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 | 10 | #include <linux/fs.h> |
b446b60e | 11 | #include <linux/kdev_t.h> |
1da177e4 LT |
12 | #include <linux/kernel.h> |
13 | #include <linux/blkdev.h> | |
66114cad | 14 | #include <linux/backing-dev.h> |
1da177e4 LT |
15 | #include <linux/init.h> |
16 | #include <linux/spinlock.h> | |
f500975a | 17 | #include <linux/proc_fs.h> |
1da177e4 LT |
18 | #include <linux/seq_file.h> |
19 | #include <linux/slab.h> | |
20 | #include <linux/kmod.h> | |
b81e0c23 | 21 | #include <linux/major.h> |
58383af6 | 22 | #include <linux/mutex.h> |
bcce3de1 | 23 | #include <linux/idr.h> |
77ea887e | 24 | #include <linux/log2.h> |
25e823c8 | 25 | #include <linux/pm_runtime.h> |
99e6608c | 26 | #include <linux/badblocks.h> |
82d981d4 | 27 | #include <linux/part_stat.h> |
8f9e7b65 | 28 | #include "blk-throttle.h" |
1da177e4 | 29 | |
ff88972c | 30 | #include "blk.h" |
2aa7745b | 31 | #include "blk-mq-sched.h" |
8e141f9e | 32 | #include "blk-rq-qos.h" |
1059699f | 33 | #include "blk-cgroup.h" |
ff88972c | 34 | |
31eb6186 | 35 | static struct kobject *block_depr; |
1da177e4 | 36 | |
cf179948 MC |
37 | /* |
38 | * Unique, monotonically increasing sequential number associated with block | |
39 | * devices instances (i.e. incremented each time a device is attached). | |
40 | * Associating uevents with block devices in userspace is difficult and racy: | |
41 | * the uevent netlink socket is lossy, and on slow and overloaded systems has | |
42 | * a very high latency. | |
43 | * Block devices do not have exclusive owners in userspace, any process can set | |
44 | * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0 | |
45 | * can be reused again and again). | |
46 | * A userspace process setting up a block device and watching for its events | |
47 | * cannot thus reliably tell whether an event relates to the device it just set | |
48 | * up or another earlier instance with the same name. | |
49 | * This sequential number allows userspace processes to solve this problem, and | |
50 | * uniquely associate an uevent to the lifetime to a device. | |
51 | */ | |
52 | static atomic64_t diskseq; | |
53 | ||
bcce3de1 | 54 | /* for extended dynamic devt allocation, currently only one major is used */ |
ce23bba8 | 55 | #define NR_EXT_DEVT (1 << MINORBITS) |
22ae8ce8 | 56 | static DEFINE_IDA(ext_devt_ida); |
bcce3de1 | 57 | |
a782483c CH |
58 | void set_capacity(struct gendisk *disk, sector_t sectors) |
59 | { | |
cb8432d6 | 60 | struct block_device *bdev = disk->part0; |
a782483c | 61 | |
0f472277 | 62 | spin_lock(&bdev->bd_size_lock); |
a782483c | 63 | i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); |
f09313c5 | 64 | bdev->bd_nr_sectors = sectors; |
0f472277 | 65 | spin_unlock(&bdev->bd_size_lock); |
a782483c CH |
66 | } |
67 | EXPORT_SYMBOL(set_capacity); | |
68 | ||
e598a72f | 69 | /* |
449f4ec9 CH |
70 | * Set disk capacity and notify if the size is not currently zero and will not |
71 | * be set to zero. Returns true if a uevent was sent, otherwise false. | |
e598a72f | 72 | */ |
449f4ec9 | 73 | bool set_capacity_and_notify(struct gendisk *disk, sector_t size) |
e598a72f BS |
74 | { |
75 | sector_t capacity = get_capacity(disk); | |
a782483c | 76 | char *envp[] = { "RESIZE=1", NULL }; |
e598a72f BS |
77 | |
78 | set_capacity(disk, size); | |
e598a72f | 79 | |
a782483c CH |
80 | /* |
81 | * Only print a message and send a uevent if the gendisk is user visible | |
82 | * and alive. This avoids spamming the log and udev when setting the | |
83 | * initial capacity during probing. | |
84 | */ | |
85 | if (size == capacity || | |
50b4aecf CH |
86 | !disk_live(disk) || |
87 | (disk->flags & GENHD_FL_HIDDEN)) | |
a782483c | 88 | return false; |
e598a72f | 89 | |
a782483c | 90 | pr_info("%s: detected capacity change from %lld to %lld\n", |
452c0bf8 | 91 | disk->disk_name, capacity, size); |
7e890c37 | 92 | |
a782483c CH |
93 | /* |
94 | * Historically we did not send a uevent for changes to/from an empty | |
95 | * device. | |
96 | */ | |
97 | if (!capacity || !size) | |
98 | return false; | |
99 | kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); | |
100 | return true; | |
e598a72f | 101 | } |
449f4ec9 | 102 | EXPORT_SYMBOL_GPL(set_capacity_and_notify); |
e598a72f | 103 | |
0d02129e CH |
104 | static void part_stat_read_all(struct block_device *part, |
105 | struct disk_stats *stat) | |
ea18e0f0 KK |
106 | { |
107 | int cpu; | |
108 | ||
109 | memset(stat, 0, sizeof(struct disk_stats)); | |
110 | for_each_possible_cpu(cpu) { | |
0d02129e | 111 | struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu); |
ea18e0f0 KK |
112 | int group; |
113 | ||
114 | for (group = 0; group < NR_STAT_GROUPS; group++) { | |
115 | stat->nsecs[group] += ptr->nsecs[group]; | |
116 | stat->sectors[group] += ptr->sectors[group]; | |
117 | stat->ios[group] += ptr->ios[group]; | |
118 | stat->merges[group] += ptr->merges[group]; | |
119 | } | |
120 | ||
121 | stat->io_ticks += ptr->io_ticks; | |
ea18e0f0 KK |
122 | } |
123 | } | |
ea18e0f0 | 124 | |
8446fe92 | 125 | static unsigned int part_in_flight(struct block_device *part) |
f299b7c7 | 126 | { |
b2f609e1 | 127 | unsigned int inflight = 0; |
1226b8dd | 128 | int cpu; |
f299b7c7 | 129 | |
1226b8dd | 130 | for_each_possible_cpu(cpu) { |
e016b782 MP |
131 | inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + |
132 | part_stat_local_read_cpu(part, in_flight[1], cpu); | |
1226b8dd | 133 | } |
e016b782 MP |
134 | if ((int)inflight < 0) |
135 | inflight = 0; | |
1226b8dd | 136 | |
e016b782 | 137 | return inflight; |
f299b7c7 JA |
138 | } |
139 | ||
8446fe92 CH |
140 | static void part_in_flight_rw(struct block_device *part, |
141 | unsigned int inflight[2]) | |
bf0ddaba | 142 | { |
1226b8dd MP |
143 | int cpu; |
144 | ||
1226b8dd MP |
145 | inflight[0] = 0; |
146 | inflight[1] = 0; | |
147 | for_each_possible_cpu(cpu) { | |
148 | inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); | |
149 | inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); | |
150 | } | |
151 | if ((int)inflight[0] < 0) | |
152 | inflight[0] = 0; | |
153 | if ((int)inflight[1] < 0) | |
154 | inflight[1] = 0; | |
bf0ddaba OS |
155 | } |
156 | ||
1da177e4 LT |
157 | /* |
158 | * Can be deleted altogether. Later. | |
159 | * | |
160 | */ | |
133d55cd | 161 | #define BLKDEV_MAJOR_HASH_SIZE 255 |
1da177e4 LT |
162 | static struct blk_major_name { |
163 | struct blk_major_name *next; | |
164 | int major; | |
165 | char name[16]; | |
fbdee71b | 166 | #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD |
a160c615 | 167 | void (*probe)(dev_t devt); |
fbdee71b | 168 | #endif |
68eef3b4 | 169 | } *major_names[BLKDEV_MAJOR_HASH_SIZE]; |
e49fbbbf | 170 | static DEFINE_MUTEX(major_names_lock); |
dfbb3409 | 171 | static DEFINE_SPINLOCK(major_names_spinlock); |
1da177e4 LT |
172 | |
173 | /* index in the above - for now: assume no multimajor ranges */ | |
e61eb2e9 | 174 | static inline int major_to_index(unsigned major) |
1da177e4 | 175 | { |
68eef3b4 | 176 | return major % BLKDEV_MAJOR_HASH_SIZE; |
7170be5f NH |
177 | } |
178 | ||
68eef3b4 | 179 | #ifdef CONFIG_PROC_FS |
cf771cb5 | 180 | void blkdev_show(struct seq_file *seqf, off_t offset) |
7170be5f | 181 | { |
68eef3b4 | 182 | struct blk_major_name *dp; |
7170be5f | 183 | |
dfbb3409 | 184 | spin_lock(&major_names_spinlock); |
133d55cd LG |
185 | for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) |
186 | if (dp->major == offset) | |
cf771cb5 | 187 | seq_printf(seqf, "%3d %s\n", dp->major, dp->name); |
dfbb3409 | 188 | spin_unlock(&major_names_spinlock); |
1da177e4 | 189 | } |
68eef3b4 | 190 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 191 | |
9e8c0bcc | 192 | /** |
e2b6b301 | 193 | * __register_blkdev - register a new block device |
9e8c0bcc | 194 | * |
f33ff110 SB |
195 | * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If |
196 | * @major = 0, try to allocate any unused major number. | |
9e8c0bcc | 197 | * @name: the name of the new block device as a zero terminated string |
26e06f5b LC |
198 | * @probe: pre-devtmpfs / pre-udev callback used to create disks when their |
199 | * pre-created device node is accessed. When a probe call uses | |
200 | * add_disk() and it fails the driver must cleanup resources. This | |
201 | * interface may soon be removed. | |
9e8c0bcc MN |
202 | * |
203 | * The @name must be unique within the system. | |
204 | * | |
0e056eb5 | 205 | * The return value depends on the @major input parameter: |
206 | * | |
f33ff110 SB |
207 | * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] |
208 | * then the function returns zero on success, or a negative error code | |
0e056eb5 | 209 | * - if any unused major number was requested with @major = 0 parameter |
9e8c0bcc | 210 | * then the return value is the allocated major number in range |
f33ff110 SB |
211 | * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise |
212 | * | |
213 | * See Documentation/admin-guide/devices.txt for the list of allocated | |
214 | * major numbers. | |
e2b6b301 CH |
215 | * |
216 | * Use register_blkdev instead for any new code. | |
9e8c0bcc | 217 | */ |
a160c615 CH |
218 | int __register_blkdev(unsigned int major, const char *name, |
219 | void (*probe)(dev_t devt)) | |
1da177e4 LT |
220 | { |
221 | struct blk_major_name **n, *p; | |
222 | int index, ret = 0; | |
223 | ||
e49fbbbf | 224 | mutex_lock(&major_names_lock); |
1da177e4 LT |
225 | |
226 | /* temporary */ | |
227 | if (major == 0) { | |
228 | for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { | |
229 | if (major_names[index] == NULL) | |
230 | break; | |
231 | } | |
232 | ||
233 | if (index == 0) { | |
dfc76d11 KP |
234 | printk("%s: failed to get major for %s\n", |
235 | __func__, name); | |
1da177e4 LT |
236 | ret = -EBUSY; |
237 | goto out; | |
238 | } | |
239 | major = index; | |
240 | ret = major; | |
241 | } | |
242 | ||
133d55cd | 243 | if (major >= BLKDEV_MAJOR_MAX) { |
dfc76d11 KP |
244 | pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", |
245 | __func__, major, BLKDEV_MAJOR_MAX-1, name); | |
133d55cd LG |
246 | |
247 | ret = -EINVAL; | |
248 | goto out; | |
249 | } | |
250 | ||
1da177e4 LT |
251 | p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); |
252 | if (p == NULL) { | |
253 | ret = -ENOMEM; | |
254 | goto out; | |
255 | } | |
256 | ||
257 | p->major = major; | |
fbdee71b | 258 | #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD |
a160c615 | 259 | p->probe = probe; |
fbdee71b | 260 | #endif |
1da177e4 LT |
261 | strlcpy(p->name, name, sizeof(p->name)); |
262 | p->next = NULL; | |
263 | index = major_to_index(major); | |
264 | ||
dfbb3409 | 265 | spin_lock(&major_names_spinlock); |
1da177e4 LT |
266 | for (n = &major_names[index]; *n; n = &(*n)->next) { |
267 | if ((*n)->major == major) | |
268 | break; | |
269 | } | |
270 | if (!*n) | |
271 | *n = p; | |
272 | else | |
273 | ret = -EBUSY; | |
dfbb3409 | 274 | spin_unlock(&major_names_spinlock); |
1da177e4 LT |
275 | |
276 | if (ret < 0) { | |
f33ff110 | 277 | printk("register_blkdev: cannot get major %u for %s\n", |
1da177e4 LT |
278 | major, name); |
279 | kfree(p); | |
280 | } | |
281 | out: | |
e49fbbbf | 282 | mutex_unlock(&major_names_lock); |
1da177e4 LT |
283 | return ret; |
284 | } | |
a160c615 | 285 | EXPORT_SYMBOL(__register_blkdev); |
1da177e4 | 286 | |
f4480240 | 287 | void unregister_blkdev(unsigned int major, const char *name) |
1da177e4 LT |
288 | { |
289 | struct blk_major_name **n; | |
290 | struct blk_major_name *p = NULL; | |
291 | int index = major_to_index(major); | |
1da177e4 | 292 | |
e49fbbbf | 293 | mutex_lock(&major_names_lock); |
dfbb3409 | 294 | spin_lock(&major_names_spinlock); |
1da177e4 LT |
295 | for (n = &major_names[index]; *n; n = &(*n)->next) |
296 | if ((*n)->major == major) | |
297 | break; | |
294462a5 AM |
298 | if (!*n || strcmp((*n)->name, name)) { |
299 | WARN_ON(1); | |
294462a5 | 300 | } else { |
1da177e4 LT |
301 | p = *n; |
302 | *n = p->next; | |
303 | } | |
dfbb3409 | 304 | spin_unlock(&major_names_spinlock); |
e49fbbbf | 305 | mutex_unlock(&major_names_lock); |
1da177e4 | 306 | kfree(p); |
1da177e4 LT |
307 | } |
308 | ||
309 | EXPORT_SYMBOL(unregister_blkdev); | |
310 | ||
7c3f828b | 311 | int blk_alloc_ext_minor(void) |
bcce3de1 | 312 | { |
bab998d6 | 313 | int idx; |
bcce3de1 | 314 | |
d1868328 | 315 | idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT - 1, GFP_KERNEL); |
c4b2b7d1 CH |
316 | if (idx == -ENOSPC) |
317 | return -EBUSY; | |
318 | return idx; | |
bcce3de1 TH |
319 | } |
320 | ||
7c3f828b | 321 | void blk_free_ext_minor(unsigned int minor) |
bcce3de1 | 322 | { |
c4b2b7d1 | 323 | ida_free(&ext_devt_ida, minor); |
6fcc44d1 YY |
324 | } |
325 | ||
1f014290 TH |
326 | static char *bdevt_str(dev_t devt, char *buf) |
327 | { | |
328 | if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { | |
329 | char tbuf[BDEVT_SIZE]; | |
330 | snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); | |
331 | snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); | |
332 | } else | |
333 | snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); | |
334 | ||
335 | return buf; | |
336 | } | |
337 | ||
bc359d03 CH |
338 | void disk_uevent(struct gendisk *disk, enum kobject_action action) |
339 | { | |
bc359d03 | 340 | struct block_device *part; |
3212135a | 341 | unsigned long idx; |
bc359d03 | 342 | |
3212135a CH |
343 | rcu_read_lock(); |
344 | xa_for_each(&disk->part_tbl, idx, part) { | |
345 | if (bdev_is_partition(part) && !bdev_nr_sectors(part)) | |
346 | continue; | |
498dcc13 | 347 | if (!kobject_get_unless_zero(&part->bd_device.kobj)) |
3212135a CH |
348 | continue; |
349 | ||
350 | rcu_read_unlock(); | |
bc359d03 | 351 | kobject_uevent(bdev_kobj(part), action); |
498dcc13 | 352 | put_device(&part->bd_device); |
3212135a CH |
353 | rcu_read_lock(); |
354 | } | |
355 | rcu_read_unlock(); | |
bc359d03 CH |
356 | } |
357 | EXPORT_SYMBOL_GPL(disk_uevent); | |
358 | ||
0f77b29a | 359 | int disk_scan_partitions(struct gendisk *disk, fmode_t mode) |
9301fe73 CH |
360 | { |
361 | struct block_device *bdev; | |
e5cfefa9 | 362 | int ret = 0; |
9301fe73 | 363 | |
9f18db57 | 364 | if (disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) |
e16e506c | 365 | return -EINVAL; |
b9684a71 CH |
366 | if (test_bit(GD_SUPPRESS_PART_SCAN, &disk->state)) |
367 | return -EINVAL; | |
e16e506c CH |
368 | if (disk->open_partitions) |
369 | return -EBUSY; | |
9301fe73 | 370 | |
e5cfefa9 YK |
371 | /* |
372 | * If the device is opened exclusively by current thread already, it's | |
373 | * safe to scan partitons, otherwise, use bd_prepare_to_claim() to | |
374 | * synchronize with other exclusive openers and other partition | |
375 | * scanners. | |
376 | */ | |
377 | if (!(mode & FMODE_EXCL)) { | |
378 | ret = bd_prepare_to_claim(disk->part0, disk_scan_partitions); | |
379 | if (ret) | |
380 | return ret; | |
381 | } | |
382 | ||
3723091e | 383 | set_bit(GD_NEED_PART_SCAN, &disk->state); |
e5cfefa9 | 384 | bdev = blkdev_get_by_dev(disk_devt(disk), mode & ~FMODE_EXCL, NULL); |
e16e506c | 385 | if (IS_ERR(bdev)) |
e5cfefa9 YK |
386 | ret = PTR_ERR(bdev); |
387 | else | |
428913bc | 388 | blkdev_put(bdev, mode & ~FMODE_EXCL); |
e5cfefa9 | 389 | |
3723091e YK |
390 | /* |
391 | * If blkdev_get_by_dev() failed early, GD_NEED_PART_SCAN is still set, | |
392 | * and this will cause that re-assemble partitioned raid device will | |
393 | * creat partition for underlying disk. | |
394 | */ | |
395 | clear_bit(GD_NEED_PART_SCAN, &disk->state); | |
e5cfefa9 YK |
396 | if (!(mode & FMODE_EXCL)) |
397 | bd_abort_claiming(disk->part0, disk_scan_partitions); | |
398 | return ret; | |
9301fe73 CH |
399 | } |
400 | ||
1da177e4 | 401 | /** |
d1254a87 | 402 | * device_add_disk - add disk information to kernel list |
e63a46be | 403 | * @parent: parent device for the disk |
1da177e4 | 404 | * @disk: per-device partitioning information |
fef912bf | 405 | * @groups: Additional per-device sysfs groups |
1da177e4 LT |
406 | * |
407 | * This function registers the partitioning information in @disk | |
408 | * with the kernel. | |
409 | */ | |
278167fd LC |
410 | int __must_check device_add_disk(struct device *parent, struct gendisk *disk, |
411 | const struct attribute_group **groups) | |
d1254a87 | 412 | |
1da177e4 | 413 | { |
52b85909 | 414 | struct device *ddev = disk_to_dev(disk); |
7c3f828b | 415 | int ret; |
cf0ca9fe | 416 | |
69fe0f29 ML |
417 | /* Only makes sense for bio-based to set ->poll_bio */ |
418 | if (queue_is_mq(disk->queue) && disk->fops->poll_bio) | |
419 | return -EINVAL; | |
420 | ||
737eb78e DLM |
421 | /* |
422 | * The disk queue should now be all set with enough information about | |
423 | * the device for the elevator code to pick an adequate default | |
424 | * elevator if one is needed, that is, for devices requesting queue | |
425 | * registration. | |
426 | */ | |
d1254a87 | 427 | elevator_init_mq(disk->queue); |
737eb78e | 428 | |
9f4107b0 JA |
429 | /* Mark bdev as having a submit_bio, if needed */ |
430 | disk->part0->bd_has_submit_bio = disk->fops->submit_bio != NULL; | |
431 | ||
7c3f828b CH |
432 | /* |
433 | * If the driver provides an explicit major number it also must provide | |
434 | * the number of minors numbers supported, and those will be used to | |
435 | * setup the gendisk. | |
436 | * Otherwise just allocate the device numbers for both the whole device | |
437 | * and all partitions from the extended dev_t space. | |
3e1a7ff8 | 438 | */ |
02341a08 | 439 | ret = -EINVAL; |
7c3f828b | 440 | if (disk->major) { |
83cbce95 | 441 | if (WARN_ON(!disk->minors)) |
02341a08 | 442 | goto out_exit_elevator; |
2e3c73fa CH |
443 | |
444 | if (disk->minors > DISK_MAX_PARTS) { | |
445 | pr_err("block: can't allocate more than %d partitions\n", | |
446 | DISK_MAX_PARTS); | |
447 | disk->minors = DISK_MAX_PARTS; | |
448 | } | |
e338924b | 449 | if (disk->first_minor + disk->minors > MINORMASK + 1) |
02341a08 | 450 | goto out_exit_elevator; |
7c3f828b | 451 | } else { |
83cbce95 | 452 | if (WARN_ON(disk->minors)) |
02341a08 | 453 | goto out_exit_elevator; |
3e1a7ff8 | 454 | |
7c3f828b | 455 | ret = blk_alloc_ext_minor(); |
83cbce95 | 456 | if (ret < 0) |
02341a08 | 457 | goto out_exit_elevator; |
7c3f828b | 458 | disk->major = BLOCK_EXT_MAJOR; |
539711d7 | 459 | disk->first_minor = ret; |
3e1a7ff8 | 460 | } |
7c3f828b | 461 | |
52b85909 CH |
462 | /* delay uevents, until we scanned partition table */ |
463 | dev_set_uevent_suppress(ddev, 1); | |
464 | ||
465 | ddev->parent = parent; | |
466 | ddev->groups = groups; | |
467 | dev_set_name(ddev, "%s", disk->disk_name); | |
8235b5c1 CH |
468 | if (!(disk->flags & GENHD_FL_HIDDEN)) |
469 | ddev->devt = MKDEV(disk->major, disk->first_minor); | |
83cbce95 LC |
470 | ret = device_add(ddev); |
471 | if (ret) | |
99d8690a CH |
472 | goto out_free_ext_minor; |
473 | ||
474 | ret = disk_alloc_events(disk); | |
475 | if (ret) | |
476 | goto out_device_del; | |
477 | ||
52b85909 CH |
478 | if (!sysfs_deprecated) { |
479 | ret = sysfs_create_link(block_depr, &ddev->kobj, | |
480 | kobject_name(&ddev->kobj)); | |
83cbce95 LC |
481 | if (ret) |
482 | goto out_device_del; | |
52b85909 CH |
483 | } |
484 | ||
485 | /* | |
486 | * avoid probable deadlock caused by allocating memory with | |
487 | * GFP_KERNEL in runtime_resume callback of its all ancestor | |
488 | * devices | |
489 | */ | |
490 | pm_runtime_set_memalloc_noio(ddev, true); | |
491 | ||
178fa7d4 CH |
492 | ret = blk_integrity_add(disk); |
493 | if (ret) | |
b6553bef | 494 | goto out_del_block_link; |
178fa7d4 | 495 | |
52b85909 CH |
496 | disk->part0->bd_holder_dir = |
497 | kobject_create_and_add("holders", &ddev->kobj); | |
fe7d064f LC |
498 | if (!disk->part0->bd_holder_dir) { |
499 | ret = -ENOMEM; | |
83cbce95 | 500 | goto out_del_integrity; |
fe7d064f | 501 | } |
52b85909 | 502 | disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); |
fe7d064f LC |
503 | if (!disk->slave_dir) { |
504 | ret = -ENOMEM; | |
83cbce95 | 505 | goto out_put_holder_dir; |
fe7d064f | 506 | } |
52b85909 | 507 | |
83cbce95 LC |
508 | ret = blk_register_queue(disk); |
509 | if (ret) | |
510 | goto out_put_slave_dir; | |
75f4dca5 | 511 | |
9f18db57 | 512 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8235b5c1 CH |
513 | ret = bdi_register(disk->bdi, "%u:%u", |
514 | disk->major, disk->first_minor); | |
83cbce95 LC |
515 | if (ret) |
516 | goto out_unregister_queue; | |
8235b5c1 | 517 | bdi_set_owner(disk->bdi, ddev); |
83cbce95 LC |
518 | ret = sysfs_create_link(&ddev->kobj, |
519 | &disk->bdi->dev->kobj, "bdi"); | |
520 | if (ret) | |
521 | goto out_unregister_bdi; | |
8235b5c1 | 522 | |
e5cfefa9 YK |
523 | /* Make sure the first partition scan will be proceed */ |
524 | if (get_capacity(disk) && !(disk->flags & GENHD_FL_NO_PART) && | |
525 | !test_bit(GD_SUPPRESS_PART_SCAN, &disk->state)) | |
526 | set_bit(GD_NEED_PART_SCAN, &disk->state); | |
527 | ||
9d5ee676 | 528 | bdev_add(disk->part0, ddev->devt); |
e16e506c | 529 | if (get_capacity(disk)) |
0f77b29a | 530 | disk_scan_partitions(disk, FMODE_READ); |
52b85909 CH |
531 | |
532 | /* | |
533 | * Announce the disk and partitions after all partitions are | |
8235b5c1 | 534 | * created. (for hidden disks uevents remain suppressed forever) |
52b85909 CH |
535 | */ |
536 | dev_set_uevent_suppress(ddev, 0); | |
537 | disk_uevent(disk, KOBJ_ADD); | |
a0a6314a CH |
538 | } else { |
539 | /* | |
540 | * Even if the block_device for a hidden gendisk is not | |
541 | * registered, it needs to have a valid bd_dev so that the | |
542 | * freeing of the dynamic major works. | |
543 | */ | |
544 | disk->part0->bd_dev = MKDEV(disk->major, disk->first_minor); | |
52b85909 CH |
545 | } |
546 | ||
75f4dca5 | 547 | disk_update_readahead(disk); |
77ea887e | 548 | disk_add_events(disk); |
76792055 | 549 | set_bit(GD_ADDED, &disk->state); |
83cbce95 LC |
550 | return 0; |
551 | ||
552 | out_unregister_bdi: | |
553 | if (!(disk->flags & GENHD_FL_HIDDEN)) | |
554 | bdi_unregister(disk->bdi); | |
555 | out_unregister_queue: | |
556 | blk_unregister_queue(disk); | |
fa81cbaf | 557 | rq_qos_exit(disk->queue); |
83cbce95 LC |
558 | out_put_slave_dir: |
559 | kobject_put(disk->slave_dir); | |
d90db3b1 | 560 | disk->slave_dir = NULL; |
83cbce95 LC |
561 | out_put_holder_dir: |
562 | kobject_put(disk->part0->bd_holder_dir); | |
563 | out_del_integrity: | |
564 | blk_integrity_del(disk); | |
565 | out_del_block_link: | |
566 | if (!sysfs_deprecated) | |
567 | sysfs_remove_link(block_depr, dev_name(ddev)); | |
568 | out_device_del: | |
569 | device_del(ddev); | |
83cbce95 LC |
570 | out_free_ext_minor: |
571 | if (disk->major == BLOCK_EXT_MAJOR) | |
572 | blk_free_ext_minor(disk->first_minor); | |
02341a08 YK |
573 | out_exit_elevator: |
574 | if (disk->queue->elevator) | |
575 | elevator_exit(disk->queue); | |
278167fd | 576 | return ret; |
1da177e4 | 577 | } |
e63a46be | 578 | EXPORT_SYMBOL(device_add_disk); |
1da177e4 | 579 | |
7a5428dc CH |
580 | /** |
581 | * blk_mark_disk_dead - mark a disk as dead | |
582 | * @disk: disk to mark as dead | |
583 | * | |
584 | * Mark as disk as dead (e.g. surprise removed) and don't accept any new I/O | |
585 | * to this disk. | |
586 | */ | |
587 | void blk_mark_disk_dead(struct gendisk *disk) | |
588 | { | |
589 | set_bit(GD_DEAD, &disk->state); | |
590 | blk_queue_start_drain(disk->queue); | |
71b26083 CH |
591 | |
592 | /* | |
593 | * Stop buffered writers from dirtying pages that can't be written out. | |
594 | */ | |
595 | set_capacity_and_notify(disk, 0); | |
7a5428dc CH |
596 | } |
597 | EXPORT_SYMBOL_GPL(blk_mark_disk_dead); | |
598 | ||
b5bd357c LC |
599 | /** |
600 | * del_gendisk - remove the gendisk | |
601 | * @disk: the struct gendisk to remove | |
602 | * | |
603 | * Removes the gendisk and all its associated resources. This deletes the | |
604 | * partitions associated with the gendisk, and unregisters the associated | |
605 | * request_queue. | |
606 | * | |
607 | * This is the counter to the respective __device_add_disk() call. | |
608 | * | |
609 | * The final removal of the struct gendisk happens when its refcount reaches 0 | |
610 | * with put_disk(), which should be called after del_gendisk(), if | |
611 | * __device_add_disk() was used. | |
e8c7d14a LC |
612 | * |
613 | * Drivers exist which depend on the release of the gendisk to be synchronous, | |
614 | * it should not be deferred. | |
615 | * | |
616 | * Context: can sleep | |
b5bd357c | 617 | */ |
d2bf1b67 | 618 | void del_gendisk(struct gendisk *disk) |
1da177e4 | 619 | { |
8e141f9e CH |
620 | struct request_queue *q = disk->queue; |
621 | ||
e8c7d14a LC |
622 | might_sleep(); |
623 | ||
9f286992 | 624 | if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN))) |
6b3ba976 CH |
625 | return; |
626 | ||
25520d55 | 627 | blk_integrity_del(disk); |
77ea887e TH |
628 | disk_del_events(disk); |
629 | ||
a8698707 | 630 | mutex_lock(&disk->open_mutex); |
d7a66574 | 631 | remove_inode_hash(disk->part0->bd_inode); |
d3c4a43d | 632 | blk_drop_partitions(disk); |
a8698707 | 633 | mutex_unlock(&disk->open_mutex); |
c76f48eb | 634 | |
45611837 CH |
635 | fsync_bdev(disk->part0); |
636 | __invalidate_device(disk->part0, true); | |
637 | ||
8e141f9e CH |
638 | /* |
639 | * Fail any new I/O. | |
640 | */ | |
641 | set_bit(GD_DEAD, &disk->state); | |
6f8191fd CH |
642 | if (test_bit(GD_OWNS_QUEUE, &disk->state)) |
643 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); | |
d2bf1b67 | 644 | set_capacity(disk, 0); |
d2bf1b67 | 645 | |
8e141f9e CH |
646 | /* |
647 | * Prevent new I/O from crossing bio_queue_enter(). | |
648 | */ | |
649 | blk_queue_start_drain(q); | |
8e141f9e | 650 | |
6b3ba976 | 651 | if (!(disk->flags & GENHD_FL_HIDDEN)) { |
8ddcd653 | 652 | sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); |
6b3ba976 | 653 | |
90f16fdd JK |
654 | /* |
655 | * Unregister bdi before releasing device numbers (as they can | |
656 | * get reused and we'd get clashes in sysfs). | |
657 | */ | |
edb0872f | 658 | bdi_unregister(disk->bdi); |
90f16fdd | 659 | } |
d2bf1b67 | 660 | |
6b3ba976 | 661 | blk_unregister_queue(disk); |
d2bf1b67 | 662 | |
cb8432d6 | 663 | kobject_put(disk->part0->bd_holder_dir); |
d2bf1b67 | 664 | kobject_put(disk->slave_dir); |
d90db3b1 | 665 | disk->slave_dir = NULL; |
d2bf1b67 | 666 | |
8446fe92 | 667 | part_stat_set_all(disk->part0, 0); |
cb8432d6 | 668 | disk->part0->bd_stamp = 0; |
d2bf1b67 TH |
669 | if (!sysfs_deprecated) |
670 | sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); | |
25e823c8 | 671 | pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); |
d2bf1b67 | 672 | device_del(disk_to_dev(disk)); |
d308ae0d | 673 | |
4c66a326 CH |
674 | blk_mq_freeze_queue_wait(q); |
675 | ||
cad9266a | 676 | blk_throtl_cancel_bios(disk); |
8f9e7b65 | 677 | |
d308ae0d ML |
678 | blk_sync_queue(q); |
679 | blk_flush_integrity(); | |
219cf43c JC |
680 | |
681 | if (queue_is_mq(q)) | |
682 | blk_mq_cancel_work_sync(q); | |
50e34d78 CH |
683 | |
684 | blk_mq_quiesce_queue(q); | |
685 | if (q->elevator) { | |
686 | mutex_lock(&q->sysfs_lock); | |
687 | elevator_exit(q); | |
688 | mutex_unlock(&q->sysfs_lock); | |
689 | } | |
690 | rq_qos_exit(q); | |
691 | blk_mq_unquiesce_queue(q); | |
692 | ||
d308ae0d | 693 | /* |
6f8191fd CH |
694 | * If the disk does not own the queue, allow using passthrough requests |
695 | * again. Else leave the queue frozen to fail all I/O. | |
d308ae0d | 696 | */ |
6f8191fd CH |
697 | if (!test_bit(GD_OWNS_QUEUE, &disk->state)) { |
698 | blk_queue_flag_clear(QUEUE_FLAG_INIT_DONE, q); | |
699 | __blk_mq_unfreeze_queue(q, true); | |
700 | } else { | |
701 | if (queue_is_mq(q)) | |
702 | blk_mq_exit_queue(q); | |
703 | } | |
1da177e4 | 704 | } |
d2bf1b67 | 705 | EXPORT_SYMBOL(del_gendisk); |
1da177e4 | 706 | |
f059a1d2 XY |
707 | /** |
708 | * invalidate_disk - invalidate the disk | |
709 | * @disk: the struct gendisk to invalidate | |
710 | * | |
711 | * A helper to invalidates the disk. It will clean the disk's associated | |
712 | * buffer/page caches and reset its internal states so that the disk | |
713 | * can be reused by the drivers. | |
714 | * | |
715 | * Context: can sleep | |
716 | */ | |
717 | void invalidate_disk(struct gendisk *disk) | |
718 | { | |
719 | struct block_device *bdev = disk->part0; | |
720 | ||
721 | invalidate_bdev(bdev); | |
722 | bdev->bd_inode->i_mapping->wb_err = 0; | |
723 | set_capacity(disk, 0); | |
724 | } | |
725 | EXPORT_SYMBOL(invalidate_disk); | |
726 | ||
99e6608c VV |
727 | /* sysfs access to bad-blocks list. */ |
728 | static ssize_t disk_badblocks_show(struct device *dev, | |
729 | struct device_attribute *attr, | |
730 | char *page) | |
731 | { | |
732 | struct gendisk *disk = dev_to_disk(dev); | |
733 | ||
734 | if (!disk->bb) | |
735 | return sprintf(page, "\n"); | |
736 | ||
737 | return badblocks_show(disk->bb, page, 0); | |
738 | } | |
739 | ||
740 | static ssize_t disk_badblocks_store(struct device *dev, | |
741 | struct device_attribute *attr, | |
742 | const char *page, size_t len) | |
743 | { | |
744 | struct gendisk *disk = dev_to_disk(dev); | |
745 | ||
746 | if (!disk->bb) | |
747 | return -ENXIO; | |
748 | ||
749 | return badblocks_store(disk->bb, page, len, 0); | |
750 | } | |
751 | ||
fbdee71b | 752 | #ifdef CONFIG_BLOCK_LEGACY_AUTOLOAD |
22ae8ce8 | 753 | void blk_request_module(dev_t devt) |
bd8eff3b | 754 | { |
a160c615 CH |
755 | unsigned int major = MAJOR(devt); |
756 | struct blk_major_name **n; | |
757 | ||
758 | mutex_lock(&major_names_lock); | |
759 | for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { | |
760 | if ((*n)->major == major && (*n)->probe) { | |
761 | (*n)->probe(devt); | |
762 | mutex_unlock(&major_names_lock); | |
763 | return; | |
764 | } | |
765 | } | |
766 | mutex_unlock(&major_names_lock); | |
767 | ||
bd8eff3b CH |
768 | if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) |
769 | /* Make old-style 2.4 aliases work */ | |
770 | request_module("block-major-%d", MAJOR(devt)); | |
771 | } | |
fbdee71b | 772 | #endif /* CONFIG_BLOCK_LEGACY_AUTOLOAD */ |
bd8eff3b | 773 | |
5c6f35c5 GKH |
774 | /* |
775 | * print a full list of all partitions - intended for places where the root | |
776 | * filesystem can't be mounted and thus to give the victim some idea of what | |
777 | * went wrong | |
778 | */ | |
779 | void __init printk_all_partitions(void) | |
780 | { | |
def4e38d TH |
781 | struct class_dev_iter iter; |
782 | struct device *dev; | |
783 | ||
784 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); | |
785 | while ((dev = class_dev_iter_next(&iter))) { | |
786 | struct gendisk *disk = dev_to_disk(dev); | |
ad1eaa53 | 787 | struct block_device *part; |
1f014290 | 788 | char devt_buf[BDEVT_SIZE]; |
e559f58d | 789 | unsigned long idx; |
def4e38d TH |
790 | |
791 | /* | |
792 | * Don't show empty devices or things that have been | |
25985edc | 793 | * suppressed |
def4e38d | 794 | */ |
3b5149ac | 795 | if (get_capacity(disk) == 0 || (disk->flags & GENHD_FL_HIDDEN)) |
def4e38d TH |
796 | continue; |
797 | ||
798 | /* | |
e559f58d CH |
799 | * Note, unlike /proc/partitions, I am showing the numbers in |
800 | * hex - the same format as the root= option takes. | |
def4e38d | 801 | */ |
e559f58d CH |
802 | rcu_read_lock(); |
803 | xa_for_each(&disk->part_tbl, idx, part) { | |
804 | if (!bdev_nr_sectors(part)) | |
805 | continue; | |
a9e7bc3d | 806 | printk("%s%s %10llu %pg %s", |
e559f58d | 807 | bdev_is_partition(part) ? " " : "", |
ad1eaa53 | 808 | bdevt_str(part->bd_dev, devt_buf), |
a9e7bc3d | 809 | bdev_nr_sectors(part) >> 1, part, |
ad1eaa53 CH |
810 | part->bd_meta_info ? |
811 | part->bd_meta_info->uuid : ""); | |
e559f58d | 812 | if (bdev_is_partition(part)) |
074a7aca | 813 | printk("\n"); |
e559f58d CH |
814 | else if (dev->parent && dev->parent->driver) |
815 | printk(" driver: %s\n", | |
816 | dev->parent->driver->name); | |
817 | else | |
818 | printk(" (driver?)\n"); | |
074a7aca | 819 | } |
e559f58d | 820 | rcu_read_unlock(); |
def4e38d TH |
821 | } |
822 | class_dev_iter_exit(&iter); | |
dd2a345f DG |
823 | } |
824 | ||
1da177e4 LT |
825 | #ifdef CONFIG_PROC_FS |
826 | /* iterator */ | |
def4e38d | 827 | static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) |
68c4d4a7 | 828 | { |
def4e38d TH |
829 | loff_t skip = *pos; |
830 | struct class_dev_iter *iter; | |
831 | struct device *dev; | |
68c4d4a7 | 832 | |
aeb3d3a8 | 833 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); |
def4e38d TH |
834 | if (!iter) |
835 | return ERR_PTR(-ENOMEM); | |
836 | ||
837 | seqf->private = iter; | |
838 | class_dev_iter_init(iter, &block_class, NULL, &disk_type); | |
839 | do { | |
840 | dev = class_dev_iter_next(iter); | |
841 | if (!dev) | |
842 | return NULL; | |
843 | } while (skip--); | |
844 | ||
845 | return dev_to_disk(dev); | |
68c4d4a7 GKH |
846 | } |
847 | ||
def4e38d | 848 | static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) |
1da177e4 | 849 | { |
edfaa7c3 | 850 | struct device *dev; |
1da177e4 | 851 | |
def4e38d TH |
852 | (*pos)++; |
853 | dev = class_dev_iter_next(seqf->private); | |
2ac3cee5 | 854 | if (dev) |
68c4d4a7 | 855 | return dev_to_disk(dev); |
2ac3cee5 | 856 | |
1da177e4 LT |
857 | return NULL; |
858 | } | |
859 | ||
def4e38d | 860 | static void disk_seqf_stop(struct seq_file *seqf, void *v) |
27f30251 | 861 | { |
def4e38d | 862 | struct class_dev_iter *iter = seqf->private; |
27f30251 | 863 | |
def4e38d TH |
864 | /* stop is called even after start failed :-( */ |
865 | if (iter) { | |
866 | class_dev_iter_exit(iter); | |
867 | kfree(iter); | |
77da1605 | 868 | seqf->private = NULL; |
5c0ef6d0 | 869 | } |
1da177e4 LT |
870 | } |
871 | ||
def4e38d | 872 | static void *show_partition_start(struct seq_file *seqf, loff_t *pos) |
1da177e4 | 873 | { |
06768067 | 874 | void *p; |
def4e38d TH |
875 | |
876 | p = disk_seqf_start(seqf, pos); | |
b9f985b6 | 877 | if (!IS_ERR_OR_NULL(p) && !*pos) |
def4e38d TH |
878 | seq_puts(seqf, "major minor #blocks name\n\n"); |
879 | return p; | |
1da177e4 LT |
880 | } |
881 | ||
cf771cb5 | 882 | static int show_partition(struct seq_file *seqf, void *v) |
1da177e4 LT |
883 | { |
884 | struct gendisk *sgp = v; | |
ad1eaa53 | 885 | struct block_device *part; |
ecc75a98 | 886 | unsigned long idx; |
1da177e4 | 887 | |
3b5149ac | 888 | if (!get_capacity(sgp) || (sgp->flags & GENHD_FL_HIDDEN)) |
1da177e4 LT |
889 | return 0; |
890 | ||
ecc75a98 CH |
891 | rcu_read_lock(); |
892 | xa_for_each(&sgp->part_tbl, idx, part) { | |
893 | if (!bdev_nr_sectors(part)) | |
894 | continue; | |
a291bb43 | 895 | seq_printf(seqf, "%4d %7d %10llu %pg\n", |
ad1eaa53 | 896 | MAJOR(part->bd_dev), MINOR(part->bd_dev), |
a291bb43 | 897 | bdev_nr_sectors(part) >> 1, part); |
ecc75a98 CH |
898 | } |
899 | rcu_read_unlock(); | |
1da177e4 LT |
900 | return 0; |
901 | } | |
902 | ||
f500975a | 903 | static const struct seq_operations partitions_op = { |
def4e38d TH |
904 | .start = show_partition_start, |
905 | .next = disk_seqf_next, | |
906 | .stop = disk_seqf_stop, | |
edfaa7c3 | 907 | .show = show_partition |
1da177e4 LT |
908 | }; |
909 | #endif | |
910 | ||
1da177e4 LT |
911 | static int __init genhd_device_init(void) |
912 | { | |
e105b8bf DW |
913 | int error; |
914 | ||
915 | block_class.dev_kobj = sysfs_dev_block_kobj; | |
916 | error = class_register(&block_class); | |
ee27a558 RM |
917 | if (unlikely(error)) |
918 | return error; | |
1da177e4 | 919 | blk_dev_init(); |
edfaa7c3 | 920 | |
561ec68e ZY |
921 | register_blkdev(BLOCK_EXT_MAJOR, "blkext"); |
922 | ||
edfaa7c3 | 923 | /* create top-level block dir */ |
e52eec13 AK |
924 | if (!sysfs_deprecated) |
925 | block_depr = kobject_create_and_add("block", NULL); | |
830d3cfb | 926 | return 0; |
1da177e4 LT |
927 | } |
928 | ||
929 | subsys_initcall(genhd_device_init); | |
930 | ||
edfaa7c3 KS |
931 | static ssize_t disk_range_show(struct device *dev, |
932 | struct device_attribute *attr, char *buf) | |
1da177e4 | 933 | { |
edfaa7c3 | 934 | struct gendisk *disk = dev_to_disk(dev); |
1da177e4 | 935 | |
edfaa7c3 | 936 | return sprintf(buf, "%d\n", disk->minors); |
1da177e4 LT |
937 | } |
938 | ||
1f014290 TH |
939 | static ssize_t disk_ext_range_show(struct device *dev, |
940 | struct device_attribute *attr, char *buf) | |
941 | { | |
942 | struct gendisk *disk = dev_to_disk(dev); | |
943 | ||
1ebe2e5f CH |
944 | return sprintf(buf, "%d\n", |
945 | (disk->flags & GENHD_FL_NO_PART) ? 1 : DISK_MAX_PARTS); | |
1f014290 TH |
946 | } |
947 | ||
edfaa7c3 KS |
948 | static ssize_t disk_removable_show(struct device *dev, |
949 | struct device_attribute *attr, char *buf) | |
a7fd6706 | 950 | { |
edfaa7c3 | 951 | struct gendisk *disk = dev_to_disk(dev); |
a7fd6706 | 952 | |
edfaa7c3 KS |
953 | return sprintf(buf, "%d\n", |
954 | (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); | |
a7fd6706 KS |
955 | } |
956 | ||
8ddcd653 CH |
957 | static ssize_t disk_hidden_show(struct device *dev, |
958 | struct device_attribute *attr, char *buf) | |
959 | { | |
960 | struct gendisk *disk = dev_to_disk(dev); | |
961 | ||
962 | return sprintf(buf, "%d\n", | |
963 | (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); | |
964 | } | |
965 | ||
1c9ce527 KS |
966 | static ssize_t disk_ro_show(struct device *dev, |
967 | struct device_attribute *attr, char *buf) | |
968 | { | |
969 | struct gendisk *disk = dev_to_disk(dev); | |
970 | ||
b7db9956 | 971 | return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); |
1c9ce527 KS |
972 | } |
973 | ||
3ad5cee5 CH |
974 | ssize_t part_size_show(struct device *dev, |
975 | struct device_attribute *attr, char *buf) | |
976 | { | |
0d02129e | 977 | return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev))); |
3ad5cee5 CH |
978 | } |
979 | ||
980 | ssize_t part_stat_show(struct device *dev, | |
981 | struct device_attribute *attr, char *buf) | |
982 | { | |
0d02129e | 983 | struct block_device *bdev = dev_to_bdev(dev); |
ed6cddef | 984 | struct request_queue *q = bdev_get_queue(bdev); |
ea18e0f0 | 985 | struct disk_stats stat; |
3ad5cee5 CH |
986 | unsigned int inflight; |
987 | ||
b2f609e1 | 988 | if (queue_is_mq(q)) |
0d02129e | 989 | inflight = blk_mq_in_flight(q, bdev); |
b2f609e1 | 990 | else |
0d02129e | 991 | inflight = part_in_flight(bdev); |
ea18e0f0 | 992 | |
86d73312 ZW |
993 | if (inflight) { |
994 | part_stat_lock(); | |
995 | update_io_ticks(bdev, jiffies, true); | |
996 | part_stat_unlock(); | |
997 | } | |
998 | part_stat_read_all(bdev, &stat); | |
3ad5cee5 CH |
999 | return sprintf(buf, |
1000 | "%8lu %8lu %8llu %8u " | |
1001 | "%8lu %8lu %8llu %8u " | |
1002 | "%8u %8u %8u " | |
1003 | "%8lu %8lu %8llu %8u " | |
1004 | "%8lu %8u" | |
1005 | "\n", | |
ea18e0f0 KK |
1006 | stat.ios[STAT_READ], |
1007 | stat.merges[STAT_READ], | |
1008 | (unsigned long long)stat.sectors[STAT_READ], | |
1009 | (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), | |
1010 | stat.ios[STAT_WRITE], | |
1011 | stat.merges[STAT_WRITE], | |
1012 | (unsigned long long)stat.sectors[STAT_WRITE], | |
1013 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), | |
3ad5cee5 | 1014 | inflight, |
ea18e0f0 | 1015 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1016 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1017 | stat.nsecs[STAT_WRITE] + | |
1018 | stat.nsecs[STAT_DISCARD] + | |
1019 | stat.nsecs[STAT_FLUSH], | |
1020 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1021 | stat.ios[STAT_DISCARD], |
1022 | stat.merges[STAT_DISCARD], | |
1023 | (unsigned long long)stat.sectors[STAT_DISCARD], | |
1024 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), | |
1025 | stat.ios[STAT_FLUSH], | |
1026 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); | |
3ad5cee5 CH |
1027 | } |
1028 | ||
1029 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, | |
1030 | char *buf) | |
1031 | { | |
0d02129e | 1032 | struct block_device *bdev = dev_to_bdev(dev); |
ed6cddef | 1033 | struct request_queue *q = bdev_get_queue(bdev); |
3ad5cee5 CH |
1034 | unsigned int inflight[2]; |
1035 | ||
b2f609e1 | 1036 | if (queue_is_mq(q)) |
0d02129e | 1037 | blk_mq_in_flight_rw(q, bdev, inflight); |
b2f609e1 | 1038 | else |
0d02129e | 1039 | part_in_flight_rw(bdev, inflight); |
b2f609e1 | 1040 | |
3ad5cee5 CH |
1041 | return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); |
1042 | } | |
1043 | ||
edfaa7c3 KS |
1044 | static ssize_t disk_capability_show(struct device *dev, |
1045 | struct device_attribute *attr, char *buf) | |
86ce18d7 | 1046 | { |
e81cd5a9 CH |
1047 | dev_warn_once(dev, "the capability attribute has been deprecated.\n"); |
1048 | return sprintf(buf, "0\n"); | |
86ce18d7 | 1049 | } |
edfaa7c3 | 1050 | |
c72758f3 MP |
1051 | static ssize_t disk_alignment_offset_show(struct device *dev, |
1052 | struct device_attribute *attr, | |
1053 | char *buf) | |
1054 | { | |
1055 | struct gendisk *disk = dev_to_disk(dev); | |
1056 | ||
640f2a23 | 1057 | return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0)); |
c72758f3 MP |
1058 | } |
1059 | ||
86b37281 MP |
1060 | static ssize_t disk_discard_alignment_show(struct device *dev, |
1061 | struct device_attribute *attr, | |
1062 | char *buf) | |
1063 | { | |
1064 | struct gendisk *disk = dev_to_disk(dev); | |
1065 | ||
4e1462ff | 1066 | return sprintf(buf, "%d\n", bdev_alignment_offset(disk->part0)); |
86b37281 MP |
1067 | } |
1068 | ||
13927b31 MC |
1069 | static ssize_t diskseq_show(struct device *dev, |
1070 | struct device_attribute *attr, char *buf) | |
1071 | { | |
1072 | struct gendisk *disk = dev_to_disk(dev); | |
1073 | ||
1074 | return sprintf(buf, "%llu\n", disk->diskseq); | |
1075 | } | |
1076 | ||
5657a819 JP |
1077 | static DEVICE_ATTR(range, 0444, disk_range_show, NULL); |
1078 | static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); | |
1079 | static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); | |
1080 | static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); | |
1081 | static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); | |
1082 | static DEVICE_ATTR(size, 0444, part_size_show, NULL); | |
1083 | static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); | |
1084 | static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); | |
1085 | static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); | |
1086 | static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); | |
1087 | static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); | |
1088 | static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); | |
13927b31 | 1089 | static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL); |
3ad5cee5 | 1090 | |
c17bb495 | 1091 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
3ad5cee5 CH |
1092 | ssize_t part_fail_show(struct device *dev, |
1093 | struct device_attribute *attr, char *buf) | |
1094 | { | |
0d02129e | 1095 | return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail); |
3ad5cee5 CH |
1096 | } |
1097 | ||
1098 | ssize_t part_fail_store(struct device *dev, | |
1099 | struct device_attribute *attr, | |
1100 | const char *buf, size_t count) | |
1101 | { | |
3ad5cee5 CH |
1102 | int i; |
1103 | ||
1104 | if (count > 0 && sscanf(buf, "%d", &i) > 0) | |
0d02129e | 1105 | dev_to_bdev(dev)->bd_make_it_fail = i; |
3ad5cee5 CH |
1106 | |
1107 | return count; | |
1108 | } | |
1109 | ||
edfaa7c3 | 1110 | static struct device_attribute dev_attr_fail = |
5657a819 | 1111 | __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); |
3ad5cee5 CH |
1112 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
1113 | ||
581d4e28 JA |
1114 | #ifdef CONFIG_FAIL_IO_TIMEOUT |
1115 | static struct device_attribute dev_attr_fail_timeout = | |
5657a819 | 1116 | __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); |
581d4e28 | 1117 | #endif |
edfaa7c3 KS |
1118 | |
1119 | static struct attribute *disk_attrs[] = { | |
1120 | &dev_attr_range.attr, | |
1f014290 | 1121 | &dev_attr_ext_range.attr, |
edfaa7c3 | 1122 | &dev_attr_removable.attr, |
8ddcd653 | 1123 | &dev_attr_hidden.attr, |
1c9ce527 | 1124 | &dev_attr_ro.attr, |
edfaa7c3 | 1125 | &dev_attr_size.attr, |
c72758f3 | 1126 | &dev_attr_alignment_offset.attr, |
86b37281 | 1127 | &dev_attr_discard_alignment.attr, |
edfaa7c3 KS |
1128 | &dev_attr_capability.attr, |
1129 | &dev_attr_stat.attr, | |
316d315b | 1130 | &dev_attr_inflight.attr, |
99e6608c | 1131 | &dev_attr_badblocks.attr, |
2bc8cda5 CH |
1132 | &dev_attr_events.attr, |
1133 | &dev_attr_events_async.attr, | |
1134 | &dev_attr_events_poll_msecs.attr, | |
13927b31 | 1135 | &dev_attr_diskseq.attr, |
edfaa7c3 KS |
1136 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1137 | &dev_attr_fail.attr, | |
581d4e28 JA |
1138 | #endif |
1139 | #ifdef CONFIG_FAIL_IO_TIMEOUT | |
1140 | &dev_attr_fail_timeout.attr, | |
edfaa7c3 KS |
1141 | #endif |
1142 | NULL | |
1143 | }; | |
1144 | ||
9438b3e0 DW |
1145 | static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) |
1146 | { | |
1147 | struct device *dev = container_of(kobj, typeof(*dev), kobj); | |
1148 | struct gendisk *disk = dev_to_disk(dev); | |
1149 | ||
1150 | if (a == &dev_attr_badblocks.attr && !disk->bb) | |
1151 | return 0; | |
1152 | return a->mode; | |
1153 | } | |
1154 | ||
edfaa7c3 KS |
1155 | static struct attribute_group disk_attr_group = { |
1156 | .attrs = disk_attrs, | |
9438b3e0 | 1157 | .is_visible = disk_visible, |
edfaa7c3 KS |
1158 | }; |
1159 | ||
a4dbd674 | 1160 | static const struct attribute_group *disk_attr_groups[] = { |
edfaa7c3 | 1161 | &disk_attr_group, |
cc5c516d CH |
1162 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
1163 | &blk_trace_attr_group, | |
1164 | #endif | |
edfaa7c3 | 1165 | NULL |
1da177e4 LT |
1166 | }; |
1167 | ||
b5bd357c LC |
1168 | /** |
1169 | * disk_release - releases all allocated resources of the gendisk | |
1170 | * @dev: the device representing this disk | |
1171 | * | |
1172 | * This function releases all allocated resources of the gendisk. | |
1173 | * | |
b5bd357c LC |
1174 | * Drivers which used __device_add_disk() have a gendisk with a request_queue |
1175 | * assigned. Since the request_queue sits on top of the gendisk for these | |
1176 | * drivers we also call blk_put_queue() for them, and we expect the | |
1177 | * request_queue refcount to reach 0 at this point, and so the request_queue | |
1178 | * will also be freed prior to the disk. | |
e8c7d14a LC |
1179 | * |
1180 | * Context: can sleep | |
b5bd357c | 1181 | */ |
edfaa7c3 | 1182 | static void disk_release(struct device *dev) |
1da177e4 | 1183 | { |
edfaa7c3 KS |
1184 | struct gendisk *disk = dev_to_disk(dev); |
1185 | ||
e8c7d14a | 1186 | might_sleep(); |
a2041761 | 1187 | WARN_ON_ONCE(disk_live(disk)); |
e8c7d14a | 1188 | |
c5db2cfc CH |
1189 | /* |
1190 | * To undo the all initialization from blk_mq_init_allocated_queue in | |
1191 | * case of a probe failure where add_disk is never called we have to | |
1192 | * call blk_mq_exit_queue here. We can't do this for the more common | |
1193 | * teardown case (yet) as the tagset can be gone by the time the disk | |
1194 | * is released once it was added. | |
1195 | */ | |
1196 | if (queue_is_mq(disk->queue) && | |
1197 | test_bit(GD_OWNS_QUEUE, &disk->state) && | |
1198 | !test_bit(GD_ADDED, &disk->state)) | |
1199 | blk_mq_exit_queue(disk->queue); | |
1200 | ||
b6553bef CH |
1201 | blkcg_exit_disk(disk); |
1202 | ||
46754bd0 | 1203 | bioset_exit(&disk->bio_split); |
2a19b28f | 1204 | |
77ea887e | 1205 | disk_release_events(disk); |
1da177e4 | 1206 | kfree(disk->random); |
5d400665 | 1207 | disk_free_zone_bitmaps(disk); |
a33df75c | 1208 | xa_destroy(&disk->part_tbl); |
1059699f | 1209 | |
d152c682 | 1210 | disk->queue->disk = NULL; |
61a35cfc | 1211 | blk_put_queue(disk->queue); |
76792055 CH |
1212 | |
1213 | if (test_bit(GD_ADDED, &disk->state) && disk->fops->free_disk) | |
1214 | disk->fops->free_disk(disk); | |
1215 | ||
2f4731dc | 1216 | iput(disk->part0->bd_inode); /* frees the disk */ |
1da177e4 | 1217 | } |
87eb7107 | 1218 | |
23680f0b | 1219 | static int block_uevent(const struct device *dev, struct kobj_uevent_env *env) |
87eb7107 | 1220 | { |
23680f0b | 1221 | const struct gendisk *disk = dev_to_disk(dev); |
87eb7107 MC |
1222 | |
1223 | return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq); | |
1224 | } | |
1225 | ||
edfaa7c3 KS |
1226 | struct class block_class = { |
1227 | .name = "block", | |
87eb7107 | 1228 | .dev_uevent = block_uevent, |
1da177e4 LT |
1229 | }; |
1230 | ||
a9b12f8b | 1231 | static char *block_devnode(const struct device *dev, umode_t *mode, |
050a4f34 JA |
1232 | kuid_t *uid, kgid_t *gid) |
1233 | { | |
1234 | struct gendisk *disk = dev_to_disk(dev); | |
1235 | ||
1236 | if (disk->fops->devnode) | |
1237 | return disk->fops->devnode(disk, mode); | |
1238 | return NULL; | |
1239 | } | |
1240 | ||
ef45fe47 | 1241 | const struct device_type disk_type = { |
edfaa7c3 KS |
1242 | .name = "disk", |
1243 | .groups = disk_attr_groups, | |
1244 | .release = disk_release, | |
050a4f34 | 1245 | .devnode = block_devnode, |
1da177e4 LT |
1246 | }; |
1247 | ||
a6e2ba88 | 1248 | #ifdef CONFIG_PROC_FS |
cf771cb5 TH |
1249 | /* |
1250 | * aggregate disk stat collector. Uses the same stats that the sysfs | |
1251 | * entries do, above, but makes them available through one seq_file. | |
1252 | * | |
1253 | * The output looks suspiciously like /proc/partitions with a bunch of | |
1254 | * extra fields. | |
1255 | */ | |
1256 | static int diskstats_show(struct seq_file *seqf, void *v) | |
1da177e4 LT |
1257 | { |
1258 | struct gendisk *gp = v; | |
ad1eaa53 | 1259 | struct block_device *hd; |
e016b782 | 1260 | unsigned int inflight; |
ea18e0f0 | 1261 | struct disk_stats stat; |
7fae67cc | 1262 | unsigned long idx; |
1da177e4 LT |
1263 | |
1264 | /* | |
ed9e1982 | 1265 | if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) |
cf771cb5 | 1266 | seq_puts(seqf, "major minor name" |
1da177e4 LT |
1267 | " rio rmerge rsect ruse wio wmerge " |
1268 | "wsect wuse running use aveq" | |
1269 | "\n\n"); | |
1270 | */ | |
9f5e4865 | 1271 | |
7fae67cc CH |
1272 | rcu_read_lock(); |
1273 | xa_for_each(&gp->part_tbl, idx, hd) { | |
1274 | if (bdev_is_partition(hd) && !bdev_nr_sectors(hd)) | |
1275 | continue; | |
b2f609e1 | 1276 | if (queue_is_mq(gp->queue)) |
ad1eaa53 | 1277 | inflight = blk_mq_in_flight(gp->queue, hd); |
b2f609e1 | 1278 | else |
ad1eaa53 | 1279 | inflight = part_in_flight(hd); |
ea18e0f0 | 1280 | |
86d73312 ZW |
1281 | if (inflight) { |
1282 | part_stat_lock(); | |
1283 | update_io_ticks(hd, jiffies, true); | |
1284 | part_stat_unlock(); | |
1285 | } | |
1286 | part_stat_read_all(hd, &stat); | |
26e2d7a3 | 1287 | seq_printf(seqf, "%4d %7d %pg " |
bdca3c87 MC |
1288 | "%lu %lu %lu %u " |
1289 | "%lu %lu %lu %u " | |
1290 | "%u %u %u " | |
b6866318 KK |
1291 | "%lu %lu %lu %u " |
1292 | "%lu %u" | |
1293 | "\n", | |
26e2d7a3 | 1294 | MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd, |
ea18e0f0 KK |
1295 | stat.ios[STAT_READ], |
1296 | stat.merges[STAT_READ], | |
1297 | stat.sectors[STAT_READ], | |
1298 | (unsigned int)div_u64(stat.nsecs[STAT_READ], | |
1299 | NSEC_PER_MSEC), | |
1300 | stat.ios[STAT_WRITE], | |
1301 | stat.merges[STAT_WRITE], | |
1302 | stat.sectors[STAT_WRITE], | |
1303 | (unsigned int)div_u64(stat.nsecs[STAT_WRITE], | |
1304 | NSEC_PER_MSEC), | |
e016b782 | 1305 | inflight, |
ea18e0f0 | 1306 | jiffies_to_msecs(stat.io_ticks), |
8cd5b8fc KK |
1307 | (unsigned int)div_u64(stat.nsecs[STAT_READ] + |
1308 | stat.nsecs[STAT_WRITE] + | |
1309 | stat.nsecs[STAT_DISCARD] + | |
1310 | stat.nsecs[STAT_FLUSH], | |
1311 | NSEC_PER_MSEC), | |
ea18e0f0 KK |
1312 | stat.ios[STAT_DISCARD], |
1313 | stat.merges[STAT_DISCARD], | |
1314 | stat.sectors[STAT_DISCARD], | |
1315 | (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], | |
1316 | NSEC_PER_MSEC), | |
1317 | stat.ios[STAT_FLUSH], | |
1318 | (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], | |
1319 | NSEC_PER_MSEC) | |
28f39d55 | 1320 | ); |
1da177e4 | 1321 | } |
7fae67cc | 1322 | rcu_read_unlock(); |
9f5e4865 | 1323 | |
1da177e4 LT |
1324 | return 0; |
1325 | } | |
1326 | ||
31d85ab2 | 1327 | static const struct seq_operations diskstats_op = { |
def4e38d TH |
1328 | .start = disk_seqf_start, |
1329 | .next = disk_seqf_next, | |
1330 | .stop = disk_seqf_stop, | |
1da177e4 LT |
1331 | .show = diskstats_show |
1332 | }; | |
f500975a AD |
1333 | |
1334 | static int __init proc_genhd_init(void) | |
1335 | { | |
fddda2b7 CH |
1336 | proc_create_seq("diskstats", 0, NULL, &diskstats_op); |
1337 | proc_create_seq("partitions", 0, NULL, &partitions_op); | |
f500975a AD |
1338 | return 0; |
1339 | } | |
1340 | module_init(proc_genhd_init); | |
a6e2ba88 | 1341 | #endif /* CONFIG_PROC_FS */ |
1da177e4 | 1342 | |
c97d93c3 CH |
1343 | dev_t part_devt(struct gendisk *disk, u8 partno) |
1344 | { | |
0e0ccdec | 1345 | struct block_device *part; |
c97d93c3 CH |
1346 | dev_t devt = 0; |
1347 | ||
0e0ccdec CH |
1348 | rcu_read_lock(); |
1349 | part = xa_load(&disk->part_tbl, partno); | |
1350 | if (part) | |
c97d93c3 | 1351 | devt = part->bd_dev; |
0e0ccdec | 1352 | rcu_read_unlock(); |
c97d93c3 CH |
1353 | |
1354 | return devt; | |
1355 | } | |
1356 | ||
cf771cb5 | 1357 | dev_t blk_lookup_devt(const char *name, int partno) |
a142be85 | 1358 | { |
def4e38d TH |
1359 | dev_t devt = MKDEV(0, 0); |
1360 | struct class_dev_iter iter; | |
1361 | struct device *dev; | |
a142be85 | 1362 | |
def4e38d TH |
1363 | class_dev_iter_init(&iter, &block_class, NULL, &disk_type); |
1364 | while ((dev = class_dev_iter_next(&iter))) { | |
a142be85 | 1365 | struct gendisk *disk = dev_to_disk(dev); |
a142be85 | 1366 | |
3ada8b7e | 1367 | if (strcmp(dev_name(dev), name)) |
f331c029 | 1368 | continue; |
f331c029 | 1369 | |
41b8c853 NB |
1370 | if (partno < disk->minors) { |
1371 | /* We need to return the right devno, even | |
1372 | * if the partition doesn't exist yet. | |
1373 | */ | |
1374 | devt = MKDEV(MAJOR(dev->devt), | |
1375 | MINOR(dev->devt) + partno); | |
c97d93c3 CH |
1376 | } else { |
1377 | devt = part_devt(disk, partno); | |
1378 | if (devt) | |
1379 | break; | |
def4e38d | 1380 | } |
5c0ef6d0 | 1381 | } |
def4e38d | 1382 | class_dev_iter_exit(&iter); |
edfaa7c3 KS |
1383 | return devt; |
1384 | } | |
edfaa7c3 | 1385 | |
4a1fa41d CH |
1386 | struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id, |
1387 | struct lock_class_key *lkclass) | |
1946089a CL |
1388 | { |
1389 | struct gendisk *disk; | |
1390 | ||
c1b511eb | 1391 | disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); |
f93af2a4 | 1392 | if (!disk) |
aa0c680c | 1393 | return NULL; |
6c23a968 | 1394 | |
46754bd0 CH |
1395 | if (bioset_init(&disk->bio_split, BIO_POOL_SIZE, 0, 0)) |
1396 | goto out_free_disk; | |
1397 | ||
edb0872f CH |
1398 | disk->bdi = bdi_alloc(node_id); |
1399 | if (!disk->bdi) | |
46754bd0 | 1400 | goto out_free_bioset; |
edb0872f | 1401 | |
17220ca5 PB |
1402 | /* bdev_alloc() might need the queue, set before the first call */ |
1403 | disk->queue = q; | |
1404 | ||
cb8432d6 CH |
1405 | disk->part0 = bdev_alloc(disk, 0); |
1406 | if (!disk->part0) | |
edb0872f | 1407 | goto out_free_bdi; |
22ae8ce8 | 1408 | |
f93af2a4 | 1409 | disk->node_id = node_id; |
a8698707 | 1410 | mutex_init(&disk->open_mutex); |
a33df75c CH |
1411 | xa_init(&disk->part_tbl); |
1412 | if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL)) | |
1413 | goto out_destroy_part_tbl; | |
f93af2a4 | 1414 | |
b6553bef CH |
1415 | if (blkcg_init_disk(disk)) |
1416 | goto out_erase_part0; | |
1417 | ||
f93af2a4 CH |
1418 | rand_initialize_disk(disk); |
1419 | disk_to_dev(disk)->class = &block_class; | |
1420 | disk_to_dev(disk)->type = &disk_type; | |
1421 | device_initialize(disk_to_dev(disk)); | |
cf179948 | 1422 | inc_diskseq(disk); |
d152c682 | 1423 | q->disk = disk; |
4dcc4874 | 1424 | lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0); |
0dbcfe24 CH |
1425 | #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED |
1426 | INIT_LIST_HEAD(&disk->slave_bdevs); | |
1427 | #endif | |
1da177e4 | 1428 | return disk; |
f93af2a4 | 1429 | |
b6553bef CH |
1430 | out_erase_part0: |
1431 | xa_erase(&disk->part_tbl, 0); | |
a33df75c CH |
1432 | out_destroy_part_tbl: |
1433 | xa_destroy(&disk->part_tbl); | |
06cc978d | 1434 | disk->part0->bd_disk = NULL; |
2f4731dc | 1435 | iput(disk->part0->bd_inode); |
edb0872f CH |
1436 | out_free_bdi: |
1437 | bdi_put(disk->bdi); | |
46754bd0 CH |
1438 | out_free_bioset: |
1439 | bioset_exit(&disk->bio_split); | |
f93af2a4 CH |
1440 | out_free_disk: |
1441 | kfree(disk); | |
1442 | return NULL; | |
1da177e4 | 1443 | } |
1da177e4 | 1444 | |
4dcc4874 | 1445 | struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass) |
f525464a CH |
1446 | { |
1447 | struct request_queue *q; | |
1448 | struct gendisk *disk; | |
1449 | ||
80bd4a7a | 1450 | q = blk_alloc_queue(node); |
f525464a CH |
1451 | if (!q) |
1452 | return NULL; | |
1453 | ||
4a1fa41d | 1454 | disk = __alloc_disk_node(q, node, lkclass); |
f525464a | 1455 | if (!disk) { |
6f8191fd | 1456 | blk_put_queue(q); |
f525464a CH |
1457 | return NULL; |
1458 | } | |
6f8191fd | 1459 | set_bit(GD_OWNS_QUEUE, &disk->state); |
f525464a CH |
1460 | return disk; |
1461 | } | |
1462 | EXPORT_SYMBOL(__blk_alloc_disk); | |
1463 | ||
b5bd357c LC |
1464 | /** |
1465 | * put_disk - decrements the gendisk refcount | |
0d20dcc2 | 1466 | * @disk: the struct gendisk to decrement the refcount for |
b5bd357c LC |
1467 | * |
1468 | * This decrements the refcount for the struct gendisk. When this reaches 0 | |
1469 | * we'll have disk_release() called. | |
e8c7d14a | 1470 | * |
c5db2cfc CH |
1471 | * Note: for blk-mq disk put_disk must be called before freeing the tag_set |
1472 | * when handling probe errors (that is before add_disk() is called). | |
1473 | * | |
e8c7d14a LC |
1474 | * Context: Any context, but the last reference must not be dropped from |
1475 | * atomic context. | |
b5bd357c | 1476 | */ |
1da177e4 LT |
1477 | void put_disk(struct gendisk *disk) |
1478 | { | |
1479 | if (disk) | |
efdc41c8 | 1480 | put_device(disk_to_dev(disk)); |
1da177e4 | 1481 | } |
1da177e4 LT |
1482 | EXPORT_SYMBOL(put_disk); |
1483 | ||
e3264a4d HR |
1484 | static void set_disk_ro_uevent(struct gendisk *gd, int ro) |
1485 | { | |
1486 | char event[] = "DISK_RO=1"; | |
1487 | char *envp[] = { event, NULL }; | |
1488 | ||
1489 | if (!ro) | |
1490 | event[8] = '0'; | |
1491 | kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); | |
1492 | } | |
1493 | ||
52f019d4 CH |
1494 | /** |
1495 | * set_disk_ro - set a gendisk read-only | |
1496 | * @disk: gendisk to operate on | |
7f31bee3 | 1497 | * @read_only: %true to set the disk read-only, %false set the disk read/write |
52f019d4 CH |
1498 | * |
1499 | * This function is used to indicate whether a given disk device should have its | |
1500 | * read-only flag set. set_disk_ro() is typically used by device drivers to | |
1501 | * indicate whether the underlying physical device is write-protected. | |
1502 | */ | |
1503 | void set_disk_ro(struct gendisk *disk, bool read_only) | |
1da177e4 | 1504 | { |
52f019d4 CH |
1505 | if (read_only) { |
1506 | if (test_and_set_bit(GD_READ_ONLY, &disk->state)) | |
1507 | return; | |
1508 | } else { | |
1509 | if (!test_and_clear_bit(GD_READ_ONLY, &disk->state)) | |
1510 | return; | |
e3264a4d | 1511 | } |
52f019d4 | 1512 | set_disk_ro_uevent(disk, read_only); |
1da177e4 | 1513 | } |
1da177e4 LT |
1514 | EXPORT_SYMBOL(set_disk_ro); |
1515 | ||
cf179948 MC |
1516 | void inc_diskseq(struct gendisk *disk) |
1517 | { | |
1518 | disk->diskseq = atomic64_inc_return(&diskseq); | |
1519 | } |