1 // SPDX-License-Identifier: GPL-2.0-only
3 * sd.c Copyright (C) 1992 Drew Eckhardt
4 * Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
6 * Linux scsi disk driver
7 * Initial versions: Drew Eckhardt
8 * Subsequent revisions: Eric Youngdale
9 * Modification history:
10 * - Drew Eckhardt <drew@colorado.edu> original
11 * - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12 * outstanding request, and other enhancements.
13 * Support loadable low-level scsi drivers.
14 * - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15 * eight major numbers.
16 * - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17 * - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18 * sd_init and cleanups.
19 * - Alex Davis <letmein@erols.com> Fix problem where partition info
20 * not being read in sd_open. Fix problem where removable media
21 * could be ejected after sd_open.
22 * - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23 * - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24 * <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25 * Support 32k/1M disks.
27 * Logging policy (needs CONFIG_SCSI_LOGGING defined):
28 * - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29 * - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30 * - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31 * - entering other commands: SCSI_LOG_HLQUEUE level 3
32 * Note: when the logging level is set by the user, it must be greater
33 * than the level indicated above to trigger output.
36 #include <linux/module.h>
38 #include <linux/kernel.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 #include <scsi/scsi_common.h>
73 #include "scsi_priv.h"
74 #include "scsi_logging.h"
76 MODULE_AUTHOR("Eric Youngdale");
77 MODULE_DESCRIPTION("SCSI disk (sd) driver");
78 MODULE_LICENSE("GPL");
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
103 static void sd_config_discard(struct scsi_disk *, unsigned int);
104 static void sd_config_write_same(struct scsi_disk *);
105 static int sd_revalidate_disk(struct gendisk *);
106 static void sd_unlock_native_capacity(struct gendisk *disk);
107 static void sd_shutdown(struct device *);
108 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
109 static void scsi_disk_release(struct device *cdev);
111 static DEFINE_IDA(sd_index_ida);
113 static mempool_t *sd_page_pool;
114 static struct lock_class_key sd_bio_compl_lkclass;
116 static const char *sd_cache_types[] = {
117 "write through", "none", "write back",
118 "write back, no read (daft)"
121 static void sd_set_flush_flag(struct scsi_disk *sdkp)
123 bool wc = false, fua = false;
131 blk_queue_write_cache(sdkp->disk->queue, wc, fua);
135 cache_type_store(struct device *dev, struct device_attribute *attr,
136 const char *buf, size_t count)
138 int ct, rcd, wce, sp;
139 struct scsi_disk *sdkp = to_scsi_disk(dev);
140 struct scsi_device *sdp = sdkp->device;
143 struct scsi_mode_data data;
144 struct scsi_sense_hdr sshdr;
145 static const char temp[] = "temporary ";
148 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
149 /* no cache control on RBC devices; theoretically they
150 * can do it, but there's probably so many exceptions
151 * it's not worth the risk */
154 if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
155 buf += sizeof(temp) - 1;
156 sdkp->cache_override = 1;
158 sdkp->cache_override = 0;
161 ct = sysfs_match_string(sd_cache_types, buf);
165 rcd = ct & 0x01 ? 1 : 0;
166 wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
168 if (sdkp->cache_override) {
171 sd_set_flush_flag(sdkp);
175 if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
176 sdkp->max_retries, &data, NULL))
178 len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
179 data.block_descriptor_length);
180 buffer_data = buffer + data.header_length +
181 data.block_descriptor_length;
182 buffer_data[2] &= ~0x05;
183 buffer_data[2] |= wce << 2 | rcd;
184 sp = buffer_data[0] & 0x80 ? 1 : 0;
185 buffer_data[0] &= ~0x80;
188 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
189 * received mode parameter buffer before doing MODE SELECT.
191 data.device_specific = 0;
193 ret = scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
194 sdkp->max_retries, &data, &sshdr);
196 if (ret > 0 && scsi_sense_valid(&sshdr))
197 sd_print_sense_hdr(sdkp, &sshdr);
200 sd_revalidate_disk(sdkp->disk);
205 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
208 struct scsi_disk *sdkp = to_scsi_disk(dev);
209 struct scsi_device *sdp = sdkp->device;
211 return sprintf(buf, "%u\n", sdp->manage_start_stop);
215 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
216 const char *buf, size_t count)
218 struct scsi_disk *sdkp = to_scsi_disk(dev);
219 struct scsi_device *sdp = sdkp->device;
222 if (!capable(CAP_SYS_ADMIN))
225 if (kstrtobool(buf, &v))
228 sdp->manage_start_stop = v;
232 static DEVICE_ATTR_RW(manage_start_stop);
235 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
237 struct scsi_disk *sdkp = to_scsi_disk(dev);
239 return sprintf(buf, "%u\n", sdkp->device->allow_restart);
243 allow_restart_store(struct device *dev, struct device_attribute *attr,
244 const char *buf, size_t count)
247 struct scsi_disk *sdkp = to_scsi_disk(dev);
248 struct scsi_device *sdp = sdkp->device;
250 if (!capable(CAP_SYS_ADMIN))
253 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
256 if (kstrtobool(buf, &v))
259 sdp->allow_restart = v;
263 static DEVICE_ATTR_RW(allow_restart);
266 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
268 struct scsi_disk *sdkp = to_scsi_disk(dev);
269 int ct = sdkp->RCD + 2*sdkp->WCE;
271 return sprintf(buf, "%s\n", sd_cache_types[ct]);
273 static DEVICE_ATTR_RW(cache_type);
276 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
278 struct scsi_disk *sdkp = to_scsi_disk(dev);
280 return sprintf(buf, "%u\n", sdkp->DPOFUA);
282 static DEVICE_ATTR_RO(FUA);
285 protection_type_show(struct device *dev, struct device_attribute *attr,
288 struct scsi_disk *sdkp = to_scsi_disk(dev);
290 return sprintf(buf, "%u\n", sdkp->protection_type);
294 protection_type_store(struct device *dev, struct device_attribute *attr,
295 const char *buf, size_t count)
297 struct scsi_disk *sdkp = to_scsi_disk(dev);
301 if (!capable(CAP_SYS_ADMIN))
304 err = kstrtouint(buf, 10, &val);
309 if (val <= T10_PI_TYPE3_PROTECTION)
310 sdkp->protection_type = val;
314 static DEVICE_ATTR_RW(protection_type);
317 protection_mode_show(struct device *dev, struct device_attribute *attr,
320 struct scsi_disk *sdkp = to_scsi_disk(dev);
321 struct scsi_device *sdp = sdkp->device;
322 unsigned int dif, dix;
324 dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
325 dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
327 if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
333 return sprintf(buf, "none\n");
335 return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
337 static DEVICE_ATTR_RO(protection_mode);
340 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
342 struct scsi_disk *sdkp = to_scsi_disk(dev);
344 return sprintf(buf, "%u\n", sdkp->ATO);
346 static DEVICE_ATTR_RO(app_tag_own);
349 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
352 struct scsi_disk *sdkp = to_scsi_disk(dev);
354 return sprintf(buf, "%u\n", sdkp->lbpme);
356 static DEVICE_ATTR_RO(thin_provisioning);
358 /* sysfs_match_string() requires dense arrays */
359 static const char *lbp_mode[] = {
360 [SD_LBP_FULL] = "full",
361 [SD_LBP_UNMAP] = "unmap",
362 [SD_LBP_WS16] = "writesame_16",
363 [SD_LBP_WS10] = "writesame_10",
364 [SD_LBP_ZERO] = "writesame_zero",
365 [SD_LBP_DISABLE] = "disabled",
369 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
372 struct scsi_disk *sdkp = to_scsi_disk(dev);
374 return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
378 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
379 const char *buf, size_t count)
381 struct scsi_disk *sdkp = to_scsi_disk(dev);
382 struct scsi_device *sdp = sdkp->device;
385 if (!capable(CAP_SYS_ADMIN))
388 if (sd_is_zoned(sdkp)) {
389 sd_config_discard(sdkp, SD_LBP_DISABLE);
393 if (sdp->type != TYPE_DISK)
396 mode = sysfs_match_string(lbp_mode, buf);
400 sd_config_discard(sdkp, mode);
404 static DEVICE_ATTR_RW(provisioning_mode);
406 /* sysfs_match_string() requires dense arrays */
407 static const char *zeroing_mode[] = {
408 [SD_ZERO_WRITE] = "write",
409 [SD_ZERO_WS] = "writesame",
410 [SD_ZERO_WS16_UNMAP] = "writesame_16_unmap",
411 [SD_ZERO_WS10_UNMAP] = "writesame_10_unmap",
415 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
418 struct scsi_disk *sdkp = to_scsi_disk(dev);
420 return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
424 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
425 const char *buf, size_t count)
427 struct scsi_disk *sdkp = to_scsi_disk(dev);
430 if (!capable(CAP_SYS_ADMIN))
433 mode = sysfs_match_string(zeroing_mode, buf);
437 sdkp->zeroing_mode = mode;
441 static DEVICE_ATTR_RW(zeroing_mode);
444 max_medium_access_timeouts_show(struct device *dev,
445 struct device_attribute *attr, char *buf)
447 struct scsi_disk *sdkp = to_scsi_disk(dev);
449 return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
453 max_medium_access_timeouts_store(struct device *dev,
454 struct device_attribute *attr, const char *buf,
457 struct scsi_disk *sdkp = to_scsi_disk(dev);
460 if (!capable(CAP_SYS_ADMIN))
463 err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
465 return err ? err : count;
467 static DEVICE_ATTR_RW(max_medium_access_timeouts);
470 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
473 struct scsi_disk *sdkp = to_scsi_disk(dev);
475 return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
479 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
480 const char *buf, size_t count)
482 struct scsi_disk *sdkp = to_scsi_disk(dev);
483 struct scsi_device *sdp = sdkp->device;
487 if (!capable(CAP_SYS_ADMIN))
490 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
493 err = kstrtoul(buf, 10, &max);
499 sdp->no_write_same = 1;
500 else if (max <= SD_MAX_WS16_BLOCKS) {
501 sdp->no_write_same = 0;
502 sdkp->max_ws_blocks = max;
505 sd_config_write_same(sdkp);
509 static DEVICE_ATTR_RW(max_write_same_blocks);
512 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
514 struct scsi_disk *sdkp = to_scsi_disk(dev);
516 if (sdkp->device->type == TYPE_ZBC)
517 return sprintf(buf, "host-managed\n");
518 if (sdkp->zoned == 1)
519 return sprintf(buf, "host-aware\n");
520 if (sdkp->zoned == 2)
521 return sprintf(buf, "drive-managed\n");
522 return sprintf(buf, "none\n");
524 static DEVICE_ATTR_RO(zoned_cap);
527 max_retries_store(struct device *dev, struct device_attribute *attr,
528 const char *buf, size_t count)
530 struct scsi_disk *sdkp = to_scsi_disk(dev);
531 struct scsi_device *sdev = sdkp->device;
534 err = kstrtoint(buf, 10, &retries);
538 if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
539 sdkp->max_retries = retries;
543 sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
549 max_retries_show(struct device *dev, struct device_attribute *attr,
552 struct scsi_disk *sdkp = to_scsi_disk(dev);
554 return sprintf(buf, "%d\n", sdkp->max_retries);
557 static DEVICE_ATTR_RW(max_retries);
559 static struct attribute *sd_disk_attrs[] = {
560 &dev_attr_cache_type.attr,
562 &dev_attr_allow_restart.attr,
563 &dev_attr_manage_start_stop.attr,
564 &dev_attr_protection_type.attr,
565 &dev_attr_protection_mode.attr,
566 &dev_attr_app_tag_own.attr,
567 &dev_attr_thin_provisioning.attr,
568 &dev_attr_provisioning_mode.attr,
569 &dev_attr_zeroing_mode.attr,
570 &dev_attr_max_write_same_blocks.attr,
571 &dev_attr_max_medium_access_timeouts.attr,
572 &dev_attr_zoned_cap.attr,
573 &dev_attr_max_retries.attr,
576 ATTRIBUTE_GROUPS(sd_disk);
578 static struct class sd_disk_class = {
580 .dev_release = scsi_disk_release,
581 .dev_groups = sd_disk_groups,
585 * Don't request a new module, as that could deadlock in multipath
588 static void sd_default_probe(dev_t devt)
593 * Device no to disk mapping:
595 * major disc2 disc p1
596 * |............|.............|....|....| <- dev_t
599 * Inside a major, we have 16k disks, however mapped non-
600 * contiguously. The first 16 disks are for major0, the next
601 * ones with major1, ... Disk 256 is for major0 again, disk 272
603 * As we stay compatible with our numbering scheme, we can reuse
604 * the well-know SCSI majors 8, 65--71, 136--143.
