]>
Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * gendisk handling | |
7b51e703 CH |
4 | * |
5 | * Portions Copyright (C) 2020 Christoph Hellwig | |
1da177e4 LT |
6 | */ |
7 | ||
1da177e4 | 8 | #include <linux/module.h> |
3ad5cee5 | 9 | #include <linux/ctype.h> |
1da177e4 LT |
10 | #include <linux/fs.h> |
11 | #include <linux/genhd.h> | |
b446b60e | 12 | #include <linux/kdev_t.h> |
1da177e4 LT |
13 | #include <linux/kernel.h> |
14 | #include <linux/blkdev.h> | |
66114cad | 15 | #include <linux/backing-dev.h> |
1da177e4 LT |
16 | #include <linux/init.h> |
17 | #include <linux/spinlock.h> | |
f500975a | 18 | #include <linux/proc_fs.h> |
1da177e4 LT |
19 | #include <linux/seq_file.h> |
20 | #include <linux/slab.h> | |
21 | #include <linux/kmod.h> | |
a9f215d7 | 22 | #include <linux/major.h> |
58383af6 | 23 | #include <linux/mutex.h> |
bcce3de1 | 24 | #include <linux/idr.h> |
77ea887e | 25 | #include <linux/log2.h> |
25e823c8 | 26 | #include <linux/pm_runtime.h> |
99e6608c | 27 | #include <linux/badblocks.h> |
1da177e4 | 28 | |
ff88972c | 29 | #include "blk.h" |
8e141f9e | 30 | #include "blk-rq-qos.h" |
ff88972c | 31 | |
31eb6186 | 32 | static struct kobject *block_depr; |
1da177e4 | 33 | |
cf179948 MC |
34 | /* |
35 | * Unique, monotonically increasing sequential number associated with block | |
36 | * devices instances (i.e. incremented each time a device is attached). | |
37 | * Associating uevents with block devices in userspace is difficult and racy: | |
38 | * the uevent netlink socket is lossy, and on slow and overloaded systems has | |
39 | * a very high latency. | |
40 | * Block devices do not have exclusive owners in userspace, any process can set | |
41 | * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0 | |
42 | * can be reused again and again). | |
43 | * A userspace process setting up a block device and watching for its events | |
44 | * cannot thus reliably tell whether an event relates to the device it just set | |
45 | * up or another earlier instance with the same name. | |
46 | * This sequential number allows userspace processes to solve this problem, and | |
47 | * uniquely associate an uevent to the lifetime to a device. | |
48 | */ | |
49 | static atomic64_t diskseq; | |
50 | ||
bcce3de1 | 51 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 52 | #define NR_EXT_DEVT (1 << MINORBITS) |
22ae8ce8 | 53 | static DEFINE_IDA(ext_devt_ida); |
bcce3de1 | 54 | |
a782483c CH |
55 | void set_capacity(struct gendisk *disk, sector_t sectors) |
56 | { | |
cb8432d6 | 57 | struct block_device *bdev = disk->part0; |
a782483c | 58 | |
0f472277 | 59 | spin_lock(&bdev->bd_size_lock); |
a782483c | 60 | i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); |
0f472277 | 61 | spin_unlock(&bdev->bd_size_lock); |
a782483c CH |
62 | } |
63 | EXPORT_SYMBOL(set_capacity); | |
64 | ||
e598a72f | 65 | /* |
449f4ec9 CH |
66 | * Set disk capacity and notify if the size is not currently zero and will not |
67 | * be set to zero. Returns true if a uevent was sent, otherwise false. | |
e598a72f | 68 | */ |
449f4ec9 | 69 | bool set_capacity_and_notify(struct gendisk *disk, sector_t size) |
e598a72f BS |
70 | { |
71 | sector_t capacity = get_capacity(disk); | |
a782483c | 72 | char *envp[] = { "RESIZE=1", NULL }; |
e598a72f BS |
73 | |
74 | set_capacity(disk, size); | |
e598a72f | 75 | |
a782483c CH |
76 | /* |
77 | * Only print a message and send a uevent if the gendisk is user visible | |
78 | * and alive. This avoids spamming the log and udev when setting the | |
79 | * initial capacity during probing. | |
80 | */ | |
81 | if (size == capacity || | |
50b4aecf CH |
82 | !disk_live(disk) || |
83 | (disk->flags & GENHD_FL_HIDDEN)) | |
a782483c | 84 | return false; |
e598a72f | 85 | |
a782483c | 86 | pr_info("%s: detected capacity change from %lld to %lld\n", |
452c0bf8 | 87 | disk->disk_name, capacity, size); |
7e890c37 | 88 | |
a782483c CH |
89 | /* |
90 | * Historically we did not send a uevent for changes to/from an empty | |
91 | * device. | |
92 | */ | |
93 | if (!capacity || !size) | |
94 | return false; | |
95 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
96 | return true; | |
e598a72f | 97 | } |
449f4ec9 | 98 | EXPORT_SYMBOL_GPL(set_capacity_and_notify); |
e598a72f | 99 | |
5cbd28e3 | 100 | /* |
abd2864a CH |
101 | * Format the device name of the indicated block device into the supplied buffer |
102 | * and return a pointer to that same buffer for convenience. | |
103 | * | |
104 | * Note: do not use this in new code, use the %pg specifier to sprintf and | |
105 | * printk insted. | |
5cbd28e3 | 106 | */ |
abd2864a | 107 | const char *bdevname(struct block_device *bdev, char *buf) |
5cbd28e3 | 108 | { |
abd2864a CH |
109 | struct gendisk *hd = bdev->bd_disk; |
110 | int partno = bdev->bd_partno; | |
111 | ||
5cbd28e3 CH |
112 | if (!partno) |
113 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
114 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
115 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
116 | else | |
117 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
118 | ||
119 | return buf; | |
120 | } | |
5cbd28e3 | 121 | EXPORT_SYMBOL(bdevname); |
e598a72f | 122 | |
0d02129e CH |
123 | static void part_stat_read_all(struct block_device *part, |
124 | struct disk_stats *stat) | |
ea18e0f0 KK |
125 | { |
126 | int cpu; | |
127 | ||
128 | memset(stat, 0, sizeof(struct disk_stats)); | |
129 | for_each_possible_cpu(cpu) { | |
0d02129e | 130 | struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu); |
ea18e0f0 KK |
131 | int group; |
132 | ||
133 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
134 | stat->nsecs[group] += ptr->nsecs[group]; | |
135 | stat->sectors[group] += ptr->sectors[group]; | |
136 | stat->ios[group] += ptr->ios[group]; | |
137 | stat->merges[group] += ptr->merges[group]; | |
138 | } | |
139 | ||
140 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
141 | } |
142 | } | |
ea18e0f0 | 143 | |
8446fe92 | 144 | static unsigned int part_in_flight(struct block_device *part) |
f299b7c7 | 145 | { |
b2f609e1 | 146 | unsigned int inflight = 0; |
1226b8dd | 147 | int cpu; |
f299b7c7 | 148 | |
1226b8dd | 149 | for_each_possible_cpu(cpu) { |
e016b782 MP |
150 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
151 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 152 | } |
e016b782 MP |
153 | if ((int)inflight < 0) |
154 | inflight = 0; | |
1226b8dd | 155 | |
e016b782 | 156 | return inflight; |
f299b7c7 JA |
157 | } |
158 | ||
8446fe92 CH |
159 | static void part_in_flight_rw(struct block_device *part, |
160 | unsigned int inflight[2]) | |
bf0ddaba | 161 | { |
1226b8dd MP |
162 | int cpu; |
163 | ||
1226b8dd MP |
164 | inflight[0] = 0; |
