]>
Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
6a0cb1bc HR |
2 | /* |
3 | * Zoned block device handling | |
4 | * | |
5 | * Copyright (c) 2015, Hannes Reinecke | |
6 | * Copyright (c) 2015, SUSE Linux GmbH | |
7 | * | |
8 | * Copyright (c) 2016, Damien Le Moal | |
9 | * Copyright (c) 2016, Western Digital | |
10 | */ | |
11 | ||
12 | #include <linux/kernel.h> | |
13 | #include <linux/module.h> | |
14 | #include <linux/rbtree.h> | |
15 | #include <linux/blkdev.h> | |
bf505456 | 16 | #include <linux/blk-mq.h> |
26202928 DLM |
17 | #include <linux/mm.h> |
18 | #include <linux/vmalloc.h> | |
bd976e52 | 19 | #include <linux/sched/mm.h> |
6a0cb1bc | 20 | |
a2d6b3a2 DLM |
21 | #include "blk.h" |
22 | ||
02694e86 CK |
23 | #define ZONE_COND_NAME(name) [BLK_ZONE_COND_##name] = #name |
24 | static const char *const zone_cond_name[] = { | |
25 | ZONE_COND_NAME(NOT_WP), | |
26 | ZONE_COND_NAME(EMPTY), | |
27 | ZONE_COND_NAME(IMP_OPEN), | |
28 | ZONE_COND_NAME(EXP_OPEN), | |
29 | ZONE_COND_NAME(CLOSED), | |
30 | ZONE_COND_NAME(READONLY), | |
31 | ZONE_COND_NAME(FULL), | |
32 | ZONE_COND_NAME(OFFLINE), | |
33 | }; | |
34 | #undef ZONE_COND_NAME | |
35 | ||
36 | /** | |
37 | * blk_zone_cond_str - Return string XXX in BLK_ZONE_COND_XXX. | |
38 | * @zone_cond: BLK_ZONE_COND_XXX. | |
39 | * | |
40 | * Description: Centralize block layer function to convert BLK_ZONE_COND_XXX | |
41 | * into string format. Useful in the debugging and tracing zone conditions. For | |
42 | * invalid BLK_ZONE_COND_XXX it returns string "UNKNOWN". | |
43 | */ | |
44 | const char *blk_zone_cond_str(enum blk_zone_cond zone_cond) | |
45 | { | |
46 | static const char *zone_cond_str = "UNKNOWN"; | |
47 | ||
48 | if (zone_cond < ARRAY_SIZE(zone_cond_name) && zone_cond_name[zone_cond]) | |
49 | zone_cond_str = zone_cond_name[zone_cond]; | |
50 | ||
51 | return zone_cond_str; | |
52 | } | |
53 | EXPORT_SYMBOL_GPL(blk_zone_cond_str); | |
54 | ||
6cc77e9c CH |
55 | /* |
56 | * Return true if a request is a write requests that needs zone write locking. | |
57 | */ | |
58 | bool blk_req_needs_zone_write_lock(struct request *rq) | |
59 | { | |
60 | if (!rq->q->seq_zones_wlock) | |
61 | return false; | |
62 | ||
63 | if (blk_rq_is_passthrough(rq)) | |
64 | return false; | |
65 | ||
66 | switch (req_op(rq)) { | |
67 | case REQ_OP_WRITE_ZEROES: | |
68 | case REQ_OP_WRITE_SAME: | |
69 | case REQ_OP_WRITE: | |
70 | return blk_rq_zone_is_seq(rq); | |
71 | default: | |
72 | return false; | |
73 | } | |
74 | } | |
75 | EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock); | |
76 | ||
1392d370 JT |
77 | bool blk_req_zone_write_trylock(struct request *rq) |
78 | { | |
79 | unsigned int zno = blk_rq_zone_no(rq); | |
80 | ||
81 | if (test_and_set_bit(zno, rq->q->seq_zones_wlock)) | |
82 | return false; | |
83 | ||
84 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
85 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
86 | ||
87 | return true; | |
88 | } | |
89 | EXPORT_SYMBOL_GPL(blk_req_zone_write_trylock); | |
90 | ||
6cc77e9c CH |
91 | void __blk_req_zone_write_lock(struct request *rq) |
92 | { | |
93 | if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq), | |
94 | rq->q->seq_zones_wlock))) | |
95 | return; | |
96 | ||
97 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
98 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
99 | } | |
100 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock); | |
101 | ||
102 | void __blk_req_zone_write_unlock(struct request *rq) | |
103 | { | |
104 | rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED; | |
105 | if (rq->q->seq_zones_wlock) | |
106 | WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq), | |
107 | rq->q->seq_zones_wlock)); | |
108 | } | |
109 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock); | |
110 | ||
a91e1380 DLM |
111 | /** |
112 | * blkdev_nr_zones - Get number of zones | |
9b38bb4b | 113 | * @disk: Target gendisk |
a91e1380 | 114 | * |
9b38bb4b CH |
115 | * Return the total number of zones of a zoned block device. For a block |
116 | * device without zone capabilities, the number of zones is always 0. | |
a91e1380 | 117 | */ |
9b38bb4b | 118 | unsigned int blkdev_nr_zones(struct gendisk *disk) |
a91e1380 | 119 | { |
9b38bb4b | 120 | sector_t zone_sectors = blk_queue_zone_sectors(disk->queue); |
a91e1380 | 121 | |
9b38bb4b | 122 | if (!blk_queue_is_zoned(disk->queue)) |
a91e1380 | 123 | return 0; |
9b38bb4b | 124 | return (get_capacity(disk) + zone_sectors - 1) >> ilog2(zone_sectors); |
a91e1380 DLM |
125 | } |
126 | EXPORT_SYMBOL_GPL(blkdev_nr_zones); | |
127 | ||
6a0cb1bc HR |
128 | /** |
129 | * blkdev_report_zones - Get zones information | |
130 | * @bdev: Target block device | |
131 | * @sector: Sector from which to report zones | |
d4100351 CH |
132 | * @nr_zones: Maximum number of zones to report |
133 | * @cb: Callback function called for each reported zone | |
134 | * @data: Private data for the callback | |
6a0cb1bc HR |
135 | * |
136 | * Description: | |
d4100351 CH |
137 | * Get zone information starting from the zone containing @sector for at most |
138 | * @nr_zones, and call @cb for each zone reported by the device. | |
139 | * To report all zones in a device starting from @sector, the BLK_ALL_ZONES | |
140 | * constant can be passed to @nr_zones. | |
141 | * Returns the number of zones reported by the device, or a negative errno | |
142 | * value in case of failure. | |
143 | * | |
144 | * Note: The caller must use memalloc_noXX_save/restore() calls to control | |
145 | * memory allocations done within this function. | |
6a0cb1bc | 146 | */ |
e76239a3 | 147 | int blkdev_report_zones(struct block_device *bdev, sector_t sector, |
d4100351 | 148 | unsigned int nr_zones, report_zones_cb cb, void *data) |
6a0cb1bc | 149 | { |
ceeb373a | 150 | struct gendisk *disk = bdev->bd_disk; |
5eac3eb3 | 151 | sector_t capacity = get_capacity(disk); |
6a0cb1bc | 152 | |
d4100351 CH |
153 | if (!blk_queue_is_zoned(bdev_get_queue(bdev)) || |
154 | WARN_ON_ONCE(!disk->fops->report_zones)) | |
e76239a3 | 155 | return -EOPNOTSUPP; |
6a0cb1bc | 156 | |
d4100351 | 157 | if (!nr_zones || sector >= capacity) |
6a0cb1bc | 158 | return 0; |
6a0cb1bc | 159 | |
d4100351 | 160 | return disk->fops->report_zones(disk, sector, nr_zones, cb, data); |
6a0cb1bc HR |
161 | } |
162 | EXPORT_SYMBOL_GPL(blkdev_report_zones); | |
163 | ||
6e33dbf2 | 164 | static inline bool blkdev_allow_reset_all_zones(struct block_device *bdev, |
c7a1d926 | 165 | sector_t sector, |
6e33dbf2 CK |
166 | sector_t nr_sectors) |
