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> | |
b81e0c23 | 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> |
82d981d4 | 28 | #include <linux/part_stat.h> |
1da177e4 | 29 | |
ff88972c | 30 | #include "blk.h" |
2aa7745b | 31 | #include "blk-mq-sched.h" |
8e141f9e | 32 | #include "blk-rq-qos.h" |
ff88972c | 33 | |
31eb6186 | 34 | static struct kobject *block_depr; |
1da177e4 | 35 | |
cf179948 MC |
36 | /* |
37 | * Unique, monotonically increasing sequential number associated with block | |
38 | * devices instances (i.e. incremented each time a device is attached). | |
39 | * Associating uevents with block devices in userspace is difficult and racy: | |
40 | * the uevent netlink socket is lossy, and on slow and overloaded systems has | |
41 | * a very high latency. | |
42 | * Block devices do not have exclusive owners in userspace, any process can set | |
43 | * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0 | |
44 | * can be reused again and again). | |
45 | * A userspace process setting up a block device and watching for its events | |
46 | * cannot thus reliably tell whether an event relates to the device it just set | |
47 | * up or another earlier instance with the same name. | |
48 | * This sequential number allows userspace processes to solve this problem, and | |
49 | * uniquely associate an uevent to the lifetime to a device. | |
50 | */ | |
51 | static atomic64_t diskseq; | |
52 | ||
bcce3de1 | 53 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 54 | #define NR_EXT_DEVT (1 << MINORBITS) |
22ae8ce8 | 55 | static DEFINE_IDA(ext_devt_ida); |
bcce3de1 | 56 | |
a782483c CH |
57 | void set_capacity(struct gendisk *disk, sector_t sectors) |
58 | { | |
cb8432d6 | 59 | struct block_device *bdev = disk->part0; |
a782483c | 60 | |
0f472277 | 61 | spin_lock(&bdev->bd_size_lock); |
a782483c | 62 | i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); |
f09313c5 | 63 | bdev->bd_nr_sectors = sectors; |
0f472277 | 64 | spin_unlock(&bdev->bd_size_lock); |
a782483c CH |
65 | } |
66 | EXPORT_SYMBOL(set_capacity); | |
67 | ||
e598a72f | 68 | /* |
449f4ec9 CH |
69 | * Set disk capacity and notify if the size is not currently zero and will not |
70 | * be set to zero. Returns true if a uevent was sent, otherwise false. | |
e598a72f | 71 | */ |
449f4ec9 | 72 | bool set_capacity_and_notify(struct gendisk *disk, sector_t size) |
e598a72f BS |
73 | { |
74 | sector_t capacity = get_capacity(disk); | |
a782483c | 75 | char *envp[] = { "RESIZE=1", NULL }; |
e598a72f BS |
76 | |
77 | set_capacity(disk, size); | |
e598a72f | 78 | |
a782483c CH |
79 | /* |
80 | * Only print a message and send a uevent if the gendisk is user visible | |
81 | * and alive. This avoids spamming the log and udev when setting the | |
82 | * initial capacity during probing. | |
83 | */ | |
84 | if (size == capacity || | |
50b4aecf CH |
85 | !disk_live(disk) || |
86 | (disk->flags & GENHD_FL_HIDDEN)) | |
a782483c | 87 | return false; |
e598a72f | 88 | |
a782483c | 89 | pr_info("%s: detected capacity change from %lld to %lld\n", |
452c0bf8 | 90 | disk->disk_name, capacity, size); |
7e890c37 | 91 | |
a782483c CH |
92 | /* |
93 | * Historically we did not send a uevent for changes to/from an empty | |
94 | * device. | |
95 | */ | |
96 | if (!capacity || !size) | |
97 | return false; | |
98 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
99 | return true; | |
e598a72f | 100 | } |
449f4ec9 | 101 | EXPORT_SYMBOL_GPL(set_capacity_and_notify); |
e598a72f | 102 | |
5cbd28e3 | 103 | /* |
abd2864a CH |
104 | * Format the device name of the indicated block device into the supplied buffer |
105 | * and return a pointer to that same buffer for convenience. | |
106 | * | |
107 | * Note: do not use this in new code, use the %pg specifier to sprintf and | |
108 | * printk insted. | |
5cbd28e3 | 109 | */ |
abd2864a | 110 | const char *bdevname(struct block_device *bdev, char *buf) |
5cbd28e3 | 111 | { |
abd2864a CH |
112 | struct gendisk *hd = bdev->bd_disk; |
113 | int partno = bdev->bd_partno; | |
114 | ||
5cbd28e3 CH |
115 | if (!partno) |
116 | snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); | |
117 | else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) | |
118 | snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); | |
119 | else | |
120 | snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); | |
121 | ||
122 | return buf; | |
123 | } | |
5cbd28e3 | 124 | EXPORT_SYMBOL(bdevname); |
e598a72f | 125 | |
0d02129e CH |
126 | static void part_stat_read_all(struct block_device *part, |
127 | struct disk_stats *stat) | |
ea18e0f0 KK |
128 | { |
129 | int cpu; | |
130 | ||
131 | memset(stat, 0, sizeof(struct disk_stats)); | |
132 | for_each_possible_cpu(cpu) { | |
0d02129e | 133 | struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu); |
ea18e0f0 KK |
134 | int group; |
135 | ||
136 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
137 | stat->nsecs[group] += ptr->nsecs[group]; | |
138 | stat->sectors[group] += ptr->sectors[group]; | |
139 | stat->ios[group] += ptr->ios[group]; | |
140 | stat->merges[group] += ptr->merges[group]; | |
141 | } | |
142 | ||
143 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
144 | } |
145 | } | |
ea18e0f0 | 146 | |
8446fe92 | 147 | static unsigned int part_in_flight(struct block_device *part) |
f299b7c7 | 148 | { |
b2f609e1 | 149 | unsigned int inflight = 0; |
1226b8dd | 150 | int cpu; |
f299b7c7 | 151 | |
1226b8dd | 152 | for_each_possible_cpu(cpu) { |
e016b782 MP |
153 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
154 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 155 | } |
e016b782 MP |
156 | if ((int)inflight < 0) |
157 | inflight = 0; | |
1226b8dd | 158 | |
e016b782 | 159 | return inflight; |
f299b7c7 JA |
160 | } |
161 | ||
8446fe92 CH |
162 | static void part_in_flight_rw(struct block_device *part, |
163 | unsigned int inflight[2]) | |
bf0ddaba | 164 | { |
1226b8dd MP |
165 | int cpu; |
166 | ||
1226b8dd MP |
167 | inflight[0] = 0; |
168 | inflight[1] = 0; | |
169 | for_each_possible_cpu(cpu) { | |
170 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
171 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
172 | } | |
173 | if ((int)inflight[0] < 0) | |
174 | inflight[0] = 0; | |
175 | if ((int)inflight[1] < 0) | |
176 | inflight[1] = 0; | |
bf0ddaba OS |
177 | } |
178 | ||
1da177e4 LT |
179 | /* |
180 | * Can be deleted altogether. Later. | |
181 | * | |
182 | */ | |
133d55cd | 183 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
184 | static struct blk_major_name { |
185 | struct blk_major_name *next; | |
186 | int major; | |
187 | char name[16]; | |
