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 static inline sector_t blk_zone_start(struct request_queue *q,
26 sector_t zone_mask = blk_queue_zone_sectors(q) - 1;
28 return sector & ~zone_mask;
32 * Return true if a request is a write requests that needs zone write locking.
34 bool blk_req_needs_zone_write_lock(struct request *rq)
36 if (!rq->q->seq_zones_wlock)
39 if (blk_rq_is_passthrough(rq))
43 case REQ_OP_WRITE_ZEROES:
44 case REQ_OP_WRITE_SAME:
46 return blk_rq_zone_is_seq(rq);
51 EXPORT_SYMBOL_GPL(blk_req_needs_zone_write_lock);
53 void __blk_req_zone_write_lock(struct request *rq)
55 if (WARN_ON_ONCE(test_and_set_bit(blk_rq_zone_no(rq),
56 rq->q->seq_zones_wlock)))
59 WARN_ON_ONCE(rq->rq_flags & RQF_ZONE_WRITE_LOCKED);
60 rq->rq_flags |= RQF_ZONE_WRITE_LOCKED;
62 EXPORT_SYMBOL_GPL(__blk_req_zone_write_lock);
64 void __blk_req_zone_write_unlock(struct request *rq)
66 rq->rq_flags &= ~RQF_ZONE_WRITE_LOCKED;
67 if (rq->q->seq_zones_wlock)
68 WARN_ON_ONCE(!test_and_clear_bit(blk_rq_zone_no(rq),
69 rq->q->seq_zones_wlock));
71 EXPORT_SYMBOL_GPL(__blk_req_zone_write_unlock);
74 * blkdev_nr_zones - Get number of zones
75 * @disk: Target gendisk
77 * Return the total number of zones of a zoned block device. For a block
78 * device without zone capabilities, the number of zones is always 0.
80 unsigned int blkdev_nr_zones(struct gendisk *disk)
82 sector_t zone_sectors = blk_queue_zone_sectors(disk->queue);
84 if (!blk_queue_is_zoned(disk->queue))
86 return (get_capacity(disk) + zone_sectors - 1) >> ilog2(zone_sectors);
88 EXPORT_SYMBOL_GPL(blkdev_nr_zones);
91 * blkdev_report_zones - Get zones information
92 * @bdev: Target block device
93 * @sector: Sector from which to report zones
94 * @nr_zones: Maximum number of zones to report
95 * @cb: Callback function called for each reported zone
96 * @data: Private data for the callback
99 * Get zone information starting from the zone containing @sector for at most
100 * @nr_zones, and call @cb for each zone reported by the device.
101 * To report all zones in a device starting from @sector, the BLK_ALL_ZONES
102 * constant can be passed to @nr_zones.
103 * Returns the number of zones reported by the device, or a negative errno
104 * value in case of failure.
106 * Note: The caller must use memalloc_noXX_save/restore() calls to control
107 * memory allocations done within this function.
109 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
110 unsigned int nr_zones, report_zones_cb cb, void *data)
112 struct gendisk *disk = bdev->bd_disk;
113 sector_t capacity = get_capacity(disk);
115 if (!blk_queue_is_zoned(bdev_get_queue(bdev)) ||
116 WARN_ON_ONCE(!disk->fops->report_zones))
119 if (!nr_zones || sector >= capacity)
122 return disk->fops->report_zones(disk, sector, nr_zones, cb, data);
124 EXPORT_SYMBOL_GPL(blkdev_report_zones);
126 static inline bool blkdev_allow_reset_all_zones(struct block_device *bdev,
130 if (!blk_queue_zone_resetall(bdev_get_queue(bdev)))
134 * REQ_OP_ZONE_RESET_ALL can be executed only if the number of sectors
135 * of the applicable zone range is the entire disk.
137 return !sector && nr_sectors == get_capacity(bdev->bd_disk);
141 * blkdev_zone_mgmt - Execute a zone management operation on a range of zones
142 * @bdev: Target block device
143 * @op: Operation to be performed on the zones
144 * @sector: Start sector of the first zone to operate on
145 * @nr_sectors: Number of sectors, should be at least the length of one zone and
146 * must be zone size aligned.
147 * @gfp_mask: Memory allocation flags (for bio_alloc)
150 * Perform the specified operation on the range of zones specified by
151 * @sector..@sector+@nr_sectors. Specifying the entire disk sector range
152 * is valid, but the specified range should not contain conventional zones.
