xref: /linux/drivers/block/null_blk/zoned.c (revision 9b960d8cd6f712cb2c03e2bdd4d5ca058238037f)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/vmalloc.h>
3 #include <linux/bitmap.h>
4 #include "null_blk.h"
5 
6 #define CREATE_TRACE_POINTS
7 #include "trace.h"
8 
9 #undef pr_fmt
10 #define pr_fmt(fmt)	"null_blk: " fmt
11 
12 #define NULL_ZONE_INVALID_WP	((sector_t)-1)
13 
mb_to_sects(unsigned long mb)14 static inline sector_t mb_to_sects(unsigned long mb)
15 {
16 	return ((sector_t)mb * SZ_1M) >> SECTOR_SHIFT;
17 }
18 
null_zone_no(struct nullb_device * dev,sector_t sect)19 static inline unsigned int null_zone_no(struct nullb_device *dev, sector_t sect)
20 {
21 	return sect >> ilog2(dev->zone_size_sects);
22 }
23 
null_init_zone_lock(struct nullb_device * dev,struct nullb_zone * zone)24 static inline void null_init_zone_lock(struct nullb_device *dev,
25 				       struct nullb_zone *zone)
26 {
27 	if (!dev->memory_backed)
28 		spin_lock_init(&zone->spinlock);
29 	else
30 		mutex_init(&zone->mutex);
31 }
32 
null_lock_zone(struct nullb_device * dev,struct nullb_zone * zone)33 static inline void null_lock_zone(struct nullb_device *dev,
34 				  struct nullb_zone *zone)
35 {
36 	if (!dev->memory_backed)
37 		spin_lock_irq(&zone->spinlock);
38 	else
39 		mutex_lock(&zone->mutex);
40 }
41 
null_unlock_zone(struct nullb_device * dev,struct nullb_zone * zone)42 static inline void null_unlock_zone(struct nullb_device *dev,
43 				    struct nullb_zone *zone)
44 {
45 	if (!dev->memory_backed)
46 		spin_unlock_irq(&zone->spinlock);
47 	else
48 		mutex_unlock(&zone->mutex);
49 }
50 
null_init_zoned_dev(struct nullb_device * dev,struct queue_limits * lim)51 int null_init_zoned_dev(struct nullb_device *dev,
52 			struct queue_limits *lim)
53 {
54 	sector_t dev_capacity_sects, zone_capacity_sects;
55 	struct nullb_zone *zone;
56 	sector_t sector = 0;
57 	unsigned int i;
58 
59 	if (!is_power_of_2(dev->zone_size)) {
60 		pr_err("zone_size must be power-of-two\n");
61 		return -EINVAL;
62 	}
63 	if (dev->zone_size > dev->size) {
64 		pr_err("Zone size larger than device capacity\n");
65 		return -EINVAL;
66 	}
67 
68 	if (!dev->zone_capacity)
69 		dev->zone_capacity = dev->zone_size;
70 
71 	if (dev->zone_capacity > dev->zone_size) {
72 		pr_err("zone capacity (%lu MB) larger than zone size (%lu MB)\n",
73 		       dev->zone_capacity, dev->zone_size);
74 		return -EINVAL;
75 	}
76 
77 	/*
78 	 * If a smaller zone capacity was requested, do not allow a smaller last
79 	 * zone at the same time as such zone configuration does not correspond
80 	 * to any real zoned device.