606 static int sd_major(int major_idx)
610 return SCSI_DISK0_MAJOR;
612 return SCSI_DISK1_MAJOR + major_idx - 1;
614 return SCSI_DISK8_MAJOR + major_idx - 8;
617 return 0; /* shut up gcc */
621 #ifdef CONFIG_BLK_SED_OPAL
622 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
623 size_t len, bool send)
625 struct scsi_disk *sdkp = data;
626 struct scsi_device *sdev = sdkp->device;
628 const struct scsi_exec_args exec_args = {
629 .req_flags = BLK_MQ_REQ_PM,
633 cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
635 put_unaligned_be16(spsp, &cdb[2]);
636 put_unaligned_be32(len, &cdb[6]);
638 ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
639 buffer, len, SD_TIMEOUT, sdkp->max_retries,
641 return ret <= 0 ? ret : -EIO;
643 #endif /* CONFIG_BLK_SED_OPAL */
646 * Look up the DIX operation based on whether the command is read or
647 * write and whether dix and dif are enabled.
649 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
651 /* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
652 static const unsigned int ops[] = { /* wrt dix dif */
653 SCSI_PROT_NORMAL, /* 0 0 0 */
654 SCSI_PROT_READ_STRIP, /* 0 0 1 */
655 SCSI_PROT_READ_INSERT, /* 0 1 0 */
656 SCSI_PROT_READ_PASS, /* 0 1 1 */
657 SCSI_PROT_NORMAL, /* 1 0 0 */
658 SCSI_PROT_WRITE_INSERT, /* 1 0 1 */
659 SCSI_PROT_WRITE_STRIP, /* 1 1 0 */
660 SCSI_PROT_WRITE_PASS, /* 1 1 1 */
663 return ops[write << 2 | dix << 1 | dif];
667 * Returns a mask of the protection flags that are valid for a given DIX
670 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
672 static const unsigned int flag_mask[] = {
673 [SCSI_PROT_NORMAL] = 0,
675 [SCSI_PROT_READ_STRIP] = SCSI_PROT_TRANSFER_PI |
676 SCSI_PROT_GUARD_CHECK |
677 SCSI_PROT_REF_CHECK |
678 SCSI_PROT_REF_INCREMENT,
680 [SCSI_PROT_READ_INSERT] = SCSI_PROT_REF_INCREMENT |
681 SCSI_PROT_IP_CHECKSUM,
683 [SCSI_PROT_READ_PASS] = SCSI_PROT_TRANSFER_PI |
684 SCSI_PROT_GUARD_CHECK |
685 SCSI_PROT_REF_CHECK |
686 SCSI_PROT_REF_INCREMENT |
687 SCSI_PROT_IP_CHECKSUM,
689 [SCSI_PROT_WRITE_INSERT] = SCSI_PROT_TRANSFER_PI |
690 SCSI_PROT_REF_INCREMENT,
692 [SCSI_PROT_WRITE_STRIP] = SCSI_PROT_GUARD_CHECK |
693 SCSI_PROT_REF_CHECK |
694 SCSI_PROT_REF_INCREMENT |
695 SCSI_PROT_IP_CHECKSUM,
697 [SCSI_PROT_WRITE_PASS] = SCSI_PROT_TRANSFER_PI |
698 SCSI_PROT_GUARD_CHECK |
699 SCSI_PROT_REF_CHECK |
700 SCSI_PROT_REF_INCREMENT |
701 SCSI_PROT_IP_CHECKSUM,
704 return flag_mask[prot_op];
707 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
708 unsigned int dix, unsigned int dif)
710 struct request *rq = scsi_cmd_to_rq(scmd);
711 struct bio *bio = rq->bio;
712 unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
713 unsigned int protect = 0;
715 if (dix) { /* DIX Type 0, 1, 2, 3 */
716 if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
717 scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
719 if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
720 scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
723 if (dif != T10_PI_TYPE3_PROTECTION) { /* DIX/DIF Type 0, 1, 2 */
724 scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
726 if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
727 scmd->prot_flags |= SCSI_PROT_REF_CHECK;
730 if (dif) { /* DIX/DIF Type 1, 2, 3 */
731 scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
733 if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
734 protect = 3 << 5; /* Disable target PI checking */
736 protect = 1 << 5; /* Enable target PI checking */
739 scsi_set_prot_op(scmd, prot_op);
740 scsi_set_prot_type(scmd, dif);
741 scmd->prot_flags &= sd_prot_flag_mask(prot_op);
746 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
748 struct request_queue *q = sdkp->disk->queue;
749 unsigned int logical_block_size = sdkp->device->sector_size;
750 unsigned int max_blocks = 0;
752 q->limits.discard_alignment =
753 sdkp->unmap_alignment * logical_block_size;
754 q->limits.discard_granularity =
755 max(sdkp->physical_block_size,
756 sdkp->unmap_granularity * logical_block_size);
757 sdkp->provisioning_mode = mode;
763 blk_queue_max_discard_sectors(q, 0);
767 max_blocks = min_not_zero(sdkp->max_unmap_blocks,
768 (u32)SD_MAX_WS16_BLOCKS);
772 if (sdkp->device->unmap_limit_for_ws)
773 max_blocks = sdkp->max_unmap_blocks;
775 max_blocks = sdkp->max_ws_blocks;
777 max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
781 if (sdkp->device->unmap_limit_for_ws)
782 max_blocks = sdkp->max_unmap_blocks;
784 max_blocks = sdkp->max_ws_blocks;
786 max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
790 max_blocks = min_not_zero(sdkp->max_ws_blocks,
791 (u32)SD_MAX_WS10_BLOCKS);
795 blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
798 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
802 page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
805 clear_highpage(page);
806 bvec_set_page(&rq->special_vec, page, data_len, 0);
807 rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
808 return bvec_virt(&rq->special_vec);
811 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
813 struct scsi_device *sdp = cmd->device;
814 struct request *rq = scsi_cmd_to_rq(cmd);
815 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
816 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
817 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
818 unsigned int data_len = 24;
821 buf = sd_set_special_bvec(rq, data_len);
823 return BLK_STS_RESOURCE;
826 cmd->cmnd[0] = UNMAP;
829 put_unaligned_be16(6 + 16, &buf[0]);
830 put_unaligned_be16(16, &buf[2]);
831 put_unaligned_be64(lba, &buf[8]);
832 put_unaligned_be32(nr_blocks, &buf[16]);
834 cmd->allowed = sdkp->max_retries;
835 cmd->transfersize = data_len;
836 rq->timeout = SD_TIMEOUT;
838 return scsi_alloc_sgtables(cmd);
841 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
844 struct scsi_device *sdp = cmd->device;
845 struct request *rq = scsi_cmd_to_rq(cmd);
846 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
847 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
848 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
849 u32 data_len = sdp->sector_size;
851 if (!sd_set_special_bvec(rq, data_len))
852 return BLK_STS_RESOURCE;
855 cmd->cmnd[0] = WRITE_SAME_16;
857 cmd->cmnd[1] = 0x8; /* UNMAP */
858 put_unaligned_be64(lba, &cmd->cmnd[2]);
859 put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
861 cmd->allowed = sdkp->max_retries;
862 cmd->transfersize = data_len;
863 rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
865 return scsi_alloc_sgtables(cmd);
868 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
871 struct scsi_device *sdp = cmd->device;
872 struct request *rq = scsi_cmd_to_rq(cmd);
873 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
874 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
875 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
876 u32 data_len = sdp->sector_size;
878 if (!sd_set_special_bvec(rq, data_len))
879 return BLK_STS_RESOURCE;
882 cmd->cmnd[0] = WRITE_SAME;
884 cmd->cmnd[1] = 0x8; /* UNMAP */
885 put_unaligned_be32(lba, &cmd->cmnd[2]);
886 put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
888 cmd->allowed = sdkp->max_retries;
889 cmd->transfersize = data_len;
890 rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
892 return scsi_alloc_sgtables(cmd);
895 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
897 struct request *rq = scsi_cmd_to_rq(cmd);
898 struct scsi_device *sdp = cmd->device;
899 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
900 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
901 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
903 if (!(rq->cmd_flags & REQ_NOUNMAP)) {
904 switch (sdkp->zeroing_mode) {
905 case SD_ZERO_WS16_UNMAP:
906 return sd_setup_write_same16_cmnd(cmd, true);
907 case SD_ZERO_WS10_UNMAP:
908 return sd_setup_write_same10_cmnd(cmd, true);
912 if (sdp->no_write_same) {
913 rq->rq_flags |= RQF_QUIET;
914 return BLK_STS_TARGET;
917 if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
918 return sd_setup_write_same16_cmnd(cmd, false);
920 return sd_setup_write_same10_cmnd(cmd, false);
923 static void sd_config_write_same(struct scsi_disk *sdkp)
925 struct request_queue *q = sdkp->disk->queue;
926 unsigned int logical_block_size = sdkp->device->sector_size;
928 if (sdkp->device->no_write_same) {
929 sdkp->max_ws_blocks = 0;
933 /* Some devices can not handle block counts above 0xffff despite
934 * supporting WRITE SAME(16). Consequently we default to 64k
935 * blocks per I/O unless the device explicitly advertises a
938 if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
939 sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
940 (u32)SD_MAX_WS16_BLOCKS);
941 else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
942 sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
943 (u32)SD_MAX_WS10_BLOCKS);
945 sdkp->device->no_write_same = 1;
946 sdkp->max_ws_blocks = 0;
949 if (sdkp->lbprz && sdkp->lbpws)
950 sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
951 else if (sdkp->lbprz && sdkp->lbpws10)
952 sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
953 else if (sdkp->max_ws_blocks)
954 sdkp->zeroing_mode = SD_ZERO_WS;
956 sdkp->zeroing_mode = SD_ZERO_WRITE;
958 if (sdkp->max_ws_blocks &&
959 sdkp->physical_block_size > logical_block_size) {
961 * Reporting a maximum number of blocks that is not aligned
962 * on the device physical size would cause a large write same
963 * request to be split into physically unaligned chunks by
964 * __blkdev_issue_write_zeroes() even if the caller of this
965 * functions took care to align the large request. So make sure
966 * the maximum reported is aligned to the device physical block
967 * size. This is only an optional optimization for regular
968 * disks, but this is mandatory to avoid failure of large write
969 * same requests directed at sequential write required zones of
970 * host-managed ZBC disks.