153 * The operation to execute on each zone can be a zone reset, open, close
156 int blkdev_zone_mgmt(struct block_device *bdev, enum req_opf op,
157 sector_t sector, sector_t nr_sectors,
160 struct request_queue *q = bdev_get_queue(bdev);
161 sector_t zone_sectors = blk_queue_zone_sectors(q);
162 sector_t capacity = get_capacity(bdev->bd_disk);
163 sector_t end_sector = sector + nr_sectors;
164 struct bio *bio = NULL;
167 if (!blk_queue_is_zoned(q))
170 if (bdev_read_only(bdev))
173 if (!op_is_zone_mgmt(op))
176 if (!nr_sectors || end_sector > capacity)
180 /* Check alignment (handle eventual smaller last zone) */
181 if (sector & (zone_sectors - 1))
184 if ((nr_sectors & (zone_sectors - 1)) && end_sector != capacity)
187 while (sector < end_sector) {
188 bio = blk_next_bio(bio, 0, gfp_mask);
189 bio_set_dev(bio, bdev);
192 * Special case for the zone reset operation that reset all
193 * zones, this is useful for applications like mkfs.
195 if (op == REQ_OP_ZONE_RESET &&
196 blkdev_allow_reset_all_zones(bdev, sector, nr_sectors)) {
197 bio->bi_opf = REQ_OP_ZONE_RESET_ALL;
201 bio->bi_opf = op | REQ_SYNC;
202 bio->bi_iter.bi_sector = sector;
203 sector += zone_sectors;
205 /* This may take a while, so be nice to others */
209 ret = submit_bio_wait(bio);
214 EXPORT_SYMBOL_GPL(blkdev_zone_mgmt);
216 struct zone_report_args {
217 struct blk_zone __user *zones;
220 static int blkdev_copy_zone_to_user(struct blk_zone *zone, unsigned int idx,
223 struct zone_report_args *args = data;
225 if (copy_to_user(&args->zones[idx], zone, sizeof(struct blk_zone)))
231 * BLKREPORTZONE ioctl processing.
232 * Called from blkdev_ioctl.
234 int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode,
235 unsigned int cmd, unsigned long arg)
237 void __user *argp = (void __user *)arg;
238 struct zone_report_args args;
239 struct request_queue *q;
240 struct blk_zone_report rep;
246 q = bdev_get_queue(bdev);
250 if (!blk_queue_is_zoned(q))
253 if (!capable(CAP_SYS_ADMIN))
256 if (copy_from_user(&rep, argp, sizeof(struct blk_zone_report)))
262 args.zones = argp + sizeof(struct blk_zone_report);
263 ret = blkdev_report_zones(bdev, rep.sector, rep.nr_zones,
264 blkdev_copy_zone_to_user, &args);
269 if (copy_to_user(argp, &rep, sizeof(struct blk_zone_report)))
275 * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing.
276 * Called from blkdev_ioctl.
278 int blkdev_zone_mgmt_ioctl(struct block_device *bdev, fmode_t mode,
279 unsigned int cmd, unsigned long arg)
281 void __user *argp = (void __user *)arg;
282 struct request_queue *q;
283 struct blk_zone_range zrange;
289 q = bdev_get_queue(bdev);
293 if (!blk_queue_is_zoned(q))
296 if (!capable(CAP_SYS_ADMIN))
299 if (!(mode & FMODE_WRITE))
302 if (copy_from_user(&zrange, argp, sizeof(struct blk_zone_range)))
307 op = REQ_OP_ZONE_RESET;
310 op = REQ_OP_ZONE_OPEN;
313 op = REQ_OP_ZONE_CLOSE;
316 op = REQ_OP_ZONE_FINISH;
322 return blkdev_zone_mgmt(bdev, op, zrange.sector, zrange.nr_sectors,
326 static inline unsigned long *blk_alloc_zone_bitmap(int node,
327 unsigned int nr_zones)
329 return kcalloc_node(BITS_TO_LONGS(nr_zones), sizeof(unsigned long),
333 void blk_queue_free_zone_bitmaps(struct request_queue *q)
335 kfree(q->conv_zones_bitmap);
336 q->conv_zones_bitmap = NULL;
337 kfree(q->seq_zones_wlock);
338 q->seq_zones_wlock = NULL;
341 struct blk_revalidate_zone_args {
342 struct gendisk *disk;
343 unsigned long *conv_zones_bitmap;
344 unsigned long *seq_zones_wlock;
345 unsigned int nr_zones;
346 sector_t zone_sectors;
351 * Helper function to check the validity of zones of a zoned block device.
