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
);
73 static inline unsigned int __blkdev_nr_zones(struct request_queue
*q
,
76 sector_t zone_sectors
= blk_queue_zone_sectors(q
);
78 return (nr_sectors
+ zone_sectors
- 1) >> ilog2(zone_sectors
);
82 * blkdev_nr_zones - Get number of zones
83 * @bdev: Target block device
86 * Return the total number of zones of a zoned block device.
87 * For a regular block device, the number of zones is always 0.
89 unsigned int blkdev_nr_zones(struct block_device
*bdev
)
91 struct request_queue
*q
= bdev_get_queue(bdev
);
93 if (!blk_queue_is_zoned(q
))
96 return __blkdev_nr_zones(q
, get_capacity(bdev
->bd_disk
));
98 EXPORT_SYMBOL_GPL(blkdev_nr_zones
);
101 * blkdev_report_zones - Get zones information
102 * @bdev: Target block device
103 * @sector: Sector from which to report zones
104 * @nr_zones: Maximum number of zones to report
105 * @cb: Callback function called for each reported zone
106 * @data: Private data for the callback
109 * Get zone information starting from the zone containing @sector for at most
110 * @nr_zones, and call @cb for each zone reported by the device.
111 * To report all zones in a device starting from @sector, the BLK_ALL_ZONES
112 * constant can be passed to @nr_zones.
113 * Returns the number of zones reported by the device, or a negative errno
114 * value in case of failure.
116 * Note: The caller must use memalloc_noXX_save/restore() calls to control
117 * memory allocations done within this function.
119 int blkdev_report_zones(struct block_device
*bdev
, sector_t sector
,
120 unsigned int nr_zones
, report_zones_cb cb
, void *data
)
122 struct gendisk
*disk
= bdev
->bd_disk
;
123 sector_t capacity
= get_capacity(disk
);
125 if (!blk_queue_is_zoned(bdev_get_queue(bdev
)) ||
126 WARN_ON_ONCE(!disk
->fops
->report_zones
))
129 if (!nr_zones
|| sector
>= capacity
)
132 return disk
->fops
->report_zones(disk
, sector
, nr_zones
, cb
, data
);
134 EXPORT_SYMBOL_GPL(blkdev_report_zones
);
136 static inline bool blkdev_allow_reset_all_zones(struct block_device
*bdev
,
140 if (!blk_queue_zone_resetall(bdev_get_queue(bdev
)))
144 * REQ_OP_ZONE_RESET_ALL can be executed only if the number of sectors
145 * of the applicable zone range is the entire disk.
147 return !sector
&& nr_sectors
== get_capacity(bdev
->bd_disk
);
151 * blkdev_zone_mgmt - Execute a zone management operation on a range of zones
152 * @bdev: Target block device
153 * @op: Operation to be performed on the zones
154 * @sector: Start sector of the first zone to operate on
155 * @nr_sectors: Number of sectors, should be at least the length of one zone and
156 * must be zone size aligned.
157 * @gfp_mask: Memory allocation flags (for bio_alloc)
160 * Perform the specified operation on the range of zones specified by
161 * @sector..@sector+@nr_sectors. Specifying the entire disk sector range
162 * is valid, but the specified range should not contain conventional zones.
163 * The operation to execute on each zone can be a zone reset, open, close
166 int blkdev_zone_mgmt(struct block_device
*bdev
, enum req_opf op
,
167 sector_t sector
, sector_t nr_sectors
,
170 struct request_queue
*q
= bdev_get_queue(bdev
);
171 sector_t zone_sectors
= blk_queue_zone_sectors(q
);
172 sector_t capacity
= get_capacity(bdev
->bd_disk
);
173 sector_t end_sector
= sector
+ nr_sectors
;
174 struct bio
*bio
= NULL
;
177 if (!blk_queue_is_zoned(q
))
180 if (bdev_read_only(bdev
))
183 if (!op_is_zone_mgmt(op
))
186 if (!nr_sectors
|| end_sector
> capacity
)
190 /* Check alignment (handle eventual smaller last zone) */
191 if (sector
& (zone_sectors
- 1))
194 if ((nr_sectors
& (zone_sectors
- 1)) && end_sector
!= capacity
)
197 while (sector
< end_sector
) {
198 bio
= blk_next_bio(bio
, 0, gfp_mask
);
199 bio_set_dev(bio
, bdev
);
202 * Special case for the zone reset operation that reset all
203 * zones, this is useful for applications like mkfs.