a160c615 | 188 | void (*probe)(dev_t devt); |
68eef3b4 | 189 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
e49fbbbf | 190 | static DEFINE_MUTEX(major_names_lock); |
dfbb3409 | 191 | static DEFINE_SPINLOCK(major_names_spinlock); |
1da177e4 LT |
192 | |
193 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 194 | static inline int major_to_index(unsigned major) |
1da177e4 | 195 | { |
68eef3b4 | 196 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
197 | } |
198 | ||
68eef3b4 | 199 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 200 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 201 | { |
68eef3b4 | 202 | struct blk_major_name *dp; |
7170be5f | 203 | |
dfbb3409 | 204 | spin_lock(&major_names_spinlock); |
133d55cd LG |
205 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) |
206 | if (dp->major == offset) | |
cf771cb5 | 207 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
dfbb3409 | 208 | spin_unlock(&major_names_spinlock); |
1da177e4 | 209 | } |
68eef3b4 | 210 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 211 | |
9e8c0bcc | 212 | /** |
e2b6b301 | 213 | * __register_blkdev - register a new block device |
9e8c0bcc | 214 | * |
f33ff110 SB |
215 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
216 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc | 217 | * @name: the name of the new block device as a zero terminated string |
26e06f5b LC |
218 | * @probe: pre-devtmpfs / pre-udev callback used to create disks when their |
219 | * pre-created device node is accessed. When a probe call uses | |
220 | * add_disk() and it fails the driver must cleanup resources. This | |
221 | * interface may soon be removed. | |
9e8c0bcc MN |
222 | * |
223 | * The @name must be unique within the system. | |
224 | * | |
0e056eb5 | 225 | * The return value depends on the @major input parameter: |
226 | * | |
f33ff110 SB |
227 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
228 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 229 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 230 | * then the return value is the allocated major number in range |
f33ff110 SB |
231 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
232 | * | |
233 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
234 | * major numbers. | |
e2b6b301 CH |
235 | * |
236 | * Use register_blkdev instead for any new code. | |
9e8c0bcc | 237 | */ |
a160c615 CH |
238 | int __register_blkdev(unsigned int major, const char *name, |
239 | void (*probe)(dev_t devt)) | |
1da177e4 LT |
240 | { |
241 | struct blk_major_name **n, *p; | |
242 | int index, ret = 0; | |
243 | ||
e49fbbbf | 244 | mutex_lock(&major_names_lock); |
1da177e4 LT |
245 | |
246 | /* temporary */ | |
247 | if (major == 0) { | |
248 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
249 | if (major_names[index] == NULL) | |
250 | break; | |
251 | } | |
252 | ||
253 | if (index == 0) { | |
dfc76d11 KP |
254 | printk("%s: failed to get major for %s\n", |
255 | __func__, name); | |
1da177e4 LT |
256 | ret = -EBUSY; |
257 | goto out; | |
258 | } | |
259 | major = index; | |
260 | ret = major; | |
261 | } | |
262 | ||
133d55cd | 263 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
264 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
265 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
266 | |
267 | ret = -EINVAL; | |
268 | goto out; | |
269 | } | |
270 | ||
1da177e4 LT |
271 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
272 | if (p == NULL) { | |
273 | ret = -ENOMEM; | |
274 | goto out; | |
275 | } | |
276 | ||
277 | p->major = major; | |
a160c615 | 278 | p->probe = probe; |
1da177e4 LT |
279 | strlcpy(p->name, name, sizeof(p->name)); |
280 | p->next = NULL; | |
281 | index = major_to_index(major); | |
282 | ||
dfbb3409 | 283 | spin_lock(&major_names_spinlock); |
1da177e4 LT |
284 | for (n = &major_names[index]; *n; n = &(*n)->next) { |
285 | if ((*n)->major == major) | |
286 | break; | |
287 | } | |
288 | if (!*n) | |
289 | *n = p; | |
290 | else | |
291 | ret = -EBUSY; | |
dfbb3409 | 292 | spin_unlock(&major_names_spinlock); |
1da177e4 LT |
293 | |
294 | if (ret < 0) { | |
f33ff110 | 295 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
296 | major, name); |
297 | kfree(p); | |
298 | } | |
299 | out: | |
e49fbbbf | 300 | mutex_unlock(&major_names_lock); |
1da177e4 LT |
301 | return ret; |
302 | } | |
a160c615 | 303 | EXPORT_SYMBOL(__register_blkdev); |
1da177e4 | 304 | |
f4480240 | 305 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
306 | { |
307 | struct blk_major_name **n; | |
308 | struct blk_major_name *p = NULL; | |
309 | int index = major_to_index(major); | |
1da177e4 | 310 | |
e49fbbbf | 311 | mutex_lock(&major_names_lock); |
dfbb3409 | 312 | spin_lock(&major_names_spinlock); |
1da177e4 LT |
313 | for (n = &major_names[index]; *n; n = &(*n)->next) |
314 | if ((*n)->major == major) | |
315 | break; | |
294462a5 AM |
316 | if (!*n || strcmp((*n)->name, name)) { |
317 | WARN_ON(1); | |
294462a5 | 318 | } else { |
1da177e4 LT |
319 | p = *n; |
320 | *n = p->next; | |
321 | } | |
dfbb3409 | 322 | spin_unlock(&major_names_spinlock); |
e49fbbbf | 323 | mutex_unlock(&major_names_lock); |
1da177e4 | 324 | kfree(p); |
1da177e4 LT |
325 | } |
326 | ||
327 | EXPORT_SYMBOL(unregister_blkdev); | |
328 | ||
7c3f828b | 329 | int blk_alloc_ext_minor(void) |
bcce3de1 | 330 | { |
bab998d6 | 331 | int idx; |
bcce3de1 | 332 | |
22ae8ce8 | 333 | idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL); |
c4b2b7d1 CH |
334 | if (idx == -ENOSPC) |
335 | return -EBUSY; | |
336 | return idx; | |
bcce3de1 TH |
337 | } |
338 | ||
7c3f828b | 339 | void blk_free_ext_minor(unsigned int minor) |
bcce3de1 | 340 | { |
c4b2b7d1 | 341 | ida_free(&ext_devt_ida, minor); |
6fcc44d1 YY |
342 | } |
343 | ||
1f014290 TH |
344 | static char *bdevt_str(dev_t devt, char *buf) |
345 | { | |
346 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
347 | char tbuf[BDEVT_SIZE]; | |
348 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
349 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
350 | } else | |
351 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
352 | ||
353 | return buf; | |
354 | } | |
355 | ||
bc359d03 CH |
356 | void disk_uevent(struct gendisk *disk, enum kobject_action action) |
357 | { | |
bc359d03 | 358 | struct block_device *part; |
3212135a | 359 | unsigned long idx; |
bc359d03 | 360 | |
3212135a CH |
361 | rcu_read_lock(); |
362 | xa_for_each(&disk->part_tbl, idx, part) { | |
363 | if (bdev_is_partition(part) && !bdev_nr_sectors(part)) | |
364 | continue; | |
498dcc13 | 365 | if (!kobject_get_unless_zero(&part->bd_device.kobj)) |
3212135a CH |
366 | continue; |
367 | ||
368 | rcu_read_unlock(); | |
bc359d03 | 369 | kobject_uevent(bdev_kobj(part), action); |
498dcc13 | 370 | put_device(&part->bd_device); |
3212135a CH |
371 | rcu_read_lock(); |
372 | } | |
373 | rcu_read_unlock(); | |
bc359d03 CH |
374 | } |
375 | EXPORT_SYMBOL_GPL(disk_uevent); | |
376 | ||
e16e506c | 377 | int disk_scan_partitions(struct gendisk *disk, fmode_t mode) |
9301fe73 CH |
378 | { |
379 | struct block_device *bdev; | |
380 | ||
9f18db57 | 381 | if (disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) |
e16e506c CH |
382 | return -EINVAL; |
383 | if (disk->open_partitions) | |
384 | return -EBUSY; | |
9301fe73 CH |
385 | |
386 | set_bit(GD_NEED_PART_SCAN, &disk->state); | |
e16e506c CH |
387 | bdev = blkdev_get_by_dev(disk_devt(disk), mode, NULL); |
388 | if (IS_ERR(bdev)) | |
389 | return PTR_ERR(bdev); | |
390 | blkdev_put(bdev, mode); | |
391 | return 0; | |
9301fe73 CH |
392 | } |
393 | ||
1da177e4 | 394 | /** |
d1254a87 | 395 | * device_add_disk - add disk information to kernel list |
e63a46be | 396 | * @parent: parent device for the disk |
1da177e4 | 397 | * @disk: per-device partitioning information |
fef912bf | 398 | * @groups: Additional per-device sysfs groups |
1da177e4 LT |
399 | * |
400 | * This function registers the partitioning information in @disk | |
401 | * with the kernel. | |
402 | */ | |
278167fd LC |
403 | int __must_check device_add_disk(struct device *parent, struct gendisk *disk, |
404 | const struct attribute_group **groups) | |
d1254a87 | 405 | |
1da177e4 | 406 | { |
52b85909 | 407 | struct device *ddev = disk_to_dev(disk); |
7c3f828b | 408 | int ret; |
cf0ca9fe | 409 | |
737eb78e DLM |
410 | /* |
411 | * The disk queue should now be all set with enough information about | |
412 | * the device for the elevator code to pick an adequate default | |
413 | * elevator if one is needed, that is, for devices requesting queue | |
414 | * registration. | |
415 | */ | |
d1254a87 | 416 | elevator_init_mq(disk->queue); |
737eb78e | 417 | |
7c3f828b CH |
418 | /* |
419 | * If the driver provides an explicit major number it also must provide | |
420 | * the number of minors numbers supported, and those will be used to | |
421 | * setup the gendisk. | |
422 | * Otherwise just allocate the device numbers for both the whole device | |
423 | * and all partitions from the extended dev_t space. | |
3e1a7ff8 | 424 | */ |
7c3f828b | 425 | if (disk->major) { |
83cbce95 LC |
426 | if (WARN_ON(!disk->minors)) |
427 | return -EINVAL; | |
2e3c73fa CH |
428 | |
429 | if (disk->minors > DISK_MAX_PARTS) { | |
430 | pr_err("block: can't allocate more than %d partitions\n", | |
431 | DISK_MAX_PARTS); | |
432 | disk->minors = DISK_MAX_PARTS; | |
433 | } | |
e338924b TH |
434 | if (disk->first_minor + disk->minors > MINORMASK + 1) |
435 | return -EINVAL; | |
7c3f828b | 436 | } else { |
83cbce95 LC |
437 | if (WARN_ON(disk->minors)) |
438 | return -EINVAL; | |
3e1a7ff8 | 439 | |
7c3f828b | 440 | ret = blk_alloc_ext_minor(); |
83cbce95 LC |
441 | if (ret < 0) |
442 | return ret; | |
7c3f828b | 443 | disk->major = BLOCK_EXT_MAJOR; |
539711d7 | 444 | disk->first_minor = ret; |
3e1a7ff8 | 445 | } |
7c3f828b | 446 | |
52b85909 CH |
447 | /* delay uevents, until we scanned partition table */ |
448 | dev_set_uevent_suppress(ddev, 1); | |
449 | ||
450 | ddev->parent = parent; | |
451 | ddev->groups = groups; | |
452 | dev_set_name(ddev, "%s", disk->disk_name); | |
8235b5c1 CH |
453 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
454 | ddev->devt = MKDEV(disk->major, disk->first_minor); | |
83cbce95 LC |
455 | ret = device_add(ddev); |
456 | if (ret) | |
99d8690a CH |
457 | goto out_free_ext_minor; |
458 | ||
459 | ret = disk_alloc_events(disk); | |
460 | if (ret) | |
461 | goto out_device_del; | |
462 | ||
52b85909 CH |
463 | if (!sysfs_deprecated) { |
464 | ret = sysfs_create_link(block_depr, &ddev->kobj, | |
465 | kobject_name(&ddev->kobj)); | |
83cbce95 LC |
466 | if (ret) |
467 | goto out_device_del; | |
52b85909 CH |
468 | } |
469 | ||
470 | /* | |
471 | * avoid probable deadlock caused by allocating memory with | |
472 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
473 | * devices | |
474 | */ | |
475 | pm_runtime_set_memalloc_noio(ddev, true); | |
476 | ||
83cbce95 LC |
477 | ret = blk_integrity_add(disk); |
478 | if (ret) | |
479 | goto out_del_block_link; | |
bab53f6b | 480 | |
52b85909 CH |
481 | disk->part0->bd_holder_dir = |
482 | kobject_create_and_add("holders", &ddev->kobj); | |
fe7d064f LC |
483 | if (!disk->part0->bd_holder_dir) { |
484 | ret = -ENOMEM; | |
83cbce95 | 485 | goto out_del_integrity; |
fe7d064f | 486 | } |
52b85909 | 487 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); |
fe7d064f LC |
488 | if (!disk->slave_dir) { |
489 | ret = -ENOMEM; | |
83cbce95 | 490 | goto out_put_holder_dir; |
fe7d064f | 491 | } |
52b85909 | 492 | |
83cbce95 LC |
493 | ret = bd_register_pending_holders(disk); |
494 | if (ret < 0) | |
495 | goto out_put_slave_dir; | |
52b85909 | 496 | |
83cbce95 LC |
497 | ret = blk_register_queue(disk); |
498 | if (ret) | |
499 | goto out_put_slave_dir; | |
75f4dca5 | 500 | |
9f18db57 | 501 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8235b5c1 CH |
502 | ret = bdi_register(disk->bdi, "%u:%u", |
503 | disk->major, disk->first_minor); | |
83cbce95 LC |
504 | if (ret) |
505 | goto out_unregister_queue; | |
8235b5c1 | 506 | bdi_set_owner(disk->bdi, ddev); |
83cbce95 LC |
507 | ret = sysfs_create_link(&ddev->kobj, |
508 | &disk->bdi->dev->kobj, "bdi"); | |
509 | if (ret) | |
510 | goto out_unregister_bdi; | |
8235b5c1 | 511 | |
9d5ee676 | 512 | bdev_add(disk->part0, ddev->devt); |
e16e506c CH |
513 | if (get_capacity(disk)) |
514 | disk_scan_partitions(disk, FMODE_READ); | |
52b85909 CH |
515 | |
516 | /* | |
517 | * Announce the disk and partitions after all partitions are | |
8235b5c1 | 518 | * created. (for hidden disks uevents remain suppressed forever) |
52b85909 CH |
519 | */ |
520 | dev_set_uevent_suppress(ddev, 0); | |
521 | disk_uevent(disk, KOBJ_ADD); | |
52b85909 CH |
522 | } |
523 | ||
75f4dca5 | 524 | disk_update_readahead(disk); |
77ea887e | 525 | disk_add_events(disk); |
83cbce95 LC |
526 | return 0; |
527 | ||
528 | out_unregister_bdi: | |
529 | if (!(disk->flags & GENHD_FL_HIDDEN)) | |
530 | bdi_unregister(disk->bdi); | |
531 | out_unregister_queue: | |
532 | blk_unregister_queue(disk); | |
533 | out_put_slave_dir: | |
534 | kobject_put(disk->slave_dir); | |
535 | out_put_holder_dir: | |
536 | kobject_put(disk->part0->bd_holder_dir); | |
537 | out_del_integrity: | |
538 | blk_integrity_del(disk); | |
539 | out_del_block_link: | |
540 | if (!sysfs_deprecated) | |
541 | sysfs_remove_link(block_depr, dev_name(ddev)); | |
542 | out_device_del: | |
543 | device_del(ddev); | |
83cbce95 LC |
544 | out_free_ext_minor: |
545 | if (disk->major == BLOCK_EXT_MAJOR) | |
546 | blk_free_ext_minor(disk->first_minor); | |
278167fd | 547 | return ret; |
1da177e4 | 548 | } |
e63a46be | 549 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 550 | |
b5bd357c LC |
551 | /** |
552 | * del_gendisk - remove the gendisk | |
553 | * @disk: the struct gendisk to remove | |
554 | * | |
555 | * Removes the gendisk and all its associated resources. This deletes the | |
556 | * partitions associated with the gendisk, and unregisters the associated | |
557 | * request_queue. | |
558 | * | |
559 | * This is the counter to the respective __device_add_disk() call. | |
560 | * | |
561 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
562 | * with put_disk(), which should be called after del_gendisk(), if | |
563 | * __device_add_disk() was used. | |
e8c7d14a LC |
564 | * |
565 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
566 | * it should not be deferred. | |
567 | * | |
568 | * Context: can sleep | |
b5bd357c | 569 | */ |
d2bf1b67 | 570 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 571 | { |
8e141f9e CH |
572 | struct request_queue *q = disk->queue; |
573 | ||
e8c7d14a LC |
574 | might_sleep(); |
575 | ||
9f286992 | 576 | if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN))) |
6b3ba976 CH |
577 | return; |
578 | ||
25520d55 | 579 | blk_integrity_del(disk); |
77ea887e TH |
580 | disk_del_events(disk); |
581 | ||
a8698707 | 582 | mutex_lock(&disk->open_mutex); |
d7a66574 | 583 | remove_inode_hash(disk->part0->bd_inode); |
d3c4a43d | 584 | blk_drop_partitions(disk); |
a8698707 | 585 | mutex_unlock(&disk->open_mutex); |
c76f48eb | 586 | |
45611837 CH |
587 | fsync_bdev(disk->part0); |
588 | __invalidate_device(disk->part0, true); | |
589 | ||
8e141f9e CH |
590 | /* |
591 | * Fail any new I/O. | |
592 | */ | |
593 | set_bit(GD_DEAD, &disk->state); | |
d2bf1b67 | 594 | set_capacity(disk, 0); |
d2bf1b67 | 595 | |
8e141f9e CH |
596 | /* |
597 | * Prevent new I/O from crossing bio_queue_enter(). | |
598 | */ | |
599 | blk_queue_start_drain(q); | |
8e141f9e | 600 | |
6b3ba976 | 601 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 | 602 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
6b3ba976 | 603 | |
90f16fdd JK |
604 | /* |
605 | * Unregister bdi before releasing device numbers (as they can | |
606 | * get reused and we'd get clashes in sysfs). | |
607 | */ | |
edb0872f | 608 | bdi_unregister(disk->bdi); |
90f16fdd | 609 | } |
d2bf1b67 | 610 | |
6b3ba976 | 611 | blk_unregister_queue(disk); |
d2bf1b67 | 612 | |
cb8432d6 | 613 | kobject_put(disk->part0->bd_holder_dir); |
d2bf1b67 | 614 | kobject_put(disk->slave_dir); |
d2bf1b67 | 615 | |
8446fe92 | 616 | part_stat_set_all(disk->part0, 0); |
cb8432d6 | 617 | disk->part0->bd_stamp = 0; |
d2bf1b67 TH |
618 | if (!sysfs_deprecated) |
619 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 620 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 621 | device_del(disk_to_dev(disk)); |
d308ae0d ML |
622 | |
623 | blk_mq_freeze_queue_wait(q); | |
624 | ||
625 | rq_qos_exit(q); | |
626 | blk_sync_queue(q); | |
627 | blk_flush_integrity(); | |
628 | /* | |
629 | * Allow using passthrough request again after the queue is torn down. | |
630 | */ | |
631 | blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q); | |
632 | __blk_mq_unfreeze_queue(q, true); | |
633 | ||
1da177e4 | 634 | } |
d2bf1b67 | 635 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 636 | |
f059a1d2 XY |
637 | /** |
638 | * invalidate_disk - invalidate the disk | |
639 | * @disk: the struct gendisk to invalidate | |
640 | * | |
641 | * A helper to invalidates the disk. It will clean the disk's associated | |
642 | * buffer/page caches and reset its internal states so that the disk | |
643 | * can be reused by the drivers. | |
644 | * | |
645 | * Context: can sleep | |
646 | */ | |
647 | void invalidate_disk(struct gendisk *disk) | |
648 | { | |
649 | struct block_device *bdev = disk->part0; | |
650 | ||
651 | invalidate_bdev(bdev); | |
652 | bdev->bd_inode->i_mapping->wb_err = 0; | |
653 | set_capacity(disk, 0); | |
654 | } | |
655 | EXPORT_SYMBOL(invalidate_disk); | |
656 | ||
99e6608c VV |
657 | /* sysfs access to bad-blocks list. */ |
658 | static ssize_t disk_badblocks_show(struct device *dev, | |
659 | struct device_attribute *attr, | |
660 | char *page) | |
661 | { | |
662 | struct gendisk *disk = dev_to_disk(dev); | |
663 | ||
664 | if (!disk->bb) | |
665 | return sprintf(page, "\n"); | |
666 | ||
667 | return badblocks_show(disk->bb, page, 0); | |
668 | } | |
669 | ||
670 | static ssize_t disk_badblocks_store(struct device *dev, | |
671 | struct device_attribute *attr, | |
672 | const char *page, size_t len) | |
673 | { | |
674 | struct gendisk *disk = dev_to_disk(dev); | |
675 | ||
676 | if (!disk->bb) | |
677 | return -ENXIO; | |
678 | ||
679 | return badblocks_store(disk->bb, page, len, 0); | |
680 | } | |
681 | ||
22ae8ce8 | 682 | void blk_request_module(dev_t devt) |
bd8eff3b | 683 | { |
a160c615 CH |
684 | unsigned int major = MAJOR(devt); |
685 | struct blk_major_name **n; | |
686 | ||
687 | mutex_lock(&major_names_lock); | |
688 | for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { | |
689 | if ((*n)->major == major && (*n)->probe) { | |
690 | (*n)->probe(devt); | |
691 | mutex_unlock(&major_names_lock); | |
692 | return; | |
693 | } | |
694 | } | |
695 | mutex_unlock(&major_names_lock); | |
696 | ||
bd8eff3b CH |
697 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
698 | /* Make old-style 2.4 aliases work */ | |
699 | request_module("block-major-%d", MAJOR(devt)); | |
700 | } | |
701 | ||
5c6f35c5 GKH |
702 | /* |
703 | * print a full list of all partitions - intended for places where the root | |
704 | * filesystem can't be mounted and thus to give the victim some idea of what | |
705 | * went wrong | |
706 | */ | |
707 | void __init printk_all_partitions(void) | |
708 | { | |
def4e38d TH |
709 | struct class_dev_iter iter; |
710 | struct device *dev; | |
711 | ||
712 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
713 | while ((dev = class_dev_iter_next(&iter))) { | |
714 | struct gendisk *disk = dev_to_disk(dev); | |
ad1eaa53 | 715 | struct block_device *part; |
1f014290 | 716 | char devt_buf[BDEVT_SIZE]; |
e559f58d | 717 | unsigned long idx; |
def4e38d TH |
718 | |
719 | /* | |
720 | * Don't show empty devices or things that have been | |
25985edc | 721 | * suppressed |
def4e38d | 722 | */ |
3b5149ac | 723 | if (get_capacity(disk) == 0 || (disk->flags & GENHD_FL_HIDDEN)) |
def4e38d TH |
724 | continue; |
725 | ||
726 | /* | |
e559f58d CH |
727 | * Note, unlike /proc/partitions, I am showing the numbers in |
728 | * hex - the same format as the root= option takes. | |
def4e38d | 729 | */ |
e559f58d CH |
730 | rcu_read_lock(); |
731 | xa_for_each(&disk->part_tbl, idx, part) { | |
732 | if (!bdev_nr_sectors(part)) | |
733 | continue; | |
a9e7bc3d | 734 | printk("%s%s %10llu %pg %s", |
e559f58d | 735 | bdev_is_partition(part) ? " " : "", |
ad1eaa53 | 736 | bdevt_str(part->bd_dev, devt_buf), |
a9e7bc3d | 737 | bdev_nr_sectors(part) >> 1, part, |
ad1eaa53 CH |
738 | part->bd_meta_info ? |
739 | part->bd_meta_info->uuid : ""); | |
e559f58d | 740 | if (bdev_is_partition(part)) |
074a7aca | 741 | printk("\n"); |
e559f58d CH |
742 | else if (dev->parent && dev->parent->driver) |
743 | printk(" driver: %s\n", | |
744 | dev->parent->driver->name); | |
745 | else | |
746 | printk(" (driver?)\n"); | |
074a7aca | 747 | } |
e559f58d | 748 | rcu_read_unlock(); |
def4e38d TH |
749 | } |
750 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
751 | } |
752 | ||
1da177e4 LT |
753 | #ifdef CONFIG_PROC_FS |
754 | /* iterator */ | |
def4e38d | 755 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 756 | { |
def4e38d TH |
757 | loff_t skip = *pos; |
758 | struct class_dev_iter *iter; | |
759 | struct device *dev; | |
68c4d4a7 | 760 | |
aeb3d3a8 | 761 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
762 | if (!iter) |
763 | return ERR_PTR(-ENOMEM); | |
764 | ||
765 | seqf->private = iter; | |
766 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
767 | do { | |
768 | dev = class_dev_iter_next(iter); | |
769 | if (!dev) | |
770 | return NULL; | |
771 | } while (skip--); | |
772 | ||
773 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
774 | } |
775 | ||
def4e38d | 776 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 777 | { |
edfaa7c3 | 778 | struct device *dev; |
1da177e4 | 779 | |
def4e38d TH |
780 | (*pos)++; |
781 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 782 | if (dev) |
68c4d4a7 | 783 | return dev_to_disk(dev); |
2ac3cee5 | 784 | |
1da177e4 LT |
785 | return NULL; |
786 | } | |
787 | ||
def4e38d | 788 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 789 | { |
def4e38d | 790 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 791 | |
def4e38d TH |
792 | /* stop is called even after start failed :-( */ |
793 | if (iter) { | |
794 | class_dev_iter_exit(iter); | |
795 | kfree(iter); | |
77da1605 | 796 | seqf->private = NULL; |
5c0ef6d0 | 797 | } |
1da177e4 LT |
798 | } |
799 | ||
def4e38d | 800 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 801 | { |
06768067 | 802 | void *p; |
def4e38d TH |
803 | |
804 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 805 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
806 | seq_puts(seqf, "major minor #blocks name\n\n"); |
807 | return p; | |
1da177e4 LT |
808 | } |
809 | ||
cf771cb5 | 810 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
811 | { |
812 | struct gendisk *sgp = v; | |
ad1eaa53 | 813 | struct block_device *part; |
ecc75a98 | 814 | unsigned long idx; |
1da177e4 | 815 | |
3b5149ac | 816 | if (!get_capacity(sgp) || (sgp->flags & GENHD_FL_HIDDEN)) |
1da177e4 LT |
817 | return 0; |
818 | ||
ecc75a98 CH |
819 | rcu_read_lock(); |
820 | xa_for_each(&sgp->part_tbl, idx, part) { | |
821 | if (!bdev_nr_sectors(part)) | |
822 | continue; | |
a291bb43 | 823 | seq_printf(seqf, "%4d %7d %10llu %pg\n", |
ad1eaa53 | 824 | MAJOR(part->bd_dev), MINOR(part->bd_dev), |
a291bb43 | 825 | bdev_nr_sectors(part) >> 1, part); |
ecc75a98 CH |
826 | } |
827 | rcu_read_unlock(); | |
1da177e4 LT |
828 | return 0; |
829 | } | |
830 | ||
f500975a | 831 | static const struct seq_operations partitions_op = { |
def4e38d TH |
832 | .start = show_partition_start, |
833 | .next = disk_seqf_next, | |
834 | .stop = disk_seqf_stop, | |
edfaa7c3 | 835 | .show = show_partition |
1da177e4 LT |
836 | }; |
837 | #endif | |
838 | ||
1da177e4 LT |
839 | static int __init genhd_device_init(void) |
840 | { | |
e105b8bf DW |
841 | int error; |
842 | ||
843 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
844 | error = class_register(&block_class); | |
ee27a558 RM |
845 | if (unlikely(error)) |
846 | return error; | |
1da177e4 | 847 | blk_dev_init(); |
edfaa7c3 | 848 | |
561ec68e ZY |
849 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
850 | ||
edfaa7c3 | 851 | /* create top-level block dir */ |
e52eec13 AK |
852 | if (!sysfs_deprecated) |
853 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 854 | return 0; |
1da177e4 LT |
855 | } |
856 | ||
857 | subsys_initcall(genhd_device_init); | |
858 | ||
edfaa7c3 KS |
859 | static ssize_t disk_range_show(struct device *dev, |
860 | struct device_attribute *attr, char *buf) | |
1da177e4 | 861 | { |
edfaa7c3 | 862 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 863 | |
edfaa7c3 | 864 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
865 | } |
866 | ||
1f014290 TH |
867 | static ssize_t disk_ext_range_show(struct device *dev, |
868 | struct device_attribute *attr, char *buf) | |
869 | { | |
870 | struct gendisk *disk = dev_to_disk(dev); | |
871 | ||
1ebe2e5f CH |
872 | return sprintf(buf, "%d\n", |
873 | (disk->flags & GENHD_FL_NO_PART) ? 1 : DISK_MAX_PARTS); | |
1f014290 TH |
874 | } |
875 | ||
edfaa7c3 KS |
876 | static ssize_t disk_removable_show(struct device *dev, |
877 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 878 | { |
edfaa7c3 | 879 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 880 | |
edfaa7c3 KS |
881 | return sprintf(buf, "%d\n", |
882 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
883 | } |
884 | ||
8ddcd653 CH |
885 | static ssize_t disk_hidden_show(struct device *dev, |
886 | struct device_attribute *attr, char *buf) | |
887 | { | |
888 | struct gendisk *disk = dev_to_disk(dev); | |
889 | ||
890 | return sprintf(buf, "%d\n", | |
891 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
892 | } | |
893 | ||
1c9ce527 KS |
894 | static ssize_t disk_ro_show(struct device *dev, |
895 | struct device_attribute *attr, char *buf) | |
896 | { | |
897 | struct gendisk *disk = dev_to_disk(dev); | |
898 | ||
b7db9956 | 899 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
900 | } |
901 | ||
3ad5cee5 CH |
902 | ssize_t part_size_show(struct device *dev, |
903 | struct device_attribute *attr, char *buf) | |
904 | { | |
0d02129e | 905 | return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev))); |
3ad5cee5 CH |
906 | } |
907 | ||
908 | ssize_t part_stat_show(struct device *dev, | |
909 | struct device_attribute *attr, char *buf) | |
910 | { | |
0d02129e | 911 | struct block_device *bdev = dev_to_bdev(dev); |
ed6cddef | 912 | struct request_queue *q = bdev_get_queue(bdev); |
ea18e0f0 | 913 | struct disk_stats stat; |
3ad5cee5 CH |
914 | unsigned int inflight; |
915 | ||
0d02129e | 916 | part_stat_read_all(bdev, &stat); |
b2f609e1 | 917 | if (queue_is_mq(q)) |
0d02129e | 918 | inflight = blk_mq_in_flight(q, bdev); |
b2f609e1 | 919 | else |
0d02129e | 920 | inflight = part_in_flight(bdev); |
ea18e0f0 | 921 | |
3ad5cee5 CH |
922 | return sprintf(buf, |
923 | "%8lu %8lu %8llu %8u " | |
924 | "%8lu %8lu %8llu %8u " | |
925 | "%8u %8u %8u " | |
926 | "%8lu %8lu %8llu %8u " | |
927 | "%8lu %8u" | |
928 | "\n", | |
ea18e0f0 KK |
929 | stat.ios[STAT_READ], |
930 | stat.merges[STAT_READ], | |
931 | (unsigned long long)stat.sectors[STAT_READ], | |
932 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
933 | stat.ios[STAT_WRITE], | |
934 | stat.merges[STAT_WRITE], | |
935 | (unsigned long long)stat.sectors[STAT_WRITE], | |
936 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 937 | inflight, |
ea18e0f0 | 938 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
939 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
940 | stat.nsecs[STAT_WRITE] + | |
941 | stat.nsecs[STAT_DISCARD] + | |
942 | stat.nsecs[STAT_FLUSH], | |
943 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
944 | stat.ios[STAT_DISCARD], |
945 | stat.merges[STAT_DISCARD], | |
946 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
947 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
948 | stat.ios[STAT_FLUSH], | |
949 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
950 | } |
951 | ||
952 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
953 | char *buf) | |
954 | { | |
0d02129e | 955 | struct block_device *bdev = dev_to_bdev(dev); |
ed6cddef | 956 | struct request_queue *q = bdev_get_queue(bdev); |
3ad5cee5 CH |
957 | unsigned int inflight[2]; |
958 | ||
b2f609e1 | 959 | if (queue_is_mq(q)) |
0d02129e | 960 | blk_mq_in_flight_rw(q, bdev, inflight); |
b2f609e1 | 961 | else |
0d02129e | 962 | part_in_flight_rw(bdev, inflight); |
b2f609e1 | 963 | |
3ad5cee5 CH |
964 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
965 | } | |
966 | ||
edfaa7c3 KS |
967 | static ssize_t disk_capability_show(struct device *dev, |
968 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 969 | { |
edfaa7c3 KS |
970 | struct gendisk *disk = dev_to_disk(dev); |
971 | ||
972 | return sprintf(buf, "%x\n", disk->flags); | |
86ce18d7 | 973 | } |
edfaa7c3 | 974 | |
c72758f3 MP |
975 | static ssize_t disk_alignment_offset_show(struct device *dev, |
976 | struct device_attribute *attr, | |
977 | char *buf) | |
978 | { | |
979 | struct gendisk *disk = dev_to_disk(dev); | |
980 | ||
981 | return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); | |
982 | } | |
983 | ||
86b37281 MP |
984 | static ssize_t disk_discard_alignment_show(struct device *dev, |
985 | struct device_attribute *attr, | |
986 | char *buf) | |
987 | { | |
988 | struct gendisk *disk = dev_to_disk(dev); | |
989 | ||
dd3d145d | 990 | return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); |
86b37281 MP |
991 | } |
992 | ||
13927b31 MC |
993 | static ssize_t diskseq_show(struct device *dev, |
994 | struct device_attribute *attr, char *buf) | |
995 | { | |
996 | struct gendisk *disk = dev_to_disk(dev); | |
997 | ||
998 | return sprintf(buf, "%llu\n", disk->diskseq); | |
999 | } | |
1000 | ||
5657a819 JP |
1001 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1002 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1003 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1004 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1005 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1006 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1007 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1008 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1009 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1010 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1011 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1012 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
13927b31 | 1013 | static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL); |
3ad5cee5 | 1014 | |
c17bb495 | 1015 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1016 | ssize_t part_fail_show(struct device *dev, |
1017 | struct device_attribute *attr, char *buf) | |
1018 | { | |
0d02129e | 1019 | return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail); |
3ad5cee5 CH |
1020 | } |
1021 | ||
1022 | ssize_t part_fail_store(struct device *dev, | |
1023 | struct device_attribute *attr, | |
1024 | const char *buf, size_t count) | |
1025 | { | |
3ad5cee5 CH |
1026 | int i; |
1027 | ||
1028 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
0d02129e | 1029 | dev_to_bdev(dev)->bd_make_it_fail = i; |
3ad5cee5 CH |
1030 | |
1031 | return count; | |
1032 | } | |
1033 | ||
edfaa7c3 | 1034 | static struct device_attribute dev_attr_fail = |
5657a819 | 1035 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1036 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1037 | ||
581d4e28 JA |
1038 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1039 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1040 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1041 | #endif |
edfaa7c3 KS |
1042 | |
1043 | static struct attribute *disk_attrs[] = { | |
1044 | &dev_attr_range.attr, | |
1f014290 | 1045 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1046 | &dev_attr_removable.attr, |
8ddcd653 | 1047 | &dev_attr_hidden.attr, |
1c9ce527 | 1048 | &dev_attr_ro.attr, |
edfaa7c3 | 1049 | &dev_attr_size.attr, |
c72758f3 | 1050 | &dev_attr_alignment_offset.attr, |
86b37281 | 1051 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1052 | &dev_attr_capability.attr, |
1053 | &dev_attr_stat.attr, | |
316d315b | 1054 | &dev_attr_inflight.attr, |
99e6608c | 1055 | &dev_attr_badblocks.attr, |
2bc8cda5 CH |
1056 | &dev_attr_events.attr, |
1057 | &dev_attr_events_async.attr, | |
1058 | &dev_attr_events_poll_msecs.attr, | |
13927b31 | 1059 | &dev_attr_diskseq.attr, |
edfaa7c3 KS |
1060 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1061 | &dev_attr_fail.attr, | |
581d4e28 JA |
1062 | #endif |
1063 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1064 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1065 | #endif |
1066 | NULL | |
1067 | }; | |
1068 | ||
9438b3e0 DW |
1069 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1070 | { | |
1071 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1072 | struct gendisk *disk = dev_to_disk(dev); | |
1073 | ||
1074 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1075 | return 0; | |
1076 | return a->mode; | |
1077 | } | |
1078 | ||
edfaa7c3 KS |
1079 | static struct attribute_group disk_attr_group = { |
1080 | .attrs = disk_attrs, | |
9438b3e0 | 1081 | .is_visible = disk_visible, |
edfaa7c3 KS |
1082 | }; |
1083 | ||
a4dbd674 | 1084 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 KS |
1085 | &disk_attr_group, |
1086 | NULL | |
1da177e4 LT |
1087 | }; |
1088 | ||
b5bd357c LC |
1089 | /** |
1090 | * disk_release - releases all allocated resources of the gendisk | |
1091 | * @dev: the device representing this disk | |
1092 | * | |
1093 | * This function releases all allocated resources of the gendisk. | |
1094 | * | |
b5bd357c LC |
1095 | * Drivers which used __device_add_disk() have a gendisk with a request_queue |
1096 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1097 | * drivers we also call blk_put_queue() for them, and we expect the | |
1098 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1099 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1100 | * |
1101 | * Context: can sleep | |
b5bd357c | 1102 | */ |
edfaa7c3 | 1103 | static void disk_release(struct device *dev) |
1da177e4 | 1104 | { |
edfaa7c3 KS |
1105 | struct gendisk *disk = dev_to_disk(dev); |
1106 | ||
e8c7d14a | 1107 | might_sleep(); |
a2041761 | 1108 | WARN_ON_ONCE(disk_live(disk)); |
e8c7d14a | 1109 | |
2a19b28f ML |
1110 | blk_mq_cancel_work_sync(disk->queue); |
1111 | ||
77ea887e | 1112 | disk_release_events(disk); |
1da177e4 | 1113 | kfree(disk->random); |
a33df75c | 1114 | xa_destroy(&disk->part_tbl); |
d152c682 | 1115 | disk->queue->disk = NULL; |
61a35cfc | 1116 | blk_put_queue(disk->queue); |
2f4731dc | 1117 | iput(disk->part0->bd_inode); /* frees the disk */ |
1da177e4 | 1118 | } |
87eb7107 MC |
1119 | |
1120 | static int block_uevent(struct device *dev, struct kobj_uevent_env *env) | |
1121 | { | |
1122 | struct gendisk *disk = dev_to_disk(dev); | |
1123 | ||
1124 | return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq); | |
1125 | } | |
1126 | ||
edfaa7c3 KS |
1127 | struct class block_class = { |
1128 | .name = "block", | |
87eb7107 | 1129 | .dev_uevent = block_uevent, |
1da177e4 LT |
1130 | }; |
1131 | ||
3c2670e6 | 1132 | static char *block_devnode(struct device *dev, umode_t *mode, |
4e4098a3 | 1133 | kuid_t *uid, kgid_t *gid) |
b03f38b6 KS |
1134 | { |
1135 | struct gendisk *disk = dev_to_disk(dev); | |
1136 | ||
348e114b CH |
1137 | if (disk->fops->devnode) |
1138 | return disk->fops->devnode(disk, mode); | |
b03f38b6 KS |
1139 | return NULL; |
1140 | } | |
1141 | ||
ef45fe47 | 1142 | const struct device_type disk_type = { |
edfaa7c3 KS |
1143 | .name = "disk", |
1144 | .groups = disk_attr_groups, | |
1145 | .release = disk_release, | |
e454cea2 | 1146 | .devnode = block_devnode, |
1da177e4 LT |
1147 | }; |
1148 | ||
a6e2ba88 | 1149 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1150 | /* |
1151 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1152 | * entries do, above, but makes them available through one seq_file. | |
1153 | * | |
1154 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1155 | * extra fields. | |
1156 | */ | |
1157 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1158 | { |
1159 | struct gendisk *gp = v; | |
ad1eaa53 | 1160 | struct block_device *hd; |
e016b782 | 1161 | unsigned int inflight; |
ea18e0f0 | 1162 | struct disk_stats stat; |
7fae67cc | 1163 | unsigned long idx; |
1da177e4 LT |
1164 | |
1165 | /* | |
ed9e1982 | 1166 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1167 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1168 | " rio rmerge rsect ruse wio wmerge " |
1169 | "wsect wuse running use aveq" | |
1170 | "\n\n"); | |
1171 | */ | |
9f5e4865 | 1172 | |
7fae67cc CH |
1173 | rcu_read_lock(); |
1174 | xa_for_each(&gp->part_tbl, idx, hd) { | |
1175 | if (bdev_is_partition(hd) && !bdev_nr_sectors(hd)) | |
1176 | continue; | |
0d02129e | 1177 | part_stat_read_all(hd, &stat); |
b2f609e1 | 1178 | if (queue_is_mq(gp->queue)) |
ad1eaa53 | 1179 | inflight = blk_mq_in_flight(gp->queue, hd); |
b2f609e1 | 1180 | else |
ad1eaa53 | 1181 | inflight = part_in_flight(hd); |
ea18e0f0 | 1182 | |
26e2d7a3 | 1183 | seq_printf(seqf, "%4d %7d %pg " |
bdca3c87 MC |
1184 | "%lu %lu %lu %u " |
1185 | "%lu %lu %lu %u " | |
1186 | "%u %u %u " | |
b6866318 KK |
1187 | "%lu %lu %lu %u " |
1188 | "%lu %u" | |
1189 | "\n", | |
26e2d7a3 | 1190 | MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd, |
ea18e0f0 KK |
1191 | stat.ios[STAT_READ], |
1192 | stat.merges[STAT_READ], | |
1193 | stat.sectors[STAT_READ], | |
1194 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1195 | NSEC_PER_MSEC), | |
1196 | stat.ios[STAT_WRITE], | |
1197 | stat.merges[STAT_WRITE], | |
1198 | stat.sectors[STAT_WRITE], | |
1199 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1200 | NSEC_PER_MSEC), | |
e016b782 | 1201 | inflight, |
ea18e0f0 | 1202 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1203 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1204 | stat.nsecs[STAT_WRITE] + | |
1205 | stat.nsecs[STAT_DISCARD] + | |
1206 | stat.nsecs[STAT_FLUSH], | |
1207 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1208 | stat.ios[STAT_DISCARD], |
1209 | stat.merges[STAT_DISCARD], | |
1210 | stat.sectors[STAT_DISCARD], | |
1211 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1212 | NSEC_PER_MSEC), | |
1213 | stat.ios[STAT_FLUSH], | |
1214 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1215 | NSEC_PER_MSEC) | |
28f39d55 | 1216 | ); |
1da177e4 | 1217 | } |
7fae67cc | 1218 | rcu_read_unlock(); |
9f5e4865 | 1219 | |
1da177e4 LT |
1220 | return 0; |
1221 | } | |
1222 | ||
31d85ab2 | 1223 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1224 | .start = disk_seqf_start, |
1225 | .next = disk_seqf_next, | |
1226 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1227 | .show = diskstats_show |
1228 | }; | |
f500975a AD |
1229 | |
1230 | static int __init proc_genhd_init(void) | |
1231 | { | |
fddda2b7 CH |
1232 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1233 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1234 | return 0; |
1235 | } | |
1236 | module_init(proc_genhd_init); | |
a6e2ba88 | 1237 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1238 | |
c97d93c3 CH |
1239 | dev_t part_devt(struct gendisk *disk, u8 partno) |
1240 | { | |
0e0ccdec | 1241 | struct block_device *part; |
c97d93c3 CH |
1242 | dev_t devt = 0; |
1243 | ||
0e0ccdec CH |
1244 | rcu_read_lock(); |
1245 | part = xa_load(&disk->part_tbl, partno); | |
1246 | if (part) | |
c97d93c3 | 1247 | devt = part->bd_dev; |
0e0ccdec | 1248 | rcu_read_unlock(); |
c97d93c3 CH |
1249 | |
1250 | return devt; | |
1251 | } | |
1252 | ||
cf771cb5 | 1253 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1254 | { |
def4e38d TH |
1255 | dev_t devt = MKDEV(0, 0); |
1256 | struct class_dev_iter iter; | |
1257 | struct device *dev; | |
a142be85 | 1258 | |
def4e38d TH |
1259 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1260 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1261 | struct gendisk *disk = dev_to_disk(dev); |
a142be85 | 1262 | |
3ada8b7e | 1263 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1264 | continue; |
f331c029 | 1265 | |
41b8c853 NB |
1266 | if (partno < disk->minors) { |
1267 | /* We need to return the right devno, even | |
1268 | * if the partition doesn't exist yet. | |
1269 | */ | |
1270 | devt = MKDEV(MAJOR(dev->devt), | |
1271 | MINOR(dev->devt) + partno); | |
c97d93c3 CH |
1272 | } else { |
1273 | devt = part_devt(disk, partno); | |
1274 | if (devt) | |
1275 | break; | |
def4e38d | 1276 | } |
5c0ef6d0 | 1277 | } |
def4e38d | 1278 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1279 | return devt; |
1280 | } | |
edfaa7c3 | 1281 | |
4a1fa41d CH |
1282 | struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id, |
1283 | struct lock_class_key *lkclass) | |
1946089a CL |
1284 | { |
1285 | struct gendisk *disk; | |
1286 | ||
61a35cfc CH |
1287 | if (!blk_get_queue(q)) |
1288 | return NULL; | |
1289 | ||
c1b511eb | 1290 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 | 1291 | if (!disk) |
61a35cfc | 1292 | goto out_put_queue; |
6c23a968 | 1293 | |
edb0872f CH |
1294 | disk->bdi = bdi_alloc(node_id); |
1295 | if (!disk->bdi) | |
1296 | goto out_free_disk; | |
1297 | ||
17220ca5 PB |
1298 | /* bdev_alloc() might need the queue, set before the first call */ |
1299 | disk->queue = q; | |
1300 | ||
cb8432d6 CH |
1301 | disk->part0 = bdev_alloc(disk, 0); |
1302 | if (!disk->part0) | |
edb0872f | 1303 | goto out_free_bdi; |
22ae8ce8 | 1304 | |
f93af2a4 | 1305 | disk->node_id = node_id; |
a8698707 | 1306 | mutex_init(&disk->open_mutex); |
a33df75c CH |
1307 | xa_init(&disk->part_tbl); |
1308 | if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL)) | |
1309 | goto out_destroy_part_tbl; | |
f93af2a4 | 1310 | |
f93af2a4 CH |
1311 | rand_initialize_disk(disk); |
1312 | disk_to_dev(disk)->class = &block_class; | |
1313 | disk_to_dev(disk)->type = &disk_type; | |
1314 | device_initialize(disk_to_dev(disk)); | |
cf179948 | 1315 | inc_diskseq(disk); |
d152c682 | 1316 | q->disk = disk; |
4dcc4874 | 1317 | lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0); |
0dbcfe24 CH |
1318 | #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED |
1319 | INIT_LIST_HEAD(&disk->slave_bdevs); | |
1320 | #endif | |
1da177e4 | 1321 | return disk; |
f93af2a4 | 1322 | |
a33df75c CH |
1323 | out_destroy_part_tbl: |
1324 | xa_destroy(&disk->part_tbl); | |
06cc978d | 1325 | disk->part0->bd_disk = NULL; |
2f4731dc | 1326 | iput(disk->part0->bd_inode); |
edb0872f CH |
1327 | out_free_bdi: |
1328 | bdi_put(disk->bdi); | |
f93af2a4 CH |
1329 | out_free_disk: |
1330 | kfree(disk); | |
61a35cfc CH |
1331 | out_put_queue: |
1332 | blk_put_queue(q); | |
f93af2a4 | 1333 | return NULL; |
1da177e4 | 1334 | } |
e319e1fb | 1335 | EXPORT_SYMBOL(__alloc_disk_node); |
1da177e4 | 1336 | |
4dcc4874 | 1337 | struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass) |
f525464a CH |
1338 | { |
1339 | struct request_queue *q; | |
1340 | struct gendisk *disk; | |
1341 | ||
704b914f | 1342 | q = blk_alloc_queue(node, false); |
f525464a CH |
1343 | if (!q) |
1344 | return NULL; | |
1345 | ||
4a1fa41d | 1346 | disk = __alloc_disk_node(q, node, lkclass); |
f525464a CH |
1347 | if (!disk) { |
1348 | blk_cleanup_queue(q); | |
1349 | return NULL; | |
1350 | } | |
f525464a CH |
1351 | return disk; |
1352 | } | |
1353 | EXPORT_SYMBOL(__blk_alloc_disk); | |
1354 | ||
b5bd357c LC |
1355 | /** |
1356 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1357 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1358 | * |
1359 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1360 | * we'll have disk_release() called. | |
e8c7d14a LC |
1361 | * |
1362 | * Context: Any context, but the last reference must not be dropped from | |
1363 | * atomic context. | |
b5bd357c | 1364 | */ |
1da177e4 LT |
1365 | void put_disk(struct gendisk *disk) |
1366 | { | |
1367 | if (disk) | |
efdc41c8 | 1368 | put_device(disk_to_dev(disk)); |
1da177e4 | 1369 | } |
1da177e4 LT |
1370 | EXPORT_SYMBOL(put_disk); |
1371 | ||
f525464a CH |
1372 | /** |
1373 | * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk | |
1374 | * @disk: gendisk to shutdown | |
1375 | * | |
1376 | * Mark the queue hanging off @disk DYING, drain all pending requests, then mark | |
1377 | * the queue DEAD, destroy and put it and the gendisk structure. | |
1378 | * | |
1379 | * Context: can sleep | |
1380 | */ | |
1381 | void blk_cleanup_disk(struct gendisk *disk) | |
1382 | { | |
1383 | blk_cleanup_queue(disk->queue); | |
1384 | put_disk(disk); | |
1385 | } | |
1386 | EXPORT_SYMBOL(blk_cleanup_disk); | |
1387 | ||
e3264a4d HR |
1388 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1389 | { | |
1390 | char event[] = "DISK_RO=1"; | |
1391 | char *envp[] = { event, NULL }; | |
1392 | ||
1393 | if (!ro) | |
1394 | event[8] = '0'; | |
1395 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1396 | } | |
1397 | ||
52f019d4 CH |
1398 | /** |
1399 | * set_disk_ro - set a gendisk read-only | |
1400 | * @disk: gendisk to operate on | |
7f31bee3 | 1401 | * @read_only: %true to set the disk read-only, %false set the disk read/write |
52f019d4 CH |
1402 | * |
1403 | * This function is used to indicate whether a given disk device should have its | |
1404 | * read-only flag set. set_disk_ro() is typically used by device drivers to | |
1405 | * indicate whether the underlying physical device is write-protected. | |
1406 | */ | |
1407 | void set_disk_ro(struct gendisk *disk, bool read_only) | |
1da177e4 | 1408 | { |
52f019d4 CH |
1409 | if (read_only) { |
1410 | if (test_and_set_bit(GD_READ_ONLY, &disk->state)) | |
1411 | return; | |
1412 | } else { | |
1413 | if (!test_and_clear_bit(GD_READ_ONLY, &disk->state)) | |
1414 | return; | |
e3264a4d | 1415 | } |
52f019d4 | 1416 | set_disk_ro_uevent(disk, read_only); |
1da177e4 | 1417 | } |
1da177e4 LT |
1418 | EXPORT_SYMBOL(set_disk_ro); |
1419 | ||
cf179948 MC |
1420 | void inc_diskseq(struct gendisk *disk) |
1421 | { | |
1422 | disk->diskseq = atomic64_inc_return(&diskseq); | |
1423 | } |