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: | |
6cc77e9c CH |
68 | case REQ_OP_WRITE: |
69 | return blk_rq_zone_is_seq(rq); | |
70 | default: | |
71 | return false; | |
72 | } | |
73 | } | |
74 | EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock); | |
75 | ||
1392d370 JT |
76 | bool blk_req_zone_write_trylock(struct request *rq) |
77 | { | |
78 | unsigned int zno = blk_rq_zone_no(rq); | |
79 | ||
80 | if (test_and_set_bit(zno, rq->q->seq_zones_wlock)) | |
81 | return false; | |
82 | ||
83 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
84 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
85 | ||
86 | return true; | |
87 | } | |
88 | EXPORT_SYMBOL_GPL(blk_req_zone_write_trylock); | |
89 | ||
6cc77e9c CH |
90 | void __blk_req_zone_write_lock(struct request *rq) |
91 | { | |
92 | if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq), | |
93 | rq->q->seq_zones_wlock))) | |
94 | return; | |
95 | ||
96 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
97 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
98 | } | |
99 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock); | |
100 | ||
101 | void __blk_req_zone_write_unlock(struct request *rq) | |
102 | { | |
103 | rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED; | |
104 | if (rq->q->seq_zones_wlock) | |
105 | WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq), | |
106 | rq->q->seq_zones_wlock)); | |
107 | } | |
108 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock); | |
109 | ||
a91e1380 DLM |
110 | /** |
111 | * blkdev_nr_zones - Get number of zones | |
9b38bb4b | 112 | * @disk: Target gendisk |
a91e1380 | 113 | * |
9b38bb4b CH |
114 | * Return the total number of zones of a zoned block device. For a block |
115 | * device without zone capabilities, the number of zones is always 0. | |
a91e1380 | 116 | */ |
9b38bb4b | 117 | unsigned int blkdev_nr_zones(struct gendisk *disk) |
a91e1380 | 118 | { |
9b38bb4b | 119 | sector_t zone_sectors = blk_queue_zone_sectors(disk->queue); |
a91e1380 | 120 | |
9b38bb4b | 121 | if (!blk_queue_is_zoned(disk->queue)) |
a91e1380 | 122 | return 0; |
9b38bb4b | 123 | return (get_capacity(disk) + zone_sectors - 1) >> ilog2(zone_sectors); |
a91e1380 DLM |
124 | } |
125 | EXPORT_SYMBOL_GPL(blkdev_nr_zones); | |
126 | ||
6a0cb1bc HR |
127 | /** |
128 | * blkdev_report_zones - Get zones information | |
129 | * @bdev: Target block device | |
130 | * @sector: Sector from which to report zones | |
d4100351 CH |
131 | * @nr_zones: Maximum number of zones to report |
132 | * @cb: Callback function called for each reported zone | |
133 | * @data: Private data for the callback | |
6a0cb1bc HR |
134 | * |
135 | * Description: | |
d4100351 CH |
136 | * Get zone information starting from the zone containing @sector for at most |
137 | * @nr_zones, and call @cb for each zone reported by the device. | |
138 | * To report all zones in a device starting from @sector, the BLK_ALL_ZONES | |
139 | * constant can be passed to @nr_zones. | |
140 | * Returns the number of zones reported by the device, or a negative errno | |
141 | * value in case of failure. | |
142 | * | |
143 | * Note: The caller must use memalloc_noXX_save/restore() calls to control | |
144 | * memory allocations done within this function. | |
6a0cb1bc | 145 | */ |
e76239a3 | 146 | int blkdev_report_zones(struct block_device *bdev, sector_t sector, |
d4100351 | 147 | unsigned int nr_zones, report_zones_cb cb, void *data) |
6a0cb1bc | 148 | { |
ceeb373a | 149 | struct gendisk *disk = bdev->bd_disk; |
5eac3eb3 | 150 | sector_t capacity = get_capacity(disk); |
6a0cb1bc | 151 | |
d4100351 CH |
152 | if (!blk_queue_is_zoned(bdev_get_queue(bdev)) || |
153 | WARN_ON_ONCE(!disk->fops->report_zones)) | |
e76239a3 | 154 | return -EOPNOTSUPP; |
6a0cb1bc | 155 | |
d4100351 | 156 | if (!nr_zones || sector >= capacity) |
6a0cb1bc | 157 | return 0; |
6a0cb1bc | 158 | |
d4100351 | 159 | return disk->fops->report_zones(disk, sector, nr_zones, cb, data); |
6a0cb1bc HR |
160 | } |
161 | EXPORT_SYMBOL_GPL(blkdev_report_zones); | |
162 | ||
1ee533ec DLM |
163 | static inline unsigned long *blk_alloc_zone_bitmap(int node, |
164 | unsigned int nr_zones) | |
6e33dbf2 | 165 | { |
1ee533ec DLM |
166 | return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long), |
167 | GFP_NOIO, node); | |
168 | } | |
6e33dbf2 | 169 | |
1ee533ec DLM |
170 | static int blk_zone_need_reset_cb(struct blk_zone *zone, unsigned int idx, |
171 | void *data) | |
172 | { | |
6e33dbf2 | 173 | /* |
1ee533ec DLM |
174 | * For an all-zones reset, ignore conventional, empty, read-only |
175 | * and offline zones. | |
6e33dbf2 | 176 | */ |
1ee533ec DLM |
177 | switch (zone->cond) { |
178 | case BLK_ZONE_COND_NOT_WP: | |
179 | case BLK_ZONE_COND_EMPTY: | |
180 | case BLK_ZONE_COND_READONLY: | |
181 | case BLK_ZONE_COND_OFFLINE: | |
182 | return 0; | |
183 | default: | |
184 | set_bit(idx, (unsigned long *)data); | |
185 | return 0; | |
186 | } | |
187 | } | |
188 | ||
189 | static int blkdev_zone_reset_all_emulated(struct block_device *bdev, | |
190 | gfp_t gfp_mask) | |
191 | { | |
192 | struct request_queue *q = bdev_get_queue(bdev); | |
193 | sector_t capacity = get_capacity(bdev->bd_disk); | |
194 | sector_t zone_sectors = blk_queue_zone_sectors(q); | |
195 | unsigned long *need_reset; | |
196 | struct bio *bio = NULL; | |
197 | sector_t sector = 0; | |
198 | int ret; | |
199 | ||
200 | need_reset = blk_alloc_zone_bitmap(q->node, q->nr_zones); | |
201 | if (!need_reset) | |
202 | return -ENOMEM; | |
203 | ||
204 | ret = bdev->bd_disk->fops->report_zones(bdev->bd_disk, 0, | |
205 | q->nr_zones, blk_zone_need_reset_cb, | |
206 | need_reset); | |
207 | if (ret < 0) | |
208 | goto out_free_need_reset; | |
209 | ||
210 | ret = 0; | |
211 | while (sector < capacity) { | |
212 | if (!test_bit(blk_queue_zone_no(q, sector), need_reset)) { | |
213 | sector += zone_sectors; | |
214 | continue; | |
215 | } | |
216 | ||
0a3140ea CK |
217 | bio = blk_next_bio(bio, bdev, 0, REQ_OP_ZONE_RESET | REQ_SYNC, |
218 | gfp_mask); | |
1ee533ec DLM |
219 | bio->bi_iter.bi_sector = sector; |
220 | sector += zone_sectors; | |
221 | ||
222 | /* This may take a while, so be nice to others */ | |
223 | cond_resched(); | |
224 | } | |
225 | ||
226 | if (bio) { | |
227 | ret = submit_bio_wait(bio); | |
228 | bio_put(bio); | |
229 | } | |
230 | ||
231 | out_free_need_reset: | |
232 | kfree(need_reset); | |
233 | return ret; | |
234 | } | |
235 | ||
236 | static int blkdev_zone_reset_all(struct block_device *bdev, gfp_t gfp_mask) | |
237 | { | |
238 | struct bio bio; | |
239 | ||
49add496 | 240 | bio_init(&bio, bdev, NULL, 0, REQ_OP_ZONE_RESET_ALL | REQ_SYNC); |
1ee533ec | 241 | return submit_bio_wait(&bio); |
6e33dbf2 CK |
242 | } |
243 | ||
6a0cb1bc | 244 | /** |
6c1b1da5 | 245 | * blkdev_zone_mgmt - Execute a zone management operation on a range of zones |
6a0cb1bc | 246 | * @bdev: Target block device |
6c1b1da5 AJ |
247 | * @op: Operation to be performed on the zones |
248 | * @sector: Start sector of the first zone to operate on | |
249 | * @nr_sectors: Number of sectors, should be at least the length of one zone and | |
250 | * must be zone size aligned. | |
6a0cb1bc HR |
251 | * @gfp_mask: Memory allocation flags (for bio_alloc) |
252 | * | |
253 | * Description: | |
6c1b1da5 | 254 | * Perform the specified operation on the range of zones specified by |
6a0cb1bc HR |
255 | * @sector..@sector+@nr_sectors. Specifying the entire disk sector range |
256 | * is valid, but the specified range should not contain conventional zones. | |
6c1b1da5 AJ |
257 | * The operation to execute on each zone can be a zone reset, open, close |
258 | * or finish request. | |
6a0cb1bc | 259 | */ |
6c1b1da5 AJ |
260 | int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op, |
261 | sector_t sector, sector_t nr_sectors, | |
262 | gfp_t gfp_mask) | |
6a0cb1bc HR |
263 | { |
264 | struct request_queue *q = bdev_get_queue(bdev); | |
6c1b1da5 | 265 | sector_t zone_sectors = blk_queue_zone_sectors(q); |
5eac3eb3 | 266 | sector_t capacity = get_capacity(bdev->bd_disk); |
6a0cb1bc | 267 | sector_t end_sector = sector + nr_sectors; |
a2d6b3a2 | 268 | struct bio *bio = NULL; |
1ee533ec | 269 | int ret = 0; |
6a0cb1bc | 270 | |
6a0cb1bc HR |
271 | if (!blk_queue_is_zoned(q)) |
272 | return -EOPNOTSUPP; | |
273 | ||
a2d6b3a2 DLM |
274 | if (bdev_read_only(bdev)) |
275 | return -EPERM; | |
276 | ||
6c1b1da5 AJ |
277 | if (!op_is_zone_mgmt(op)) |
278 | return -EOPNOTSUPP; | |
279 | ||
11bde986 | 280 | if (end_sector <= sector || end_sector > capacity) |
6a0cb1bc HR |
281 | /* Out of range */ |
282 | return -EINVAL; | |
283 | ||
284 | /* Check alignment (handle eventual smaller last zone) */ | |
6a0cb1bc HR |
285 | if (sector & (zone_sectors - 1)) |
286 | return -EINVAL; | |
287 | ||
5eac3eb3 | 288 | if ((nr_sectors & (zone_sectors - 1)) && end_sector != capacity) |
6a0cb1bc HR |
289 | return -EINVAL; |
290 | ||
1ee533ec DLM |
291 | /* |
292 | * In the case of a zone reset operation over all zones, | |
293 | * REQ_OP_ZONE_RESET_ALL can be used with devices supporting this | |
294 | * command. For other devices, we emulate this command behavior by | |
295 | * identifying the zones needing a reset. | |
296 | */ | |
297 | if (op == REQ_OP_ZONE_RESET && sector == 0 && nr_sectors == capacity) { | |
298 | if (!blk_queue_zone_resetall(q)) | |
299 | return blkdev_zone_reset_all_emulated(bdev, gfp_mask); | |
300 | return blkdev_zone_reset_all(bdev, gfp_mask); | |
301 | } | |
302 | ||
6a0cb1bc | 303 | while (sector < end_sector) { |
0a3140ea | 304 | bio = blk_next_bio(bio, bdev, 0, op | REQ_SYNC, gfp_mask); |
c7a1d926 | 305 | bio->bi_iter.bi_sector = sector; |
6a0cb1bc HR |
306 | sector += zone_sectors; |
307 | ||
308 | /* This may take a while, so be nice to others */ | |
309 | cond_resched(); | |
6a0cb1bc HR |
310 | } |
311 | ||
a2d6b3a2 DLM |
312 | ret = submit_bio_wait(bio); |
313 | bio_put(bio); | |
314 | ||
a2d6b3a2 | 315 | return ret; |
6a0cb1bc | 316 | } |
6c1b1da5 | 317 | EXPORT_SYMBOL_GPL(blkdev_zone_mgmt); |
3ed05a98 | 318 | |
d4100351 CH |
319 | struct zone_report_args { |
320 | struct blk_zone __user *zones; | |
321 | }; | |
322 | ||
323 | static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx, | |
324 | void *data) | |
325 | { | |
326 | struct zone_report_args *args = data; | |
327 | ||
328 | if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone))) | |
329 | return -EFAULT; | |
330 | return 0; | |
331 | } | |
332 | ||
56c4bddb | 333 | /* |
3ed05a98 ST |
334 | * BLKREPORTZONE ioctl processing. |
335 | * Called from blkdev_ioctl. | |
336 | */ | |
337 | int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
338 | unsigned int cmd, unsigned long arg) | |
339 | { | |
340 | void __user *argp = (void __user *)arg; | |
d4100351 | 341 | struct zone_report_args args; |
3ed05a98 ST |
342 | struct request_queue *q; |
343 | struct blk_zone_report rep; | |
3ed05a98 ST |
344 | int ret; |
345 | ||
346 | if (!argp) | |
347 | return -EINVAL; | |
348 | ||
349 | q = bdev_get_queue(bdev); | |
350 | if (!q) | |
351 | return -ENXIO; | |
352 | ||
353 | if (!blk_queue_is_zoned(q)) | |
354 | return -ENOTTY; | |
355 | ||
3ed05a98 ST |
356 | if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report))) |
357 | return -EFAULT; | |
358 | ||
359 | if (!rep.nr_zones) | |
360 | return -EINVAL; | |
361 | ||
d4100351 CH |
362 | args.zones = argp + sizeof(struct blk_zone_report); |
363 | ret = blkdev_report_zones(bdev, rep.sector, rep.nr_zones, | |
364 | blkdev_copy_zone_to_user, &args); | |
365 | if (ret < 0) | |
366 | return ret; | |
3ed05a98 | 367 | |
d4100351 | 368 | rep.nr_zones = ret; |
82394db7 | 369 | rep.flags = BLK_ZONE_REP_CAPACITY; |
d4100351 CH |
370 | if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report))) |
371 | return -EFAULT; | |
372 | return 0; | |
3ed05a98 ST |
373 | } |
374 | ||
e5113505 SK |
375 | static int blkdev_truncate_zone_range(struct block_device *bdev, fmode_t mode, |
376 | const struct blk_zone_range *zrange) | |
377 | { | |
378 | loff_t start, end; | |
379 | ||
380 | if (zrange->sector + zrange->nr_sectors <= zrange->sector || | |
381 | zrange->sector + zrange->nr_sectors > get_capacity(bdev->bd_disk)) | |
382 | /* Out of range */ | |
383 | return -EINVAL; | |
384 | ||
385 | start = zrange->sector << SECTOR_SHIFT; | |
386 | end = ((zrange->sector + zrange->nr_sectors) << SECTOR_SHIFT) - 1; | |
387 | ||
388 | return truncate_bdev_range(bdev, mode, start, end); | |
389 | } | |
390 | ||
56c4bddb | 391 | /* |
e876df1f | 392 | * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing. |
3ed05a98 ST |
393 | * Called from blkdev_ioctl. |
394 | */ | |
e876df1f AJ |
395 | int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode, |
396 | unsigned int cmd, unsigned long arg) | |
3ed05a98 ST |
397 | { |
398 | void __user *argp = (void __user *)arg; | |
399 | struct request_queue *q; | |
400 | struct blk_zone_range zrange; | |
e876df1f | 401 | enum req_opf op; |
e5113505 | 402 | int ret; |
3ed05a98 ST |
403 | |
404 | if (!argp) | |
405 | return -EINVAL; | |
406 | ||
407 | q = bdev_get_queue(bdev); | |
408 | if (!q) | |
409 | return -ENXIO; | |
410 | ||
411 | if (!blk_queue_is_zoned(q)) | |
412 | return -ENOTTY; | |
413 | ||
3ed05a98 ST |
414 | if (!(mode & FMODE_WRITE)) |
415 | return -EBADF; | |
416 | ||
417 | if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range))) | |
418 | return -EFAULT; | |
419 | ||
e876df1f AJ |
420 | switch (cmd) { |
421 | case BLKRESETZONE: | |
422 | op = REQ_OP_ZONE_RESET; | |
e5113505 SK |
423 | |
424 | /* Invalidate the page cache, including dirty pages. */ | |
86399ea0 | 425 | filemap_invalidate_lock(bdev->bd_inode->i_mapping); |
e5113505 SK |
426 | ret = blkdev_truncate_zone_range(bdev, mode, &zrange); |
427 | if (ret) | |
86399ea0 | 428 | goto fail; |
e876df1f AJ |
429 | break; |
430 | case BLKOPENZONE: | |
431 | op = REQ_OP_ZONE_OPEN; | |
432 | break; | |
433 | case BLKCLOSEZONE: | |
434 | op = REQ_OP_ZONE_CLOSE; | |
435 | break; | |
436 | case BLKFINISHZONE: | |
437 | op = REQ_OP_ZONE_FINISH; | |
438 | break; | |
439 | default: | |
440 | return -ENOTTY; | |
441 | } | |
442 | ||
e5113505 SK |
443 | ret = blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors, |
444 | GFP_KERNEL); | |
445 | ||
86399ea0 SK |
446 | fail: |
447 | if (cmd == BLKRESETZONE) | |
448 | filemap_invalidate_unlock(bdev->bd_inode->i_mapping); | |
e5113505 SK |
449 | |
450 | return ret; | |
3ed05a98 | 451 | } |
bf505456 | 452 | |
bf505456 DLM |
453 | void blk_queue_free_zone_bitmaps(struct request_queue *q) |
454 | { | |
f216fdd7 CH |
455 | kfree(q->conv_zones_bitmap); |
456 | q->conv_zones_bitmap = NULL; | |
bf505456 DLM |
457 | kfree(q->seq_zones_wlock); |
458 | q->seq_zones_wlock = NULL; | |
459 | } | |
460 | ||
d4100351 CH |
461 | struct blk_revalidate_zone_args { |
462 | struct gendisk *disk; | |
f216fdd7 | 463 | unsigned long *conv_zones_bitmap; |
d4100351 | 464 | unsigned long *seq_zones_wlock; |
e94f5819 | 465 | unsigned int nr_zones; |
6c6b3549 | 466 | sector_t zone_sectors; |
d4100351 CH |
467 | sector_t sector; |
468 | }; | |
469 | ||
d9dd7308 DLM |
470 | /* |
471 | * Helper function to check the validity of zones of a zoned block device. | |
472 | */ | |
d4100351 CH |
473 | static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx, |
474 | void *data) | |
d9dd7308 | 475 | { |
d4100351 CH |
476 | struct blk_revalidate_zone_args *args = data; |
477 | struct gendisk *disk = args->disk; | |
d9dd7308 | 478 | struct request_queue *q = disk->queue; |
d9dd7308 DLM |
479 | sector_t capacity = get_capacity(disk); |
480 | ||
481 | /* | |
482 | * All zones must have the same size, with the exception on an eventual | |
483 | * smaller last zone. | |
484 | */ | |
6c6b3549 CH |
485 | if (zone->start == 0) { |
486 | if (zone->len == 0 || !is_power_of_2(zone->len)) { | |
487 | pr_warn("%s: Invalid zoned device with non power of two zone size (%llu)\n", | |
488 | disk->disk_name, zone->len); | |
489 | return -ENODEV; | |
490 | } | |
d9dd7308 | 491 | |
6c6b3549 CH |
492 | args->zone_sectors = zone->len; |
493 | args->nr_zones = (capacity + zone->len - 1) >> ilog2(zone->len); | |
494 | } else if (zone->start + args->zone_sectors < capacity) { | |
495 | if (zone->len != args->zone_sectors) { | |
496 | pr_warn("%s: Invalid zoned device with non constant zone size\n", | |
497 | disk->disk_name); | |
498 | return -ENODEV; | |
499 | } | |
500 | } else { | |
501 | if (zone->len > args->zone_sectors) { | |
502 | pr_warn("%s: Invalid zoned device with larger last zone size\n", | |
503 | disk->disk_name); | |
504 | return -ENODEV; | |
505 | } | |
d9dd7308 DLM |
506 | } |
507 | ||
508 | /* Check for holes in the zone report */ | |
d4100351 | 509 | if (zone->start != args->sector) { |
d9dd7308 | 510 | pr_warn("%s: Zone gap at sectors %llu..%llu\n", |
d4100351 CH |
511 | disk->disk_name, args->sector, zone->start); |
512 | return -ENODEV; | |
d9dd7308 DLM |
513 | } |
514 | ||
515 | /* Check zone type */ | |
516 | switch (zone->type) { | |
517 | case BLK_ZONE_TYPE_CONVENTIONAL: | |
e94f5819 CH |
518 | if (!args->conv_zones_bitmap) { |
519 | args->conv_zones_bitmap = | |
520 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
521 | if (!args->conv_zones_bitmap) | |
522 | return -ENOMEM; | |
523 | } | |
524 | set_bit(idx, args->conv_zones_bitmap); | |
525 | break; | |
d9dd7308 DLM |
526 | case BLK_ZONE_TYPE_SEQWRITE_REQ: |
527 | case BLK_ZONE_TYPE_SEQWRITE_PREF: | |
e94f5819 CH |
528 | if (!args->seq_zones_wlock) { |
529 | args->seq_zones_wlock = | |
530 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
531 | if (!args->seq_zones_wlock) | |
532 | return -ENOMEM; | |
533 | } | |
d9dd7308 DLM |
534 | break; |
535 | default: | |
536 | pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n", | |
537 | disk->disk_name, (int)zone->type, zone->start); | |
d4100351 | 538 | return -ENODEV; |
d9dd7308 DLM |
539 | } |
540 | ||
d4100351 CH |
541 | args->sector += zone->len; |
542 | return 0; | |
543 | } | |
544 | ||
bf505456 DLM |
545 | /** |
546 | * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps | |
547 | * @disk: Target disk | |
e732671a | 548 | * @update_driver_data: Callback to update driver data on the frozen disk |
bf505456 DLM |
549 | * |
550 | * Helper function for low-level device drivers to (re) allocate and initialize | |
551 | * a disk request queue zone bitmaps. This functions should normally be called | |
ae58954d CH |
552 | * within the disk ->revalidate method for blk-mq based drivers. For BIO based |
553 | * drivers only q->nr_zones needs to be updated so that the sysfs exposed value | |
554 | * is correct. | |
e732671a DLM |
555 | * If the @update_driver_data callback function is not NULL, the callback is |
556 | * executed with the device request queue frozen after all zones have been | |
557 | * checked. | |
bf505456 | 558 | */ |
e732671a DLM |
559 | int blk_revalidate_disk_zones(struct gendisk *disk, |
560 | void (*update_driver_data)(struct gendisk *disk)) | |
bf505456 DLM |
561 | { |
562 | struct request_queue *q = disk->queue; | |
e94f5819 CH |
563 | struct blk_revalidate_zone_args args = { |
564 | .disk = disk, | |
e94f5819 | 565 | }; |
6c6b3549 CH |
566 | unsigned int noio_flag; |
567 | int ret; | |
bf505456 | 568 | |
c98c3d09 CH |
569 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) |
570 | return -EIO; | |
ae58954d CH |
571 | if (WARN_ON_ONCE(!queue_is_mq(q))) |
572 | return -EIO; | |
bf505456 | 573 | |
1a1206dc JT |
574 | if (!get_capacity(disk)) |
575 | return -EIO; | |
576 | ||
e94f5819 | 577 | /* |
6c6b3549 CH |
578 | * Ensure that all memory allocations in this context are done as if |
579 | * GFP_NOIO was specified. | |
e94f5819 | 580 | */ |
6c6b3549 CH |
581 | noio_flag = memalloc_noio_save(); |
582 | ret = disk->fops->report_zones(disk, 0, UINT_MAX, | |
583 | blk_revalidate_zone_cb, &args); | |
2afdeb23 DLM |
584 | if (!ret) { |
585 | pr_warn("%s: No zones reported\n", disk->disk_name); | |
586 | ret = -ENODEV; | |
587 | } | |
6c6b3549 | 588 | memalloc_noio_restore(noio_flag); |
bf505456 | 589 | |
2afdeb23 DLM |
590 | /* |
591 | * If zones where reported, make sure that the entire disk capacity | |
592 | * has been checked. | |
593 | */ | |
594 | if (ret > 0 && args.sector != get_capacity(disk)) { | |
595 | pr_warn("%s: Missing zones from sector %llu\n", | |
596 | disk->disk_name, args.sector); | |
597 | ret = -ENODEV; | |
598 | } | |
599 | ||
bf505456 | 600 | /* |
6c6b3549 CH |
601 | * Install the new bitmaps and update nr_zones only once the queue is |
602 | * stopped and all I/Os are completed (i.e. a scheduler is not | |
603 | * referencing the bitmaps). | |
bf505456 DLM |
604 | */ |
605 | blk_mq_freeze_queue(q); | |
2afdeb23 | 606 | if (ret > 0) { |
6c6b3549 | 607 | blk_queue_chunk_sectors(q, args.zone_sectors); |
e94f5819 | 608 | q->nr_zones = args.nr_zones; |
d4100351 | 609 | swap(q->seq_zones_wlock, args.seq_zones_wlock); |
f216fdd7 | 610 | swap(q->conv_zones_bitmap, args.conv_zones_bitmap); |
e732671a DLM |
611 | if (update_driver_data) |
612 | update_driver_data(disk); | |
d4100351 CH |
613 | ret = 0; |
614 | } else { | |
bf505456 | 615 | pr_warn("%s: failed to revalidate zones\n", disk->disk_name); |
bf505456 | 616 | blk_queue_free_zone_bitmaps(q); |
bf505456 | 617 | } |
d4100351 | 618 | blk_mq_unfreeze_queue(q); |
bf505456 | 619 | |
d4100351 | 620 | kfree(args.seq_zones_wlock); |
f216fdd7 | 621 | kfree(args.conv_zones_bitmap); |
bf505456 DLM |
622 | return ret; |
623 | } | |
624 | EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones); | |
508aebb8 DLM |
625 | |
626 | void blk_queue_clear_zone_settings(struct request_queue *q) | |
627 | { | |
628 | blk_mq_freeze_queue(q); | |
629 | ||
630 | blk_queue_free_zone_bitmaps(q); | |
631 | blk_queue_flag_clear(QUEUE_FLAG_ZONE_RESETALL, q); | |
632 | q->required_elevator_features &= ~ELEVATOR_F_ZBD_SEQ_WRITE; | |
633 | q->nr_zones = 0; | |
634 | q->max_open_zones = 0; | |
635 | q->max_active_zones = 0; | |
636 | q->limits.chunk_sectors = 0; | |
637 | q->limits.zone_write_granularity = 0; | |
638 | q->limits.max_zone_append_sectors = 0; | |
639 | ||
640 | blk_mq_unfreeze_queue(q); | |
641 | } |