2 * SCSI Zoned Block commands
4 * Copyright (C) 2014-2015 SUSE Linux GmbH
5 * Written by: Hannes Reinecke <hare@suse.de>
6 * Modified by: Damien Le Moal <damien.lemoal@hgst.com>
7 * Modified by: Shaun Tancheff <shaun.tancheff@seagate.com>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License version
11 * 2 as published by the Free Software Foundation.
13 * This program is distributed in the hope that it will be useful, but
14 * WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program; see the file COPYING. If not, write to
20 * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
25 #include <linux/blkdev.h>
27 #include <asm/unaligned.h>
29 #include <scsi/scsi.h>
30 #include <scsi/scsi_cmnd.h>
35 * sd_zbc_parse_report - Convert a zone descriptor to a struct blk_zone,
36 * @sdkp: The disk the report originated from
37 * @buf: Address of the report zone descriptor
38 * @zone: the destination zone structure
40 * All LBA sized values are converted to 512B sectors unit.
42 static void sd_zbc_parse_report(struct scsi_disk *sdkp, u8 *buf,
43 struct blk_zone *zone)
45 struct scsi_device *sdp = sdkp->device;
47 memset(zone, 0, sizeof(struct blk_zone));
49 zone->type = buf[0] & 0x0f;
50 zone->cond = (buf[1] >> 4) & 0xf;
56 zone->len = logical_to_sectors(sdp, get_unaligned_be64(&buf[8]));
57 zone->start = logical_to_sectors(sdp, get_unaligned_be64(&buf[16]));
58 zone->wp = logical_to_sectors(sdp, get_unaligned_be64(&buf[24]));
59 if (zone->type != ZBC_ZONE_TYPE_CONV &&
60 zone->cond == ZBC_ZONE_COND_FULL)
61 zone->wp = zone->start + zone->len;
65 * sd_zbc_report_zones - Issue a REPORT ZONES scsi command.
66 * @sdkp: The target disk
67 * @buf: Buffer to use for the reply
68 * @buflen: the buffer size
69 * @lba: Start LBA of the report
71 * For internal use during device validation.
73 static int sd_zbc_report_zones(struct scsi_disk *sdkp, unsigned char *buf,
74 unsigned int buflen, sector_t lba)
76 struct scsi_device *sdp = sdkp->device;
77 const int timeout = sdp->request_queue->rq_timeout;
78 struct scsi_sense_hdr sshdr;
79 unsigned char cmd[16];
85 cmd[1] = ZI_REPORT_ZONES;
86 put_unaligned_be64(lba, &cmd[2]);
87 put_unaligned_be32(buflen, &cmd[10]);
88 memset(buf, 0, buflen);
90 result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
92 timeout, SD_MAX_RETRIES, NULL);
94 sd_printk(KERN_ERR, sdkp,
95 "REPORT ZONES lba %llu failed with %d/%d\n",
96 (unsigned long long)lba,
97 host_byte(result), driver_byte(result));
101 rep_len = get_unaligned_be32(&buf[0]);
103 sd_printk(KERN_ERR, sdkp,
104 "REPORT ZONES report invalid length %u\n",
113 * sd_zbc_setup_report_cmnd - Prepare a REPORT ZONES scsi command
114 * @cmd: The command to setup
116 * Call in sd_init_command() for a REQ_OP_ZONE_REPORT request.
118 int sd_zbc_setup_report_cmnd(struct scsi_cmnd *cmd)
120 struct request *rq = cmd->request;
121 struct scsi_disk *sdkp = scsi_disk(rq->rq_disk);
122 sector_t lba, sector = blk_rq_pos(rq);
123 unsigned int nr_bytes = blk_rq_bytes(rq);
126 WARN_ON(nr_bytes == 0);
128 if (!sd_is_zoned(sdkp))
129 /* Not a zoned device */
132 ret = scsi_init_io(cmd);
133 if (ret != BLKPREP_OK)
137 memset(cmd->cmnd, 0, cmd->cmd_len);
138 cmd->cmnd[0] = ZBC_IN;
139 cmd->cmnd[1] = ZI_REPORT_ZONES;
140 lba = sectors_to_logical(sdkp->device, sector);
141 put_unaligned_be64(lba, &cmd->cmnd[2]);
142 put_unaligned_be32(nr_bytes, &cmd->cmnd[10]);
143 /* Do partial report for speeding things up */
144 cmd->cmnd[14] = ZBC_REPORT_ZONE_PARTIAL;
146 cmd->sc_data_direction = DMA_FROM_DEVICE;
147 cmd->sdb.length = nr_bytes;
148 cmd->transfersize = sdkp->device->sector_size;
152 * Report may return less bytes than requested. Make sure
153 * to report completion on the entire initial request.
155 rq->__data_len = nr_bytes;
161 * sd_zbc_report_zones_complete - Process a REPORT ZONES scsi command reply.
162 * @scmd: The completed report zones command
163 * @good_bytes: reply size in bytes
165 * Convert all reported zone descriptors to struct blk_zone. The conversion
166 * is done in-place, directly in the request specified sg buffer.
168 static void sd_zbc_report_zones_complete(struct scsi_cmnd *scmd,
169 unsigned int good_bytes)
171 struct request *rq = scmd->request;
172 struct scsi_disk *sdkp = scsi_disk(rq->rq_disk);
173 struct sg_mapping_iter miter;
174 struct blk_zone_report_hdr hdr;
175 struct blk_zone zone;
176 unsigned int offset, bytes = 0;
183 memset(&hdr, 0, sizeof(struct blk_zone_report_hdr));
185 sg_miter_start(&miter, scsi_sglist(scmd), scsi_sg_count(scmd),
186 SG_MITER_TO_SG | SG_MITER_ATOMIC);
188 local_irq_save(flags);
189 while (sg_miter_next(&miter) && bytes < good_bytes) {
195 /* Set the report header */
196 hdr.nr_zones = min_t(unsigned int,
197 (good_bytes - 64) / 64,
198 get_unaligned_be32(&buf[0]) / 64);
199 memcpy(buf, &hdr, sizeof(struct blk_zone_report_hdr));
204 /* Parse zone descriptors */
205 while (offset < miter.length && hdr.nr_zones) {
206 WARN_ON(offset > miter.length);
207 buf = miter.addr + offset;
208 sd_zbc_parse_report(sdkp, buf, &zone);
209 memcpy(buf, &zone, sizeof(struct blk_zone));
219 sg_miter_stop(&miter);
220 local_irq_restore(flags);
224 * sd_zbc_zone_sectors - Get the device zone size in number of 512B sectors.
225 * @sdkp: The target disk
227 static inline sector_t sd_zbc_zone_sectors(struct scsi_disk *sdkp)
229 return logical_to_sectors(sdkp->device, sdkp->zone_blocks);
233 * sd_zbc_setup_reset_cmnd - Prepare a RESET WRITE POINTER scsi command.
234 * @cmd: the command to setup
236 * Called from sd_init_command() for a REQ_OP_ZONE_RESET request.
238 int sd_zbc_setup_reset_cmnd(struct scsi_cmnd *cmd)
240 struct request *rq = cmd->request;
241 struct scsi_disk *sdkp = scsi_disk(rq->rq_disk);
242 sector_t sector = blk_rq_pos(rq);
243 sector_t block = sectors_to_logical(sdkp->device, sector);
245 if (!sd_is_zoned(sdkp))
246 /* Not a zoned device */
249 if (sdkp->device->changed)
252 if (sector & (sd_zbc_zone_sectors(sdkp) - 1))
253 /* Unaligned request */
257 memset(cmd->cmnd, 0, cmd->cmd_len);
258 cmd->cmnd[0] = ZBC_OUT;
259 cmd->cmnd[1] = ZO_RESET_WRITE_POINTER;
260 put_unaligned_be64(block, &cmd->cmnd[2]);
262 rq->timeout = SD_TIMEOUT;
263 cmd->sc_data_direction = DMA_NONE;
264 cmd->transfersize = 0;
271 * sd_zbc_complete - ZBC command post processing.
272 * @cmd: Completed command
273 * @good_bytes: Command reply bytes
274 * @sshdr: command sense header
276 * Called from sd_done(). Process report zones reply and handle reset zone
277 * and write commands errors.