353 static int blk_revalidate_zone_cb(struct blk_zone *zone, unsigned int idx,
356 struct blk_revalidate_zone_args *args = data;
357 struct gendisk *disk = args->disk;
358 struct request_queue *q = disk->queue;
359 sector_t capacity = get_capacity(disk);
362 * All zones must have the same size, with the exception on an eventual
365 if (zone->start == 0) {
366 if (zone->len == 0 || !is_power_of_2(zone->len)) {
367 pr_warn("%s: Invalid zoned device with non power of two zone size (%llu)\n",
368 disk->disk_name, zone->len);
372 args->zone_sectors = zone->len;
373 args->nr_zones = (capacity + zone->len - 1) >> ilog2(zone->len);
374 } else if (zone->start + args->zone_sectors < capacity) {
375 if (zone->len != args->zone_sectors) {
376 pr_warn("%s: Invalid zoned device with non constant zone size\n",
381 if (zone->len > args->zone_sectors) {
382 pr_warn("%s: Invalid zoned device with larger last zone size\n",
388 /* Check for holes in the zone report */
389 if (zone->start != args->sector) {
390 pr_warn("%s: Zone gap at sectors %llu..%llu\n",
391 disk->disk_name, args->sector, zone->start);
395 /* Check zone type */
396 switch (zone->type) {
397 case BLK_ZONE_TYPE_CONVENTIONAL:
398 if (!args->conv_zones_bitmap) {
399 args->conv_zones_bitmap =
400 blk_alloc_zone_bitmap(q->node, args->nr_zones);
401 if (!args->conv_zones_bitmap)
404 set_bit(idx, args->conv_zones_bitmap);
406 case BLK_ZONE_TYPE_SEQWRITE_REQ:
407 case BLK_ZONE_TYPE_SEQWRITE_PREF:
408 if (!args->seq_zones_wlock) {
409 args->seq_zones_wlock =
410 blk_alloc_zone_bitmap(q->node, args->nr_zones);
411 if (!args->seq_zones_wlock)
416 pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n",
417 disk->disk_name, (int)zone->type, zone->start);
421 args->sector += zone->len;
426 * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps
429 * Helper function for low-level device drivers to (re) allocate and initialize
430 * a disk request queue zone bitmaps. This functions should normally be called
431 * within the disk ->revalidate method for blk-mq based drivers. For BIO based
432 * drivers only q->nr_zones needs to be updated so that the sysfs exposed value
435 int blk_revalidate_disk_zones(struct gendisk *disk)
437 struct request_queue *q = disk->queue;
438 struct blk_revalidate_zone_args args = {
441 unsigned int noio_flag;
444 if (WARN_ON_ONCE(!blk_queue_is_zoned(q)))
446 if (WARN_ON_ONCE(!queue_is_mq(q)))
450 * Ensure that all memory allocations in this context are done as if
451 * GFP_NOIO was specified.
453 noio_flag = memalloc_noio_save();
454 ret = disk->fops->report_zones(disk, 0, UINT_MAX,
455 blk_revalidate_zone_cb, &args);
456 memalloc_noio_restore(noio_flag);
459 * Install the new bitmaps and update nr_zones only once the queue is
460 * stopped and all I/Os are completed (i.e. a scheduler is not
461 * referencing the bitmaps).
463 blk_mq_freeze_queue(q);
465 blk_queue_chunk_sectors(q, args.zone_sectors);
466 q->nr_zones = args.nr_zones;
467 swap(q->seq_zones_wlock, args.seq_zones_wlock);
468 swap(q->conv_zones_bitmap, args.conv_zones_bitmap);
471 pr_warn("%s: failed to revalidate zones\n", disk->disk_name);
472 blk_queue_free_zone_bitmaps(q);
474 blk_mq_unfreeze_queue(q);
476 kfree(args.seq_zones_wlock);
477 kfree(args.conv_zones_bitmap);
480 EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones);