1 // SPDX-License-Identifier: GPL-2.0-only
3 * sd.c Copyright (C) 1992 Drew Eckhardt
4 * Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
6 * Linux scsi disk driver
7 * Initial versions: Drew Eckhardt
8 * Subsequent revisions: Eric Youngdale
9 * Modification history:
10 * - Drew Eckhardt <drew@colorado.edu> original
11 * - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12 * outstanding request, and other enhancements.
13 * Support loadable low-level scsi drivers.
14 * - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15 * eight major numbers.
16 * - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17 * - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18 * sd_init and cleanups.
19 * - Alex Davis <letmein@erols.com> Fix problem where partition info
20 * not being read in sd_open. Fix problem where removable media
21 * could be ejected after sd_open.
22 * - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23 * - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24 * <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25 * Support 32k/1M disks.
27 * Logging policy (needs CONFIG_SCSI_LOGGING defined):
28 * - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29 * - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30 * - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31 * - entering other commands: SCSI_LOG_HLQUEUE level 3
32 * Note: when the logging level is set by the user, it must be greater
33 * than the level indicated above to trigger output.
36 #include <linux/module.h>
38 #include <linux/kernel.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 #include <scsi/scsi_common.h>
73 #include "scsi_priv.h"
74 #include "scsi_logging.h"
76 MODULE_AUTHOR("Eric Youngdale");
77 MODULE_DESCRIPTION("SCSI disk (sd) driver");
78 MODULE_LICENSE("GPL");
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
103 static void sd_config_discard(struct scsi_disk *, unsigned int);
104 static void sd_config_write_same(struct scsi_disk *);
105 static int sd_revalidate_disk(struct gendisk *);
106 static void sd_unlock_native_capacity(struct gendisk *disk);
107 static void sd_shutdown(struct device *);
108 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
109 static void scsi_disk_release(struct device *cdev);
111 static DEFINE_IDA(sd_index_ida);
113 static mempool_t *sd_page_pool;
114 static struct lock_class_key sd_bio_compl_lkclass;
116 static const char *sd_cache_types[] = {
117 "write through", "none", "write back",
118 "write back, no read (daft)"
121 static void sd_set_flush_flag(struct scsi_disk *sdkp)
123 bool wc = false, fua = false;
131 blk_queue_write_cache(sdkp->disk->queue, wc, fua);
135 cache_type_store(struct device *dev, struct device_attribute *attr,
136 const char *buf, size_t count)
138 int ct, rcd, wce, sp;
139 struct scsi_disk *sdkp = to_scsi_disk(dev);
140 struct scsi_device *sdp = sdkp->device;
143 struct scsi_mode_data data;
144 struct scsi_sense_hdr sshdr;
145 static const char temp[] = "temporary ";
148 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
149 /* no cache control on RBC devices; theoretically they
150 * can do it, but there's probably so many exceptions
151 * it's not worth the risk */
154 if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
155 buf += sizeof(temp) - 1;
156 sdkp->cache_override = 1;
158 sdkp->cache_override = 0;
161 ct = sysfs_match_string(sd_cache_types, buf);
165 rcd = ct & 0x01 ? 1 : 0;
166 wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
168 if (sdkp->cache_override) {
171 sd_set_flush_flag(sdkp);
175 if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
176 sdkp->max_retries, &data, NULL))
178 len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
179 data.block_descriptor_length);
180 buffer_data = buffer + data.header_length +
181 data.block_descriptor_length;
182 buffer_data[2] &= ~0x05;
183 buffer_data[2] |= wce << 2 | rcd;
184 sp = buffer_data[0] & 0x80 ? 1 : 0;
185 buffer_data[0] &= ~0x80;
188 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
189 * received mode parameter buffer before doing MODE SELECT.
191 data.device_specific = 0;
193 ret = scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
194 sdkp->max_retries, &data, &sshdr);
196 if (ret > 0 && scsi_sense_valid(&sshdr))
197 sd_print_sense_hdr(sdkp, &sshdr);
200 sd_revalidate_disk(sdkp->disk);
205 manage_start_stop_show(struct device *dev,
206 struct device_attribute *attr, char *buf)
208 struct scsi_disk *sdkp = to_scsi_disk(dev);
209 struct scsi_device *sdp = sdkp->device;
211 return sysfs_emit(buf, "%u\n",
212 sdp->manage_system_start_stop &&
213 sdp->manage_runtime_start_stop &&
214 sdp->manage_shutdown);
216 static DEVICE_ATTR_RO(manage_start_stop);
219 manage_system_start_stop_show(struct device *dev,
220 struct device_attribute *attr, char *buf)
222 struct scsi_disk *sdkp = to_scsi_disk(dev);
223 struct scsi_device *sdp = sdkp->device;
225 return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
229 manage_system_start_stop_store(struct device *dev,
230 struct device_attribute *attr,
231 const char *buf, size_t count)
233 struct scsi_disk *sdkp = to_scsi_disk(dev);
234 struct scsi_device *sdp = sdkp->device;
237 if (!capable(CAP_SYS_ADMIN))
240 if (kstrtobool(buf, &v))
243 sdp->manage_system_start_stop = v;
247 static DEVICE_ATTR_RW(manage_system_start_stop);
250 manage_runtime_start_stop_show(struct device *dev,
251 struct device_attribute *attr, char *buf)
253 struct scsi_disk *sdkp = to_scsi_disk(dev);
254 struct scsi_device *sdp = sdkp->device;
256 return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
260 manage_runtime_start_stop_store(struct device *dev,
261 struct device_attribute *attr,
262 const char *buf, size_t count)
264 struct scsi_disk *sdkp = to_scsi_disk(dev);
265 struct scsi_device *sdp = sdkp->device;
268 if (!capable(CAP_SYS_ADMIN))
271 if (kstrtobool(buf, &v))
274 sdp->manage_runtime_start_stop = v;
278 static DEVICE_ATTR_RW(manage_runtime_start_stop);
280 static ssize_t manage_shutdown_show(struct device *dev,
281 struct device_attribute *attr, char *buf)
283 struct scsi_disk *sdkp = to_scsi_disk(dev);
284 struct scsi_device *sdp = sdkp->device;
286 return sysfs_emit(buf, "%u\n", sdp->manage_shutdown);
289 static ssize_t manage_shutdown_store(struct device *dev,
290 struct device_attribute *attr,
291 const char *buf, size_t count)
293 struct scsi_disk *sdkp = to_scsi_disk(dev);
294 struct scsi_device *sdp = sdkp->device;
297 if (!capable(CAP_SYS_ADMIN))
300 if (kstrtobool(buf, &v))
303 sdp->manage_shutdown = v;
307 static DEVICE_ATTR_RW(manage_shutdown);
310 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
312 struct scsi_disk *sdkp = to_scsi_disk(dev);
314 return sprintf(buf, "%u\n", sdkp->device->allow_restart);
318 allow_restart_store(struct device *dev, struct device_attribute *attr,
319 const char *buf, size_t count)
322 struct scsi_disk *sdkp = to_scsi_disk(dev);
323 struct scsi_device *sdp = sdkp->device;
325 if (!capable(CAP_SYS_ADMIN))
328 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
331 if (kstrtobool(buf, &v))
334 sdp->allow_restart = v;
338 static DEVICE_ATTR_RW(allow_restart);
341 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
343 struct scsi_disk *sdkp = to_scsi_disk(dev);
344 int ct = sdkp->RCD + 2*sdkp->WCE;
346 return sprintf(buf, "%s\n", sd_cache_types[ct]);
348 static DEVICE_ATTR_RW(cache_type);
351 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
353 struct scsi_disk *sdkp = to_scsi_disk(dev);
355 return sprintf(buf, "%u\n", sdkp->DPOFUA);
357 static DEVICE_ATTR_RO(FUA);
360 protection_type_show(struct device *dev, struct device_attribute *attr,
363 struct scsi_disk *sdkp = to_scsi_disk(dev);
365 return sprintf(buf, "%u\n", sdkp->protection_type);
369 protection_type_store(struct device *dev, struct device_attribute *attr,
370 const char *buf, size_t count)
372 struct scsi_disk *sdkp = to_scsi_disk(dev);
376 if (!capable(CAP_SYS_ADMIN))
379 err = kstrtouint(buf, 10, &val);
384 if (val <= T10_PI_TYPE3_PROTECTION)
385 sdkp->protection_type = val;
389 static DEVICE_ATTR_RW(protection_type);
392 protection_mode_show(struct device *dev, struct device_attribute *attr,
395 struct scsi_disk *sdkp = to_scsi_disk(dev);
396 struct scsi_device *sdp = sdkp->device;
397 unsigned int dif, dix;
399 dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
400 dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
402 if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
408 return sprintf(buf, "none\n");
410 return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
412 static DEVICE_ATTR_RO(protection_mode);
415 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
417 struct scsi_disk *sdkp = to_scsi_disk(dev);
419 return sprintf(buf, "%u\n", sdkp->ATO);
421 static DEVICE_ATTR_RO(app_tag_own);
424 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
427 struct scsi_disk *sdkp = to_scsi_disk(dev);
429 return sprintf(buf, "%u\n", sdkp->lbpme);
431 static DEVICE_ATTR_RO(thin_provisioning);
433 /* sysfs_match_string() requires dense arrays */
434 static const char *lbp_mode[] = {
435 [SD_LBP_FULL] = "full",
436 [SD_LBP_UNMAP] = "unmap",
437 [SD_LBP_WS16] = "writesame_16",
438 [SD_LBP_WS10] = "writesame_10",
439 [SD_LBP_ZERO] = "writesame_zero",
440 [SD_LBP_DISABLE] = "disabled",
444 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
447 struct scsi_disk *sdkp = to_scsi_disk(dev);
449 return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
453 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
454 const char *buf, size_t count)
456 struct scsi_disk *sdkp = to_scsi_disk(dev);
457 struct scsi_device *sdp = sdkp->device;
460 if (!capable(CAP_SYS_ADMIN))
463 if (sd_is_zoned(sdkp)) {
464 sd_config_discard(sdkp, SD_LBP_DISABLE);
468 if (sdp->type != TYPE_DISK)
471 mode = sysfs_match_string(lbp_mode, buf);
475 sd_config_discard(sdkp, mode);
479 static DEVICE_ATTR_RW(provisioning_mode);
481 /* sysfs_match_string() requires dense arrays */
482 static const char *zeroing_mode[] = {
483 [SD_ZERO_WRITE] = "write",
484 [SD_ZERO_WS] = "writesame",
485 [SD_ZERO_WS16_UNMAP] = "writesame_16_unmap",
486 [SD_ZERO_WS10_UNMAP] = "writesame_10_unmap",
490 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
493 struct scsi_disk *sdkp = to_scsi_disk(dev);
495 return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
499 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
500 const char *buf, size_t count)
502 struct scsi_disk *sdkp = to_scsi_disk(dev);
505 if (!capable(CAP_SYS_ADMIN))
508 mode = sysfs_match_string(zeroing_mode, buf);
512 sdkp->zeroing_mode = mode;
516 static DEVICE_ATTR_RW(zeroing_mode);
519 max_medium_access_timeouts_show(struct device *dev,
520 struct device_attribute *attr, char *buf)
522 struct scsi_disk *sdkp = to_scsi_disk(dev);
524 return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
528 max_medium_access_timeouts_store(struct device *dev,
529 struct device_attribute *attr, const char *buf,
532 struct scsi_disk *sdkp = to_scsi_disk(dev);
535 if (!capable(CAP_SYS_ADMIN))
538 err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
540 return err ? err : count;
542 static DEVICE_ATTR_RW(max_medium_access_timeouts);
545 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
548 struct scsi_disk *sdkp = to_scsi_disk(dev);
550 return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
554 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
555 const char *buf, size_t count)
557 struct scsi_disk *sdkp = to_scsi_disk(dev);
558 struct scsi_device *sdp = sdkp->device;
562 if (!capable(CAP_SYS_ADMIN))
565 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
568 err = kstrtoul(buf, 10, &max);
574 sdp->no_write_same = 1;
575 else if (max <= SD_MAX_WS16_BLOCKS) {
576 sdp->no_write_same = 0;
577 sdkp->max_ws_blocks = max;
580 sd_config_write_same(sdkp);
584 static DEVICE_ATTR_RW(max_write_same_blocks);
587 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
589 struct scsi_disk *sdkp = to_scsi_disk(dev);
591 if (sdkp->device->type == TYPE_ZBC)
592 return sprintf(buf, "host-managed\n");
593 if (sdkp->zoned == 1)
594 return sprintf(buf, "host-aware\n");
595 if (sdkp->zoned == 2)
596 return sprintf(buf, "drive-managed\n");
597 return sprintf(buf, "none\n");
599 static DEVICE_ATTR_RO(zoned_cap);
602 max_retries_store(struct device *dev, struct device_attribute *attr,
603 const char *buf, size_t count)
605 struct scsi_disk *sdkp = to_scsi_disk(dev);
606 struct scsi_device *sdev = sdkp->device;
609 err = kstrtoint(buf, 10, &retries);
613 if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
614 sdkp->max_retries = retries;
618 sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
624 max_retries_show(struct device *dev, struct device_attribute *attr,
627 struct scsi_disk *sdkp = to_scsi_disk(dev);
629 return sprintf(buf, "%d\n", sdkp->max_retries);
632 static DEVICE_ATTR_RW(max_retries);
634 static struct attribute *sd_disk_attrs[] = {
635 &dev_attr_cache_type.attr,
637 &dev_attr_allow_restart.attr,
638 &dev_attr_manage_start_stop.attr,
639 &dev_attr_manage_system_start_stop.attr,
640 &dev_attr_manage_runtime_start_stop.attr,
641 &dev_attr_manage_shutdown.attr,
642 &dev_attr_protection_type.attr,
643 &dev_attr_protection_mode.attr,
644 &dev_attr_app_tag_own.attr,
645 &dev_attr_thin_provisioning.attr,
646 &dev_attr_provisioning_mode.attr,
647 &dev_attr_zeroing_mode.attr,
648 &dev_attr_max_write_same_blocks.attr,
649 &dev_attr_max_medium_access_timeouts.attr,
650 &dev_attr_zoned_cap.attr,
651 &dev_attr_max_retries.attr,
654 ATTRIBUTE_GROUPS(sd_disk);
656 static struct class sd_disk_class = {
658 .dev_release = scsi_disk_release,
659 .dev_groups = sd_disk_groups,
663 * Don't request a new module, as that could deadlock in multipath
666 static void sd_default_probe(dev_t devt)
671 * Device no to disk mapping:
673 * major disc2 disc p1
674 * |............|.............|....|....| <- dev_t
677 * Inside a major, we have 16k disks, however mapped non-
678 * contiguously. The first 16 disks are for major0, the next
679 * ones with major1, ... Disk 256 is for major0 again, disk 272
681 * As we stay compatible with our numbering scheme, we can reuse
682 * the well-know SCSI majors 8, 65--71, 136--143.