205 if (op
== REQ_OP_ZONE_RESET
&&
206 blkdev_allow_reset_all_zones(bdev
, sector
, nr_sectors
)) {
207 bio
->bi_opf
= REQ_OP_ZONE_RESET_ALL
;
212 bio
->bi_iter
.bi_sector
= sector
;
213 sector
+= zone_sectors
;
215 /* This may take a while, so be nice to others */
219 ret
= submit_bio_wait(bio
);
224 EXPORT_SYMBOL_GPL(blkdev_zone_mgmt
);
226 struct zone_report_args
{
227 struct blk_zone __user
*zones
;
230 static int blkdev_copy_zone_to_user(struct blk_zone
*zone
, unsigned int idx
,
233 struct zone_report_args
*args
= data
;
235 if (copy_to_user(&args
->zones
[idx
], zone
, sizeof(struct blk_zone
)))
241 * BLKREPORTZONE ioctl processing.
242 * Called from blkdev_ioctl.
244 int blkdev_report_zones_ioctl(struct block_device
*bdev
, fmode_t mode
,
245 unsigned int cmd
, unsigned long arg
)
247 void __user
*argp
= (void __user
*)arg
;
248 struct zone_report_args args
;
249 struct request_queue
*q
;
250 struct blk_zone_report rep
;
256 q
= bdev_get_queue(bdev
);
260 if (!blk_queue_is_zoned(q
))
263 if (!capable(CAP_SYS_ADMIN
))
266 if (copy_from_user(&rep
, argp
, sizeof(struct blk_zone_report
)))
272 args
.zones
= argp
+ sizeof(struct blk_zone_report
);
273 ret
= blkdev_report_zones(bdev
, rep
.sector
, rep
.nr_zones
,
274 blkdev_copy_zone_to_user
, &args
);
279 if (copy_to_user(argp
, &rep
, sizeof(struct blk_zone_report
)))
285 * BLKRESETZONE, BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE ioctl processing.
286 * Called from blkdev_ioctl.
288 int blkdev_zone_mgmt_ioctl(struct block_device
*bdev
, fmode_t mode
,
289 unsigned int cmd
, unsigned long arg
)
291 void __user
*argp
= (void __user
*)arg
;
292 struct request_queue
*q
;
293 struct blk_zone_range zrange
;
299 q
= bdev_get_queue(bdev
);
303 if (!blk_queue_is_zoned(q
))
306 if (!capable(CAP_SYS_ADMIN
))
309 if (!(mode
& FMODE_WRITE
))
312 if (copy_from_user(&zrange
, argp
, sizeof(struct blk_zone_range
)))
317 op
= REQ_OP_ZONE_RESET
;
320 op
= REQ_OP_ZONE_OPEN
;
323 op
= REQ_OP_ZONE_CLOSE
;
326 op
= REQ_OP_ZONE_FINISH
;
332 return blkdev_zone_mgmt(bdev
, op
, zrange
.sector
, zrange
.nr_sectors
,
336 static inline unsigned long *blk_alloc_zone_bitmap(int node
,
337 unsigned int nr_zones
)
339 return kcalloc_node(BITS_TO_LONGS(nr_zones
), sizeof(unsigned long),
343 void blk_queue_free_zone_bitmaps(struct request_queue
*q
)
345 kfree(q
->seq_zones_bitmap
);
346 q
->seq_zones_bitmap
= NULL
;
347 kfree(q
->seq_zones_wlock
);
348 q
->seq_zones_wlock
= NULL
;
351 struct blk_revalidate_zone_args
{
352 struct gendisk
*disk
;
353 unsigned long *seq_zones_bitmap
;
354 unsigned long *seq_zones_wlock
;
359 * Helper function to check the validity of zones of a zoned block device.