167 | { | |
168 | if (!blk_queue_zone_resetall(bdev_get_queue(bdev))) | |
169 | return false; | |
170 | ||
6e33dbf2 | 171 | /* |
5eac3eb3 DLM |
172 | * REQ_OP_ZONE_RESET_ALL can be executed only if the number of sectors |
173 | * of the applicable zone range is the entire disk. | |
6e33dbf2 | 174 | */ |
5eac3eb3 | 175 | return !sector && nr_sectors == get_capacity(bdev->bd_disk); |
6e33dbf2 CK |
176 | } |
177 | ||
6a0cb1bc | 178 | /** |
6c1b1da5 | 179 | * blkdev_zone_mgmt - Execute a zone management operation on a range of zones |
6a0cb1bc | 180 | * @bdev: Target block device |
6c1b1da5 AJ |
181 | * @op: Operation to be performed on the zones |
182 | * @sector: Start sector of the first zone to operate on | |
183 | * @nr_sectors: Number of sectors, should be at least the length of one zone and | |
184 | * must be zone size aligned. | |
6a0cb1bc HR |
185 | * @gfp_mask: Memory allocation flags (for bio_alloc) |
186 | * | |
187 | * Description: | |
6c1b1da5 | 188 | * Perform the specified operation on the range of zones specified by |
6a0cb1bc HR |
189 | * @sector..@sector+@nr_sectors. Specifying the entire disk sector range |
190 | * is valid, but the specified range should not contain conventional zones. | |
6c1b1da5 AJ |
191 | * The operation to execute on each zone can be a zone reset, open, close |
192 | * or finish request. | |
6a0cb1bc | 193 | */ |
6c1b1da5 AJ |
194 | int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op, |
195 | sector_t sector, sector_t nr_sectors, | |
196 | gfp_t gfp_mask) | |
6a0cb1bc HR |
197 | { |
198 | struct request_queue *q = bdev_get_queue(bdev); | |
6c1b1da5 | 199 | sector_t zone_sectors = blk_queue_zone_sectors(q); |
5eac3eb3 | 200 | sector_t capacity = get_capacity(bdev->bd_disk); |
6a0cb1bc | 201 | sector_t end_sector = sector + nr_sectors; |
a2d6b3a2 | 202 | struct bio *bio = NULL; |
6a0cb1bc HR |
203 | int ret; |
204 | ||
6a0cb1bc HR |
205 | if (!blk_queue_is_zoned(q)) |
206 | return -EOPNOTSUPP; | |
207 | ||
a2d6b3a2 DLM |
208 | if (bdev_read_only(bdev)) |
209 | return -EPERM; | |
210 | ||
6c1b1da5 AJ |
211 | if (!op_is_zone_mgmt(op)) |
212 | return -EOPNOTSUPP; | |
213 | ||
11bde986 | 214 | if (end_sector <= sector || end_sector > capacity) |
6a0cb1bc HR |
215 | /* Out of range */ |
216 | return -EINVAL; | |
217 | ||
218 | /* Check alignment (handle eventual smaller last zone) */ | |
6a0cb1bc HR |
219 | if (sector & (zone_sectors - 1)) |
220 | return -EINVAL; | |
221 | ||
5eac3eb3 | 222 | if ((nr_sectors & (zone_sectors - 1)) && end_sector != capacity) |
6a0cb1bc HR |
223 | return -EINVAL; |
224 | ||
225 | while (sector < end_sector) { | |
a2d6b3a2 | 226 | bio = blk_next_bio(bio, 0, gfp_mask); |
74d46992 | 227 | bio_set_dev(bio, bdev); |
6a0cb1bc | 228 | |
c7a1d926 DLM |
229 | /* |
230 | * Special case for the zone reset operation that reset all | |
231 | * zones, this is useful for applications like mkfs. | |
232 | */ | |
6c1b1da5 AJ |
233 | if (op == REQ_OP_ZONE_RESET && |
234 | blkdev_allow_reset_all_zones(bdev, sector, nr_sectors)) { | |
faa44c69 | 235 | bio->bi_opf = REQ_OP_ZONE_RESET_ALL | REQ_SYNC; |
c7a1d926 DLM |
236 | break; |
237 | } | |
238 | ||
8e42d239 | 239 | bio->bi_opf = op | REQ_SYNC; |
c7a1d926 | 240 | bio->bi_iter.bi_sector = sector; |
6a0cb1bc HR |
241 | sector += zone_sectors; |
242 | ||
243 | /* This may take a while, so be nice to others */ | |
244 | cond_resched(); | |
6a0cb1bc HR |
245 | } |
246 | ||
a2d6b3a2 DLM |
247 | ret = submit_bio_wait(bio); |
248 | bio_put(bio); | |
249 | ||
a2d6b3a2 | 250 | return ret; |
6a0cb1bc | 251 | } |
6c1b1da5 | 252 | EXPORT_SYMBOL_GPL(blkdev_zone_mgmt); |
3ed05a98 | 253 | |
d4100351 CH |
254 | struct zone_report_args { |
255 | struct blk_zone __user *zones; | |
256 | }; | |
257 | ||
258 | static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx, | |
259 | void *data) | |
260 | { | |
261 | struct zone_report_args *args = data; | |
262 | ||
263 | if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone))) | |
264 | return -EFAULT; | |
265 | return 0; | |
266 | } | |
267 | ||
56c4bddb | 268 | /* |
3ed05a98 ST |
269 | * BLKREPORTZONE ioctl processing. |
270 | * Called from blkdev_ioctl. | |
271 | */ | |
272 | int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
273 | unsigned int cmd, unsigned long arg) | |
274 | { | |
275 | void __user *argp = (void __user *)arg; | |
d4100351 | 276 | struct zone_report_args args; |
3ed05a98 ST |
277 | struct request_queue *q; |
278 | struct blk_zone_report rep; | |
3ed05a98 ST |
279 | int ret; |
280 | ||
281 | if (!argp) | |
282 | return -EINVAL; | |
283 | ||
284 | q = bdev_get_queue(bdev); | |
285 | if (!q) | |
286 | return -ENXIO; | |
287 | ||
288 | if (!blk_queue_is_zoned(q)) | |
289 | return -ENOTTY; | |
290 | ||
291 | if (!capable(CAP_SYS_ADMIN)) | |
292 | return -EACCES; | |
293 | ||
294 | if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report))) | |
295 | return -EFAULT; | |
296 | ||
297 | if (!rep.nr_zones) | |
298 | return -EINVAL; | |
299 | ||
d4100351 CH |
300 | args.zones = argp + sizeof(struct blk_zone_report); |
301 | ret = blkdev_report_zones(bdev, rep.sector, rep.nr_zones, | |
302 | blkdev_copy_zone_to_user, &args); | |
303 | if (ret < 0) | |
304 | return ret; | |
3ed05a98 | 305 | |
d4100351 | 306 | rep.nr_zones = ret; |
82394db7 | 307 | rep.flags = BLK_ZONE_REP_CAPACITY; |
d4100351 CH |
308 | if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report))) |
309 | return -EFAULT; | |
310 | return 0; | |
3ed05a98 ST |
311 | } |
312 | ||
e5113505 SK |
313 | static int blkdev_truncate_zone_range(struct block_device *bdev, fmode_t mode, |
314 | const struct blk_zone_range *zrange) | |
315 | { | |
316 | loff_t start, end; | |
317 | ||
318 | if (zrange->sector + zrange->nr_sectors <= zrange->sector || | |
319 | zrange->sector + zrange->nr_sectors > get_capacity(bdev->bd_disk)) | |
320 | /* Out of range */ | |
321 | return -EINVAL; | |
322 | ||
323 | start = zrange->sector << SECTOR_SHIFT; | |
324 | end = ((zrange->sector + zrange->nr_sectors) << SECTOR_SHIFT) - 1; | |
325 | ||
326 | return truncate_bdev_range(bdev, mode, start, end); | |
327 | } | |
328 | ||
56c4bddb | 329 | /* |
e876df1f | 330 | * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing. |
3ed05a98 ST |
331 | * Called from blkdev_ioctl. |
332 | */ | |
e876df1f AJ |
333 | int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode, |
334 | unsigned int cmd, unsigned long arg) | |
3ed05a98 ST |
335 | { |
336 | void __user *argp = (void __user *)arg; | |
337 | struct request_queue *q; | |
338 | struct blk_zone_range zrange; | |
e876df1f | 339 | enum req_opf op; |
e5113505 | 340 | int ret; |
3ed05a98 ST |
341 | |
342 | if (!argp) | |
343 | return -EINVAL; | |
344 | ||
345 | q = bdev_get_queue(bdev); | |
346 | if (!q) | |
347 | return -ENXIO; | |
348 | ||
349 | if (!blk_queue_is_zoned(q)) | |
350 | return -ENOTTY; | |
351 | ||
352 | if (!capable(CAP_SYS_ADMIN)) | |
353 | return -EACCES; | |
354 | ||
355 | if (!(mode & FMODE_WRITE)) | |
356 | return -EBADF; | |
357 | ||
358 | if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range))) | |
359 | return -EFAULT; | |
360 | ||
e876df1f AJ |
361 | switch (cmd) { |
362 | case BLKRESETZONE: | |
363 | op = REQ_OP_ZONE_RESET; | |
e5113505 SK |
364 | |
365 | /* Invalidate the page cache, including dirty pages. */ | |
366 | ret = blkdev_truncate_zone_range(bdev, mode, &zrange); | |
367 | if (ret) | |
368 | return ret; | |
e876df1f AJ |
369 | break; |
370 | case BLKOPENZONE: | |
371 | op = REQ_OP_ZONE_OPEN; | |
372 | break; | |
373 | case BLKCLOSEZONE: | |
374 | op = REQ_OP_ZONE_CLOSE; | |
375 | break; | |
376 | case BLKFINISHZONE: | |
377 | op = REQ_OP_ZONE_FINISH; | |
378 | break; | |
379 | default: | |
380 | return -ENOTTY; | |
381 | } | |
382 | ||
e5113505 SK |
383 | ret = blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors, |
384 | GFP_KERNEL); | |
385 | ||
386 | /* | |
387 | * Invalidate the page cache again for zone reset: writes can only be | |
388 | * direct for zoned devices so concurrent writes would not add any page | |
389 | * to the page cache after/during reset. The page cache may be filled | |
390 | * again due to concurrent reads though and dropping the pages for | |
391 | * these is fine. | |
392 | */ | |
393 | if (!ret && cmd == BLKRESETZONE) | |
394 | ret = blkdev_truncate_zone_range(bdev, mode, &zrange); | |
395 | ||
396 | return ret; | |
3ed05a98 | 397 | } |
bf505456 DLM |
398 | |
399 | static inline unsigned long *blk_alloc_zone_bitmap(int node, | |
400 | unsigned int nr_zones) | |
401 | { | |
402 | return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long), | |
403 | GFP_NOIO, node); | |
404 | } | |
405 | ||
bf505456 DLM |
406 | void blk_queue_free_zone_bitmaps(struct request_queue *q) |
407 | { | |
f216fdd7 CH |
408 | kfree(q->conv_zones_bitmap); |
409 | q->conv_zones_bitmap = NULL; | |
bf505456 DLM |
410 | kfree(q->seq_zones_wlock); |
411 | q->seq_zones_wlock = NULL; | |
412 | } | |
413 | ||
d4100351 CH |
414 | struct blk_revalidate_zone_args { |
415 | struct gendisk *disk; | |
f216fdd7 | 416 | unsigned long *conv_zones_bitmap; |
d4100351 | 417 | unsigned long *seq_zones_wlock; |
e94f5819 | 418 | unsigned int nr_zones; |
6c6b3549 | 419 | sector_t zone_sectors; |
d4100351 CH |
420 | sector_t sector; |
421 | }; | |
422 | ||
d9dd7308 DLM |
423 | /* |
424 | * Helper function to check the validity of zones of a zoned block device. | |
425 | */ | |
d4100351 CH |
426 | static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx, |
427 | void *data) | |
d9dd7308 | 428 | { |
d4100351 CH |
429 | struct blk_revalidate_zone_args *args = data; |
430 | struct gendisk *disk = args->disk; | |
d9dd7308 | 431 | struct request_queue *q = disk->queue; |
d9dd7308 DLM |
432 | sector_t capacity = get_capacity(disk); |
433 | ||
434 | /* | |
435 | * All zones must have the same size, with the exception on an eventual | |
436 | * smaller last zone. | |
437 | */ | |
6c6b3549 CH |
438 | if (zone->start == 0) { |
439 | if (zone->len == 0 || !is_power_of_2(zone->len)) { | |
440 | pr_warn("%s: Invalid zoned device with non power of two zone size (%llu)\n", | |
441 | disk->disk_name, zone->len); | |
442 | return -ENODEV; | |
443 | } | |
d9dd7308 | 444 | |
6c6b3549 CH |
445 | args->zone_sectors = zone->len; |
446 | args->nr_zones = (capacity + zone->len - 1) >> ilog2(zone->len); | |
447 | } else if (zone->start + args->zone_sectors < capacity) { | |
448 | if (zone->len != args->zone_sectors) { | |
449 | pr_warn("%s: Invalid zoned device with non constant zone size\n", | |
450 | disk->disk_name); | |
451 | return -ENODEV; | |
452 | } | |
453 | } else { | |
454 | if (zone->len > args->zone_sectors) { | |
455 | pr_warn("%s: Invalid zoned device with larger last zone size\n", | |
456 | disk->disk_name); | |
457 | return -ENODEV; | |
458 | } | |
d9dd7308 DLM |
459 | } |
460 | ||
461 | /* Check for holes in the zone report */ | |
d4100351 | 462 | if (zone->start != args->sector) { |
d9dd7308 | 463 | pr_warn("%s: Zone gap at sectors %llu..%llu\n", |
d4100351 CH |
464 | disk->disk_name, args->sector, zone->start); |
465 | return -ENODEV; | |
d9dd7308 DLM |
466 | } |
467 | ||
468 | /* Check zone type */ | |
469 | switch (zone->type) { | |
470 | case BLK_ZONE_TYPE_CONVENTIONAL: | |
e94f5819 CH |
471 | if (!args->conv_zones_bitmap) { |
472 | args->conv_zones_bitmap = | |
473 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
474 | if (!args->conv_zones_bitmap) | |
475 | return -ENOMEM; | |
476 | } | |
477 | set_bit(idx, args->conv_zones_bitmap); | |
478 | break; | |
d9dd7308 DLM |
479 | case BLK_ZONE_TYPE_SEQWRITE_REQ: |
480 | case BLK_ZONE_TYPE_SEQWRITE_PREF: | |
e94f5819 CH |
481 | if (!args->seq_zones_wlock) { |
482 | args->seq_zones_wlock = | |
483 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
484 | if (!args->seq_zones_wlock) | |
485 | return -ENOMEM; | |
486 | } | |
d9dd7308 DLM |
487 | break; |
488 | default: | |
489 | pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n", | |
490 | disk->disk_name, (int)zone->type, zone->start); | |
d4100351 | 491 | return -ENODEV; |
d9dd7308 DLM |
492 | } |
493 | ||
d4100351 CH |
494 | args->sector += zone->len; |
495 | return 0; | |
496 | } | |
497 | ||
bf505456 DLM |
498 | /** |
499 | * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps | |
500 | * @disk: Target disk | |
e732671a | 501 | * @update_driver_data: Callback to update driver data on the frozen disk |
bf505456 DLM |
502 | * |
503 | * Helper function for low-level device drivers to (re) allocate and initialize | |
504 | * a disk request queue zone bitmaps. This functions should normally be called | |
ae58954d CH |
505 | * within the disk ->revalidate method for blk-mq based drivers. For BIO based |
506 | * drivers only q->nr_zones needs to be updated so that the sysfs exposed value | |
507 | * is correct. | |
e732671a DLM |
508 | * If the @update_driver_data callback function is not NULL, the callback is |
509 | * executed with the device request queue frozen after all zones have been | |
510 | * checked. | |
bf505456 | 511 | */ |
e732671a DLM |