81 	 */
82 	if (dev->zone_capacity != dev->zone_size &&
83 	    dev->size & (dev->zone_size - 1)) {
84 		pr_err("A smaller last zone is not allowed with zone capacity smaller than zone size.\n");
85 		return -EINVAL;
86 	}
87 
88 	zone_capacity_sects = mb_to_sects(dev->zone_capacity);
89 	dev_capacity_sects = mb_to_sects(dev->size);
90 	dev->zone_size_sects = mb_to_sects(dev->zone_size);
91 	dev->nr_zones = round_up(dev_capacity_sects, dev->zone_size_sects)
92 		>> ilog2(dev->zone_size_sects);
93 
94 	dev->zones = kvmalloc_array(dev->nr_zones, sizeof(struct nullb_zone),
95 				    GFP_KERNEL | __GFP_ZERO);
96 	if (!dev->zones)
97 		return -ENOMEM;
98 
99 	spin_lock_init(&dev->zone_res_lock);
100 
101 	if (dev->zone_nr_conv >= dev->nr_zones) {
102 		dev->zone_nr_conv = dev->nr_zones - 1;
103 		pr_info("changed the number of conventional zones to %u",
104 			dev->zone_nr_conv);
105 	}
106 
107 	dev->zone_append_max_sectors =
108 		min(ALIGN_DOWN(dev->zone_append_max_sectors,
109 			       dev->blocksize >> SECTOR_SHIFT),
110 		    zone_capacity_sects);
111 
112 	/* Max active zones has to be < nbr of seq zones in order to be enforceable */
113 	if (dev->zone_max_active >= dev->nr_zones - dev->zone_nr_conv) {
114 		dev->zone_max_active = 0;
115 		pr_info("zone_max_active limit disabled, limit >= zone count\n");
116 	}
117 
118 	/* Max open zones has to be <= max active zones */
119 	if (dev->zone_max_active && dev->zone_max_open > dev->zone_max_active) {
120 		dev->zone_max_open = dev->zone_max_active;
121 		pr_info("changed the maximum number of open zones to %u\n",
122 			dev->zone_max_open);
123 	} else if (dev->zone_max_open >= dev->nr_zones - dev->zone_nr_conv) {
124 		dev->zone_max_open = 0;
125 		pr_info("zone_max_open limit disabled, limit >= zone count\n");
126 	}
127 	dev->need_zone_res_mgmt = dev->zone_max_active || dev->zone_max_open;
128 	dev->imp_close_zone_no = dev->zone_nr_conv;
129 
130 	for (i = 0; i <  dev->zone_nr_conv; i++) {
131 		zone = &dev->zones[i];
132 
133 		null_init_zone_lock(dev, zone);
134 		zone->start = sector;
135 		zone->len = dev->zone_size_sects;
136 		zone->capacity = zone->len;
137 		zone->wp = zone->start + zone->len;
138 		zone->type = BLK_ZONE_TYPE_CONVENTIONAL;
139 		zone->cond = BLK_ZONE_COND_NOT_WP;
140 
141 		sector += dev->zone_size_sects;
142 	}
143 
144 	for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
145 		zone = &dev->zones[i];
146 
147 		null_init_zone_lock(dev, zone);
148 		zone->start = sector;
149 		if (zone->start + dev->zone_size_sects > dev_capacity_sects)
150 			zone->len = dev_capacity_sects - zone->start;
151 		else
152 			zone->len = dev->zone_size_sects;
153 		zone->capacity =
154 			min_t(sector_t, zone->len, zone_capacity_sects);
155 		zone->type = BLK_ZONE_TYPE_SEQWRITE_REQ;
156 		if (dev->zone_full) {
157 			zone->cond = BLK_ZONE_COND_FULL;
158 			zone->wp = zone->start + zone->capacity;
159 		} else{
160 			zone->cond = BLK_ZONE_COND_EMPTY;
161 			zone->wp = zone->start;
162 		}
163 
164 		sector += dev->zone_size_sects;
165 	}
166 
167 	lim->features |= BLK_FEAT_ZONED;
168 	lim->chunk_sectors = dev->zone_size_sects;
169 	lim->max_hw_zone_append_sectors = dev->zone_append_max_sectors;
170 	lim->max_open_zones = dev->zone_max_open;
171 	lim->max_active_zones = dev->zone_max_active;
172 	return 0;
173 }
174 
null_register_zoned_dev(struct nullb * nullb)175 int null_register_zoned_dev(struct nullb *nullb)
176 {
177 	struct request_queue *q = nullb->q;
178 	struct gendisk *disk = nullb->disk;
179 
180 	pr_info("%s: using %s zone append\n",
181 		disk->disk_name,
182 		queue_emulates_zone_append(q) ? "emulated" : "native");
183 
184 	return blk_revalidate_disk_zones(disk);
185 }
186 
null_free_zoned_dev(struct nullb_device * dev)187 void null_free_zoned_dev(struct nullb_device *dev)
188 {
189 	kvfree(dev->zones);
190 	dev->zones = NULL;
191 }
192 
null_report_zones(struct gendisk * disk,sector_t sector,unsigned int nr_zones,report_zones_cb cb,void * data)193 int null_report_zones(struct gendisk *disk, sector_t sector,
194 		unsigned int nr_zones, report_zones_cb cb, void *data)
195 {
196 	struct nullb *nullb = disk->private_data;
197 	struct nullb_device *dev = nullb->dev;
198 	unsigned int first_zone, i;
199 	struct nullb_zone *zone;
200 	struct blk_zone blkz;
201 	int error;
202 
203 	first_zone = null_zone_no(dev, sector);
204 	if (first_zone >= dev->nr_zones)
205 		return 0;
206 
207 	nr_zones = min(nr_zones, dev->nr_zones - first_zone);
208 	trace_nullb_report_zones(nullb, nr_zones);
209 
210 	memset(&blkz, 0, sizeof(struct blk_zone));
211 	zone = &dev->zones[first_zone];
212 	for (i = 0; i < nr_zones; i++, zone++) {
213 		/*
214 		 * Stacked DM target drivers will remap the zone information by
215 		 * modifying the zone information passed to the report callback.
216 		 * So use a local copy to avoid corruption of the device zone
217 		 * array.
218 		 */
219 		null_lock_zone(dev, zone);
220 		blkz.start = zone->start;
221 		blkz.len = zone->len;
222 		blkz.wp = zone->wp;
223 		blkz.type = zone->type;
224 		blkz.cond = zone->cond;
225 		blkz.capacity = zone->capacity;
226 		null_unlock_zone(dev, zone);
227 
228 		error = cb(&blkz, i, data);
229 		if (error)
230 			return error;
231 	}
232 
233 	return nr_zones;
234 }
235 
236 /*
237  * This is called in the case of memory backing from null_process_cmd()
238  * with the target zone already locked.
239  */
null_zone_valid_read_len(struct nullb * nullb,sector_t sector,unsigned int len)240 size_t null_zone_valid_read_len(struct nullb *nullb,
241 				sector_t sector, unsigned int len)
242 {
243 	struct nullb_device *dev = nullb->dev;
244 	struct nullb_zone *zone = &dev->zones[null_zone_no(dev, sector)];
245 	unsigned int nr_sectors = len >> SECTOR_SHIFT;
246 
247 	/* Read must be below the write pointer position */
248 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL ||
249 	    sector + nr_sectors <= zone->wp)
250 		return len;
251 
252 	if (sector > zone->wp)
253 		return 0;
254 
255 	return (zone->wp - sector) << SECTOR_SHIFT;
256 }
257 
null_close_imp_open_zone(struct nullb_device * dev)258 static void null_close_imp_open_zone(struct nullb_device *dev)
259 {
260 	struct nullb_zone *zone;
261 	unsigned int zno, i;
262 
263 	zno = dev->imp_close_zone_no;
264 	if (zno >= dev->nr_zones)
265 		zno = dev->zone_nr_conv;
266 
267 	for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
268 		zone = &dev->zones[zno];
269 		zno++;
270 		if (zno >= dev->nr_zones)
271 			zno = dev->zone_nr_conv;
272 
273 		if (zone->cond == BLK_ZONE_COND_IMP_OPEN) {
274 			dev->nr_zones_imp_open--;
275 			if (zone->wp == zone->start) {
276 				zone->cond = BLK_ZONE_COND_EMPTY;
277 			} else {
278 				zone->cond = BLK_ZONE_COND_CLOSED;
279 				dev->nr_zones_closed++;
280 			}
281 			dev->imp_close_zone_no = zno;
282 			return;
283 		}
284 	}
285 }
286 
null_check_active(struct nullb_device * dev)287 static blk_status_t null_check_active(struct nullb_device *dev)
288 {
289 	if (!dev->zone_max_active)
290 		return BLK_STS_OK;
291 
292 	if (dev->nr_zones_exp_open + dev->nr_zones_imp_open +
293 			dev->nr_zones_closed < dev->zone_max_active)
294 		return BLK_STS_OK;
295 
296 	return BLK_STS_ZONE_ACTIVE_RESOURCE;
297 }
298 
null_check_open(struct nullb_device * dev)299 static blk_status_t null_check_open(struct nullb_device *dev)
300 {
301 	if (!dev->zone_max_open)
302 		return BLK_STS_OK;
303 
304 	if (dev->nr_zones_exp_open + dev->nr_zones_imp_open < dev->zone_max_open)
305 		return BLK_STS_OK;
306 
307 	if (dev->nr_zones_imp_open) {
308 		if (null_check_active(dev) == BLK_STS_OK) {
309 			null_close_imp_open_zone(dev);
310 			return BLK_STS_OK;
311 		}
312 	}
313 
314 	return BLK_STS_ZONE_OPEN_RESOURCE;
315 }
316 
317 /*
318  * This function matches the manage open zone resources function in the ZBC standard,
319  * with the addition of max active zones support (added in the ZNS standard).