279 void sd_zbc_complete(struct scsi_cmnd *cmd, unsigned int good_bytes,
280 struct scsi_sense_hdr *sshdr)
282 int result = cmd->result;
283 struct request *rq = cmd->request;
285 switch (req_op(rq)) {
286 case REQ_OP_ZONE_RESET:
289 sshdr->sense_key == ILLEGAL_REQUEST &&
292 * INVALID FIELD IN CDB error: reset of a conventional
293 * zone was attempted. Nothing to worry about, so be
294 * quiet about the error.
296 rq->rq_flags |= RQF_QUIET;
300 case REQ_OP_WRITE_ZEROES:
301 case REQ_OP_WRITE_SAME:
304 case REQ_OP_ZONE_REPORT:
307 sd_zbc_report_zones_complete(cmd, good_bytes);
314 * sd_zbc_read_zoned_characteristics - Read zoned block device characteristics
316 * @buf: Buffer where to store the VPD page data
320 static int sd_zbc_read_zoned_characteristics(struct scsi_disk *sdkp,
324 if (scsi_get_vpd_page(sdkp->device, 0xb6, buf, 64)) {
325 sd_printk(KERN_NOTICE, sdkp,
326 "Unconstrained-read check failed\n");
330 if (sdkp->device->type != TYPE_ZBC) {
333 sdkp->zones_optimal_open = get_unaligned_be32(&buf[8]);
334 sdkp->zones_optimal_nonseq = get_unaligned_be32(&buf[12]);
335 sdkp->zones_max_open = 0;
338 sdkp->urswrz = buf[4] & 1;
339 sdkp->zones_optimal_open = 0;
340 sdkp->zones_optimal_nonseq = 0;
341 sdkp->zones_max_open = get_unaligned_be32(&buf[16]);
348 * sd_zbc_check_capacity - Check reported capacity.
350 * @buf: Buffer to use for commands
352 * ZBC drive may report only the capacity of the first conventional zones at
353 * LBA 0. This is indicated by the RC_BASIS field of the read capacity reply.
354 * Check this here. If the disk reported only its conventional zones capacity,
355 * get the total capacity by doing a report zones.
357 static int sd_zbc_check_capacity(struct scsi_disk *sdkp, unsigned char *buf)
362 if (sdkp->rc_basis != 0)
365 /* Do a report zone to get the maximum LBA to check capacity */
366 ret = sd_zbc_report_zones(sdkp, buf, SD_BUF_SIZE, 0);
370 /* The max_lba field is the capacity of this device */
371 lba = get_unaligned_be64(&buf[8]);
372 if (lba + 1 == sdkp->capacity)
375 if (sdkp->first_scan)
376 sd_printk(KERN_WARNING, sdkp,
377 "Changing capacity from %llu to max LBA+1 %llu\n",
378 (unsigned long long)sdkp->capacity,
379 (unsigned long long)lba + 1);
380 sdkp->capacity = lba + 1;
385 #define SD_ZBC_BUF_SIZE 131072U
388 * sd_zbc_check_zone_size - Check the device zone sizes
391 * Check that all zones of the device are equal. The last zone can however
392 * be smaller. The zone size must also be a power of two number of LBAs.
394 * Returns the zone size in bytes upon success or an error code upon failure.
396 static s64 sd_zbc_check_zone_size(struct scsi_disk *sdkp)
402 unsigned int buf_len;
403 unsigned int list_length;
408 buf = kmalloc(SD_ZBC_BUF_SIZE, GFP_KERNEL);
412 /* Do a report zone to get the same field */
413 ret = sd_zbc_report_zones(sdkp, buf, SD_ZBC_BUF_SIZE, 0);
417 same = buf[4] & 0x0f;
420 zone_blocks = get_unaligned_be64(&rec[8]);
425 * Check the size of all zones: all zones must be of
426 * equal size, except the last zone which can be smaller
431 /* Parse REPORT ZONES header */
432 list_length = get_unaligned_be32(&buf[0]) + 64;
434 buf_len = min(list_length, SD_ZBC_BUF_SIZE);
436 /* Parse zone descriptors */
437 while (rec < buf + buf_len) {
438 u64 this_zone_blocks = get_unaligned_be64(&rec[8]);
440 if (zone_blocks == 0) {
441 zone_blocks = this_zone_blocks;
442 } else if (this_zone_blocks != zone_blocks &&
443 (block + this_zone_blocks < sdkp->capacity
444 || this_zone_blocks > zone_blocks)) {
445 this_zone_blocks = 0;
448 block += this_zone_blocks;
452 if (block < sdkp->capacity) {
453 ret = sd_zbc_report_zones(sdkp, buf,
454 SD_ZBC_BUF_SIZE, block);
459 } while (block < sdkp->capacity);
463 if (sdkp->first_scan)
464 sd_printk(KERN_NOTICE, sdkp,
465 "Devices with non constant zone "
466 "size are not supported\n");
468 } else if (!is_power_of_2(zone_blocks)) {
469 if (sdkp->first_scan)
470 sd_printk(KERN_NOTICE, sdkp,
471 "Devices with non power of 2 zone "
472 "size are not supported\n");
474 } else if (logical_to_sectors(sdkp->device, zone_blocks) > UINT_MAX) {
475 if (sdkp->first_scan)
476 sd_printk(KERN_NOTICE, sdkp,
477 "Zone size too large\n");
490 * sd_zbc_alloc_zone_bitmap - Allocate a zone bitmap (one bit per zone).