684 static int sd_major(int major_idx)
688 return SCSI_DISK0_MAJOR;
690 return SCSI_DISK1_MAJOR + major_idx - 1;
692 return SCSI_DISK8_MAJOR + major_idx - 8;
695 return 0; /* shut up gcc */
699 #ifdef CONFIG_BLK_SED_OPAL
700 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
701 size_t len, bool send)
703 struct scsi_disk *sdkp = data;
704 struct scsi_device *sdev = sdkp->device;
706 const struct scsi_exec_args exec_args = {
707 .req_flags = BLK_MQ_REQ_PM,
711 cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
713 put_unaligned_be16(spsp, &cdb[2]);
714 put_unaligned_be32(len, &cdb[6]);
716 ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
717 buffer, len, SD_TIMEOUT, sdkp->max_retries,
719 return ret <= 0 ? ret : -EIO;
721 #endif /* CONFIG_BLK_SED_OPAL */
724 * Look up the DIX operation based on whether the command is read or
725 * write and whether dix and dif are enabled.
727 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
729 /* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
730 static const unsigned int ops[] = { /* wrt dix dif */
731 SCSI_PROT_NORMAL, /* 0 0 0 */
732 SCSI_PROT_READ_STRIP, /* 0 0 1 */
733 SCSI_PROT_READ_INSERT, /* 0 1 0 */
734 SCSI_PROT_READ_PASS, /* 0 1 1 */
735 SCSI_PROT_NORMAL, /* 1 0 0 */
736 SCSI_PROT_WRITE_INSERT, /* 1 0 1 */
737 SCSI_PROT_WRITE_STRIP, /* 1 1 0 */
738 SCSI_PROT_WRITE_PASS, /* 1 1 1 */
741 return ops[write << 2 | dix << 1 | dif];
745 * Returns a mask of the protection flags that are valid for a given DIX
748 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
750 static const unsigned int flag_mask[] = {
751 [SCSI_PROT_NORMAL] = 0,
753 [SCSI_PROT_READ_STRIP] = SCSI_PROT_TRANSFER_PI |
754 SCSI_PROT_GUARD_CHECK |
755 SCSI_PROT_REF_CHECK |
756 SCSI_PROT_REF_INCREMENT,
758 [SCSI_PROT_READ_INSERT] = SCSI_PROT_REF_INCREMENT |
759 SCSI_PROT_IP_CHECKSUM,
761 [SCSI_PROT_READ_PASS] = SCSI_PROT_TRANSFER_PI |
762 SCSI_PROT_GUARD_CHECK |
763 SCSI_PROT_REF_CHECK |
764 SCSI_PROT_REF_INCREMENT |
765 SCSI_PROT_IP_CHECKSUM,
767 [SCSI_PROT_WRITE_INSERT] = SCSI_PROT_TRANSFER_PI |
768 SCSI_PROT_REF_INCREMENT,
770 [SCSI_PROT_WRITE_STRIP] = SCSI_PROT_GUARD_CHECK |
771 SCSI_PROT_REF_CHECK |
772 SCSI_PROT_REF_INCREMENT |
773 SCSI_PROT_IP_CHECKSUM,
775 [SCSI_PROT_WRITE_PASS] = SCSI_PROT_TRANSFER_PI |
776 SCSI_PROT_GUARD_CHECK |
777 SCSI_PROT_REF_CHECK |
778 SCSI_PROT_REF_INCREMENT |
779 SCSI_PROT_IP_CHECKSUM,
782 return flag_mask[prot_op];
785 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
786 unsigned int dix, unsigned int dif)
788 struct request *rq = scsi_cmd_to_rq(scmd);
789 struct bio *bio = rq->bio;
790 unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
791 unsigned int protect = 0;
793 if (dix) { /* DIX Type 0, 1, 2, 3 */
794 if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
795 scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
797 if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
798 scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
801 if (dif != T10_PI_TYPE3_PROTECTION) { /* DIX/DIF Type 0, 1, 2 */
802 scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
804 if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
805 scmd->prot_flags |= SCSI_PROT_REF_CHECK;
808 if (dif) { /* DIX/DIF Type 1, 2, 3 */
809 scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
811 if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
812 protect = 3 << 5; /* Disable target PI checking */
814 protect = 1 << 5; /* Enable target PI checking */
817 scsi_set_prot_op(scmd, prot_op);
818 scsi_set_prot_type(scmd, dif);
819 scmd->prot_flags &= sd_prot_flag_mask(prot_op);
824 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
826 struct request_queue *q = sdkp->disk->queue;
827 unsigned int logical_block_size = sdkp->device->sector_size;
828 unsigned int max_blocks = 0;
830 q->limits.discard_alignment =
831 sdkp->unmap_alignment * logical_block_size;
832 q->limits.discard_granularity =
833 max(sdkp->physical_block_size,
834 sdkp->unmap_granularity * logical_block_size);
835 sdkp->provisioning_mode = mode;
841 blk_queue_max_discard_sectors(q, 0);
845 max_blocks = min_not_zero(sdkp->max_unmap_blocks,
846 (u32)SD_MAX_WS16_BLOCKS);
850 if (sdkp->device->unmap_limit_for_ws)
851 max_blocks = sdkp->max_unmap_blocks;
853 max_blocks = sdkp->max_ws_blocks;
855 max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
859 if (sdkp->device->unmap_limit_for_ws)
860 max_blocks = sdkp->max_unmap_blocks;
862 max_blocks = sdkp->max_ws_blocks;
864 max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
868 max_blocks = min_not_zero(sdkp->max_ws_blocks,
869 (u32)SD_MAX_WS10_BLOCKS);
873 blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
876 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
880 page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
883 clear_highpage(page);
884 bvec_set_page(&rq->special_vec, page, data_len, 0);
885 rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
886 return bvec_virt(&rq->special_vec);
889 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
891 struct scsi_device *sdp = cmd->device;
892 struct request *rq = scsi_cmd_to_rq(cmd);
893 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
894 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
895 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
896 unsigned int data_len = 24;
899 buf = sd_set_special_bvec(rq, data_len);
901 return BLK_STS_RESOURCE;
904 cmd->cmnd[0] = UNMAP;
907 put_unaligned_be16(6 + 16, &buf[0]);
908 put_unaligned_be16(16, &buf[2]);
909 put_unaligned_be64(lba, &buf[8]);
910 put_unaligned_be32(nr_blocks, &buf[16]);
912 cmd->allowed = sdkp->max_retries;
913 cmd->transfersize = data_len;
914 rq->timeout = SD_TIMEOUT;
916 return scsi_alloc_sgtables(cmd);
919 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
922 struct scsi_device *sdp = cmd->device;
923 struct request *rq = scsi_cmd_to_rq(cmd);
924 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
925 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
926 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
927 u32 data_len = sdp->sector_size;
929 if (!sd_set_special_bvec(rq, data_len))
930 return BLK_STS_RESOURCE;
933 cmd->cmnd[0] = WRITE_SAME_16;
935 cmd->cmnd[1] = 0x8; /* UNMAP */
936 put_unaligned_be64(lba, &cmd->cmnd[2]);
937 put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
939 cmd->allowed = sdkp->max_retries;
940 cmd->transfersize = data_len;
941 rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
943 return scsi_alloc_sgtables(cmd);
946 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
949 struct scsi_device *sdp = cmd->device;
950 struct request *rq = scsi_cmd_to_rq(cmd);
951 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
952 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
953 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
954 u32 data_len = sdp->sector_size;
956 if (!sd_set_special_bvec(rq, data_len))
957 return BLK_STS_RESOURCE;
960 cmd->cmnd[0] = WRITE_SAME;
962 cmd->cmnd[1] = 0x8; /* UNMAP */
963 put_unaligned_be32(lba, &cmd->cmnd[2]);
964 put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
966 cmd->allowed = sdkp->max_retries;
967 cmd->transfersize = data_len;
968 rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
970 return scsi_alloc_sgtables(cmd);
973 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
975 struct request *rq = scsi_cmd_to_rq(cmd);
976 struct scsi_device *sdp = cmd->device;
977 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
978 u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
979 u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
981 if (!(rq->cmd_flags & REQ_NOUNMAP)) {
982 switch (sdkp->zeroing_mode) {
983 case SD_ZERO_WS16_UNMAP:
984 return sd_setup_write_same16_cmnd(cmd, true);
985 case SD_ZERO_WS10_UNMAP:
986 return sd_setup_write_same10_cmnd(cmd, true);
990 if (sdp->no_write_same) {
991 rq->rq_flags |= RQF_QUIET;
992 return BLK_STS_TARGET;
995 if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
996 return sd_setup_write_same16_cmnd(cmd, false);
998 return sd_setup_write_same10_cmnd(cmd, false);
1001 static void sd_config_write_same(struct scsi_disk *sdkp)
1003 struct request_queue *q = sdkp->disk->queue;
1004 unsigned int logical_block_size = sdkp->device->sector_size;
1006 if (sdkp->device->no_write_same) {
1007 sdkp->max_ws_blocks = 0;
1011 /* Some devices can not handle block counts above 0xffff despite
1012 * supporting WRITE SAME(16). Consequently we default to 64k
1013 * blocks per I/O unless the device explicitly advertises a
1016 if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
1017 sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1018 (u32)SD_MAX_WS16_BLOCKS);
1019 else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
1020 sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1021 (u32)SD_MAX_WS10_BLOCKS);
1023 sdkp->device->no_write_same = 1;
1024 sdkp->max_ws_blocks = 0;
1027 if (sdkp->lbprz && sdkp->lbpws)
1028 sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
1029 else if (sdkp->lbprz && sdkp->lbpws10)
1030 sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
1031 else if (sdkp->max_ws_blocks)
1032 sdkp->zeroing_mode = SD_ZERO_WS;
1034 sdkp->zeroing_mode = SD_ZERO_WRITE;
1036 if (sdkp->max_ws_blocks &&
1037 sdkp->physical_block_size > logical_block_size) {
1039 * Reporting a maximum number of blocks that is not aligned
1040 * on the device physical size would cause a large write same
1041 * request to be split into physically unaligned chunks by
1042 * __blkdev_issue_write_zeroes() even if the caller of this
1043 * functions took care to align the large request. So make sure
1044 * the maximum reported is aligned to the device physical block
1045 * size. This is only an optional optimization for regular
1046 * disks, but this is mandatory to avoid failure of large write
1047 * same requests directed at sequential write required zones of
1048 * host-managed ZBC disks.
