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 (!rq->q->disk->seq_zones_wlock) |
6cc77e9c CH |
61 | return false; |
62 | ||
19821fee | 63 | return blk_rq_is_seq_zoned_write(rq); |
6cc77e9c CH |
64 | } |
65 | EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock); | |
66 | ||
1392d370 JT |
67 | bool blk_req_zone_write_trylock(struct request *rq) |
68 | { | |
69 | unsigned int zno = blk_rq_zone_no(rq); | |
70 | ||
d86e716a | 71 | if (test_and_set_bit(zno, rq->q->disk->seq_zones_wlock)) |
1392d370 JT |
72 | return false; |
73 | ||
74 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
75 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
76 | ||
77 | return true; | |
78 | } | |
79 | EXPORT_SYMBOL_GPL(blk_req_zone_write_trylock); | |
80 | ||
6cc77e9c CH |
81 | void __blk_req_zone_write_lock(struct request *rq) |
82 | { | |
83 | if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq), | |
d86e716a | 84 | rq->q->disk->seq_zones_wlock))) |
6cc77e9c CH |
85 | return; |
86 | ||
87 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
88 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
89 | } | |
90 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock); | |
91 | ||
92 | void __blk_req_zone_write_unlock(struct request *rq) | |
93 | { | |
94 | rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED; | |
d86e716a | 95 | if (rq->q->disk->seq_zones_wlock) |
6cc77e9c | 96 | WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq), |
d86e716a | 97 | rq->q->disk->seq_zones_wlock)); |
6cc77e9c CH |
98 | } |
99 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock); | |
100 | ||
a91e1380 | 101 | /** |
b623e347 CH |
102 | * bdev_nr_zones - Get number of zones |
103 | * @bdev: Target device | |
a91e1380 | 104 | * |
9b38bb4b CH |
105 | * Return the total number of zones of a zoned block device. For a block |
106 | * device without zone capabilities, the number of zones is always 0. | |
a91e1380 | 107 | */ |
b623e347 | 108 | unsigned int bdev_nr_zones(struct block_device *bdev) |
a91e1380 | 109 | { |
b623e347 | 110 | sector_t zone_sectors = bdev_zone_sectors(bdev); |
a91e1380 | 111 | |
b623e347 | 112 | if (!bdev_is_zoned(bdev)) |
a91e1380 | 113 | return 0; |
b623e347 CH |
114 | return (bdev_nr_sectors(bdev) + zone_sectors - 1) >> |
115 | ilog2(zone_sectors); | |
a91e1380 | 116 | } |
b623e347 | 117 | EXPORT_SYMBOL_GPL(bdev_nr_zones); |
a91e1380 | 118 | |
6a0cb1bc HR |
119 | /** |
120 | * blkdev_report_zones - Get zones information | |
121 | * @bdev: Target block device | |
122 | * @sector: Sector from which to report zones | |
d4100351 CH |
123 | * @nr_zones: Maximum number of zones to report |
124 | * @cb: Callback function called for each reported zone | |
125 | * @data: Private data for the callback | |
6a0cb1bc HR |
126 | * |
127 | * Description: | |
d4100351 CH |
128 | * Get zone information starting from the zone containing @sector for at most |
129 | * @nr_zones, and call @cb for each zone reported by the device. | |
130 | * To report all zones in a device starting from @sector, the BLK_ALL_ZONES | |
131 | * constant can be passed to @nr_zones. | |
132 | * Returns the number of zones reported by the device, or a negative errno | |
