1 // SPDX-License-Identifier: GPL-2.0
3 * Zoned block device handling
5 * Copyright (c) 2015, Hannes Reinecke
6 * Copyright (c) 2015, SUSE Linux GmbH
8 * Copyright (c) 2016, Damien Le Moal
9 * Copyright (c) 2016, Western Digital
12 #include <linux/kernel.h>
13 #include <linux/module.h>
14 #include <linux/rbtree.h>
15 #include <linux/blkdev.h>
16 #include <linux/blk-mq.h>
18 #include <linux/vmalloc.h>
19 #include <linux/sched/mm.h>
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),
32 ZONE_COND_NAME(OFFLINE),
37 * blk_zone_cond_str - Return string XXX in BLK_ZONE_COND_XXX.
38 * @zone_cond: BLK_ZONE_COND_XXX.
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".
44 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond)
46 static const char *zone_cond_str = "UNKNOWN";
48 if (zone_cond < ARRAY_SIZE(zone_cond_name) && zone_cond_name[zone_cond])
49 zone_cond_str = zone_cond_name[zone_cond];
53 EXPORT_SYMBOL_GPL(blk_zone_cond_str);
55 static inline sector_t blk_zone_start(struct request_queue *q,
58 sector_t zone_mask = blk_queue_zone_sectors(q) - 1;
60 return sector & ~zone_mask;
64 * Return true if a request is a write requests that needs zone write locking.
66 bool blk_req_needs_zone_write_lock(struct request *rq)
68 if (!rq->q->seq_zones_wlock)
71 if (blk_rq_is_passthrough(rq))
75 case REQ_OP_WRITE_ZEROES:
76 case REQ_OP_WRITE_SAME:
78 return blk_rq_zone_is_seq(rq);
83 EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock);
85 void __blk_req_zone_write_lock(struct request *rq)
87 if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq),
88 rq->q->seq_zones_wlock)))
91 WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED);
92 rq->rq_flags |= RQF_ZONE_WRITE_LOCKED;
94 EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock);
96 void __blk_req_zone_write_unlock(struct request *rq)
98 rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED;
99 if (rq->q->seq_zones_wlock)
100 WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq),
101 rq->q->seq_zones_wlock));
103 EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock);
106 * blkdev_nr_zones - Get number of zones
107 * @disk: Target gendisk
109 * Return the total number of zones of a zoned block device. For a block
110 * device without zone capabilities, the number of zones is always 0.
112 unsigned int blkdev_nr_zones(struct gendisk *disk)
114 sector_t zone_sectors = blk_queue_zone_sectors(disk->queue);
116 if (!blk_queue_is_zoned(disk->queue))
118 return (get_capacity(disk) + zone_sectors - 1) >> ilog2(zone_sectors);
120 EXPORT_SYMBOL_GPL(blkdev_nr_zones);
123 * blkdev_report_zones - Get zones information
124 * @bdev: Target block device
125 * @sector: Sector from which to report zones
126 * @nr_zones: Maximum number of zones to report
127 * @cb: Callback function called for each reported zone
128 * @data: Private data for the callback
131 * Get zone information starting from the zone containing @sector for at most
132 * @nr_zones, and call @cb for each zone reported by the device.
133 * To report all zones in a device starting from @sector, the BLK_ALL_ZONES
134 * constant can be passed to @nr_zones.
135 * Returns the number of zones reported by the device, or a negative errno
136 * value in case of failure.
138 * Note: The caller must use memalloc_noXX_save/restore() calls to control
139 * memory allocations done within this function.
141 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
142 unsigned int nr_zones, report_zones_cb cb, void *data)
144 struct gendisk *disk = bdev->bd_disk;
145 sector_t capacity = get_capacity(disk);
147 if (!blk_queue_is_zoned(bdev_get_queue(bdev)) ||
148 WARN_ON_ONCE(!disk->fops->report_zones))
151 if (!nr_zones || sector >= capacity)
154 return disk->fops->report_zones(disk, sector, nr_zones, cb, data);
156 EXPORT_SYMBOL_GPL(blkdev_report_zones);
158 static inline bool blkdev_allow_reset_all_zones(struct block_device *bdev,
162 if (!blk_queue_zone_resetall(bdev_get_queue(bdev)))
166 * REQ_OP_ZONE_RESET_ALL can be executed only if the number of sectors
167 * of the applicable zone range is the entire disk.