320  *
321  * The function determines if a zone can transition to implicit open or explicit open,
322  * while maintaining the max open zone (and max active zone) limit(s). It may close an
323  * implicit open zone in order to make additional zone resources available.
324  *
325  * ZBC states that an implicit open zone shall be closed only if there is not
326  * room within the open limit. However, with the addition of an active limit,
327  * it is not certain that closing an implicit open zone will allow a new zone
328  * to be opened, since we might already be at the active limit capacity.
329  */
null_check_zone_resources(struct nullb_device * dev,struct nullb_zone * zone)330 static blk_status_t null_check_zone_resources(struct nullb_device *dev,
331 					      struct nullb_zone *zone)
332 {
333 	blk_status_t ret;
334 
335 	switch (zone->cond) {
336 	case BLK_ZONE_COND_EMPTY:
337 		ret = null_check_active(dev);
338 		if (ret != BLK_STS_OK)
339 			return ret;
340 		fallthrough;
341 	case BLK_ZONE_COND_CLOSED:
342 		return null_check_open(dev);
343 	default:
344 		/* Should never be called for other states */
345 		WARN_ON(1);
346 		return BLK_STS_IOERR;
347 	}
348 }
349 
null_zone_write(struct nullb_cmd * cmd,sector_t sector,unsigned int nr_sectors,bool append)350 static blk_status_t null_zone_write(struct nullb_cmd *cmd, sector_t sector,
351 				    unsigned int nr_sectors, bool append)
352 {
353 	struct nullb_device *dev = cmd->nq->dev;
354 	unsigned int zno = null_zone_no(dev, sector);
355 	struct nullb_zone *zone = &dev->zones[zno];
356 	blk_status_t badblocks_ret = BLK_STS_OK;
357 	blk_status_t ret;
358 
359 	trace_nullb_zone_op(cmd, zno, zone->cond);
360 
361 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) {
362 		if (append)
363 			return BLK_STS_IOERR;
364 		return null_process_cmd(cmd, REQ_OP_WRITE, sector, nr_sectors);
365 	}
366 
367 	null_lock_zone(dev, zone);
368 
369 	/*
370 	 * Regular writes must be at the write pointer position. Zone append
371 	 * writes are automatically issued at the write pointer and the position
372 	 * returned using the request sector. Note that we do not check the zone
373 	 * condition because for FULL, READONLY and OFFLINE zones, the sector
374 	 * check against the zone write pointer will always result in failing
375 	 * the command.