361 static int blk_revalidate_zone_cb(struct blk_zone
*zone
, unsigned int idx
,
364 struct blk_revalidate_zone_args
*args
= data
;
365 struct gendisk
*disk
= args
->disk
;
366 struct request_queue
*q
= disk
->queue
;
367 sector_t zone_sectors
= blk_queue_zone_sectors(q
);
368 sector_t capacity
= get_capacity(disk
);
371 * All zones must have the same size, with the exception on an eventual
374 if (zone
->start
+ zone_sectors
< capacity
&&
375 zone
->len
!= zone_sectors
) {
376 pr_warn("%s: Invalid zoned device with non constant zone size\n",
381 if (zone
->start
+ zone
->len
>= capacity
&&
382 zone
->len
> zone_sectors
) {
383 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 case BLK_ZONE_TYPE_SEQWRITE_REQ
:
399 case BLK_ZONE_TYPE_SEQWRITE_PREF
:
402 pr_warn("%s: Invalid zone type 0x%x at sectors %llu\n",
403 disk
->disk_name
, (int)zone
->type
, zone
->start
);
407 if (zone
->type
!= BLK_ZONE_TYPE_CONVENTIONAL
)
408 set_bit(idx
, args
->seq_zones_bitmap
);
410 args
->sector
+= zone
->len
;
414 static int blk_update_zone_info(struct gendisk
*disk
, unsigned int nr_zones
,
415 struct blk_revalidate_zone_args
*args
)
418 * Ensure that all memory allocations in this context are done as
419 * if GFP_NOIO was specified.
421 unsigned int noio_flag
= memalloc_noio_save();
422 struct request_queue
*q
= disk
->queue
;
425 args
->seq_zones_wlock
= blk_alloc_zone_bitmap(q
->node
, nr_zones
);
426 if (!args
->seq_zones_wlock
)
428 args
->seq_zones_bitmap
= blk_alloc_zone_bitmap(q
->node
, nr_zones
);
429 if (!args
->seq_zones_bitmap
)
432 ret
= disk
->fops
->report_zones(disk
, 0, nr_zones
,
433 blk_revalidate_zone_cb
, args
);
434 memalloc_noio_restore(noio_flag
);
439 * blk_revalidate_disk_zones - (re)allocate and initialize zone bitmaps
442 * Helper function for low-level device drivers to (re) allocate and initialize
443 * a disk request queue zone bitmaps. This functions should normally be called
444 * within the disk ->revalidate method. For BIO based queues, no zone bitmap
447 int blk_revalidate_disk_zones(struct gendisk
*disk
)
449 struct request_queue
*q
= disk
->queue
;
450 unsigned int nr_zones
= __blkdev_nr_zones(q
, get_capacity(disk
));
451 struct blk_revalidate_zone_args args
= { .disk
= disk
};
454 if (WARN_ON_ONCE(!blk_queue_is_zoned(q
)))
458 * BIO based queues do not use a scheduler so only q->nr_zones
459 * needs to be updated so that the sysfs exposed value is correct.
461 if (!queue_is_mq(q
)) {
462 q
->nr_zones
= nr_zones
;
467 ret
= blk_update_zone_info(disk
, nr_zones
, &args
);
470 * Install the new bitmaps, making sure the queue is stopped and
471 * all I/Os are completed (i.e. a scheduler is not referencing the
474 blk_mq_freeze_queue(q
);
476 q
->nr_zones
= nr_zones
;
477 swap(q
->seq_zones_wlock
, args
.seq_zones_wlock
);
478 swap(q
->seq_zones_bitmap
, args
.seq_zones_bitmap
);
481 pr_warn("%s: failed to revalidate zones\n", disk
->disk_name
);
482 blk_queue_free_zone_bitmaps(q
);
484 blk_mq_unfreeze_queue(q
);
486 kfree(args
.seq_zones_wlock
);
487 kfree(args
.seq_zones_bitmap
);
490 EXPORT_SYMBOL_GPL(blk_revalidate_disk_zones
);