491 * @nr_zones: Number of zones to allocate space for.
492 * @numa_node: NUMA node to allocate the memory from.
494 static inline unsigned long *
495 sd_zbc_alloc_zone_bitmap(u32 nr_zones, int numa_node)
497 return kzalloc_node(BITS_TO_LONGS(nr_zones) * sizeof(unsigned long),
498 GFP_KERNEL, numa_node);
502 * sd_zbc_get_seq_zones - Parse report zones reply to identify sequential zones
504 * @buf: report reply buffer
505 * @buflen: length of @buf
506 * @zone_shift: logarithm base 2 of the number of blocks in a zone
507 * @seq_zones_bitmap: bitmap of sequential zones to set
509 * Parse reported zone descriptors in @buf to identify sequential zones and
510 * set the reported zone bit in @seq_zones_bitmap accordingly.
511 * Since read-only and offline zones cannot be written, do not
512 * mark them as sequential in the bitmap.
513 * Return the LBA after the last zone reported.
515 static sector_t sd_zbc_get_seq_zones(struct scsi_disk *sdkp, unsigned char *buf,
516 unsigned int buflen, u32 zone_shift,
517 unsigned long *seq_zones_bitmap)
519 sector_t lba, next_lba = sdkp->capacity;
520 unsigned int buf_len, list_length;
524 list_length = get_unaligned_be32(&buf[0]) + 64;
525 buf_len = min(list_length, buflen);
528 while (rec < buf + buf_len) {
529 type = rec[0] & 0x0f;
530 cond = (rec[1] >> 4) & 0xf;
531 lba = get_unaligned_be64(&rec[16]);
532 if (type != ZBC_ZONE_TYPE_CONV &&
533 cond != ZBC_ZONE_COND_READONLY &&
534 cond != ZBC_ZONE_COND_OFFLINE)
535 set_bit(lba >> zone_shift, seq_zones_bitmap);
536 next_lba = lba + get_unaligned_be64(&rec[8]);
544 * sd_zbc_setup_seq_zones_bitmap - Initialize a seq zone bitmap.
546 * @zone_shift: logarithm base 2 of the number of blocks in a zone
547 * @nr_zones: number of zones to set up a seq zone bitmap for
549 * Allocate a zone bitmap and initialize it by identifying sequential zones.
551 static unsigned long *
552 sd_zbc_setup_seq_zones_bitmap(struct scsi_disk *sdkp, u32 zone_shift,
555 struct request_queue *q = sdkp->disk->queue;
556 unsigned long *seq_zones_bitmap;
561 seq_zones_bitmap = sd_zbc_alloc_zone_bitmap(nr_zones, q->node);
562 if (!seq_zones_bitmap)
563 return ERR_PTR(-ENOMEM);
565 buf = kmalloc(SD_ZBC_BUF_SIZE, GFP_KERNEL);
569 while (lba < sdkp->capacity) {
570 ret = sd_zbc_report_zones(sdkp, buf, SD_ZBC_BUF_SIZE, lba);
573 lba = sd_zbc_get_seq_zones(sdkp, buf, SD_ZBC_BUF_SIZE,
574 zone_shift, seq_zones_bitmap);
577 if (lba != sdkp->capacity) {
578 /* Something went wrong */
585 kfree(seq_zones_bitmap);
588 return seq_zones_bitmap;
591 static void sd_zbc_cleanup(struct scsi_disk *sdkp)
593 struct request_queue *q = sdkp->disk->queue;
595 kfree(q->seq_zones_bitmap);
596 q->seq_zones_bitmap = NULL;
598 kfree(q->seq_zones_wlock);
599 q->seq_zones_wlock = NULL;
604 static int sd_zbc_setup(struct scsi_disk *sdkp, u32 zone_blocks)
606 struct request_queue *q = sdkp->disk->queue;
607 u32 zone_shift = ilog2(zone_blocks);
611 /* chunk_sectors indicates the zone size */
612 blk_queue_chunk_sectors(q,
613 logical_to_sectors(sdkp->device, zone_blocks));
614 nr_zones = round_up(sdkp->capacity, zone_blocks) >> zone_shift;
617 * Initialize the device request queue information if the number
620 if (nr_zones != sdkp->nr_zones || nr_zones != q->nr_zones) {
621 unsigned long *seq_zones_wlock = NULL, *seq_zones_bitmap = NULL;
622 size_t zone_bitmap_size;
625 seq_zones_wlock = sd_zbc_alloc_zone_bitmap(nr_zones,
627 if (!seq_zones_wlock) {
632 seq_zones_bitmap = sd_zbc_setup_seq_zones_bitmap(sdkp,
633 zone_shift, nr_zones);
634 if (IS_ERR(seq_zones_bitmap)) {
635 ret = PTR_ERR(seq_zones_bitmap);
636 kfree(seq_zones_wlock);
640 zone_bitmap_size = BITS_TO_LONGS(nr_zones) *
641 sizeof(unsigned long);
642 blk_mq_freeze_queue(q);
643 if (q->nr_zones != nr_zones) {
644 /* READ16/WRITE16 is mandatory for ZBC disks */
645 sdkp->device->use_16_for_rw = 1;
646 sdkp->device->use_10_for_rw = 0;
648 sdkp->zone_blocks = zone_blocks;
649 sdkp->zone_shift = zone_shift;
650 sdkp->nr_zones = nr_zones;
651 q->nr_zones = nr_zones;
652 swap(q->seq_zones_wlock, seq_zones_wlock);
653 swap(q->seq_zones_bitmap, seq_zones_bitmap);
654 } else if (memcmp(q->seq_zones_bitmap, seq_zones_bitmap,
655 zone_bitmap_size) != 0) {
656 memcpy(q->seq_zones_bitmap, seq_zones_bitmap,
659 blk_mq_unfreeze_queue(q);
660 kfree(seq_zones_wlock);
661 kfree(seq_zones_bitmap);
667 sd_zbc_cleanup(sdkp);
671 int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
676 if (!sd_is_zoned(sdkp))
678 * Device managed or normal SCSI disk,
679 * no special handling required
683 /* Get zoned block device characteristics */
684 ret = sd_zbc_read_zoned_characteristics(sdkp, buf);
689 * Check for unconstrained reads: host-managed devices with
690 * constrained reads (drives failing read after write pointer)
694 if (sdkp->first_scan)
695 sd_printk(KERN_NOTICE, sdkp,
696 "constrained reads devices are not supported\n");
702 ret = sd_zbc_check_capacity(sdkp, buf);
707 * Check zone size: only devices with a constant zone size (except
708 * an eventual last runt zone) that is a power of 2 are supported.
710 zone_blocks = sd_zbc_check_zone_size(sdkp);
712 if (zone_blocks != (u32)zone_blocks)
718 /* The drive satisfies the kernel restrictions: set it up */
719 ret = sd_zbc_setup(sdkp, zone_blocks);
727 sd_zbc_cleanup(sdkp);
732 void sd_zbc_remove(struct scsi_disk *sdkp)
734 sd_zbc_cleanup(sdkp);
737 void sd_zbc_print_zones(struct scsi_disk *sdkp)
739 if (!sd_is_zoned(sdkp) || !sdkp->capacity)
742 if (sdkp->capacity & (sdkp->zone_blocks - 1))
743 sd_printk(KERN_NOTICE, sdkp,
744 "%u zones of %u logical blocks + 1 runt zone\n",
748 sd_printk(KERN_NOTICE, sdkp,
749 "%u zones of %u logical blocks\n",