376 	 */
377 	if (append) {
378 		if (WARN_ON_ONCE(!dev->zone_append_max_sectors) ||
379 		    zone->wp == NULL_ZONE_INVALID_WP) {
380 			ret = BLK_STS_IOERR;
381 			goto unlock_zone;
382 		}
383 		sector = zone->wp;
384 		blk_mq_rq_from_pdu(cmd)->__sector = sector;
385 	}
386 
387 	if (sector != zone->wp ||
388 	    zone->wp + nr_sectors > zone->start + zone->capacity) {
389 		ret = BLK_STS_IOERR;
390 		goto unlock_zone;
391 	}
392 
393 	if (zone->cond == BLK_ZONE_COND_CLOSED ||
394 	    zone->cond == BLK_ZONE_COND_EMPTY) {
395 		if (dev->need_zone_res_mgmt) {
396 			spin_lock(&dev->zone_res_lock);
397 
398 			ret = null_check_zone_resources(dev, zone);
399 			if (ret != BLK_STS_OK) {
400 				spin_unlock(&dev->zone_res_lock);
401 				goto unlock_zone;
402 			}
403 			if (zone->cond == BLK_ZONE_COND_CLOSED) {
404 				dev->nr_zones_closed--;
405 				dev->nr_zones_imp_open++;
406 			} else if (zone->cond == BLK_ZONE_COND_EMPTY) {
407 				dev->nr_zones_imp_open++;
408 			}
409 
410 			spin_unlock(&dev->zone_res_lock);
411 		}
412 
413 		zone->cond = BLK_ZONE_COND_IMP_OPEN;
414 	}
415 
416 	if (dev->badblocks.shift != -1) {
417 		badblocks_ret = null_handle_badblocks(cmd, sector, &nr_sectors);
418 		if (badblocks_ret != BLK_STS_OK && !nr_sectors) {
419 			ret = badblocks_ret;
420 			goto unlock_zone;
421 		}
422 	}
423 
424 	if (dev->memory_backed) {
425 		ret = null_handle_memory_backed(cmd, REQ_OP_WRITE, sector,
426 						nr_sectors);
427 		if (ret != BLK_STS_OK)
428 			goto unlock_zone;
429 	}
430 
431 	zone->wp += nr_sectors;
432 	if (zone->wp == zone->start + zone->capacity) {
433 		if (dev->need_zone_res_mgmt) {
434 			spin_lock(&dev->zone_res_lock);
435 			if (zone->cond == BLK_ZONE_COND_EXP_OPEN)
436 				dev->nr_zones_exp_open--;
437 			else if (zone->cond == BLK_ZONE_COND_IMP_OPEN)
438 				dev->nr_zones_imp_open--;
439 			spin_unlock(&dev->zone_res_lock);
440 		}
441 		zone->cond = BLK_ZONE_COND_FULL;
442 	}
443 
444 	ret = badblocks_ret;
445 
446 unlock_zone:
447 	null_unlock_zone(dev, zone);
448 
449 	return ret;
450 }
451 
null_open_zone(struct nullb_device * dev,struct nullb_zone * zone)452 static blk_status_t null_open_zone(struct nullb_device *dev,
453 				   struct nullb_zone *zone)
454 {
455 	blk_status_t ret = BLK_STS_OK;
456 
457 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
458 		return BLK_STS_IOERR;
459 
460 	switch (zone->cond) {
461 	case BLK_ZONE_COND_EXP_OPEN:
462 		/* Open operation on exp open is not an error */
463 		return BLK_STS_OK;
464 	case BLK_ZONE_COND_EMPTY:
465 	case BLK_ZONE_COND_IMP_OPEN:
466 	case BLK_ZONE_COND_CLOSED:
467 		break;
468 	case BLK_ZONE_COND_FULL:
469 	default:
470 		return BLK_STS_IOERR;
471 	}
472 
473 	if (dev->need_zone_res_mgmt) {
474 		spin_lock(&dev->zone_res_lock);
475 
476 		switch (zone->cond) {
477 		case BLK_ZONE_COND_EMPTY:
478 			ret = null_check_zone_resources(dev, zone);
479 			if (ret != BLK_STS_OK) {
480 				spin_unlock(&dev->zone_res_lock);
481 				return ret;
482 			}
483 			break;
484 		case BLK_ZONE_COND_IMP_OPEN:
485 			dev->nr_zones_imp_open--;
486 			break;
487 		case BLK_ZONE_COND_CLOSED:
488 			ret = null_check_zone_resources(dev, zone);
489 			if (ret != BLK_STS_OK) {
490 				spin_unlock(&dev->zone_res_lock);
491 				return ret;
492 			}
493 			dev->nr_zones_closed--;
494 			break;
495 		default:
496 			break;
497 		}
498 
499 		dev->nr_zones_exp_open++;
500 
501 		spin_unlock(&dev->zone_res_lock);
502 	}
503 
504 	zone->cond = BLK_ZONE_COND_EXP_OPEN;
505 
506 	return BLK_STS_OK;
507 }
508 
null_close_zone(struct nullb_device * dev,struct nullb_zone * zone)509 static blk_status_t null_close_zone(struct nullb_device *dev,
510 				    struct nullb_zone *zone)
511 {
512 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
513 		return BLK_STS_IOERR;
514 
515 	switch (zone->cond) {
516 	case BLK_ZONE_COND_CLOSED:
517 		/* close operation on closed is not an error */
518 		return BLK_STS_OK;
519 	case BLK_ZONE_COND_IMP_OPEN:
520 	case BLK_ZONE_COND_EXP_OPEN:
521 		break;
522 	case BLK_ZONE_COND_EMPTY:
523 	case BLK_ZONE_COND_FULL:
524 	default:
525 		return BLK_STS_IOERR;
526 	}
527 
528 	if (dev->need_zone_res_mgmt) {
529 		spin_lock(&dev->zone_res_lock);
530 
531 		switch (zone->cond) {
532 		case BLK_ZONE_COND_IMP_OPEN:
533 			dev->nr_zones_imp_open--;
534 			break;
535 		case BLK_ZONE_COND_EXP_OPEN:
536 			dev->nr_zones_exp_open--;
537 			break;
538 		default:
539 			break;
540 		}
541 
542 		if (zone->wp > zone->start)
543 			dev->nr_zones_closed++;
544 
545 		spin_unlock(&dev->zone_res_lock);
546 	}
547 
548 	if (zone->wp == zone->start)
549 		zone->cond = BLK_ZONE_COND_EMPTY;
550 	else
551 		zone->cond = BLK_ZONE_COND_CLOSED;
552 
553 	return BLK_STS_OK;
554 }
555 
null_finish_zone(struct nullb_device * dev,struct nullb_zone * zone)556 static blk_status_t null_finish_zone(struct nullb_device *dev,
557 				     struct nullb_zone *zone)
558 {
559 	blk_status_t ret = BLK_STS_OK;
560 
561 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
562 		return BLK_STS_IOERR;
563 
564 	if (dev->need_zone_res_mgmt) {
565 		spin_lock(&dev->zone_res_lock);
566 
567 		switch (zone->cond) {
568 		case BLK_ZONE_COND_FULL:
569 			/* Finish operation on full is not an error */
570 			spin_unlock(&dev->zone_res_lock);
571 			return BLK_STS_OK;
572 		case BLK_ZONE_COND_EMPTY:
573 			ret = null_check_zone_resources(dev, zone);
574 			if (ret != BLK_STS_OK) {
575 				spin_unlock(&dev->zone_res_lock);
576 				return ret;
577 			}
578 			break;
579 		case BLK_ZONE_COND_IMP_OPEN:
580 			dev->nr_zones_imp_open--;
581 			break;
582 		case BLK_ZONE_COND_EXP_OPEN:
583 			dev->nr_zones_exp_open--;
584 			break;
585 		case BLK_ZONE_COND_CLOSED:
586 			ret = null_check_zone_resources(dev, zone);
587 			if (ret != BLK_STS_OK) {
588 				spin_unlock(&dev->zone_res_lock);
589 				return ret;
590 			}
591 			dev->nr_zones_closed--;
592 			break;
593 		default:
594 			spin_unlock(&dev->zone_res_lock);
595 			return BLK_STS_IOERR;
596 		}
597 
598 		spin_unlock(&dev->zone_res_lock);
599 	}
600 
601 	zone->cond = BLK_ZONE_COND_FULL;
602 	zone->wp = zone->start + zone->len;
603 
604 	return BLK_STS_OK;
605 }
606 
null_reset_zone(struct nullb_device * dev,struct nullb_zone * zone)607 static blk_status_t null_reset_zone(struct nullb_device *dev,
608 				    struct nullb_zone *zone)
609 {
610 	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
611 		return BLK_STS_IOERR;
612 
613 	if (dev->need_zone_res_mgmt) {
614 		spin_lock(&dev->zone_res_lock);
615 
616 		switch (zone->cond) {
617 		case BLK_ZONE_COND_IMP_OPEN:
618 			dev->nr_zones_imp_open--;
619 			break;
620 		case BLK_ZONE_COND_EXP_OPEN:
621 			dev->nr_zones_exp_open--;
622 			break;
623 		case BLK_ZONE_COND_CLOSED:
624 			dev->nr_zones_closed--;
625 			break;
626 		case BLK_ZONE_COND_EMPTY:
627 		case BLK_ZONE_COND_FULL:
628 			break;
629 		default:
630 			spin_unlock(&dev->zone_res_lock);
631 			return BLK_STS_IOERR;
632 		}
633 
634 		spin_unlock(&dev->zone_res_lock);
635 	}
636 
637 	zone->cond = BLK_ZONE_COND_EMPTY;
638 	zone->wp = zone->start;
639 
640 	if (dev->memory_backed)
641 		return null_handle_discard(dev, zone->start, zone->len);
642 
643 	return BLK_STS_OK;
644 }
645 
null_zone_mgmt(struct nullb_cmd * cmd,enum req_op op,sector_t sector)646 static blk_status_t null_zone_mgmt(struct nullb_cmd *cmd, enum req_op op,
647 				   sector_t sector)
648 {
649 	struct nullb_device *dev = cmd->nq->dev;
650 	unsigned int zone_no;
651 	struct nullb_zone *zone;
652 	blk_status_t ret;
653 	size_t i;
654 
655 	if (op == REQ_OP_ZONE_RESET_ALL) {
656 		for (i = dev->zone_nr_conv; i < dev->nr_zones; i++) {
657 			zone = &dev->zones[i];
658 			null_lock_zone(dev, zone);
659 			if (zone->cond != BLK_ZONE_COND_EMPTY &&
660 			    zone->cond != BLK_ZONE_COND_READONLY &&
661 			    zone->cond != BLK_ZONE_COND_OFFLINE) {
662 				null_reset_zone(dev, zone);
663 				trace_nullb_zone_op(cmd, i, zone->cond);
664 			}
665 			null_unlock_zone(dev, zone);
666 		}
667 		return BLK_STS_OK;
668 	}
669 
670 	zone_no = null_zone_no(dev, sector);
671 	zone = &dev->zones[zone_no];
672 
673 	null_lock_zone(dev, zone);
674 
675 	if (zone->cond == BLK_ZONE_COND_READONLY ||
676 	    zone->cond == BLK_ZONE_COND_OFFLINE) {
677 		ret = BLK_STS_IOERR;
678 		goto unlock;
679 	}
680 
681 	switch (op) {
682 	case REQ_OP_ZONE_RESET:
683 		ret = null_reset_zone(dev, zone);
684 		break;
685 	case REQ_OP_ZONE_OPEN:
686 		ret = null_open_zone(dev, zone);
687 		break;
688 	case REQ_OP_ZONE_CLOSE:
689 		ret = null_close_zone(dev, zone);
690 		break;
691 	case REQ_OP_ZONE_FINISH:
692 		ret = null_finish_zone(dev, zone);
693 		break;
694 	default:
695 		ret = BLK_STS_NOTSUPP;
696 		break;
697 	}
698 
699 	if (ret == BLK_STS_OK)
700 		trace_nullb_zone_op(cmd, zone_no, zone->cond);
701 
702 unlock:
703 	null_unlock_zone(dev, zone);
704 
705 	return ret;
706 }
707 
null_process_zoned_cmd(struct nullb_cmd * cmd,enum req_op op,sector_t sector,sector_t nr_sectors)708 blk_status_t null_process_zoned_cmd(struct nullb_cmd *cmd, enum req_op op,
709 				    sector_t sector, sector_t nr_sectors)
710 {
711 	struct nullb_device *dev;
712 	struct nullb_zone *zone;
713 	blk_status_t sts;
714 
715 	switch (op) {
716 	case REQ_OP_WRITE:
717 		return null_zone_write(cmd, sector, nr_sectors, false);
718 	case REQ_OP_ZONE_APPEND:
719 		return null_zone_write(cmd, sector, nr_sectors, true);
720 	case REQ_OP_ZONE_RESET:
721 	case REQ_OP_ZONE_RESET_ALL:
722 	case REQ_OP_ZONE_OPEN:
723 	case REQ_OP_ZONE_CLOSE:
724 	case REQ_OP_ZONE_FINISH:
725 		return null_zone_mgmt(cmd, op, sector);
726 	default:
727 		dev = cmd->nq->dev;
728 		zone = &dev->zones[null_zone_no(dev, sector)];
729 		if (zone->cond == BLK_ZONE_COND_OFFLINE)
730 			return BLK_STS_IOERR;
731 
732 		null_lock_zone(dev, zone);
733 		sts = null_process_cmd(cmd, op, sector, nr_sectors);
734 		null_unlock_zone(dev, zone);
735 		return sts;
736 	}
737 }
738 
739 /*
740  * Set a zone in the read-only or offline condition.
741  */
null_set_zone_cond(struct nullb_device * dev,struct nullb_zone * zone,enum blk_zone_cond cond)742 static void null_set_zone_cond(struct nullb_device *dev,
743 			       struct nullb_zone *zone, enum blk_zone_cond cond)
744 {
745 	if (WARN_ON_ONCE(cond != BLK_ZONE_COND_READONLY &&
746 			 cond != BLK_ZONE_COND_OFFLINE))
747 		return;
748 
749 	null_lock_zone(dev, zone);
750 
751 	/*
752 	 * If the read-only condition is requested again to zones already in
753 	 * read-only condition, restore back normal empty condition. Do the same
754 	 * if the offline condition is requested for offline zones. Otherwise,
755 	 * set the specified zone condition to the zones. Finish the zones
756 	 * beforehand to free up zone resources.
757 	 */
758 	if (zone->cond == cond) {
759 		zone->cond = BLK_ZONE_COND_EMPTY;
760 		zone->wp = zone->start;
761 		if (dev->memory_backed)
762 			null_handle_discard(dev, zone->start, zone->len);
763 	} else {
764 		if (zone->cond != BLK_ZONE_COND_READONLY &&
765 		    zone->cond != BLK_ZONE_COND_OFFLINE)
766 			null_finish_zone(dev, zone);
767 		zone->cond = cond;
768 		zone->wp = NULL_ZONE_INVALID_WP;
769 	}
770 
771 	null_unlock_zone(dev, zone);
772 }
773 
774 /*
775  * Identify a zone from the sector written to configfs file. Then set zone
776  * condition to the zone.
777  */
zone_cond_store(struct nullb_device * dev,const char * page,size_t count,enum blk_zone_cond cond)778 ssize_t zone_cond_store(struct nullb_device *dev, const char *page,
779 			size_t count, enum blk_zone_cond cond)
780 {
781 	unsigned long long sector;
782 	unsigned int zone_no;
783 	int ret;
784 
785 	if (!dev->zoned) {
786 		pr_err("null_blk device is not zoned\n");
787 		return -EINVAL;
788 	}
789 
790 	if (!dev->zones) {
791 		pr_err("null_blk device is not yet powered\n");
792 		return -EINVAL;
793 	}
794 
795 	ret = kstrtoull(page, 0, &sector);
796 	if (ret < 0)
797 		return ret;
798 
799 	zone_no = null_zone_no(dev, sector);
800 	if (zone_no >= dev->nr_zones) {
801 		pr_err("Sector out of range\n");
802 		return -EINVAL;
803 	}
804 
805 	if (dev->zones[zone_no].type == BLK_ZONE_TYPE_CONVENTIONAL) {
806 		pr_err("Can not change condition of conventional zones\n");
807 		return -EINVAL;
808 	}
809 
810 	null_set_zone_cond(dev, &dev->zones[zone_no], cond);
811 
812 	return count;
813 }
814