133 | * value in case of failure. | |
134 | * | |
135 | * Note: The caller must use memalloc_noXX_save/restore() calls to control | |
136 | * memory allocations done within this function. | |
6a0cb1bc | 137 | */ |
e76239a3 | 138 | int blkdev_report_zones(struct block_device *bdev, sector_t sector, |
d4100351 | 139 | unsigned int nr_zones, report_zones_cb cb, void *data) |
6a0cb1bc | 140 | { |
ceeb373a | 141 | struct gendisk *disk = bdev->bd_disk; |
5eac3eb3 | 142 | sector_t capacity = get_capacity(disk); |
6a0cb1bc | 143 | |
edd1dbc8 | 144 | if (!bdev_is_zoned(bdev) || WARN_ON_ONCE(!disk->fops->report_zones)) |
e76239a3 | 145 | return -EOPNOTSUPP; |
6a0cb1bc | 146 | |
d4100351 | 147 | if (!nr_zones || sector >= capacity) |
6a0cb1bc | 148 | return 0; |
6a0cb1bc | 149 | |
d4100351 | 150 | return disk->fops->report_zones(disk, sector, nr_zones, cb, data); |
6a0cb1bc HR |
151 | } |
152 | EXPORT_SYMBOL_GPL(blkdev_report_zones); | |
153 | ||
1ee533ec DLM |
154 | static inline unsigned long *blk_alloc_zone_bitmap(int node, |
155 | unsigned int nr_zones) | |
6e33dbf2 | 156 | { |
1ee533ec DLM |
157 | return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long), |
158 | GFP_NOIO, node); | |
159 | } | |
6e33dbf2 | 160 | |
1ee533ec DLM |
161 | static int blk_zone_need_reset_cb(struct blk_zone *zone, unsigned int idx, |
162 | void *data) | |
163 | { | |
6e33dbf2 | 164 | /* |
1ee533ec DLM |
165 | * For an all-zones reset, ignore conventional, empty, read-only |
166 | * and offline zones. | |
6e33dbf2 | 167 | */ |
1ee533ec DLM |
168 | switch (zone->cond) { |
169 | case BLK_ZONE_COND_NOT_WP: | |
170 | case BLK_ZONE_COND_EMPTY: | |
171 | case BLK_ZONE_COND_READONLY: | |
172 | case BLK_ZONE_COND_OFFLINE: | |
173 | return 0; | |
174 | default: | |
175 | set_bit(idx, (unsigned long *)data); | |
176 | return 0; | |
177 | } | |
178 | } | |
179 | ||
180 | static int blkdev_zone_reset_all_emulated(struct block_device *bdev, | |
181 | gfp_t gfp_mask) | |
182 | { | |
d86e716a | 183 | struct gendisk *disk = bdev->bd_disk; |
375c140c CH |
184 | sector_t capacity = bdev_nr_sectors(bdev); |
185 | sector_t zone_sectors = bdev_zone_sectors(bdev); | |
1ee533ec DLM |
186 | unsigned long *need_reset; |
187 | struct bio *bio = NULL; | |
188 | sector_t sector = 0; | |
189 | int ret; | |
190 | ||
d86e716a | 191 | need_reset = blk_alloc_zone_bitmap(disk->queue->node, disk->nr_zones); |
1ee533ec DLM |
192 | if (!need_reset) |
193 | return -ENOMEM; | |
194 | ||
d86e716a CH |
195 | ret = disk->fops->report_zones(disk, 0, disk->nr_zones, |
196 | blk_zone_need_reset_cb, need_reset); | |
1ee533ec DLM |
197 | if (ret < 0) |
198 | goto out_free_need_reset; | |
199 | ||
200 | ret = 0; | |
201 | while (sector < capacity) { | |
d86e716a | 202 | if (!test_bit(disk_zone_no(disk, sector), need_reset)) { |
1ee533ec DLM |
203 | sector += zone_sectors; |
204 | continue; | |
205 | } | |
206 | ||
0a3140ea CK |
207 | bio = blk_next_bio(bio, bdev, 0, REQ_OP_ZONE_RESET | REQ_SYNC, |
208 | gfp_mask); | |
1ee533ec DLM |
209 | bio->bi_iter.bi_sector = sector; |
210 | sector += zone_sectors; | |
211 | ||
212 | /* This may take a while, so be nice to others */ | |
213 | cond_resched(); | |
214 | } | |
215 | ||
216 | if (bio) { | |
217 | ret = submit_bio_wait(bio); | |
218 | bio_put(bio); | |
219 | } | |
220 | ||
221 | out_free_need_reset: | |
222 | kfree(need_reset); | |
223 | return ret; | |
224 | } | |
225 | ||
226 | static int blkdev_zone_reset_all(struct block_device *bdev, gfp_t gfp_mask) | |
227 | { | |
228 | struct bio bio; | |
229 | ||
49add496 | 230 | bio_init(&bio, bdev, NULL, 0, REQ_OP_ZONE_RESET_ALL | REQ_SYNC); |
1ee533ec | 231 | return submit_bio_wait(&bio); |
6e33dbf2 CK |
232 | } |
233 | ||
6a0cb1bc | 234 | /** |
6c1b1da5 | 235 | * blkdev_zone_mgmt - Execute a zone management operation on a range of zones |
6a0cb1bc | 236 | * @bdev: Target block device |
6c1b1da5 AJ |
237 | * @op: Operation to be performed on the zones |
238 | * @sector: Start sector of the first zone to operate on | |
239 | * @nr_sectors: Number of sectors, should be at least the length of one zone and | |
240 | * must be zone size aligned. | |
6a0cb1bc HR |
241 | * @gfp_mask: Memory allocation flags (for bio_alloc) |
242 | * | |
243 | * Description: | |
6c1b1da5 | 244 | * Perform the specified operation on the range of zones specified by |
6a0cb1bc HR |
245 | * @sector..@sector+@nr_sectors. Specifying the entire disk sector range |
246 | * is valid, but the specified range should not contain conventional zones. | |
6c1b1da5 AJ |
247 | * The operation to execute on each zone can be a zone reset, open, close |
248 | * or finish request. | |
6a0cb1bc | 249 | */ |
ff07a02e BVA |
250 | int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, |
251 | sector_t sector, sector_t nr_sectors, gfp_t gfp_mask) | |
6a0cb1bc HR |
252 | { |
253 | struct request_queue *q = bdev_get_queue(bdev); | |
375c140c CH |
254 | sector_t zone_sectors = bdev_zone_sectors(bdev); |
255 | sector_t capacity = bdev_nr_sectors(bdev); | |
6a0cb1bc | 256 | sector_t end_sector = sector + nr_sectors; |
a2d6b3a2 | 257 | struct bio *bio = NULL; |
1ee533ec | 258 | int ret = 0; |
6a0cb1bc | 259 | |
edd1dbc8 | 260 | if (!bdev_is_zoned(bdev)) |
6a0cb1bc HR |
261 | return -EOPNOTSUPP; |
262 | ||
a2d6b3a2 DLM |
263 | if (bdev_read_only(bdev)) |
264 | return -EPERM; | |
265 | ||
6c1b1da5 AJ |
266 | if (!op_is_zone_mgmt(op)) |
267 | return -EOPNOTSUPP; | |
268 | ||
11bde986 | 269 | if (end_sector <= sector || end_sector > capacity) |
6a0cb1bc HR |
270 | /* Out of range */ |
271 | return -EINVAL; | |
272 | ||
273 | /* Check alignment (handle eventual smaller last zone) */ | |
e29b2100 | 274 | if (!bdev_is_zone_start(bdev, sector)) |
6a0cb1bc HR |
275 | return -EINVAL; |
276 | ||
e29b2100 | 277 | if (!bdev_is_zone_start(bdev, nr_sectors) && end_sector != capacity) |
6a0cb1bc HR |
278 | return -EINVAL; |
279 | ||
1ee533ec DLM |
280 | /* |
281 | * In the case of a zone reset operation over all zones, | |
282 | * REQ_OP_ZONE_RESET_ALL can be used with devices supporting this | |
283 | * command. For other devices, we emulate this command behavior by | |
284 | * identifying the zones needing a reset. | |
285 | */ | |
286 | if (op == REQ_OP_ZONE_RESET && sector == 0 && nr_sectors == capacity) { | |
287 | if (!blk_queue_zone_resetall(q)) | |
288 | return blkdev_zone_reset_all_emulated(bdev, gfp_mask); | |
289 | return blkdev_zone_reset_all(bdev, gfp_mask); | |
290 | } | |
291 | ||
6a0cb1bc | 292 | while (sector < end_sector) { |
0a3140ea | 293 | bio = blk_next_bio(bio, bdev, 0, op | REQ_SYNC, gfp_mask); |
c7a1d926 | 294 | bio->bi_iter.bi_sector = sector; |
6a0cb1bc HR |
295 | sector += zone_sectors; |
296 | ||
297 | /* This may take a while, so be nice to others */ | |
298 | cond_resched(); | |
6a0cb1bc HR |
299 | } |
300 | ||
a2d6b3a2 DLM |
301 | ret = submit_bio_wait(bio); |
302 | bio_put(bio); | |
303 | ||
a2d6b3a2 | 304 | return ret; |
6a0cb1bc | 305 | } |
6c1b1da5 | 306 | EXPORT_SYMBOL_GPL(blkdev_zone_mgmt); |
3ed05a98 | 307 | |
d4100351 CH |
308 | struct zone_report_args { |
309 | struct blk_zone __user *zones; | |
310 | }; | |
311 | ||
312 | static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx, | |
313 | void *data) | |
314 | { | |
315 | struct zone_report_args *args = data; | |
316 | ||
317 | if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone))) | |
318 | return -EFAULT; | |
319 | return 0; | |
320 | } | |
321 | ||
56c4bddb | 322 | /* |
3ed05a98 ST |
323 | * BLKREPORTZONE ioctl processing. |
324 | * Called from blkdev_ioctl. | |
325 | */ | |
5e4ea834 CH |
326 | int blkdev_report_zones_ioctl(struct block_device *bdev, unsigned int cmd, |
327 | unsigned long arg) | |
3ed05a98 ST |
328 | { |
329 | void __user *argp = (void __user *)arg; | |
d4100351 | 330 | struct zone_report_args args; |
3ed05a98 | 331 | struct blk_zone_report rep; |
3ed05a98 ST |
332 | int ret; |
333 | ||
334 | if (!argp) | |
335 | return -EINVAL; | |
336 | ||
edd1dbc8 | 337 | if (!bdev_is_zoned(bdev)) |
3ed05a98 ST |
338 | return -ENOTTY; |
339 | ||
3ed05a98 ST |
340 | if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report))) |
341 | return -EFAULT; | |
342 | ||
343 | if (!rep.nr_zones) | |
344 | return -EINVAL; | |
345 | ||
d4100351 CH |
346 | args.zones = argp + sizeof(struct blk_zone_report); |
347 | ret = blkdev_report_zones(bdev, rep.sector, rep.nr_zones, | |
348 | blkdev_copy_zone_to_user, &args); | |
349 | if (ret < 0) | |
350 | return ret; | |
3ed05a98 | 351 | |
d4100351 | 352 | rep.nr_zones = ret; |
82394db7 | 353 | rep.flags = BLK_ZONE_REP_CAPACITY; |
d4100351 CH |
354 | if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report))) |
355 | return -EFAULT; | |
356 | return 0; | |
3ed05a98 ST |
357 | } |
358 | ||
05bdb996 CH |
359 | static int blkdev_truncate_zone_range(struct block_device *bdev, |
360 | blk_mode_t mode, const struct blk_zone_range *zrange) | |
e5113505 SK |
361 | { |
362 | loff_t start, end; | |
363 | ||
364 | if (zrange->sector + zrange->nr_sectors <= zrange->sector || | |
365 | zrange->sector + zrange->nr_sectors > get_capacity(bdev->bd_disk)) | |
366 | /* Out of range */ | |
367 | return -EINVAL; | |
368 | ||
369 | start = zrange->sector << SECTOR_SHIFT; | |
370 | end = ((zrange->sector + zrange->nr_sectors) << SECTOR_SHIFT) - 1; | |
371 | ||
372 | return truncate_bdev_range(bdev, mode, start, end); | |
373 | } | |
374 | ||
56c4bddb | 375 | /* |
e876df1f | 376 | * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing. |
3ed05a98 ST |
377 | * Called from blkdev_ioctl. |
378 | */ | |
05bdb996 | 379 | int blkdev_zone_mgmt_ioctl(struct block_device *bdev, blk_mode_t mode, |
e876df1f | 380 | unsigned int cmd, unsigned long arg) |
3ed05a98 ST |
381 | { |
382 | void __user *argp = (void __user *)arg; | |
3ed05a98 | 383 | struct blk_zone_range zrange; |
ff07a02e | 384 | enum req_op op; |
e5113505 | 385 | int ret; |
3ed05a98 ST |
386 | |
387 | if (!argp) | |
388 | return -EINVAL; | |
389 | ||
edd1dbc8 | 390 | if (!bdev_is_zoned(bdev)) |
3ed05a98 ST |
391 | return -ENOTTY; |
392 | ||
05bdb996 | 393 | if (!(mode & BLK_OPEN_WRITE)) |
3ed05a98 ST |
394 | return -EBADF; |
395 | ||
396 | if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range))) | |
397 | return -EFAULT; | |
398 | ||
e876df1f AJ |
399 | switch (cmd) { |
400 | case BLKRESETZONE: | |
401 | op = REQ_OP_ZONE_RESET; | |
e5113505 SK |
402 | |
403 | /* Invalidate the page cache, including dirty pages. */ | |
86399ea0 | 404 | filemap_invalidate_lock(bdev->bd_inode->i_mapping); |
e5113505 SK |
405 | ret = blkdev_truncate_zone_range(bdev, mode, &zrange); |
406 | if (ret) | |
86399ea0 | 407 | goto fail; |
e876df1f AJ |
408 | break; |
409 | case BLKOPENZONE: | |
410 | op = REQ_OP_ZONE_OPEN; | |
411 | break; | |
412 | case BLKCLOSEZONE: | |
413 | op = REQ_OP_ZONE_CLOSE; | |
414 | break; | |
415 | case BLKFINISHZONE: | |
416 | op = REQ_OP_ZONE_FINISH; | |
417 | break; | |
418 | default: | |
419 | return -ENOTTY; | |
420 | } | |
421 | ||
e5113505 SK |
422 | ret = blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors, |
423 | GFP_KERNEL); | |
424 | ||
86399ea0 SK |
425 | fail: |
426 | if (cmd == BLKRESETZONE) | |
427 | filemap_invalidate_unlock(bdev->bd_inode->i_mapping); | |
e5113505 SK |
428 | |
429 | return ret; | |
3ed05a98 | 430 | } |
bf505456 | 431 | |
5d400665 | 432 | void disk_free_zone_bitmaps(struct gendisk *disk) |
bf505456 | 433 | { |
d86e716a CH |
434 | kfree(disk->conv_zones_bitmap); |
435 | disk->conv_zones_bitmap = NULL; | |
436 | kfree(disk->seq_zones_wlock); | |
437 | disk->seq_zones_wlock = NULL; | |
bf505456 DLM |
438 | } |
439 | ||
d4100351 CH |
440 | struct blk_revalidate_zone_args { |
441 | struct gendisk *disk; | |
f216fdd7 | 442 | unsigned long *conv_zones_bitmap; |
d4100351 | 443 | unsigned long *seq_zones_wlock; |
e94f5819 | 444 | unsigned int nr_zones; |
6c6b3549 | 445 | sector_t zone_sectors; |
d4100351 CH |
446 | sector_t sector; |
447 | }; | |
448 | ||
d9dd7308 DLM |
449 | /* |
450 | * Helper function to check the validity of zones of a zoned block device. | |
451 | */ | |
d4100351 CH |
452 | static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx, |
453 | void *data) | |
d9dd7308 | 454 | { |
d4100351 CH |
455 | struct blk_revalidate_zone_args *args = data; |
456 | struct gendisk *disk = args->disk; | |
d9dd7308 | 457 | struct request_queue *q = disk->queue; |
d9dd7308 DLM |
458 | sector_t capacity = get_capacity(disk); |
459 | ||
460 | /* | |
461 | * All zones must have the same size, with the exception on an eventual | |
462 | * smaller last zone. | |
463 | */ | |
6c6b3549 CH |
464 | if (zone->start == 0) { |
465 | if (zone->len == 0 || !is_power_of_2(zone->len)) { | |
466 | pr_warn("%s: Invalid zoned device with non power of two zone size (%llu)\n", | |
467 | disk->disk_name, zone->len); | |
468 | return -ENODEV; | |
469 | } | |
d9dd7308 | 470 | |
6c6b3549 CH |
471 | args->zone_sectors = zone->len; |
472 | args->nr_zones = (capacity + zone->len - 1) >> ilog2(zone->len); | |
473 | } else if (zone->start + args->zone_sectors < capacity) { | |
474 | if (zone->len != args->zone_sectors) { | |
475 | pr_warn("%s: Invalid zoned device with non constant zone size\n", | |
476 | disk->disk_name); | |
477 | return -ENODEV; | |
478 | } | |
479 | } else { | |
480 | if (zone->len > args->zone_sectors) { | |
481 | pr_warn("%s: Invalid zoned device with larger last zone size\n", | |
482 | disk->disk_name); | |
483 | return -ENODEV; | |
484 | } | |
d9dd7308 DLM |
485 | } |
486 | ||
487 | /* Check for holes in the zone report */ | |
d4100351 | 488 | if (zone->start != args->sector) { |
d9dd7308 | 489 | pr_warn("%s: Zone gap at sectors %llu..%llu\n", |
d4100351 CH |
490 | disk->disk_name, args->sector, zone->start); |
491 | return -ENODEV; | |
d9dd7308 DLM |
492 | } |
493 | ||
494 | /* Check zone type */ | |
495 | switch (zone->type) { | |
496 | case BLK_ZONE_TYPE_CONVENTIONAL: | |
e94f5819 CH |
497 | if (!args->conv_zones_bitmap) { |
498 | args->conv_zones_bitmap = | |
499 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
500 | if (!args->conv_zones_bitmap) | |
501 | return -ENOMEM; | |
502 | } | |
503 | set_bit(idx, args->conv_zones_bitmap); | |
504 | break; | |
d9dd7308 DLM |
505 | case BLK_ZONE_TYPE_SEQWRITE_REQ: |
506 | case BLK_ZONE_TYPE_SEQWRITE_PREF: | |
e94f5819 CH |
507 | if (!args->seq_zones_wlock) { |
508 | args->seq_zones_wlock = | |
509 | blk_alloc_zone_bitmap(q->node, args->nr_zones); | |
510 | if (!args->seq_zones_wlock) | |
511 | return -ENOMEM; | |
512 | } | |
d9dd7308 DLM |
513 | break; |
514 | default: | |
515 | pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n", | |
516 | disk->disk_name, (int)zone->type, zone->start); | |
d4100351 | 517 | return -ENODEV; |
d9dd7308 DLM |
518 | } |
519 | ||
d4100351 CH |
520 | args->sector += zone->len; |
521 | return 0; | |
522 | } | |
523 | ||
bf505456 DLM |
524 | /** |
525 | * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps | |
526 | * @disk: Target disk | |
e732671a | 527 | * @update_driver_data: Callback to update driver data on the frozen disk |
bf505456 DLM |
528 | * |
529 | * Helper function for low-level device drivers to (re) allocate and initialize | |
530 | * a disk request queue zone bitmaps. This functions should normally be called | |
ae58954d CH |
531 | * within the disk ->revalidate method for blk-mq based drivers. For BIO based |
532 | * drivers only q->nr_zones needs to be updated so that the sysfs exposed value | |
533 | * is correct. | |
e732671a DLM |
534 | * If the @update_driver_data callback function is not NULL, the callback is |
535 | * executed with the device request queue frozen after all zones have been | |
536 | * checked. | |
bf505456 | 537 | */ |
e732671a DLM |
538 | int blk_revalidate_disk_zones(struct gendisk *disk, |
539 | void (*update_driver_data)(struct gendisk *disk)) | |
bf505456 DLM |
540 | { |
541 | struct request_queue *q = disk->queue; | |
e94f5819 CH |
542 | struct blk_revalidate_zone_args args = { |
543 | .disk = disk, | |
e94f5819 | 544 | }; |
6c6b3549 CH |
545 | unsigned int noio_flag; |
546 | int ret; | |
bf505456 | 547 | |
c98c3d09 CH |
548 | if (WARN_ON_ONCE(!blk_queue_is_zoned(q))) |
549 | return -EIO; | |
ae58954d CH |
550 | if (WARN_ON_ONCE(!queue_is_mq(q))) |
551 | return -EIO; | |
bf505456 | 552 | |
1a1206dc JT |
553 | if (!get_capacity(disk)) |
554 | return -EIO; | |
555 | ||
e94f5819 | 556 | /* |
6c6b3549 CH |
557 | * Ensure that all memory allocations in this context are done as if |
558 | * GFP_NOIO was specified. | |
e94f5819 | 559 | */ |
6c6b3549 CH |
560 | noio_flag = memalloc_noio_save(); |
561 | ret = disk->fops->report_zones(disk, 0, UINT_MAX, | |
562 | blk_revalidate_zone_cb, &args); | |
2afdeb23 DLM |
563 | if (!ret) { |
564 | pr_warn("%s: No zones reported\n", disk->disk_name); | |
565 | ret = -ENODEV; | |
566 | } | |
6c6b3549 | 567 | memalloc_noio_restore(noio_flag); |
bf505456 | 568 | |
2afdeb23 DLM |
569 | /* |
570 | * If zones where reported, make sure that the entire disk capacity | |
571 | * has been checked. | |
572 | */ | |
573 | if (ret > 0 && args.sector != get_capacity(disk)) { | |
574 | pr_warn("%s: Missing zones from sector %llu\n", | |
575 | disk->disk_name, args.sector); | |
576 | ret = -ENODEV; | |
577 | } | |
578 | ||
bf505456 | 579 | /* |
6c6b3549 CH |
580 | * Install the new bitmaps and update nr_zones only once the queue is |
581 | * stopped and all I/Os are completed (i.e. a scheduler is not | |
582 | * referencing the bitmaps). | |
bf505456 DLM |
583 | */ |
584 | blk_mq_freeze_queue(q); | |
2afdeb23 | 585 | if (ret > 0) { |
6c6b3549 | 586 | blk_queue_chunk_sectors(q, args.zone_sectors); |
d86e716a CH |
587 | disk->nr_zones = args.nr_zones; |
588 | swap(disk->seq_zones_wlock, args.seq_zones_wlock); | |
589 | swap(disk->conv_zones_bitmap, args.conv_zones_bitmap); | |
e732671a DLM |
590 | if (update_driver_data) |
591 | update_driver_data(disk); | |
d4100351 CH |
592 | ret = 0; |
593 | } else { | |
bf505456 | 594 | pr_warn("%s: failed to revalidate zones\n", disk->disk_name); |
5d400665 | 595 | disk_free_zone_bitmaps(disk); |
bf505456 | 596 | } |
d4100351 | 597 | blk_mq_unfreeze_queue(q); |
bf505456 | 598 | |
d4100351 | 599 | kfree(args.seq_zones_wlock); |
f216fdd7 | 600 | kfree(args.conv_zones_bitmap); |
bf505456 DLM |
601 | return ret; |
602 | } | |
603 | EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones); | |
508aebb8 | 604 | |
b3c72f81 | 605 | void disk_clear_zone_settings(struct gendisk *disk) |
508aebb8 | 606 | { |
b3c72f81 CH |
607 | struct request_queue *q = disk->queue; |
608 | ||
508aebb8 DLM |
609 | blk_mq_freeze_queue(q); |
610 | ||
5d400665 | 611 | disk_free_zone_bitmaps(disk); |
508aebb8 DLM |
612 | blk_queue_flag_clear(QUEUE_FLAG_ZONE_RESETALL, q); |
613 | q->required_elevator_features &= ~ELEVATOR_F_ZBD_SEQ_WRITE; | |
d86e716a CH |
614 | disk->nr_zones = 0; |
615 | disk->max_open_zones = 0; | |
616 | disk->max_active_zones = 0; | |
508aebb8 DLM |
617 | q->limits.chunk_sectors = 0; |
618 | q->limits.zone_write_granularity = 0; | |
619 | q->limits.max_zone_append_sectors = 0; | |
620 | ||
621 | blk_mq_unfreeze_queue(q); | |
622 | } |