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 | { | |
d86e716a | 60 | if (blk_rq_is_passthrough(rq)) |
6cc77e9c CH |
61 | return false; |
62 | ||
d86e716a | 63 | if (!rq->q->disk->seq_zones_wlock) |
6cc77e9c CH |
64 | return false; |
65 | ||
8cafdb5a | 66 | if (bdev_op_is_zoned_write(rq->q->disk->part0, req_op(rq))) |
6cc77e9c | 67 | return blk_rq_zone_is_seq(rq); |
8cafdb5a PR |
68 | |
69 | return false; | |
6cc77e9c CH |
70 | } |
71 | EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock); | |
72 | ||
1392d370 JT |
73 | bool blk_req_zone_write_trylock(struct request *rq) |
74 | { | |
75 | unsigned int zno = blk_rq_zone_no(rq); | |
76 | ||
d86e716a | 77 | if (test_and_set_bit(zno, rq->q->disk->seq_zones_wlock)) |
1392d370 JT |
78 | return false; |
79 | ||
80 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
81 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
82 | ||
83 | return true; | |
84 | } | |
85 | EXPORT_SYMBOL_GPL(blk_req_zone_write_trylock); | |
86 | ||
6cc77e9c CH |
87 | void __blk_req_zone_write_lock(struct request *rq) |
88 | { | |
89 | if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq), | |
d86e716a | 90 | rq->q->disk->seq_zones_wlock))) |
6cc77e9c CH |
91 | return; |
92 | ||
93 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
94 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
95 | } | |
96 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock); | |
97 | ||
98 | void __blk_req_zone_write_unlock(struct request *rq) | |
99 | { | |
100 | rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED; | |
d86e716a | 101 | if (rq->q->disk->seq_zones_wlock) |
6cc77e9c | 102 | WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq), |
d86e716a | 103 | rq->q->disk->seq_zones_wlock)); |
6cc77e9c CH |
104 | } |
105 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock); | |
106 | ||
a91e1380 | 107 | /** |
b623e347 CH |
108 | * bdev_nr_zones - Get number of zones |
109 | * @bdev: Target device | |
a91e1380 | 110 | * |
9b38bb4b CH |
111 | * Return the total number of zones of a zoned block device. For a block |
112 | * device without zone capabilities, the number of zones is always 0. | |
a91e1380 | 113 | */ |
b623e347 | 114 | unsigned int bdev_nr_zones(struct block_device *bdev) |
a91e1380 | 115 | { |
b623e347 | 116 | sector_t zone_sectors = bdev_zone_sectors(bdev); |
a91e1380 | 117 | |
b623e347 | 118 | if (!bdev_is_zoned(bdev)) |
a91e1380 | 119 | return 0; |
b623e347 CH |
120 | return (bdev_nr_sectors(bdev) + zone_sectors - 1) >> |
121 | ilog2(zone_sectors); | |
a91e1380 | 122 | } |
b623e347 | 123 | EXPORT_SYMBOL_GPL(bdev_nr_zones); |
a91e1380 | 124 | |
6a0cb1bc HR |
125 | /** |
126 | * blkdev_report_zones - Get zones information | |
127 | * @bdev: Target block device | |
128 | * @sector: Sector from which to report zones | |
d4100351 CH |
129 | * @nr_zones: Maximum number of zones to report |
130 | * @cb: Callback function called for each reported zone | |
131 | * @data: Private data for the callback | |
6a0cb1bc HR |
132 | * |
133 | * Description: | |
d4100351 CH |
134 | * Get zone information starting from the zone containing @sector for at most |
135 | * @nr_zones, and call @cb for each zone reported by the device. | |
136 | * To report all zones in a device starting from @sector, the BLK_ALL_ZONES | |
137 | * constant can be passed to @nr_zones. | |
138 | * Returns the number of zones reported by the device, or a negative errno | |
139 | * value in case of failure. | |
140 | * | |
141 | * Note: The caller must use memalloc_noXX_save/restore() calls to control | |
142 | * memory allocations done within this function. | |
6a0cb1bc | 143 | */ |
e76239a3 | 144 | int blkdev_report_zones(struct block_device *bdev, sector_t sector, |
d4100351 | 145 | unsigned int nr_zones, report_zones_cb cb, void *data) |
6a0cb1bc | 146 | { |
ceeb373a | 147 | struct gendisk *disk = bdev->bd_disk; |
5eac3eb3 | 148 | sector_t capacity = get_capacity(disk); |
6a0cb1bc | 149 | |
edd1dbc8 | 150 | if (!bdev_is_zoned(bdev) || WARN_ON_ONCE(!disk->fops->report_zones)) |
e76239a3 | 151 | return -EOPNOTSUPP; |
6a0cb1bc | 152 | |
d4100351 | 153 | if (!nr_zones || sector >= capacity) |
6a0cb1bc | 154 | return 0; |
6a0cb1bc | 155 | |
d4100351 | 156 | return disk->fops->report_zones(disk, sector, nr_zones, cb, data); |
6a0cb1bc HR |
157 | } |
158 | EXPORT_SYMBOL_GPL(blkdev_report_zones); | |
159 | ||
1ee533ec DLM |
160 | static inline unsigned long *blk_alloc_zone_bitmap(int node, |
161 | unsigned int nr_zones) | |
6e33dbf2 | 162 | { |
1ee533ec DLM |
163 | return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long), |
164 | GFP_NOIO, node); | |
165 | } | |
6e33dbf2 | 166 | |
1ee533ec DLM |
167 | static int blk_zone_need_reset_cb(struct blk_zone *zone, unsigned int idx, |
168 | void *data) | |
169 | { | |
6e33dbf2 | 170 | /* |
1ee533ec DLM |
171 | * For an all-zones reset, ignore conventional, empty, read-only |
172 | * and offline zones. | |
6e33dbf2 | 173 | */ |
1ee533ec DLM |
174 | switch (zone->cond) { |
175 | case BLK_ZONE_COND_NOT_WP: | |
176 | case BLK_ZONE_COND_EMPTY: | |
177 | case BLK_ZONE_COND_READONLY: | |
178 | case BLK_ZONE_COND_OFFLINE: | |
179 | return 0; | |
180 | default: | |
181 | set_bit(idx, (unsigned long *)data); | |
182 | return 0; | |
183 | } | |
184 | } | |
185 | ||
186 | static int blkdev_zone_reset_all_emulated(struct block_device *bdev, | |
187 | gfp_t gfp_mask) | |
188 | { | |
d86e716a | 189 | struct gendisk *disk = bdev->bd_disk; |
375c140c CH |
190 | sector_t capacity = bdev_nr_sectors(bdev); |
191 | sector_t zone_sectors = bdev_zone_sectors(bdev); | |
1ee533ec DLM |
192 | unsigned long *need_reset; |
193 | struct bio *bio = NULL; | |
194 | sector_t sector = 0; | |
195 | int ret; | |
196 | ||
d86e716a | 197 | need_reset = blk_alloc_zone_bitmap(disk->queue->node, disk->nr_zones); |
1ee533ec DLM |
198 | if (!need_reset) |
199 | return -ENOMEM; | |
200 | ||
d86e716a CH |
201 | ret = disk->fops->report_zones(disk, 0, disk->nr_zones, |
202 | blk_zone_need_reset_cb, need_reset); | |
1ee533ec DLM |
203 | if (ret < 0) |
204 | goto out_free_need_reset; | |
205 | ||
206 | ret = 0; | |
207 | while (sector < capacity) { | |
d86e716a | 208 | if (!test_bit(disk_zone_no(disk, sector), need_reset)) { |
1ee533ec DLM |
209 | sector += zone_sectors; |
210 | continue; | |
211 | } | |
212 | ||
0a3140ea CK |
213 | bio = blk_next_bio(bio, bdev, 0, REQ_OP_ZONE_RESET | REQ_SYNC, |
214 | gfp_mask); | |
1ee533ec DLM |
215 | bio->bi_iter.bi_sector = sector; |
216 | sector += zone_sectors; | |
217 | ||
218 | /* This may take a while, so be nice to others */ | |
219 | cond_resched(); | |
220 | } | |
221 | ||
222 | if (bio) { | |
223 | ret = submit_bio_wait(bio); | |
224 | bio_put(bio); | |
225 | } | |
226 | ||
227 | out_free_need_reset: | |
228 | kfree(need_reset); | |
229 | return ret; | |
230 | } | |
231 | ||
232 | static int blkdev_zone_reset_all(struct block_device *bdev, gfp_t gfp_mask) | |
233 | { | |
234 | struct bio bio; | |
235 | ||
49add496 | 236 | bio_init(&bio, bdev, NULL, 0, REQ_OP_ZONE_RESET_ALL | REQ_SYNC); |
1ee533ec | 237 | return submit_bio_wait(&bio); |
6e33dbf2 CK |
238 | } |
239 | ||
6a0cb1bc | 240 | /** |
6c1b1da5 | 241 | * blkdev_zone_mgmt - Execute a zone management operation on a range of zones |
6a0cb1bc | 242 | * @bdev: Target block device |
6c1b1da5 AJ |
243 | * @op: Operation to be performed on the zones |
244 | * @sector: Start sector of the first zone to operate on | |
245 | * @nr_sectors: Number of sectors, should be at least the length of one zone and | |
246 | * must be zone size aligned. | |
6a0cb1bc HR |
247 | * @gfp_mask: Memory allocation flags (for bio_alloc) |
248 | * | |
249 | * Description: | |
6c1b1da5 | 250 | * Perform the specified operation on the range of zones specified by |
6a0cb1bc HR |
251 | * @sector..@sector+@nr_sectors. Specifying the entire disk sector range |
252 | * is valid, but the specified range should not contain conventional zones. | |
6c1b1da5 AJ |
253 | * The operation to execute on each zone can be a zone reset, open, close |
254 | * or finish request. | |
6a0cb1bc | 255 | */ |
ff07a02e BVA |
256 | int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, |
257 | sector_t sector, sector_t nr_sectors, gfp_t gfp_mask) | |
6a0cb1bc HR |
258 | { |
259 | struct request_queue *q = bdev_get_queue(bdev); | |
375c140c CH |
260 | sector_t zone_sectors = bdev_zone_sectors(bdev); |
261 | sector_t capacity = bdev_nr_sectors(bdev); | |
6a0cb1bc | 262 | sector_t end_sector = sector + nr_sectors; |
a2d6b3a2 | 263 | struct bio *bio = NULL; |
1ee533ec | 264 | int ret = 0; |
6a0cb1bc | 265 | |
edd1dbc8 | 266 | if (!bdev_is_zoned(bdev)) |
6a0cb1bc HR |
267 | return -EOPNOTSUPP; |
268 | ||
a2d6b3a2 DLM |
269 | if (bdev_read_only(bdev)) |
270 | return -EPERM; | |
271 | ||
6c1b1da5 AJ |
272 | if (!op_is_zone_mgmt(op)) |
273 | return -EOPNOTSUPP; | |
274 | ||
11bde986 | 275 | if (end_sector <= sector || end_sector > capacity) |
6a0cb1bc HR |
276 | /* Out of range */ |
277 | return -EINVAL; | |
278 | ||
279 | /* Check alignment (handle eventual smaller last zone) */ | |
e29b2100 | 280 | if (!bdev_is_zone_start(bdev, sector)) |
6a0cb1bc HR |
281 | return -EINVAL; |
282 | ||
e29b2100 | 283 | if (!bdev_is_zone_start(bdev, nr_sectors) && end_sector != capacity) |
6a0cb1bc HR |
284 | return -EINVAL; |
285 | ||
1ee533ec DLM |
286 | /* |
287 | * In the case of a zone reset operation over all zones, | |
288 | * REQ_OP_ZONE_RESET_ALL can be used with devices supporting this | |
289 | * command. For other devices, we emulate this command behavior by | |
290 | * identifying the zones needing a reset. | |
291 | */ | |
292 | if (op == REQ_OP_ZONE_RESET && sector == 0 && nr_sectors == capacity) { | |
293 | if (!blk_queue_zone_resetall(q)) | |
294 | return blkdev_zone_reset_all_emulated(bdev, gfp_mask); | |
295 | return blkdev_zone_reset_all(bdev, gfp_mask); | |
296 | } | |
297 | ||
6a0cb1bc | 298 | while (sector < end_sector) { |
0a3140ea | 299 | bio = blk_next_bio(bio, bdev, 0, op | REQ_SYNC, gfp_mask); |
c7a1d926 | 300 | bio->bi_iter.bi_sector = sector; |
6a0cb1bc HR |
301 | sector += zone_sectors; |
302 | ||
303 | /* This may take a while, so be nice to others */ | |
304 | cond_resched(); | |
6a0cb1bc HR |
305 | } |
306 | ||
a2d6b3a2 DLM |
307 | ret = submit_bio_wait(bio); |
308 | bio_put(bio); | |
309 | ||
a2d6b3a2 | 310 | return ret; |
6a0cb1bc | 311 | } |
6c1b1da5 | 312 | EXPORT_SYMBOL_GPL(blkdev_zone_mgmt); |
3ed05a98 | 313 | |
d4100351 CH |
314 | struct zone_report_args { |
315 | struct blk_zone __user *zones; | |
316 | }; | |
317 | ||
318 | static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx, | |
319 | void *data) | |
320 | { | |
321 | struct zone_report_args *args = data; | |
322 | ||
323 | if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone))) | |
324 | return -EFAULT; | |
325 | return 0; | |
326 | } | |
327 | ||
56c4bddb | 328 | /* |
3ed05a98 ST |
329 | * BLKREPORTZONE ioctl processing. |
330 | * Called from blkdev_ioctl. | |
331 | */ | |
332 | int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
333 | unsigned int cmd, unsigned long arg) | |
334 | { | |
335 | void __user *argp = (void __user *)arg; | |
d4100351 | 336 | struct zone_report_args args; |
3ed05a98 | 337 | struct blk_zone_report rep; |
3ed05a98 ST |
338 | int ret; |
339 | ||
340 | if (!argp) | |
341 | return -EINVAL; | |
342 | ||
edd1dbc8 | 343 | if (!bdev_is_zoned(bdev)) |
3ed05a98 ST |
344 | return -ENOTTY; |
345 | ||
3ed05a98 ST |
346 | if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report))) |
347 | return -EFAULT; | |
348 | ||
349 | if (!rep.nr_zones) | |
350 | return -EINVAL; | |
351 | ||
d4100351 CH |
352 | args.zones = argp + sizeof(struct blk_zone_report); |
353 | ret = blkdev_report_zones(bdev, rep.sector, rep.nr_zones, | |
354 | blkdev_copy_zone_to_user, &args); | |
355 | if (ret < 0) | |
356 | return ret; | |
3ed05a98 | 357 | |
d4100351 | 358 | rep.nr_zones = ret; |
82394db7 | 359 | rep.flags = BLK_ZONE_REP_CAPACITY; |
d4100351 CH |
360 | if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report))) |
361 | return -EFAULT; | |
362 | return 0; | |
3ed05a98 ST |
363 | } |
364 | ||
e5113505 SK |
365 | static int blkdev_truncate_zone_range(struct block_device *bdev, fmode_t mode, |
366 | const struct blk_zone_range *zrange) | |
367 | { | |
368 | loff_t start, end; | |
369 | ||
370 | if (zrange->sector + zrange->nr_sectors <= zrange->sector || | |
371 | zrange->sector + zrange->nr_sectors > get_capacity(bdev->bd_disk)) | |
372 | /* Out of range */ | |
373 | return -EINVAL; | |
374 | ||
375 | start = zrange->sector << SECTOR_SHIFT; | |
376 | end = ((zrange->sector + zrange->nr_sectors) << SECTOR_SHIFT) - 1; | |
377 | ||
378 | return truncate_bdev_range(bdev, mode, start, end); | |
379 | } | |
380 | ||
56c4bddb | 381 | /* |
e876df1f | 382 | * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing. |
3ed05a98 ST |
383 | * Called from blkdev_ioctl. |
384 | */ | |
e876df1f AJ |
385 | int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode, |
386 | unsigned int cmd, unsigned long arg) | |
3ed05a98 ST |
387 | { |
388 | void __user *argp = (void __user *)arg; | |
3ed05a98 | 389 | struct blk_zone_range zrange; |
ff07a02e | 390 | enum req_op op; |
e5113505 | 391 | int ret; |
3ed05a98 ST |
392 | |
393 | if (!argp) | |
394 | return -EINVAL; | |
395 | ||
edd1dbc8 | 396 | if (!bdev_is_zoned(bdev)) |
3ed05a98 ST |
397 | return -ENOTTY; |
398 | ||
3ed05a98 ST |
399 | if (!(mode & FMODE_WRITE)) |
400 | return -EBADF; | |
401 | ||
402 | if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range))) | |
403 | return -EFAULT; | |
404 | ||
e876df1f AJ |
405 | switch (cmd) { |
406 | case BLKRESETZONE: | |
407 | op = REQ_OP_ZONE_RESET; | |
e5113505 SK |
408 | |
409 | /* Invalidate the page cache, including dirty pages. */ | |
86399ea0 | 410 | filemap_invalidate_lock(bdev->bd_inode->i_mapping); |
e5113505 SK |
411 | ret = blkdev_truncate_zone_range(bdev, mode, &zrange); |
412 | if (ret) | |
86399ea0 | 413 | goto fail; |
e876df1f AJ |
414 | break; |
415 | case BLKOPENZONE: | |
416 | op = REQ_OP_ZONE_OPEN; | |
417 | break; | |
418 | case BLKCLOSEZONE: | |
419 | op = REQ_OP_ZONE_CLOSE; | |
420 | break; | |
421 | case BLKFINISHZONE: | |
422 | op = REQ_OP_ZONE_FINISH; | |
423 | break; | |
424 | default: | |
425 | return -ENOTTY; | |
426 | } | |
427 | ||
e5113505 SK |
428 | ret = blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors, |
429 | GFP_KERNEL); | |
430 | ||
86399ea0 SK |
431 | fail: |
432 | if (cmd == BLKRESETZONE) | |
433 | filemap_invalidate_unlock(bdev->bd_inode->i_mapping); | |
e5113505 SK |
434 | |
435 | return ret; | |
3ed05a98 | 436 | } |
bf505456 | 437 | |
5d400665 | 438 | void disk_free_zone_bitmaps(struct gendisk *disk) |
bf505456 | 439 | { |
d86e716a CH |
440 | kfree(disk->conv_zones_bitmap); |
441 | disk->conv_zones_bitmap = NULL; | |
442 | kfree(disk->seq_zones_wlock); | |
443 | disk->seq_zones_wlock = NULL; | |
bf505456 DLM |
444 | } |
445 | ||
d4100351 CH |
446 | struct blk_revalidate_zone_args { |
447 | struct gendisk *disk; | |
f216fdd7 | 448 | unsigned long *conv_zones_bitmap; |
d4100351 | 449 | unsigned long *seq_zones_wlock; |
e94f5819 | 450 | unsigned int nr_zones; |
6c6b3549 | 451 | sector_t zone_sectors; |
d4100351 CH |
452 | sector_t sector; |
453 | }; | |
454 | ||
d9dd7308 DLM |
455 | /* |
456 | * Helper function to check the validity of zones of a zoned block device. | |
457 | */ | |
d4100351 CH |
458 | static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx, |
459 | void *data) | |
d9dd7308 | 460 | { |
d4100351 CH |
461 | struct blk_revalidate_zone_args *args = data; |
462 | struct gendisk *disk = args->disk; | |
d9dd7308 | 463 | struct request_queue *q = disk->queue; |
d9dd7308 DLM |
464 | sector_t capacity = get_capacity(disk); |
465 | ||
466 | /* | |
467 | * All zones must have the same size, with the exception on an eventual | |
468 | * smaller last zone. | |
469 | */ | |
6c6b3549 CH |
470 | if (zone->start == 0) { |
471 | if (zone->len == 0 || !is_power_of_2(zone->len)) { | |
472 | pr_warn("%s: Invalid zoned device with non power of two zone size (%llu)\n", | |
473 | disk->disk_name, zone->len); | |
474 | return -ENODEV; | |
475 | } | |
d9dd7308 | 476 | |
6c6b3549 CH |
477 | args->zone_sectors = zone->len; |
478 | args->nr_zones = (capacity + zone->len - 1) >> ilog2(zone->len); | |
479 | } else if (zone->start + args->zone_sectors < capacity) { | |
480 | if (zone->len != args->zone_sectors) { | |
481 | pr_warn("%s: Invalid zoned device with non constant zone size\n", | |
482 | disk->disk_name); | |
483 | return -ENODEV; | |
484 | } | |
485 | } else { | |
486 | if (zone->len > args->zone_sectors) { | |
487 | pr_warn("%s: Invalid zoned device with larger last zone size\n", | |
488 | disk->disk_name); | |
489 | return -ENODEV; | |
490 | } | |
d9dd7308 DLM |
491 | } |
492 | ||
493 | /* Check for holes in the zone report */ | |
d4100351 | 494 | if (zone->start != args->sector) { |
d9dd7308 | 495 | pr_warn("%s: Zone gap at sectors %llu..%llu\n", |
d4100351 CH |
496 | disk->disk_name, args->sector, zone->start); |
497 | return -ENODEV; | |
d9dd7308 DLM |
498 | } |
499 | ||
500 | /* Check zone type */ | |
501 | switch (zone->type) { | |
502 | case BLK_ZONE_TYPE_CONVENTIONAL: | |
e94f5819 CH |
503 | if (!args->conv_zones_bitmap) { |
504 | args->conv_zones_bitmap = | |
505 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
506 | if (!args->conv_zones_bitmap) | |
507 | return -ENOMEM; | |
508 | } | |
509 | set_bit(idx, args->conv_zones_bitmap); | |
510 | break; | |
d9dd7308 DLM |
511 | case BLK_ZONE_TYPE_SEQWRITE_REQ: |
512 | case BLK_ZONE_TYPE_SEQWRITE_PREF: | |
e94f5819 CH |
513 | if (!args->seq_zones_wlock) { |
514 | args->seq_zones_wlock = | |
515 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
516 | if (!args->seq_zones_wlock) | |
517 | return -ENOMEM; | |
518 | } | |
d9dd7308 DLM |
519 | break; |
520 | default: | |
521 | pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n", | |
522 | disk->disk_name, (int)zone->type, zone->start); | |
d4100351 | 523 | return -ENODEV; |
d9dd7308 DLM |
524 | } |
525 | ||
d4100351 CH |
526 | args->sector += zone->len; |
527 | return 0; | |
528 | } | |
529 | ||
bf505456 DLM |
530 | /** |
531 | * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps | |
532 | * @disk: Target disk | |
e732671a | 533 | * @update_driver_data: Callback to update driver data on the frozen disk |
bf505456 DLM |
534 | * |
535 | * Helper function for low-level device drivers to (re) allocate and initialize | |
536 | * a disk request queue zone bitmaps. This functions should normally be called | |
ae58954d CH |
537 | * within the disk ->revalidate method for blk-mq based drivers. For BIO based |
538 | * drivers only q->nr_zones needs to be updated so that the sysfs exposed value | |
539 | * is correct. | |
e732671a DLM |
540 | * If the @update_driver_data callback function is not NULL, the callback is |
541 | * executed with the device request queue frozen after all zones have been | |
542 | * checked. | |
bf505456 | 543 | */ |
e732671a DLM |
544 | int blk_revalidate_disk_zones(struct gendisk *disk, |
545 | void (*update_driver_data)(struct gendisk *disk)) | |
bf505456 DLM |
546 | { |
547 | struct request_queue *q = disk->queue; | |
e94f5819 CH |
548 | struct blk_revalidate_zone_args args = { |
549 | .disk = disk, | |
e94f5819 | 550 | }; |
6c6b3549 CH |
551 | unsigned int noio_flag; |
552 | int ret; | |
bf505456 | 553 | |
c98c3d09 CH |
554 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) |
555 | return -EIO; | |
ae58954d CH |
556 | if (WARN_ON_ONCE(!queue_is_mq(q))) |
557 | return -EIO; | |
bf505456 | 558 | |
1a1206dc JT |
559 | if (!get_capacity(disk)) |
560 | return -EIO; | |
561 | ||
e94f5819 | 562 | /* |
6c6b3549 CH |
563 | * Ensure that all memory allocations in this context are done as if |
564 | * GFP_NOIO was specified. | |
e94f5819 | 565 | */ |
6c6b3549 CH |
566 | noio_flag = memalloc_noio_save(); |
567 | ret = disk->fops->report_zones(disk, 0, UINT_MAX, | |
568 | blk_revalidate_zone_cb, &args); | |
2afdeb23 DLM |
569 | if (!ret) { |
570 | pr_warn("%s: No zones reported\n", disk->disk_name); | |
571 | ret = -ENODEV; | |
572 | } | |
6c6b3549 | 573 | memalloc_noio_restore(noio_flag); |
bf505456 | 574 | |
2afdeb23 DLM |
575 | /* |
576 | * If zones where reported, make sure that the entire disk capacity | |
577 | * has been checked. | |
578 | */ | |
579 | if (ret > 0 && args.sector != get_capacity(disk)) { | |
580 | pr_warn("%s: Missing zones from sector %llu\n", | |
581 | disk->disk_name, args.sector); | |
582 | ret = -ENODEV; | |
583 | } | |
584 | ||
bf505456 | 585 | /* |
6c6b3549 CH |
586 | * Install the new bitmaps and update nr_zones only once the queue is |
587 | * stopped and all I/Os are completed (i.e. a scheduler is not | |
588 | * referencing the bitmaps). | |
bf505456 DLM |
589 | */ |
590 | blk_mq_freeze_queue(q); | |
2afdeb23 | 591 | if (ret > 0) { |
6c6b3549 | 592 | blk_queue_chunk_sectors(q, args.zone_sectors); |
d86e716a CH |
593 | disk->nr_zones = args.nr_zones; |
594 | swap(disk->seq_zones_wlock, args.seq_zones_wlock); | |
595 | swap(disk->conv_zones_bitmap, args.conv_zones_bitmap); | |
e732671a DLM |
596 | if (update_driver_data) |
597 | update_driver_data(disk); | |
d4100351 CH |
598 | ret = 0; |
599 | } else { | |
bf505456 | 600 | pr_warn("%s: failed to revalidate zones\n", disk->disk_name); |
5d400665 | 601 | disk_free_zone_bitmaps(disk); |
bf505456 | 602 | } |
d4100351 | 603 | blk_mq_unfreeze_queue(q); |
bf505456 | 604 | |
d4100351 | 605 | kfree(args.seq_zones_wlock); |
f216fdd7 | 606 | kfree(args.conv_zones_bitmap); |
bf505456 DLM |
607 | return ret; |
608 | } | |
609 | EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones); | |
508aebb8 | 610 | |
b3c72f81 | 611 | void disk_clear_zone_settings(struct gendisk *disk) |
508aebb8 | 612 | { |
b3c72f81 CH |
613 | struct request_queue *q = disk->queue; |
614 | ||
508aebb8 DLM |
615 | blk_mq_freeze_queue(q); |
616 | ||
5d400665 | 617 | disk_free_zone_bitmaps(disk); |
508aebb8 DLM |
618 | blk_queue_flag_clear(QUEUE_FLAG_ZONE_RESETALL, q); |
619 | q->required_elevator_features &= ~ELEVATOR_F_ZBD_SEQ_WRITE; | |
d86e716a CH |
620 | disk->nr_zones = 0; |
621 | disk->max_open_zones = 0; | |
622 | disk->max_active_zones = 0; | |
508aebb8 DLM |
623 | q->limits.chunk_sectors = 0; |
624 | q->limits.zone_write_granularity = 0; | |
625 | q->limits.max_zone_append_sectors = 0; | |
626 | ||
627 | blk_mq_unfreeze_queue(q); | |
628 | } |