165 | inflight[1] = 0; | |
166 | for_each_possible_cpu(cpu) { | |
167 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
168 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
169 | } | |
170 | if ((int)inflight[0] < 0) | |
171 | inflight[0] = 0; | |
172 | if ((int)inflight[1] < 0) | |
173 | inflight[1] = 0; | |
bf0ddaba OS |
174 | } |
175 | ||
1da177e4 LT |
176 | /* |
177 | * Can be deleted altogether. Later. | |
178 | * | |
179 | */ | |
133d55cd | 180 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
181 | static struct blk_major_name { |
182 | struct blk_major_name *next; | |
183 | int major; | |
184 | char name[16]; | |
a160c615 | 185 | void (*probe)(dev_t devt); |
68eef3b4 | 186 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
e49fbbbf | 187 | static DEFINE_MUTEX(major_names_lock); |
dfbb3409 | 188 | static DEFINE_SPINLOCK(major_names_spinlock); |
1da177e4 LT |
189 | |
190 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 191 | static inline int major_to_index(unsigned major) |
1da177e4 | 192 | { |
68eef3b4 | 193 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
194 | } |
195 | ||
68eef3b4 | 196 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 197 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 198 | { |
68eef3b4 | 199 | struct blk_major_name *dp; |
7170be5f | 200 | |
dfbb3409 | 201 | spin_lock(&major_names_spinlock); |
133d55cd LG |
202 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) |
203 | if (dp->major == offset) | |
cf771cb5 | 204 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
dfbb3409 | 205 | spin_unlock(&major_names_spinlock); |
1da177e4 | 206 | } |
68eef3b4 | 207 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 208 | |
9e8c0bcc | 209 | /** |
e2b6b301 | 210 | * __register_blkdev - register a new block device |
9e8c0bcc | 211 | * |
f33ff110 SB |
212 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
213 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc | 214 | * @name: the name of the new block device as a zero terminated string |
e2b6b301 | 215 | * @probe: allback that is called on access to any minor number of @major |
9e8c0bcc MN |
216 | * |
217 | * The @name must be unique within the system. | |
218 | * | |
0e056eb5 MCC |
219 | * The return value depends on the @major input parameter: |
220 | * | |
f33ff110 SB |
221 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
222 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 223 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 224 | * then the return value is the allocated major number in range |
f33ff110 SB |
225 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
226 | * | |
227 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
228 | * major numbers. | |
e2b6b301 CH |
229 | * |
230 | * Use register_blkdev instead for any new code. | |
9e8c0bcc | 231 | */ |
a160c615 CH |
232 | int __register_blkdev(unsigned int major, const char *name, |
233 | void (*probe)(dev_t devt)) | |
1da177e4 LT |
234 | { |
235 | struct blk_major_name **n, *p; | |
236 | int index, ret = 0; | |
237 | ||
e49fbbbf | 238 | mutex_lock(&major_names_lock); |
1da177e4 LT |
239 | |
240 | /* temporary */ | |
241 | if (major == 0) { | |
242 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
243 | if (major_names[index] == NULL) | |
244 | break; | |
245 | } | |
246 | ||
247 | if (index == 0) { | |
dfc76d11 KP |
248 | printk("%s: failed to get major for %s\n", |
249 | __func__, name); | |
1da177e4 LT |
250 | ret = -EBUSY; |
251 | goto out; | |
252 | } | |
253 | major = index; | |
254 | ret = major; | |
255 | } | |
256 | ||
133d55cd | 257 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
258 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
259 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
260 | |
261 | ret = -EINVAL; | |
262 | goto out; | |
263 | } | |
264 | ||
1da177e4 LT |
265 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
266 | if (p == NULL) { | |
267 | ret = -ENOMEM; | |
268 | goto out; | |
269 | } | |
270 | ||
271 | p->major = major; | |
a160c615 | 272 | p->probe = probe; |
1da177e4 LT |
273 | strlcpy(p->name, name, sizeof(p->name)); |
274 | p->next = NULL; | |
275 | index = major_to_index(major); | |
276 | ||
dfbb3409 | 277 | spin_lock(&major_names_spinlock); |
1da177e4 LT |
278 | for (n = &major_names[index]; *n; n = &(*n)->next) { |
279 | if ((*n)->major == major) | |
280 | break; | |
281 | } | |
282 | if (!*n) | |
283 | *n = p; | |
284 | else | |
285 | ret = -EBUSY; | |
dfbb3409 | 286 | spin_unlock(&major_names_spinlock); |
1da177e4 LT |
287 | |
288 | if (ret < 0) { | |
f33ff110 | 289 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
290 | major, name); |
291 | kfree(p); | |
292 | } | |
293 | out: | |
e49fbbbf | 294 | mutex_unlock(&major_names_lock); |
1da177e4 LT |
295 | return ret; |
296 | } | |
a160c615 | 297 | EXPORT_SYMBOL(__register_blkdev); |
1da177e4 | 298 | |
f4480240 | 299 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
300 | { |
301 | struct blk_major_name **n; | |
302 | struct blk_major_name *p = NULL; | |
303 | int index = major_to_index(major); | |
1da177e4 | 304 | |
e49fbbbf | 305 | mutex_lock(&major_names_lock); |
dfbb3409 | 306 | spin_lock(&major_names_spinlock); |
1da177e4 LT |
307 | for (n = &major_names[index]; *n; n = &(*n)->next) |
308 | if ((*n)->major == major) | |
309 | break; | |
294462a5 AM |
310 | if (!*n || strcmp((*n)->name, name)) { |
311 | WARN_ON(1); | |
294462a5 | 312 | } else { |
1da177e4 LT |
313 | p = *n; |
314 | *n = p->next; | |
315 | } | |
dfbb3409 | 316 | spin_unlock(&major_names_spinlock); |
e49fbbbf | 317 | mutex_unlock(&major_names_lock); |
1da177e4 | 318 | kfree(p); |
1da177e4 LT |
319 | } |
320 | ||
321 | EXPORT_SYMBOL(unregister_blkdev); | |
322 | ||
7c3f828b | 323 | int blk_alloc_ext_minor(void) |
bcce3de1 | 324 | { |
bab998d6 | 325 | int idx; |
bcce3de1 | 326 | |
f4980e91 | 327 | idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL); |
c4b2b7d1 CH |
328 | if (idx == -ENOSPC) |
329 | return -EBUSY; | |
330 | return idx; | |
bcce3de1 TH |
331 | } |
332 | ||
7c3f828b | 333 | void blk_free_ext_minor(unsigned int minor) |
bcce3de1 | 334 | { |
c4b2b7d1 | 335 | ida_free(&ext_devt_ida, minor); |
6fcc44d1 YY |
336 | } |
337 | ||
1f014290 TH |
338 | static char *bdevt_str(dev_t devt, char *buf) |
339 | { | |
340 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
341 | char tbuf[BDEVT_SIZE]; | |
342 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
343 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
344 | } else | |
345 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
346 | ||
347 | return buf; | |
348 | } | |
349 | ||
bc359d03 CH |
350 | void disk_uevent(struct gendisk *disk, enum kobject_action action) |