1050 sdkp->max_ws_blocks =
1051 round_down(sdkp->max_ws_blocks,
1052 bytes_to_logical(sdkp->device,
1053 sdkp->physical_block_size));
1057 blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1058 (logical_block_size >> 9));
1061 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1063 struct request *rq = scsi_cmd_to_rq(cmd);
1064 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1066 /* flush requests don't perform I/O, zero the S/G table */
1067 memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1069 if (cmd->device->use_16_for_sync) {
1070 cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1073 cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1076 cmd->transfersize = 0;
1077 cmd->allowed = sdkp->max_retries;
1079 rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1083 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1084 sector_t lba, unsigned int nr_blocks,
1085 unsigned char flags, unsigned int dld)
1087 cmd->cmd_len = SD_EXT_CDB_SIZE;
1088 cmd->cmnd[0] = VARIABLE_LENGTH_CMD;
1089 cmd->cmnd[7] = 0x18; /* Additional CDB len */
1090 cmd->cmnd[9] = write ? WRITE_32 : READ_32;
1091 cmd->cmnd[10] = flags;
1092 cmd->cmnd[11] = dld & 0x07;
1093 put_unaligned_be64(lba, &cmd->cmnd[12]);
1094 put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1095 put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1100 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1101 sector_t lba, unsigned int nr_blocks,
1102 unsigned char flags, unsigned int dld)
1105 cmd->cmnd[0] = write ? WRITE_16 : READ_16;
1106 cmd->cmnd[1] = flags | ((dld >> 2) & 0x01);
1107 cmd->cmnd[14] = (dld & 0x03) << 6;
1109 put_unaligned_be64(lba, &cmd->cmnd[2]);
1110 put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1115 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1116 sector_t lba, unsigned int nr_blocks,
1117 unsigned char flags)
1120 cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1121 cmd->cmnd[1] = flags;
1124 put_unaligned_be32(lba, &cmd->cmnd[2]);
1125 put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1130 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1131 sector_t lba, unsigned int nr_blocks,
1132 unsigned char flags)
1134 /* Avoid that 0 blocks gets translated into 256 blocks. */
1135 if (WARN_ON_ONCE(nr_blocks == 0))
1136 return BLK_STS_IOERR;
1138 if (unlikely(flags & 0x8)) {
1140 * This happens only if this drive failed 10byte rw
1141 * command with ILLEGAL_REQUEST during operation and
1142 * thus turned off use_10_for_rw.
1144 scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1145 return BLK_STS_IOERR;
1149 cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1150 cmd->cmnd[1] = (lba >> 16) & 0x1f;
1151 cmd->cmnd[2] = (lba >> 8) & 0xff;
1152 cmd->cmnd[3] = lba & 0xff;
1153 cmd->cmnd[4] = nr_blocks;
1160 * Check if a command has a duration limit set. If it does, and the target
1161 * device supports CDL and the feature is enabled, return the limit
1162 * descriptor index to use. Return 0 (no limit) otherwise.
1164 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1166 struct scsi_device *sdp = sdkp->device;
1169 if (!sdp->cdl_supported || !sdp->cdl_enable)
1173 * Use "no limit" if the request ioprio does not specify a duration
1176 hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1177 if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1178 hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1181 return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1184 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1186 struct request *rq = scsi_cmd_to_rq(cmd);
1187 struct scsi_device *sdp = cmd->device;
1188 struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1189 sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1191 unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1192 unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1193 bool write = rq_data_dir(rq) == WRITE;
1194 unsigned char protect, fua;
1200 ret = scsi_alloc_sgtables(cmd);
1201 if (ret != BLK_STS_OK)
1204 ret = BLK_STS_IOERR;
1205 if (!scsi_device_online(sdp) || sdp->changed) {
1206 scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1210 if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1211 scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1215 if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1216 scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1221 * Some SD card readers can't handle accesses which touch the
1222 * last one or two logical blocks. Split accesses as needed.
1224 threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1226 if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1227 if (lba < threshold) {
1228 /* Access up to the threshold but not beyond */
1229 nr_blocks = threshold - lba;
1231 /* Access only a single logical block */
1236 if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1237 ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1242 fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1243 dix = scsi_prot_sg_count(cmd);
1244 dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1245 dld = sd_cdl_dld(sdkp, cmd);
1248 protect = sd_setup_protect_cmnd(cmd, dix, dif);
1252 if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1253 ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1254 protect | fua, dld);
1255 } else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1256 ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1257 protect | fua, dld);
1258 } else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1259 sdp->use_10_for_rw || protect) {
1260 ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1263 ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1267 if (unlikely(ret != BLK_STS_OK))
1271 * We shouldn't disconnect in the middle of a sector, so with a dumb
1272 * host adapter, it's safe to assume that we can at least transfer
1273 * this many bytes between each connect / disconnect.
1275 cmd->transfersize = sdp->sector_size;
1276 cmd->underflow = nr_blocks << 9;
1277 cmd->allowed = sdkp->max_retries;
1278 cmd->sdb.length = nr_blocks * sdp->sector_size;
1281 scmd_printk(KERN_INFO, cmd,
1282 "%s: block=%llu, count=%d\n", __func__,
1283 (unsigned long long)blk_rq_pos(rq),
1284 blk_rq_sectors(rq)));
1286 scmd_printk(KERN_INFO, cmd,
1287 "%s %d/%u 512 byte blocks.\n",
1288 write ? "writing" : "reading", nr_blocks,
1289 blk_rq_sectors(rq)));
1292 * This indicates that the command is ready from our end to be queued.
1296 scsi_free_sgtables(cmd);
1300 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1302 struct request *rq = scsi_cmd_to_rq(cmd);
1304 switch (req_op(rq)) {
1305 case REQ_OP_DISCARD:
1306 switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1308 return sd_setup_unmap_cmnd(cmd);
1310 return sd_setup_write_same16_cmnd(cmd, true);
1312 return sd_setup_write_same10_cmnd(cmd, true);
1314 return sd_setup_write_same10_cmnd(cmd, false);
1316 return BLK_STS_TARGET;
1318 case REQ_OP_WRITE_ZEROES:
1319 return sd_setup_write_zeroes_cmnd(cmd);
1321 return sd_setup_flush_cmnd(cmd);
1324 case REQ_OP_ZONE_APPEND:
1325 return sd_setup_read_write_cmnd(cmd);
1326 case REQ_OP_ZONE_RESET:
1327 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1329 case REQ_OP_ZONE_RESET_ALL:
1330 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1332 case REQ_OP_ZONE_OPEN:
1333 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1334 case REQ_OP_ZONE_CLOSE:
1335 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1336 case REQ_OP_ZONE_FINISH:
1337 return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1340 return BLK_STS_NOTSUPP;
1344 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1346 struct request *rq = scsi_cmd_to_rq(SCpnt);
1348 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1349 mempool_free(rq->special_vec.bv_page, sd_page_pool);
1352 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1354 if (sdkp->device->removable || sdkp->write_prot) {
1355 if (disk_check_media_change(disk))
1360 * Force a full rescan after ioctl(BLKRRPART). While the disk state has
1361 * nothing to do with partitions, BLKRRPART is used to force a full
1362 * revalidate after things like a format for historical reasons.
1364 return test_bit(GD_NEED_PART_SCAN, &disk->state);
1368 * sd_open - open a scsi disk device
1369 * @disk: disk to open
1372 * Returns 0 if successful. Returns a negated errno value in case
1375 * Note: This can be called from a user context (e.g. fsck(1) )
1376 * or from within the kernel (e.g. as a result of a mount(1) ).
1377 * In the latter case @inode and @filp carry an abridged amount
1378 * of information as noted above.
1380 * Locking: called with disk->open_mutex held.
1382 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1384 struct scsi_disk *sdkp = scsi_disk(disk);
1385 struct scsi_device *sdev = sdkp->device;
1388 if (scsi_device_get(sdev))
1391 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1394 * If the device is in error recovery, wait until it is done.
1395 * If the device is offline, then disallow any access to it.
1398 if (!scsi_block_when_processing_errors(sdev))
1401 if (sd_need_revalidate(disk, sdkp))
1402 sd_revalidate_disk(disk);
1405 * If the drive is empty, just let the open fail.
1407 retval = -ENOMEDIUM;
1408 if (sdev->removable && !sdkp->media_present &&
1409 !(mode & BLK_OPEN_NDELAY))
1413 * If the device has the write protect tab set, have the open fail
1414 * if the user expects to be able to write to the thing.
1417 if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1421 * It is possible that the disk changing stuff resulted in
1422 * the device being taken offline. If this is the case,
1423 * report this to the user, and don't pretend that the
1424 * open actually succeeded.
1427 if (!scsi_device_online(sdev))
1430 if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1431 if (scsi_block_when_processing_errors(sdev))
1432 scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1438 scsi_device_put(sdev);
1443 * sd_release - invoked when the (last) close(2) is called on this
1445 * @disk: disk to release
1449 * Note: may block (uninterruptible) if error recovery is underway
1452 * Locking: called with disk->open_mutex held.
1454 static void sd_release(struct gendisk *disk)
1456 struct scsi_disk *sdkp = scsi_disk(disk);
1457 struct scsi_device *sdev = sdkp->device;
1459 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1461 if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1462 if (scsi_block_when_processing_errors(sdev))
1463 scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1466 scsi_device_put(sdev);
1469 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1471 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1472 struct scsi_device *sdp = sdkp->device;
1473 struct Scsi_Host *host = sdp->host;
1474 sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1477 /* default to most commonly used values */
1478 diskinfo[0] = 0x40; /* 1 << 6 */
1479 diskinfo[1] = 0x20; /* 1 << 5 */
1480 diskinfo[2] = capacity >> 11;
1482 /* override with calculated, extended default, or driver values */
1483 if (host->hostt->bios_param)
1484 host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1486 scsicam_bios_param(bdev, capacity, diskinfo);
1488 geo->heads = diskinfo[0];
1489 geo->sectors = diskinfo[1];
1490 geo->cylinders = diskinfo[2];
1495 * sd_ioctl - process an ioctl
1496 * @bdev: target block device
1498 * @cmd: ioctl command number
1499 * @arg: this is third argument given to ioctl(2) system call.
1500 * Often contains a pointer.
1502 * Returns 0 if successful (some ioctls return positive numbers on
1503 * success as well). Returns a negated errno value in case of error.
1505 * Note: most ioctls are forward onto the block subsystem or further
1506 * down in the scsi subsystem.
1508 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1509 unsigned int cmd, unsigned long arg)
1511 struct gendisk *disk = bdev->bd_disk;
1512 struct scsi_disk *sdkp = scsi_disk(disk);
1513 struct scsi_device *sdp = sdkp->device;
1514 void __user *p = (void __user *)arg;
1517 SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1518 "cmd=0x%x\n", disk->disk_name, cmd));
1520 if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1521 return -ENOIOCTLCMD;
1524 * If we are in the middle of error recovery, don't let anyone
1525 * else try and use this device. Also, if error recovery fails, it
1526 * may try and take the device offline, in which case all further
1527 * access to the device is prohibited.
1529 error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1530 (mode & BLK_OPEN_NDELAY));
1534 if (is_sed_ioctl(cmd))
1535 return sed_ioctl(sdkp->opal_dev, cmd, p);
1536 return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1539 static void set_media_not_present(struct scsi_disk *sdkp)
1541 if (sdkp->media_present)
1542 sdkp->device->changed = 1;
1544 if (sdkp->device->removable) {
1545 sdkp->media_present = 0;
1550 static int media_not_present(struct scsi_disk *sdkp,
1551 struct scsi_sense_hdr *sshdr)
1553 if (!scsi_sense_valid(sshdr))
1556 /* not invoked for commands that could return deferred errors */
1557 switch (sshdr->sense_key) {
1558 case UNIT_ATTENTION:
1560 /* medium not present */
1561 if (sshdr->asc == 0x3A) {
1562 set_media_not_present(sdkp);
1570 * sd_check_events - check media events
1571 * @disk: kernel device descriptor
1572 * @clearing: disk events currently being cleared
1574 * Returns mask of DISK_EVENT_*.
1576 * Note: this function is invoked from the block subsystem.
1578 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1580 struct scsi_disk *sdkp = disk->private_data;
1581 struct scsi_device *sdp;
1589 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1592 * If the device is offline, don't send any commands - just pretend as
1593 * if the command failed. If the device ever comes back online, we
1594 * can deal with it then. It is only because of unrecoverable errors
1595 * that we would ever take a device offline in the first place.
1597 if (!scsi_device_online(sdp)) {
1598 set_media_not_present(sdkp);
1603 * Using TEST_UNIT_READY enables differentiation between drive with
1604 * no cartridge loaded - NOT READY, drive with changed cartridge -
1605 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1607 * Drives that auto spin down. eg iomega jaz 1G, will be started
1608 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1609 * sd_revalidate() is called.
1611 if (scsi_block_when_processing_errors(sdp)) {
1612 struct scsi_sense_hdr sshdr = { 0, };
1614 retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1617 /* failed to execute TUR, assume media not present */
1618 if (retval < 0 || host_byte(retval)) {
1619 set_media_not_present(sdkp);
1623 if (media_not_present(sdkp, &sshdr))
1628 * For removable scsi disk we have to recognise the presence
1629 * of a disk in the drive.
1631 if (!sdkp->media_present)
1633 sdkp->media_present = 1;
1636 * sdp->changed is set under the following conditions:
1638 * Medium present state has changed in either direction.
1639 * Device has indicated UNIT_ATTENTION.
1641 disk_changed = sdp->changed;
1643 return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1646 static int sd_sync_cache(struct scsi_disk *sdkp)
1649 struct scsi_device *sdp = sdkp->device;
1650 const int timeout = sdp->request_queue->rq_timeout
1651 * SD_FLUSH_TIMEOUT_MULTIPLIER;
1652 /* Leave the rest of the command zero to indicate flush everything. */
1653 const unsigned char cmd[16] = { sdp->use_16_for_sync ?
1654 SYNCHRONIZE_CACHE_16 : SYNCHRONIZE_CACHE };
1655 struct scsi_sense_hdr sshdr;
1656 struct scsi_failure failure_defs[] = {
1659 .result = SCMD_FAILURE_RESULT_ANY,
1663 struct scsi_failures failures = {
1664 .failure_definitions = failure_defs,
1666 const struct scsi_exec_args exec_args = {
1667 .req_flags = BLK_MQ_REQ_PM,
1669 .failures = &failures,
1672 if (!scsi_device_online(sdp))
1675 res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, timeout,
1676 sdkp->max_retries, &exec_args);
1678 sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1683 if (scsi_status_is_check_condition(res) &&
1684 scsi_sense_valid(&sshdr)) {
1685 sd_print_sense_hdr(sdkp, &sshdr);
1687 /* we need to evaluate the error return */
1688 if (sshdr.asc == 0x3a || /* medium not present */
1689 sshdr.asc == 0x20 || /* invalid command */
1690 (sshdr.asc == 0x74 && sshdr.ascq == 0x71)) /* drive is password locked */
1691 /* this is no error here */
1694 * This drive doesn't support sync and there's not much
1695 * we can do because this is called during shutdown
1696 * or suspend so just return success so those operations
1699 if (sshdr.sense_key == ILLEGAL_REQUEST)
1703 switch (host_byte(res)) {
1704 /* ignore errors due to racing a disconnection */
1705 case DID_BAD_TARGET:
1706 case DID_NO_CONNECT:
1708 /* signal the upper layer it might try again */
1712 case DID_SOFT_ERROR:
1721 static void sd_rescan(struct device *dev)
1723 struct scsi_disk *sdkp = dev_get_drvdata(dev);
1725 sd_revalidate_disk(sdkp->disk);
1728 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1729 enum blk_unique_id type)
1731 struct scsi_device *sdev = scsi_disk(disk)->device;
1732 const struct scsi_vpd *vpd;
1733 const unsigned char *d;
1734 int ret = -ENXIO, len;
1737 vpd = rcu_dereference(sdev->vpd_pg83);
1742 for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1743 /* we only care about designators with LU association */
1744 if (((d[1] >> 4) & 0x3) != 0x00)
1746 if ((d[1] & 0xf) != type)
1750 * Only exit early if a 16-byte descriptor was found. Otherwise
1751 * keep looking as one with more entropy might still show up.
1754 if (len != 8 && len != 12 && len != 16)
1757 memcpy(id, d + 4, len);
1766 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1768 switch (host_byte(result)) {
1769 case DID_TRANSPORT_MARGINAL:
1770 case DID_TRANSPORT_DISRUPTED:
1772 return PR_STS_RETRY_PATH_FAILURE;
1773 case DID_NO_CONNECT:
1774 return PR_STS_PATH_FAILED;
1775 case DID_TRANSPORT_FAILFAST:
1776 return PR_STS_PATH_FAST_FAILED;
1779 switch (status_byte(result)) {
1780 case SAM_STAT_RESERVATION_CONFLICT:
1781 return PR_STS_RESERVATION_CONFLICT;
1782 case SAM_STAT_CHECK_CONDITION:
1783 if (!scsi_sense_valid(sshdr))
1784 return PR_STS_IOERR;
1786 if (sshdr->sense_key == ILLEGAL_REQUEST &&
1787 (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1792 return PR_STS_IOERR;
1796 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1797 unsigned char *data, int data_len)
1799 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1800 struct scsi_device *sdev = sdkp->device;
1801 struct scsi_sense_hdr sshdr;
1802 u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1803 struct scsi_failure failure_defs[] = {
1805 .sense = UNIT_ATTENTION,
1806 .asc = SCMD_FAILURE_ASC_ANY,
1807 .ascq = SCMD_FAILURE_ASCQ_ANY,
1809 .result = SAM_STAT_CHECK_CONDITION,
1813 struct scsi_failures failures = {
1814 .failure_definitions = failure_defs,
1816 const struct scsi_exec_args exec_args = {
1818 .failures = &failures,
1822 put_unaligned_be16(data_len, &cmd[7]);
1824 result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1825 SD_TIMEOUT, sdkp->max_retries, &exec_args);
1826 if (scsi_status_is_check_condition(result) &&
1827 scsi_sense_valid(&sshdr)) {
1828 sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1829 scsi_print_sense_hdr(sdev, NULL, &sshdr);
1835 return sd_scsi_to_pr_err(&sshdr, result);
1838 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1840 int result, i, data_offset, num_copy_keys;
1841 u32 num_keys = keys_info->num_keys;
1842 int data_len = num_keys * 8 + 8;
1845 data = kzalloc(data_len, GFP_KERNEL);
1849 result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1853 keys_info->generation = get_unaligned_be32(&data[0]);
1854 keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1857 num_copy_keys = min(num_keys, keys_info->num_keys);
1859 for (i = 0; i < num_copy_keys; i++) {
1860 keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1869 static int sd_pr_read_reservation(struct block_device *bdev,
1870 struct pr_held_reservation *rsv)
1872 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1873 struct scsi_device *sdev = sdkp->device;
1877 result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1881 len = get_unaligned_be32(&data[4]);
1885 /* Make sure we have at least the key and type */
1887 sdev_printk(KERN_INFO, sdev,
1888 "READ RESERVATION failed due to short return buffer of %d bytes\n",
1893 rsv->generation = get_unaligned_be32(&data[0]);
1894 rsv->key = get_unaligned_be64(&data[8]);
1895 rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1899 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1900 u64 sa_key, enum scsi_pr_type type, u8 flags)
1902 struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1903 struct scsi_device *sdev = sdkp->device;
1904 struct scsi_sense_hdr sshdr;
1905 struct scsi_failure failure_defs[] = {
1907 .sense = UNIT_ATTENTION,
1908 .asc = SCMD_FAILURE_ASC_ANY,
1909 .ascq = SCMD_FAILURE_ASCQ_ANY,
1911 .result = SAM_STAT_CHECK_CONDITION,
1915 struct scsi_failures failures = {
1916 .failure_definitions = failure_defs,
1918 const struct scsi_exec_args exec_args = {
1920 .failures = &failures,
1923 u8 cmd[16] = { 0, };
1924 u8 data[24] = { 0, };
1926 cmd[0] = PERSISTENT_RESERVE_OUT;
1929 put_unaligned_be32(sizeof(data), &cmd[5]);
1931 put_unaligned_be64(key, &data[0]);
1932 put_unaligned_be64(sa_key, &data[8]);
1935 result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1936 sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1939 if (scsi_status_is_check_condition(result) &&
1940 scsi_sense_valid(&sshdr)) {
1941 sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1942 scsi_print_sense_hdr(sdev, NULL, &sshdr);
1948 return sd_scsi_to_pr_err(&sshdr, result);
1951 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1954 if (flags & ~PR_FL_IGNORE_KEY)
1956 return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1957 old_key, new_key, 0,
1958 (1 << 0) /* APTPL */);
1961 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1966 return sd_pr_out_command(bdev, 0x01, key, 0,
1967 block_pr_type_to_scsi(type), 0);
1970 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1972 return sd_pr_out_command(bdev, 0x02, key, 0,
1973 block_pr_type_to_scsi(type), 0);
1976 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1977 enum pr_type type, bool abort)
1979 return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1980 block_pr_type_to_scsi(type), 0);
1983 static int sd_pr_clear(struct block_device *bdev, u64 key)
1985 return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
1988 static const struct pr_ops sd_pr_ops = {
1989 .pr_register = sd_pr_register,
1990 .pr_reserve = sd_pr_reserve,
1991 .pr_release = sd_pr_release,
1992 .pr_preempt = sd_pr_preempt,
1993 .pr_clear = sd_pr_clear,
1994 .pr_read_keys = sd_pr_read_keys,
1995 .pr_read_reservation = sd_pr_read_reservation,
1998 static void scsi_disk_free_disk(struct gendisk *disk)
2000 struct scsi_disk *sdkp = scsi_disk(disk);
2002 put_device(&sdkp->disk_dev);
2005 static const struct block_device_operations sd_fops = {
2006 .owner = THIS_MODULE,
2008 .release = sd_release,
2010 .getgeo = sd_getgeo,
2011 .compat_ioctl = blkdev_compat_ptr_ioctl,
2012 .check_events = sd_check_events,
2013 .unlock_native_capacity = sd_unlock_native_capacity,
2014 .report_zones = sd_zbc_report_zones,
2015 .get_unique_id = sd_get_unique_id,
2016 .free_disk = scsi_disk_free_disk,
2017 .pr_ops = &sd_pr_ops,
2021 * sd_eh_reset - reset error handling callback
2022 * @scmd: sd-issued command that has failed
2024 * This function is called by the SCSI midlayer before starting
2025 * SCSI EH. When counting medium access failures we have to be
2026 * careful to register it only only once per device and SCSI EH run;
2027 * there might be several timed out commands which will cause the
2028 * 'max_medium_access_timeouts' counter to trigger after the first
2029 * SCSI EH run already and set the device to offline.
2030 * So this function resets the internal counter before starting SCSI EH.
2032 static void sd_eh_reset(struct scsi_cmnd *scmd)
2034 struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2036 /* New SCSI EH run, reset gate variable */
2037 sdkp->ignore_medium_access_errors = false;
2041 * sd_eh_action - error handling callback
2042 * @scmd: sd-issued command that has failed
2043 * @eh_disp: The recovery disposition suggested by the midlayer
2045 * This function is called by the SCSI midlayer upon completion of an
2046 * error test command (currently TEST UNIT READY). The result of sending
2047 * the eh command is passed in eh_disp. We're looking for devices that
2048 * fail medium access commands but are OK with non access commands like
2049 * test unit ready (so wrongly see the device as having a successful
2052 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
2054 struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2055 struct scsi_device *sdev = scmd->device;
2057 if (!scsi_device_online(sdev) ||
2058 !scsi_medium_access_command(scmd) ||
2059 host_byte(scmd->result) != DID_TIME_OUT ||
2064 * The device has timed out executing a medium access command.
2065 * However, the TEST UNIT READY command sent during error
2066 * handling completed successfully. Either the device is in the
2067 * process of recovering or has it suffered an internal failure
2068 * that prevents access to the storage medium.
2070 if (!sdkp->ignore_medium_access_errors) {
2071 sdkp->medium_access_timed_out++;
2072 sdkp->ignore_medium_access_errors = true;
2076 * If the device keeps failing read/write commands but TEST UNIT
2077 * READY always completes successfully we assume that medium
2078 * access is no longer possible and take the device offline.
2080 if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2081 scmd_printk(KERN_ERR, scmd,
2082 "Medium access timeout failure. Offlining disk!\n");
2083 mutex_lock(&sdev->state_mutex);
2084 scsi_device_set_state(sdev, SDEV_OFFLINE);
2085 mutex_unlock(&sdev->state_mutex);
2093 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2095 struct request *req = scsi_cmd_to_rq(scmd);
2096 struct scsi_device *sdev = scmd->device;
2097 unsigned int transferred, good_bytes;
2098 u64 start_lba, end_lba, bad_lba;
2101 * Some commands have a payload smaller than the device logical
2102 * block size (e.g. INQUIRY on a 4K disk).
2104 if (scsi_bufflen(scmd) <= sdev->sector_size)
2107 /* Check if we have a 'bad_lba' information */
2108 if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2109 SCSI_SENSE_BUFFERSIZE,
2114 * If the bad lba was reported incorrectly, we have no idea where
2117 start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2118 end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2119 if (bad_lba < start_lba || bad_lba >= end_lba)
2123 * resid is optional but mostly filled in. When it's unused,
2124 * its value is zero, so we assume the whole buffer transferred
2126 transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2128 /* This computation should always be done in terms of the
2129 * resolution of the device's medium.
2131 good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2133 return min(good_bytes, transferred);
2137 * sd_done - bottom half handler: called when the lower level
2138 * driver has completed (successfully or otherwise) a scsi command.
2139 * @SCpnt: mid-level's per command structure.
2141 * Note: potentially run from within an ISR. Must not block.
2143 static int sd_done(struct scsi_cmnd *SCpnt)
2145 int result = SCpnt->result;
2146 unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2147 unsigned int sector_size = SCpnt->device->sector_size;
2149 struct scsi_sense_hdr sshdr;
2150 struct request *req = scsi_cmd_to_rq(SCpnt);
2151 struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2152 int sense_valid = 0;
2153 int sense_deferred = 0;
2155 switch (req_op(req)) {
2156 case REQ_OP_DISCARD:
2157 case REQ_OP_WRITE_ZEROES:
2158 case REQ_OP_ZONE_RESET:
2159 case REQ_OP_ZONE_RESET_ALL:
2160 case REQ_OP_ZONE_OPEN:
2161 case REQ_OP_ZONE_CLOSE:
2162 case REQ_OP_ZONE_FINISH:
2164 good_bytes = blk_rq_bytes(req);
2165 scsi_set_resid(SCpnt, 0);
2168 scsi_set_resid(SCpnt, blk_rq_bytes(req));
2173 * In case of bogus fw or device, we could end up having
2174 * an unaligned partial completion. Check this here and force
2177 resid = scsi_get_resid(SCpnt);
2178 if (resid & (sector_size - 1)) {
2179 sd_printk(KERN_INFO, sdkp,
2180 "Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2181 resid, sector_size);
2182 scsi_print_command(SCpnt);
2183 resid = min(scsi_bufflen(SCpnt),
2184 round_up(resid, sector_size));
2185 scsi_set_resid(SCpnt, resid);
2190 sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2192 sense_deferred = scsi_sense_is_deferred(&sshdr);
2194 sdkp->medium_access_timed_out = 0;
2196 if (!scsi_status_is_check_condition(result) &&
2197 (!sense_valid || sense_deferred))
2200 switch (sshdr.sense_key) {
2201 case HARDWARE_ERROR:
2203 good_bytes = sd_completed_bytes(SCpnt);
2205 case RECOVERED_ERROR:
2206 good_bytes = scsi_bufflen(SCpnt);
2209 /* This indicates a false check condition, so ignore it. An
2210 * unknown amount of data was transferred so treat it as an
2214 memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2216 case ABORTED_COMMAND:
2217 if (sshdr.asc == 0x10) /* DIF: Target detected corruption */
2218 good_bytes = sd_completed_bytes(SCpnt);
2220 case ILLEGAL_REQUEST:
2221 switch (sshdr.asc) {
2222 case 0x10: /* DIX: Host detected corruption */
2223 good_bytes = sd_completed_bytes(SCpnt);
2225 case 0x20: /* INVALID COMMAND OPCODE */
2226 case 0x24: /* INVALID FIELD IN CDB */
2227 switch (SCpnt->cmnd[0]) {
2229 sd_config_discard(sdkp, SD_LBP_DISABLE);
2233 if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2234 sd_config_discard(sdkp, SD_LBP_DISABLE);
2236 sdkp->device->no_write_same = 1;
2237 sd_config_write_same(sdkp);
2238 req->rq_flags |= RQF_QUIET;
2249 if (sd_is_zoned(sdkp))
2250 good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2252 SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2253 "sd_done: completed %d of %d bytes\n",
2254 good_bytes, scsi_bufflen(SCpnt)));
2260 * spinup disk - called only in sd_revalidate_disk()
2263 sd_spinup_disk(struct scsi_disk *sdkp)
2265 static const u8 cmd[10] = { TEST_UNIT_READY };
2266 unsigned long spintime_expire = 0;
2267 int spintime, sense_valid = 0;
2268 unsigned int the_result;
2269 struct scsi_sense_hdr sshdr;
2270 struct scsi_failure failure_defs[] = {
2271 /* Do not retry Medium Not Present */
2273 .sense = UNIT_ATTENTION,
2275 .ascq = SCMD_FAILURE_ASCQ_ANY,
2276 .result = SAM_STAT_CHECK_CONDITION,
2281 .ascq = SCMD_FAILURE_ASCQ_ANY,
2282 .result = SAM_STAT_CHECK_CONDITION,
2284 /* Retry when scsi_status_is_good would return false 3 times */
2286 .result = SCMD_FAILURE_STAT_ANY,
2291 struct scsi_failures failures = {
2292 .failure_definitions = failure_defs,
2294 const struct scsi_exec_args exec_args = {
2296 .failures = &failures,
2301 /* Spin up drives, as required. Only do this at boot time */
2302 /* Spinup needs to be done for module loads too. */
2304 bool media_was_present = sdkp->media_present;
2306 scsi_failures_reset_retries(&failures);
2308 the_result = scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN,
2309 NULL, 0, SD_TIMEOUT,
2310 sdkp->max_retries, &exec_args);
2313 if (the_result > 0) {
2315 * If the drive has indicated to us that it doesn't
2316 * have any media in it, don't bother with any more
2319 if (media_not_present(sdkp, &sshdr)) {
2320 if (media_was_present)
2321 sd_printk(KERN_NOTICE, sdkp,
2322 "Media removed, stopped polling\n");
2325 sense_valid = scsi_sense_valid(&sshdr);
2328 if (!scsi_status_is_check_condition(the_result)) {
2329 /* no sense, TUR either succeeded or failed
2330 * with a status error */
2331 if(!spintime && !scsi_status_is_good(the_result)) {
2332 sd_print_result(sdkp, "Test Unit Ready failed",
2339 * The device does not want the automatic start to be issued.
2341 if (sdkp->device->no_start_on_add)
2344 if (sense_valid && sshdr.sense_key == NOT_READY) {
2345 if (sshdr.asc == 4 && sshdr.ascq == 3)
2346 break; /* manual intervention required */
2347 if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2348 break; /* standby */
2349 if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2350 break; /* unavailable */
2351 if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2352 break; /* sanitize in progress */
2353 if (sshdr.asc == 4 && sshdr.ascq == 0x24)
2354 break; /* depopulation in progress */
2355 if (sshdr.asc == 4 && sshdr.ascq == 0x25)
2356 break; /* depopulation restoration in progress */
2358 * Issue command to spin up drive when not ready
2361 /* Return immediately and start spin cycle */
2362 const u8 start_cmd[10] = {
2365 [4] = sdkp->device->start_stop_pwr_cond ?
2369 sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2370 scsi_execute_cmd(sdkp->device, start_cmd,
2371 REQ_OP_DRV_IN, NULL, 0,
2372 SD_TIMEOUT, sdkp->max_retries,
2374 spintime_expire = jiffies + 100 * HZ;
2377 /* Wait 1 second for next try */
2379 printk(KERN_CONT ".");
2382 * Wait for USB flash devices with slow firmware.
2383 * Yes, this sense key/ASC combination shouldn't
2384 * occur here. It's characteristic of these devices.
2386 } else if (sense_valid &&
2387 sshdr.sense_key == UNIT_ATTENTION &&
2388 sshdr.asc == 0x28) {
2390 spintime_expire = jiffies + 5 * HZ;
2393 /* Wait 1 second for next try */
2396 /* we don't understand the sense code, so it's
2397 * probably pointless to loop */
2399 sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2400 sd_print_sense_hdr(sdkp, &sshdr);
2405 } while (spintime && time_before_eq(jiffies, spintime_expire));
2408 if (scsi_status_is_good(the_result))
2409 printk(KERN_CONT "ready\n");
2411 printk(KERN_CONT "not responding...\n");
2416 * Determine whether disk supports Data Integrity Field.
2418 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2420 struct scsi_device *sdp = sdkp->device;
2423 if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2424 sdkp->protection_type = 0;
2428 type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2430 if (type > T10_PI_TYPE3_PROTECTION) {
2431 sd_printk(KERN_ERR, sdkp, "formatted with unsupported" \
2432 " protection type %u. Disabling disk!\n",
2434 sdkp->protection_type = 0;
2438 sdkp->protection_type = type;
2443 static void sd_config_protection(struct scsi_disk *sdkp)
2445 struct scsi_device *sdp = sdkp->device;
2447 sd_dif_config_host(sdkp);
2449 if (!sdkp->protection_type)
2452 if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2453 sd_first_printk(KERN_NOTICE, sdkp,
2454 "Disabling DIF Type %u protection\n",
2455 sdkp->protection_type);
2456 sdkp->protection_type = 0;
2459 sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2460 sdkp->protection_type);
2463 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2464 struct scsi_sense_hdr *sshdr, int sense_valid,
2468 sd_print_sense_hdr(sdkp, sshdr);
2470 sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2473 * Set dirty bit for removable devices if not ready -
2474 * sometimes drives will not report this properly.
2476 if (sdp->removable &&
2477 sense_valid && sshdr->sense_key == NOT_READY)
2478 set_media_not_present(sdkp);
2481 * We used to set media_present to 0 here to indicate no media
2482 * in the drive, but some drives fail read capacity even with
2483 * media present, so we can't do that.
2485 sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2489 #if RC16_LEN > SD_BUF_SIZE
2490 #error RC16_LEN must not be more than SD_BUF_SIZE
2493 #define READ_CAPACITY_RETRIES_ON_RESET 10
2495 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2496 unsigned char *buffer)
2498 unsigned char cmd[16];
2499 struct scsi_sense_hdr sshdr;
2500 const struct scsi_exec_args exec_args = {
2503 int sense_valid = 0;
2505 int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2506 unsigned int alignment;
2507 unsigned long long lba;
2508 unsigned sector_size;
2510 if (sdp->no_read_capacity_16)
2515 cmd[0] = SERVICE_ACTION_IN_16;
2516 cmd[1] = SAI_READ_CAPACITY_16;
2518 memset(buffer, 0, RC16_LEN);
2520 the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2521 buffer, RC16_LEN, SD_TIMEOUT,
2522 sdkp->max_retries, &exec_args);
2523 if (the_result > 0) {
2524 if (media_not_present(sdkp, &sshdr))
2527 sense_valid = scsi_sense_valid(&sshdr);
2529 sshdr.sense_key == ILLEGAL_REQUEST &&
2530 (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2532 /* Invalid Command Operation Code or
2533 * Invalid Field in CDB, just retry
2534 * silently with RC10 */
2537 sshdr.sense_key == UNIT_ATTENTION &&
2538 sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2539 /* Device reset might occur several times,
2540 * give it one more chance */
2541 if (--reset_retries > 0)
2546 } while (the_result && retries);
2549 sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2550 read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2554 sector_size = get_unaligned_be32(&buffer[8]);
2555 lba = get_unaligned_be64(&buffer[0]);
2557 if (sd_read_protection_type(sdkp, buffer) < 0) {
2562 /* Logical blocks per physical block exponent */
2563 sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2566 sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2568 /* Lowest aligned logical block */
2569 alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2570 blk_queue_alignment_offset(sdp->request_queue, alignment);
2571 if (alignment && sdkp->first_scan)
2572 sd_printk(KERN_NOTICE, sdkp,
2573 "physical block alignment offset: %u\n", alignment);
2575 if (buffer[14] & 0x80) { /* LBPME */
2578 if (buffer[14] & 0x40) /* LBPRZ */
2581 sd_config_discard(sdkp, SD_LBP_WS16);
2584 sdkp->capacity = lba + 1;
2588 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2589 unsigned char *buffer)
2591 static const u8 cmd[10] = { READ_CAPACITY };
2592 struct scsi_sense_hdr sshdr;
2593 struct scsi_failure failure_defs[] = {
2594 /* Do not retry Medium Not Present */
2596 .sense = UNIT_ATTENTION,
2598 .result = SAM_STAT_CHECK_CONDITION,
2603 .result = SAM_STAT_CHECK_CONDITION,
2605 /* Device reset might occur several times so retry a lot */
2607 .sense = UNIT_ATTENTION,
2609 .allowed = READ_CAPACITY_RETRIES_ON_RESET,
2610 .result = SAM_STAT_CHECK_CONDITION,
2612 /* Any other error not listed above retry 3 times */
2614 .result = SCMD_FAILURE_RESULT_ANY,
2619 struct scsi_failures failures = {
2620 .failure_definitions = failure_defs,
2622 const struct scsi_exec_args exec_args = {
2624 .failures = &failures,
2626 int sense_valid = 0;
2629 unsigned sector_size;
2631 memset(buffer, 0, 8);
2633 the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2634 8, SD_TIMEOUT, sdkp->max_retries,
2637 if (the_result > 0) {
2638 sense_valid = scsi_sense_valid(&sshdr);
2640 if (media_not_present(sdkp, &sshdr))
2645 sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2646 read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2650 sector_size = get_unaligned_be32(&buffer[4]);
2651 lba = get_unaligned_be32(&buffer[0]);
2653 if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2654 /* Some buggy (usb cardreader) devices return an lba of
2655 0xffffffff when the want to report a size of 0 (with
2656 which they really mean no media is present) */
2658 sdkp->physical_block_size = sector_size;
2662 sdkp->capacity = lba + 1;
2663 sdkp->physical_block_size = sector_size;
2667 static int sd_try_rc16_first(struct scsi_device *sdp)
2669 if (sdp->host->max_cmd_len < 16)
2671 if (sdp->try_rc_10_first)
2673 if (sdp->scsi_level > SCSI_SPC_2)
2675 if (scsi_device_protection(sdp))
2681 * read disk capacity
2684 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2687 struct scsi_device *sdp = sdkp->device;
2689 if (sd_try_rc16_first(sdp)) {
2690 sector_size = read_capacity_16(sdkp, sdp, buffer);
2691 if (sector_size == -EOVERFLOW)
2693 if (sector_size == -ENODEV)
2695 if (sector_size < 0)
2696 sector_size = read_capacity_10(sdkp, sdp, buffer);
2697 if (sector_size < 0)
2700 sector_size = read_capacity_10(sdkp, sdp, buffer);
2701 if (sector_size == -EOVERFLOW)
2703 if (sector_size < 0)
2705 if ((sizeof(sdkp->capacity) > 4) &&
2706 (sdkp->capacity > 0xffffffffULL)) {
2707 int old_sector_size = sector_size;
2708 sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2709 "Trying to use READ CAPACITY(16).\n");
2710 sector_size = read_capacity_16(sdkp, sdp, buffer);
2711 if (sector_size < 0) {
2712 sd_printk(KERN_NOTICE, sdkp,
2713 "Using 0xffffffff as device size\n");
2714 sdkp->capacity = 1 + (sector_t) 0xffffffff;
2715 sector_size = old_sector_size;
2718 /* Remember that READ CAPACITY(16) succeeded */
2719 sdp->try_rc_10_first = 0;
2723 /* Some devices are known to return the total number of blocks,
2724 * not the highest block number. Some devices have versions
2725 * which do this and others which do not. Some devices we might
2726 * suspect of doing this but we don't know for certain.
2728 * If we know the reported capacity is wrong, decrement it. If
2729 * we can only guess, then assume the number of blocks is even
2730 * (usually true but not always) and err on the side of lowering
2733 if (sdp->fix_capacity ||
2734 (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2735 sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2736 "from its reported value: %llu\n",
2737 (unsigned long long) sdkp->capacity);
2742 if (sector_size == 0) {
2744 sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2748 if (sector_size != 512 &&
2749 sector_size != 1024 &&
2750 sector_size != 2048 &&
2751 sector_size != 4096) {
2752 sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2755 * The user might want to re-format the drive with
2756 * a supported sectorsize. Once this happens, it
2757 * would be relatively trivial to set the thing up.
2758 * For this reason, we leave the thing in the table.
2762 * set a bogus sector size so the normal read/write
2763 * logic in the block layer will eventually refuse any
2764 * request on this device without tripping over power
2765 * of two sector size assumptions
2769 blk_queue_logical_block_size(sdp->request_queue, sector_size);
2770 blk_queue_physical_block_size(sdp->request_queue,
2771 sdkp->physical_block_size);
2772 sdkp->device->sector_size = sector_size;
2774 if (sdkp->capacity > 0xffffffff)
2775 sdp->use_16_for_rw = 1;
2780 * Print disk capacity
2783 sd_print_capacity(struct scsi_disk *sdkp,
2784 sector_t old_capacity)
2786 int sector_size = sdkp->device->sector_size;
2787 char cap_str_2[10], cap_str_10[10];
2789 if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2792 string_get_size(sdkp->capacity, sector_size,
2793 STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2794 string_get_size(sdkp->capacity, sector_size,
2795 STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2797 sd_printk(KERN_NOTICE, sdkp,
2798 "%llu %d-byte logical blocks: (%s/%s)\n",
2799 (unsigned long long)sdkp->capacity,
2800 sector_size, cap_str_10, cap_str_2);
2802 if (sdkp->physical_block_size != sector_size)
2803 sd_printk(KERN_NOTICE, sdkp,
2804 "%u-byte physical blocks\n",
2805 sdkp->physical_block_size);
2808 /* called with buffer of length 512 */
2810 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2811 unsigned char *buffer, int len, struct scsi_mode_data *data,
2812 struct scsi_sense_hdr *sshdr)
2815 * If we must use MODE SENSE(10), make sure that the buffer length
2816 * is at least 8 bytes so that the mode sense header fits.
2818 if (sdkp->device->use_10_for_ms && len < 8)
2821 return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2822 SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2826 * read write protect setting, if possible - called only in sd_revalidate_disk()
2827 * called with buffer of length SD_BUF_SIZE
2830 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2833 struct scsi_device *sdp = sdkp->device;
2834 struct scsi_mode_data data;
2835 int old_wp = sdkp->write_prot;
2837 set_disk_ro(sdkp->disk, 0);
2838 if (sdp->skip_ms_page_3f) {
2839 sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2843 if (sdp->use_192_bytes_for_3f) {
2844 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2847 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2848 * We have to start carefully: some devices hang if we ask
2849 * for more than is available.
2851 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2854 * Second attempt: ask for page 0 When only page 0 is
2855 * implemented, a request for page 3F may return Sense Key
2856 * 5: Illegal Request, Sense Code 24: Invalid field in
2860 res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2863 * Third attempt: ask 255 bytes, as we did earlier.
2866 res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2871 sd_first_printk(KERN_WARNING, sdkp,
2872 "Test WP failed, assume Write Enabled\n");
2874 sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2875 set_disk_ro(sdkp->disk, sdkp->write_prot);
2876 if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2877 sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2878 sdkp->write_prot ? "on" : "off");
2879 sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2885 * sd_read_cache_type - called only from sd_revalidate_disk()
2886 * called with buffer of length SD_BUF_SIZE
2889 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2892 struct scsi_device *sdp = sdkp->device;
2897 struct scsi_mode_data data;
2898 struct scsi_sense_hdr sshdr;
2899 int old_wce = sdkp->WCE;
2900 int old_rcd = sdkp->RCD;
2901 int old_dpofua = sdkp->DPOFUA;
2904 if (sdkp->cache_override)
2908 if (sdp->skip_ms_page_8) {
2909 if (sdp->type == TYPE_RBC)
2912 if (sdp->skip_ms_page_3f)
2915 if (sdp->use_192_bytes_for_3f)
2919 } else if (sdp->type == TYPE_RBC) {
2927 /* cautiously ask */
2928 res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2934 if (!data.header_length) {
2937 sd_first_printk(KERN_ERR, sdkp,
2938 "Missing header in MODE_SENSE response\n");
2941 /* that went OK, now ask for the proper length */
2945 * We're only interested in the first three bytes, actually.
2946 * But the data cache page is defined for the first 20.
2950 else if (len > SD_BUF_SIZE) {
2951 sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2952 "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2955 if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2959 if (len > first_len)
2960 res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2964 int offset = data.header_length + data.block_descriptor_length;
2966 while (offset < len) {
2967 u8 page_code = buffer[offset] & 0x3F;
2968 u8 spf = buffer[offset] & 0x40;
2970 if (page_code == 8 || page_code == 6) {
2971 /* We're interested only in the first 3 bytes.
2973 if (len - offset <= 2) {
2974 sd_first_printk(KERN_ERR, sdkp,
2975 "Incomplete mode parameter "
2979 modepage = page_code;
2983 /* Go to the next page */
2984 if (spf && len - offset > 3)
2985 offset += 4 + (buffer[offset+2] << 8) +
2987 else if (!spf && len - offset > 1)
2988 offset += 2 + buffer[offset+1];
2990 sd_first_printk(KERN_ERR, sdkp,
2992 "parameter data\n");
2998 sd_first_printk(KERN_WARNING, sdkp,
2999 "No Caching mode page found\n");
3003 if (modepage == 8) {
3004 sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
3005 sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
3007 sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
3011 sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
3012 if (sdp->broken_fua) {
3013 sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
3015 } else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
3016 !sdkp->device->use_16_for_rw) {
3017 sd_first_printk(KERN_NOTICE, sdkp,
3018 "Uses READ/WRITE(6), disabling FUA\n");
3022 /* No cache flush allowed for write protected devices */
3023 if (sdkp->WCE && sdkp->write_prot)
3026 if (sdkp->first_scan || old_wce != sdkp->WCE ||
3027 old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
3028 sd_printk(KERN_NOTICE, sdkp,
3029 "Write cache: %s, read cache: %s, %s\n",
3030 sdkp->WCE ? "enabled" : "disabled",
3031 sdkp->RCD ? "disabled" : "enabled",
3032 sdkp->DPOFUA ? "supports DPO and FUA"
3033 : "doesn't support DPO or FUA");
3039 if (res == -EIO && scsi_sense_valid(&sshdr) &&
3040 sshdr.sense_key == ILLEGAL_REQUEST &&
3041 sshdr.asc == 0x24 && sshdr.ascq == 0x0)
3042 /* Invalid field in CDB */
3043 sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
3045 sd_first_printk(KERN_ERR, sdkp,
3046 "Asking for cache data failed\n");
3049 if (sdp->wce_default_on) {
3050 sd_first_printk(KERN_NOTICE, sdkp,
3051 "Assuming drive cache: write back\n");
3054 sd_first_printk(KERN_WARNING, sdkp,
3055 "Assuming drive cache: write through\n");
3063 * The ATO bit indicates whether the DIF application tag is available
3064 * for use by the operating system.
3066 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
3069 struct scsi_device *sdp = sdkp->device;
3070 struct scsi_mode_data data;
3071 struct scsi_sense_hdr sshdr;
3073 if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
3076 if (sdkp->protection_type == 0)
3079 res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
3080 sdkp->max_retries, &data, &sshdr);
3082 if (res < 0 || !data.header_length ||
3084 sd_first_printk(KERN_WARNING, sdkp,
3085 "getting Control mode page failed, assume no ATO\n");
3087 if (res == -EIO && scsi_sense_valid(&sshdr))
3088 sd_print_sense_hdr(sdkp, &sshdr);
3093 offset = data.header_length + data.block_descriptor_length;
3095 if ((buffer[offset] & 0x3f) != 0x0a) {
3096 sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
3100 if ((buffer[offset + 5] & 0x80) == 0)
3109 * sd_read_block_limits - Query disk device for preferred I/O sizes.
3110 * @sdkp: disk to query
3112 static void sd_read_block_limits(struct scsi_disk *sdkp)
3114 struct scsi_vpd *vpd;
3118 vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3119 if (!vpd || vpd->len < 16)
3122 sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3123 sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3124 sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3126 if (vpd->len >= 64) {
3127 unsigned int lba_count, desc_count;
3129 sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3134 lba_count = get_unaligned_be32(&vpd->data[20]);
3135 desc_count = get_unaligned_be32(&vpd->data[24]);
3137 if (lba_count && desc_count)
3138 sdkp->max_unmap_blocks = lba_count;
3140 sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3142 if (vpd->data[32] & 0x80)
3143 sdkp->unmap_alignment =
3144 get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3146 if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
3148 if (sdkp->max_unmap_blocks)
3149 sd_config_discard(sdkp, SD_LBP_UNMAP);
3151 sd_config_discard(sdkp, SD_LBP_WS16);
3153 } else { /* LBP VPD page tells us what to use */
3154 if (sdkp->lbpu && sdkp->max_unmap_blocks)
3155 sd_config_discard(sdkp, SD_LBP_UNMAP);
3156 else if (sdkp->lbpws)
3157 sd_config_discard(sdkp, SD_LBP_WS16);
3158 else if (sdkp->lbpws10)
3159 sd_config_discard(sdkp, SD_LBP_WS10);
3161 sd_config_discard(sdkp, SD_LBP_DISABLE);
3170 * sd_read_block_characteristics - Query block dev. characteristics
3171 * @sdkp: disk to query
3173 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
3175 struct request_queue *q = sdkp->disk->queue;
3176 struct scsi_vpd *vpd;
3180 vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3182 if (!vpd || vpd->len < 8) {
3187 rot = get_unaligned_be16(&vpd->data[4]);
3188 sdkp->zoned = (vpd->data[8] >> 4) & 3;
3192 blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3193 blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3197 #ifdef CONFIG_BLK_DEV_ZONED /* sd_probe rejects ZBD devices early otherwise */
3198 if (sdkp->device->type == TYPE_ZBC) {
3202 disk_set_zoned(sdkp->disk);
3205 * Per ZBC and ZAC specifications, writes in sequential write
3206 * required zones of host-managed devices must be aligned to
3207 * the device physical block size.
3209 blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3212 * Host-aware devices are treated as conventional.
3214 WARN_ON_ONCE(blk_queue_is_zoned(q));
3216 #endif /* CONFIG_BLK_DEV_ZONED */
3218 if (!sdkp->first_scan)
3221 if (blk_queue_is_zoned(q))
3222 sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n");
3223 else if (sdkp->zoned == 1)
3224 sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n");
3225 else if (sdkp->zoned == 2)
3226 sd_printk(KERN_NOTICE, sdkp, "Drive-managed SMR disk\n");
3230 * sd_read_block_provisioning - Query provisioning VPD page
3231 * @sdkp: disk to query
3233 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3235 struct scsi_vpd *vpd;
3237 if (sdkp->lbpme == 0)
3241 vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3243 if (!vpd || vpd->len < 8) {
3249 sdkp->lbpu = (vpd->data[5] >> 7) & 1; /* UNMAP */
3250 sdkp->lbpws = (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3251 sdkp->lbpws10 = (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3255 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3257 struct scsi_device *sdev = sdkp->device;
3259 if (sdev->host->no_write_same) {
3260 sdev->no_write_same = 1;
3265 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3266 struct scsi_vpd *vpd;
3268 sdev->no_report_opcodes = 1;
3270 /* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3271 * CODES is unsupported and the device has an ATA
3272 * Information VPD page (SAT).
3275 vpd = rcu_dereference(sdev->vpd_pg89);
3277 sdev->no_write_same = 1;
3281 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3284 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3288 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3290 struct scsi_device *sdev = sdkp->device;
3292 if (!sdev->security_supported)
3295 if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3296 SECURITY_PROTOCOL_IN, 0) == 1 &&
3297 scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3298 SECURITY_PROTOCOL_OUT, 0) == 1)
3302 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3304 return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3308 * sd_read_cpr - Query concurrent positioning ranges
3309 * @sdkp: disk to query
3311 static void sd_read_cpr(struct scsi_disk *sdkp)
3313 struct blk_independent_access_ranges *iars = NULL;
3314 unsigned char *buffer = NULL;
3315 unsigned int nr_cpr = 0;
3316 int i, vpd_len, buf_len = SD_BUF_SIZE;
3320 * We need to have the capacity set first for the block layer to be
3321 * able to check the ranges.
3323 if (sdkp->first_scan)
3326 if (!sdkp->capacity)
3330 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3331 * leading to a maximum page size of 64 + 256*32 bytes.
3333 buf_len = 64 + 256*32;
3334 buffer = kmalloc(buf_len, GFP_KERNEL);
3335 if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3338 /* We must have at least a 64B header and one 32B range descriptor */
3339 vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3340 if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3341 sd_printk(KERN_ERR, sdkp,
3342 "Invalid Concurrent Positioning Ranges VPD page\n");
3346 nr_cpr = (vpd_len - 64) / 32;
3352 iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3359 for (i = 0; i < nr_cpr; i++, desc += 32) {
3361 sd_printk(KERN_ERR, sdkp,
3362 "Invalid Concurrent Positioning Range number\n");
3367 iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3368 iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3372 disk_set_independent_access_ranges(sdkp->disk, iars);
3373 if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3374 sd_printk(KERN_NOTICE, sdkp,
3375 "%u concurrent positioning ranges\n", nr_cpr);
3376 sdkp->nr_actuators = nr_cpr;
3382 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3384 struct scsi_device *sdp = sdkp->device;
3385 unsigned int min_xfer_bytes =
3386 logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3388 if (sdkp->min_xfer_blocks == 0)
3391 if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3392 sd_first_printk(KERN_WARNING, sdkp,
3393 "Preferred minimum I/O size %u bytes not a " \
3394 "multiple of physical block size (%u bytes)\n",
3395 min_xfer_bytes, sdkp->physical_block_size);
3396 sdkp->min_xfer_blocks = 0;
3400 sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3406 * Determine the device's preferred I/O size for reads and writes
3407 * unless the reported value is unreasonably small, large, not a
3408 * multiple of the physical block size, or simply garbage.
3410 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3411 unsigned int dev_max)
3413 struct scsi_device *sdp = sdkp->device;
3414 unsigned int opt_xfer_bytes =
3415 logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3416 unsigned int min_xfer_bytes =
3417 logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3419 if (sdkp->opt_xfer_blocks == 0)
3422 if (sdkp->opt_xfer_blocks > dev_max) {
3423 sd_first_printk(KERN_WARNING, sdkp,
3424 "Optimal transfer size %u logical blocks " \
3425 "> dev_max (%u logical blocks)\n",
3426 sdkp->opt_xfer_blocks, dev_max);
3430 if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3431 sd_first_printk(KERN_WARNING, sdkp,
3432 "Optimal transfer size %u logical blocks " \
3433 "> sd driver limit (%u logical blocks)\n",
3434 sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3438 if (opt_xfer_bytes < PAGE_SIZE) {
3439 sd_first_printk(KERN_WARNING, sdkp,
3440 "Optimal transfer size %u bytes < " \
3441 "PAGE_SIZE (%u bytes)\n",
3442 opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3446 if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3447 sd_first_printk(KERN_WARNING, sdkp,
3448 "Optimal transfer size %u bytes not a " \
3449 "multiple of preferred minimum block " \
3450 "size (%u bytes)\n",
3451 opt_xfer_bytes, min_xfer_bytes);
3455 if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3456 sd_first_printk(KERN_WARNING, sdkp,
3457 "Optimal transfer size %u bytes not a " \
3458 "multiple of physical block size (%u bytes)\n",
3459 opt_xfer_bytes, sdkp->physical_block_size);
3463 sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3468 static void sd_read_block_zero(struct scsi_disk *sdkp)
3470 unsigned int buf_len = sdkp->device->sector_size;
3471 char *buffer, cmd[10] = { };
3473 buffer = kmalloc(buf_len, GFP_KERNEL);
3478 put_unaligned_be32(0, &cmd[2]); /* Logical block address 0 */
3479 put_unaligned_be16(1, &cmd[7]); /* Transfer 1 logical block */
3481 scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN, buffer, buf_len,
3482 SD_TIMEOUT, sdkp->max_retries, NULL);
3487 * sd_revalidate_disk - called the first time a new disk is seen,
3488 * performs disk spin up, read_capacity, etc.
3489 * @disk: struct gendisk we care about
3491 static int sd_revalidate_disk(struct gendisk *disk)
3493 struct scsi_disk *sdkp = scsi_disk(disk);
3494 struct scsi_device *sdp = sdkp->device;
3495 struct request_queue *q = sdkp->disk->queue;
3496 sector_t old_capacity = sdkp->capacity;
3497 unsigned char *buffer;
3498 unsigned int dev_max, rw_max;
3500 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3501 "sd_revalidate_disk\n"));
3504 * If the device is offline, don't try and read capacity or any
3505 * of the other niceties.
3507 if (!scsi_device_online(sdp))
3510 buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3512 sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3513 "allocation failure.\n");
3517 sd_spinup_disk(sdkp);
3520 * Without media there is no reason to ask; moreover, some devices
3521 * react badly if we do.
3523 if (sdkp->media_present) {
3524 sd_read_capacity(sdkp, buffer);
3526 * Some USB/UAS devices return generic values for mode pages
3527 * until the media has been accessed. Trigger a READ operation
3528 * to force the device to populate mode pages.
3530 if (sdp->read_before_ms)
3531 sd_read_block_zero(sdkp);
3533 * set the default to rotational. All non-rotational devices
3534 * support the block characteristics VPD page, which will
3535 * cause this to be updated correctly and any device which
3536 * doesn't support it should be treated as rotational.
3538 blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3539 blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3541 if (scsi_device_supports_vpd(sdp)) {
3542 sd_read_block_provisioning(sdkp);
3543 sd_read_block_limits(sdkp);
3544 sd_read_block_characteristics(sdkp);
3545 sd_zbc_read_zones(sdkp, buffer);
3549 sd_print_capacity(sdkp, old_capacity);
3551 sd_read_write_protect_flag(sdkp, buffer);
3552 sd_read_cache_type(sdkp, buffer);
3553 sd_read_app_tag_own(sdkp, buffer);
3554 sd_read_write_same(sdkp, buffer);
3555 sd_read_security(sdkp, buffer);
3556 sd_config_protection(sdkp);
3560 * We now have all cache related info, determine how we deal
3561 * with flush requests.
3563 sd_set_flush_flag(sdkp);
3565 /* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3566 dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3568 /* Some devices report a maximum block count for READ/WRITE requests. */
3569 dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3570 q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3572 if (sd_validate_min_xfer_size(sdkp))
3573 blk_queue_io_min(sdkp->disk->queue,
3574 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3576 blk_queue_io_min(sdkp->disk->queue, 0);
3578 if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3579 q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3580 rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3582 q->limits.io_opt = 0;
3583 rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3584 (sector_t)BLK_DEF_MAX_SECTORS_CAP);
3588 * Limit default to SCSI host optimal sector limit if set. There may be
3589 * an impact on performance for when the size of a request exceeds this
3592 rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3594 /* Do not exceed controller limit */
3595 rw_max = min(rw_max, queue_max_hw_sectors(q));
3598 * Only update max_sectors if previously unset or if the current value
3599 * exceeds the capabilities of the hardware.
3601 if (sdkp->first_scan ||
3602 q->limits.max_sectors > q->limits.max_dev_sectors ||
3603 q->limits.max_sectors > q->limits.max_hw_sectors)
3604 q->limits.max_sectors = rw_max;
3606 sdkp->first_scan = 0;
3608 set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3609 sd_config_write_same(sdkp);
3613 * For a zoned drive, revalidating the zones can be done only once
3614 * the gendisk capacity is set. So if this fails, set back the gendisk
3617 if (sd_zbc_revalidate_zones(sdkp))
3618 set_capacity_and_notify(disk, 0);
3625 * sd_unlock_native_capacity - unlock native capacity
3626 * @disk: struct gendisk to set capacity for
3628 * Block layer calls this function if it detects that partitions
3629 * on @disk reach beyond the end of the device. If the SCSI host
3630 * implements ->unlock_native_capacity() method, it's invoked to
3631 * give it a chance to adjust the device capacity.
3634 * Defined by block layer. Might sleep.
3636 static void sd_unlock_native_capacity(struct gendisk *disk)
3638 struct scsi_device *sdev = scsi_disk(disk)->device;
3640 if (sdev->host->hostt->unlock_native_capacity)
3641 sdev->host->hostt->unlock_native_capacity(sdev);
3645 * sd_format_disk_name - format disk name
3646 * @prefix: name prefix - ie. "sd" for SCSI disks
3647 * @index: index of the disk to format name for
3648 * @buf: output buffer
3649 * @buflen: length of the output buffer
3651 * SCSI disk names starts at sda. The 26th device is sdz and the
3652 * 27th is sdaa. The last one for two lettered suffix is sdzz
3653 * which is followed by sdaaa.
3655 * This is basically 26 base counting with one extra 'nil' entry
3656 * at the beginning from the second digit on and can be
3657 * determined using similar method as 26 base conversion with the
3658 * index shifted -1 after each digit is computed.
3664 * 0 on success, -errno on failure.
3666 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3668 const int base = 'z' - 'a' + 1;
3669 char *begin = buf + strlen(prefix);
3670 char *end = buf + buflen;
3680 *--p = 'a' + (index % unit);
3681 index = (index / unit) - 1;
3682 } while (index >= 0);
3684 memmove(begin, p, end - p);
3685 memcpy(buf, prefix, strlen(prefix));
3691 * sd_probe - called during driver initialization and whenever a
3692 * new scsi device is attached to the system. It is called once
3693 * for each scsi device (not just disks) present.
3694 * @dev: pointer to device object
3696 * Returns 0 if successful (or not interested in this scsi device
3697 * (e.g. scanner)); 1 when there is an error.
3699 * Note: this function is invoked from the scsi mid-level.
3700 * This function sets up the mapping between a given
3701 * <host,channel,id,lun> (found in sdp) and new device name
3702 * (e.g. /dev/sda). More precisely it is the block device major
3703 * and minor number that is chosen here.
3705 * Assume sd_probe is not re-entrant (for time being)
3706 * Also think about sd_probe() and sd_remove() running coincidentally.
3708 static int sd_probe(struct device *dev)
3710 struct scsi_device *sdp = to_scsi_device(dev);
3711 struct scsi_disk *sdkp;
3716 scsi_autopm_get_device(sdp);
3718 if (sdp->type != TYPE_DISK &&
3719 sdp->type != TYPE_ZBC &&
3720 sdp->type != TYPE_MOD &&
3721 sdp->type != TYPE_RBC)
3724 if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3725 sdev_printk(KERN_WARNING, sdp,
3726 "Unsupported ZBC host-managed device.\n");
3730 SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3734 sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3738 gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3739 &sd_bio_compl_lkclass);
3743 index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3745 sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3749 error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3751 sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3752 goto out_free_index;
3757 sdkp->index = index;
3758 sdkp->max_retries = SD_MAX_RETRIES;
3759 atomic_set(&sdkp->openers, 0);
3760 atomic_set(&sdkp->device->ioerr_cnt, 0);
3762 if (!sdp->request_queue->rq_timeout) {
3763 if (sdp->type != TYPE_MOD)
3764 blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3766 blk_queue_rq_timeout(sdp->request_queue,
3770 device_initialize(&sdkp->disk_dev);
3771 sdkp->disk_dev.parent = get_device(dev);
3772 sdkp->disk_dev.class = &sd_disk_class;
3773 dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3775 error = device_add(&sdkp->disk_dev);
3777 put_device(&sdkp->disk_dev);
3781 dev_set_drvdata(dev, sdkp);
3783 gd->major = sd_major((index & 0xf0) >> 4);
3784 gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3785 gd->minors = SD_MINORS;
3787 gd->fops = &sd_fops;
3788 gd->private_data = sdkp;
3790 /* defaults, until the device tells us otherwise */
3791 sdp->sector_size = 512;
3793 sdkp->media_present = 1;
3794 sdkp->write_prot = 0;
3795 sdkp->cache_override = 0;
3799 sdkp->first_scan = 1;
3800 sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3802 sd_revalidate_disk(gd);
3804 if (sdp->removable) {
3805 gd->flags |= GENHD_FL_REMOVABLE;
3806 gd->events |= DISK_EVENT_MEDIA_CHANGE;
3807 gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3810 blk_pm_runtime_init(sdp->request_queue, dev);
3811 if (sdp->rpm_autosuspend) {
3812 pm_runtime_set_autosuspend_delay(dev,
3813 sdp->host->rpm_autosuspend_delay);
3816 error = device_add_disk(dev, gd, NULL);
3818 put_device(&sdkp->disk_dev);
3823 if (sdkp->security) {
3824 sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3826 sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3829 sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3830 sdp->removable ? "removable " : "");
3831 scsi_autopm_put_device(sdp);
3836 ida_free(&sd_index_ida, index);
3842 scsi_autopm_put_device(sdp);
3847 * sd_remove - called whenever a scsi disk (previously recognized by
3848 * sd_probe) is detached from the system. It is called (potentially
3849 * multiple times) during sd module unload.
3850 * @dev: pointer to device object
3852 * Note: this function is invoked from the scsi mid-level.
3853 * This function potentially frees up a device name (e.g. /dev/sdc)
3854 * that could be re-used by a subsequent sd_probe().
3855 * This function is not called when the built-in sd driver is "exit-ed".
3857 static int sd_remove(struct device *dev)
3859 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3861 scsi_autopm_get_device(sdkp->device);
3863 device_del(&sdkp->disk_dev);
3864 del_gendisk(sdkp->disk);
3865 if (!sdkp->suspended)
3868 put_disk(sdkp->disk);
3872 static void scsi_disk_release(struct device *dev)
3874 struct scsi_disk *sdkp = to_scsi_disk(dev);
3876 ida_free(&sd_index_ida, sdkp->index);
3877 sd_zbc_free_zone_info(sdkp);
3878 put_device(&sdkp->device->sdev_gendev);
3879 free_opal_dev(sdkp->opal_dev);
3884 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3886 unsigned char cmd[6] = { START_STOP }; /* START_VALID */
3887 struct scsi_sense_hdr sshdr;
3888 const struct scsi_exec_args exec_args = {
3890 .req_flags = BLK_MQ_REQ_PM,
3892 struct scsi_device *sdp = sdkp->device;
3896 cmd[4] |= 1; /* START */
3898 if (sdp->start_stop_pwr_cond)
3899 cmd[4] |= start ? 1 << 4 : 3 << 4; /* Active or Standby */
3901 if (!scsi_device_online(sdp))
3904 res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3905 sdkp->max_retries, &exec_args);
3907 sd_print_result(sdkp, "Start/Stop Unit failed", res);
3908 if (res > 0 && scsi_sense_valid(&sshdr)) {
3909 sd_print_sense_hdr(sdkp, &sshdr);
3910 /* 0x3a is medium not present */
3911 if (sshdr.asc == 0x3a)
3916 /* SCSI error codes must not go to the generic layer */
3924 * Send a SYNCHRONIZE CACHE instruction down to the device through
3925 * the normal SCSI command structure. Wait for the command to
3928 static void sd_shutdown(struct device *dev)
3930 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3933 return; /* this can happen */
3935 if (pm_runtime_suspended(dev))
3938 if (sdkp->WCE && sdkp->media_present) {
3939 sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3940 sd_sync_cache(sdkp);
3943 if ((system_state != SYSTEM_RESTART &&
3944 sdkp->device->manage_system_start_stop) ||
3945 (system_state == SYSTEM_POWER_OFF &&
3946 sdkp->device->manage_shutdown)) {
3947 sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3948 sd_start_stop_device(sdkp, 0);
3952 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
3954 return (sdev->manage_system_start_stop && !runtime) ||
3955 (sdev->manage_runtime_start_stop && runtime);
3958 static int sd_suspend_common(struct device *dev, bool runtime)
3960 struct scsi_disk *sdkp = dev_get_drvdata(dev);
3963 if (!sdkp) /* E.g.: runtime suspend following sd_remove() */
3966 if (sdkp->WCE && sdkp->media_present) {
3967 if (!sdkp->device->silence_suspend)
3968 sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3969 ret = sd_sync_cache(sdkp);
3970 /* ignore OFFLINE device */
3978 if (sd_do_start_stop(sdkp->device, runtime)) {
3979 if (!sdkp->device->silence_suspend)
3980 sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3981 /* an error is not worth aborting a system sleep */
3982 ret = sd_start_stop_device(sdkp, 0);
3988 sdkp->suspended = true;
3993 static int sd_suspend_system(struct device *dev)
3995 if (pm_runtime_suspended(dev))
3998 return sd_suspend_common(dev, false);
4001 static int sd_suspend_runtime(struct device *dev)
4003 return sd_suspend_common(dev, true);
4006 static int sd_resume(struct device *dev, bool runtime)
4008 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4011 if (!sdkp) /* E.g.: runtime resume at the start of sd_probe() */
4014 if (!sd_do_start_stop(sdkp->device, runtime)) {
4015 sdkp->suspended = false;
4019 sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4020 ret = sd_start_stop_device(sdkp, 1);
4022 opal_unlock_from_suspend(sdkp->opal_dev);
4023 sdkp->suspended = false;
4029 static int sd_resume_system(struct device *dev)
4031 if (pm_runtime_suspended(dev)) {
4032 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4033 struct scsi_device *sdp = sdkp ? sdkp->device : NULL;
4035 if (sdp && sdp->force_runtime_start_on_system_start)
4036 pm_request_resume(dev);
4041 return sd_resume(dev, false);
4044 static int sd_resume_runtime(struct device *dev)
4046 struct scsi_disk *sdkp = dev_get_drvdata(dev);
4047 struct scsi_device *sdp;
4049 if (!sdkp) /* E.g.: runtime resume at the start of sd_probe() */
4054 if (sdp->ignore_media_change) {
4055 /* clear the device's sense data */
4056 static const u8 cmd[10] = { REQUEST_SENSE };
4057 const struct scsi_exec_args exec_args = {
4058 .req_flags = BLK_MQ_REQ_PM,
4061 if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
4062 sdp->request_queue->rq_timeout, 1,
4064 sd_printk(KERN_NOTICE, sdkp,
4065 "Failed to clear sense data\n");
4068 return sd_resume(dev, true);
4071 static const struct dev_pm_ops sd_pm_ops = {
4072 .suspend = sd_suspend_system,
4073 .resume = sd_resume_system,
4074 .poweroff = sd_suspend_system,
4075 .restore = sd_resume_system,
4076 .runtime_suspend = sd_suspend_runtime,
4077 .runtime_resume = sd_resume_runtime,
4080 static struct scsi_driver sd_template = {
4083 .owner = THIS_MODULE,
4085 .probe_type = PROBE_PREFER_ASYNCHRONOUS,
4086 .remove = sd_remove,
4087 .shutdown = sd_shutdown,
4090 .rescan = sd_rescan,
4091 .init_command = sd_init_command,
4092 .uninit_command = sd_uninit_command,
4094 .eh_action = sd_eh_action,
4095 .eh_reset = sd_eh_reset,
4099 * init_sd - entry point for this driver (both when built in or when
4102 * Note: this function registers this driver with the scsi mid-level.
4104 static int __init init_sd(void)
4106 int majors = 0, i, err;
4108 SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
4110 for (i = 0; i < SD_MAJORS; i++) {
4111 if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
4119 err = class_register(&sd_disk_class);
4123 sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4124 if (!sd_page_pool) {
4125 printk(KERN_ERR "sd: can't init discard page pool\n");
4130 err = scsi_register_driver(&sd_template.gendrv);
4132 goto err_out_driver;
4137 mempool_destroy(sd_page_pool);
4139 class_unregister(&sd_disk_class);
4141 for (i = 0; i < SD_MAJORS; i++)
4142 unregister_blkdev(sd_major(i), "sd");
4147 * exit_sd - exit point for this driver (when it is a module).
4149 * Note: this function unregisters this driver from the scsi mid-level.
4151 static void __exit exit_sd(void)
4155 SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4157 scsi_unregister_driver(&sd_template.gendrv);
4158 mempool_destroy(sd_page_pool);
4160 class_unregister(&sd_disk_class);
4162 for (i = 0; i < SD_MAJORS; i++)
4163 unregister_blkdev(sd_major(i), "sd");
4166 module_init(init_sd);
4167 module_exit(exit_sd);
4169 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4171 scsi_print_sense_hdr(sdkp->device,
4172 sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4175 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4177 const char *hb_string = scsi_hostbyte_string(result);
4180 sd_printk(KERN_INFO, sdkp,
4181 "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4182 hb_string ? hb_string : "invalid",
4185 sd_printk(KERN_INFO, sdkp,
4186 "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4187 msg, host_byte(result), "DRIVER_OK");