(struct gendisk *disk)
197 blk_unregister_queue(disk);
198 blk_unregister_region(MKDEV(disk->major, disk->first_minor),
199 disk->minors);
203 * get_gendisk - get partitioning information for a given device
204 * @dev: device to get partitioning information for
206 * This function gets the structure containing partitioning
207 * information for the given device @dev.
209 struct gendisk *get_gendisk(dev_t devt, int *part)
211 struct kobject *kobj = kobj_lookup(bdev_map, devt, part);
212 struct device *dev = kobj_to_dev(kobj);
214 return kobj ? dev_to_disk(dev) : NULL;
218 * print a full list of all partitions - intended for places where the root
219 * filesystem can't be mounted and thus to give the victim some idea of what
220 * went wrong
222 void __init printk_all_partitions(void)
224 struct device *dev;
225 struct gendisk *sgp;
226 char buf[BDEVNAME_SIZE];
227 int n;
229 mutex_lock(&block_class_lock);
230 /* For each block device... */
231 list_for_each_entry(dev, &block_class.devices, node) {
232 if (dev->type != &disk_type)
233 continue;
234 sgp = dev_to_disk(dev);
236 * Don't show empty devices or things that have been surpressed
238 if (get_capacity(sgp) == 0 ||
239 (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
240 continue;
243 * Note, unlike /proc/partitions, I am showing the numbers in
244 * hex - the same format as the root= option takes.
246 printk("%02x%02x %10llu %s",
247 sgp->major, sgp->first_minor,
248 (unsigned long long)get_capacity(sgp) >> 1,
249 disk_name(sgp, 0, buf));
250 if (sgp->driverfs_dev != NULL &&
251 sgp->driverfs_dev->driver != NULL)
252 printk(" driver: %s\n",
253 sgp->driverfs_dev->driver->name);
254 else
255 printk(" (driver?)\n");
257 /* now show the partitions */
258 for (n = 0; n < sgp->minors - 1; ++n) {
259 if (sgp->part[n] == NULL)
260 continue;
261 if (sgp->part[n]->nr_sects == 0)
262 continue;
263 printk(" %02x%02x %10llu %s\n",
264 sgp->major, n + 1 + sgp->first_minor,
265 (unsigned long long)sgp->part[n]->nr_sects >> 1,
266 disk_name(sgp, n + 1, buf));
270 mutex_unlock(&block_class_lock);
273 #ifdef CONFIG_PROC_FS
274 /* iterator */
275 static void *part_start(struct seq_file *part, loff_t *pos)
277 loff_t k = *pos;
278 struct device *dev;
280 mutex_lock(&block_class_lock);
281 list_for_each_entry(dev, &block_class.devices, node) {
282 if (dev->type != &disk_type)
283 continue;
284 if (!k--)
285 return dev_to_disk(dev);
287 return NULL;
290 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
292 struct gendisk *gp = v;
293 struct device *dev;
294 ++*pos;
295 list_for_each_entry(dev, &gp->dev.node, node) {
296 if (&dev->node == &block_class.devices)
297 return NULL;
298 if (dev->type == &disk_type)
299 return dev_to_disk(dev);
301 return NULL;
304 static void part_stop(struct seq_file *part, void *v)
306 mutex_unlock(&block_class_lock);
309 static int show_partition(struct seq_file *part, void *v)
311 struct gendisk *sgp = v;
312 int n;
313 char buf[BDEVNAME_SIZE];
315 if (&sgp->dev.node == block_class.devices.next)
316 seq_puts(part, "major minor #blocks name\n\n");
318 /* Don't show non-partitionable removeable devices or empty devices */
319 if (!get_capacity(sgp) ||
320 (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
321 return 0;
322 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
323 return 0;
325 /* show the full disk and all non-0 size partitions of it */
326 seq_printf(part, "%4d %4d %10llu %s\n",
327 sgp->major, sgp->first_minor,
328 (unsigned long long)get_capacity(sgp) >> 1,
329 disk_name(sgp, 0, buf));
330 for (n = 0; n < sgp->minors - 1; n++) {
331 if (!sgp->part[n])
332 continue;
333 if (sgp->part[n]->nr_sects == 0)
334 continue;
335 seq_printf(part, "%4d %4d %10llu %s\n",
336 sgp->major, n + 1 + sgp->first_minor,
337 (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
338 disk_name(sgp, n + 1, buf));
341 return 0;
344 const struct seq_operations partitions_op = {
345 .start = part_start,
346 .next = part_next,
347 .stop = part_stop,
348 .show = show_partition
350 #endif
353 static struct kobject *base_probe(dev_t devt, int *part, void *data)
355 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
356 /* Make old-style 2.4 aliases work */
357 request_module("block-major-%d", MAJOR(devt));
358 return NULL;
361 static int __init genhd_device_init(void)
363 int error = class_register(&block_class);
364 if (unlikely(error))
365 return error;
366 bdev_map = kobj_map_init(base_probe, &block_class_lock);
367 blk_dev_init();
369 #ifndef CONFIG_SYSFS_DEPRECATED
370 /* create top-level block dir */
371 block_depr = kobject_create_and_add("block", NULL);
372 #endif
373 return 0;
376 subsys_initcall(genhd_device_init);
378 static ssize_t disk_range_show(struct device *dev,
379 struct device_attribute *attr, char *buf)
381 struct gendisk *disk = dev_to_disk(dev);
383 return sprintf(buf, "%d\n", disk->minors);
386 static ssize_t disk_removable_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",
392 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
395 static ssize_t disk_size_show(struct device *dev,
396 struct device_attribute *attr, char *buf)
398 struct gendisk *disk = dev_to_disk(dev);
400 return sprintf(buf, "%llu\n", (unsigned long long)get_capacity(disk));
403 static ssize_t disk_capability_show(struct device *dev,
404 struct device_attribute *attr, char *buf)
406 struct gendisk *disk = dev_to_disk(dev);
408 return sprintf(buf, "%x\n", disk->flags);
411 static ssize_t disk_stat_show(struct device *dev,
412 struct device_attribute *attr, char *buf)
414 struct gendisk *disk = dev_to_disk(dev);
416 preempt_disable();
417 disk_round_stats(disk);
418 preempt_enable();
419 return sprintf(buf,
420 "%8lu %8lu %8llu %8u "
421 "%8lu %8lu %8llu %8u "
422 "%8u %8u %8u"
423 "\n",
424 disk_stat_read(disk, ios[READ]),
425 disk_stat_read(disk, merges[READ]),
426 (unsigned long long)disk_stat_read(disk, sectors[READ]),
427 jiffies_to_msecs(disk_stat_read(disk, ticks[READ])),
428 disk_stat_read(disk, ios[WRITE]),
429 disk_stat_read(disk, merges[WRITE]),
430 (unsigned long long)disk_stat_read(disk, sectors[WRITE]),
431 jiffies_to_msecs(disk_stat_read(disk, ticks[WRITE])),
432 disk->in_flight,
433 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
434 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
437 #ifdef CONFIG_FAIL_MAKE_REQUEST
438 static ssize_t disk_fail_show(struct device *dev,
439 struct device_attribute *attr, char *buf)
441 struct gendisk *disk = dev_to_disk(dev);
443 return sprintf(buf, "%d\n", disk->flags & GENHD_FL_FAIL ? 1 : 0);
446 static ssize_t disk_fail_store(struct device *dev,
447 struct device_attribute *attr,
448 const char *buf, size_t count)
450 struct gendisk *disk = dev_to_disk(dev);
451 int i;
453 if (count > 0 && sscanf(buf, "%d", &i) > 0) {
454 if (i == 0)
455 disk->flags &= ~GENHD_FL_FAIL;
456 else
457 disk->flags |= GENHD_FL_FAIL;
460 return count;
463 #endif
465 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
466 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
467 static DEVICE_ATTR(size, S_IRUGO, disk_size_show, NULL);
468 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
469 static DEVICE_ATTR(stat, S_IRUGO, disk_stat_show, NULL);
470 #ifdef CONFIG_FAIL_MAKE_REQUEST
471 static struct device_attribute dev_attr_fail =
472 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, disk_fail_show, disk_fail_store);
473 #endif
475 static struct attribute *disk_attrs[] = {
476 &dev_attr_range.attr,
477 &dev_attr_removable.attr,
478 &dev_attr_size.attr,
479 &dev_attr_capability.attr,
480 &dev_attr_stat.attr,
481 #ifdef CONFIG_FAIL_MAKE_REQUEST
482 &dev_attr_fail.attr,
483 #endif
484 NULL
487 static struct attribute_group disk_attr_group = {
488 .attrs = disk_attrs,
491 static struct attribute_group *disk_attr_groups[] = {
492 &disk_attr_group,
493 NULL
496 static void disk_release(struct device *dev)
498 struct gendisk *disk = dev_to_disk(dev);
500 kfree(disk->random);
501 kfree(disk->part);
502 free_disk_stats(disk);
503 kfree(disk);
505 struct class block_class = {
506 .name = "block",
509 static struct device_type disk_type = {
510 .name = "disk",
511 .groups = disk_attr_groups,
512 .release = disk_release,
516 * aggregate disk stat collector. Uses the same stats that the sysfs
517 * entries do, above, but makes them available through one seq_file.
519 * The output looks suspiciously like /proc/partitions with a bunch of
520 * extra fields.
523 static void *diskstats_start(struct seq_file *part, loff_t *pos)
525 loff_t k = *pos;
526 struct device *dev;
528 mutex_lock(&block_class_lock);
529 list_for_each_entry(dev, &block_class.devices, node) {
530 if (dev->type != &disk_type)
531 continue;
532 if (!k--)
533 return dev_to_disk(dev);
535 return NULL;
538 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
540 struct gendisk *gp = v;
541 struct device *dev;
543 ++*pos;
544 list_for_each_entry(dev, &gp->dev.node, node) {
545 if (&dev->node == &block_class.devices)
546 return NULL;
547 if (dev->type == &disk_type)
548 return dev_to_disk(dev);
550 return NULL;
553 static void diskstats_stop(struct seq_file *part, void *v)
555 mutex_unlock(&block_class_lock);
558 static int diskstats_show(struct seq_file *s, void *v)
560 struct gendisk *gp = v;
561 char buf[BDEVNAME_SIZE];
562 int n = 0;
565 if (&gp->dev.kobj.entry == block_class.devices.next)
566 seq_puts(s, "major minor name"
567 " rio rmerge rsect ruse wio wmerge "
568 "wsect wuse running use aveq"
569 "\n\n");
572 preempt_disable();
573 disk_round_stats(gp);
574 preempt_enable();
575 seq_printf(s, "%4d %4d %s %lu %lu %llu %u %lu %lu %llu %u %u %u %u\n",
576 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
577 disk_stat_read(gp, ios[0]), disk_stat_read(gp, merges[0]),
578 (unsigned long long)disk_stat_read(gp, sectors[0]),
579 jiffies_to_msecs(disk_stat_read(gp, ticks[0])),
580 disk_stat_read(gp, ios[1]), disk_stat_read(gp, merges[1]),
581 (unsigned long long)disk_stat_read(gp, sectors[1]),
582 jiffies_to_msecs(disk_stat_read(gp, ticks[1])),
583 gp->in_flight,
584 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
585 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
587 /* now show all non-0 size partitions of it */
588 for (n = 0; n < gp->minors - 1; n++) {
589 struct hd_struct *hd = gp->part[n];
591 if (!hd || !hd->nr_sects)
592 continue;
594 preempt_disable();
595 part_round_stats(hd);
596 preempt_enable();
597 seq_printf(s, "%4d %4d %s %lu %lu %llu "
598 "%u %lu %lu %llu %u %u %u %u\n",
599 gp->major, n + gp->first_minor + 1,
600 disk_name(gp, n + 1, buf),
601 part_stat_read(hd, ios[0]),
602 part_stat_read(hd, merges[0]),
603 (unsigned long long)part_stat_read(hd, sectors[0]),
604 jiffies_to_msecs(part_stat_read(hd, ticks[0])),
605 part_stat_read(hd, ios[1]),
606 part_stat_read(hd, merges[1]),
607 (unsigned long long)part_stat_read(hd, sectors[1]),
608 jiffies_to_msecs(part_stat_read(hd, ticks[1])),
609 hd->in_flight,
610 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
611 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
615 return 0;
618 const struct seq_operations diskstats_op = {
619 .start = diskstats_start,
620 .next = diskstats_next,
621 .stop = diskstats_stop,
622 .show = diskstats_show
625 static void media_change_notify_thread(struct work_struct *work)
627 struct gendisk *gd = container_of(work, struct gendisk, async_notify);
628 char event[] = "MEDIA_CHANGE=1";
629 char *envp[] = { event, NULL };
632 * set enviroment vars to indicate which event this is for
633 * so that user space will know to go check the media status.
635 kobject_uevent_env(&gd->dev.kobj, KOBJ_CHANGE, envp);
636 put_device(gd->driverfs_dev);
639 #if 0
640 void genhd_media_change_notify(struct gendisk *disk)
642 get_device(disk->driverfs_dev);
643 schedule_work(&disk->async_notify);
645 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
646 #endif /* 0 */
648 dev_t blk_lookup_devt(const char *name)
650 struct device *dev;
651 dev_t devt = MKDEV(0, 0);
653 mutex_lock(&block_class_lock);
654 list_for_each_entry(dev, &block_class.devices, node) {
655 if (strcmp(dev->bus_id, name) == 0) {
656 devt = dev->devt;
657 break;
660 mutex_unlock(&block_class_lock);
662 return devt;
665 EXPORT_SYMBOL(blk_lookup_devt);
667 struct gendisk *alloc_disk(int minors)
669 return alloc_disk_node(minors, -1);
672 struct gendisk *alloc_disk_node(int minors, int node_id)
674 struct gendisk *disk;
676 disk = kmalloc_node(sizeof(struct gendisk),
677 GFP_KERNEL | __GFP_ZERO, node_id);
678 if (disk) {
679 if (!init_disk_stats(disk)) {
680 kfree(disk);
681 return NULL;
683 if (minors > 1) {
684 int size = (minors - 1) * sizeof(struct hd_struct *);
685 disk->part = kmalloc_node(size,
686 GFP_KERNEL | __GFP_ZERO, node_id);
687 if (!disk->part) {
688 free_disk_stats(disk);
689 kfree(disk);
690 return NULL;
693 disk->minors = minors;
694 rand_initialize_disk(disk);
695 disk->dev.class = &block_class;
696 disk->dev.type = &disk_type;
697 device_initialize(&disk->dev);
698 INIT_WORK(&disk->async_notify,
699 media_change_notify_thread);
701 return disk;
704 EXPORT_SYMBOL(alloc_disk);
705 EXPORT_SYMBOL(alloc_disk_node);
707 struct kobject *get_disk(struct gendisk *disk)
709 struct module *owner;
710 struct kobject *kobj;
712 if (!disk->fops)
713 return NULL;
714 owner = disk->fops->owner;
715 if (owner && !try_module_get(owner))
716 return NULL;
717 kobj = kobject_get(&disk->dev.kobj);
718 if (kobj == NULL) {
719 module_put(owner);
720 return NULL;
722 return kobj;
726 EXPORT_SYMBOL(get_disk);
728 void put_disk(struct gendisk *disk)
730 if (disk)
731 kobject_put(&disk->dev.kobj);
734 EXPORT_SYMBOL(put_disk);
736 void set_device_ro(struct block_device *bdev, int flag)
738 if (bdev->bd_contains != bdev)
739 bdev->bd_part->policy = flag;
740 else
741 bdev->bd_disk->policy = flag;
744 EXPORT_SYMBOL(set_device_ro);
746 void set_disk_ro(struct gendisk *disk, int flag)
748 int i;
749 disk->policy = flag;
750 for (i = 0; i < disk->minors - 1; i++)
751 if (disk->part[i]) disk->part[i]->policy = flag;
754 EXPORT_SYMBOL(set_disk_ro);
756 int bdev_read_only(struct block_device *bdev)
758 if (!bdev)
759 return 0;
760 else if (bdev->bd_contains != bdev)
761 return bdev->bd_part->policy;
762 else
763 return bdev->bd_disk->policy;
766 EXPORT_SYMBOL(bdev_read_only);
768 int invalidate_partition(struct gendisk *disk, int index)
770 int res = 0;
771 struct block_device *bdev = bdget_disk(disk, index);
772 if (bdev) {
773 fsync_bdev(bdev);
774 res = __invalidate_device(bdev);
775 bdput(bdev);
777 return res;
780 EXPORT_SYMBOL(invalidate_partition);