Commit | Line | Data |
---|---|---|
6a0cb1bc HR |
1 | /* |
2 | * Zoned block device handling | |
3 | * | |
4 | * Copyright (c) 2015, Hannes Reinecke | |
5 | * Copyright (c) 2015, SUSE Linux GmbH | |
6 | * | |
7 | * Copyright (c) 2016, Damien Le Moal | |
8 | * Copyright (c) 2016, Western Digital | |
9 | */ | |
10 | ||
11 | #include <linux/kernel.h> | |
12 | #include <linux/module.h> | |
13 | #include <linux/rbtree.h> | |
14 | #include <linux/blkdev.h> | |
15 | ||
a2d6b3a2 DLM |
16 | #include "blk.h" |
17 | ||
6a0cb1bc HR |
18 | static inline sector_t blk_zone_start(struct request_queue *q, |
19 | sector_t sector) | |
20 | { | |
f99e8648 | 21 | sector_t zone_mask = blk_queue_zone_sectors(q) - 1; |
6a0cb1bc HR |
22 | |
23 | return sector & ~zone_mask; | |
24 | } | |
25 | ||
6cc77e9c CH |
26 | /* |
27 | * Return true if a request is a write requests that needs zone write locking. | |
28 | */ | |
29 | bool blk_req_needs_zone_write_lock(struct request *rq) | |
30 | { | |
31 | if (!rq->q->seq_zones_wlock) | |
32 | return false; | |
33 | ||
34 | if (blk_rq_is_passthrough(rq)) | |
35 | return false; | |
36 | ||
37 | switch (req_op(rq)) { | |
38 | case REQ_OP_WRITE_ZEROES: | |
39 | case REQ_OP_WRITE_SAME: | |
40 | case REQ_OP_WRITE: | |
41 | return blk_rq_zone_is_seq(rq); | |
42 | default: | |
43 | return false; | |
44 | } | |
45 | } | |
46 | EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock); | |
47 | ||
48 | void __blk_req_zone_write_lock(struct request *rq) | |
49 | { | |
50 | if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq), | |
51 | rq->q->seq_zones_wlock))) | |
52 | return; | |
53 | ||
54 | WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED); | |
55 | rq->rq_flags |= RQF_ZONE_WRITE_LOCKED; | |
56 | } | |
57 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock); | |
58 | ||
59 | void __blk_req_zone_write_unlock(struct request *rq) | |
60 | { | |
61 | rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED; | |
62 | if (rq->q->seq_zones_wlock) | |
63 | WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq), | |
64 | rq->q->seq_zones_wlock)); | |
65 | } | |
66 | EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock); | |
67 | ||
a91e1380 DLM |
68 | static inline unsigned int __blkdev_nr_zones(struct request_queue *q, |
69 | sector_t nr_sectors) | |
70 | { | |
71 | unsigned long zone_sectors = blk_queue_zone_sectors(q); | |
72 | ||
73 | return (nr_sectors + zone_sectors - 1) >> ilog2(zone_sectors); | |
74 | } | |
75 | ||
76 | /** | |
77 | * blkdev_nr_zones - Get number of zones | |
78 | * @bdev: Target block device | |
79 | * | |
80 | * Description: | |
81 | * Return the total number of zones of a zoned block device. | |
82 | * For a regular block device, the number of zones is always 0. | |
83 | */ | |
84 | unsigned int blkdev_nr_zones(struct block_device *bdev) | |
85 | { | |
86 | struct request_queue *q = bdev_get_queue(bdev); | |
87 | ||
88 | if (!blk_queue_is_zoned(q)) | |
89 | return 0; | |
90 | ||
91 | return __blkdev_nr_zones(q, bdev->bd_part->nr_sects); | |
92 | } | |
93 | EXPORT_SYMBOL_GPL(blkdev_nr_zones); | |
94 | ||
6a0cb1bc | 95 | /* |
e76239a3 CH |
96 | * Check that a zone report belongs to this partition, and if yes, fix its start |
97 | * sector and write pointer and return true. Return false otherwise. | |
6a0cb1bc | 98 | */ |
e76239a3 | 99 | static bool blkdev_report_zone(struct block_device *bdev, struct blk_zone *rep) |
6a0cb1bc HR |
100 | { |
101 | sector_t offset = get_start_sect(bdev); | |
102 | ||
103 | if (rep->start < offset) | |
104 | return false; | |
105 | ||
106 | rep->start -= offset; | |
107 | if (rep->start + rep->len > bdev->bd_part->nr_sects) | |
108 | return false; | |
109 | ||
110 | if (rep->type == BLK_ZONE_TYPE_CONVENTIONAL) | |
111 | rep->wp = rep->start + rep->len; | |
112 | else | |
113 | rep->wp -= offset; | |
6a0cb1bc HR |
114 | return true; |
115 | } | |
116 | ||
e76239a3 CH |
117 | static int blk_report_zones(struct gendisk *disk, sector_t sector, |
118 | struct blk_zone *zones, unsigned int *nr_zones, | |
119 | gfp_t gfp_mask) | |
120 | { | |
121 | struct request_queue *q = disk->queue; | |
122 | unsigned int z = 0, n, nrz = *nr_zones; | |
123 | sector_t capacity = get_capacity(disk); | |
124 | int ret; | |
125 | ||
126 | while (z < nrz && sector < capacity) { | |
127 | n = nrz - z; | |
128 | ret = disk->fops->report_zones(disk, sector, &zones[z], &n, | |
129 | gfp_mask); | |
130 | if (ret) | |
131 | return ret; | |
132 | if (!n) | |
133 | break; | |
134 | sector += blk_queue_zone_sectors(q) * n; | |
135 | z += n; | |
136 | } | |
137 | ||
138 | WARN_ON(z > *nr_zones); | |
139 | *nr_zones = z; | |
140 | ||
141 | return 0; | |
142 | } | |
143 | ||
6a0cb1bc HR |
144 | /** |
145 | * blkdev_report_zones - Get zones information | |
146 | * @bdev: Target block device | |
147 | * @sector: Sector from which to report zones | |
148 | * @zones: Array of zone structures where to return the zones information | |
149 | * @nr_zones: Number of zone structures in the zone array | |
150 | * @gfp_mask: Memory allocation flags (for bio_alloc) | |
151 | * | |
152 | * Description: | |
153 | * Get zone information starting from the zone containing @sector. | |
154 | * The number of zone information reported may be less than the number | |
155 | * requested by @nr_zones. The number of zones actually reported is | |
156 | * returned in @nr_zones. | |
157 | */ | |
e76239a3 CH |
158 | int blkdev_report_zones(struct block_device *bdev, sector_t sector, |
159 | struct blk_zone *zones, unsigned int *nr_zones, | |
6a0cb1bc HR |
160 | gfp_t gfp_mask) |
161 | { | |
162 | struct request_queue *q = bdev_get_queue(bdev); | |
e76239a3 | 163 | unsigned int i, nrz; |
3c4da758 | 164 | int ret; |
6a0cb1bc | 165 | |
6a0cb1bc HR |
166 | if (!blk_queue_is_zoned(q)) |
167 | return -EOPNOTSUPP; | |
168 | ||
e76239a3 CH |
169 | /* |
170 | * A block device that advertized itself as zoned must have a | |
171 | * report_zones method. If it does not have one defined, the device | |
172 | * driver has a bug. So warn about that. | |
173 | */ | |
174 | if (WARN_ON_ONCE(!bdev->bd_disk->fops->report_zones)) | |
175 | return -EOPNOTSUPP; | |
6a0cb1bc | 176 | |
e76239a3 | 177 | if (!*nr_zones || sector >= bdev->bd_part->nr_sects) { |
6a0cb1bc HR |
178 | *nr_zones = 0; |
179 | return 0; | |
180 | } | |
181 | ||
e76239a3 CH |
182 | nrz = min(*nr_zones, |
183 | __blkdev_nr_zones(q, bdev->bd_part->nr_sects - sector)); | |
184 | ret = blk_report_zones(bdev->bd_disk, get_start_sect(bdev) + sector, | |
185 | zones, &nrz, gfp_mask); | |
6a0cb1bc | 186 | if (ret) |
e76239a3 | 187 | return ret; |
6a0cb1bc | 188 | |
e76239a3 CH |
189 | for (i = 0; i < nrz; i++) { |
190 | if (!blkdev_report_zone(bdev, zones)) | |
6a0cb1bc | 191 | break; |
e76239a3 | 192 | zones++; |
6a0cb1bc HR |
193 | } |
194 | ||
e76239a3 | 195 | *nr_zones = i; |
6a0cb1bc | 196 | |
e76239a3 | 197 | return 0; |
6a0cb1bc HR |
198 | } |
199 | EXPORT_SYMBOL_GPL(blkdev_report_zones); | |
200 | ||
201 | /** | |
202 | * blkdev_reset_zones - Reset zones write pointer | |
203 | * @bdev: Target block device | |
204 | * @sector: Start sector of the first zone to reset | |
205 | * @nr_sectors: Number of sectors, at least the length of one zone | |
206 | * @gfp_mask: Memory allocation flags (for bio_alloc) | |
207 | * | |
208 | * Description: | |
209 | * Reset the write pointer of the zones contained in the range | |
210 | * @sector..@sector+@nr_sectors. Specifying the entire disk sector range | |
211 | * is valid, but the specified range should not contain conventional zones. | |
212 | */ | |
213 | int blkdev_reset_zones(struct block_device *bdev, | |
214 | sector_t sector, sector_t nr_sectors, | |
215 | gfp_t gfp_mask) | |
216 | { | |
217 | struct request_queue *q = bdev_get_queue(bdev); | |
218 | sector_t zone_sectors; | |
219 | sector_t end_sector = sector + nr_sectors; | |
a2d6b3a2 DLM |
220 | struct bio *bio = NULL; |
221 | struct blk_plug plug; | |
6a0cb1bc HR |
222 | int ret; |
223 | ||
6a0cb1bc HR |
224 | if (!blk_queue_is_zoned(q)) |
225 | return -EOPNOTSUPP; | |
226 | ||
a2d6b3a2 DLM |
227 | if (bdev_read_only(bdev)) |
228 | return -EPERM; | |
229 | ||
230 | if (!nr_sectors || end_sector > bdev->bd_part->nr_sects) | |
6a0cb1bc HR |
231 | /* Out of range */ |
232 | return -EINVAL; | |
233 | ||
234 | /* Check alignment (handle eventual smaller last zone) */ | |
f99e8648 | 235 | zone_sectors = blk_queue_zone_sectors(q); |
6a0cb1bc HR |
236 | if (sector & (zone_sectors - 1)) |
237 | return -EINVAL; | |
238 | ||
239 | if ((nr_sectors & (zone_sectors - 1)) && | |
240 | end_sector != bdev->bd_part->nr_sects) | |
241 | return -EINVAL; | |
242 | ||
a2d6b3a2 | 243 | blk_start_plug(&plug); |
6a0cb1bc HR |
244 | while (sector < end_sector) { |
245 | ||
a2d6b3a2 | 246 | bio = blk_next_bio(bio, 0, gfp_mask); |
6a0cb1bc | 247 | bio->bi_iter.bi_sector = sector; |
74d46992 | 248 | bio_set_dev(bio, bdev); |
6a0cb1bc HR |
249 | bio_set_op_attrs(bio, REQ_OP_ZONE_RESET, 0); |
250 | ||
6a0cb1bc HR |
251 | sector += zone_sectors; |
252 | ||
253 | /* This may take a while, so be nice to others */ | |
254 | cond_resched(); | |
255 | ||
256 | } | |
257 | ||
a2d6b3a2 DLM |
258 | ret = submit_bio_wait(bio); |
259 | bio_put(bio); | |
260 | ||
261 | blk_finish_plug(&plug); | |
262 | ||
263 | return ret; | |
6a0cb1bc HR |
264 | } |
265 | EXPORT_SYMBOL_GPL(blkdev_reset_zones); | |
3ed05a98 | 266 | |
56c4bddb | 267 | /* |
3ed05a98 ST |
268 | * BLKREPORTZONE ioctl processing. |
269 | * Called from blkdev_ioctl. | |
270 | */ | |
271 | int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
272 | unsigned int cmd, unsigned long arg) | |
273 | { | |
274 | void __user *argp = (void __user *)arg; | |
275 | struct request_queue *q; | |
276 | struct blk_zone_report rep; | |
277 | struct blk_zone *zones; | |
278 | int ret; | |
279 | ||
280 | if (!argp) | |
281 | return -EINVAL; | |
282 | ||
283 | q = bdev_get_queue(bdev); | |
284 | if (!q) | |
285 | return -ENXIO; | |
286 | ||
287 | if (!blk_queue_is_zoned(q)) | |
288 | return -ENOTTY; | |
289 | ||
290 | if (!capable(CAP_SYS_ADMIN)) | |
291 | return -EACCES; | |
292 | ||
293 | if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report))) | |
294 | return -EFAULT; | |
295 | ||
296 | if (!rep.nr_zones) | |
297 | return -EINVAL; | |
298 | ||
2e85fbaf | 299 | rep.nr_zones = min(blkdev_nr_zones(bdev), rep.nr_zones); |
327ea4ad | 300 | |
344476e1 KC |
301 | zones = kvmalloc_array(rep.nr_zones, sizeof(struct blk_zone), |
302 | GFP_KERNEL | __GFP_ZERO); | |
3ed05a98 ST |
303 | if (!zones) |
304 | return -ENOMEM; | |
305 | ||
306 | ret = blkdev_report_zones(bdev, rep.sector, | |
307 | zones, &rep.nr_zones, | |
308 | GFP_KERNEL); | |
309 | if (ret) | |
310 | goto out; | |
311 | ||
312 | if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report))) { | |
313 | ret = -EFAULT; | |
314 | goto out; | |
315 | } | |
316 | ||
317 | if (rep.nr_zones) { | |
318 | if (copy_to_user(argp + sizeof(struct blk_zone_report), zones, | |
319 | sizeof(struct blk_zone) * rep.nr_zones)) | |
320 | ret = -EFAULT; | |
321 | } | |
322 | ||
323 | out: | |
327ea4ad | 324 | kvfree(zones); |
3ed05a98 ST |
325 | |
326 | return ret; | |
327 | } | |
328 | ||
56c4bddb | 329 | /* |
3ed05a98 ST |
330 | * BLKRESETZONE ioctl processing. |
331 | * Called from blkdev_ioctl. | |
332 | */ | |
333 | int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, | |
334 | unsigned int cmd, unsigned long arg) | |
335 | { | |
336 | void __user *argp = (void __user *)arg; | |
337 | struct request_queue *q; | |
338 | struct blk_zone_range zrange; | |
339 | ||
340 | if (!argp) | |
341 | return -EINVAL; | |
342 | ||
343 | q = bdev_get_queue(bdev); | |
344 | if (!q) | |
345 | return -ENXIO; | |
346 | ||
347 | if (!blk_queue_is_zoned(q)) | |
348 | return -ENOTTY; | |
349 | ||
350 | if (!capable(CAP_SYS_ADMIN)) | |
351 | return -EACCES; | |
352 | ||
353 | if (!(mode & FMODE_WRITE)) | |
354 | return -EBADF; | |
355 | ||
356 | if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range))) | |
357 | return -EFAULT; | |
358 | ||
359 | return blkdev_reset_zones(bdev, zrange.sector, zrange.nr_sectors, | |
360 | GFP_KERNEL); | |
361 | } |