1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/vmalloc.h>
5 #define CREATE_TRACE_POINTS
6 #include "null_blk_trace.h"
8 /* zone_size in MBs to sectors. */
9 #define ZONE_SIZE_SHIFT 11
11 static inline unsigned int null_zone_no(struct nullb_device *dev, sector_t sect)
13 return sect >> ilog2(dev->zone_size_sects);
16 int null_init_zoned_dev(struct nullb_device *dev, struct request_queue *q)
18 sector_t dev_size = (sector_t)dev->size * 1024 * 1024;
22 if (!is_power_of_2(dev->zone_size)) {
23 pr_err("zone_size must be power-of-two\n");
26 if (dev->zone_size > dev->size) {
27 pr_err("Zone size larger than device capacity\n");
31 if (!dev->zone_capacity)
32 dev->zone_capacity = dev->zone_size;
34 if (dev->zone_capacity > dev->zone_size) {
35 pr_err("null_blk: zone capacity (%lu MB) larger than zone size (%lu MB)\n",
36 dev->zone_capacity, dev->zone_size);
40 dev->zone_size_sects = dev->zone_size << ZONE_SIZE_SHIFT;
41 dev->nr_zones = dev_size >>
42 (SECTOR_SHIFT + ilog2(dev->zone_size_sects));
43 dev->zones = kvmalloc_array(dev->nr_zones, sizeof(struct blk_zone),
44 GFP_KERNEL | __GFP_ZERO);
48 spin_lock_init(&dev->zone_lock);
49 if (dev->zone_nr_conv >= dev->nr_zones) {
50 dev->zone_nr_conv = dev->nr_zones - 1;
51 pr_info("changed the number of conventional zones to %u",
55 /* Max active zones has to be < nbr of seq zones in order to be enforceable */
56 if (dev->zone_max_active >= dev->nr_zones - dev->zone_nr_conv) {
57 dev->zone_max_active = 0;
58 pr_info("zone_max_active limit disabled, limit >= zone count\n");
61 /* Max open zones has to be <= max active zones */
62 if (dev->zone_max_active && dev->zone_max_open > dev->zone_max_active) {
63 dev->zone_max_open = dev->zone_max_active;
64 pr_info("changed the maximum number of open zones to %u\n",
66 } else if (dev->zone_max_open >= dev->nr_zones - dev->zone_nr_conv) {
67 dev->zone_max_open = 0;
68 pr_info("zone_max_open limit disabled, limit >= zone count\n");
71 for (i = 0; i < dev->zone_nr_conv; i++) {
72 struct blk_zone *zone = &dev->zones[i];
75 zone->len = dev->zone_size_sects;
76 zone->capacity = zone->len;
77 zone->wp = zone->start + zone->len;
78 zone->type = BLK_ZONE_TYPE_CONVENTIONAL;
79 zone->cond = BLK_ZONE_COND_NOT_WP;
81 sector += dev->zone_size_sects;
84 for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
85 struct blk_zone *zone = &dev->zones[i];
87 zone->start = zone->wp = sector;
88 zone->len = dev->zone_size_sects;
89 zone->capacity = dev->zone_capacity << ZONE_SIZE_SHIFT;
90 zone->type = BLK_ZONE_TYPE_SEQWRITE_REQ;
91 zone->cond = BLK_ZONE_COND_EMPTY;
93 sector += dev->zone_size_sects;
96 q->limits.zoned = BLK_ZONED_HM;
97 blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
98 blk_queue_required_elevator_features(q, ELEVATOR_F_ZBD_SEQ_WRITE);
103 int null_register_zoned_dev(struct nullb *nullb)
105 struct nullb_device *dev = nullb->dev;
106 struct request_queue *q = nullb->q;
108 if (queue_is_mq(q)) {
109 int ret = blk_revalidate_disk_zones(nullb->disk, NULL);
114 blk_queue_chunk_sectors(q, dev->zone_size_sects);
115 q->nr_zones = blkdev_nr_zones(nullb->disk);
118 blk_queue_max_zone_append_sectors(q, dev->zone_size_sects);
119 blk_queue_max_open_zones(q, dev->zone_max_open);
120 blk_queue_max_active_zones(q, dev->zone_max_active);
125 void null_free_zoned_dev(struct nullb_device *dev)
130 int null_report_zones(struct gendisk *disk, sector_t sector,
131 unsigned int nr_zones, report_zones_cb cb, void *data)
133 struct nullb *nullb = disk->private_data;
134 struct nullb_device *dev = nullb->dev;
135 unsigned int first_zone, i;
136 struct blk_zone zone;
139 first_zone = null_zone_no(dev, sector);
140 if (first_zone >= dev->nr_zones)
143 nr_zones = min(nr_zones, dev->nr_zones - first_zone);
144 trace_nullb_report_zones(nullb, nr_zones);
146 for (i = 0; i < nr_zones; i++) {
148 * Stacked DM target drivers will remap the zone information by
149 * modifying the zone information passed to the report callback.
150 * So use a local copy to avoid corruption of the device zone
153 spin_lock_irq(&dev->zone_lock);
154 memcpy(&zone, &dev->zones[first_zone + i],
155 sizeof(struct blk_zone));
156 spin_unlock_irq(&dev->zone_lock);
158 error = cb(&zone, i, data);
166 size_t null_zone_valid_read_len(struct nullb *nullb,
167 sector_t sector, unsigned int len)
169 struct nullb_device *dev = nullb->dev;
170 struct blk_zone *zone = &dev->zones[null_zone_no(dev, sector)];
171 unsigned int nr_sectors = len >> SECTOR_SHIFT;
173 /* Read must be below the write pointer position */
174 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL ||
175 sector + nr_sectors <= zone->wp)
178 if (sector > zone->wp)
181 return (zone->wp - sector) << SECTOR_SHIFT;
184 static blk_status_t null_close_zone(struct nullb_device *dev, struct blk_zone *zone)
186 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
187 return BLK_STS_IOERR;
189 switch (zone->cond) {
190 case BLK_ZONE_COND_CLOSED:
191 /* close operation on closed is not an error */
193 case BLK_ZONE_COND_IMP_OPEN:
194 dev->nr_zones_imp_open--;
196 case BLK_ZONE_COND_EXP_OPEN:
197 dev->nr_zones_exp_open--;
199 case BLK_ZONE_COND_EMPTY:
200 case BLK_ZONE_COND_FULL:
202 return BLK_STS_IOERR;
205 if (zone->wp == zone->start) {
206 zone->cond = BLK_ZONE_COND_EMPTY;
208 zone->cond = BLK_ZONE_COND_CLOSED;
209 dev->nr_zones_closed++;
215 static void null_close_first_imp_zone(struct nullb_device *dev)
219 for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
220 if (dev->zones[i].cond == BLK_ZONE_COND_IMP_OPEN) {
221 null_close_zone(dev, &dev->zones[i]);
227 static blk_status_t null_check_active(struct nullb_device *dev)
229 if (!dev->zone_max_active)
232 if (dev->nr_zones_exp_open + dev->nr_zones_imp_open +
233 dev->nr_zones_closed < dev->zone_max_active)
236 return BLK_STS_ZONE_ACTIVE_RESOURCE;
239 static blk_status_t null_check_open(struct nullb_device *dev)
241 if (!dev->zone_max_open)
244 if (dev->nr_zones_exp_open + dev->nr_zones_imp_open < dev->zone_max_open)
247 if (dev->nr_zones_imp_open) {
248 if (null_check_active(dev) == BLK_STS_OK) {
249 null_close_first_imp_zone(dev);
254 return BLK_STS_ZONE_OPEN_RESOURCE;
258 * This function matches the manage open zone resources function in the ZBC standard,
259 * with the addition of max active zones support (added in the ZNS standard).
261 * The function determines if a zone can transition to implicit open or explicit open,
262 * while maintaining the max open zone (and max active zone) limit(s). It may close an
263 * implicit open zone in order to make additional zone resources available.