351 | { | |
bc359d03 | 352 | struct block_device *part; |
3212135a | 353 | unsigned long idx; |
bc359d03 | 354 | |
3212135a CH |
355 | rcu_read_lock(); |
356 | xa_for_each(&disk->part_tbl, idx, part) { | |
357 | if (bdev_is_partition(part) && !bdev_nr_sectors(part)) | |
358 | continue; | |
498dcc13 | 359 | if (!kobject_get_unless_zero(&part->bd_device.kobj)) |
3212135a CH |
360 | continue; |
361 | ||
362 | rcu_read_unlock(); | |
bc359d03 | 363 | kobject_uevent(bdev_kobj(part), action); |
498dcc13 | 364 | put_device(&part->bd_device); |
3212135a CH |
365 | rcu_read_lock(); |
366 | } | |
367 | rcu_read_unlock(); | |
bc359d03 CH |
368 | } |
369 | EXPORT_SYMBOL_GPL(disk_uevent); | |
370 | ||
9301fe73 CH |
371 | static void disk_scan_partitions(struct gendisk *disk) |
372 | { | |
373 | struct block_device *bdev; | |
374 | ||
375 | if (!get_capacity(disk) || !disk_part_scan_enabled(disk)) | |
376 | return; | |
377 | ||
378 | set_bit(GD_NEED_PART_SCAN, &disk->state); | |
379 | bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL); | |
380 | if (!IS_ERR(bdev)) | |
381 | blkdev_put(bdev, FMODE_READ); | |
382 | } | |
383 | ||
1da177e4 | 384 | /** |
d1254a87 | 385 | * device_add_disk - add disk information to kernel list |
e63a46be | 386 | * @parent: parent device for the disk |
1da177e4 | 387 | * @disk: per-device partitioning information |
fef912bf | 388 | * @groups: Additional per-device sysfs groups |
1da177e4 LT |
389 | * |
390 | * This function registers the partitioning information in @disk | |
391 | * with the kernel. | |
392 | */ | |
83cbce95 | 393 | int device_add_disk(struct device *parent, struct gendisk *disk, |
d1254a87 CH |
394 | const struct attribute_group **groups) |
395 | ||
1da177e4 | 396 | { |
52b85909 | 397 | struct device *ddev = disk_to_dev(disk); |
7c3f828b | 398 | int ret; |
cf0ca9fe | 399 | |
737eb78e DLM |
400 | /* |
401 | * The disk queue should now be all set with enough information about | |
402 | * the device for the elevator code to pick an adequate default | |
403 | * elevator if one is needed, that is, for devices requesting queue | |
404 | * registration. | |
405 | */ | |
d1254a87 | 406 | elevator_init_mq(disk->queue); |
737eb78e | 407 | |
7c3f828b CH |
408 | /* |
409 | * If the driver provides an explicit major number it also must provide | |
410 | * the number of minors numbers supported, and those will be used to | |
411 | * setup the gendisk. | |
412 | * Otherwise just allocate the device numbers for both the whole device | |
413 | * and all partitions from the extended dev_t space. | |
3e1a7ff8 | 414 | */ |
7c3f828b | 415 | if (disk->major) { |
83cbce95 LC |
416 | if (WARN_ON(!disk->minors)) |
417 | return -EINVAL; | |
2e3c73fa CH |
418 | |
419 | if (disk->minors > DISK_MAX_PARTS) { | |
420 | pr_err("block: can't allocate more than %d partitions\n", | |
421 | DISK_MAX_PARTS); | |
422 | disk->minors = DISK_MAX_PARTS; | |
423 | } | |
90afd1e3 TH |
424 | if (disk->first_minor + disk->minors > MINORMASK + 1) |
425 | return -EINVAL; | |
7c3f828b | 426 | } else { |
83cbce95 LC |
427 | if (WARN_ON(disk->minors)) |
428 | return -EINVAL; | |
3e1a7ff8 | 429 | |
7c3f828b | 430 | ret = blk_alloc_ext_minor(); |
83cbce95 LC |
431 | if (ret < 0) |
432 | return ret; | |
7c3f828b | 433 | disk->major = BLOCK_EXT_MAJOR; |
539711d7 | 434 | disk->first_minor = ret; |
0d1feb72 | 435 | disk->flags |= GENHD_FL_EXT_DEVT; |
3e1a7ff8 | 436 | } |
7c3f828b | 437 | |
52b85909 CH |
438 | /* delay uevents, until we scanned partition table */ |
439 | dev_set_uevent_suppress(ddev, 1); | |
440 | ||
441 | ddev->parent = parent; | |
442 | ddev->groups = groups; | |
443 | dev_set_name(ddev, "%s", disk->disk_name); | |
8235b5c1 CH |
444 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
445 | ddev->devt = MKDEV(disk->major, disk->first_minor); | |
83cbce95 LC |
446 | ret = device_add(ddev); |
447 | if (ret) | |
35bda087 CH |
448 | goto out_free_ext_minor; |
449 | ||
450 | ret = disk_alloc_events(disk); | |
451 | if (ret) | |
452 | goto out_device_del; | |
453 | ||
52b85909 CH |
454 | if (!sysfs_deprecated) { |
455 | ret = sysfs_create_link(block_depr, &ddev->kobj, | |
456 | kobject_name(&ddev->kobj)); | |
83cbce95 LC |
457 | if (ret) |
458 | goto out_device_del; | |
52b85909 CH |
459 | } |
460 | ||
461 | /* | |
462 | * avoid probable deadlock caused by allocating memory with | |
463 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
464 | * devices | |
465 | */ | |
466 | pm_runtime_set_memalloc_noio(ddev, true); | |
467 | ||
83cbce95 LC |
468 | ret = blk_integrity_add(disk); |
469 | if (ret) | |
470 | goto out_del_block_link; | |
bab53f6b | 471 | |
52b85909 CH |
472 | disk->part0->bd_holder_dir = |
473 | kobject_create_and_add("holders", &ddev->kobj); | |
385ccb70 LC |
474 | if (!disk->part0->bd_holder_dir) { |
475 | ret = -ENOMEM; | |
83cbce95 | 476 | goto out_del_integrity; |
385ccb70 | 477 | } |
52b85909 | 478 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); |
385ccb70 LC |
479 | if (!disk->slave_dir) { |
480 | ret = -ENOMEM; | |
83cbce95 | 481 | goto out_put_holder_dir; |
385ccb70 | 482 | } |
52b85909 | 483 | |
83cbce95 LC |
484 | ret = bd_register_pending_holders(disk); |
485 | if (ret < 0) | |
486 | goto out_put_slave_dir; | |
52b85909 | 487 | |
83cbce95 LC |
488 | ret = blk_register_queue(disk); |
489 | if (ret) | |
490 | goto out_put_slave_dir; | |
75f4dca5 | 491 | |
8235b5c1 CH |
492 | if (disk->flags & GENHD_FL_HIDDEN) { |
493 | /* | |
494 | * Don't let hidden disks show up in /proc/partitions, | |
495 | * and don't bother scanning for partitions either. | |
496 | */ | |
497 | disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; | |
498 | disk->flags |= GENHD_FL_NO_PART_SCAN; | |
499 | } else { | |
500 | ret = bdi_register(disk->bdi, "%u:%u", | |
501 | disk->major, disk->first_minor); | |
83cbce95 LC |
502 | if (ret) |
503 | goto out_unregister_queue; | |
8235b5c1 | 504 | bdi_set_owner(disk->bdi, ddev); |
83cbce95 LC |
505 | ret = sysfs_create_link(&ddev->kobj, |
506 | &disk->bdi->dev->kobj, "bdi"); | |
507 | if (ret) | |
508 | goto out_unregister_bdi; | |
8235b5c1 | 509 | |
9d5ee676 | 510 | bdev_add(disk->part0, ddev->devt); |
52b85909 CH |
511 | disk_scan_partitions(disk); |
512 | ||
513 | /* | |
514 | * Announce the disk and partitions after all partitions are | |
8235b5c1 | 515 | * created. (for hidden disks uevents remain suppressed forever) |
52b85909 CH |
516 | */ |
517 | dev_set_uevent_suppress(ddev, 0); | |
518 | disk_uevent(disk, KOBJ_ADD); | |
52b85909 CH |
519 | } |
520 | ||
75f4dca5 | 521 | disk_update_readahead(disk); |
77ea887e | 522 | disk_add_events(disk); |
83cbce95 LC |
523 | return 0; |
524 | ||
525 | out_unregister_bdi: | |
526 | if (!(disk->flags & GENHD_FL_HIDDEN)) | |
527 | bdi_unregister(disk->bdi); | |
528 | out_unregister_queue: | |
529 | blk_unregister_queue(disk); | |
530 | out_put_slave_dir: | |
531 | kobject_put(disk->slave_dir); | |
532 | out_put_holder_dir: | |
533 | kobject_put(disk->part0->bd_holder_dir); | |
534 | out_del_integrity: | |
535 | blk_integrity_del(disk); | |
536 | out_del_block_link: | |
537 | if (!sysfs_deprecated) | |
538 | sysfs_remove_link(block_depr, dev_name(ddev)); | |
539 | out_device_del: | |
540 | device_del(ddev); | |
83cbce95 LC |
541 | out_free_ext_minor: |
542 | if (disk->major == BLOCK_EXT_MAJOR) | |
543 | blk_free_ext_minor(disk->first_minor); | |
544 | return WARN_ON_ONCE(ret); /* keep until all callers handle errors */ | |
1da177e4 | 545 | } |
e63a46be | 546 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 547 | |
b142c5ae CH |
548 | /** |
549 | * blk_mark_disk_dead - mark a disk as dead | |
550 | * @disk: disk to mark as dead | |
551 | * | |
552 | * Mark as disk as dead (e.g. surprise removed) and don't accept any new I/O | |
553 | * to this disk. | |
554 | */ | |
555 | void blk_mark_disk_dead(struct gendisk *disk) | |
556 | { | |
557 | set_bit(GD_DEAD, &disk->state); | |
558 | blk_queue_start_drain(disk->queue); | |
559 | } | |
560 | EXPORT_SYMBOL_GPL(blk_mark_disk_dead); | |
561 | ||
b5bd357c LC |
562 | /** |
563 | * del_gendisk - remove the gendisk | |
564 | * @disk: the struct gendisk to remove | |
565 | * | |
566 | * Removes the gendisk and all its associated resources. This deletes the | |
567 | * partitions associated with the gendisk, and unregisters the associated | |
568 | * request_queue. | |
569 | * | |
570 | * This is the counter to the respective __device_add_disk() call. | |
571 | * | |
572 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
573 | * with put_disk(), which should be called after del_gendisk(), if | |
574 | * __device_add_disk() was used. | |
e8c7d14a LC |
575 | * |
576 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
577 | * it should not be deferred. | |
578 | * | |
579 | * Context: can sleep | |
b5bd357c | 580 | */ |
d2bf1b67 | 581 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 582 | { |
8e141f9e CH |
583 | struct request_queue *q = disk->queue; |
584 | ||
e8c7d14a LC |
585 | might_sleep(); |
586 | ||
9f286992 | 587 | if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN))) |
6b3ba976 CH |
588 | return; |
589 | ||
25520d55 | 590 | blk_integrity_del(disk); |
77ea887e TH |
591 | disk_del_events(disk); |
592 | ||
a8698707 | 593 | mutex_lock(&disk->open_mutex); |
d7a66574 | 594 | remove_inode_hash(disk->part0->bd_inode); |
d3c4a43d | 595 | blk_drop_partitions(disk); |
a8698707 | 596 | mutex_unlock(&disk->open_mutex); |
c76f48eb | 597 | |
45611837 CH |
598 | fsync_bdev(disk->part0); |
599 | __invalidate_device(disk->part0, true); | |
600 | ||
8e141f9e CH |
601 | /* |
602 | * Fail any new I/O. | |
603 | */ | |
604 | set_bit(GD_DEAD, &disk->state); | |
d2bf1b67 | 605 | set_capacity(disk, 0); |
d2bf1b67 | 606 | |
8e141f9e CH |
607 | /* |
608 | * Prevent new I/O from crossing bio_queue_enter(). | |
609 | */ | |
610 | blk_queue_start_drain(q); | |
8e141f9e | 611 | |
6b3ba976 | 612 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 | 613 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
6b3ba976 | 614 | |
90f16fdd JK |
615 | /* |
616 | * Unregister bdi before releasing device numbers (as they can | |
617 | * get reused and we'd get clashes in sysfs). | |
618 | */ | |
edb0872f | 619 | bdi_unregister(disk->bdi); |
90f16fdd | 620 | } |
d2bf1b67 | 621 | |
6b3ba976 | 622 | blk_unregister_queue(disk); |
d2bf1b67 | 623 | |
cb8432d6 | 624 | kobject_put(disk->part0->bd_holder_dir); |
d2bf1b67 | 625 | kobject_put(disk->slave_dir); |
d2bf1b67 | 626 | |
8446fe92 | 627 | part_stat_set_all(disk->part0, 0); |
cb8432d6 | 628 | disk->part0->bd_stamp = 0; |
d2bf1b67 TH |
629 | if (!sysfs_deprecated) |
630 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 631 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 632 | device_del(disk_to_dev(disk)); |
d308ae0d ML |
633 | |
634 | blk_mq_freeze_queue_wait(q); | |
635 | ||
636 | rq_qos_exit(q); | |
637 | blk_sync_queue(q); | |
638 | blk_flush_integrity(); | |
639 | /* | |
640 | * Allow using passthrough request again after the queue is torn down. | |
641 | */ | |
642 | blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q); | |
643 | __blk_mq_unfreeze_queue(q, true); | |
644 | ||
1da177e4 | 645 | } |
d2bf1b67 | 646 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 647 | |
99e6608c VV |
648 | /* sysfs access to bad-blocks list. */ |
649 | static ssize_t disk_badblocks_show(struct device *dev, | |
650 | struct device_attribute *attr, | |
651 | char *page) | |
652 | { | |
653 | struct gendisk *disk = dev_to_disk(dev); | |
654 | ||
655 | if (!disk->bb) | |
656 | return sprintf(page, "\n"); | |
657 | ||
658 | return badblocks_show(disk->bb, page, 0); | |
659 | } | |
660 | ||
661 | static ssize_t disk_badblocks_store(struct device *dev, | |
662 | struct device_attribute *attr, | |
663 | const char *page, size_t len) | |
664 | { | |
665 | struct gendisk *disk = dev_to_disk(dev); | |
666 | ||
667 | if (!disk->bb) | |
668 | return -ENXIO; | |
669 | ||
670 | return badblocks_store(disk->bb, page, len, 0); | |
671 | } | |
672 | ||
22ae8ce8 | 673 | void blk_request_module(dev_t devt) |
bd8eff3b | 674 | { |
a160c615 CH |
675 | unsigned int major = MAJOR(devt); |
676 | struct blk_major_name **n; | |
677 | ||
678 | mutex_lock(&major_names_lock); | |
679 | for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { | |
680 | if ((*n)->major == major && (*n)->probe) { | |
681 | (*n)->probe(devt); | |
682 | mutex_unlock(&major_names_lock); | |
683 | return; | |
684 | } | |
685 | } | |
686 | mutex_unlock(&major_names_lock); | |
687 | ||
bd8eff3b CH |
688 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
689 | /* Make old-style 2.4 aliases work */ | |
690 | request_module("block-major-%d", MAJOR(devt)); | |
691 | } | |
692 | ||
5c6f35c5 GKH |
693 | /* |
694 | * print a full list of all partitions - intended for places where the root | |
695 | * filesystem can't be mounted and thus to give the victim some idea of what | |
696 | * went wrong | |
697 | */ | |
698 | void __init printk_all_partitions(void) | |
699 | { | |
def4e38d TH |
700 | struct class_dev_iter iter; |
701 | struct device *dev; | |
702 | ||
703 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
704 | while ((dev = class_dev_iter_next(&iter))) { | |
705 | struct gendisk *disk = dev_to_disk(dev); | |
ad1eaa53 | 706 | struct block_device *part; |
1f014290 | 707 | char devt_buf[BDEVT_SIZE]; |
e559f58d | 708 | unsigned long idx; |
def4e38d TH |
709 | |
710 | /* | |
711 | * Don't show empty devices or things that have been | |
25985edc | 712 | * suppressed |
def4e38d TH |
713 | */ |
714 | if (get_capacity(disk) == 0 || | |
715 | (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) | |
716 | continue; | |
717 | ||
718 | /* | |
e559f58d CH |
719 | * Note, unlike /proc/partitions, I am showing the numbers in |
720 | * hex - the same format as the root= option takes. | |
def4e38d | 721 | */ |
e559f58d CH |
722 | rcu_read_lock(); |
723 | xa_for_each(&disk->part_tbl, idx, part) { | |
724 | if (!bdev_nr_sectors(part)) | |
725 | continue; | |
a9e7bc3d | 726 | printk("%s%s %10llu %pg %s", |
e559f58d | 727 | bdev_is_partition(part) ? " " : "", |
ad1eaa53 | 728 | bdevt_str(part->bd_dev, devt_buf), |
a9e7bc3d | 729 | bdev_nr_sectors(part) >> 1, part, |
ad1eaa53 CH |
730 | part->bd_meta_info ? |
731 | part->bd_meta_info->uuid : ""); | |
e559f58d | 732 | if (bdev_is_partition(part)) |
074a7aca | 733 | printk("\n"); |
e559f58d CH |
734 | else if (dev->parent && dev->parent->driver) |
735 | printk(" driver: %s\n", | |
736 | dev->parent->driver->name); | |
737 | else | |
738 | printk(" (driver?)\n"); | |
074a7aca | 739 | } |
e559f58d | 740 | rcu_read_unlock(); |
def4e38d TH |
741 | } |
742 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
743 | } |
744 | ||
1da177e4 LT |
745 | #ifdef CONFIG_PROC_FS |
746 | /* iterator */ | |
def4e38d | 747 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 748 | { |
def4e38d TH |
749 | loff_t skip = *pos; |
750 | struct class_dev_iter *iter; | |
751 | struct device *dev; | |
68c4d4a7 | 752 | |
aeb3d3a8 | 753 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
754 | if (!iter) |
755 | return ERR_PTR(-ENOMEM); | |
756 | ||
757 | seqf->private = iter; | |
758 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
759 | do { | |
760 | dev = class_dev_iter_next(iter); | |
761 | if (!dev) | |
762 | return NULL; | |
763 | } while (skip--); | |
764 | ||
765 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
766 | } |
767 | ||
def4e38d | 768 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 769 | { |
edfaa7c3 | 770 | struct device *dev; |
1da177e4 | 771 | |
def4e38d TH |
772 | (*pos)++; |
773 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 774 | if (dev) |
68c4d4a7 | 775 | return dev_to_disk(dev); |
2ac3cee5 | 776 | |
1da177e4 LT |
777 | return NULL; |
778 | } | |
779 | ||
def4e38d | 780 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 781 | { |
def4e38d | 782 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 783 | |
def4e38d TH |
784 | /* stop is called even after start failed :-( */ |
785 | if (iter) { | |
786 | class_dev_iter_exit(iter); | |
787 | kfree(iter); | |
77da1605 | 788 | seqf->private = NULL; |
5c0ef6d0 | 789 | } |
1da177e4 LT |
790 | } |
791 | ||
def4e38d | 792 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 793 | { |
06768067 | 794 | void *p; |
def4e38d TH |
795 | |
796 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 797 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
798 | seq_puts(seqf, "major minor #blocks name\n\n"); |
799 | return p; | |
1da177e4 LT |
800 | } |
801 | ||
cf771cb5 | 802 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
803 | { |
804 | struct gendisk *sgp = v; | |
ad1eaa53 | 805 | struct block_device *part; |
ecc75a98 | 806 | unsigned long idx; |
1da177e4 | 807 | |
1da177e4 | 808 | /* Don't show non-partitionable removeable devices or empty devices */ |
d27769ec | 809 | if (!get_capacity(sgp) || (!disk_max_parts(sgp) && |
f331c029 | 810 | (sgp->flags & GENHD_FL_REMOVABLE))) |
1da177e4 LT |
811 | return 0; |
812 | if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) | |
813 | return 0; | |
814 | ||
ecc75a98 CH |
815 | rcu_read_lock(); |
816 | xa_for_each(&sgp->part_tbl, idx, part) { | |
817 | if (!bdev_nr_sectors(part)) | |
818 | continue; | |
a291bb43 | 819 | seq_printf(seqf, "%4d %7d %10llu %pg\n", |
ad1eaa53 | 820 | MAJOR(part->bd_dev), MINOR(part->bd_dev), |
a291bb43 | 821 | bdev_nr_sectors(part) >> 1, part); |
ecc75a98 CH |
822 | } |
823 | rcu_read_unlock(); | |
1da177e4 LT |
824 | return 0; |
825 | } | |
826 | ||
f500975a | 827 | static const struct seq_operations partitions_op = { |
def4e38d TH |
828 | .start = show_partition_start, |
829 | .next = disk_seqf_next, | |
830 | .stop = disk_seqf_stop, | |
edfaa7c3 | 831 | .show = show_partition |
1da177e4 LT |
832 | }; |
833 | #endif | |
834 | ||
1da177e4 LT |
835 | static int __init genhd_device_init(void) |
836 | { | |
e105b8bf DW |
837 | int error; |
838 | ||
839 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
840 | error = class_register(&block_class); | |
ee27a558 RM |
841 | if (unlikely(error)) |
842 | return error; | |
1da177e4 | 843 | blk_dev_init(); |
edfaa7c3 | 844 | |
561ec68e ZY |
845 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
846 | ||
edfaa7c3 | 847 | /* create top-level block dir */ |
e52eec13 AK |
848 | if (!sysfs_deprecated) |
849 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 850 | return 0; |
1da177e4 LT |
851 | } |
852 | ||
853 | subsys_initcall(genhd_device_init); | |
854 | ||
edfaa7c3 KS |
855 | static ssize_t disk_range_show(struct device *dev, |
856 | struct device_attribute *attr, char *buf) | |
1da177e4 | 857 | { |
edfaa7c3 | 858 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 859 | |
edfaa7c3 | 860 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
861 | } |
862 | ||
1f014290 TH |
863 | static ssize_t disk_ext_range_show(struct device *dev, |
864 | struct device_attribute *attr, char *buf) | |
865 | { | |
866 | struct gendisk *disk = dev_to_disk(dev); | |
867 | ||
b5d0b9df | 868 | return sprintf(buf, "%d\n", disk_max_parts(disk)); |
1f014290 TH |
869 | } |
870 | ||
edfaa7c3 KS |
871 | static ssize_t disk_removable_show(struct device *dev, |
872 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 873 | { |
edfaa7c3 | 874 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 875 | |
edfaa7c3 KS |
876 | return sprintf(buf, "%d\n", |
877 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
878 | } |
879 | ||
8ddcd653 CH |
880 | static ssize_t disk_hidden_show(struct device *dev, |
881 | struct device_attribute *attr, char *buf) | |
882 | { | |
883 | struct gendisk *disk = dev_to_disk(dev); | |
884 | ||
885 | return sprintf(buf, "%d\n", | |
886 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
887 | } | |
888 | ||
1c9ce527 KS |
889 | static ssize_t disk_ro_show(struct device *dev, |
890 | struct device_attribute *attr, char *buf) | |
891 | { | |
892 | struct gendisk *disk = dev_to_disk(dev); | |
893 | ||
b7db9956 | 894 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
895 | } |
896 | ||
3ad5cee5 CH |
897 | ssize_t part_size_show(struct device *dev, |
898 | struct device_attribute *attr, char *buf) | |
899 | { | |
0d02129e | 900 | return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev))); |
3ad5cee5 CH |
901 | } |
902 | ||
903 | ssize_t part_stat_show(struct device *dev, | |
904 | struct device_attribute *attr, char *buf) | |
905 | { | |
0d02129e CH |
906 | struct block_device *bdev = dev_to_bdev(dev); |
907 | struct request_queue *q = bdev->bd_disk->queue; | |
ea18e0f0 | 908 | struct disk_stats stat; |
3ad5cee5 CH |
909 | unsigned int inflight; |
910 | ||
0d02129e | 911 | part_stat_read_all(bdev, &stat); |
b2f609e1 | 912 | if (queue_is_mq(q)) |
0d02129e | 913 | inflight = blk_mq_in_flight(q, bdev); |
b2f609e1 | 914 | else |
0d02129e | 915 | inflight = part_in_flight(bdev); |
ea18e0f0 | 916 | |
3ad5cee5 CH |
917 | return sprintf(buf, |
918 | "%8lu %8lu %8llu %8u " | |
919 | "%8lu %8lu %8llu %8u " | |
920 | "%8u %8u %8u " | |
921 | "%8lu %8lu %8llu %8u " | |
922 | "%8lu %8u" | |
923 | "\n", | |
ea18e0f0 KK |
924 | stat.ios[STAT_READ], |
925 | stat.merges[STAT_READ], | |
926 | (unsigned long long)stat.sectors[STAT_READ], | |
927 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
928 | stat.ios[STAT_WRITE], | |
929 | stat.merges[STAT_WRITE], | |
930 | (unsigned long long)stat.sectors[STAT_WRITE], | |
931 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 932 | inflight, |
ea18e0f0 | 933 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
934 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
935 | stat.nsecs[STAT_WRITE] + | |
936 | stat.nsecs[STAT_DISCARD] + | |
937 | stat.nsecs[STAT_FLUSH], | |
938 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
939 | stat.ios[STAT_DISCARD], |
940 | stat.merges[STAT_DISCARD], | |
941 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
942 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
943 | stat.ios[STAT_FLUSH], | |
944 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
945 | } |
946 | ||
947 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
948 | char *buf) | |
949 | { | |
0d02129e CH |
950 | struct block_device *bdev = dev_to_bdev(dev); |
951 | struct request_queue *q = bdev->bd_disk->queue; | |
3ad5cee5 CH |
952 | unsigned int inflight[2]; |
953 | ||
b2f609e1 | 954 | if (queue_is_mq(q)) |
0d02129e | 955 | blk_mq_in_flight_rw(q, bdev, inflight); |
b2f609e1 | 956 | else |
0d02129e | 957 | part_in_flight_rw(bdev, inflight); |
b2f609e1 | 958 | |
3ad5cee5 CH |
959 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
960 | } | |
961 | ||
edfaa7c3 KS |
962 | static ssize_t disk_capability_show(struct device *dev, |
963 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 964 | { |
edfaa7c3 KS |
965 | struct gendisk *disk = dev_to_disk(dev); |
966 | ||
967 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 968 | } |
edfaa7c3 | 969 | |
c72758f3 MP |
970 | static ssize_t disk_alignment_offset_show(struct device *dev, |
971 | struct device_attribute *attr, | |
972 | char *buf) | |
973 | { | |
974 | struct gendisk *disk = dev_to_disk(dev); | |
975 | ||
976 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
977 | } | |
978 | ||
86b37281 MP |
979 | static ssize_t disk_discard_alignment_show(struct device *dev, |
980 | struct device_attribute *attr, | |
981 | char *buf) | |
982 | { | |
983 | struct gendisk *disk = dev_to_disk(dev); | |
984 | ||
dd3d145d | 985 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
986 | } |
987 | ||
13927b31 MC |
988 | static ssize_t diskseq_show(struct device *dev, |
989 | struct device_attribute *attr, char *buf) | |
990 | { | |
991 | struct gendisk *disk = dev_to_disk(dev); | |
992 | ||
993 | return sprintf(buf, "%llu\n", disk->diskseq); | |
994 | } | |
995 | ||
5657a819 JP |
996 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
997 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
998 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
999 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1000 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1001 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1002 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1003 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1004 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1005 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1006 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1007 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
13927b31 | 1008 | static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL); |
3ad5cee5 | 1009 | |
c17bb495 | 1010 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1011 | ssize_t part_fail_show(struct device *dev, |
1012 | struct device_attribute *attr, char *buf) | |
1013 | { | |
0d02129e | 1014 | return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail); |
3ad5cee5 CH |
1015 | } |
1016 | ||
1017 | ssize_t part_fail_store(struct device *dev, | |
1018 | struct device_attribute *attr, | |
1019 | const char *buf, size_t count) | |
1020 | { | |
3ad5cee5 CH |
1021 | int i; |
1022 | ||
1023 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
0d02129e | 1024 | dev_to_bdev(dev)->bd_make_it_fail = i; |
3ad5cee5 CH |
1025 | |
1026 | return count; | |
1027 | } | |
1028 | ||
edfaa7c3 | 1029 | static struct device_attribute dev_attr_fail = |
5657a819 | 1030 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1031 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1032 | ||
581d4e28 JA |
1033 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1034 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1035 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1036 | #endif |
edfaa7c3 KS |
1037 | |
1038 | static struct attribute *disk_attrs[] = { | |
1039 | &dev_attr_range.attr, | |
1f014290 | 1040 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1041 | &dev_attr_removable.attr, |
8ddcd653 | 1042 | &dev_attr_hidden.attr, |
1c9ce527 | 1043 | &dev_attr_ro.attr, |
edfaa7c3 | 1044 | &dev_attr_size.attr, |
c72758f3 | 1045 | &dev_attr_alignment_offset.attr, |
86b37281 | 1046 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1047 | &dev_attr_capability.attr, |
1048 | &dev_attr_stat.attr, | |
316d315b | 1049 | &dev_attr_inflight.attr, |
99e6608c | 1050 | &dev_attr_badblocks.attr, |
2bc8cda5 CH |
1051 | &dev_attr_events.attr, |
1052 | &dev_attr_events_async.attr, | |
1053 | &dev_attr_events_poll_msecs.attr, | |
13927b31 | 1054 | &dev_attr_diskseq.attr, |
edfaa7c3 KS |
1055 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1056 | &dev_attr_fail.attr, | |
581d4e28 JA |
1057 | #endif |
1058 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1059 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1060 | #endif |
1061 | NULL | |
1062 | }; | |
1063 | ||
9438b3e0 DW |
1064 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1065 | { | |
1066 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1067 | struct gendisk *disk = dev_to_disk(dev); | |
1068 | ||
1069 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1070 | return 0; | |
1071 | return a->mode; | |
1072 | } | |
1073 | ||
edfaa7c3 KS |
1074 | static struct attribute_group disk_attr_group = { |
1075 | .attrs = disk_attrs, | |
9438b3e0 | 1076 | .is_visible = disk_visible, |
edfaa7c3 KS |
1077 | }; |
1078 | ||
a4dbd674 | 1079 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1080 | &disk_attr_group, |
1081 | NULL | |
1da177e4 LT |
1082 | }; |
1083 | ||
b5bd357c LC |
1084 | /** |
1085 | * disk_release - releases all allocated resources of the gendisk | |
1086 | * @dev: the device representing this disk | |
1087 | * | |
1088 | * This function releases all allocated resources of the gendisk. | |
1089 | * | |
b5bd357c LC |
1090 | * Drivers which used __device_add_disk() have a gendisk with a request_queue |
1091 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1092 | * drivers we also call blk_put_queue() for them, and we expect the | |
1093 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1094 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1095 | * |
1096 | * Context: can sleep | |
b5bd357c | 1097 | */ |
edfaa7c3 | 1098 | static void disk_release(struct device *dev) |
1da177e4 | 1099 | { |
edfaa7c3 KS |
1100 | struct gendisk *disk = dev_to_disk(dev); |
1101 | ||
e8c7d14a | 1102 | might_sleep(); |
a2041761 | 1103 | WARN_ON_ONCE(disk_live(disk)); |
e8c7d14a | 1104 | |
9a2b5cd4 ML |
1105 | blk_mq_cancel_work_sync(disk->queue); |
1106 | ||
77ea887e | 1107 | disk_release_events(disk); |
1da177e4 | 1108 | kfree(disk->random); |
a33df75c | 1109 | xa_destroy(&disk->part_tbl); |
d152c682 | 1110 | disk->queue->disk = NULL; |
61a35cfc | 1111 | blk_put_queue(disk->queue); |
2f4731dc | 1112 | iput(disk->part0->bd_inode); /* frees the disk */ |
1da177e4 | 1113 | } |
87eb7107 MC |
1114 | |
1115 | static int block_uevent(struct device *dev, struct kobj_uevent_env *env) | |
1116 | { | |
1117 | struct gendisk *disk = dev_to_disk(dev); | |
1118 | ||
1119 | return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq); | |
1120 | } | |
1121 | ||
edfaa7c3 KS |
1122 | struct class block_class = { |
1123 | .name = "block", | |
87eb7107 | 1124 | .dev_uevent = block_uevent, |
1da177e4 LT |
1125 | }; |
1126 | ||
3c2670e6 | 1127 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1128 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1129 | { |
1130 | struct gendisk *disk = dev_to_disk(dev); | |
1131 | ||
348e114b CH |
1132 | if (disk->fops->devnode) |
1133 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1134 | return NULL; |
1135 | } | |
1136 | ||
ef45fe47 | 1137 | const struct device_type disk_type = { |
edfaa7c3 KS |
1138 | .name = "disk", |
1139 | .groups = disk_attr_groups, | |
1140 | .release = disk_release, | |
e454cea2 | 1141 | .devnode = block_devnode, |
1da177e4 LT |
1142 | }; |
1143 | ||
a6e2ba88 | 1144 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1145 | /* |
1146 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1147 | * entries do, above, but makes them available through one seq_file. | |
1148 | * | |
1149 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1150 | * extra fields. | |
1151 | */ | |
1152 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1153 | { |
1154 | struct gendisk *gp = v; | |
ad1eaa53 | 1155 | struct block_device *hd; |
e016b782 | 1156 | unsigned int inflight; |
ea18e0f0 | 1157 | struct disk_stats stat; |
7fae67cc | 1158 | unsigned long idx; |
1da177e4 LT |
1159 | |
1160 | /* | |
ed9e1982 | 1161 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1162 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1163 | " rio rmerge rsect ruse wio wmerge " |
1164 | "wsect wuse running use aveq" | |
1165 | "\n\n"); | |
1166 | */ | |
9f5e4865 | 1167 | |
7fae67cc CH |
1168 | rcu_read_lock(); |
1169 | xa_for_each(&gp->part_tbl, idx, hd) { | |
1170 | if (bdev_is_partition(hd) && !bdev_nr_sectors(hd)) | |
1171 | continue; | |
0d02129e | 1172 | part_stat_read_all(hd, &stat); |
b2f609e1 | 1173 | if (queue_is_mq(gp->queue)) |
ad1eaa53 | 1174 | inflight = blk_mq_in_flight(gp->queue, hd); |
b2f609e1 | 1175 | else |
ad1eaa53 | 1176 | inflight = part_in_flight(hd); |
ea18e0f0 | 1177 | |
26e2d7a3 | 1178 | seq_printf(seqf, "%4d %7d %pg " |
bdca3c87 MC |
1179 | "%lu %lu %lu %u " |
1180 | "%lu %lu %lu %u " | |
1181 | "%u %u %u " | |
b6866318 KK |
1182 | "%lu %lu %lu %u " |
1183 | "%lu %u" | |
1184 | "\n", | |
26e2d7a3 | 1185 | MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd, |
ea18e0f0 KK |
1186 | stat.ios[STAT_READ], |
1187 | stat.merges[STAT_READ], | |
1188 | stat.sectors[STAT_READ], | |
1189 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1190 | NSEC_PER_MSEC), | |
1191 | stat.ios[STAT_WRITE], | |
1192 | stat.merges[STAT_WRITE], | |
1193 | stat.sectors[STAT_WRITE], | |
1194 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1195 | NSEC_PER_MSEC), | |
e016b782 | 1196 | inflight, |
ea18e0f0 | 1197 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1198 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1199 | stat.nsecs[STAT_WRITE] + | |
1200 | stat.nsecs[STAT_DISCARD] + | |
1201 | stat.nsecs[STAT_FLUSH], | |
1202 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1203 | stat.ios[STAT_DISCARD], |
1204 | stat.merges[STAT_DISCARD], | |
1205 | stat.sectors[STAT_DISCARD], | |
1206 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1207 | NSEC_PER_MSEC), | |
1208 | stat.ios[STAT_FLUSH], | |
1209 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1210 | NSEC_PER_MSEC) | |
28f39d55 | 1211 | ); |
1da177e4 | 1212 | } |
7fae67cc | 1213 | rcu_read_unlock(); |
9f5e4865 | 1214 | |
1da177e4 LT |
1215 | return 0; |
1216 | } | |
1217 | ||
31d85ab2 | 1218 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1219 | .start = disk_seqf_start, |
1220 | .next = disk_seqf_next, | |
1221 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1222 | .show = diskstats_show |
1223 | }; | |
f500975a AD |
1224 | |
1225 | static int __init proc_genhd_init(void) | |
1226 | { | |
fddda2b7 CH |
1227 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1228 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1229 | return 0; |
1230 | } | |
1231 | module_init(proc_genhd_init); | |
a6e2ba88 | 1232 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1233 | |
c97d93c3 CH |
1234 | dev_t part_devt(struct gendisk *disk, u8 partno) |
1235 | { | |
0e0ccdec | 1236 | struct block_device *part; |
c97d93c3 CH |
1237 | dev_t devt = 0; |
1238 | ||
0e0ccdec CH |
1239 | rcu_read_lock(); |
1240 | part = xa_load(&disk->part_tbl, partno); | |
1241 | if (part) | |
c97d93c3 | 1242 | devt = part->bd_dev; |
0e0ccdec | 1243 | rcu_read_unlock(); |
c97d93c3 CH |
1244 | |
1245 | return devt; | |
1246 | } | |
1247 | ||
cf771cb5 | 1248 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1249 | { |
def4e38d TH |
1250 | dev_t devt = MKDEV(0, 0); |
1251 | struct class_dev_iter iter; | |
1252 | struct device *dev; | |
a142be85 | 1253 | |
def4e38d TH |
1254 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1255 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1256 | struct gendisk *disk = dev_to_disk(dev); |
a142be85 | 1257 | |
3ada8b7e | 1258 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1259 | continue; |
f331c029 | 1260 | |
41b8c853 NB |
1261 | if (partno < disk->minors) { |
1262 | /* We need to return the right devno, even | |
1263 | * if the partition doesn't exist yet. | |
1264 | */ | |
1265 | devt = MKDEV(MAJOR(dev->devt), | |
1266 | MINOR(dev->devt) + partno); | |
c97d93c3 CH |
1267 | } else { |
1268 | devt = part_devt(disk, partno); | |
1269 | if (devt) | |
1270 | break; | |
def4e38d | 1271 | } |
5c0ef6d0 | 1272 | } |
def4e38d | 1273 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1274 | return devt; |
1275 | } | |
edfaa7c3 | 1276 | |
4a1fa41d CH |
1277 | struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id, |
1278 | struct lock_class_key *lkclass) | |
1946089a CL |
1279 | { |
1280 | struct gendisk *disk; | |
1281 | ||
61a35cfc CH |
1282 | if (!blk_get_queue(q)) |
1283 | return NULL; | |
1284 | ||
c1b511eb | 1285 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 | 1286 | if (!disk) |
61a35cfc | 1287 | goto out_put_queue; |
6c23a968 | 1288 | |
edb0872f CH |
1289 | disk->bdi = bdi_alloc(node_id); |
1290 | if (!disk->bdi) | |
1291 | goto out_free_disk; | |
1292 | ||
cb8432d6 CH |
1293 | disk->part0 = bdev_alloc(disk, 0); |
1294 | if (!disk->part0) | |
edb0872f | 1295 | goto out_free_bdi; |
22ae8ce8 | 1296 | |
f93af2a4 | 1297 | disk->node_id = node_id; |
a8698707 | 1298 | mutex_init(&disk->open_mutex); |
a33df75c CH |
1299 | xa_init(&disk->part_tbl); |
1300 | if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL)) | |
1301 | goto out_destroy_part_tbl; | |
f93af2a4 | 1302 | |
f93af2a4 CH |
1303 | rand_initialize_disk(disk); |
1304 | disk_to_dev(disk)->class = &block_class; | |
1305 | disk_to_dev(disk)->type = &disk_type; | |
1306 | device_initialize(disk_to_dev(disk)); | |
cf179948 | 1307 | inc_diskseq(disk); |
4a1fa41d | 1308 | disk->queue = q; |
d152c682 | 1309 | q->disk = disk; |
4dcc4874 | 1310 | lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0); |
0dbcfe24 CH |
1311 | #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED |
1312 | INIT_LIST_HEAD(&disk->slave_bdevs); | |
1313 | #endif | |
1da177e4 | 1314 | return disk; |
f93af2a4 | 1315 | |
a33df75c CH |
1316 | out_destroy_part_tbl: |
1317 | xa_destroy(&disk->part_tbl); | |
06cc978d | 1318 | disk->part0->bd_disk = NULL; |
2f4731dc | 1319 | iput(disk->part0->bd_inode); |
edb0872f CH |
1320 | out_free_bdi: |
1321 | bdi_put(disk->bdi); | |
f93af2a4 CH |
1322 | out_free_disk: |
1323 | kfree(disk); | |
61a35cfc CH |
1324 | out_put_queue: |
1325 | blk_put_queue(q); | |
f93af2a4 | 1326 | return NULL; |
1da177e4 | 1327 | } |
e319e1fb | 1328 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1329 | |
4dcc4874 | 1330 | struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass) |
f525464a CH |
1331 | { |
1332 | struct request_queue *q; | |
1333 | struct gendisk *disk; | |
1334 | ||
1335 | q = blk_alloc_queue(node); | |
1336 | if (!q) | |
1337 | return NULL; | |
1338 | ||
4a1fa41d | 1339 | disk = __alloc_disk_node(q, node, lkclass); |
f525464a CH |
1340 | if (!disk) { |
1341 | blk_cleanup_queue(q); | |
1342 | return NULL; | |
1343 | } | |
f525464a CH |
1344 | return disk; |
1345 | } | |
1346 | EXPORT_SYMBOL(__blk_alloc_disk); | |
1347 | ||
b5bd357c LC |
1348 | /** |
1349 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1350 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1351 | * |
1352 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1353 | * we'll have disk_release() called. | |
e8c7d14a LC |
1354 | * |
1355 | * Context: Any context, but the last reference must not be dropped from | |
1356 | * atomic context. | |
b5bd357c | 1357 | */ |
1da177e4 LT |
1358 | void put_disk(struct gendisk *disk) |
1359 | { | |
1360 | if (disk) | |
efdc41c8 | 1361 | put_device(disk_to_dev(disk)); |
1da177e4 | 1362 | } |
1da177e4 LT |
1363 | EXPORT_SYMBOL(put_disk); |
1364 | ||
f525464a CH |
1365 | /** |
1366 | * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk | |
1367 | * @disk: gendisk to shutdown | |
1368 | * | |
1369 | * Mark the queue hanging off @disk DYING, drain all pending requests, then mark | |
1370 | * the queue DEAD, destroy and put it and the gendisk structure. | |
1371 | * | |
1372 | * Context: can sleep | |
1373 | */ | |
1374 | void blk_cleanup_disk(struct gendisk *disk) | |
1375 | { | |
1376 | blk_cleanup_queue(disk->queue); | |
1377 | put_disk(disk); | |
1378 | } | |
1379 | EXPORT_SYMBOL(blk_cleanup_disk); | |
1380 | ||
e3264a4d HR |
1381 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1382 | { | |
1383 | char event[] = "DISK_RO=1"; | |
1384 | char *envp[] = { event, NULL }; | |
1385 | ||
1386 | if (!ro) | |
1387 | event[8] = '0'; | |
1388 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1389 | } | |
1390 | ||
52f019d4 CH |
1391 | /** |
1392 | * set_disk_ro - set a gendisk read-only | |
1393 | * @disk: gendisk to operate on | |
7f31bee3 | 1394 | * @read_only: %true to set the disk read-only, %false set the disk read/write |
52f019d4 CH |
1395 | * |
1396 | * This function is used to indicate whether a given disk device should have its | |
1397 | * read-only flag set. set_disk_ro() is typically used by device drivers to | |
1398 | * indicate whether the underlying physical device is write-protected. | |
1399 | */ | |
1400 | void set_disk_ro(struct gendisk *disk, bool read_only) | |
1da177e4 | 1401 | { |
52f019d4 CH |
1402 | if (read_only) { |
1403 | if (test_and_set_bit(GD_READ_ONLY, &disk->state)) | |
1404 | return; | |
1405 | } else { | |
1406 | if (!test_and_clear_bit(GD_READ_ONLY, &disk->state)) | |
1407 | return; | |
e3264a4d | 1408 | } |
52f019d4 | 1409 | set_disk_ro_uevent(disk, read_only); |
1da177e4 | 1410 | } |
1da177e4 LT |
1411 | EXPORT_SYMBOL(set_disk_ro); |
1412 | ||
1413 | int bdev_read_only(struct block_device *bdev) | |
1414 | { | |
947139bf | 1415 | return bdev->bd_read_only || get_disk_ro(bdev->bd_disk); |
1da177e4 | 1416 | } |
1da177e4 | 1417 | EXPORT_SYMBOL(bdev_read_only); |
cf179948 MC |
1418 | |
1419 | void inc_diskseq(struct gendisk *disk) | |
1420 | { | |
1421 | disk->diskseq = atomic64_inc_return(&diskseq); | |
1422 | } |