972 sdkp->max_ws_blocks =
973 round_down(sdkp->max_ws_blocks,
974 bytes_to_logical(sdkp->device,
975 sdkp->physical_block_size));
979 blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
980 (logical_block_size >> 9));
983 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
985 struct request *rq = scsi_cmd_to_rq(cmd);
986 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
988 /* flush requests don't perform I/O, zero the S/G table */
989 memset(&cmd->sdb, 0, sizeof(cmd->sdb));
991 if (cmd->device->use_16_for_sync) {
992 cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
995 cmd->cmnd[0] = SYNCHRONIZE_CACHE;
998 cmd->transfersize = 0;
999 cmd->allowed = sdkp->max_retries;
1001 rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1005 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1006 sector_t lba, unsigned int nr_blocks,
1007 unsigned char flags, unsigned int dld)
1009 cmd->cmd_len = SD_EXT_CDB_SIZE;
1010 cmd->cmnd[0] = VARIABLE_LENGTH_CMD;
1011 cmd->cmnd[7] = 0x18; /* Additional CDB len */
1012 cmd->cmnd[9] = write ? WRITE_32 : READ_32;
1013 cmd->cmnd[10] = flags;
1014 cmd->cmnd[11] = dld & 0x07;
1015 put_unaligned_be64(lba, &cmd->cmnd[12]);
1016 put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1017 put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1022 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1023 sector_t lba, unsigned int nr_blocks,
1024 unsigned char flags, unsigned int dld)
1027 cmd->cmnd[0] = write ? WRITE_16 : READ_16;
1028 cmd->cmnd[1] = flags | ((dld >> 2) & 0x01);
1029 cmd->cmnd[14] = (dld & 0x03) << 6;
1031 put_unaligned_be64(lba, &cmd->cmnd[2]);
1032 put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1037 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1038 sector_t lba, unsigned int nr_blocks,
1039 unsigned char flags)
1042 cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1043 cmd->cmnd[1] = flags;
1046 put_unaligned_be32(lba, &cmd->cmnd[2]);
1047 put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1052 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1053 sector_t lba, unsigned int nr_blocks,
1054 unsigned char flags)
1056 /* Avoid that 0 blocks gets translated into 256 blocks. */
1057 if (WARN_ON_ONCE(nr_blocks == 0))
1058 return BLK_STS_IOERR;
1060 if (unlikely(flags & 0x8)) {
1062 * This happens only if this drive failed 10byte rw
1063 * command with ILLEGAL_REQUEST during operation and
1064 * thus turned off use_10_for_rw.
1066 scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1067 return BLK_STS_IOERR;
1071 cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1072 cmd->cmnd[1] = (lba >> 16) & 0x1f;
1073 cmd->cmnd[2] = (lba >> 8) & 0xff;
1074 cmd->cmnd[3] = lba & 0xff;
1075 cmd->cmnd[4] = nr_blocks;
1082 * Check if a command has a duration limit set. If it does, and the target
1083 * device supports CDL and the feature is enabled, return the limit
1084 * descriptor index to use. Return 0 (no limit) otherwise.
1086 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1088 struct scsi_device *sdp = sdkp->device;
1091 if (!sdp->cdl_supported || !sdp->cdl_enable)
1095 * Use "no limit" if the request ioprio does not specify a duration
1098 hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1099 if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1100 hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1103 return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1106 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1108 struct request *rq = scsi_cmd_to_rq(cmd);
1109 struct scsi_device *sdp = cmd->device;
1110 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1111 sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1113 unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1114 unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1115 bool write = rq_data_dir(rq) == WRITE;
1116 unsigned char protect, fua;
1122 ret = scsi_alloc_sgtables(cmd);
1123 if (ret != BLK_STS_OK)
1126 ret = BLK_STS_IOERR;
1127 if (!scsi_device_online(sdp) || sdp->changed) {
1128 scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1132 if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1133 scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1137 if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1138 scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1143 * Some SD card readers can't handle accesses which touch the
1144 * last one or two logical blocks. Split accesses as needed.
1146 threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1148 if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1149 if (lba < threshold) {
1150 /* Access up to the threshold but not beyond */
1151 nr_blocks = threshold - lba;
1153 /* Access only a single logical block */
1158 if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1159 ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1164 fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1165 dix = scsi_prot_sg_count(cmd);
1166 dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1167 dld = sd_cdl_dld(sdkp, cmd);
1170 protect = sd_setup_protect_cmnd(cmd, dix, dif);
1174 if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1175 ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1176 protect | fua, dld);
1177 } else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1178 ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1179 protect | fua, dld);
1180 } else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1181 sdp->use_10_for_rw || protect) {
1182 ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1185 ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1189 if (unlikely(ret != BLK_STS_OK))
1193 * We shouldn't disconnect in the middle of a sector, so with a dumb
1194 * host adapter, it's safe to assume that we can at least transfer
1195 * this many bytes between each connect / disconnect.
1197 cmd->transfersize = sdp->sector_size;
1198 cmd->underflow = nr_blocks << 9;
1199 cmd->allowed = sdkp->max_retries;
1200 cmd->sdb.length = nr_blocks * sdp->sector_size;
1203 scmd_printk(KERN_INFO, cmd,
1204 "%s: block=%llu, count=%d\n", __func__,
1205 (unsigned long long)blk_rq_pos(rq),
1206 blk_rq_sectors(rq)));
1208 scmd_printk(KERN_INFO, cmd,
1209 "%s %d/%u 512 byte blocks.\n",
1210 write ? "writing" : "reading", nr_blocks,
1211 blk_rq_sectors(rq)));
1214 * This indicates that the command is ready from our end to be queued.
1218 scsi_free_sgtables(cmd);
1222 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1224 struct request *rq = scsi_cmd_to_rq(cmd);
1226 switch (req_op(rq)) {
1227 case REQ_OP_DISCARD:
1228 switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1230 return sd_setup_unmap_cmnd(cmd);
1232 return sd_setup_write_same16_cmnd(cmd, true);
1234 return sd_setup_write_same10_cmnd(cmd, true);
1236 return sd_setup_write_same10_cmnd(cmd, false);
1238 return BLK_STS_TARGET;
1240 case REQ_OP_WRITE_ZEROES:
1241 return sd_setup_write_zeroes_cmnd(cmd);
1243 return sd_setup_flush_cmnd(cmd);
1246 case REQ_OP_ZONE_APPEND:
1247 return sd_setup_read_write_cmnd(cmd);
1248 case REQ_OP_ZONE_RESET:
1249 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1251 case REQ_OP_ZONE_RESET_ALL:
1252 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1254 case REQ_OP_ZONE_OPEN:
1255 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1256 case REQ_OP_ZONE_CLOSE:
1257 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1258 case REQ_OP_ZONE_FINISH:
1259 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1262 return BLK_STS_NOTSUPP;
1266 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1268 struct request *rq = scsi_cmd_to_rq(SCpnt);
1270 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1271 mempool_free(rq->special_vec.bv_page, sd_page_pool);
1274 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1276 if (sdkp->device->removable || sdkp->write_prot) {
1277 if (disk_check_media_change(disk))
1282 * Force a full rescan after ioctl(BLKRRPART). While the disk state has
1283 * nothing to do with partitions, BLKRRPART is used to force a full
1284 * revalidate after things like a format for historical reasons.
1286 return test_bit(GD_NEED_PART_SCAN, &disk->state);
1290 * sd_open - open a scsi disk device
1291 * @disk: disk to open
1294 * Returns 0 if successful. Returns a negated errno value in case
1297 * Note: This can be called from a user context (e.g. fsck(1) )
1298 * or from within the kernel (e.g. as a result of a mount(1) ).
1299 * In the latter case @inode and @filp carry an abridged amount
1300 * of information as noted above.
1302 * Locking: called with disk->open_mutex held.
1304 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1306 struct scsi_disk *sdkp = scsi_disk(disk);
1307 struct scsi_device *sdev = sdkp->device;
1310 if (scsi_device_get(sdev))
1313 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1316 * If the device is in error recovery, wait until it is done.
1317 * If the device is offline, then disallow any access to it.
1320 if (!scsi_block_when_processing_errors(sdev))
1323 if (sd_need_revalidate(disk, sdkp))
1324 sd_revalidate_disk(disk);
1327 * If the drive is empty, just let the open fail.
1329 retval = -ENOMEDIUM;
1330 if (sdev->removable && !sdkp->media_present &&
1331 !(mode & BLK_OPEN_NDELAY))
1335 * If the device has the write protect tab set, have the open fail
1336 * if the user expects to be able to write to the thing.
1339 if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1343 * It is possible that the disk changing stuff resulted in
1344 * the device being taken offline. If this is the case,
1345 * report this to the user, and don't pretend that the
1346 * open actually succeeded.
1349 if (!scsi_device_online(sdev))
1352 if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1353 if (scsi_block_when_processing_errors(sdev))
1354 scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1360 scsi_device_put(sdev);
1365 * sd_release - invoked when the (last) close(2) is called on this
1367 * @disk: disk to release
1371 * Note: may block (uninterruptible) if error recovery is underway
1374 * Locking: called with disk->open_mutex held.
1376 static void sd_release(struct gendisk *disk)
1378 struct scsi_disk *sdkp = scsi_disk(disk);
1379 struct scsi_device *sdev = sdkp->device;
1381 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1383 if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1384 if (scsi_block_when_processing_errors(sdev))
1385 scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1388 scsi_device_put(sdev);
1391 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1393 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1394 struct scsi_device *sdp = sdkp->device;
1395 struct Scsi_Host *host = sdp->host;
1396 sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1399 /* default to most commonly used values */
1400 diskinfo[0] = 0x40; /* 1 << 6 */
1401 diskinfo[1] = 0x20; /* 1 << 5 */
1402 diskinfo[2] = capacity >> 11;
1404 /* override with calculated, extended default, or driver values */
1405 if (host->hostt->bios_param)
1406 host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1408 scsicam_bios_param(bdev, capacity, diskinfo);
1410 geo->heads = diskinfo[0];
1411 geo->sectors = diskinfo[1];
1412 geo->cylinders = diskinfo[2];
1417 * sd_ioctl - process an ioctl
1418 * @bdev: target block device
1420 * @cmd: ioctl command number
1421 * @arg: this is third argument given to ioctl(2) system call.
1422 * Often contains a pointer.
1424 * Returns 0 if successful (some ioctls return positive numbers on
1425 * success as well). Returns a negated errno value in case of error.
1427 * Note: most ioctls are forward onto the block subsystem or further
1428 * down in the scsi subsystem.
1430 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1431 unsigned int cmd, unsigned long arg)
1433 struct gendisk *disk = bdev->bd_disk;
1434 struct scsi_disk *sdkp = scsi_disk(disk);
1435 struct scsi_device *sdp = sdkp->device;
1436 void __user *p = (void __user *)arg;
1439 SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1440 "cmd=0x%x\n", disk->disk_name, cmd));
1442 if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1443 return -ENOIOCTLCMD;
1446 * If we are in the middle of error recovery, don't let anyone
1447 * else try and use this device. Also, if error recovery fails, it
1448 * may try and take the device offline, in which case all further
1449 * access to the device is prohibited.
1451 error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1452 (mode & BLK_OPEN_NDELAY));
1456 if (is_sed_ioctl(cmd))
1457 return sed_ioctl(sdkp->opal_dev, cmd, p);
1458 return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1461 static void set_media_not_present(struct scsi_disk *sdkp)
1463 if (sdkp->media_present)
1464 sdkp->device->changed = 1;
1466 if (sdkp->device->removable) {
1467 sdkp->media_present = 0;
1472 static int media_not_present(struct scsi_disk *sdkp,
1473 struct scsi_sense_hdr *sshdr)
1475 if (!scsi_sense_valid(sshdr))
1478 /* not invoked for commands that could return deferred errors */
1479 switch (sshdr->sense_key) {
1480 case UNIT_ATTENTION:
1482 /* medium not present */
1483 if (sshdr->asc == 0x3A) {
1484 set_media_not_present(sdkp);
1492 * sd_check_events - check media events
1493 * @disk: kernel device descriptor
1494 * @clearing: disk events currently being cleared
1496 * Returns mask of DISK_EVENT_*.
