1 // SPDX-License-Identifier: GPL-2.0-only
3 * SCSI Zoned Block commands
5 * Copyright (C) 2014-2015 SUSE Linux GmbH
6 * Written by: Hannes Reinecke <hare@suse.de>
7 * Modified by: Damien Le Moal <damien.lemoal@hgst.com>
8 * Modified by: Shaun Tancheff <shaun.tancheff@seagate.com>
11 #include <linux/blkdev.h>
12 #include <linux/vmalloc.h>
13 #include <linux/sched/mm.h>
15 #include <asm/unaligned.h>
17 #include <scsi/scsi.h>
18 #include <scsi/scsi_cmnd.h>
23 * sd_zbc_parse_report - Convert a zone descriptor to a struct blk_zone,
24 * @sdkp: The disk the report originated from
25 * @buf: Address of the report zone descriptor
26 * @zone: the destination zone structure
28 * All LBA sized values are converted to 512B sectors unit.
30 static void sd_zbc_parse_report(struct scsi_disk *sdkp, u8 *buf,
31 struct blk_zone *zone)
33 struct scsi_device *sdp = sdkp->device;
35 memset(zone, 0, sizeof(struct blk_zone));
37 zone->type = buf[0] & 0x0f;
38 zone->cond = (buf[1] >> 4) & 0xf;
44 zone->len = logical_to_sectors(sdp, get_unaligned_be64(&buf[8]));
45 zone->start = logical_to_sectors(sdp, get_unaligned_be64(&buf[16]));
46 zone->wp = logical_to_sectors(sdp, get_unaligned_be64(&buf[24]));
47 if (zone->type != ZBC_ZONE_TYPE_CONV &&
48 zone->cond == ZBC_ZONE_COND_FULL)
49 zone->wp = zone->start + zone->len;
53 * sd_zbc_do_report_zones - Issue a REPORT ZONES scsi command.
54 * @sdkp: The target disk
55 * @buf: vmalloc-ed buffer to use for the reply
56 * @buflen: the buffer size
57 * @lba: Start LBA of the report
58 * @partial: Do partial report
60 * For internal use during device validation.
61 * Using partial=true can significantly speed up execution of a report zones
62 * command because the disk does not have to count all possible report matching
63 * zones and will only report the count of zones fitting in the command reply
66 static int sd_zbc_do_report_zones(struct scsi_disk *sdkp, unsigned char *buf,
67 unsigned int buflen, sector_t lba,
70 struct scsi_device *sdp = sdkp->device;
71 const int timeout = sdp->request_queue->rq_timeout;
72 struct scsi_sense_hdr sshdr;
73 unsigned char cmd[16];
79 cmd[1] = ZI_REPORT_ZONES;
80 put_unaligned_be64(lba, &cmd[2]);
81 put_unaligned_be32(buflen, &cmd[10]);
83 cmd[14] = ZBC_REPORT_ZONE_PARTIAL;
85 result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
87 timeout, SD_MAX_RETRIES, NULL);
89 sd_printk(KERN_ERR, sdkp,
90 "REPORT ZONES lba %llu failed with %d/%d\n",
91 (unsigned long long)lba,
92 host_byte(result), driver_byte(result));
96 rep_len = get_unaligned_be32(&buf[0]);
98 sd_printk(KERN_ERR, sdkp,
99 "REPORT ZONES report invalid length %u\n",
108 * Maximum number of zones to get with one report zones command.
110 #define SD_ZBC_REPORT_MAX_ZONES 8192U
113 * Allocate a buffer for report zones reply.
114 * @sdkp: The target disk
115 * @nr_zones: Maximum number of zones to report
116 * @buflen: Size of the buffer allocated
118 * Try to allocate a reply buffer for the number of requested zones.
119 * The size of the buffer allocated may be smaller than requested to
120 * satify the device constraint (max_hw_sectors, max_segments, etc).
122 * Return the address of the allocated buffer and update @buflen with
123 * the size of the allocated buffer.
125 static void *sd_zbc_alloc_report_buffer(struct scsi_disk *sdkp,
126 unsigned int nr_zones, size_t *buflen)
128 struct request_queue *q = sdkp->disk->queue;
133 * Report zone buffer size should be at most 64B times the number of
134 * zones requested plus the 64B reply header, but should be at least
135 * SECTOR_SIZE for ATA devices.
136 * Make sure that this size does not exceed the hardware capabilities.
137 * Furthermore, since the report zone command cannot be split, make
138 * sure that the allocated buffer can always be mapped by limiting the
139 * number of pages allocated to the HBA max segments limit.
141 nr_zones = min(nr_zones, SD_ZBC_REPORT_MAX_ZONES);
142 bufsize = roundup((nr_zones + 1) * 64, 512);
143 bufsize = min_t(size_t, bufsize,
144 queue_max_hw_sectors(q) << SECTOR_SHIFT);
145 bufsize = min_t(size_t, bufsize, queue_max_segments(q) << PAGE_SHIFT);
147 buf = vzalloc(bufsize);
155 * sd_zbc_report_zones - Disk report zones operation.
156 * @disk: The target disk
157 * @sector: Start 512B sector of the report
158 * @zones: Array of zone descriptors
159 * @nr_zones: Number of descriptors in the array
161 * Execute a report zones command on the target disk.
163 int sd_zbc_report_zones(struct gendisk *disk, sector_t sector,
164 struct blk_zone *zones, unsigned int *nr_zones)
166 struct scsi_disk *sdkp = scsi_disk(disk);
167 unsigned int i, nrz = *nr_zones;
169 size_t buflen = 0, offset = 0;
172 if (!sd_is_zoned(sdkp))
173 /* Not a zoned device */
176 buf = sd_zbc_alloc_report_buffer(sdkp, nrz, &buflen);
180 ret = sd_zbc_do_report_zones(sdkp, buf, buflen,
181 sectors_to_logical(sdkp->device, sector), true);
185 nrz = min(nrz, get_unaligned_be32(&buf[0]) / 64);
186 for (i = 0; i < nrz; i++) {
188 sd_zbc_parse_report(sdkp, buf + offset, zones);
201 * sd_zbc_zone_sectors - Get the device zone size in number of 512B sectors.
202 * @sdkp: The target disk
204 static inline sector_t sd_zbc_zone_sectors(struct scsi_disk *sdkp)
206 return logical_to_sectors(sdkp->device, sdkp->zone_blocks);
210 * sd_zbc_setup_reset_cmnd - Prepare a RESET WRITE POINTER scsi command.
211 * @cmd: the command to setup
212 * @all: Reset all zones control.
214 * Called from sd_init_command() for a REQ_OP_ZONE_RESET request.
216 blk_status_t sd_zbc_setup_reset_cmnd(struct scsi_cmnd *cmd, bool all)
218 struct request *rq = cmd->request;
219 struct scsi_disk *sdkp = scsi_disk(rq->rq_disk);
220 sector_t sector = blk_rq_pos(rq);
221 sector_t block = sectors_to_logical(sdkp->device, sector);
223 if (!sd_is_zoned(sdkp))
224 /* Not a zoned device */
225 return BLK_STS_IOERR;
227 if (sdkp->device->changed)
228 return BLK_STS_IOERR;
230 if (sector & (sd_zbc_zone_sectors(sdkp) - 1))
231 /* Unaligned request */
232 return BLK_STS_IOERR;
235 memset(cmd->cmnd, 0, cmd->cmd_len);
236 cmd->cmnd[0] = ZBC_OUT;
237 cmd->cmnd[1] = ZO_RESET_WRITE_POINTER;
241 put_unaligned_be64(block, &cmd->cmnd[2]);
243 rq->timeout = SD_TIMEOUT;
244 cmd->sc_data_direction = DMA_NONE;
245 cmd->transfersize = 0;
252 * sd_zbc_complete - ZBC command post processing.
253 * @cmd: Completed command
254 * @good_bytes: Command reply bytes
255 * @sshdr: command sense header
257 * Called from sd_done(). Process report zones reply and handle reset zone
258 * and write commands errors.
260 void sd_zbc_complete(struct scsi_cmnd *cmd, unsigned int good_bytes,
261 struct scsi_sense_hdr *sshdr)
263 int result = cmd->result;
264 struct request *rq = cmd->request;
266 switch (req_op(rq)) {
267 case REQ_OP_ZONE_RESET:
268 case REQ_OP_ZONE_RESET_ALL:
271 sshdr->sense_key == ILLEGAL_REQUEST &&
274 * INVALID FIELD IN CDB error: reset of a conventional
275 * zone was attempted. Nothing to worry about, so be
276 * quiet about the error.
278 rq->rq_flags |= RQF_QUIET;
282 case REQ_OP_WRITE_ZEROES:
283 case REQ_OP_WRITE_SAME:
289 * sd_zbc_check_zoned_characteristics - Check zoned block device characteristics
291 * @buf: Buffer where to store the VPD page data
293 * Read VPD page B6, get information and check that reads are unconstrained.
295 static int sd_zbc_check_zoned_characteristics(struct scsi_disk *sdkp,
299 if (scsi_get_vpd_page(sdkp->device, 0xb6, buf, 64)) {
300 sd_printk(KERN_NOTICE, sdkp,
301 "Read zoned characteristics VPD page failed\n");
305 if (sdkp->device->type != TYPE_ZBC) {
308 sdkp->zones_optimal_open = get_unaligned_be32(&buf[8]);
309 sdkp->zones_optimal_nonseq = get_unaligned_be32(&buf[12]);
310 sdkp->zones_max_open = 0;
313 sdkp->urswrz = buf[4] & 1;
314 sdkp->zones_optimal_open = 0;
315 sdkp->zones_optimal_nonseq = 0;
316 sdkp->zones_max_open = get_unaligned_be32(&buf[16]);
320 * Check for unconstrained reads: host-managed devices with
321 * constrained reads (drives failing read after write pointer)
325 if (sdkp->first_scan)
326 sd_printk(KERN_NOTICE, sdkp,
327 "constrained reads devices are not supported\n");
335 * sd_zbc_check_zones - Check the device capacity and zone sizes
338 * Check that the device capacity as reported by READ CAPACITY matches the
339 * max_lba value (plus one)of the report zones command reply. Also check that
340 * all zones of the device have an equal size, only allowing the last zone of
341 * the disk to have a smaller size (runt zone). The zone size must also be a
344 * Returns the zone size in number of blocks upon success or an error code
347 static int sd_zbc_check_zones(struct scsi_disk *sdkp, u32 *zblocks)
349 size_t bufsize, buflen;
350 unsigned int noio_flag;
352 sector_t max_lba, block = 0;
358 /* Do all memory allocations as if GFP_NOIO was specified */
359 noio_flag = memalloc_noio_save();
362 buf = sd_zbc_alloc_report_buffer(sdkp, SD_ZBC_REPORT_MAX_ZONES,
369 /* Do a report zone to get max_lba and the same field */
370 ret = sd_zbc_do_report_zones(sdkp, buf, bufsize, 0, false);
374 if (sdkp->rc_basis == 0) {
375 /* The max_lba field is the capacity of this device */
376 max_lba = get_unaligned_be64(&buf[8]);
377 if (sdkp->capacity != max_lba + 1) {
378 if (sdkp->first_scan)
379 sd_printk(KERN_WARNING, sdkp,
380 "Changing capacity from %llu to max LBA+1 %llu\n",
381 (unsigned long long)sdkp->capacity,
382 (unsigned long long)max_lba + 1);
383 sdkp->capacity = max_lba + 1;
388 * Check same field: for any value other than 0, we know that all zones
389 * have the same size.
391 same = buf[4] & 0x0f;
394 zone_blocks = get_unaligned_be64(&rec[8]);
399 * Check the size of all zones: all zones must be of
400 * equal size, except the last zone which can be smaller
405 /* Parse REPORT ZONES header */
406 buflen = min_t(size_t, get_unaligned_be32(&buf[0]) + 64,
410 /* Parse zone descriptors */
411 while (rec < buf + buflen) {
412 u64 this_zone_blocks = get_unaligned_be64(&rec[8]);
414 if (zone_blocks == 0) {
415 zone_blocks = this_zone_blocks;
416 } else if (this_zone_blocks != zone_blocks &&
417 (block + this_zone_blocks < sdkp->capacity
418 || this_zone_blocks > zone_blocks)) {
422 block += this_zone_blocks;
426 if (block < sdkp->capacity) {
427 ret = sd_zbc_do_report_zones(sdkp, buf, bufsize, block,
433 } while (block < sdkp->capacity);
437 if (sdkp->first_scan)
438 sd_printk(KERN_NOTICE, sdkp,
439 "Devices with non constant zone "
440 "size are not supported\n");
442 } else if (!is_power_of_2(zone_blocks)) {
443 if (sdkp->first_scan)
444 sd_printk(KERN_NOTICE, sdkp,
445 "Devices with non power of 2 zone "
446 "size are not supported\n");
448 } else if (logical_to_sectors(sdkp->device, zone_blocks) > UINT_MAX) {
449 if (sdkp->first_scan)
450 sd_printk(KERN_NOTICE, sdkp,
451 "Zone size too large\n");
454 *zblocks = zone_blocks;
459 memalloc_noio_restore(noio_flag);
465 int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
467 struct gendisk *disk = sdkp->disk;
468 unsigned int nr_zones;
472 if (!sd_is_zoned(sdkp))
474 * Device managed or normal SCSI disk,
475 * no special handling required
479 /* Check zoned block device characteristics (unconstrained reads) */
480 ret = sd_zbc_check_zoned_characteristics(sdkp, buf);
485 * Check zone size: only devices with a constant zone size (except
486 * an eventual last runt zone) that is a power of 2 are supported.
488 ret = sd_zbc_check_zones(sdkp, &zone_blocks);
492 /* The drive satisfies the kernel restrictions: set it up */
493 blk_queue_chunk_sectors(sdkp->disk->queue,
494 logical_to_sectors(sdkp->device, zone_blocks));
495 blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, sdkp->disk->queue);
496 blk_queue_required_elevator_features(sdkp->disk->queue,
497 ELEVATOR_F_ZBD_SEQ_WRITE);
498 nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks);
500 /* READ16/WRITE16 is mandatory for ZBC disks */
501 sdkp->device->use_16_for_rw = 1;
502 sdkp->device->use_10_for_rw = 0;
505 * Revalidate the disk zone bitmaps once the block device capacity is
506 * set on the second revalidate execution during disk scan and if
507 * something changed when executing a normal revalidate.
509 if (sdkp->first_scan) {
510 sdkp->zone_blocks = zone_blocks;
511 sdkp->nr_zones = nr_zones;
515 if (sdkp->zone_blocks != zone_blocks ||
516 sdkp->nr_zones != nr_zones ||
517 disk->queue->nr_zones != nr_zones) {
518 ret = blk_revalidate_disk_zones(disk);
521 sdkp->zone_blocks = zone_blocks;
522 sdkp->nr_zones = nr_zones;
533 void sd_zbc_print_zones(struct scsi_disk *sdkp)
535 if (!sd_is_zoned(sdkp) || !sdkp->capacity)
538 if (sdkp->capacity & (sdkp->zone_blocks - 1))
539 sd_printk(KERN_NOTICE, sdkp,
540 "%u zones of %u logical blocks + 1 runt zone\n",
544 sd_printk(KERN_NOTICE, sdkp,
545 "%u zones of %u logical blocks\n",