265 * ZBC states that an implicit open zone shall be closed only if there is not
266 * room within the open limit. However, with the addition of an active limit,
267 * it is not certain that closing an implicit open zone will allow a new zone
268 * to be opened, since we might already be at the active limit capacity.
270 static blk_status_t null_check_zone_resources(struct nullb_device *dev, struct blk_zone *zone)
274 switch (zone->cond) {
275 case BLK_ZONE_COND_EMPTY:
276 ret = null_check_active(dev);
277 if (ret != BLK_STS_OK)
280 case BLK_ZONE_COND_CLOSED:
281 return null_check_open(dev);
283 /* Should never be called for other states */
285 return BLK_STS_IOERR;
289 static blk_status_t null_zone_write(struct nullb_cmd *cmd, sector_t sector,
290 unsigned int nr_sectors, bool append)
292 struct nullb_device *dev = cmd->nq->dev;
293 unsigned int zno = null_zone_no(dev, sector);
294 struct blk_zone *zone = &dev->zones[zno];
297 trace_nullb_zone_op(cmd, zno, zone->cond);
299 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
300 return null_process_cmd(cmd, REQ_OP_WRITE, sector, nr_sectors);
302 switch (zone->cond) {
303 case BLK_ZONE_COND_FULL:
304 /* Cannot write to a full zone */
305 return BLK_STS_IOERR;
306 case BLK_ZONE_COND_EMPTY:
307 case BLK_ZONE_COND_CLOSED:
308 ret = null_check_zone_resources(dev, zone);
309 if (ret != BLK_STS_OK)
312 case BLK_ZONE_COND_IMP_OPEN:
313 case BLK_ZONE_COND_EXP_OPEN:
316 /* Invalid zone condition */
317 return BLK_STS_IOERR;
321 * Regular writes must be at the write pointer position.
322 * Zone append writes are automatically issued at the write
323 * pointer and the position returned using the request or BIO
329 cmd->bio->bi_iter.bi_sector = sector;
331 cmd->rq->__sector = sector;
332 } else if (sector != zone->wp) {
333 return BLK_STS_IOERR;
336 if (zone->wp + nr_sectors > zone->start + zone->capacity)
337 return BLK_STS_IOERR;
339 if (zone->cond == BLK_ZONE_COND_CLOSED) {
340 dev->nr_zones_closed--;
341 dev->nr_zones_imp_open++;
342 } else if (zone->cond == BLK_ZONE_COND_EMPTY) {
343 dev->nr_zones_imp_open++;
345 if (zone->cond != BLK_ZONE_COND_EXP_OPEN)
346 zone->cond = BLK_ZONE_COND_IMP_OPEN;
348 ret = null_process_cmd(cmd, REQ_OP_WRITE, sector, nr_sectors);
349 if (ret != BLK_STS_OK)
352 zone->wp += nr_sectors;
353 if (zone->wp == zone->start + zone->capacity) {
354 if (zone->cond == BLK_ZONE_COND_EXP_OPEN)
355 dev->nr_zones_exp_open--;
356 else if (zone->cond == BLK_ZONE_COND_IMP_OPEN)
357 dev->nr_zones_imp_open--;
358 zone->cond = BLK_ZONE_COND_FULL;
363 static blk_status_t null_open_zone(struct nullb_device *dev, struct blk_zone *zone)
367 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
368 return BLK_STS_IOERR;
370 switch (zone->cond) {
371 case BLK_ZONE_COND_EXP_OPEN:
372 /* open operation on exp open is not an error */
374 case BLK_ZONE_COND_EMPTY:
375 ret = null_check_zone_resources(dev, zone);
376 if (ret != BLK_STS_OK)
379 case BLK_ZONE_COND_IMP_OPEN:
380 dev->nr_zones_imp_open--;
382 case BLK_ZONE_COND_CLOSED:
383 ret = null_check_zone_resources(dev, zone);
384 if (ret != BLK_STS_OK)
386 dev->nr_zones_closed--;
388 case BLK_ZONE_COND_FULL:
390 return BLK_STS_IOERR;
393 zone->cond = BLK_ZONE_COND_EXP_OPEN;
394 dev->nr_zones_exp_open++;
399 static blk_status_t null_finish_zone(struct nullb_device *dev, struct blk_zone *zone)
403 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
404 return BLK_STS_IOERR;
406 switch (zone->cond) {
407 case BLK_ZONE_COND_FULL:
408 /* finish operation on full is not an error */
410 case BLK_ZONE_COND_EMPTY:
411 ret = null_check_zone_resources(dev, zone);
412 if (ret != BLK_STS_OK)
415 case BLK_ZONE_COND_IMP_OPEN:
416 dev->nr_zones_imp_open--;
418 case BLK_ZONE_COND_EXP_OPEN:
419 dev->nr_zones_exp_open--;
421 case BLK_ZONE_COND_CLOSED:
422 ret = null_check_zone_resources(dev, zone);
423 if (ret != BLK_STS_OK)
425 dev->nr_zones_closed--;
428 return BLK_STS_IOERR;
431 zone->cond = BLK_ZONE_COND_FULL;
432 zone->wp = zone->start + zone->len;
437 static blk_status_t null_reset_zone(struct nullb_device *dev, struct blk_zone *zone)
439 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
440 return BLK_STS_IOERR;
442 switch (zone->cond) {
443 case BLK_ZONE_COND_EMPTY:
444 /* reset operation on empty is not an error */
446 case BLK_ZONE_COND_IMP_OPEN:
447 dev->nr_zones_imp_open--;
449 case BLK_ZONE_COND_EXP_OPEN:
450 dev->nr_zones_exp_open--;
452 case BLK_ZONE_COND_CLOSED:
453 dev->nr_zones_closed--;
455 case BLK_ZONE_COND_FULL:
458 return BLK_STS_IOERR;
461 zone->cond = BLK_ZONE_COND_EMPTY;
462 zone->wp = zone->start;
467 static blk_status_t null_zone_mgmt(struct nullb_cmd *cmd, enum req_opf op,
470 struct nullb_device *dev = cmd->nq->dev;
471 unsigned int zone_no = null_zone_no(dev, sector);
472 struct blk_zone *zone = &dev->zones[zone_no];
473 blk_status_t ret = BLK_STS_OK;
477 case REQ_OP_ZONE_RESET_ALL:
478 for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
479 zone = &dev->zones[i];
480 if (zone->cond != BLK_ZONE_COND_EMPTY) {
481 null_reset_zone(dev, zone);
482 trace_nullb_zone_op(cmd, i, zone->cond);
486 case REQ_OP_ZONE_RESET:
487 ret = null_reset_zone(dev, zone);
489 case REQ_OP_ZONE_OPEN:
490 ret = null_open_zone(dev, zone);
492 case REQ_OP_ZONE_CLOSE:
493 ret = null_close_zone(dev, zone);
495 case REQ_OP_ZONE_FINISH:
496 ret = null_finish_zone(dev, zone);
499 return BLK_STS_NOTSUPP;
502 if (ret == BLK_STS_OK)
503 trace_nullb_zone_op(cmd, zone_no, zone->cond);
508 blk_status_t null_process_zoned_cmd(struct nullb_cmd *cmd, enum req_opf op,
509 sector_t sector, sector_t nr_sectors)
512 struct nullb_device *dev = cmd->nq->dev;
514 spin_lock_irq(&dev->zone_lock);
517 sts = null_zone_write(cmd, sector, nr_sectors, false);
519 case REQ_OP_ZONE_APPEND:
520 sts = null_zone_write(cmd, sector, nr_sectors, true);
522 case REQ_OP_ZONE_RESET:
523 case REQ_OP_ZONE_RESET_ALL:
524 case REQ_OP_ZONE_OPEN:
525 case REQ_OP_ZONE_CLOSE:
526 case REQ_OP_ZONE_FINISH:
527 sts = null_zone_mgmt(cmd, op, sector);
530 sts = null_process_cmd(cmd, op, sector, nr_sectors);
532 spin_unlock_irq(&dev->zone_lock);