1498 * Note: this function is invoked from the block subsystem.
1500 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1502 struct scsi_disk *sdkp = disk->private_data;
1503 struct scsi_device *sdp;
1511 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1514 * If the device is offline, don't send any commands - just pretend as
1515 * if the command failed. If the device ever comes back online, we
1516 * can deal with it then. It is only because of unrecoverable errors
1517 * that we would ever take a device offline in the first place.
1519 if (!scsi_device_online(sdp)) {
1520 set_media_not_present(sdkp);
1525 * Using TEST_UNIT_READY enables differentiation between drive with
1526 * no cartridge loaded - NOT READY, drive with changed cartridge -
1527 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1529 * Drives that auto spin down. eg iomega jaz 1G, will be started
1530 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1531 * sd_revalidate() is called.
1533 if (scsi_block_when_processing_errors(sdp)) {
1534 struct scsi_sense_hdr sshdr = { 0, };
1536 retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1539 /* failed to execute TUR, assume media not present */
1540 if (retval < 0 || host_byte(retval)) {
1541 set_media_not_present(sdkp);
1545 if (media_not_present(sdkp, &sshdr))
1550 * For removable scsi disk we have to recognise the presence
1551 * of a disk in the drive.
1553 if (!sdkp->media_present)
1555 sdkp->media_present = 1;
1558 * sdp->changed is set under the following conditions:
1560 * Medium present state has changed in either direction.
1561 * Device has indicated UNIT_ATTENTION.
1563 disk_changed = sdp->changed;
1565 return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1568 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1571 struct scsi_device *sdp = sdkp->device;
1572 const int timeout = sdp->request_queue->rq_timeout
1573 * SD_FLUSH_TIMEOUT_MULTIPLIER;
1574 struct scsi_sense_hdr my_sshdr;
1575 const struct scsi_exec_args exec_args = {
1576 .req_flags = BLK_MQ_REQ_PM,
1577 /* caller might not be interested in sense, but we need it */
1578 .sshdr = sshdr ? : &my_sshdr,
1581 if (!scsi_device_online(sdp))
1584 sshdr = exec_args.sshdr;
1586 for (retries = 3; retries > 0; --retries) {
1587 unsigned char cmd[16] = { 0 };
1589 if (sdp->use_16_for_sync)
1590 cmd[0] = SYNCHRONIZE_CACHE_16;
1592 cmd[0] = SYNCHRONIZE_CACHE;
1594 * Leave the rest of the command zero to indicate
1597 res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1598 timeout, sdkp->max_retries, &exec_args);
1604 sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1609 if (scsi_status_is_check_condition(res) &&
1610 scsi_sense_valid(sshdr)) {
1611 sd_print_sense_hdr(sdkp, sshdr);
1613 /* we need to evaluate the error return */
1614 if (sshdr->asc == 0x3a || /* medium not present */
1615 sshdr->asc == 0x20 || /* invalid command */
1616 (sshdr->asc == 0x74 && sshdr->ascq == 0x71)) /* drive is password locked */
1617 /* this is no error here */
1621 switch (host_byte(res)) {
1622 /* ignore errors due to racing a disconnection */
1623 case DID_BAD_TARGET:
1624 case DID_NO_CONNECT:
1626 /* signal the upper layer it might try again */
1630 case DID_SOFT_ERROR:
1639 static void sd_rescan(struct device *dev)
1641 struct scsi_disk *sdkp = dev_get_drvdata(dev);
1643 sd_revalidate_disk(sdkp->disk);
1646 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1647 enum blk_unique_id type)
1649 struct scsi_device *sdev = scsi_disk(disk)->device;
1650 const struct scsi_vpd *vpd;
1651 const unsigned char *d;
1652 int ret = -ENXIO, len;
1655 vpd = rcu_dereference(sdev->vpd_pg83);
1660 for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1661 /* we only care about designators with LU association */
1662 if (((d[1] >> 4) & 0x3) != 0x00)
1664 if ((d[1] & 0xf) != type)
1668 * Only exit early if a 16-byte descriptor was found. Otherwise
1669 * keep looking as one with more entropy might still show up.
1672 if (len != 8 && len != 12 && len != 16)
1675 memcpy(id, d + 4, len);
1684 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1686 switch (host_byte(result)) {
1687 case DID_TRANSPORT_MARGINAL:
1688 case DID_TRANSPORT_DISRUPTED:
1690 return PR_STS_RETRY_PATH_FAILURE;
1691 case DID_NO_CONNECT:
1692 return PR_STS_PATH_FAILED;
1693 case DID_TRANSPORT_FAILFAST:
1694 return PR_STS_PATH_FAST_FAILED;
1697 switch (status_byte(result)) {
1698 case SAM_STAT_RESERVATION_CONFLICT:
1699 return PR_STS_RESERVATION_CONFLICT;
1700 case SAM_STAT_CHECK_CONDITION:
1701 if (!scsi_sense_valid(sshdr))
1702 return PR_STS_IOERR;
1704 if (sshdr->sense_key == ILLEGAL_REQUEST &&
1705 (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1710 return PR_STS_IOERR;
1714 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1715 unsigned char *data, int data_len)
1717 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1718 struct scsi_device *sdev = sdkp->device;
1719 struct scsi_sense_hdr sshdr;
1720 u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1721 const struct scsi_exec_args exec_args = {
1726 put_unaligned_be16(data_len, &cmd[7]);
1728 result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1729 SD_TIMEOUT, sdkp->max_retries, &exec_args);
1730 if (scsi_status_is_check_condition(result) &&
1731 scsi_sense_valid(&sshdr)) {
1732 sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1733 scsi_print_sense_hdr(sdev, NULL, &sshdr);
1739 return sd_scsi_to_pr_err(&sshdr, result);
1742 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1744 int result, i, data_offset, num_copy_keys;
1745 u32 num_keys = keys_info->num_keys;
1746 int data_len = num_keys * 8 + 8;
1749 data = kzalloc(data_len, GFP_KERNEL);
1753 result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1757 keys_info->generation = get_unaligned_be32(&data[0]);
1758 keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1761 num_copy_keys = min(num_keys, keys_info->num_keys);
1763 for (i = 0; i < num_copy_keys; i++) {
1764 keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1773 static int sd_pr_read_reservation(struct block_device *bdev,
1774 struct pr_held_reservation *rsv)
1776 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1777 struct scsi_device *sdev = sdkp->device;
1781 result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1785 len = get_unaligned_be32(&data[4]);
1789 /* Make sure we have at least the key and type */
1791 sdev_printk(KERN_INFO, sdev,
1792 "READ RESERVATION failed due to short return buffer of %d bytes\n",
1797 rsv->generation = get_unaligned_be32(&data[0]);
1798 rsv->key = get_unaligned_be64(&data[8]);
1799 rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1803 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1804 u64 sa_key, enum scsi_pr_type type, u8 flags)
1806 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1807 struct scsi_device *sdev = sdkp->device;
1808 struct scsi_sense_hdr sshdr;
1809 const struct scsi_exec_args exec_args = {
1813 u8 cmd[16] = { 0, };
1814 u8 data[24] = { 0, };
1816 cmd[0] = PERSISTENT_RESERVE_OUT;
1819 put_unaligned_be32(sizeof(data), &cmd[5]);
1821 put_unaligned_be64(key, &data[0]);
1822 put_unaligned_be64(sa_key, &data[8]);
1825 result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1826 sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1829 if (scsi_status_is_check_condition(result) &&
1830 scsi_sense_valid(&sshdr)) {
1831 sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1832 scsi_print_sense_hdr(sdev, NULL, &sshdr);
1838 return sd_scsi_to_pr_err(&sshdr, result);
1841 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1844 if (flags & ~PR_FL_IGNORE_KEY)
1846 return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1847 old_key, new_key, 0,
1848 (1 << 0) /* APTPL */);
1851 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1856 return sd_pr_out_command(bdev, 0x01, key, 0,
1857 block_pr_type_to_scsi(type), 0);
1860 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1862 return sd_pr_out_command(bdev, 0x02, key, 0,
1863 block_pr_type_to_scsi(type), 0);
1866 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1867 enum pr_type type, bool abort)
1869 return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1870 block_pr_type_to_scsi(type), 0);
1873 static int sd_pr_clear(struct block_device *bdev, u64 key)
1875 return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
1878 static const struct pr_ops sd_pr_ops = {
1879 .pr_register = sd_pr_register,
1880 .pr_reserve = sd_pr_reserve,
1881 .pr_release = sd_pr_release,
1882 .pr_preempt = sd_pr_preempt,
1883 .pr_clear = sd_pr_clear,
1884 .pr_read_keys = sd_pr_read_keys,
1885 .pr_read_reservation = sd_pr_read_reservation,
1888 static void scsi_disk_free_disk(struct gendisk *disk)
1890 struct scsi_disk *sdkp = scsi_disk(disk);
1892 put_device(&sdkp->disk_dev);
1895 static const struct block_device_operations sd_fops = {
1896 .owner = THIS_MODULE,
1898 .release = sd_release,
1900 .getgeo = sd_getgeo,
1901 .compat_ioctl = blkdev_compat_ptr_ioctl,
1902 .check_events = sd_check_events,
1903 .unlock_native_capacity = sd_unlock_native_capacity,
1904 .report_zones = sd_zbc_report_zones,
1905 .get_unique_id = sd_get_unique_id,
1906 .free_disk = scsi_disk_free_disk,
1907 .pr_ops = &sd_pr_ops,
1911 * sd_eh_reset - reset error handling callback
1912 * @scmd: sd-issued command that has failed
1914 * This function is called by the SCSI midlayer before starting
1915 * SCSI EH. When counting medium access failures we have to be
1916 * careful to register it only only once per device and SCSI EH run;
1917 * there might be several timed out commands which will cause the
1918 * 'max_medium_access_timeouts' counter to trigger after the first
1919 * SCSI EH run already and set the device to offline.
1920 * So this function resets the internal counter before starting SCSI EH.
1922 static void sd_eh_reset(struct scsi_cmnd *scmd)
1924 struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1926 /* New SCSI EH run, reset gate variable */
1927 sdkp->ignore_medium_access_errors = false;
1931 * sd_eh_action - error handling callback
1932 * @scmd: sd-issued command that has failed
1933 * @eh_disp: The recovery disposition suggested by the midlayer
1935 * This function is called by the SCSI midlayer upon completion of an
1936 * error test command (currently TEST UNIT READY). The result of sending
1937 * the eh command is passed in eh_disp. We're looking for devices that
1938 * fail medium access commands but are OK with non access commands like
1939 * test unit ready (so wrongly see the device as having a successful
1942 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1944 struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1945 struct scsi_device *sdev = scmd->device;
1947 if (!scsi_device_online(sdev) ||
1948 !scsi_medium_access_command(scmd) ||
1949 host_byte(scmd->result) != DID_TIME_OUT ||
1954 * The device has timed out executing a medium access command.
1955 * However, the TEST UNIT READY command sent during error
1956 * handling completed successfully. Either the device is in the
1957 * process of recovering or has it suffered an internal failure
1958 * that prevents access to the storage medium.
1960 if (!sdkp->ignore_medium_access_errors) {
1961 sdkp->medium_access_timed_out++;
1962 sdkp->ignore_medium_access_errors = true;
1966 * If the device keeps failing read/write commands but TEST UNIT
1967 * READY always completes successfully we assume that medium
1968 * access is no longer possible and take the device offline.
1970 if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1971 scmd_printk(KERN_ERR, scmd,
1972 "Medium access timeout failure. Offlining disk!\n");
1973 mutex_lock(&sdev->state_mutex);
1974 scsi_device_set_state(sdev, SDEV_OFFLINE);
1975 mutex_unlock(&sdev->state_mutex);
1983 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1985 struct request *req = scsi_cmd_to_rq(scmd);
1986 struct scsi_device *sdev = scmd->device;
1987 unsigned int transferred, good_bytes;
1988 u64 start_lba, end_lba, bad_lba;
1991 * Some commands have a payload smaller than the device logical
1992 * block size (e.g. INQUIRY on a 4K disk).
1994 if (scsi_bufflen(scmd) <= sdev->sector_size)
1997 /* Check if we have a 'bad_lba' information */
1998 if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1999 SCSI_SENSE_BUFFERSIZE,
2004 * If the bad lba was reported incorrectly, we have no idea where
2007 start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2008 end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2009 if (bad_lba < start_lba || bad_lba >= end_lba)
2013 * resid is optional but mostly filled in. When it's unused,
2014 * its value is zero, so we assume the whole buffer transferred
2016 transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2018 /* This computation should always be done in terms of the
2019 * resolution of the device's medium.
2021 good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2023 return min(good_bytes, transferred);
2027 * sd_done - bottom half handler: called when the lower level
2028 * driver has completed (successfully or otherwise) a scsi command.
2029 * @SCpnt: mid-level's per command structure.
2031 * Note: potentially run from within an ISR. Must not block.
2033 static int sd_done(struct scsi_cmnd *SCpnt)
2035 int result = SCpnt->result;
2036 unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2037 unsigned int sector_size = SCpnt->device->sector_size;
2039 struct scsi_sense_hdr sshdr;
2040 struct request *req = scsi_cmd_to_rq(SCpnt);
2041 struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2042 int sense_valid = 0;
2043 int sense_deferred = 0;
2045 switch (req_op(req)) {
2046 case REQ_OP_DISCARD:
2047 case REQ_OP_WRITE_ZEROES:
2048 case REQ_OP_ZONE_RESET:
2049 case REQ_OP_ZONE_RESET_ALL:
2050 case REQ_OP_ZONE_OPEN:
2051 case REQ_OP_ZONE_CLOSE:
2052 case REQ_OP_ZONE_FINISH:
2054 good_bytes = blk_rq_bytes(req);
2055 scsi_set_resid(SCpnt, 0);
2058 scsi_set_resid(SCpnt, blk_rq_bytes(req));
2063 * In case of bogus fw or device, we could end up having
2064 * an unaligned partial completion. Check this here and force
2067 resid = scsi_get_resid(SCpnt);
2068 if (resid & (sector_size - 1)) {
2069 sd_printk(KERN_INFO, sdkp,
2070 "Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2071 resid, sector_size);
2072 scsi_print_command(SCpnt);
2073 resid = min(scsi_bufflen(SCpnt),
2074 round_up(resid, sector_size));
2075 scsi_set_resid(SCpnt, resid);
2080 sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2082 sense_deferred = scsi_sense_is_deferred(&sshdr);
2084 sdkp->medium_access_timed_out = 0;
2086 if (!scsi_status_is_check_condition(result) &&
2087 (!sense_valid || sense_deferred))
2090 switch (sshdr.sense_key) {
2091 case HARDWARE_ERROR:
2093 good_bytes = sd_completed_bytes(SCpnt);
2095 case RECOVERED_ERROR:
2096 good_bytes = scsi_bufflen(SCpnt);
2099 /* This indicates a false check condition, so ignore it. An
2100 * unknown amount of data was transferred so treat it as an
2104 memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2106 case ABORTED_COMMAND:
2107 if (sshdr.asc == 0x10) /* DIF: Target detected corruption */
2108 good_bytes = sd_completed_bytes(SCpnt);
2110 case ILLEGAL_REQUEST:
2111 switch (sshdr.asc) {
2112 case 0x10: /* DIX: Host detected corruption */
2113 good_bytes = sd_completed_bytes(SCpnt);
2115 case 0x20: /* INVALID COMMAND OPCODE */
2116 case 0x24: /* INVALID FIELD IN CDB */
2117 switch (SCpnt->cmnd[0]) {
2119 sd_config_discard(sdkp, SD_LBP_DISABLE);
2123 if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2124 sd_config_discard(sdkp, SD_LBP_DISABLE);
2126 sdkp->device->no_write_same = 1;
2127 sd_config_write_same(sdkp);
2128 req->rq_flags |= RQF_QUIET;
2139 if (sd_is_zoned(sdkp))
2140 good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2142 SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2143 "sd_done: completed %d of %d bytes\n",
2144 good_bytes, scsi_bufflen(SCpnt)));
2150 * spinup disk - called only in sd_revalidate_disk()
2153 sd_spinup_disk(struct scsi_disk *sdkp)
2155 unsigned char cmd[10];
2156 unsigned long spintime_expire = 0;
2157 int retries, spintime;
2158 unsigned int the_result;
2159 struct scsi_sense_hdr sshdr;
2160 const struct scsi_exec_args exec_args = {
2163 int sense_valid = 0;
2167 /* Spin up drives, as required. Only do this at boot time */
2168 /* Spinup needs to be done for module loads too. */
2173 bool media_was_present = sdkp->media_present;
2175 cmd[0] = TEST_UNIT_READY;
2176 memset((void *) &cmd[1], 0, 9);
2178 the_result = scsi_execute_cmd(sdkp->device, cmd,
2179 REQ_OP_DRV_IN, NULL, 0,
2184 if (the_result > 0) {
2186 * If the drive has indicated to us that it
2187 * doesn't have any media in it, don't bother
2188 * with any more polling.
2190 if (media_not_present(sdkp, &sshdr)) {
2191 if (media_was_present)
2192 sd_printk(KERN_NOTICE, sdkp,
2193 "Media removed, stopped polling\n");
2197 sense_valid = scsi_sense_valid(&sshdr);
2200 } while (retries < 3 &&
2201 (!scsi_status_is_good(the_result) ||
2202 (scsi_status_is_check_condition(the_result) &&
2203 sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2205 if (!scsi_status_is_check_condition(the_result)) {
2206 /* no sense, TUR either succeeded or failed
2207 * with a status error */
2208 if(!spintime && !scsi_status_is_good(the_result)) {
2209 sd_print_result(sdkp, "Test Unit Ready failed",
2216 * The device does not want the automatic start to be issued.
2218 if (sdkp->device->no_start_on_add)
2221 if (sense_valid && sshdr.sense_key == NOT_READY) {
2222 if (sshdr.asc == 4 && sshdr.ascq == 3)
2223 break; /* manual intervention required */
2224 if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2225 break; /* standby */
2226 if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2227 break; /* unavailable */
2228 if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2229 break; /* sanitize in progress */
2231 * Issue command to spin up drive when not ready
2234 sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2235 cmd[0] = START_STOP;
2236 cmd[1] = 1; /* Return immediately */
2237 memset((void *) &cmd[2], 0, 8);
2238 cmd[4] = 1; /* Start spin cycle */
2239 if (sdkp->device->start_stop_pwr_cond)
2241 scsi_execute_cmd(sdkp->device, cmd,
2242 REQ_OP_DRV_IN, NULL, 0,
2243 SD_TIMEOUT, sdkp->max_retries,
2245 spintime_expire = jiffies + 100 * HZ;
2248 /* Wait 1 second for next try */
2250 printk(KERN_CONT ".");
2253 * Wait for USB flash devices with slow firmware.
2254 * Yes, this sense key/ASC combination shouldn't
2255 * occur here. It's characteristic of these devices.
2257 } else if (sense_valid &&
2258 sshdr.sense_key == UNIT_ATTENTION &&
2259 sshdr.asc == 0x28) {
2261 spintime_expire = jiffies + 5 * HZ;
2264 /* Wait 1 second for next try */
2267 /* we don't understand the sense code, so it's
2268 * probably pointless to loop */
2270 sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2271 sd_print_sense_hdr(sdkp, &sshdr);
2276 } while (spintime && time_before_eq(jiffies, spintime_expire));
2279 if (scsi_status_is_good(the_result))
2280 printk(KERN_CONT "ready\n");
2282 printk(KERN_CONT "not responding...\n");
2287 * Determine whether disk supports Data Integrity Field.
2289 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2291 struct scsi_device *sdp = sdkp->device;
2294 if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2295 sdkp->protection_type = 0;
2299 type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2301 if (type > T10_PI_TYPE3_PROTECTION) {
2302 sd_printk(KERN_ERR, sdkp, "formatted with unsupported" \
2303 " protection type %u. Disabling disk!\n",
2305 sdkp->protection_type = 0;
2309 sdkp->protection_type = type;
2314 static void sd_config_protection(struct scsi_disk *sdkp)
2316 struct scsi_device *sdp = sdkp->device;
2318 sd_dif_config_host(sdkp);
2320 if (!sdkp->protection_type)
2323 if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2324 sd_first_printk(KERN_NOTICE, sdkp,
2325 "Disabling DIF Type %u protection\n",
2326 sdkp->protection_type);
2327 sdkp->protection_type = 0;
2330 sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2331 sdkp->protection_type);
2334 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2335 struct scsi_sense_hdr *sshdr, int sense_valid,
2339 sd_print_sense_hdr(sdkp, sshdr);
2341 sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2344 * Set dirty bit for removable devices if not ready -
2345 * sometimes drives will not report this properly.
2347 if (sdp->removable &&
2348 sense_valid && sshdr->sense_key == NOT_READY)
2349 set_media_not_present(sdkp);
2352 * We used to set media_present to 0 here to indicate no media
2353 * in the drive, but some drives fail read capacity even with
2354 * media present, so we can't do that.
2356 sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2360 #if RC16_LEN > SD_BUF_SIZE
2361 #error RC16_LEN must not be more than SD_BUF_SIZE
2364 #define READ_CAPACITY_RETRIES_ON_RESET 10
2366 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2367 unsigned char *buffer)
2369 unsigned char cmd[16];
2370 struct scsi_sense_hdr sshdr;
2371 const struct scsi_exec_args exec_args = {
2374 int sense_valid = 0;
2376 int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2377 unsigned int alignment;
2378 unsigned long long lba;
2379 unsigned sector_size;
2381 if (sdp->no_read_capacity_16)
2386 cmd[0] = SERVICE_ACTION_IN_16;
2387 cmd[1] = SAI_READ_CAPACITY_16;
2389 memset(buffer, 0, RC16_LEN);
2391 the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2392 buffer, RC16_LEN, SD_TIMEOUT,
2393 sdkp->max_retries, &exec_args);
2394 if (the_result > 0) {
2395 if (media_not_present(sdkp, &sshdr))
2398 sense_valid = scsi_sense_valid(&sshdr);
2400 sshdr.sense_key == ILLEGAL_REQUEST &&
2401 (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2403 /* Invalid Command Operation Code or
2404 * Invalid Field in CDB, just retry
2405 * silently with RC10 */
2408 sshdr.sense_key == UNIT_ATTENTION &&
2409 sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2410 /* Device reset might occur several times,
2411 * give it one more chance */
2412 if (--reset_retries > 0)
2417 } while (the_result && retries);
2420 sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2421 read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2425 sector_size = get_unaligned_be32(&buffer[8]);
2426 lba = get_unaligned_be64(&buffer[0]);
2428 if (sd_read_protection_type(sdkp, buffer) < 0) {
2433 /* Logical blocks per physical block exponent */
2434 sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2437 sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2439 /* Lowest aligned logical block */
2440 alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2441 blk_queue_alignment_offset(sdp->request_queue, alignment);
2442 if (alignment && sdkp->first_scan)
2443 sd_printk(KERN_NOTICE, sdkp,
2444 "physical block alignment offset: %u\n", alignment);
2446 if (buffer[14] & 0x80) { /* LBPME */
2449 if (buffer[14] & 0x40) /* LBPRZ */
2452 sd_config_discard(sdkp, SD_LBP_WS16);
2455 sdkp->capacity = lba + 1;
2459 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2460 unsigned char *buffer)
2462 unsigned char cmd[16];
2463 struct scsi_sense_hdr sshdr;
2464 const struct scsi_exec_args exec_args = {
2467 int sense_valid = 0;
2469 int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2471 unsigned sector_size;
2474 cmd[0] = READ_CAPACITY;
2475 memset(&cmd[1], 0, 9);
2476 memset(buffer, 0, 8);
2478 the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2479 8, SD_TIMEOUT, sdkp->max_retries,
2482 if (media_not_present(sdkp, &sshdr))
2485 if (the_result > 0) {
2486 sense_valid = scsi_sense_valid(&sshdr);
2488 sshdr.sense_key == UNIT_ATTENTION &&
2489 sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2490 /* Device reset might occur several times,
2491 * give it one more chance */
2492 if (--reset_retries > 0)
2497 } while (the_result && retries);
2500 sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2501 read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2505 sector_size = get_unaligned_be32(&buffer[4]);
2506 lba = get_unaligned_be32(&buffer[0]);
2508 if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2509 /* Some buggy (usb cardreader) devices return an lba of
2510 0xffffffff when the want to report a size of 0 (with
2511 which they really mean no media is present) */
2513 sdkp->physical_block_size = sector_size;
2517 sdkp->capacity = lba + 1;
2518 sdkp->physical_block_size = sector_size;
2522 static int sd_try_rc16_first(struct scsi_device *sdp)
2524 if (sdp->host->max_cmd_len < 16)
2526 if (sdp->try_rc_10_first)
2528 if (sdp->scsi_level > SCSI_SPC_2)
2530 if (scsi_device_protection(sdp))
2536 * read disk capacity
2539 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2542 struct scsi_device *sdp = sdkp->device;
2544 if (sd_try_rc16_first(sdp)) {
2545 sector_size = read_capacity_16(sdkp, sdp, buffer);
2546 if (sector_size == -EOVERFLOW)
2548 if (sector_size == -ENODEV)
2550 if (sector_size < 0)
2551 sector_size = read_capacity_10(sdkp, sdp, buffer);
2552 if (sector_size < 0)
2555 sector_size = read_capacity_10(sdkp, sdp, buffer);
2556 if (sector_size == -EOVERFLOW)
2558 if (sector_size < 0)
2560 if ((sizeof(sdkp->capacity) > 4) &&
2561 (sdkp->capacity > 0xffffffffULL)) {
2562 int old_sector_size = sector_size;
2563 sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2564 "Trying to use READ CAPACITY(16).\n");
2565 sector_size = read_capacity_16(sdkp, sdp, buffer);
2566 if (sector_size < 0) {
2567 sd_printk(KERN_NOTICE, sdkp,
2568 "Using 0xffffffff as device size\n");
2569 sdkp->capacity = 1 + (sector_t) 0xffffffff;
2570 sector_size = old_sector_size;
2573 /* Remember that READ CAPACITY(16) succeeded */
2574 sdp->try_rc_10_first = 0;
2578 /* Some devices are known to return the total number of blocks,
2579 * not the highest block number. Some devices have versions
2580 * which do this and others which do not. Some devices we might
2581 * suspect of doing this but we don't know for certain.
2583 * If we know the reported capacity is wrong, decrement it. If
2584 * we can only guess, then assume the number of blocks is even
2585 * (usually true but not always) and err on the side of lowering
2588 if (sdp->fix_capacity ||
2589 (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2590 sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2591 "from its reported value: %llu\n",
2592 (unsigned long long) sdkp->capacity);
2597 if (sector_size == 0) {
2599 sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2603 if (sector_size != 512 &&
2604 sector_size != 1024 &&
2605 sector_size != 2048 &&
2606 sector_size != 4096) {
2607 sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2610 * The user might want to re-format the drive with
2611 * a supported sectorsize. Once this happens, it
2612 * would be relatively trivial to set the thing up.
2613 * For this reason, we leave the thing in the table.
2617 * set a bogus sector size so the normal read/write
2618 * logic in the block layer will eventually refuse any
2619 * request on this device without tripping over power
2620 * of two sector size assumptions
2624 blk_queue_logical_block_size(sdp->request_queue, sector_size);
2625 blk_queue_physical_block_size(sdp->request_queue,
2626 sdkp->physical_block_size);
2627 sdkp->device->sector_size = sector_size;
2629 if (sdkp->capacity > 0xffffffff)
2630 sdp->use_16_for_rw = 1;
2635 * Print disk capacity
2638 sd_print_capacity(struct scsi_disk *sdkp,
2639 sector_t old_capacity)
2641 int sector_size = sdkp->device->sector_size;
2642 char cap_str_2[10], cap_str_10[10];
2644 if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2647 string_get_size(sdkp->capacity, sector_size,
2648 STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2649 string_get_size(sdkp->capacity, sector_size,
2650 STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2652 sd_printk(KERN_NOTICE, sdkp,
2653 "%llu %d-byte logical blocks: (%s/%s)\n",
2654 (unsigned long long)sdkp->capacity,
2655 sector_size, cap_str_10, cap_str_2);
2657 if (sdkp->physical_block_size != sector_size)
2658 sd_printk(KERN_NOTICE, sdkp,
2659 "%u-byte physical blocks\n",
2660 sdkp->physical_block_size);
2663 /* called with buffer of length 512 */
2665 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2666 unsigned char *buffer, int len, struct scsi_mode_data *data,
2667 struct scsi_sense_hdr *sshdr)
2670 * If we must use MODE SENSE(10), make sure that the buffer length
2671 * is at least 8 bytes so that the mode sense header fits.
2673 if (sdkp->device->use_10_for_ms && len < 8)
2676 return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2677 SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2681 * read write protect setting, if possible - called only in sd_revalidate_disk()
2682 * called with buffer of length SD_BUF_SIZE
2685 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2688 struct scsi_device *sdp = sdkp->device;
2689 struct scsi_mode_data data;
2690 int old_wp = sdkp->write_prot;
2692 set_disk_ro(sdkp->disk, 0);
2693 if (sdp->skip_ms_page_3f) {
2694 sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2698 if (sdp->use_192_bytes_for_3f) {
2699 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2702 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2703 * We have to start carefully: some devices hang if we ask
2704 * for more than is available.
2706 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2709 * Second attempt: ask for page 0 When only page 0 is
2710 * implemented, a request for page 3F may return Sense Key
2711 * 5: Illegal Request, Sense Code 24: Invalid field in
2715 res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2718 * Third attempt: ask 255 bytes, as we did earlier.
2721 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2726 sd_first_printk(KERN_WARNING, sdkp,
2727 "Test WP failed, assume Write Enabled\n");
2729 sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2730 set_disk_ro(sdkp->disk, sdkp->write_prot);
2731 if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2732 sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2733 sdkp->write_prot ? "on" : "off");
2734 sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2740 * sd_read_cache_type - called only from sd_revalidate_disk()
2741 * called with buffer of length SD_BUF_SIZE
2744 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2747 struct scsi_device *sdp = sdkp->device;
2752 struct scsi_mode_data data;
2753 struct scsi_sense_hdr sshdr;
2754 int old_wce = sdkp->WCE;
2755 int old_rcd = sdkp->RCD;
2756 int old_dpofua = sdkp->DPOFUA;
2759 if (sdkp->cache_override)
2763 if (sdp->skip_ms_page_8) {
2764 if (sdp->type == TYPE_RBC)
2767 if (sdp->skip_ms_page_3f)
2770 if (sdp->use_192_bytes_for_3f)
2774 } else if (sdp->type == TYPE_RBC) {
2782 /* cautiously ask */
2783 res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2789 if (!data.header_length) {
2792 sd_first_printk(KERN_ERR, sdkp,
2793 "Missing header in MODE_SENSE response\n");
2796 /* that went OK, now ask for the proper length */
2800 * We're only interested in the first three bytes, actually.
2801 * But the data cache page is defined for the first 20.
2805 else if (len > SD_BUF_SIZE) {
2806 sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2807 "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2810 if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2814 if (len > first_len)
2815 res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2819 int offset = data.header_length + data.block_descriptor_length;
2821 while (offset < len) {
2822 u8 page_code = buffer[offset] & 0x3F;
2823 u8 spf = buffer[offset] & 0x40;
2825 if (page_code == 8 || page_code == 6) {
2826 /* We're interested only in the first 3 bytes.
2828 if (len - offset <= 2) {
2829 sd_first_printk(KERN_ERR, sdkp,
2830 "Incomplete mode parameter "
2834 modepage = page_code;
2838 /* Go to the next page */
2839 if (spf && len - offset > 3)
2840 offset += 4 + (buffer[offset+2] << 8) +
2842 else if (!spf && len - offset > 1)
2843 offset += 2 + buffer[offset+1];
2845 sd_first_printk(KERN_ERR, sdkp,
2847 "parameter data\n");
2853 sd_first_printk(KERN_WARNING, sdkp,
2854 "No Caching mode page found\n");
2858 if (modepage == 8) {
2859 sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2860 sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2862 sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2866 sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2867 if (sdp->broken_fua) {
2868 sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2870 } else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2871 !sdkp->device->use_16_for_rw) {
2872 sd_first_printk(KERN_NOTICE, sdkp,
2873 "Uses READ/WRITE(6), disabling FUA\n");
2877 /* No cache flush allowed for write protected devices */
2878 if (sdkp->WCE && sdkp->write_prot)
2881 if (sdkp->first_scan || old_wce != sdkp->WCE ||
2882 old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2883 sd_printk(KERN_NOTICE, sdkp,
2884 "Write cache: %s, read cache: %s, %s\n",
2885 sdkp->WCE ? "enabled" : "disabled",
2886 sdkp->RCD ? "disabled" : "enabled",
2887 sdkp->DPOFUA ? "supports DPO and FUA"
2888 : "doesn't support DPO or FUA");
2894 if (res == -EIO && scsi_sense_valid(&sshdr) &&
2895 sshdr.sense_key == ILLEGAL_REQUEST &&
2896 sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2897 /* Invalid field in CDB */
2898 sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2900 sd_first_printk(KERN_ERR, sdkp,
2901 "Asking for cache data failed\n");
2904 if (sdp->wce_default_on) {
2905 sd_first_printk(KERN_NOTICE, sdkp,
2906 "Assuming drive cache: write back\n");
2909 sd_first_printk(KERN_WARNING, sdkp,
2910 "Assuming drive cache: write through\n");
2918 * The ATO bit indicates whether the DIF application tag is available
2919 * for use by the operating system.
2921 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2924 struct scsi_device *sdp = sdkp->device;
2925 struct scsi_mode_data data;
2926 struct scsi_sense_hdr sshdr;
2928 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2931 if (sdkp->protection_type == 0)
2934 res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
2935 sdkp->max_retries, &data, &sshdr);
2937 if (res < 0 || !data.header_length ||
2939 sd_first_printk(KERN_WARNING, sdkp,
2940 "getting Control mode page failed, assume no ATO\n");
2942 if (res == -EIO && scsi_sense_valid(&sshdr))
2943 sd_print_sense_hdr(sdkp, &sshdr);
2948 offset = data.header_length + data.block_descriptor_length;
2950 if ((buffer[offset] & 0x3f) != 0x0a) {
2951 sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2955 if ((buffer[offset + 5] & 0x80) == 0)
2964 * sd_read_block_limits - Query disk device for preferred I/O sizes.
2965 * @sdkp: disk to query
2967 static void sd_read_block_limits(struct scsi_disk *sdkp)
2969 struct scsi_vpd *vpd;
2973 vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2974 if (!vpd || vpd->len < 16)
2977 sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2978 sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2979 sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2981 if (vpd->len >= 64) {
2982 unsigned int lba_count, desc_count;
2984 sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2989 lba_count = get_unaligned_be32(&vpd->data[20]);
2990 desc_count = get_unaligned_be32(&vpd->data[24]);
2992 if (lba_count && desc_count)
2993 sdkp->max_unmap_blocks = lba_count;
2995 sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2997 if (vpd->data[32] & 0x80)
2998 sdkp->unmap_alignment =
2999 get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3001 if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
3003 if (sdkp->max_unmap_blocks)
3004 sd_config_discard(sdkp, SD_LBP_UNMAP);
3006 sd_config_discard(sdkp, SD_LBP_WS16);
3008 } else { /* LBP VPD page tells us what to use */
3009 if (sdkp->lbpu && sdkp->max_unmap_blocks)
3010 sd_config_discard(sdkp, SD_LBP_UNMAP);
3011 else if (sdkp->lbpws)
3012 sd_config_discard(sdkp, SD_LBP_WS16);
3013 else if (sdkp->lbpws10)
3014 sd_config_discard(sdkp, SD_LBP_WS10);
3016 sd_config_discard(sdkp, SD_LBP_DISABLE);
3025 * sd_read_block_characteristics - Query block dev. characteristics
3026 * @sdkp: disk to query
3028 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
3030 struct request_queue *q = sdkp->disk->queue;
3031 struct scsi_vpd *vpd;
3036 vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3038 if (!vpd || vpd->len < 8) {
3043 rot = get_unaligned_be16(&vpd->data[4]);
3044 zoned = (vpd->data[8] >> 4) & 3;
3048 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3049 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3052 if (sdkp->device->type == TYPE_ZBC) {
3054 * Host-managed: Per ZBC and ZAC specifications, writes in
3055 * sequential write required zones of host-managed devices must
3056 * be aligned to the device physical block size.
3058 disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
3059 blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3061 sdkp->zoned = zoned;
3062 if (sdkp->zoned == 1) {
3064 disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
3066 /* Regular disk or drive managed disk */
3067 disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3071 if (!sdkp->first_scan)
3074 if (blk_queue_is_zoned(q)) {
3075 sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3076 q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3078 if (sdkp->zoned == 1)
3079 sd_printk(KERN_NOTICE, sdkp,
3080 "Host-aware SMR disk used as regular disk\n");
3081 else if (sdkp->zoned == 2)
3082 sd_printk(KERN_NOTICE, sdkp,
3083 "Drive-managed SMR disk\n");
3088 * sd_read_block_provisioning - Query provisioning VPD page
3089 * @sdkp: disk to query
3091 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3093 struct scsi_vpd *vpd;
3095 if (sdkp->lbpme == 0)
3099 vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3101 if (!vpd || vpd->len < 8) {
3107 sdkp->lbpu = (vpd->data[5] >> 7) & 1; /* UNMAP */
3108 sdkp->lbpws = (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3109 sdkp->lbpws10 = (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3113 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3115 struct scsi_device *sdev = sdkp->device;
3117 if (sdev->host->no_write_same) {
3118 sdev->no_write_same = 1;
3123 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3124 struct scsi_vpd *vpd;
3126 sdev->no_report_opcodes = 1;
3128 /* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3129 * CODES is unsupported and the device has an ATA
3130 * Information VPD page (SAT).
3133 vpd = rcu_dereference(sdev->vpd_pg89);
3135 sdev->no_write_same = 1;
3139 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3142 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3146 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3148 struct scsi_device *sdev = sdkp->device;
3150 if (!sdev->security_supported)
3153 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3154 SECURITY_PROTOCOL_IN, 0) == 1 &&
3155 scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3156 SECURITY_PROTOCOL_OUT, 0) == 1)
3160 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3162 return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3166 * sd_read_cpr - Query concurrent positioning ranges
3167 * @sdkp: disk to query
3169 static void sd_read_cpr(struct scsi_disk *sdkp)
3171 struct blk_independent_access_ranges *iars = NULL;
3172 unsigned char *buffer = NULL;
3173 unsigned int nr_cpr = 0;
3174 int i, vpd_len, buf_len = SD_BUF_SIZE;
3178 * We need to have the capacity set first for the block layer to be
3179 * able to check the ranges.
3181 if (sdkp->first_scan)
3184 if (!sdkp->capacity)
3188 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3189 * leading to a maximum page size of 64 + 256*32 bytes.
3191 buf_len = 64 + 256*32;
3192 buffer = kmalloc(buf_len, GFP_KERNEL);
3193 if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3196 /* We must have at least a 64B header and one 32B range descriptor */
3197 vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3198 if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3199 sd_printk(KERN_ERR, sdkp,
3200 "Invalid Concurrent Positioning Ranges VPD page\n");
3204 nr_cpr = (vpd_len - 64) / 32;
3210 iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3217 for (i = 0; i < nr_cpr; i++, desc += 32) {
3219 sd_printk(KERN_ERR, sdkp,
3220 "Invalid Concurrent Positioning Range number\n");
3225 iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3226 iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3230 disk_set_independent_access_ranges(sdkp->disk, iars);
3231 if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3232 sd_printk(KERN_NOTICE, sdkp,
3233 "%u concurrent positioning ranges\n", nr_cpr);
3234 sdkp->nr_actuators = nr_cpr;
3240 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3242 struct scsi_device *sdp = sdkp->device;
3243 unsigned int min_xfer_bytes =
3244 logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3246 if (sdkp->min_xfer_blocks == 0)
3249 if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3250 sd_first_printk(KERN_WARNING, sdkp,
3251 "Preferred minimum I/O size %u bytes not a " \
3252 "multiple of physical block size (%u bytes)\n",
3253 min_xfer_bytes, sdkp->physical_block_size);
3254 sdkp->min_xfer_blocks = 0;
3258 sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3264 * Determine the device's preferred I/O size for reads and writes
3265 * unless the reported value is unreasonably small, large, not a
3266 * multiple of the physical block size, or simply garbage.
3268 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3269 unsigned int dev_max)
3271 struct scsi_device *sdp = sdkp->device;
3272 unsigned int opt_xfer_bytes =
3273 logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3274 unsigned int min_xfer_bytes =
3275 logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3277 if (sdkp->opt_xfer_blocks == 0)
3280 if (sdkp->opt_xfer_blocks > dev_max) {
3281 sd_first_printk(KERN_WARNING, sdkp,
3282 "Optimal transfer size %u logical blocks " \
3283 "> dev_max (%u logical blocks)\n",
3284 sdkp->opt_xfer_blocks, dev_max);
3288 if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3289 sd_first_printk(KERN_WARNING, sdkp,
3290 "Optimal transfer size %u logical blocks " \
3291 "> sd driver limit (%u logical blocks)\n",
3292 sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3296 if (opt_xfer_bytes < PAGE_SIZE) {
3297 sd_first_printk(KERN_WARNING, sdkp,
3298 "Optimal transfer size %u bytes < " \
3299 "PAGE_SIZE (%u bytes)\n",
3300 opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3304 if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3305 sd_first_printk(KERN_WARNING, sdkp,
3306 "Optimal transfer size %u bytes not a " \
3307 "multiple of preferred minimum block " \
3308 "size (%u bytes)\n",
3309 opt_xfer_bytes, min_xfer_bytes);
3313 if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3314 sd_first_printk(KERN_WARNING, sdkp,
3315 "Optimal transfer size %u bytes not a " \
3316 "multiple of physical block size (%u bytes)\n",
3317 opt_xfer_bytes, sdkp->physical_block_size);
3321 sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3327 * sd_revalidate_disk - called the first time a new disk is seen,
3328 * performs disk spin up, read_capacity, etc.
3329 * @disk: struct gendisk we care about
3331 static int sd_revalidate_disk(struct gendisk *disk)
3333 struct scsi_disk *sdkp = scsi_disk(disk);
3334 struct scsi_device *sdp = sdkp->device;
3335 struct request_queue *q = sdkp->disk->queue;
3336 sector_t old_capacity = sdkp->capacity;
3337 unsigned char *buffer;
3338 unsigned int dev_max, rw_max;
3340 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3341 "sd_revalidate_disk\n"));
3344 * If the device is offline, don't try and read capacity or any
3345 * of the other niceties.
3347 if (!scsi_device_online(sdp))
3350 buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3352 sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3353 "allocation failure.\n");
3357 sd_spinup_disk(sdkp);
3360 * Without media there is no reason to ask; moreover, some devices
3361 * react badly if we do.
3363 if (sdkp->media_present) {
3364 sd_read_capacity(sdkp, buffer);
3367 * set the default to rotational. All non-rotational devices
3368 * support the block characteristics VPD page, which will
3369 * cause this to be updated correctly and any device which
3370 * doesn't support it should be treated as rotational.
3372 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3373 blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3375 if (scsi_device_supports_vpd(sdp)) {
3376 sd_read_block_provisioning(sdkp);
3377 sd_read_block_limits(sdkp);
3378 sd_read_block_characteristics(sdkp);
3379 sd_zbc_read_zones(sdkp, buffer);
3383 sd_print_capacity(sdkp, old_capacity);
3385 sd_read_write_protect_flag(sdkp, buffer);
3386 sd_read_cache_type(sdkp, buffer);
3387 sd_read_app_tag_own(sdkp, buffer);
3388 sd_read_write_same(sdkp, buffer);
3389 sd_read_security(sdkp, buffer);
3390 sd_config_protection(sdkp);
3394 * We now have all cache related info, determine how we deal
3395 * with flush requests.
3397 sd_set_flush_flag(sdkp);
3399 /* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3400 dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3402 /* Some devices report a maximum block count for READ/WRITE requests. */
3403 dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3404 q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3406 if (sd_validate_min_xfer_size(sdkp))
3407 blk_queue_io_min(sdkp->disk->queue,
3408 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3410 blk_queue_io_min(sdkp->disk->queue, 0);
3412 if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3413 q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3414 rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3416 q->limits.io_opt = 0;
3417 rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3418 (sector_t)BLK_DEF_MAX_SECTORS);
3422 * Limit default to SCSI host optimal sector limit if set. There may be
3423 * an impact on performance for when the size of a request exceeds this
3426 rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3428 /* Do not exceed controller limit */
3429 rw_max = min(rw_max, queue_max_hw_sectors(q));
3432 * Only update max_sectors if previously unset or if the current value
3433 * exceeds the capabilities of the hardware.
3435 if (sdkp->first_scan ||
3436 q->limits.max_sectors > q->limits.max_dev_sectors ||
3437 q->limits.max_sectors > q->limits.max_hw_sectors)
3438 q->limits.max_sectors = rw_max;
3440 sdkp->first_scan = 0;
3442 set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3443 sd_config_write_same(sdkp);
3447 * For a zoned drive, revalidating the zones can be done only once
3448 * the gendisk capacity is set. So if this fails, set back the gendisk
3451 if (sd_zbc_revalidate_zones(sdkp))
3452 set_capacity_and_notify(disk, 0);
3459 * sd_unlock_native_capacity - unlock native capacity
3460 * @disk: struct gendisk to set capacity for
3462 * Block layer calls this function if it detects that partitions
3463 * on @disk reach beyond the end of the device. If the SCSI host
3464 * implements ->unlock_native_capacity() method, it's invoked to
3465 * give it a chance to adjust the device capacity.
3468 * Defined by block layer. Might sleep.
3470 static void sd_unlock_native_capacity(struct gendisk *disk)
3472 struct scsi_device *sdev = scsi_disk(disk)->device;
3474 if (sdev->host->hostt->unlock_native_capacity)
3475 sdev->host->hostt->unlock_native_capacity(sdev);
3479 * sd_format_disk_name - format disk name
3480 * @prefix: name prefix - ie. "sd" for SCSI disks
3481 * @index: index of the disk to format name for
3482 * @buf: output buffer
3483 * @buflen: length of the output buffer
3485 * SCSI disk names starts at sda. The 26th device is sdz and the
3486 * 27th is sdaa. The last one for two lettered suffix is sdzz
3487 * which is followed by sdaaa.
3489 * This is basically 26 base counting with one extra 'nil' entry
3490 * at the beginning from the second digit on and can be
3491 * determined using similar method as 26 base conversion with the
3492 * index shifted -1 after each digit is computed.
3498 * 0 on success, -errno on failure.
3500 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3502 const int base = 'z' - 'a' + 1;
3503 char *begin = buf + strlen(prefix);
3504 char *end = buf + buflen;
3514 *--p = 'a' + (index % unit);
3515 index = (index / unit) - 1;
3516 } while (index >= 0);
3518 memmove(begin, p, end - p);
3519 memcpy(buf, prefix, strlen(prefix));
3525 * sd_probe - called during driver initialization and whenever a
3526 * new scsi device is attached to the system. It is called once
3527 * for each scsi device (not just disks) present.
3528 * @dev: pointer to device object
3530 * Returns 0 if successful (or not interested in this scsi device
3531 * (e.g. scanner)); 1 when there is an error.
3533 * Note: this function is invoked from the scsi mid-level.
3534 * This function sets up the mapping between a given
3535 * <host,channel,id,lun> (found in sdp) and new device name
3536 * (e.g. /dev/sda). More precisely it is the block device major
3537 * and minor number that is chosen here.
3539 * Assume sd_probe is not re-entrant (for time being)
3540 * Also think about sd_probe() and sd_remove() running coincidentally.
3542 static int sd_probe(struct device *dev)
3544 struct scsi_device *sdp = to_scsi_device(dev);
3545 struct scsi_disk *sdkp;
3550 scsi_autopm_get_device(sdp);
3552 if (sdp->type != TYPE_DISK &&
3553 sdp->type != TYPE_ZBC &&
3554 sdp->type != TYPE_MOD &&
3555 sdp->type != TYPE_RBC)
3558 if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3559 sdev_printk(KERN_WARNING, sdp,
3560 "Unsupported ZBC host-managed device.\n");
3564 SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3568 sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3572 gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3573 &sd_bio_compl_lkclass);
3577 index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3579 sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3583 error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3585 sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3586 goto out_free_index;
3591 sdkp->index = index;
3592 sdkp->max_retries = SD_MAX_RETRIES;
3593 atomic_set(&sdkp->openers, 0);
3594 atomic_set(&sdkp->device->ioerr_cnt, 0);
3596 if (!sdp->request_queue->rq_timeout) {
3597 if (sdp->type != TYPE_MOD)
3598 blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3600 blk_queue_rq_timeout(sdp->request_queue,
3604 device_initialize(&sdkp->disk_dev);
3605 sdkp->disk_dev.parent = get_device(dev);
3606 sdkp->disk_dev.class = &sd_disk_class;
3607 dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3609 error = device_add(&sdkp->disk_dev);
3611 put_device(&sdkp->disk_dev);
3615 dev_set_drvdata(dev, sdkp);
3617 gd->major = sd_major((index & 0xf0) >> 4);
3618 gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3619 gd->minors = SD_MINORS;
3621 gd->fops = &sd_fops;
3622 gd->private_data = sdkp;
3624 /* defaults, until the device tells us otherwise */
3625 sdp->sector_size = 512;
3627 sdkp->media_present = 1;
3628 sdkp->write_prot = 0;
3629 sdkp->cache_override = 0;
3633 sdkp->first_scan = 1;
3634 sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3636 sd_revalidate_disk(gd);
3638 if (sdp->removable) {
3639 gd->flags |= GENHD_FL_REMOVABLE;
3640 gd->events |= DISK_EVENT_MEDIA_CHANGE;
3641 gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3644 blk_pm_runtime_init(sdp->request_queue, dev);
3645 if (sdp->rpm_autosuspend) {
3646 pm_runtime_set_autosuspend_delay(dev,
3647 sdp->host->hostt->rpm_autosuspend_delay);
3650 error = device_add_disk(dev, gd, NULL);
3652 put_device(&sdkp->disk_dev);
3657 if (sdkp->security) {
3658 sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3660 sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3663 sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3664 sdp->removable ? "removable " : "");
3665 scsi_autopm_put_device(sdp);
3670 ida_free(&sd_index_ida, index);
3676 scsi_autopm_put_device(sdp);
3681 * sd_remove - called whenever a scsi disk (previously recognized by
3682 * sd_probe) is detached from the system. It is called (potentially
3683 * multiple times) during sd module unload.
3684 * @dev: pointer to device object
3686 * Note: this function is invoked from the scsi mid-level.
3687 * This function potentially frees up a device name (e.g. /dev/sdc)
3688 * that could be re-used by a subsequent sd_probe().
3689 * This function is not called when the built-in sd driver is "exit-ed".
3691 static int sd_remove(struct device *dev)
3693 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3695 scsi_autopm_get_device(sdkp->device);
3697 device_del(&sdkp->disk_dev);
3698 del_gendisk(sdkp->disk);
3701 put_disk(sdkp->disk);
3705 static void scsi_disk_release(struct device *dev)
3707 struct scsi_disk *sdkp = to_scsi_disk(dev);
3709 ida_free(&sd_index_ida, sdkp->index);
3710 sd_zbc_free_zone_info(sdkp);
3711 put_device(&sdkp->device->sdev_gendev);
3712 free_opal_dev(sdkp->opal_dev);
3717 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3719 unsigned char cmd[6] = { START_STOP }; /* START_VALID */
3720 struct scsi_sense_hdr sshdr;
3721 const struct scsi_exec_args exec_args = {
3723 .req_flags = BLK_MQ_REQ_PM,
3725 struct scsi_device *sdp = sdkp->device;
3729 cmd[4] |= 1; /* START */
3731 if (sdp->start_stop_pwr_cond)
3732 cmd[4] |= start ? 1 << 4 : 3 << 4; /* Active or Standby */
3734 if (!scsi_device_online(sdp))
3737 res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3738 sdkp->max_retries, &exec_args);
3740 sd_print_result(sdkp, "Start/Stop Unit failed", res);
3741 if (res > 0 && scsi_sense_valid(&sshdr)) {
3742 sd_print_sense_hdr(sdkp, &sshdr);
3743 /* 0x3a is medium not present */
3744 if (sshdr.asc == 0x3a)
3749 /* SCSI error codes must not go to the generic layer */
3757 * Send a SYNCHRONIZE CACHE instruction down to the device through
3758 * the normal SCSI command structure. Wait for the command to
3761 static void sd_shutdown(struct device *dev)
3763 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3766 return; /* this can happen */
3768 if (pm_runtime_suspended(dev))
3771 if (sdkp->WCE && sdkp->media_present) {
3772 sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3773 sd_sync_cache(sdkp, NULL);
3776 if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3777 sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3778 sd_start_stop_device(sdkp, 0);
3782 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3784 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3785 struct scsi_sense_hdr sshdr;
3788 if (!sdkp) /* E.g.: runtime suspend following sd_remove() */
3791 if (sdkp->WCE && sdkp->media_present) {
3792 if (!sdkp->device->silence_suspend)
3793 sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3794 ret = sd_sync_cache(sdkp, &sshdr);
3797 /* ignore OFFLINE device */
3801 if (!scsi_sense_valid(&sshdr) ||
3802 sshdr.sense_key != ILLEGAL_REQUEST)
3806 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3807 * doesn't support sync. There's not much to do and
3808 * suspend shouldn't fail.
3814 if (sdkp->device->manage_start_stop) {
3815 if (!sdkp->device->silence_suspend)
3816 sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3817 /* an error is not worth aborting a system sleep */
3818 ret = sd_start_stop_device(sdkp, 0);
3819 if (ignore_stop_errors)
3826 static int sd_suspend_system(struct device *dev)
3828 if (pm_runtime_suspended(dev))
3831 return sd_suspend_common(dev, true);
3834 static int sd_suspend_runtime(struct device *dev)
3836 return sd_suspend_common(dev, false);
3839 static int sd_resume(struct device *dev)
3841 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3844 if (!sdkp) /* E.g.: runtime resume at the start of sd_probe() */
3847 if (!sdkp->device->manage_start_stop)
3850 if (!sdkp->device->no_start_on_resume) {
3851 sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3852 ret = sd_start_stop_device(sdkp, 1);
3856 opal_unlock_from_suspend(sdkp->opal_dev);
3860 static int sd_resume_system(struct device *dev)
3862 if (pm_runtime_suspended(dev))
3865 return sd_resume(dev);
3868 static int sd_resume_runtime(struct device *dev)
3870 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3871 struct scsi_device *sdp;
3873 if (!sdkp) /* E.g.: runtime resume at the start of sd_probe() */
3878 if (sdp->ignore_media_change) {
3879 /* clear the device's sense data */
3880 static const u8 cmd[10] = { REQUEST_SENSE };
3881 const struct scsi_exec_args exec_args = {
3882 .req_flags = BLK_MQ_REQ_PM,
3885 if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
3886 sdp->request_queue->rq_timeout, 1,
3888 sd_printk(KERN_NOTICE, sdkp,
3889 "Failed to clear sense data\n");
3892 return sd_resume(dev);
3895 static const struct dev_pm_ops sd_pm_ops = {
3896 .suspend = sd_suspend_system,
3897 .resume = sd_resume_system,
3898 .poweroff = sd_suspend_system,
3899 .restore = sd_resume_system,
3900 .runtime_suspend = sd_suspend_runtime,
3901 .runtime_resume = sd_resume_runtime,
3904 static struct scsi_driver sd_template = {
3907 .owner = THIS_MODULE,
3909 .probe_type = PROBE_PREFER_ASYNCHRONOUS,
3910 .remove = sd_remove,
3911 .shutdown = sd_shutdown,
3914 .rescan = sd_rescan,
3915 .init_command = sd_init_command,
3916 .uninit_command = sd_uninit_command,
3918 .eh_action = sd_eh_action,
3919 .eh_reset = sd_eh_reset,
3923 * init_sd - entry point for this driver (both when built in or when
3926 * Note: this function registers this driver with the scsi mid-level.
3928 static int __init init_sd(void)
3930 int majors = 0, i, err;
3932 SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3934 for (i = 0; i < SD_MAJORS; i++) {
3935 if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3943 err = class_register(&sd_disk_class);
3947 sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3948 if (!sd_page_pool) {
3949 printk(KERN_ERR "sd: can't init discard page pool\n");
3954 err = scsi_register_driver(&sd_template.gendrv);
3956 goto err_out_driver;
3961 mempool_destroy(sd_page_pool);
3963 class_unregister(&sd_disk_class);
3965 for (i = 0; i < SD_MAJORS; i++)
3966 unregister_blkdev(sd_major(i), "sd");
3971 * exit_sd - exit point for this driver (when it is a module).
3973 * Note: this function unregisters this driver from the scsi mid-level.
3975 static void __exit exit_sd(void)
3979 SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3981 scsi_unregister_driver(&sd_template.gendrv);
3982 mempool_destroy(sd_page_pool);
3984 class_unregister(&sd_disk_class);
3986 for (i = 0; i < SD_MAJORS; i++)
3987 unregister_blkdev(sd_major(i), "sd");
3990 module_init(init_sd);
3991 module_exit(exit_sd);
3993 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3995 scsi_print_sense_hdr(sdkp->device,
3996 sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3999 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4001 const char *hb_string = scsi_hostbyte_string(result);
4004 sd_printk(KERN_INFO, sdkp,
4005 "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4006 hb_string ? hb_string : "invalid",
4009 sd_printk(KERN_INFO, sdkp,
4010 "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4011 msg, host_byte(result), "DRIVER_OK");