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:
12 * outstanding request, and other enhancements.
13 * Support loadable low-level scsi drivers.
15 * eight major numbers.
18 * sd_init and cleanups.
20 * not being read in sd_open. Fix problem where removable media
21 * could be ejected after sd_open.
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/rw_hint.h>
51 #include <linux/major.h>
52 #include <linux/mutex.h>
53 #include <linux/string_helpers.h>
54 #include <linux/slab.h>
55 #include <linux/sed-opal.h>
56 #include <linux/pm_runtime.h>
58 #include <linux/t10-pi.h>
59 #include <linux/uaccess.h>
60 #include <asm/unaligned.h>
62 #include <scsi/scsi.h>
63 #include <scsi/scsi_cmnd.h>
64 #include <scsi/scsi_dbg.h>
65 #include <scsi/scsi_device.h>
66 #include <scsi/scsi_devinfo.h>
67 #include <scsi/scsi_driver.h>
68 #include <scsi/scsi_eh.h>
69 #include <scsi/scsi_host.h>
70 #include <scsi/scsi_ioctl.h>
71 #include <scsi/scsicam.h>
72 #include <scsi/scsi_common.h>
75 #include "scsi_priv.h"
76 #include "scsi_logging.h"
78 MODULE_AUTHOR("Eric Youngdale");
79 MODULE_DESCRIPTION("SCSI disk (sd) driver");
80 MODULE_LICENSE("GPL");
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
96 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
97 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
100 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
101 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
105 static void sd_config_discard(struct scsi_disk *, unsigned int);
106 static void sd_config_write_same(struct scsi_disk *);
107 static int sd_revalidate_disk(struct gendisk *);
108 static void sd_unlock_native_capacity(struct gendisk *disk);
109 static void sd_shutdown(struct device *);
110 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
111 static void scsi_disk_release(struct device *cdev);
113 static DEFINE_IDA(sd_index_ida);
115 static mempool_t *sd_page_pool;
116 static struct lock_class_key sd_bio_compl_lkclass;
118 static const char *sd_cache_types[] = {
119 "write through", "none", "write back",
120 "write back, no read (daft)"
123 static void sd_set_flush_flag(struct scsi_disk *sdkp)
125 bool wc = false, fua = false;
133 blk_queue_write_cache(sdkp->disk->queue, wc, fua);
137 cache_type_store(struct device *dev, struct device_attribute *attr,
138 const char *buf, size_t count)
140 int ct, rcd, wce, sp;
141 struct scsi_disk *sdkp = to_scsi_disk(dev);
142 struct scsi_device *sdp = sdkp->device;
145 struct scsi_mode_data data;
146 struct scsi_sense_hdr sshdr;
147 static const char temp[] = "temporary ";
150 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
151 /* no cache control on RBC devices; theoretically they
152 * can do it, but there's probably so many exceptions
153 * it's not worth the risk */
156 if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
157 buf += sizeof(temp) - 1;
158 sdkp->cache_override = 1;
160 sdkp->cache_override = 0;
163 ct = sysfs_match_string(sd_cache_types, buf);
167 rcd = ct & 0x01 ? 1 : 0;
168 wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
170 if (sdkp->cache_override) {
173 sd_set_flush_flag(sdkp);
177 if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
178 sdkp->max_retries, &data, NULL))
180 len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
181 data.block_descriptor_length);
182 buffer_data = buffer + data.header_length +
183 data.block_descriptor_length;
184 buffer_data[2] &= ~0x05;
185 buffer_data[2] |= wce << 2 | rcd;
186 sp = buffer_data[0] & 0x80 ? 1 : 0;
187 buffer_data[0] &= ~0x80;
190 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
191 * received mode parameter buffer before doing MODE SELECT.
193 data.device_specific = 0;
195 ret = scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
196 sdkp->max_retries, &data, &sshdr);
198 if (ret > 0 && scsi_sense_valid(&sshdr))
199 sd_print_sense_hdr(sdkp, &sshdr);
202 sd_revalidate_disk(sdkp->disk);
207 manage_start_stop_show(struct device *dev,
208 struct device_attribute *attr, char *buf)
210 struct scsi_disk *sdkp = to_scsi_disk(dev);
211 struct scsi_device *sdp = sdkp->device;
213 return sysfs_emit(buf, "%u\n",
214 sdp->manage_system_start_stop &&
215 sdp->manage_runtime_start_stop &&
216 sdp->manage_shutdown);
218 static DEVICE_ATTR_RO(manage_start_stop);
221 manage_system_start_stop_show(struct device *dev,
222 struct device_attribute *attr, char *buf)
224 struct scsi_disk *sdkp = to_scsi_disk(dev);
225 struct scsi_device *sdp = sdkp->device;
227 return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
231 manage_system_start_stop_store(struct device *dev,
232 struct device_attribute *attr,
233 const char *buf, size_t count)
235 struct scsi_disk *sdkp = to_scsi_disk(dev);
236 struct scsi_device *sdp = sdkp->device;
239 if (!capable(CAP_SYS_ADMIN))
242 if (kstrtobool(buf, &v))
245 sdp->manage_system_start_stop = v;
249 static DEVICE_ATTR_RW(manage_system_start_stop);
252 manage_runtime_start_stop_show(struct device *dev,
253 struct device_attribute *attr, char *buf)
255 struct scsi_disk *sdkp = to_scsi_disk(dev);
256 struct scsi_device *sdp = sdkp->device;
258 return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
262 manage_runtime_start_stop_store(struct device *dev,
263 struct device_attribute *attr,
264 const char *buf, size_t count)
266 struct scsi_disk *sdkp = to_scsi_disk(dev);
267 struct scsi_device *sdp = sdkp->device;
270 if (!capable(CAP_SYS_ADMIN))
273 if (kstrtobool(buf, &v))
276 sdp->manage_runtime_start_stop = v;
280 static DEVICE_ATTR_RW(manage_runtime_start_stop);
282 static ssize_t manage_shutdown_show(struct device *dev,
283 struct device_attribute *attr, char *buf)
285 struct scsi_disk *sdkp = to_scsi_disk(dev);
286 struct scsi_device *sdp = sdkp->device;
288 return sysfs_emit(buf, "%u\n", sdp->manage_shutdown);
291 static ssize_t manage_shutdown_store(struct device *dev,
292 struct device_attribute *attr,
293 const char *buf, size_t count)
295 struct scsi_disk *sdkp = to_scsi_disk(dev);
296 struct scsi_device *sdp = sdkp->device;
299 if (!capable(CAP_SYS_ADMIN))
302 if (kstrtobool(buf, &v))
305 sdp->manage_shutdown = v;
309 static DEVICE_ATTR_RW(manage_shutdown);
312 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
314 struct scsi_disk *sdkp = to_scsi_disk(dev);
316 return sprintf(buf, "%u\n", sdkp->device->allow_restart);
320 allow_restart_store(struct device *dev, struct device_attribute *attr,
321 const char *buf, size_t count)
324 struct scsi_disk *sdkp = to_scsi_disk(dev);
325 struct scsi_device *sdp = sdkp->device;
327 if (!capable(CAP_SYS_ADMIN))
330 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
333 if (kstrtobool(buf, &v))
336 sdp->allow_restart = v;
340 static DEVICE_ATTR_RW(allow_restart);
343 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
345 struct scsi_disk *sdkp = to_scsi_disk(dev);
346 int ct = sdkp->RCD + 2*sdkp->WCE;
348 return sprintf(buf, "%s\n", sd_cache_types[ct]);
350 static DEVICE_ATTR_RW(cache_type);
353 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
355 struct scsi_disk *sdkp = to_scsi_disk(dev);
357 return sprintf(buf, "%u\n", sdkp->DPOFUA);
359 static DEVICE_ATTR_RO(FUA);
362 protection_type_show(struct device *dev, struct device_attribute *attr,
365 struct scsi_disk *sdkp = to_scsi_disk(dev);
367 return sprintf(buf, "%u\n", sdkp->protection_type);
371 protection_type_store(struct device *dev, struct device_attribute *attr,
372 const char *buf, size_t count)
374 struct scsi_disk *sdkp = to_scsi_disk(dev);
378 if (!capable(CAP_SYS_ADMIN))
381 err = kstrtouint(buf, 10, &val);
386 if (val <= T10_PI_TYPE3_PROTECTION)
387 sdkp->protection_type = val;
391 static DEVICE_ATTR_RW(protection_type);
394 protection_mode_show(struct device *dev, struct device_attribute *attr,
397 struct scsi_disk *sdkp = to_scsi_disk(dev);
398 struct scsi_device *sdp = sdkp->device;
399 unsigned int dif, dix;
401 dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
402 dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
404 if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
410 return sprintf(buf, "none\n");
412 return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
414 static DEVICE_ATTR_RO(protection_mode);
417 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
419 struct scsi_disk *sdkp = to_scsi_disk(dev);
421 return sprintf(buf, "%u\n", sdkp->ATO);
423 static DEVICE_ATTR_RO(app_tag_own);
426 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
429 struct scsi_disk *sdkp = to_scsi_disk(dev);
431 return sprintf(buf, "%u\n", sdkp->lbpme);
433 static DEVICE_ATTR_RO(thin_provisioning);
435 /* sysfs_match_string() requires dense arrays */
436 static const char *lbp_mode[] = {
437 [SD_LBP_FULL] = "full",
438 [SD_LBP_UNMAP] = "unmap",
439 [SD_LBP_WS16] = "writesame_16",
440 [SD_LBP_WS10] = "writesame_10",
441 [SD_LBP_ZERO] = "writesame_zero",
442 [SD_LBP_DISABLE] = "disabled",
446 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
449 struct scsi_disk *sdkp = to_scsi_disk(dev);
451 return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
455 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
456 const char *buf, size_t count)
458 struct scsi_disk *sdkp = to_scsi_disk(dev);
459 struct scsi_device *sdp = sdkp->device;
462 if (!capable(CAP_SYS_ADMIN))
465 if (sd_is_zoned(sdkp)) {
466 sd_config_discard(sdkp, SD_LBP_DISABLE);
470 if (sdp->type != TYPE_DISK)
473 mode = sysfs_match_string(lbp_mode, buf);
477 sd_config_discard(sdkp, mode);
481 static DEVICE_ATTR_RW(provisioning_mode);
483 /* sysfs_match_string() requires dense arrays */
484 static const char *zeroing_mode[] = {
485 [SD_ZERO_WRITE] = "write",
486 [SD_ZERO_WS] = "writesame",
487 [SD_ZERO_WS16_UNMAP] = "writesame_16_unmap",
488 [SD_ZERO_WS10_UNMAP] = "writesame_10_unmap",
492 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
495 struct scsi_disk *sdkp = to_scsi_disk(dev);
497 return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
501 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
502 const char *buf, size_t count)
504 struct scsi_disk *sdkp = to_scsi_disk(dev);
507 if (!capable(CAP_SYS_ADMIN))
510 mode = sysfs_match_string(zeroing_mode, buf);
514 sdkp->zeroing_mode = mode;
518 static DEVICE_ATTR_RW(zeroing_mode);
521 max_medium_access_timeouts_show(struct device *dev,
522 struct device_attribute *attr, char *buf)
524 struct scsi_disk *sdkp = to_scsi_disk(dev);
526 return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
530 max_medium_access_timeouts_store(struct device *dev,
531 struct device_attribute *attr, const char *buf,
534 struct scsi_disk *sdkp = to_scsi_disk(dev);
537 if (!capable(CAP_SYS_ADMIN))
540 err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
542 return err ? err : count;
544 static DEVICE_ATTR_RW(max_medium_access_timeouts);
547 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
550 struct scsi_disk *sdkp = to_scsi_disk(dev);
552 return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
556 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
557 const char *buf, size_t count)
559 struct scsi_disk *sdkp = to_scsi_disk(dev);
560 struct scsi_device *sdp = sdkp->device;
564 if (!capable(CAP_SYS_ADMIN))
567 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
570 err = kstrtoul(buf, 10, &max);
576 sdp->no_write_same = 1;
577 else if (max <= SD_MAX_WS16_BLOCKS) {
578 sdp->no_write_same = 0;
579 sdkp->max_ws_blocks = max;
582 sd_config_write_same(sdkp);
586 static DEVICE_ATTR_RW(max_write_same_blocks);
589 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
591 struct scsi_disk *sdkp = to_scsi_disk(dev);
593 if (sdkp->device->type == TYPE_ZBC)
594 return sprintf(buf, "host-managed\n");
595 if (sdkp->zoned == 1)
596 return sprintf(buf, "host-aware\n");
597 if (sdkp->zoned == 2)
598 return sprintf(buf, "drive-managed\n");
599 return sprintf(buf, "none\n");
601 static DEVICE_ATTR_RO(zoned_cap);
604 max_retries_store(struct device *dev, struct device_attribute *attr,
605 const char *buf, size_t count)
607 struct scsi_disk *sdkp = to_scsi_disk(dev);
608 struct scsi_device *sdev = sdkp->device;
611 err = kstrtoint(buf, 10, &retries);
615 if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
616 sdkp->max_retries = retries;
620 sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
626 max_retries_show(struct device *dev, struct device_attribute *attr,
629 struct scsi_disk *sdkp = to_scsi_disk(dev);
631 return sprintf(buf, "%d\n", sdkp->max_retries);
634 static DEVICE_ATTR_RW(max_retries);
636 static struct attribute *sd_disk_attrs[] = {
637 &dev_attr_cache_type.attr,
639 &dev_attr_allow_restart.attr,
640 &dev_attr_manage_start_stop.attr,
641 &dev_attr_manage_system_start_stop.attr,
642 &dev_attr_manage_runtime_start_stop.attr,
643 &dev_attr_manage_shutdown.attr,
644 &dev_attr_protection_type.attr,
645 &dev_attr_protection_mode.attr,
646 &dev_attr_app_tag_own.attr,
647 &dev_attr_thin_provisioning.attr,
648 &dev_attr_provisioning_mode.attr,
649 &dev_attr_zeroing_mode.attr,
650 &dev_attr_max_write_same_blocks.attr,
651 &dev_attr_max_medium_access_timeouts.attr,
652 &dev_attr_zoned_cap.attr,
653 &dev_attr_max_retries.attr,
656 ATTRIBUTE_GROUPS(sd_disk);
658 static struct class sd_disk_class = {
660 .dev_release = scsi_disk_release,
661 .dev_groups = sd_disk_groups,
665 * Don't request a new module, as that could deadlock in multipath
668 static void sd_default_probe(dev_t devt)
673 * Device no to disk mapping:
675 * major disc2 disc p1
676 * |............|.............|....|....| <- dev_t
679 * Inside a major, we have 16k disks, however mapped non-
680 * contiguously. The first 16 disks are for major0, the next
681 * ones with major1, ... Disk 256 is for major0 again, disk 272
683 * As we stay compatible with our numbering scheme, we can reuse
684 * the well-know SCSI majors 8, 65--71, 136--143.
686 static int sd_major(int major_idx)
690 return SCSI_DISK0_MAJOR;
692 return SCSI_DISK1_MAJOR + major_idx - 1;
694 return SCSI_DISK8_MAJOR + major_idx - 8;
697 return 0; /* shut up gcc */
701 #ifdef CONFIG_BLK_SED_OPAL
702 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
703 size_t len, bool send)
705 struct scsi_disk *sdkp = data;
706 struct scsi_device *sdev = sdkp->device;
708 const struct scsi_exec_args exec_args = {
709 .req_flags = BLK_MQ_REQ_PM,
713 cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
715 put_unaligned_be16(spsp, &cdb[2]);
716 put_unaligned_be32(len, &cdb[6]);
718 ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
719 buffer, len, SD_TIMEOUT, sdkp->max_retries,
721 return ret <= 0 ? ret : -EIO;
723 #endif /* CONFIG_BLK_SED_OPAL */
726 * Look up the DIX operation based on whether the command is read or
727 * write and whether dix and dif are enabled.
729 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
731 /* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
732 static const unsigned int ops[] = { /* wrt dix dif */
733 SCSI_PROT_NORMAL, /* 0 0 0 */
734 SCSI_PROT_READ_STRIP, /* 0 0 1 */
735 SCSI_PROT_READ_INSERT, /* 0 1 0 */
736 SCSI_PROT_READ_PASS, /* 0 1 1 */
737 SCSI_PROT_NORMAL, /* 1 0 0 */
738 SCSI_PROT_WRITE_INSERT, /* 1 0 1 */
739 SCSI_PROT_WRITE_STRIP, /* 1 1 0 */
740 SCSI_PROT_WRITE_PASS, /* 1 1 1 */
743 return ops[write << 2 | dix << 1 | dif];
747 * Returns a mask of the protection flags that are valid for a given DIX
750 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
752 static const unsigned int flag_mask[] = {
753 [SCSI_PROT_NORMAL] = 0,
755 [SCSI_PROT_READ_STRIP] = SCSI_PROT_TRANSFER_PI |
756 SCSI_PROT_GUARD_CHECK |
757 SCSI_PROT_REF_CHECK |
758 SCSI_PROT_REF_INCREMENT,
760 [SCSI_PROT_READ_INSERT] = SCSI_PROT_REF_INCREMENT |
761 SCSI_PROT_IP_CHECKSUM,
763 [SCSI_PROT_READ_PASS] = SCSI_PROT_TRANSFER_PI |
764 SCSI_PROT_GUARD_CHECK |
765 SCSI_PROT_REF_CHECK |
766 SCSI_PROT_REF_INCREMENT |
767 SCSI_PROT_IP_CHECKSUM,
769 [SCSI_PROT_WRITE_INSERT] = SCSI_PROT_TRANSFER_PI |
770 SCSI_PROT_REF_INCREMENT,
772 [SCSI_PROT_WRITE_STRIP] = SCSI_PROT_GUARD_CHECK |
773 SCSI_PROT_REF_CHECK |
774 SCSI_PROT_REF_INCREMENT |
775 SCSI_PROT_IP_CHECKSUM,
777 [SCSI_PROT_WRITE_PASS] = SCSI_PROT_TRANSFER_PI |
778 SCSI_PROT_GUARD_CHECK |
779 SCSI_PROT_REF_CHECK |
780 SCSI_PROT_REF_INCREMENT |
781 SCSI_PROT_IP_CHECKSUM,
784 return flag_mask[prot_op];
787 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
788 unsigned int dix, unsigned int dif)
790 struct request *rq = scsi_cmd_to_rq(scmd);
791 struct bio *bio = rq->bio;
792 unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
793 unsigned int protect = 0;
795 if (dix) { /* DIX Type 0, 1, 2, 3 */
796 if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
797 scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
799 if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
800 scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
803 if (dif != T10_PI_TYPE3_PROTECTION) { /* DIX/DIF Type 0, 1, 2 */
804 scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
806 if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
807 scmd->prot_flags |= SCSI_PROT_REF_CHECK;
810 if (dif) { /* DIX/DIF Type 1, 2, 3 */
811 scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
813 if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
814 protect = 3 << 5; /* Disable target PI checking */
816 protect = 1 << 5; /* Enable target PI checking */
819 scsi_set_prot_op(scmd, prot_op);
820 scsi_set_prot_type(scmd, dif);
821 scmd->prot_flags &= sd_prot_flag_mask(prot_op);
826 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
828 struct request_queue *q = sdkp->disk->queue;
829 unsigned int logical_block_size = sdkp->device->sector_size;
830 unsigned int max_blocks = 0;
832 q->limits.discard_alignment =
833 sdkp->unmap_alignment * logical_block_size;
834 q->limits.discard_granularity =
835 max(sdkp->physical_block_size,
836 sdkp->unmap_granularity * logical_block_size);
837 sdkp->provisioning_mode = mode;
843 blk_queue_max_discard_sectors(q, 0);
847 max_blocks = min_not_zero(sdkp->max_unmap_blocks,
848 (u32)SD_MAX_WS16_BLOCKS);
852 if (sdkp->device->unmap_limit_for_ws)
853 max_blocks = sdkp->max_unmap_blocks;
855 max_blocks = sdkp->max_ws_blocks;
857 max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
861 if (sdkp->device->unmap_limit_for_ws)
862 max_blocks = sdkp->max_unmap_blocks;
864 max_blocks = sdkp->max_ws_blocks;
866 max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
870 max_blocks = min_not_zero(sdkp->max_ws_blocks,
871 (u32)SD_MAX_WS10_BLOCKS);
875 blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
878 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
882 page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
885 clear_highpage(page);
886 bvec_set_page(&rq->special_vec, page, data_len, 0);
887 rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
888 return bvec_virt(&rq->special_vec);
891 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
893 struct scsi_device *sdp = cmd->device;
894 struct request *rq = scsi_cmd_to_rq(cmd);
895 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
896 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
897 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
898 unsigned int data_len = 24;
901 buf = sd_set_special_bvec(rq, data_len);
903 return BLK_STS_RESOURCE;
906 cmd->cmnd[0] = UNMAP;
909 put_unaligned_be16(6 + 16, &buf[0]);
910 put_unaligned_be16(16, &buf[2]);
911 put_unaligned_be64(lba, &buf[8]);
912 put_unaligned_be32(nr_blocks, &buf[16]);
914 cmd->allowed = sdkp->max_retries;
915 cmd->transfersize = data_len;
916 rq->timeout = SD_TIMEOUT;
918 return scsi_alloc_sgtables(cmd);
921 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
924 struct scsi_device *sdp = cmd->device;
925 struct request *rq = scsi_cmd_to_rq(cmd);
926 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
927 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
928 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
929 u32 data_len = sdp->sector_size;
931 if (!sd_set_special_bvec(rq, data_len))
932 return BLK_STS_RESOURCE;
935 cmd->cmnd[0] = WRITE_SAME_16;
937 cmd->cmnd[1] = 0x8; /* UNMAP */
938 put_unaligned_be64(lba, &cmd->cmnd[2]);
939 put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
941 cmd->allowed = sdkp->max_retries;
942 cmd->transfersize = data_len;
943 rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
945 return scsi_alloc_sgtables(cmd);
948 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
951 struct scsi_device *sdp = cmd->device;
952 struct request *rq = scsi_cmd_to_rq(cmd);
953 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
954 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
955 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
956 u32 data_len = sdp->sector_size;
958 if (!sd_set_special_bvec(rq, data_len))
959 return BLK_STS_RESOURCE;
962 cmd->cmnd[0] = WRITE_SAME;
964 cmd->cmnd[1] = 0x8; /* UNMAP */
965 put_unaligned_be32(lba, &cmd->cmnd[2]);
966 put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
968 cmd->allowed = sdkp->max_retries;
969 cmd->transfersize = data_len;
970 rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
972 return scsi_alloc_sgtables(cmd);
975 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
977 struct request *rq = scsi_cmd_to_rq(cmd);
978 struct scsi_device *sdp = cmd->device;
979 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
980 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
981 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
983 if (!(rq->cmd_flags & REQ_NOUNMAP)) {
984 switch (sdkp->zeroing_mode) {
985 case SD_ZERO_WS16_UNMAP:
986 return sd_setup_write_same16_cmnd(cmd, true);
987 case SD_ZERO_WS10_UNMAP:
988 return sd_setup_write_same10_cmnd(cmd, true);
992 if (sdp->no_write_same) {
993 rq->rq_flags |= RQF_QUIET;
994 return BLK_STS_TARGET;
997 if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
998 return sd_setup_write_same16_cmnd(cmd, false);
1000 return sd_setup_write_same10_cmnd(cmd, false);
1003 static void sd_config_write_same(struct scsi_disk *sdkp)
1005 struct request_queue *q = sdkp->disk->queue;
1006 unsigned int logical_block_size = sdkp->device->sector_size;
1008 if (sdkp->device->no_write_same) {
1009 sdkp->max_ws_blocks = 0;
1013 /* Some devices can not handle block counts above 0xffff despite
1014 * supporting WRITE SAME(16). Consequently we default to 64k
1015 * blocks per I/O unless the device explicitly advertises a
1018 if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
1019 sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1020 (u32)SD_MAX_WS16_BLOCKS);
1021 else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
1022 sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1023 (u32)SD_MAX_WS10_BLOCKS);
1025 sdkp->device->no_write_same = 1;
1026 sdkp->max_ws_blocks = 0;
1029 if (sdkp->lbprz && sdkp->lbpws)
1030 sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
1031 else if (sdkp->lbprz && sdkp->lbpws10)
1032 sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
1033 else if (sdkp->max_ws_blocks)
1034 sdkp->zeroing_mode = SD_ZERO_WS;
1036 sdkp->zeroing_mode = SD_ZERO_WRITE;
1038 if (sdkp->max_ws_blocks &&
1039 sdkp->physical_block_size > logical_block_size) {
1041 * Reporting a maximum number of blocks that is not aligned
1042 * on the device physical size would cause a large write same
1043 * request to be split into physically unaligned chunks by
1044 * __blkdev_issue_write_zeroes() even if the caller of this
1045 * functions took care to align the large request. So make sure
1046 * the maximum reported is aligned to the device physical block
1047 * size. This is only an optional optimization for regular
1048 * disks, but this is mandatory to avoid failure of large write
1049 * same requests directed at sequential write required zones of
1050 * host-managed ZBC disks.
1052 sdkp->max_ws_blocks =
1053 round_down(sdkp->max_ws_blocks,
1054 bytes_to_logical(sdkp->device,
1055 sdkp->physical_block_size));
1059 blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1060 (logical_block_size >> 9));
1063 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1065 struct request *rq = scsi_cmd_to_rq(cmd);
1066 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1068 /* flush requests don't perform I/O, zero the S/G table */
1069 memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1071 if (cmd->device->use_16_for_sync) {
1072 cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1075 cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1078 cmd->transfersize = 0;
1079 cmd->allowed = sdkp->max_retries;
1081 rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1086 * sd_group_number() - Compute the GROUP NUMBER field
1087 * @cmd: SCSI command for which to compute the value of the six-bit GROUP NUMBER
1090 * From SBC-5 r05 (https://www.t10.org/cgi-bin/ac.pl?t=f&f=sbc5r05.pdf):
1091 * 0: no relative lifetime.
1092 * 1: shortest relative lifetime.
1093 * 2: second shortest relative lifetime.
1094 * 3 - 0x3d: intermediate relative lifetimes.
1095 * 0x3e: second longest relative lifetime.
1096 * 0x3f: longest relative lifetime.
1098 static u8 sd_group_number(struct scsi_cmnd *cmd)
1100 const struct request *rq = scsi_cmd_to_rq(cmd);
1101 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1106 return min3((u32)rq->write_hint, (u32)sdkp->permanent_stream_count,
1110 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1111 sector_t lba, unsigned int nr_blocks,
1112 unsigned char flags, unsigned int dld)
1114 cmd->cmd_len = SD_EXT_CDB_SIZE;
1115 cmd->cmnd[0] = VARIABLE_LENGTH_CMD;
1116 cmd->cmnd[6] = sd_group_number(cmd);
1117 cmd->cmnd[7] = 0x18; /* Additional CDB len */
1118 cmd->cmnd[9] = write ? WRITE_32 : READ_32;
1119 cmd->cmnd[10] = flags;
1120 cmd->cmnd[11] = dld & 0x07;
1121 put_unaligned_be64(lba, &cmd->cmnd[12]);
1122 put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1123 put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1128 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1129 sector_t lba, unsigned int nr_blocks,
1130 unsigned char flags, unsigned int dld)
1133 cmd->cmnd[0] = write ? WRITE_16 : READ_16;
1134 cmd->cmnd[1] = flags | ((dld >> 2) & 0x01);
1135 cmd->cmnd[14] = ((dld & 0x03) << 6) | sd_group_number(cmd);
1137 put_unaligned_be64(lba, &cmd->cmnd[2]);
1138 put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1143 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1144 sector_t lba, unsigned int nr_blocks,
1145 unsigned char flags)
1148 cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1149 cmd->cmnd[1] = flags;
1150 cmd->cmnd[6] = sd_group_number(cmd);
1152 put_unaligned_be32(lba, &cmd->cmnd[2]);
1153 put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1158 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1159 sector_t lba, unsigned int nr_blocks,
1160 unsigned char flags)
1162 /* Avoid that 0 blocks gets translated into 256 blocks. */
1163 if (WARN_ON_ONCE(nr_blocks == 0))
1164 return BLK_STS_IOERR;
1166 if (unlikely(flags & 0x8)) {
1168 * This happens only if this drive failed 10byte rw
1169 * command with ILLEGAL_REQUEST during operation and
1170 * thus turned off use_10_for_rw.
1172 scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1173 return BLK_STS_IOERR;
1177 cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1178 cmd->cmnd[1] = (lba >> 16) & 0x1f;
1179 cmd->cmnd[2] = (lba >> 8) & 0xff;
1180 cmd->cmnd[3] = lba & 0xff;
1181 cmd->cmnd[4] = nr_blocks;
1188 * Check if a command has a duration limit set. If it does, and the target
1189 * device supports CDL and the feature is enabled, return the limit
1190 * descriptor index to use. Return 0 (no limit) otherwise.
1192 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1194 struct scsi_device *sdp = sdkp->device;
1197 if (!sdp->cdl_supported || !sdp->cdl_enable)
1201 * Use "no limit" if the request ioprio does not specify a duration
1204 hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1205 if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1206 hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1209 return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1212 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1214 struct request *rq = scsi_cmd_to_rq(cmd);
1215 struct scsi_device *sdp = cmd->device;
1216 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1217 sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1219 unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1220 unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1221 bool write = rq_data_dir(rq) == WRITE;
1222 unsigned char protect, fua;
1228 ret = scsi_alloc_sgtables(cmd);
1229 if (ret != BLK_STS_OK)
1232 ret = BLK_STS_IOERR;
1233 if (!scsi_device_online(sdp) || sdp->changed) {
1234 scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1238 if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1239 scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1243 if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1244 scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1249 * Some SD card readers can't handle accesses which touch the
1250 * last one or two logical blocks. Split accesses as needed.
1252 threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1254 if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1255 if (lba < threshold) {
1256 /* Access up to the threshold but not beyond */
1257 nr_blocks = threshold - lba;
1259 /* Access only a single logical block */
1264 fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1265 dix = scsi_prot_sg_count(cmd);
1266 dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1267 dld = sd_cdl_dld(sdkp, cmd);
1270 protect = sd_setup_protect_cmnd(cmd, dix, dif);
1274 if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1275 ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1276 protect | fua, dld);
1277 } else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1278 ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1279 protect | fua, dld);
1280 } else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1281 sdp->use_10_for_rw || protect || rq->write_hint) {
1282 ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1285 ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1289 if (unlikely(ret != BLK_STS_OK))
1293 * We shouldn't disconnect in the middle of a sector, so with a dumb
1294 * host adapter, it's safe to assume that we can at least transfer
1295 * this many bytes between each connect / disconnect.
1297 cmd->transfersize = sdp->sector_size;
1298 cmd->underflow = nr_blocks << 9;
1299 cmd->allowed = sdkp->max_retries;
1300 cmd->sdb.length = nr_blocks * sdp->sector_size;
1303 scmd_printk(KERN_INFO, cmd,
1304 "%s: block=%llu, count=%d\n", __func__,
1305 (unsigned long long)blk_rq_pos(rq),
1306 blk_rq_sectors(rq)));
1308 scmd_printk(KERN_INFO, cmd,
1309 "%s %d/%u 512 byte blocks.\n",
1310 write ? "writing" : "reading", nr_blocks,
1311 blk_rq_sectors(rq)));
1314 * This indicates that the command is ready from our end to be queued.
1318 scsi_free_sgtables(cmd);
1322 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1324 struct request *rq = scsi_cmd_to_rq(cmd);
1326 switch (req_op(rq)) {
1327 case REQ_OP_DISCARD:
1328 switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1330 return sd_setup_unmap_cmnd(cmd);
1332 return sd_setup_write_same16_cmnd(cmd, true);
1334 return sd_setup_write_same10_cmnd(cmd, true);
1336 return sd_setup_write_same10_cmnd(cmd, false);
1338 return BLK_STS_TARGET;
1340 case REQ_OP_WRITE_ZEROES:
1341 return sd_setup_write_zeroes_cmnd(cmd);
1343 return sd_setup_flush_cmnd(cmd);
1346 return sd_setup_read_write_cmnd(cmd);
1347 case REQ_OP_ZONE_RESET:
1348 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1350 case REQ_OP_ZONE_RESET_ALL:
1351 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1353 case REQ_OP_ZONE_OPEN:
1354 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1355 case REQ_OP_ZONE_CLOSE:
1356 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1357 case REQ_OP_ZONE_FINISH:
1358 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1361 return BLK_STS_NOTSUPP;
1365 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1367 struct request *rq = scsi_cmd_to_rq(SCpnt);
1369 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1370 mempool_free(rq->special_vec.bv_page, sd_page_pool);
1373 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1375 if (sdkp->device->removable || sdkp->write_prot) {
1376 if (disk_check_media_change(disk))
1381 * Force a full rescan after ioctl(BLKRRPART). While the disk state has
1382 * nothing to do with partitions, BLKRRPART is used to force a full
1383 * revalidate after things like a format for historical reasons.
1385 return test_bit(GD_NEED_PART_SCAN, &disk->state);
1389 * sd_open - open a scsi disk device
1390 * @disk: disk to open
1393 * Returns 0 if successful. Returns a negated errno value in case
1396 * Note: This can be called from a user context (e.g. fsck(1) )
1397 * or from within the kernel (e.g. as a result of a mount(1) ).
1398 * In the latter case @inode and @filp carry an abridged amount
1399 * of information as noted above.
1401 * Locking: called with disk->open_mutex held.
1403 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1405 struct scsi_disk *sdkp = scsi_disk(disk);
1406 struct scsi_device *sdev = sdkp->device;
1409 if (scsi_device_get(sdev))
1412 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1415 * If the device is in error recovery, wait until it is done.
1416 * If the device is offline, then disallow any access to it.
1419 if (!scsi_block_when_processing_errors(sdev))
1422 if (sd_need_revalidate(disk, sdkp))
1423 sd_revalidate_disk(disk);
1426 * If the drive is empty, just let the open fail.
1428 retval = -ENOMEDIUM;
1429 if (sdev->removable && !sdkp->media_present &&
1430 !(mode & BLK_OPEN_NDELAY))
1434 * If the device has the write protect tab set, have the open fail
1435 * if the user expects to be able to write to the thing.
1438 if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1442 * It is possible that the disk changing stuff resulted in
1443 * the device being taken offline. If this is the case,
1444 * report this to the user, and don't pretend that the
1445 * open actually succeeded.
1448 if (!scsi_device_online(sdev))
1451 if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1452 if (scsi_block_when_processing_errors(sdev))
1453 scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1459 scsi_device_put(sdev);
1464 * sd_release - invoked when the (last) close(2) is called on this
1466 * @disk: disk to release
1470 * Note: may block (uninterruptible) if error recovery is underway
1473 * Locking: called with disk->open_mutex held.
1475 static void sd_release(struct gendisk *disk)
1477 struct scsi_disk *sdkp = scsi_disk(disk);
1478 struct scsi_device *sdev = sdkp->device;
1480 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1482 if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1483 if (scsi_block_when_processing_errors(sdev))
1484 scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1487 scsi_device_put(sdev);
1490 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1492 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1493 struct scsi_device *sdp = sdkp->device;
1494 struct Scsi_Host *host = sdp->host;
1495 sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1498 /* default to most commonly used values */
1499 diskinfo[0] = 0x40; /* 1 << 6 */
1500 diskinfo[1] = 0x20; /* 1 << 5 */
1501 diskinfo[2] = capacity >> 11;
1503 /* override with calculated, extended default, or driver values */
1504 if (host->hostt->bios_param)
1505 host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1507 scsicam_bios_param(bdev, capacity, diskinfo);
1509 geo->heads = diskinfo[0];
1510 geo->sectors = diskinfo[1];
1511 geo->cylinders = diskinfo[2];
1516 * sd_ioctl - process an ioctl
1517 * @bdev: target block device
1519 * @cmd: ioctl command number
1520 * @arg: this is third argument given to ioctl(2) system call.
1521 * Often contains a pointer.
1523 * Returns 0 if successful (some ioctls return positive numbers on
1524 * success as well). Returns a negated errno value in case of error.
1526 * Note: most ioctls are forward onto the block subsystem or further
1527 * down in the scsi subsystem.
1529 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1530 unsigned int cmd, unsigned long arg)
1532 struct gendisk *disk = bdev->bd_disk;
1533 struct scsi_disk *sdkp = scsi_disk(disk);
1534 struct scsi_device *sdp = sdkp->device;
1535 void __user *p = (void __user *)arg;
1538 SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1539 "cmd=0x%x\n", disk->disk_name, cmd));
1541 if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1542 return -ENOIOCTLCMD;
1545 * If we are in the middle of error recovery, don't let anyone
1546 * else try and use this device. Also, if error recovery fails, it
1547 * may try and take the device offline, in which case all further
1548 * access to the device is prohibited.
1550 error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1551 (mode & BLK_OPEN_NDELAY));
1555 if (is_sed_ioctl(cmd))
1556 return sed_ioctl(sdkp->opal_dev, cmd, p);
1557 return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1560 static void set_media_not_present(struct scsi_disk *sdkp)
1562 if (sdkp->media_present)
1563 sdkp->device->changed = 1;
1565 if (sdkp->device->removable) {
1566 sdkp->media_present = 0;
1571 static int media_not_present(struct scsi_disk *sdkp,
1572 struct scsi_sense_hdr *sshdr)
1574 if (!scsi_sense_valid(sshdr))
1577 /* not invoked for commands that could return deferred errors */
1578 switch (sshdr->sense_key) {
1579 case UNIT_ATTENTION:
1581 /* medium not present */
1582 if (sshdr->asc == 0x3A) {
1583 set_media_not_present(sdkp);
1591 * sd_check_events - check media events
1592 * @disk: kernel device descriptor
1593 * @clearing: disk events currently being cleared
1595 * Returns mask of DISK_EVENT_*.
1597 * Note: this function is invoked from the block subsystem.
1599 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1601 struct scsi_disk *sdkp = disk->private_data;
1602 struct scsi_device *sdp;
1610 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1613 * If the device is offline, don't send any commands - just pretend as
1614 * if the command failed. If the device ever comes back online, we
1615 * can deal with it then. It is only because of unrecoverable errors
1616 * that we would ever take a device offline in the first place.
1618 if (!scsi_device_online(sdp)) {
1619 set_media_not_present(sdkp);
1624 * Using TEST_UNIT_READY enables differentiation between drive with
1625 * no cartridge loaded - NOT READY, drive with changed cartridge -
1626 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1628 * Drives that auto spin down. eg iomega jaz 1G, will be started
1629 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1630 * sd_revalidate() is called.
1632 if (scsi_block_when_processing_errors(sdp)) {
1633 struct scsi_sense_hdr sshdr = { 0, };
1635 retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1638 /* failed to execute TUR, assume media not present */
1639 if (retval < 0 || host_byte(retval)) {
1640 set_media_not_present(sdkp);
1644 if (media_not_present(sdkp, &sshdr))
1649 * For removable scsi disk we have to recognise the presence
1650 * of a disk in the drive.
1652 if (!sdkp->media_present)
1654 sdkp->media_present = 1;
1657 * sdp->changed is set under the following conditions:
1659 * Medium present state has changed in either direction.
1660 * Device has indicated UNIT_ATTENTION.
1662 disk_changed = sdp->changed;
1664 return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1667 static int sd_sync_cache(struct scsi_disk *sdkp)
1670 struct scsi_device *sdp = sdkp->device;
1671 const int timeout = sdp->request_queue->rq_timeout
1672 * SD_FLUSH_TIMEOUT_MULTIPLIER;
1673 /* Leave the rest of the command zero to indicate flush everything. */
1674 const unsigned char cmd[16] = { sdp->use_16_for_sync ?
1675 SYNCHRONIZE_CACHE_16 : SYNCHRONIZE_CACHE };
1676 struct scsi_sense_hdr sshdr;
1677 struct scsi_failure failure_defs[] = {
1680 .result = SCMD_FAILURE_RESULT_ANY,
1684 struct scsi_failures failures = {
1685 .failure_definitions = failure_defs,
1687 const struct scsi_exec_args exec_args = {
1688 .req_flags = BLK_MQ_REQ_PM,
1690 .failures = &failures,
1693 if (!scsi_device_online(sdp))
1696 res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, timeout,
1697 sdkp->max_retries, &exec_args);
1699 sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1704 if (scsi_status_is_check_condition(res) &&
1705 scsi_sense_valid(&sshdr)) {
1706 sd_print_sense_hdr(sdkp, &sshdr);
1708 /* we need to evaluate the error return */
1709 if (sshdr.asc == 0x3a || /* medium not present */
1710 sshdr.asc == 0x20 || /* invalid command */
1711 (sshdr.asc == 0x74 && sshdr.ascq == 0x71)) /* drive is password locked */
1712 /* this is no error here */
1715 * This drive doesn't support sync and there's not much
1716 * we can do because this is called during shutdown
1717 * or suspend so just return success so those operations
1720 if (sshdr.sense_key == ILLEGAL_REQUEST)
1724 switch (host_byte(res)) {
1725 /* ignore errors due to racing a disconnection */
1726 case DID_BAD_TARGET:
1727 case DID_NO_CONNECT:
1729 /* signal the upper layer it might try again */
1733 case DID_SOFT_ERROR:
1742 static void sd_rescan(struct device *dev)
1744 struct scsi_disk *sdkp = dev_get_drvdata(dev);
1746 sd_revalidate_disk(sdkp->disk);
1749 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1750 enum blk_unique_id type)
1752 struct scsi_device *sdev = scsi_disk(disk)->device;
1753 const struct scsi_vpd *vpd;
1754 const unsigned char *d;
1755 int ret = -ENXIO, len;
1758 vpd = rcu_dereference(sdev->vpd_pg83);
1763 for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1764 /* we only care about designators with LU association */
1765 if (((d[1] >> 4) & 0x3) != 0x00)
1767 if ((d[1] & 0xf) != type)
1771 * Only exit early if a 16-byte descriptor was found. Otherwise
1772 * keep looking as one with more entropy might still show up.
1775 if (len != 8 && len != 12 && len != 16)
1778 memcpy(id, d + 4, len);
1787 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1789 switch (host_byte(result)) {
1790 case DID_TRANSPORT_MARGINAL:
1791 case DID_TRANSPORT_DISRUPTED:
1793 return PR_STS_RETRY_PATH_FAILURE;
1794 case DID_NO_CONNECT:
1795 return PR_STS_PATH_FAILED;
1796 case DID_TRANSPORT_FAILFAST:
1797 return PR_STS_PATH_FAST_FAILED;
1800 switch (status_byte(result)) {
1801 case SAM_STAT_RESERVATION_CONFLICT:
1802 return PR_STS_RESERVATION_CONFLICT;
1803 case SAM_STAT_CHECK_CONDITION:
1804 if (!scsi_sense_valid(sshdr))
1805 return PR_STS_IOERR;
1807 if (sshdr->sense_key == ILLEGAL_REQUEST &&
1808 (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1813 return PR_STS_IOERR;
1817 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1818 unsigned char *data, int data_len)
1820 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1821 struct scsi_device *sdev = sdkp->device;
1822 struct scsi_sense_hdr sshdr;
1823 u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1824 struct scsi_failure failure_defs[] = {
1826 .sense = UNIT_ATTENTION,
1827 .asc = SCMD_FAILURE_ASC_ANY,
1828 .ascq = SCMD_FAILURE_ASCQ_ANY,
1830 .result = SAM_STAT_CHECK_CONDITION,
1834 struct scsi_failures failures = {
1835 .failure_definitions = failure_defs,
1837 const struct scsi_exec_args exec_args = {
1839 .failures = &failures,
1843 put_unaligned_be16(data_len, &cmd[7]);
1845 result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1846 SD_TIMEOUT, sdkp->max_retries, &exec_args);
1847 if (scsi_status_is_check_condition(result) &&
1848 scsi_sense_valid(&sshdr)) {
1849 sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1850 scsi_print_sense_hdr(sdev, NULL, &sshdr);
1856 return sd_scsi_to_pr_err(&sshdr, result);
1859 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1861 int result, i, data_offset, num_copy_keys;
1862 u32 num_keys = keys_info->num_keys;
1863 int data_len = num_keys * 8 + 8;
1866 data = kzalloc(data_len, GFP_KERNEL);
1870 result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1874 keys_info->generation = get_unaligned_be32(&data[0]);
1875 keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1878 num_copy_keys = min(num_keys, keys_info->num_keys);
1880 for (i = 0; i < num_copy_keys; i++) {
1881 keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1890 static int sd_pr_read_reservation(struct block_device *bdev,
1891 struct pr_held_reservation *rsv)
1893 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1894 struct scsi_device *sdev = sdkp->device;
1898 result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1902 len = get_unaligned_be32(&data[4]);
1906 /* Make sure we have at least the key and type */
1908 sdev_printk(KERN_INFO, sdev,
1909 "READ RESERVATION failed due to short return buffer of %d bytes\n",
1914 rsv->generation = get_unaligned_be32(&data[0]);
1915 rsv->key = get_unaligned_be64(&data[8]);
1916 rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1920 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1921 u64 sa_key, enum scsi_pr_type type, u8 flags)
1923 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1924 struct scsi_device *sdev = sdkp->device;
1925 struct scsi_sense_hdr sshdr;
1926 struct scsi_failure failure_defs[] = {
1928 .sense = UNIT_ATTENTION,
1929 .asc = SCMD_FAILURE_ASC_ANY,
1930 .ascq = SCMD_FAILURE_ASCQ_ANY,
1932 .result = SAM_STAT_CHECK_CONDITION,
1936 struct scsi_failures failures = {
1937 .failure_definitions = failure_defs,
1939 const struct scsi_exec_args exec_args = {
1941 .failures = &failures,
1944 u8 cmd[16] = { 0, };
1945 u8 data[24] = { 0, };
1947 cmd[0] = PERSISTENT_RESERVE_OUT;
1950 put_unaligned_be32(sizeof(data), &cmd[5]);
1952 put_unaligned_be64(key, &data[0]);
1953 put_unaligned_be64(sa_key, &data[8]);
1956 result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1957 sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1960 if (scsi_status_is_check_condition(result) &&
1961 scsi_sense_valid(&sshdr)) {
1962 sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1963 scsi_print_sense_hdr(sdev, NULL, &sshdr);
1969 return sd_scsi_to_pr_err(&sshdr, result);
1972 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1975 if (flags & ~PR_FL_IGNORE_KEY)
1977 return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1978 old_key, new_key, 0,
1979 (1 << 0) /* APTPL */);
1982 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1987 return sd_pr_out_command(bdev, 0x01, key, 0,
1988 block_pr_type_to_scsi(type), 0);
1991 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1993 return sd_pr_out_command(bdev, 0x02, key, 0,
1994 block_pr_type_to_scsi(type), 0);
1997 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1998 enum pr_type type, bool abort)
2000 return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
2001 block_pr_type_to_scsi(type), 0);
2004 static int sd_pr_clear(struct block_device *bdev, u64 key)
2006 return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
2009 static const struct pr_ops sd_pr_ops = {
2010 .pr_register = sd_pr_register,
2011 .pr_reserve = sd_pr_reserve,
2012 .pr_release = sd_pr_release,
2013 .pr_preempt = sd_pr_preempt,
2014 .pr_clear = sd_pr_clear,
2015 .pr_read_keys = sd_pr_read_keys,
2016 .pr_read_reservation = sd_pr_read_reservation,
2019 static void scsi_disk_free_disk(struct gendisk *disk)
2021 struct scsi_disk *sdkp = scsi_disk(disk);
2023 put_device(&sdkp->disk_dev);
2026 static const struct block_device_operations sd_fops = {
2027 .owner = THIS_MODULE,
2029 .release = sd_release,
2031 .getgeo = sd_getgeo,
2032 .compat_ioctl = blkdev_compat_ptr_ioctl,
2033 .check_events = sd_check_events,
2034 .unlock_native_capacity = sd_unlock_native_capacity,
2035 .report_zones = sd_zbc_report_zones,
2036 .get_unique_id = sd_get_unique_id,
2037 .free_disk = scsi_disk_free_disk,
2038 .pr_ops = &sd_pr_ops,
2042 * sd_eh_reset - reset error handling callback
2043 * @scmd: sd-issued command that has failed
2045 * This function is called by the SCSI midlayer before starting
2046 * SCSI EH. When counting medium access failures we have to be
2047 * careful to register it only only once per device and SCSI EH run;
2048 * there might be several timed out commands which will cause the
2049 * 'max_medium_access_timeouts' counter to trigger after the first
2050 * SCSI EH run already and set the device to offline.
2051 * So this function resets the internal counter before starting SCSI EH.
2053 static void sd_eh_reset(struct scsi_cmnd *scmd)
2055 struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2057 /* New SCSI EH run, reset gate variable */
2058 sdkp->ignore_medium_access_errors = false;
2062 * sd_eh_action - error handling callback
2063 * @scmd: sd-issued command that has failed
2064 * @eh_disp: The recovery disposition suggested by the midlayer
2066 * This function is called by the SCSI midlayer upon completion of an
2067 * error test command (currently TEST UNIT READY). The result of sending
2068 * the eh command is passed in eh_disp. We're looking for devices that
2069 * fail medium access commands but are OK with non access commands like
2070 * test unit ready (so wrongly see the device as having a successful
2073 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
2075 struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2076 struct scsi_device *sdev = scmd->device;
2078 if (!scsi_device_online(sdev) ||
2079 !scsi_medium_access_command(scmd) ||
2080 host_byte(scmd->result) != DID_TIME_OUT ||
2085 * The device has timed out executing a medium access command.
2086 * However, the TEST UNIT READY command sent during error
2087 * handling completed successfully. Either the device is in the
2088 * process of recovering or has it suffered an internal failure
2089 * that prevents access to the storage medium.
2091 if (!sdkp->ignore_medium_access_errors) {
2092 sdkp->medium_access_timed_out++;
2093 sdkp->ignore_medium_access_errors = true;
2097 * If the device keeps failing read/write commands but TEST UNIT
2098 * READY always completes successfully we assume that medium
2099 * access is no longer possible and take the device offline.
2101 if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2102 scmd_printk(KERN_ERR, scmd,
2103 "Medium access timeout failure. Offlining disk!\n");
2104 mutex_lock(&sdev->state_mutex);
2105 scsi_device_set_state(sdev, SDEV_OFFLINE);
2106 mutex_unlock(&sdev->state_mutex);
2114 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2116 struct request *req = scsi_cmd_to_rq(scmd);
2117 struct scsi_device *sdev = scmd->device;
2118 unsigned int transferred, good_bytes;
2119 u64 start_lba, end_lba, bad_lba;
2122 * Some commands have a payload smaller than the device logical
2123 * block size (e.g. INQUIRY on a 4K disk).
2125 if (scsi_bufflen(scmd) <= sdev->sector_size)
2128 /* Check if we have a 'bad_lba' information */
2129 if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2130 SCSI_SENSE_BUFFERSIZE,
2135 * If the bad lba was reported incorrectly, we have no idea where
2138 start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2139 end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2140 if (bad_lba < start_lba || bad_lba >= end_lba)
2144 * resid is optional but mostly filled in. When it's unused,
2145 * its value is zero, so we assume the whole buffer transferred
2147 transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2149 /* This computation should always be done in terms of the
2150 * resolution of the device's medium.
2152 good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2154 return min(good_bytes, transferred);
2158 * sd_done - bottom half handler: called when the lower level
2159 * driver has completed (successfully or otherwise) a scsi command.
2160 * @SCpnt: mid-level's per command structure.
2162 * Note: potentially run from within an ISR. Must not block.
2164 static int sd_done(struct scsi_cmnd *SCpnt)
2166 int result = SCpnt->result;
2167 unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2168 unsigned int sector_size = SCpnt->device->sector_size;
2170 struct scsi_sense_hdr sshdr;
2171 struct request *req = scsi_cmd_to_rq(SCpnt);
2172 struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2173 int sense_valid = 0;
2174 int sense_deferred = 0;
2176 switch (req_op(req)) {
2177 case REQ_OP_DISCARD:
2178 case REQ_OP_WRITE_ZEROES:
2179 case REQ_OP_ZONE_RESET:
2180 case REQ_OP_ZONE_RESET_ALL:
2181 case REQ_OP_ZONE_OPEN:
2182 case REQ_OP_ZONE_CLOSE:
2183 case REQ_OP_ZONE_FINISH:
2185 good_bytes = blk_rq_bytes(req);
2186 scsi_set_resid(SCpnt, 0);
2189 scsi_set_resid(SCpnt, blk_rq_bytes(req));
2194 * In case of bogus fw or device, we could end up having
2195 * an unaligned partial completion. Check this here and force
2198 resid = scsi_get_resid(SCpnt);
2199 if (resid & (sector_size - 1)) {
2200 sd_printk(KERN_INFO, sdkp,
2201 "Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2202 resid, sector_size);
2203 scsi_print_command(SCpnt);
2204 resid = min(scsi_bufflen(SCpnt),
2205 round_up(resid, sector_size));
2206 scsi_set_resid(SCpnt, resid);
2211 sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2213 sense_deferred = scsi_sense_is_deferred(&sshdr);
2215 sdkp->medium_access_timed_out = 0;
2217 if (!scsi_status_is_check_condition(result) &&
2218 (!sense_valid || sense_deferred))
2221 switch (sshdr.sense_key) {
2222 case HARDWARE_ERROR:
2224 good_bytes = sd_completed_bytes(SCpnt);
2226 case RECOVERED_ERROR:
2227 good_bytes = scsi_bufflen(SCpnt);
2230 /* This indicates a false check condition, so ignore it. An
2231 * unknown amount of data was transferred so treat it as an
2235 memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2237 case ABORTED_COMMAND:
2238 if (sshdr.asc == 0x10) /* DIF: Target detected corruption */
2239 good_bytes = sd_completed_bytes(SCpnt);
2241 case ILLEGAL_REQUEST:
2242 switch (sshdr.asc) {
2243 case 0x10: /* DIX: Host detected corruption */
2244 good_bytes = sd_completed_bytes(SCpnt);
2246 case 0x20: /* INVALID COMMAND OPCODE */
2247 case 0x24: /* INVALID FIELD IN CDB */
2248 switch (SCpnt->cmnd[0]) {
2250 sd_config_discard(sdkp, SD_LBP_DISABLE);
2254 if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2255 sd_config_discard(sdkp, SD_LBP_DISABLE);
2257 sdkp->device->no_write_same = 1;
2258 sd_config_write_same(sdkp);
2259 req->rq_flags |= RQF_QUIET;
2270 if (sd_is_zoned(sdkp))
2271 good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2273 SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2274 "sd_done: completed %d of %d bytes\n",
2275 good_bytes, scsi_bufflen(SCpnt)));
2281 * spinup disk - called only in sd_revalidate_disk()
2284 sd_spinup_disk(struct scsi_disk *sdkp)
2286 static const u8 cmd[10] = { TEST_UNIT_READY };
2287 unsigned long spintime_expire = 0;
2288 int spintime, sense_valid = 0;
2289 unsigned int the_result;
2290 struct scsi_sense_hdr sshdr;
2291 struct scsi_failure failure_defs[] = {
2292 /* Do not retry Medium Not Present */
2294 .sense = UNIT_ATTENTION,
2296 .ascq = SCMD_FAILURE_ASCQ_ANY,
2297 .result = SAM_STAT_CHECK_CONDITION,
2302 .ascq = SCMD_FAILURE_ASCQ_ANY,
2303 .result = SAM_STAT_CHECK_CONDITION,
2305 /* Retry when scsi_status_is_good would return false 3 times */
2307 .result = SCMD_FAILURE_STAT_ANY,
2312 struct scsi_failures failures = {
2313 .failure_definitions = failure_defs,
2315 const struct scsi_exec_args exec_args = {
2317 .failures = &failures,
2322 /* Spin up drives, as required. Only do this at boot time */
2323 /* Spinup needs to be done for module loads too. */
2325 bool media_was_present = sdkp->media_present;
2327 scsi_failures_reset_retries(&failures);
2329 the_result = scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN,
2330 NULL, 0, SD_TIMEOUT,
2331 sdkp->max_retries, &exec_args);
2334 if (the_result > 0) {
2336 * If the drive has indicated to us that it doesn't
2337 * have any media in it, don't bother with any more
2340 if (media_not_present(sdkp, &sshdr)) {
2341 if (media_was_present)
2342 sd_printk(KERN_NOTICE, sdkp,
2343 "Media removed, stopped polling\n");
2346 sense_valid = scsi_sense_valid(&sshdr);
2349 if (!scsi_status_is_check_condition(the_result)) {
2350 /* no sense, TUR either succeeded or failed
2351 * with a status error */
2352 if(!spintime && !scsi_status_is_good(the_result)) {
2353 sd_print_result(sdkp, "Test Unit Ready failed",
2360 * The device does not want the automatic start to be issued.
2362 if (sdkp->device->no_start_on_add)
2365 if (sense_valid && sshdr.sense_key == NOT_READY) {
2366 if (sshdr.asc == 4 && sshdr.ascq == 3)
2367 break; /* manual intervention required */
2368 if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2369 break; /* standby */
2370 if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2371 break; /* unavailable */
2372 if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2373 break; /* sanitize in progress */
2374 if (sshdr.asc == 4 && sshdr.ascq == 0x24)
2375 break; /* depopulation in progress */
2376 if (sshdr.asc == 4 && sshdr.ascq == 0x25)
2377 break; /* depopulation restoration in progress */
2379 * Issue command to spin up drive when not ready
2382 /* Return immediately and start spin cycle */
2383 const u8 start_cmd[10] = {
2386 [4] = sdkp->device->start_stop_pwr_cond ?
2390 sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2391 scsi_execute_cmd(sdkp->device, start_cmd,
2392 REQ_OP_DRV_IN, NULL, 0,
2393 SD_TIMEOUT, sdkp->max_retries,
2395 spintime_expire = jiffies + 100 * HZ;
2398 /* Wait 1 second for next try */
2400 printk(KERN_CONT ".");
2403 * Wait for USB flash devices with slow firmware.
2404 * Yes, this sense key/ASC combination shouldn't
2405 * occur here. It's characteristic of these devices.
2407 } else if (sense_valid &&
2408 sshdr.sense_key == UNIT_ATTENTION &&
2409 sshdr.asc == 0x28) {
2411 spintime_expire = jiffies + 5 * HZ;
2414 /* Wait 1 second for next try */
2417 /* we don't understand the sense code, so it's
2418 * probably pointless to loop */
2420 sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2421 sd_print_sense_hdr(sdkp, &sshdr);
2426 } while (spintime && time_before_eq(jiffies, spintime_expire));
2429 if (scsi_status_is_good(the_result))
2430 printk(KERN_CONT "ready\n");
2432 printk(KERN_CONT "not responding...\n");
2437 * Determine whether disk supports Data Integrity Field.
2439 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2441 struct scsi_device *sdp = sdkp->device;
2444 if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2445 sdkp->protection_type = 0;
2449 type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2451 if (type > T10_PI_TYPE3_PROTECTION) {
2452 sd_printk(KERN_ERR, sdkp, "formatted with unsupported" \
2453 " protection type %u. Disabling disk!\n",
2455 sdkp->protection_type = 0;
2459 sdkp->protection_type = type;
2464 static void sd_config_protection(struct scsi_disk *sdkp)
2466 struct scsi_device *sdp = sdkp->device;
2468 sd_dif_config_host(sdkp);
2470 if (!sdkp->protection_type)
2473 if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2474 sd_first_printk(KERN_NOTICE, sdkp,
2475 "Disabling DIF Type %u protection\n",
2476 sdkp->protection_type);
2477 sdkp->protection_type = 0;
2480 sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2481 sdkp->protection_type);
2484 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2485 struct scsi_sense_hdr *sshdr, int sense_valid,
2489 sd_print_sense_hdr(sdkp, sshdr);
2491 sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2494 * Set dirty bit for removable devices if not ready -
2495 * sometimes drives will not report this properly.
2497 if (sdp->removable &&
2498 sense_valid && sshdr->sense_key == NOT_READY)
2499 set_media_not_present(sdkp);
2502 * We used to set media_present to 0 here to indicate no media
2503 * in the drive, but some drives fail read capacity even with
2504 * media present, so we can't do that.
2506 sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2510 #if RC16_LEN > SD_BUF_SIZE
2511 #error RC16_LEN must not be more than SD_BUF_SIZE
2514 #define READ_CAPACITY_RETRIES_ON_RESET 10
2516 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2517 unsigned char *buffer)
2519 unsigned char cmd[16];
2520 struct scsi_sense_hdr sshdr;
2521 const struct scsi_exec_args exec_args = {
2524 int sense_valid = 0;
2526 int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2527 unsigned int alignment;
2528 unsigned long long lba;
2529 unsigned sector_size;
2531 if (sdp->no_read_capacity_16)
2536 cmd[0] = SERVICE_ACTION_IN_16;
2537 cmd[1] = SAI_READ_CAPACITY_16;
2539 memset(buffer, 0, RC16_LEN);
2541 the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2542 buffer, RC16_LEN, SD_TIMEOUT,
2543 sdkp->max_retries, &exec_args);
2544 if (the_result > 0) {
2545 if (media_not_present(sdkp, &sshdr))
2548 sense_valid = scsi_sense_valid(&sshdr);
2550 sshdr.sense_key == ILLEGAL_REQUEST &&
2551 (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2553 /* Invalid Command Operation Code or
2554 * Invalid Field in CDB, just retry
2555 * silently with RC10 */
2558 sshdr.sense_key == UNIT_ATTENTION &&
2559 sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2560 /* Device reset might occur several times,
2561 * give it one more chance */
2562 if (--reset_retries > 0)
2567 } while (the_result && retries);
2570 sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2571 read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2575 sector_size = get_unaligned_be32(&buffer[8]);
2576 lba = get_unaligned_be64(&buffer[0]);
2578 if (sd_read_protection_type(sdkp, buffer) < 0) {
2583 /* Logical blocks per physical block exponent */
2584 sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2587 sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2589 /* Lowest aligned logical block */
2590 alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2591 blk_queue_alignment_offset(sdp->request_queue, alignment);
2592 if (alignment && sdkp->first_scan)
2593 sd_printk(KERN_NOTICE, sdkp,
2594 "physical block alignment offset: %u\n", alignment);
2596 if (buffer[14] & 0x80) { /* LBPME */
2599 if (buffer[14] & 0x40) /* LBPRZ */
2602 sd_config_discard(sdkp, SD_LBP_WS16);
2605 sdkp->capacity = lba + 1;
2609 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2610 unsigned char *buffer)
2612 static const u8 cmd[10] = { READ_CAPACITY };
2613 struct scsi_sense_hdr sshdr;
2614 struct scsi_failure failure_defs[] = {
2615 /* Do not retry Medium Not Present */
2617 .sense = UNIT_ATTENTION,
2619 .result = SAM_STAT_CHECK_CONDITION,
2624 .result = SAM_STAT_CHECK_CONDITION,
2626 /* Device reset might occur several times so retry a lot */
2628 .sense = UNIT_ATTENTION,
2630 .allowed = READ_CAPACITY_RETRIES_ON_RESET,
2631 .result = SAM_STAT_CHECK_CONDITION,
2633 /* Any other error not listed above retry 3 times */
2635 .result = SCMD_FAILURE_RESULT_ANY,
2640 struct scsi_failures failures = {
2641 .failure_definitions = failure_defs,
2643 const struct scsi_exec_args exec_args = {
2645 .failures = &failures,
2647 int sense_valid = 0;
2650 unsigned sector_size;
2652 memset(buffer, 0, 8);
2654 the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2655 8, SD_TIMEOUT, sdkp->max_retries,
2658 if (the_result > 0) {
2659 sense_valid = scsi_sense_valid(&sshdr);
2661 if (media_not_present(sdkp, &sshdr))
2666 sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2667 read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2671 sector_size = get_unaligned_be32(&buffer[4]);
2672 lba = get_unaligned_be32(&buffer[0]);
2674 if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2675 /* Some buggy (usb cardreader) devices return an lba of
2676 0xffffffff when the want to report a size of 0 (with
2677 which they really mean no media is present) */
2679 sdkp->physical_block_size = sector_size;
2683 sdkp->capacity = lba + 1;
2684 sdkp->physical_block_size = sector_size;
2688 static int sd_try_rc16_first(struct scsi_device *sdp)
2690 if (sdp->host->max_cmd_len < 16)
2692 if (sdp->try_rc_10_first)
2694 if (sdp->scsi_level > SCSI_SPC_2)
2696 if (scsi_device_protection(sdp))
2702 * read disk capacity
2705 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2708 struct scsi_device *sdp = sdkp->device;
2710 if (sd_try_rc16_first(sdp)) {
2711 sector_size = read_capacity_16(sdkp, sdp, buffer);
2712 if (sector_size == -EOVERFLOW)
2714 if (sector_size == -ENODEV)
2716 if (sector_size < 0)
2717 sector_size = read_capacity_10(sdkp, sdp, buffer);
2718 if (sector_size < 0)
2721 sector_size = read_capacity_10(sdkp, sdp, buffer);
2722 if (sector_size == -EOVERFLOW)
2724 if (sector_size < 0)
2726 if ((sizeof(sdkp->capacity) > 4) &&
2727 (sdkp->capacity > 0xffffffffULL)) {
2728 int old_sector_size = sector_size;
2729 sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2730 "Trying to use READ CAPACITY(16).\n");
2731 sector_size = read_capacity_16(sdkp, sdp, buffer);
2732 if (sector_size < 0) {
2733 sd_printk(KERN_NOTICE, sdkp,
2734 "Using 0xffffffff as device size\n");
2735 sdkp->capacity = 1 + (sector_t) 0xffffffff;
2736 sector_size = old_sector_size;
2739 /* Remember that READ CAPACITY(16) succeeded */
2740 sdp->try_rc_10_first = 0;
2744 /* Some devices are known to return the total number of blocks,
2745 * not the highest block number. Some devices have versions
2746 * which do this and others which do not. Some devices we might
2747 * suspect of doing this but we don't know for certain.
2749 * If we know the reported capacity is wrong, decrement it. If
2750 * we can only guess, then assume the number of blocks is even
2751 * (usually true but not always) and err on the side of lowering
2754 if (sdp->fix_capacity ||
2755 (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2756 sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2757 "from its reported value: %llu\n",
2758 (unsigned long long) sdkp->capacity);
2763 if (sector_size == 0) {
2765 sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2769 if (sector_size != 512 &&
2770 sector_size != 1024 &&
2771 sector_size != 2048 &&
2772 sector_size != 4096) {
2773 sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2776 * The user might want to re-format the drive with
2777 * a supported sectorsize. Once this happens, it
2778 * would be relatively trivial to set the thing up.
2779 * For this reason, we leave the thing in the table.
2783 * set a bogus sector size so the normal read/write
2784 * logic in the block layer will eventually refuse any
2785 * request on this device without tripping over power
2786 * of two sector size assumptions
2790 blk_queue_logical_block_size(sdp->request_queue, sector_size);
2791 blk_queue_physical_block_size(sdp->request_queue,
2792 sdkp->physical_block_size);
2793 sdkp->device->sector_size = sector_size;
2795 if (sdkp->capacity > 0xffffffff)
2796 sdp->use_16_for_rw = 1;
2801 * Print disk capacity
2804 sd_print_capacity(struct scsi_disk *sdkp,
2805 sector_t old_capacity)
2807 int sector_size = sdkp->device->sector_size;
2808 char cap_str_2[10], cap_str_10[10];
2810 if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2813 string_get_size(sdkp->capacity, sector_size,
2814 STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2815 string_get_size(sdkp->capacity, sector_size,
2816 STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2818 sd_printk(KERN_NOTICE, sdkp,
2819 "%llu %d-byte logical blocks: (%s/%s)\n",
2820 (unsigned long long)sdkp->capacity,
2821 sector_size, cap_str_10, cap_str_2);
2823 if (sdkp->physical_block_size != sector_size)
2824 sd_printk(KERN_NOTICE, sdkp,
2825 "%u-byte physical blocks\n",
2826 sdkp->physical_block_size);
2829 /* called with buffer of length 512 */
2831 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2832 unsigned char *buffer, int len, struct scsi_mode_data *data,
2833 struct scsi_sense_hdr *sshdr)
2836 * If we must use MODE SENSE(10), make sure that the buffer length
2837 * is at least 8 bytes so that the mode sense header fits.
2839 if (sdkp->device->use_10_for_ms && len < 8)
2842 return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2843 SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2847 * read write protect setting, if possible - called only in sd_revalidate_disk()
2848 * called with buffer of length SD_BUF_SIZE
2851 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2854 struct scsi_device *sdp = sdkp->device;
2855 struct scsi_mode_data data;
2856 int old_wp = sdkp->write_prot;
2858 set_disk_ro(sdkp->disk, 0);
2859 if (sdp->skip_ms_page_3f) {
2860 sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2864 if (sdp->use_192_bytes_for_3f) {
2865 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2868 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2869 * We have to start carefully: some devices hang if we ask
2870 * for more than is available.
2872 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2875 * Second attempt: ask for page 0 When only page 0 is
2876 * implemented, a request for page 3F may return Sense Key
2877 * 5: Illegal Request, Sense Code 24: Invalid field in
2881 res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2884 * Third attempt: ask 255 bytes, as we did earlier.
2887 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2892 sd_first_printk(KERN_WARNING, sdkp,
2893 "Test WP failed, assume Write Enabled\n");
2895 sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2896 set_disk_ro(sdkp->disk, sdkp->write_prot);
2897 if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2898 sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2899 sdkp->write_prot ? "on" : "off");
2900 sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2906 * sd_read_cache_type - called only from sd_revalidate_disk()
2907 * called with buffer of length SD_BUF_SIZE
2910 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2913 struct scsi_device *sdp = sdkp->device;
2918 struct scsi_mode_data data;
2919 struct scsi_sense_hdr sshdr;
2920 int old_wce = sdkp->WCE;
2921 int old_rcd = sdkp->RCD;
2922 int old_dpofua = sdkp->DPOFUA;
2925 if (sdkp->cache_override)
2929 if (sdp->skip_ms_page_8) {
2930 if (sdp->type == TYPE_RBC)
2933 if (sdp->skip_ms_page_3f)
2936 if (sdp->use_192_bytes_for_3f)
2940 } else if (sdp->type == TYPE_RBC) {
2948 /* cautiously ask */
2949 res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2955 if (!data.header_length) {
2958 sd_first_printk(KERN_ERR, sdkp,
2959 "Missing header in MODE_SENSE response\n");
2962 /* that went OK, now ask for the proper length */
2966 * We're only interested in the first three bytes, actually.
2967 * But the data cache page is defined for the first 20.
2971 else if (len > SD_BUF_SIZE) {
2972 sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2973 "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2976 if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2980 if (len > first_len)
2981 res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2985 int offset = data.header_length + data.block_descriptor_length;
2987 while (offset < len) {
2988 u8 page_code = buffer[offset] & 0x3F;
2989 u8 spf = buffer[offset] & 0x40;
2991 if (page_code == 8 || page_code == 6) {
2992 /* We're interested only in the first 3 bytes.
2994 if (len - offset <= 2) {
2995 sd_first_printk(KERN_ERR, sdkp,
2996 "Incomplete mode parameter "
3000 modepage = page_code;
3004 /* Go to the next page */
3005 if (spf && len - offset > 3)
3006 offset += 4 + (buffer[offset+2] << 8) +
3008 else if (!spf && len - offset > 1)
3009 offset += 2 + buffer[offset+1];
3011 sd_first_printk(KERN_ERR, sdkp,
3013 "parameter data\n");
3019 sd_first_printk(KERN_WARNING, sdkp,
3020 "No Caching mode page found\n");
3024 if (modepage == 8) {
3025 sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
3026 sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
3028 sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
3032 sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
3033 if (sdp->broken_fua) {
3034 sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
3036 } else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
3037 !sdkp->device->use_16_for_rw) {
3038 sd_first_printk(KERN_NOTICE, sdkp,
3039 "Uses READ/WRITE(6), disabling FUA\n");
3043 /* No cache flush allowed for write protected devices */
3044 if (sdkp->WCE && sdkp->write_prot)
3047 if (sdkp->first_scan || old_wce != sdkp->WCE ||
3048 old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
3049 sd_printk(KERN_NOTICE, sdkp,
3050 "Write cache: %s, read cache: %s, %s\n",
3051 sdkp->WCE ? "enabled" : "disabled",
3052 sdkp->RCD ? "disabled" : "enabled",
3053 sdkp->DPOFUA ? "supports DPO and FUA"
3054 : "doesn't support DPO or FUA");
3060 if (res == -EIO && scsi_sense_valid(&sshdr) &&
3061 sshdr.sense_key == ILLEGAL_REQUEST &&
3062 sshdr.asc == 0x24 && sshdr.ascq == 0x0)
3063 /* Invalid field in CDB */
3064 sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
3066 sd_first_printk(KERN_ERR, sdkp,
3067 "Asking for cache data failed\n");
3070 if (sdp->wce_default_on) {
3071 sd_first_printk(KERN_NOTICE, sdkp,
3072 "Assuming drive cache: write back\n");
3075 sd_first_printk(KERN_WARNING, sdkp,
3076 "Assuming drive cache: write through\n");
3083 static bool sd_is_perm_stream(struct scsi_disk *sdkp, unsigned int stream_id)
3085 u8 cdb[16] = { SERVICE_ACTION_IN_16, SAI_GET_STREAM_STATUS };
3087 struct scsi_stream_status_header h;
3088 struct scsi_stream_status s;
3090 struct scsi_device *sdev = sdkp->device;
3091 struct scsi_sense_hdr sshdr;
3092 const struct scsi_exec_args exec_args = {
3097 put_unaligned_be16(stream_id, &cdb[4]);
3098 put_unaligned_be32(sizeof(buf), &cdb[10]);
3100 res = scsi_execute_cmd(sdev, cdb, REQ_OP_DRV_IN, &buf, sizeof(buf),
3101 SD_TIMEOUT, sdkp->max_retries, &exec_args);
3104 if (scsi_status_is_check_condition(res) && scsi_sense_valid(&sshdr))
3105 sd_print_sense_hdr(sdkp, &sshdr);
3108 if (get_unaligned_be32(&buf.h.len) < sizeof(struct scsi_stream_status))
3110 return buf.h.stream_status[0].perm;
3113 static void sd_read_io_hints(struct scsi_disk *sdkp, unsigned char *buffer)
3115 struct scsi_device *sdp = sdkp->device;
3116 const struct scsi_io_group_descriptor *desc, *start, *end;
3117 u16 permanent_stream_count_old;
3118 struct scsi_sense_hdr sshdr;
3119 struct scsi_mode_data data;
3122 if (sdp->sdev_bflags & BLIST_SKIP_IO_HINTS)
3125 res = scsi_mode_sense(sdp, /*dbd=*/0x8, /*modepage=*/0x0a,
3126 /*subpage=*/0x05, buffer, SD_BUF_SIZE, SD_TIMEOUT,
3127 sdkp->max_retries, &data, &sshdr);
3130 start = (void *)buffer + data.header_length + 16;
3131 end = (void *)buffer + ALIGN_DOWN(data.header_length + data.length,
3134 * From "SBC-5 Constrained Streams with Data Lifetimes": Device severs
3135 * should assign the lowest numbered stream identifiers to permanent
3138 for (desc = start; desc < end; desc++)
3139 if (!desc->st_enble || !sd_is_perm_stream(sdkp, desc - start))
3141 permanent_stream_count_old = sdkp->permanent_stream_count;
3142 sdkp->permanent_stream_count = desc - start;
3143 if (sdkp->rscs && sdkp->permanent_stream_count < 2)
3144 sd_printk(KERN_INFO, sdkp,
3145 "Unexpected: RSCS has been set and the permanent stream count is %u\n",
3146 sdkp->permanent_stream_count);
3147 else if (sdkp->permanent_stream_count != permanent_stream_count_old)
3148 sd_printk(KERN_INFO, sdkp, "permanent stream count = %d\n",
3149 sdkp->permanent_stream_count);
3153 * The ATO bit indicates whether the DIF application tag is available
3154 * for use by the operating system.
3156 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
3159 struct scsi_device *sdp = sdkp->device;
3160 struct scsi_mode_data data;
3161 struct scsi_sense_hdr sshdr;
3163 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
3166 if (sdkp->protection_type == 0)
3169 res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
3170 sdkp->max_retries, &data, &sshdr);
3172 if (res < 0 || !data.header_length ||
3174 sd_first_printk(KERN_WARNING, sdkp,
3175 "getting Control mode page failed, assume no ATO\n");
3177 if (res == -EIO && scsi_sense_valid(&sshdr))
3178 sd_print_sense_hdr(sdkp, &sshdr);
3183 offset = data.header_length + data.block_descriptor_length;
3185 if ((buffer[offset] & 0x3f) != 0x0a) {
3186 sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
3190 if ((buffer[offset + 5] & 0x80) == 0)
3199 * sd_read_block_limits - Query disk device for preferred I/O sizes.
3200 * @sdkp: disk to query
3202 static void sd_read_block_limits(struct scsi_disk *sdkp)
3204 struct scsi_vpd *vpd;
3208 vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3209 if (!vpd || vpd->len < 16)
3212 sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3213 sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3214 sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3216 if (vpd->len >= 64) {
3217 unsigned int lba_count, desc_count;
3219 sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3224 lba_count = get_unaligned_be32(&vpd->data[20]);
3225 desc_count = get_unaligned_be32(&vpd->data[24]);
3227 if (lba_count && desc_count)
3228 sdkp->max_unmap_blocks = lba_count;
3230 sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3232 if (vpd->data[32] & 0x80)
3233 sdkp->unmap_alignment =
3234 get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3236 if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
3238 if (sdkp->max_unmap_blocks)
3239 sd_config_discard(sdkp, SD_LBP_UNMAP);
3241 sd_config_discard(sdkp, SD_LBP_WS16);
3243 } else { /* LBP VPD page tells us what to use */
3244 if (sdkp->lbpu && sdkp->max_unmap_blocks)
3245 sd_config_discard(sdkp, SD_LBP_UNMAP);
3246 else if (sdkp->lbpws)
3247 sd_config_discard(sdkp, SD_LBP_WS16);
3248 else if (sdkp->lbpws10)
3249 sd_config_discard(sdkp, SD_LBP_WS10);
3251 sd_config_discard(sdkp, SD_LBP_DISABLE);
3259 /* Parse the Block Limits Extension VPD page (0xb7) */
3260 static void sd_read_block_limits_ext(struct scsi_disk *sdkp)
3262 struct scsi_vpd *vpd;
3265 vpd = rcu_dereference(sdkp->device->vpd_pgb7);
3266 if (vpd && vpd->len >= 2)
3267 sdkp->rscs = vpd->data[5] & 1;
3272 * sd_read_block_characteristics - Query block dev. characteristics
3273 * @sdkp: disk to query
3275 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
3277 struct request_queue *q = sdkp->disk->queue;
3278 struct scsi_vpd *vpd;
3282 vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3284 if (!vpd || vpd->len < 8) {
3289 rot = get_unaligned_be16(&vpd->data[4]);
3290 sdkp->zoned = (vpd->data[8] >> 4) & 3;
3294 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3295 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3299 #ifdef CONFIG_BLK_DEV_ZONED /* sd_probe rejects ZBD devices early otherwise */
3300 if (sdkp->device->type == TYPE_ZBC) {
3304 disk_set_zoned(sdkp->disk);
3307 * Per ZBC and ZAC specifications, writes in sequential write
3308 * required zones of host-managed devices must be aligned to
3309 * the device physical block size.
3311 blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3314 * Host-aware devices are treated as conventional.
3316 WARN_ON_ONCE(blk_queue_is_zoned(q));
3318 #endif /* CONFIG_BLK_DEV_ZONED */
3320 if (!sdkp->first_scan)
3323 if (blk_queue_is_zoned(q))
3324 sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n");
3325 else if (sdkp->zoned == 1)
3326 sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n");
3327 else if (sdkp->zoned == 2)
3328 sd_printk(KERN_NOTICE, sdkp, "Drive-managed SMR disk\n");
3332 * sd_read_block_provisioning - Query provisioning VPD page
3333 * @sdkp: disk to query
3335 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3337 struct scsi_vpd *vpd;
3339 if (sdkp->lbpme == 0)
3343 vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3345 if (!vpd || vpd->len < 8) {
3351 sdkp->lbpu = (vpd->data[5] >> 7) & 1; /* UNMAP */
3352 sdkp->lbpws = (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3353 sdkp->lbpws10 = (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3357 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3359 struct scsi_device *sdev = sdkp->device;
3361 if (sdev->host->no_write_same) {
3362 sdev->no_write_same = 1;
3367 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3368 struct scsi_vpd *vpd;
3370 sdev->no_report_opcodes = 1;
3372 /* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3373 * CODES is unsupported and the device has an ATA
3374 * Information VPD page (SAT).
3377 vpd = rcu_dereference(sdev->vpd_pg89);
3379 sdev->no_write_same = 1;
3383 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3386 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3390 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3392 struct scsi_device *sdev = sdkp->device;
3394 if (!sdev->security_supported)
3397 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3398 SECURITY_PROTOCOL_IN, 0) == 1 &&
3399 scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3400 SECURITY_PROTOCOL_OUT, 0) == 1)
3404 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3406 return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3410 * sd_read_cpr - Query concurrent positioning ranges
3411 * @sdkp: disk to query
3413 static void sd_read_cpr(struct scsi_disk *sdkp)
3415 struct blk_independent_access_ranges *iars = NULL;
3416 unsigned char *buffer = NULL;
3417 unsigned int nr_cpr = 0;
3418 int i, vpd_len, buf_len = SD_BUF_SIZE;
3422 * We need to have the capacity set first for the block layer to be
3423 * able to check the ranges.
3425 if (sdkp->first_scan)
3428 if (!sdkp->capacity)
3432 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3433 * leading to a maximum page size of 64 + 256*32 bytes.
3435 buf_len = 64 + 256*32;
3436 buffer = kmalloc(buf_len, GFP_KERNEL);
3437 if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3440 /* We must have at least a 64B header and one 32B range descriptor */
3441 vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3442 if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3443 sd_printk(KERN_ERR, sdkp,
3444 "Invalid Concurrent Positioning Ranges VPD page\n");
3448 nr_cpr = (vpd_len - 64) / 32;
3454 iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3461 for (i = 0; i < nr_cpr; i++, desc += 32) {
3463 sd_printk(KERN_ERR, sdkp,
3464 "Invalid Concurrent Positioning Range number\n");
3469 iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3470 iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3474 disk_set_independent_access_ranges(sdkp->disk, iars);
3475 if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3476 sd_printk(KERN_NOTICE, sdkp,
3477 "%u concurrent positioning ranges\n", nr_cpr);
3478 sdkp->nr_actuators = nr_cpr;
3484 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3486 struct scsi_device *sdp = sdkp->device;
3487 unsigned int min_xfer_bytes =
3488 logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3490 if (sdkp->min_xfer_blocks == 0)
3493 if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3494 sd_first_printk(KERN_WARNING, sdkp,
3495 "Preferred minimum I/O size %u bytes not a " \
3496 "multiple of physical block size (%u bytes)\n",
3497 min_xfer_bytes, sdkp->physical_block_size);
3498 sdkp->min_xfer_blocks = 0;
3502 sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3508 * Determine the device's preferred I/O size for reads and writes
3509 * unless the reported value is unreasonably small, large, not a
3510 * multiple of the physical block size, or simply garbage.
3512 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3513 unsigned int dev_max)
3515 struct scsi_device *sdp = sdkp->device;
3516 unsigned int opt_xfer_bytes =
3517 logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3518 unsigned int min_xfer_bytes =
3519 logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3521 if (sdkp->opt_xfer_blocks == 0)
3524 if (sdkp->opt_xfer_blocks > dev_max) {
3525 sd_first_printk(KERN_WARNING, sdkp,
3526 "Optimal transfer size %u logical blocks " \
3527 "> dev_max (%u logical blocks)\n",
3528 sdkp->opt_xfer_blocks, dev_max);
3532 if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3533 sd_first_printk(KERN_WARNING, sdkp,
3534 "Optimal transfer size %u logical blocks " \
3535 "> sd driver limit (%u logical blocks)\n",
3536 sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3540 if (opt_xfer_bytes < PAGE_SIZE) {
3541 sd_first_printk(KERN_WARNING, sdkp,
3542 "Optimal transfer size %u bytes < " \
3543 "PAGE_SIZE (%u bytes)\n",
3544 opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3548 if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3549 sd_first_printk(KERN_WARNING, sdkp,
3550 "Optimal transfer size %u bytes not a " \
3551 "multiple of preferred minimum block " \
3552 "size (%u bytes)\n",
3553 opt_xfer_bytes, min_xfer_bytes);
3557 if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3558 sd_first_printk(KERN_WARNING, sdkp,
3559 "Optimal transfer size %u bytes not a " \
3560 "multiple of physical block size (%u bytes)\n",
3561 opt_xfer_bytes, sdkp->physical_block_size);
3565 sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3570 static void sd_read_block_zero(struct scsi_disk *sdkp)
3572 struct scsi_device *sdev = sdkp->device;
3573 unsigned int buf_len = sdev->sector_size;
3574 u8 *buffer, cmd[16] = { };
3576 buffer = kmalloc(buf_len, GFP_KERNEL);
3580 if (sdev->use_16_for_rw) {
3582 put_unaligned_be64(0, &cmd[2]); /* Logical block address 0 */
3583 put_unaligned_be32(1, &cmd[10]);/* Transfer 1 logical block */
3586 put_unaligned_be32(0, &cmd[2]); /* Logical block address 0 */
3587 put_unaligned_be16(1, &cmd[7]); /* Transfer 1 logical block */
3590 scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN, buffer, buf_len,
3591 SD_TIMEOUT, sdkp->max_retries, NULL);
3596 * sd_revalidate_disk - called the first time a new disk is seen,
3597 * performs disk spin up, read_capacity, etc.
3598 * @disk: struct gendisk we care about
3600 static int sd_revalidate_disk(struct gendisk *disk)
3602 struct scsi_disk *sdkp = scsi_disk(disk);
3603 struct scsi_device *sdp = sdkp->device;
3604 struct request_queue *q = sdkp->disk->queue;
3605 sector_t old_capacity = sdkp->capacity;
3606 unsigned char *buffer;
3607 unsigned int dev_max, rw_max;
3609 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3610 "sd_revalidate_disk\n"));
3613 * If the device is offline, don't try and read capacity or any
3614 * of the other niceties.
3616 if (!scsi_device_online(sdp))
3619 buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3621 sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3622 "allocation failure.\n");
3626 sd_spinup_disk(sdkp);
3629 * Without media there is no reason to ask; moreover, some devices
3630 * react badly if we do.
3632 if (sdkp->media_present) {
3633 sd_read_capacity(sdkp, buffer);
3635 * Some USB/UAS devices return generic values for mode pages
3636 * until the media has been accessed. Trigger a READ operation
3637 * to force the device to populate mode pages.
3639 if (sdp->read_before_ms)
3640 sd_read_block_zero(sdkp);
3642 * set the default to rotational. All non-rotational devices
3643 * support the block characteristics VPD page, which will
3644 * cause this to be updated correctly and any device which
3645 * doesn't support it should be treated as rotational.
3647 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3648 blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3650 if (scsi_device_supports_vpd(sdp)) {
3651 sd_read_block_provisioning(sdkp);
3652 sd_read_block_limits(sdkp);
3653 sd_read_block_limits_ext(sdkp);
3654 sd_read_block_characteristics(sdkp);
3655 sd_zbc_read_zones(sdkp, buffer);
3659 sd_print_capacity(sdkp, old_capacity);
3661 sd_read_write_protect_flag(sdkp, buffer);
3662 sd_read_cache_type(sdkp, buffer);
3663 sd_read_io_hints(sdkp, buffer);
3664 sd_read_app_tag_own(sdkp, buffer);
3665 sd_read_write_same(sdkp, buffer);
3666 sd_read_security(sdkp, buffer);
3667 sd_config_protection(sdkp);
3671 * We now have all cache related info, determine how we deal
3672 * with flush requests.
3674 sd_set_flush_flag(sdkp);
3676 /* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3677 dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3679 /* Some devices report a maximum block count for READ/WRITE requests. */
3680 dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3681 q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3683 if (sd_validate_min_xfer_size(sdkp))
3684 blk_queue_io_min(sdkp->disk->queue,
3685 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3687 blk_queue_io_min(sdkp->disk->queue, 0);
3689 if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3690 q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3691 rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3693 q->limits.io_opt = 0;
3694 rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3695 (sector_t)BLK_DEF_MAX_SECTORS_CAP);
3699 * Limit default to SCSI host optimal sector limit if set. There may be
3700 * an impact on performance for when the size of a request exceeds this
3703 rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3705 /* Do not exceed controller limit */
3706 rw_max = min(rw_max, queue_max_hw_sectors(q));
3709 * Only update max_sectors if previously unset or if the current value
3710 * exceeds the capabilities of the hardware.
3712 if (sdkp->first_scan ||
3713 q->limits.max_sectors > q->limits.max_dev_sectors ||
3714 q->limits.max_sectors > q->limits.max_hw_sectors) {
3715 q->limits.max_sectors = rw_max;
3716 q->limits.max_user_sectors = rw_max;
3719 sdkp->first_scan = 0;
3721 set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3722 sd_config_write_same(sdkp);
3726 * For a zoned drive, revalidating the zones can be done only once
3727 * the gendisk capacity is set. So if this fails, set back the gendisk
3730 if (sd_zbc_revalidate_zones(sdkp))
3731 set_capacity_and_notify(disk, 0);
3738 * sd_unlock_native_capacity - unlock native capacity
3739 * @disk: struct gendisk to set capacity for
3741 * Block layer calls this function if it detects that partitions
3742 * on @disk reach beyond the end of the device. If the SCSI host
3743 * implements ->unlock_native_capacity() method, it's invoked to
3744 * give it a chance to adjust the device capacity.
3747 * Defined by block layer. Might sleep.
3749 static void sd_unlock_native_capacity(struct gendisk *disk)
3751 struct scsi_device *sdev = scsi_disk(disk)->device;
3753 if (sdev->host->hostt->unlock_native_capacity)
3754 sdev->host->hostt->unlock_native_capacity(sdev);
3758 * sd_format_disk_name - format disk name
3759 * @prefix: name prefix - ie. "sd" for SCSI disks
3760 * @index: index of the disk to format name for
3761 * @buf: output buffer
3762 * @buflen: length of the output buffer
3764 * SCSI disk names starts at sda. The 26th device is sdz and the
3765 * 27th is sdaa. The last one for two lettered suffix is sdzz
3766 * which is followed by sdaaa.
3768 * This is basically 26 base counting with one extra 'nil' entry
3769 * at the beginning from the second digit on and can be
3770 * determined using similar method as 26 base conversion with the
3771 * index shifted -1 after each digit is computed.
3777 * 0 on success, -errno on failure.
3779 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3781 const int base = 'z' - 'a' + 1;
3782 char *begin = buf + strlen(prefix);
3783 char *end = buf + buflen;
3793 *--p = 'a' + (index % unit);
3794 index = (index / unit) - 1;
3795 } while (index >= 0);
3797 memmove(begin, p, end - p);
3798 memcpy(buf, prefix, strlen(prefix));
3804 * sd_probe - called during driver initialization and whenever a
3805 * new scsi device is attached to the system. It is called once
3806 * for each scsi device (not just disks) present.
3807 * @dev: pointer to device object
3809 * Returns 0 if successful (or not interested in this scsi device
3810 * (e.g. scanner)); 1 when there is an error.
3812 * Note: this function is invoked from the scsi mid-level.
3813 * This function sets up the mapping between a given
3814 * <host,channel,id,lun> (found in sdp) and new device name
3815 * (e.g. /dev/sda). More precisely it is the block device major
3816 * and minor number that is chosen here.
3818 * Assume sd_probe is not re-entrant (for time being)
3819 * Also think about sd_probe() and sd_remove() running coincidentally.
3821 static int sd_probe(struct device *dev)
3823 struct scsi_device *sdp = to_scsi_device(dev);
3824 struct scsi_disk *sdkp;
3829 scsi_autopm_get_device(sdp);
3831 if (sdp->type != TYPE_DISK &&
3832 sdp->type != TYPE_ZBC &&
3833 sdp->type != TYPE_MOD &&
3834 sdp->type != TYPE_RBC)
3837 if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3838 sdev_printk(KERN_WARNING, sdp,
3839 "Unsupported ZBC host-managed device.\n");
3843 SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3847 sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3851 gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3852 &sd_bio_compl_lkclass);
3856 index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3858 sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3862 error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3864 sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3865 goto out_free_index;
3870 sdkp->index = index;
3871 sdkp->max_retries = SD_MAX_RETRIES;
3872 atomic_set(&sdkp->openers, 0);
3873 atomic_set(&sdkp->device->ioerr_cnt, 0);
3875 if (!sdp->request_queue->rq_timeout) {
3876 if (sdp->type != TYPE_MOD)
3877 blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3879 blk_queue_rq_timeout(sdp->request_queue,
3883 device_initialize(&sdkp->disk_dev);
3884 sdkp->disk_dev.parent = get_device(dev);
3885 sdkp->disk_dev.class = &sd_disk_class;
3886 dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3888 error = device_add(&sdkp->disk_dev);
3890 put_device(&sdkp->disk_dev);
3894 dev_set_drvdata(dev, sdkp);
3896 gd->major = sd_major((index & 0xf0) >> 4);
3897 gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3898 gd->minors = SD_MINORS;
3900 gd->fops = &sd_fops;
3901 gd->private_data = sdkp;
3903 /* defaults, until the device tells us otherwise */
3904 sdp->sector_size = 512;
3906 sdkp->media_present = 1;
3907 sdkp->write_prot = 0;
3908 sdkp->cache_override = 0;
3912 sdkp->first_scan = 1;
3913 sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3915 sd_revalidate_disk(gd);
3917 if (sdp->removable) {
3918 gd->flags |= GENHD_FL_REMOVABLE;
3919 gd->events |= DISK_EVENT_MEDIA_CHANGE;
3920 gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3923 blk_pm_runtime_init(sdp->request_queue, dev);
3924 if (sdp->rpm_autosuspend) {
3925 pm_runtime_set_autosuspend_delay(dev,
3926 sdp->host->rpm_autosuspend_delay);
3929 error = device_add_disk(dev, gd, NULL);
3931 device_unregister(&sdkp->disk_dev);
3936 if (sdkp->security) {
3937 sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3939 sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3942 sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3943 sdp->removable ? "removable " : "");
3944 scsi_autopm_put_device(sdp);
3949 ida_free(&sd_index_ida, index);
3955 scsi_autopm_put_device(sdp);
3960 * sd_remove - called whenever a scsi disk (previously recognized by
3961 * sd_probe) is detached from the system. It is called (potentially
3962 * multiple times) during sd module unload.
3963 * @dev: pointer to device object
3965 * Note: this function is invoked from the scsi mid-level.
3966 * This function potentially frees up a device name (e.g. /dev/sdc)
3967 * that could be re-used by a subsequent sd_probe().
3968 * This function is not called when the built-in sd driver is "exit-ed".
3970 static int sd_remove(struct device *dev)
3972 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3974 scsi_autopm_get_device(sdkp->device);
3976 device_del(&sdkp->disk_dev);
3977 del_gendisk(sdkp->disk);
3978 if (!sdkp->suspended)
3981 put_disk(sdkp->disk);
3985 static void scsi_disk_release(struct device *dev)
3987 struct scsi_disk *sdkp = to_scsi_disk(dev);
3989 ida_free(&sd_index_ida, sdkp->index);
3990 put_device(&sdkp->device->sdev_gendev);
3991 free_opal_dev(sdkp->opal_dev);
3996 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3998 unsigned char cmd[6] = { START_STOP }; /* START_VALID */
3999 struct scsi_sense_hdr sshdr;
4000 const struct scsi_exec_args exec_args = {
4002 .req_flags = BLK_MQ_REQ_PM,
4004 struct scsi_device *sdp = sdkp->device;
4008 cmd[4] |= 1; /* START */
4010 if (sdp->start_stop_pwr_cond)
4011 cmd[4] |= start ? 1 << 4 : 3 << 4; /* Active or Standby */
4013 if (!scsi_device_online(sdp))
4016 res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
4017 sdkp->max_retries, &exec_args);
4019 sd_print_result(sdkp, "Start/Stop Unit failed", res);
4020 if (res > 0 && scsi_sense_valid(&sshdr)) {
4021 sd_print_sense_hdr(sdkp, &sshdr);
4022 /* 0x3a is medium not present */
4023 if (sshdr.asc == 0x3a)
4028 /* SCSI error codes must not go to the generic layer */
4036 * Send a SYNCHRONIZE CACHE instruction down to the device through
4037 * the normal SCSI command structure. Wait for the command to
4040 static void sd_shutdown(struct device *dev)
4042 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4045 return; /* this can happen */
4047 if (pm_runtime_suspended(dev))
4050 if (sdkp->WCE && sdkp->media_present) {
4051 sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4052 sd_sync_cache(sdkp);
4055 if ((system_state != SYSTEM_RESTART &&
4056 sdkp->device->manage_system_start_stop) ||
4057 (system_state == SYSTEM_POWER_OFF &&
4058 sdkp->device->manage_shutdown)) {
4059 sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4060 sd_start_stop_device(sdkp, 0);
4064 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
4066 return (sdev->manage_system_start_stop && !runtime) ||
4067 (sdev->manage_runtime_start_stop && runtime);
4070 static int sd_suspend_common(struct device *dev, bool runtime)
4072 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4075 if (!sdkp) /* E.g.: runtime suspend following sd_remove() */
4078 if (sdkp->WCE && sdkp->media_present) {
4079 if (!sdkp->device->silence_suspend)
4080 sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4081 ret = sd_sync_cache(sdkp);
4082 /* ignore OFFLINE device */
4090 if (sd_do_start_stop(sdkp->device, runtime)) {
4091 if (!sdkp->device->silence_suspend)
4092 sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4093 /* an error is not worth aborting a system sleep */
4094 ret = sd_start_stop_device(sdkp, 0);
4100 sdkp->suspended = true;
4105 static int sd_suspend_system(struct device *dev)
4107 if (pm_runtime_suspended(dev))
4110 return sd_suspend_common(dev, false);
4113 static int sd_suspend_runtime(struct device *dev)
4115 return sd_suspend_common(dev, true);
4118 static int sd_resume(struct device *dev)
4120 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4122 sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4124 if (opal_unlock_from_suspend(sdkp->opal_dev)) {
4125 sd_printk(KERN_NOTICE, sdkp, "OPAL unlock failed\n");
4132 static int sd_resume_common(struct device *dev, bool runtime)
4134 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4137 if (!sdkp) /* E.g.: runtime resume at the start of sd_probe() */
4140 if (!sd_do_start_stop(sdkp->device, runtime)) {
4141 sdkp->suspended = false;
4145 sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4146 ret = sd_start_stop_device(sdkp, 1);
4149 sdkp->suspended = false;
4155 static int sd_resume_system(struct device *dev)
4157 if (pm_runtime_suspended(dev)) {
4158 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4159 struct scsi_device *sdp = sdkp ? sdkp->device : NULL;
4161 if (sdp && sdp->force_runtime_start_on_system_start)
4162 pm_request_resume(dev);
4167 return sd_resume_common(dev, false);
4170 static int sd_resume_runtime(struct device *dev)
4172 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4173 struct scsi_device *sdp;
4175 if (!sdkp) /* E.g.: runtime resume at the start of sd_probe() */
4180 if (sdp->ignore_media_change) {
4181 /* clear the device's sense data */
4182 static const u8 cmd[10] = { REQUEST_SENSE };
4183 const struct scsi_exec_args exec_args = {
4184 .req_flags = BLK_MQ_REQ_PM,
4187 if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
4188 sdp->request_queue->rq_timeout, 1,
4190 sd_printk(KERN_NOTICE, sdkp,
4191 "Failed to clear sense data\n");
4194 return sd_resume_common(dev, true);
4197 static const struct dev_pm_ops sd_pm_ops = {
4198 .suspend = sd_suspend_system,
4199 .resume = sd_resume_system,
4200 .poweroff = sd_suspend_system,
4201 .restore = sd_resume_system,
4202 .runtime_suspend = sd_suspend_runtime,
4203 .runtime_resume = sd_resume_runtime,
4206 static struct scsi_driver sd_template = {
4210 .probe_type = PROBE_PREFER_ASYNCHRONOUS,
4211 .remove = sd_remove,
4212 .shutdown = sd_shutdown,
4215 .rescan = sd_rescan,
4216 .resume = sd_resume,
4217 .init_command = sd_init_command,
4218 .uninit_command = sd_uninit_command,
4220 .eh_action = sd_eh_action,
4221 .eh_reset = sd_eh_reset,
4225 * init_sd - entry point for this driver (both when built in or when
4228 * Note: this function registers this driver with the scsi mid-level.
4230 static int __init init_sd(void)
4232 int majors = 0, i, err;
4234 SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
4236 for (i = 0; i < SD_MAJORS; i++) {
4237 if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
4245 err = class_register(&sd_disk_class);
4249 sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4250 if (!sd_page_pool) {
4251 printk(KERN_ERR "sd: can't init discard page pool\n");
4256 err = scsi_register_driver(&sd_template.gendrv);
4258 goto err_out_driver;
4263 mempool_destroy(sd_page_pool);
4265 class_unregister(&sd_disk_class);
4267 for (i = 0; i < SD_MAJORS; i++)
4268 unregister_blkdev(sd_major(i), "sd");
4273 * exit_sd - exit point for this driver (when it is a module).
4275 * Note: this function unregisters this driver from the scsi mid-level.
4277 static void __exit exit_sd(void)
4281 SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4283 scsi_unregister_driver(&sd_template.gendrv);
4284 mempool_destroy(sd_page_pool);
4286 class_unregister(&sd_disk_class);
4288 for (i = 0; i < SD_MAJORS; i++)
4289 unregister_blkdev(sd_major(i), "sd");
4292 module_init(init_sd);
4293 module_exit(exit_sd);
4295 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4297 scsi_print_sense_hdr(sdkp->device,
4298 sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4301 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4303 const char *hb_string = scsi_hostbyte_string(result);
4306 sd_printk(KERN_INFO, sdkp,
4307 "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4308 hb_string ? hb_string : "invalid",
4311 sd_printk(KERN_INFO, sdkp,
4312 "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4313 msg, host_byte(result), "DRIVER_OK");