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 if (dev->zone_nr_conv >= dev->nr_zones) {
49 dev->zone_nr_conv = dev->nr_zones - 1;
50 pr_info("changed the number of conventional zones to %u",
54 /* Max active zones has to be < nbr of seq zones in order to be enforceable */
55 if (dev->zone_max_active >= dev->nr_zones - dev->zone_nr_conv) {
56 dev->zone_max_active = 0;
57 pr_info("zone_max_active limit disabled, limit >= zone count\n");
60 /* Max open zones has to be <= max active zones */
61 if (dev->zone_max_active && dev->zone_max_open > dev->zone_max_active) {
62 dev->zone_max_open = dev->zone_max_active;
63 pr_info("changed the maximum number of open zones to %u\n",
65 } else if (dev->zone_max_open >= dev->nr_zones - dev->zone_nr_conv) {
66 dev->zone_max_open = 0;
67 pr_info("zone_max_open limit disabled, limit >= zone count\n");
70 for (i = 0; i < dev->zone_nr_conv; i++) {
71 struct blk_zone *zone = &dev->zones[i];
74 zone->len = dev->zone_size_sects;
75 zone->capacity = zone->len;
76 zone->wp = zone->start + zone->len;
77 zone->type = BLK_ZONE_TYPE_CONVENTIONAL;
78 zone->cond = BLK_ZONE_COND_NOT_WP;
80 sector += dev->zone_size_sects;
83 for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
84 struct blk_zone *zone = &dev->zones[i];
86 zone->start = zone->wp = sector;
87 zone->len = dev->zone_size_sects;
88 zone->capacity = dev->zone_capacity << ZONE_SIZE_SHIFT;
89 zone->type = BLK_ZONE_TYPE_SEQWRITE_REQ;
90 zone->cond = BLK_ZONE_COND_EMPTY;
92 sector += dev->zone_size_sects;
95 q->limits.zoned = BLK_ZONED_HM;
96 blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
97 blk_queue_required_elevator_features(q, ELEVATOR_F_ZBD_SEQ_WRITE);
102 int null_register_zoned_dev(struct nullb *nullb)
104 struct nullb_device *dev = nullb->dev;
105 struct request_queue *q = nullb->q;
107 if (queue_is_mq(q)) {
108 int ret = blk_revalidate_disk_zones(nullb->disk, NULL);
113 blk_queue_chunk_sectors(q, dev->zone_size_sects);
114 q->nr_zones = blkdev_nr_zones(nullb->disk);
117 blk_queue_max_zone_append_sectors(q, dev->zone_size_sects);
118 blk_queue_max_open_zones(q, dev->zone_max_open);
119 blk_queue_max_active_zones(q, dev->zone_max_active);
124 void null_free_zoned_dev(struct nullb_device *dev)
129 int null_report_zones(struct gendisk *disk, sector_t sector,
130 unsigned int nr_zones, report_zones_cb cb, void *data)
132 struct nullb *nullb = disk->private_data;
133 struct nullb_device *dev = nullb->dev;
134 unsigned int first_zone, i;
135 struct blk_zone zone;
138 first_zone = null_zone_no(dev, sector);
139 if (first_zone >= dev->nr_zones)
142 nr_zones = min(nr_zones, dev->nr_zones - first_zone);
143 trace_nullb_report_zones(nullb, nr_zones);
145 for (i = 0; i < nr_zones; i++) {
147 * Stacked DM target drivers will remap the zone information by
148 * modifying the zone information passed to the report callback.
149 * So use a local copy to avoid corruption of the device zone
152 memcpy(&zone, &dev->zones[first_zone + i],
153 sizeof(struct blk_zone));
154 error = cb(&zone, i, data);
162 size_t null_zone_valid_read_len(struct nullb *nullb,
163 sector_t sector, unsigned int len)
165 struct nullb_device *dev = nullb->dev;
166 struct blk_zone *zone = &dev->zones[null_zone_no(dev, sector)];
167 unsigned int nr_sectors = len >> SECTOR_SHIFT;
169 /* Read must be below the write pointer position */
170 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL ||
171 sector + nr_sectors <= zone->wp)
174 if (sector > zone->wp)
177 return (zone->wp - sector) << SECTOR_SHIFT;
180 static blk_status_t null_close_zone(struct nullb_device *dev, struct blk_zone *zone)
182 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
183 return BLK_STS_IOERR;
185 switch (zone->cond) {
186 case BLK_ZONE_COND_CLOSED:
187 /* close operation on closed is not an error */
189 case BLK_ZONE_COND_IMP_OPEN:
190 dev->nr_zones_imp_open--;
192 case BLK_ZONE_COND_EXP_OPEN:
193 dev->nr_zones_exp_open--;
195 case BLK_ZONE_COND_EMPTY:
196 case BLK_ZONE_COND_FULL:
198 return BLK_STS_IOERR;
201 if (zone->wp == zone->start) {
202 zone->cond = BLK_ZONE_COND_EMPTY;
204 zone->cond = BLK_ZONE_COND_CLOSED;
205 dev->nr_zones_closed++;
211 static void null_close_first_imp_zone(struct nullb_device *dev)
215 for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
216 if (dev->zones[i].cond == BLK_ZONE_COND_IMP_OPEN) {
217 null_close_zone(dev, &dev->zones[i]);
223 static bool null_can_set_active(struct nullb_device *dev)
225 if (!dev->zone_max_active)
228 return dev->nr_zones_exp_open + dev->nr_zones_imp_open +
229 dev->nr_zones_closed < dev->zone_max_active;
232 static bool null_can_open(struct nullb_device *dev)
234 if (!dev->zone_max_open)
237 if (dev->nr_zones_exp_open + dev->nr_zones_imp_open < dev->zone_max_open)
240 if (dev->nr_zones_imp_open && null_can_set_active(dev)) {
241 null_close_first_imp_zone(dev);
249 * This function matches the manage open zone resources function in the ZBC standard,
250 * with the addition of max active zones support (added in the ZNS standard).
252 * The function determines if a zone can transition to implicit open or explicit open,
253 * while maintaining the max open zone (and max active zone) limit(s). It may close an
254 * implicit open zone in order to make additional zone resources available.
256 * ZBC states that an implicit open zone shall be closed only if there is not
257 * room within the open limit. However, with the addition of an active limit,
258 * it is not certain that closing an implicit open zone will allow a new zone
259 * to be opened, since we might already be at the active limit capacity.
261 static bool null_has_zone_resources(struct nullb_device *dev, struct blk_zone *zone)
263 switch (zone->cond) {
264 case BLK_ZONE_COND_EMPTY:
265 if (!null_can_set_active(dev))
268 case BLK_ZONE_COND_CLOSED:
269 return null_can_open(dev);
271 /* Should never be called for other states */
277 static blk_status_t null_zone_write(struct nullb_cmd *cmd, sector_t sector,
278 unsigned int nr_sectors, bool append)
280 struct nullb_device *dev = cmd->nq->dev;
281 unsigned int zno = null_zone_no(dev, sector);
282 struct blk_zone *zone = &dev->zones[zno];
285 trace_nullb_zone_op(cmd, zno, zone->cond);
287 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
288 return null_process_cmd(cmd, REQ_OP_WRITE, sector, nr_sectors);
290 switch (zone->cond) {
291 case BLK_ZONE_COND_FULL:
292 /* Cannot write to a full zone */
293 return BLK_STS_IOERR;
294 case BLK_ZONE_COND_EMPTY:
295 case BLK_ZONE_COND_CLOSED:
296 if (!null_has_zone_resources(dev, zone))
297 return BLK_STS_IOERR;
299 case BLK_ZONE_COND_IMP_OPEN:
300 case BLK_ZONE_COND_EXP_OPEN:
303 /* Invalid zone condition */
304 return BLK_STS_IOERR;
308 * Regular writes must be at the write pointer position.
309 * Zone append writes are automatically issued at the write
310 * pointer and the position returned using the request or BIO
316 cmd->bio->bi_iter.bi_sector = sector;
318 cmd->rq->__sector = sector;
319 } else if (sector != zone->wp) {
320 return BLK_STS_IOERR;
323 if (zone->wp + nr_sectors > zone->start + zone->capacity)
324 return BLK_STS_IOERR;
326 if (zone->cond == BLK_ZONE_COND_CLOSED) {
327 dev->nr_zones_closed--;
328 dev->nr_zones_imp_open++;
329 } else if (zone->cond == BLK_ZONE_COND_EMPTY) {
330 dev->nr_zones_imp_open++;
332 if (zone->cond != BLK_ZONE_COND_EXP_OPEN)
333 zone->cond = BLK_ZONE_COND_IMP_OPEN;
335 ret = null_process_cmd(cmd, REQ_OP_WRITE, sector, nr_sectors);
336 if (ret != BLK_STS_OK)
339 zone->wp += nr_sectors;
340 if (zone->wp == zone->start + zone->capacity) {
341 if (zone->cond == BLK_ZONE_COND_EXP_OPEN)
342 dev->nr_zones_exp_open--;
343 else if (zone->cond == BLK_ZONE_COND_IMP_OPEN)
344 dev->nr_zones_imp_open--;
345 zone->cond = BLK_ZONE_COND_FULL;
350 static blk_status_t null_open_zone(struct nullb_device *dev, struct blk_zone *zone)
352 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
353 return BLK_STS_IOERR;
355 switch (zone->cond) {
356 case BLK_ZONE_COND_EXP_OPEN:
357 /* open operation on exp open is not an error */
359 case BLK_ZONE_COND_EMPTY:
360 if (!null_has_zone_resources(dev, zone))
361 return BLK_STS_IOERR;
363 case BLK_ZONE_COND_IMP_OPEN:
364 dev->nr_zones_imp_open--;
366 case BLK_ZONE_COND_CLOSED:
367 if (!null_has_zone_resources(dev, zone))
368 return BLK_STS_IOERR;
369 dev->nr_zones_closed--;
371 case BLK_ZONE_COND_FULL:
373 return BLK_STS_IOERR;
376 zone->cond = BLK_ZONE_COND_EXP_OPEN;
377 dev->nr_zones_exp_open++;
382 static blk_status_t null_finish_zone(struct nullb_device *dev, struct blk_zone *zone)
384 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
385 return BLK_STS_IOERR;
387 switch (zone->cond) {
388 case BLK_ZONE_COND_FULL:
389 /* finish operation on full is not an error */
391 case BLK_ZONE_COND_EMPTY:
392 if (!null_has_zone_resources(dev, zone))
393 return BLK_STS_IOERR;
395 case BLK_ZONE_COND_IMP_OPEN:
396 dev->nr_zones_imp_open--;
398 case BLK_ZONE_COND_EXP_OPEN:
399 dev->nr_zones_exp_open--;
401 case BLK_ZONE_COND_CLOSED:
402 if (!null_has_zone_resources(dev, zone))
403 return BLK_STS_IOERR;
404 dev->nr_zones_closed--;
407 return BLK_STS_IOERR;
410 zone->cond = BLK_ZONE_COND_FULL;
411 zone->wp = zone->start + zone->len;
416 static blk_status_t null_reset_zone(struct nullb_device *dev, struct blk_zone *zone)
418 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
419 return BLK_STS_IOERR;
421 switch (zone->cond) {
422 case BLK_ZONE_COND_EMPTY:
423 /* reset operation on empty is not an error */
425 case BLK_ZONE_COND_IMP_OPEN:
426 dev->nr_zones_imp_open--;
428 case BLK_ZONE_COND_EXP_OPEN:
429 dev->nr_zones_exp_open--;
431 case BLK_ZONE_COND_CLOSED:
432 dev->nr_zones_closed--;
434 case BLK_ZONE_COND_FULL:
437 return BLK_STS_IOERR;
440 zone->cond = BLK_ZONE_COND_EMPTY;
441 zone->wp = zone->start;
446 static blk_status_t null_zone_mgmt(struct nullb_cmd *cmd, enum req_opf op,
449 struct nullb_device *dev = cmd->nq->dev;
450 unsigned int zone_no = null_zone_no(dev, sector);
451 struct blk_zone *zone = &dev->zones[zone_no];
452 blk_status_t ret = BLK_STS_OK;
456 case REQ_OP_ZONE_RESET_ALL:
457 for (i = dev->zone_nr_conv; i < dev->nr_zones; i++)
458 null_reset_zone(dev, &dev->zones[i]);
460 case REQ_OP_ZONE_RESET:
461 ret = null_reset_zone(dev, zone);
463 case REQ_OP_ZONE_OPEN:
464 ret = null_open_zone(dev, zone);
466 case REQ_OP_ZONE_CLOSE:
467 ret = null_close_zone(dev, zone);
469 case REQ_OP_ZONE_FINISH:
470 ret = null_finish_zone(dev, zone);
473 return BLK_STS_NOTSUPP;
476 if (ret == BLK_STS_OK)
477 trace_nullb_zone_op(cmd, zone_no, zone->cond);
482 blk_status_t null_process_zoned_cmd(struct nullb_cmd *cmd, enum req_opf op,
483 sector_t sector, sector_t nr_sectors)
487 return null_zone_write(cmd, sector, nr_sectors, false);
488 case REQ_OP_ZONE_APPEND:
489 return null_zone_write(cmd, sector, nr_sectors, true);
490 case REQ_OP_ZONE_RESET:
491 case REQ_OP_ZONE_RESET_ALL:
492 case REQ_OP_ZONE_OPEN:
493 case REQ_OP_ZONE_CLOSE:
494 case REQ_OP_ZONE_FINISH:
495 return null_zone_mgmt(cmd, op, sector);
497 return null_process_cmd(cmd, op, sector, nr_sectors);