512 | int blk_revalidate_disk_zones(struct gendisk *disk, |
513 | void (*update_driver_data)(struct gendisk *disk)) | |
bf505456 DLM |
514 | { |
515 | struct request_queue *q = disk->queue; | |
e94f5819 CH |
516 | struct blk_revalidate_zone_args args = { |
517 | .disk = disk, | |
e94f5819 | 518 | }; |
6c6b3549 CH |
519 | unsigned int noio_flag; |
520 | int ret; | |
bf505456 | 521 | |
c98c3d09 CH |
522 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) |
523 | return -EIO; | |
ae58954d CH |
524 | if (WARN_ON_ONCE(!queue_is_mq(q))) |
525 | return -EIO; | |
bf505456 | 526 | |
1a1206dc JT |
527 | if (!get_capacity(disk)) |
528 | return -EIO; | |
529 | ||
e94f5819 | 530 | /* |
6c6b3549 CH |
531 | * Ensure that all memory allocations in this context are done as if |
532 | * GFP_NOIO was specified. | |
e94f5819 | 533 | */ |
6c6b3549 CH |
534 | noio_flag = memalloc_noio_save(); |
535 | ret = disk->fops->report_zones(disk, 0, UINT_MAX, | |
536 | blk_revalidate_zone_cb, &args); | |
2afdeb23 DLM |
537 | if (!ret) { |
538 | pr_warn("%s: No zones reported\n", disk->disk_name); | |
539 | ret = -ENODEV; | |
540 | } | |
6c6b3549 | 541 | memalloc_noio_restore(noio_flag); |
bf505456 | 542 | |
2afdeb23 DLM |
543 | /* |
544 | * If zones where reported, make sure that the entire disk capacity | |
545 | * has been checked. | |
546 | */ | |
547 | if (ret > 0 && args.sector != get_capacity(disk)) { | |
548 | pr_warn("%s: Missing zones from sector %llu\n", | |
549 | disk->disk_name, args.sector); | |
550 | ret = -ENODEV; | |
551 | } | |
552 | ||
bf505456 | 553 | /* |
6c6b3549 CH |
554 | * Install the new bitmaps and update nr_zones only once the queue is |
555 | * stopped and all I/Os are completed (i.e. a scheduler is not | |
556 | * referencing the bitmaps). | |
bf505456 DLM |
557 | */ |
558 | blk_mq_freeze_queue(q); | |
2afdeb23 | 559 | if (ret > 0) { |
6c6b3549 | 560 | blk_queue_chunk_sectors(q, args.zone_sectors); |
e94f5819 | 561 | q->nr_zones = args.nr_zones; |
d4100351 | 562 | swap(q->seq_zones_wlock, args.seq_zones_wlock); |
f216fdd7 | 563 | swap(q->conv_zones_bitmap, args.conv_zones_bitmap); |
e732671a DLM |
564 | if (update_driver_data) |
565 | update_driver_data(disk); | |
d4100351 CH |
566 | ret = 0; |
567 | } else { | |
bf505456 | 568 | pr_warn("%s: failed to revalidate zones\n", disk->disk_name); |
bf505456 | 569 | blk_queue_free_zone_bitmaps(q); |
bf505456 | 570 | } |
d4100351 | 571 | blk_mq_unfreeze_queue(q); |
bf505456 | 572 | |
d4100351 | 573 | kfree(args.seq_zones_wlock); |
f216fdd7 | 574 | kfree(args.conv_zones_bitmap); |
bf505456 DLM |
575 | return ret; |
576 | } | |
577 | EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones); | |
508aebb8 DLM |
578 | |
579 | void blk_queue_clear_zone_settings(struct request_queue *q) | |
580 | { | |
581 | blk_mq_freeze_queue(q); | |
582 | ||
583 | blk_queue_free_zone_bitmaps(q); | |
584 | blk_queue_flag_clear(QUEUE_FLAG_ZONE_RESETALL, q); | |
585 | q->required_elevator_features &= ~ELEVATOR_F_ZBD_SEQ_WRITE; | |
586 | q->nr_zones = 0; | |
587 | q->max_open_zones = 0; | |
588 | q->max_active_zones = 0; | |
589 | q->limits.chunk_sectors = 0; | |
590 | q->limits.zone_write_granularity = 0; | |
591 | q->limits.max_zone_append_sectors = 0; | |
592 | ||
593 | blk_mq_unfreeze_queue(q); | |
594 | } |