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);
56 * Return true if a request is a write requests that needs zone write locking.
58 bool blk_req_needs_zone_write_lock(struct request *rq)
60 if (!rq->q->seq_zones_wlock)
63 if (blk_rq_is_passthrough(rq))
67 case REQ_OP_WRITE_ZEROES:
69 return blk_rq_zone_is_seq(rq);
74 EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock);
76 bool blk_req_zone_write_trylock(struct request *rq)
78 unsigned int zno = blk_rq_zone_no(rq);
80 if (test_and_set_bit(zno, rq->q->seq_zones_wlock))
83 WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED);
84 rq->rq_flags |= RQF_ZONE_WRITE_LOCKED;
88 EXPORT_SYMBOL_GPL(blk_req_zone_write_trylock);
90 void __blk_req_zone_write_lock(struct request *rq)
92 if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq),
93 rq->q->seq_zones_wlock)))
96 WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED);
97 rq->rq_flags |= RQF_ZONE_WRITE_LOCKED;
99 EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock);
101 void __blk_req_zone_write_unlock(struct request *rq)
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));
108 EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock);
111 * blkdev_nr_zones - Get number of zones
112 * @disk: Target gendisk
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.
117 unsigned int blkdev_nr_zones(struct gendisk *disk)
119 sector_t zone_sectors = blk_queue_zone_sectors(disk->queue);
121 if (!blk_queue_is_zoned(disk->queue))
123 return (get_capacity(disk) + zone_sectors - 1) >> ilog2(zone_sectors);
125 EXPORT_SYMBOL_GPL(blkdev_nr_zones);
128 * blkdev_report_zones - Get zones information
129 * @bdev: Target block device
130 * @sector: Sector from which to report zones
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
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.
143 * Note: The caller must use memalloc_noXX_save/restore() calls to control
144 * memory allocations done within this function.
146 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
147 unsigned int nr_zones, report_zones_cb cb, void *data)
149 struct gendisk *disk = bdev->bd_disk;
150 sector_t capacity = get_capacity(disk);
152 if (!blk_queue_is_zoned(bdev_get_queue(bdev)) ||
153 WARN_ON_ONCE(!disk->fops->report_zones))
156 if (!nr_zones || sector >= capacity)
159 return disk->fops->report_zones(disk, sector, nr_zones, cb, data);
161 EXPORT_SYMBOL_GPL(blkdev_report_zones);
163 static inline unsigned long *blk_alloc_zone_bitmap(int node,
164 unsigned int nr_zones)
166 return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long),
170 static int blk_zone_need_reset_cb(struct blk_zone *zone, unsigned int idx,
174 * For an all-zones reset, ignore conventional, empty, read-only
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:
184 set_bit(idx, (unsigned long *)data);
189 static int blkdev_zone_reset_all_emulated(struct block_device *bdev,
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;
200 need_reset = blk_alloc_zone_bitmap(q->node, q->nr_zones);
204 ret = bdev->bd_disk->fops->report_zones(bdev->bd_disk, 0,
205 q->nr_zones, blk_zone_need_reset_cb,
208 goto out_free_need_reset;
211 while (sector < capacity) {
212 if (!test_bit(blk_queue_zone_no(q, sector), need_reset)) {
213 sector += zone_sectors;
217 bio = blk_next_bio(bio, bdev, 0, REQ_OP_ZONE_RESET | REQ_SYNC,
219 bio->bi_iter.bi_sector = sector;
220 sector += zone_sectors;
222 /* This may take a while, so be nice to others */
227 ret = submit_bio_wait(bio);
236 static int blkdev_zone_reset_all(struct block_device *bdev, gfp_t gfp_mask)
240 bio_init(&bio, bdev, NULL, 0, REQ_OP_ZONE_RESET_ALL | REQ_SYNC);
241 return submit_bio_wait(&bio);
245 * blkdev_zone_mgmt - Execute a zone management operation on a range of zones
246 * @bdev: Target block device
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.
251 * @gfp_mask: Memory allocation flags (for bio_alloc)
254 * Perform the specified operation on the range of zones specified by
255 * @sector..@sector+@nr_sectors. Specifying the entire disk sector range
256 * is valid, but the specified range should not contain conventional zones.
257 * The operation to execute on each zone can be a zone reset, open, close
260 int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
261 sector_t sector, sector_t nr_sectors,
264 struct request_queue *q = bdev_get_queue(bdev);
265 sector_t zone_sectors = blk_queue_zone_sectors(q);
266 sector_t capacity = get_capacity(bdev->bd_disk);
267 sector_t end_sector = sector + nr_sectors;
268 struct bio *bio = NULL;
271 if (!blk_queue_is_zoned(q))
274 if (bdev_read_only(bdev))
277 if (!op_is_zone_mgmt(op))
280 if (end_sector <= sector || end_sector > capacity)
284 /* Check alignment (handle eventual smaller last zone) */
285 if (sector & (zone_sectors - 1))
288 if ((nr_sectors & (zone_sectors - 1)) && end_sector != capacity)
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.
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);
303 while (sector < end_sector) {
304 bio = blk_next_bio(bio, bdev, 0, op | REQ_SYNC, gfp_mask);
305 bio->bi_iter.bi_sector = sector;
306 sector += zone_sectors;
308 /* This may take a while, so be nice to others */
312 ret = submit_bio_wait(bio);
317 EXPORT_SYMBOL_GPL(blkdev_zone_mgmt);
319 struct zone_report_args {
320 struct blk_zone __user *zones;
323 static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx,
326 struct zone_report_args *args = data;
328 if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone)))
334 * BLKREPORTZONE ioctl processing.
335 * Called from blkdev_ioctl.
337 int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
338 unsigned int cmd, unsigned long arg)
340 void __user *argp = (void __user *)arg;
341 struct zone_report_args args;
342 struct request_queue *q;
343 struct blk_zone_report rep;
349 q = bdev_get_queue(bdev);
353 if (!blk_queue_is_zoned(q))
356 if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report)))
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);
369 rep.flags = BLK_ZONE_REP_CAPACITY;
370 if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report)))
375 static int blkdev_truncate_zone_range(struct block_device *bdev, fmode_t mode,
376 const struct blk_zone_range *zrange)
380 if (zrange->sector + zrange->nr_sectors <= zrange->sector ||
381 zrange->sector + zrange->nr_sectors > get_capacity(bdev->bd_disk))
385 start = zrange->sector << SECTOR_SHIFT;
386 end = ((zrange->sector + zrange->nr_sectors) << SECTOR_SHIFT) - 1;
388 return truncate_bdev_range(bdev, mode, start, end);
392 * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing.
393 * Called from blkdev_ioctl.
395 int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
396 unsigned int cmd, unsigned long arg)
398 void __user *argp = (void __user *)arg;
399 struct request_queue *q;
400 struct blk_zone_range zrange;
407 q = bdev_get_queue(bdev);
411 if (!blk_queue_is_zoned(q))
414 if (!(mode & FMODE_WRITE))
417 if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range)))
422 op = REQ_OP_ZONE_RESET;
424 /* Invalidate the page cache, including dirty pages. */
425 filemap_invalidate_lock(bdev->bd_inode->i_mapping);
426 ret = blkdev_truncate_zone_range(bdev, mode, &zrange);
431 op = REQ_OP_ZONE_OPEN;
434 op = REQ_OP_ZONE_CLOSE;
437 op = REQ_OP_ZONE_FINISH;
443 ret = blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors,
447 if (cmd == BLKRESETZONE)
448 filemap_invalidate_unlock(bdev->bd_inode->i_mapping);
453 void blk_queue_free_zone_bitmaps(struct request_queue *q)
455 kfree(q->conv_zones_bitmap);
456 q->conv_zones_bitmap = NULL;
457 kfree(q->seq_zones_wlock);
458 q->seq_zones_wlock = NULL;
461 struct blk_revalidate_zone_args {
462 struct gendisk *disk;
463 unsigned long *conv_zones_bitmap;
464 unsigned long *seq_zones_wlock;
465 unsigned int nr_zones;
466 sector_t zone_sectors;
471 * Helper function to check the validity of zones of a zoned block device.
473 static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx,
476 struct blk_revalidate_zone_args *args = data;
477 struct gendisk *disk = args->disk;
478 struct request_queue *q = disk->queue;
479 sector_t capacity = get_capacity(disk);
482 * All zones must have the same size, with the exception on an eventual
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);
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",
501 if (zone->len > args->zone_sectors) {
502 pr_warn("%s: Invalid zoned device with larger last zone size\n",
508 /* Check for holes in the zone report */
509 if (zone->start != args->sector) {
510 pr_warn("%s: Zone gap at sectors %llu..%llu\n",
511 disk->disk_name, args->sector, zone->start);
515 /* Check zone type */
516 switch (zone->type) {
517 case BLK_ZONE_TYPE_CONVENTIONAL:
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)
524 set_bit(idx, args->conv_zones_bitmap);
526 case BLK_ZONE_TYPE_SEQWRITE_REQ:
527 case BLK_ZONE_TYPE_SEQWRITE_PREF:
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)
536 pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n",
537 disk->disk_name, (int)zone->type, zone->start);
541 args->sector += zone->len;
546 * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps
548 * @update_driver_data: Callback to update driver data on the frozen disk
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
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
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
559 int blk_revalidate_disk_zones(struct gendisk *disk,
560 void (*update_driver_data)(struct gendisk *disk))
562 struct request_queue *q = disk->queue;
563 struct blk_revalidate_zone_args args = {
566 unsigned int noio_flag;
569 if (WARN_ON_ONCE(!blk_queue_is_zoned(q)))
571 if (WARN_ON_ONCE(!queue_is_mq(q)))
574 if (!get_capacity(disk))
578 * Ensure that all memory allocations in this context are done as if
579 * GFP_NOIO was specified.
581 noio_flag = memalloc_noio_save();
582 ret = disk->fops->report_zones(disk, 0, UINT_MAX,
583 blk_revalidate_zone_cb, &args);
585 pr_warn("%s: No zones reported\n", disk->disk_name);
588 memalloc_noio_restore(noio_flag);
591 * If zones where reported, make sure that the entire disk capacity
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);
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).
605 blk_mq_freeze_queue(q);
607 blk_queue_chunk_sectors(q, args.zone_sectors);
608 q->nr_zones = args.nr_zones;
609 swap(q->seq_zones_wlock, args.seq_zones_wlock);
610 swap(q->conv_zones_bitmap, args.conv_zones_bitmap);
611 if (update_driver_data)
612 update_driver_data(disk);
615 pr_warn("%s: failed to revalidate zones\n", disk->disk_name);
616 blk_queue_free_zone_bitmaps(q);
618 blk_mq_unfreeze_queue(q);
620 kfree(args.seq_zones_wlock);
621 kfree(args.conv_zones_bitmap);
624 EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones);
626 void blk_queue_clear_zone_settings(struct request_queue *q)
628 blk_mq_freeze_queue(q);
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;
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;
640 blk_mq_unfreeze_queue(q);