169 return !sector && nr_sectors == get_capacity(bdev->bd_disk);
173 * blkdev_zone_mgmt - Execute a zone management operation on a range of zones
174 * @bdev: Target block device
175 * @op: Operation to be performed on the zones
176 * @sector: Start sector of the first zone to operate on
177 * @nr_sectors: Number of sectors, should be at least the length of one zone and
178 * must be zone size aligned.
179 * @gfp_mask: Memory allocation flags (for bio_alloc)
182 * Perform the specified operation on the range of zones specified by
183 * @sector..@sector+@nr_sectors. Specifying the entire disk sector range
184 * is valid, but the specified range should not contain conventional zones.
185 * The operation to execute on each zone can be a zone reset, open, close
188 int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
189 sector_t sector, sector_t nr_sectors,
192 struct request_queue *q = bdev_get_queue(bdev);
193 sector_t zone_sectors = blk_queue_zone_sectors(q);
194 sector_t capacity = get_capacity(bdev->bd_disk);
195 sector_t end_sector = sector + nr_sectors;
196 struct bio *bio = NULL;
199 if (!blk_queue_is_zoned(q))
202 if (bdev_read_only(bdev))
205 if (!op_is_zone_mgmt(op))
208 if (end_sector <= sector || end_sector > capacity)
212 /* Check alignment (handle eventual smaller last zone) */
213 if (sector & (zone_sectors - 1))
216 if ((nr_sectors & (zone_sectors - 1)) && end_sector != capacity)
219 while (sector < end_sector) {
220 bio = blk_next_bio(bio, 0, gfp_mask);
221 bio_set_dev(bio, bdev);
224 * Special case for the zone reset operation that reset all
225 * zones, this is useful for applications like mkfs.
227 if (op == REQ_OP_ZONE_RESET &&
228 blkdev_allow_reset_all_zones(bdev, sector, nr_sectors)) {
229 bio->bi_opf = REQ_OP_ZONE_RESET_ALL;
233 bio->bi_opf = op | REQ_SYNC;
234 bio->bi_iter.bi_sector = sector;
235 sector += zone_sectors;
237 /* This may take a while, so be nice to others */
241 ret = submit_bio_wait(bio);
246 EXPORT_SYMBOL_GPL(blkdev_zone_mgmt);
248 struct zone_report_args {
249 struct blk_zone __user *zones;
252 static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx,
255 struct zone_report_args *args = data;
257 if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone)))
263 * BLKREPORTZONE ioctl processing.
264 * Called from blkdev_ioctl.
266 int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
267 unsigned int cmd, unsigned long arg)
269 void __user *argp = (void __user *)arg;
270 struct zone_report_args args;
271 struct request_queue *q;
272 struct blk_zone_report rep;
278 q = bdev_get_queue(bdev);
282 if (!blk_queue_is_zoned(q))
285 if (!capable(CAP_SYS_ADMIN))
288 if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report)))
294 args.zones = argp + sizeof(struct blk_zone_report);
295 ret = blkdev_report_zones(bdev, rep.sector, rep.nr_zones,
296 blkdev_copy_zone_to_user, &args);
301 if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report)))
307 * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing.
308 * Called from blkdev_ioctl.
310 int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
311 unsigned int cmd, unsigned long arg)
313 void __user *argp = (void __user *)arg;
314 struct request_queue *q;
315 struct blk_zone_range zrange;
321 q = bdev_get_queue(bdev);
325 if (!blk_queue_is_zoned(q))
328 if (!capable(CAP_SYS_ADMIN))
331 if (!(mode & FMODE_WRITE))
334 if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range)))
339 op = REQ_OP_ZONE_RESET;
342 op = REQ_OP_ZONE_OPEN;
345 op = REQ_OP_ZONE_CLOSE;
348 op = REQ_OP_ZONE_FINISH;
354 return blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors,
358 static inline unsigned long *blk_alloc_zone_bitmap(int node,
359 unsigned int nr_zones)
361 return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long),
365 void blk_queue_free_zone_bitmaps(struct request_queue *q)
367 kfree(q->conv_zones_bitmap);
368 q->conv_zones_bitmap = NULL;
369 kfree(q->seq_zones_wlock);
370 q->seq_zones_wlock = NULL;
373 struct blk_revalidate_zone_args {
374 struct gendisk *disk;
375 unsigned long *conv_zones_bitmap;
376 unsigned long *seq_zones_wlock;
377 unsigned int nr_zones;
378 sector_t zone_sectors;
383 * Helper function to check the validity of zones of a zoned block device.
385 static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx,
388 struct blk_revalidate_zone_args *args = data;
389 struct gendisk *disk = args->disk;
390 struct request_queue *q = disk->queue;
391 sector_t capacity = get_capacity(disk);
394 * All zones must have the same size, with the exception on an eventual
397 if (zone->start == 0) {
398 if (zone->len == 0 || !is_power_of_2(zone->len)) {
399 pr_warn("%s: Invalid zoned device with non power of two zone size (%llu)\n",
400 disk->disk_name, zone->len);
404 args->zone_sectors = zone->len;
405 args->nr_zones = (capacity + zone->len - 1) >> ilog2(zone->len);
406 } else if (zone->start + args->zone_sectors < capacity) {
407 if (zone->len != args->zone_sectors) {
408 pr_warn("%s: Invalid zoned device with non constant zone size\n",
413 if (zone->len > args->zone_sectors) {
414 pr_warn("%s: Invalid zoned device with larger last zone size\n",
420 /* Check for holes in the zone report */
421 if (zone->start != args->sector) {
422 pr_warn("%s: Zone gap at sectors %llu..%llu\n",
423 disk->disk_name, args->sector, zone->start);
427 /* Check zone type */
428 switch (zone->type) {
429 case BLK_ZONE_TYPE_CONVENTIONAL:
430 if (!args->conv_zones_bitmap) {
431 args->conv_zones_bitmap =
432 blk_alloc_zone_bitmap(q->node, args->nr_zones);
433 if (!args->conv_zones_bitmap)
436 set_bit(idx, args->conv_zones_bitmap);
438 case BLK_ZONE_TYPE_SEQWRITE_REQ:
439 case BLK_ZONE_TYPE_SEQWRITE_PREF:
440 if (!args->seq_zones_wlock) {
441 args->seq_zones_wlock =
442 blk_alloc_zone_bitmap(q->node, args->nr_zones);
443 if (!args->seq_zones_wlock)
448 pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n",
449 disk->disk_name, (int)zone->type, zone->start);
453 args->sector += zone->len;
458 * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps
461 * Helper function for low-level device drivers to (re) allocate and initialize
462 * a disk request queue zone bitmaps. This functions should normally be called
463 * within the disk ->revalidate method for blk-mq based drivers. For BIO based
464 * drivers only q->nr_zones needs to be updated so that the sysfs exposed value
467 int blk_revalidate_disk_zones(struct gendisk *disk)
469 struct request_queue *q = disk->queue;
470 struct blk_revalidate_zone_args args = {
473 unsigned int noio_flag;
476 if (WARN_ON_ONCE(!blk_queue_is_zoned(q)))
478 if (WARN_ON_ONCE(!queue_is_mq(q)))
482 * Ensure that all memory allocations in this context are done as if
483 * GFP_NOIO was specified.
485 noio_flag = memalloc_noio_save();
486 ret = disk->fops->report_zones(disk, 0, UINT_MAX,
487 blk_revalidate_zone_cb, &args);
488 memalloc_noio_restore(noio_flag);
491 * Install the new bitmaps and update nr_zones only once the queue is
492 * stopped and all I/Os are completed (i.e. a scheduler is not
493 * referencing the bitmaps).
495 blk_mq_freeze_queue(q);
497 blk_queue_chunk_sectors(q, args.zone_sectors);
498 q->nr_zones = args.nr_zones;
499 swap(q->seq_zones_wlock, args.seq_zones_wlock);
500 swap(q->conv_zones_bitmap, args.conv_zones_bitmap);
503 pr_warn("%s: failed to revalidate zones\n", disk->disk_name);
504 blk_queue_free_zone_bitmaps(q);
506 blk_mq_unfreeze_queue(q);
508 kfree(args.seq_zones_wlock);
509 kfree(args.conv_zones_bitmap);
512 EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones);