2 * sd.c Copyright (C) 1992 Drew Eckhardt
3 * Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5 * Linux scsi disk driver
6 * Initial versions: Drew Eckhardt
7 * Subsequent revisions: Eric Youngdale
8 * Modification history:
9 * - Drew Eckhardt <drew@colorado.edu> original
10 * - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
11 * outstanding request, and other enhancements.
12 * Support loadable low-level scsi drivers.
13 * - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
14 * eight major numbers.
15 * - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
16 * - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
17 * sd_init and cleanups.
18 * - Alex Davis <letmein@erols.com> Fix problem where partition info
19 * not being read in sd_open. Fix problem where removable media
20 * could be ejected after sd_open.
21 * - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
22 * - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
23 * <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
24 * Support 32k/1M disks.
26 * Logging policy (needs CONFIG_SCSI_LOGGING defined):
27 * - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
28 * - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
29 * - entering sd_ioctl: SCSI_LOG_IOCTL level 1
30 * - entering other commands: SCSI_LOG_HLQUEUE level 3
31 * Note: when the logging level is set by the user, it must be greater
32 * than the level indicated above to trigger output.
35 #include <linux/module.h>
37 #include <linux/kernel.h>
39 #include <linux/bio.h>
40 #include <linux/genhd.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/delay.h>
49 #include <linux/mutex.h>
50 #include <linux/string_helpers.h>
51 #include <linux/async.h>
52 #include <linux/slab.h>
53 #include <linux/pm_runtime.h>
55 #include <linux/t10-pi.h>
56 #include <linux/uaccess.h>
57 #include <asm/unaligned.h>
59 #include <scsi/scsi.h>
60 #include <scsi/scsi_cmnd.h>
61 #include <scsi/scsi_dbg.h>
62 #include <scsi/scsi_device.h>
63 #include <scsi/scsi_driver.h>
64 #include <scsi/scsi_eh.h>
65 #include <scsi/scsi_host.h>
66 #include <scsi/scsi_ioctl.h>
67 #include <scsi/scsicam.h>
70 #include "scsi_priv.h"
71 #include "scsi_logging.h"
73 MODULE_AUTHOR("Eric Youngdale");
74 MODULE_DESCRIPTION("SCSI disk (sd) driver");
75 MODULE_LICENSE("GPL");
77 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR
);
78 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR
);
79 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR
);
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR
);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR
);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR
);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR
);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR
);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR
);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR
);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR
);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR
);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR
);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR
);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR
);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR
);
93 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK
);
94 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD
);
95 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC
);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC
);
98 #if !defined(CONFIG_DEBUG_BLOCK_EXT_DEVT)
104 static void sd_config_discard(struct scsi_disk
*, unsigned int);
105 static void sd_config_write_same(struct scsi_disk
*);
106 static int sd_revalidate_disk(struct gendisk
*);
107 static void sd_unlock_native_capacity(struct gendisk
*disk
);
108 static int sd_probe(struct device
*);
109 static int sd_remove(struct device
*);
110 static void sd_shutdown(struct device
*);
111 static int sd_suspend_system(struct device
*);
112 static int sd_suspend_runtime(struct device
*);
113 static int sd_resume(struct device
*);
114 static void sd_rescan(struct device
*);
115 static int sd_init_command(struct scsi_cmnd
*SCpnt
);
116 static void sd_uninit_command(struct scsi_cmnd
*SCpnt
);
117 static int sd_done(struct scsi_cmnd
*);
118 static void sd_eh_reset(struct scsi_cmnd
*);
119 static int sd_eh_action(struct scsi_cmnd
*, int);
120 static void sd_read_capacity(struct scsi_disk
*sdkp
, unsigned char *buffer
);
121 static void scsi_disk_release(struct device
*cdev
);
122 static void sd_print_sense_hdr(struct scsi_disk
*, struct scsi_sense_hdr
*);
123 static void sd_print_result(const struct scsi_disk
*, const char *, int);
125 static DEFINE_SPINLOCK(sd_index_lock
);
126 static DEFINE_IDA(sd_index_ida
);
128 /* This semaphore is used to mediate the 0->1 reference get in the
129 * face of object destruction (i.e. we can't allow a get on an
130 * object after last put) */
131 static DEFINE_MUTEX(sd_ref_mutex
);
133 static struct kmem_cache
*sd_cdb_cache
;
134 static mempool_t
*sd_cdb_pool
;
136 static const char *sd_cache_types
[] = {
137 "write through", "none", "write back",
138 "write back, no read (daft)"
141 static void sd_set_flush_flag(struct scsi_disk
*sdkp
)
143 bool wc
= false, fua
= false;
151 blk_queue_write_cache(sdkp
->disk
->queue
, wc
, fua
);
155 cache_type_store(struct device
*dev
, struct device_attribute
*attr
,
156 const char *buf
, size_t count
)
158 int i
, ct
= -1, rcd
, wce
, sp
;
159 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
160 struct scsi_device
*sdp
= sdkp
->device
;
163 struct scsi_mode_data data
;
164 struct scsi_sense_hdr sshdr
;
165 static const char temp
[] = "temporary ";
168 if (sdp
->type
!= TYPE_DISK
&& sdp
->type
!= TYPE_ZBC
)
169 /* no cache control on RBC devices; theoretically they
170 * can do it, but there's probably so many exceptions
171 * it's not worth the risk */
174 if (strncmp(buf
, temp
, sizeof(temp
) - 1) == 0) {
175 buf
+= sizeof(temp
) - 1;
176 sdkp
->cache_override
= 1;
178 sdkp
->cache_override
= 0;
181 for (i
= 0; i
< ARRAY_SIZE(sd_cache_types
); i
++) {
182 len
= strlen(sd_cache_types
[i
]);
183 if (strncmp(sd_cache_types
[i
], buf
, len
) == 0 &&
191 rcd
= ct
& 0x01 ? 1 : 0;
192 wce
= (ct
& 0x02) && !sdkp
->write_prot
? 1 : 0;
194 if (sdkp
->cache_override
) {
197 sd_set_flush_flag(sdkp
);
201 if (scsi_mode_sense(sdp
, 0x08, 8, buffer
, sizeof(buffer
), SD_TIMEOUT
,
202 SD_MAX_RETRIES
, &data
, NULL
))
204 len
= min_t(size_t, sizeof(buffer
), data
.length
- data
.header_length
-
205 data
.block_descriptor_length
);
206 buffer_data
= buffer
+ data
.header_length
+
207 data
.block_descriptor_length
;
208 buffer_data
[2] &= ~0x05;
209 buffer_data
[2] |= wce
<< 2 | rcd
;
210 sp
= buffer_data
[0] & 0x80 ? 1 : 0;
211 buffer_data
[0] &= ~0x80;
213 if (scsi_mode_select(sdp
, 1, sp
, 8, buffer_data
, len
, SD_TIMEOUT
,
214 SD_MAX_RETRIES
, &data
, &sshdr
)) {
215 if (scsi_sense_valid(&sshdr
))
216 sd_print_sense_hdr(sdkp
, &sshdr
);
219 revalidate_disk(sdkp
->disk
);
224 manage_start_stop_show(struct device
*dev
, struct device_attribute
*attr
,
227 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
228 struct scsi_device
*sdp
= sdkp
->device
;
230 return snprintf(buf
, 20, "%u\n", sdp
->manage_start_stop
);
234 manage_start_stop_store(struct device
*dev
, struct device_attribute
*attr
,
235 const char *buf
, size_t count
)
237 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
238 struct scsi_device
*sdp
= sdkp
->device
;
240 if (!capable(CAP_SYS_ADMIN
))
243 sdp
->manage_start_stop
= simple_strtoul(buf
, NULL
, 10);
247 static DEVICE_ATTR_RW(manage_start_stop
);
250 allow_restart_show(struct device
*dev
, struct device_attribute
*attr
, char *buf
)
252 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
254 return snprintf(buf
, 40, "%d\n", sdkp
->device
->allow_restart
);
258 allow_restart_store(struct device
*dev
, struct device_attribute
*attr
,
259 const char *buf
, size_t count
)
261 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
262 struct scsi_device
*sdp
= sdkp
->device
;
264 if (!capable(CAP_SYS_ADMIN
))
267 if (sdp
->type
!= TYPE_DISK
&& sdp
->type
!= TYPE_ZBC
)
270 sdp
->allow_restart
= simple_strtoul(buf
, NULL
, 10);
274 static DEVICE_ATTR_RW(allow_restart
);
277 cache_type_show(struct device
*dev
, struct device_attribute
*attr
, char *buf
)
279 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
280 int ct
= sdkp
->RCD
+ 2*sdkp
->WCE
;
282 return snprintf(buf
, 40, "%s\n", sd_cache_types
[ct
]);
284 static DEVICE_ATTR_RW(cache_type
);
287 FUA_show(struct device
*dev
, struct device_attribute
*attr
, char *buf
)
289 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
291 return snprintf(buf
, 20, "%u\n", sdkp
->DPOFUA
);
293 static DEVICE_ATTR_RO(FUA
);
296 protection_type_show(struct device
*dev
, struct device_attribute
*attr
,
299 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
301 return snprintf(buf
, 20, "%u\n", sdkp
->protection_type
);
305 protection_type_store(struct device
*dev
, struct device_attribute
*attr
,
306 const char *buf
, size_t count
)
308 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
312 if (!capable(CAP_SYS_ADMIN
))
315 err
= kstrtouint(buf
, 10, &val
);
320 if (val
>= 0 && val
<= T10_PI_TYPE3_PROTECTION
)
321 sdkp
->protection_type
= val
;
325 static DEVICE_ATTR_RW(protection_type
);
328 protection_mode_show(struct device
*dev
, struct device_attribute
*attr
,
331 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
332 struct scsi_device
*sdp
= sdkp
->device
;
333 unsigned int dif
, dix
;
335 dif
= scsi_host_dif_capable(sdp
->host
, sdkp
->protection_type
);
336 dix
= scsi_host_dix_capable(sdp
->host
, sdkp
->protection_type
);
338 if (!dix
&& scsi_host_dix_capable(sdp
->host
, T10_PI_TYPE0_PROTECTION
)) {
344 return snprintf(buf
, 20, "none\n");
346 return snprintf(buf
, 20, "%s%u\n", dix
? "dix" : "dif", dif
);
348 static DEVICE_ATTR_RO(protection_mode
);
351 app_tag_own_show(struct device
*dev
, struct device_attribute
*attr
, char *buf
)
353 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
355 return snprintf(buf
, 20, "%u\n", sdkp
->ATO
);
357 static DEVICE_ATTR_RO(app_tag_own
);
360 thin_provisioning_show(struct device
*dev
, struct device_attribute
*attr
,
363 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
365 return snprintf(buf
, 20, "%u\n", sdkp
->lbpme
);
367 static DEVICE_ATTR_RO(thin_provisioning
);
369 static const char *lbp_mode
[] = {
370 [SD_LBP_FULL
] = "full",
371 [SD_LBP_UNMAP
] = "unmap",
372 [SD_LBP_WS16
] = "writesame_16",
373 [SD_LBP_WS10
] = "writesame_10",
374 [SD_LBP_ZERO
] = "writesame_zero",
375 [SD_LBP_DISABLE
] = "disabled",
379 provisioning_mode_show(struct device
*dev
, struct device_attribute
*attr
,
382 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
384 return snprintf(buf
, 20, "%s\n", lbp_mode
[sdkp
->provisioning_mode
]);
388 provisioning_mode_store(struct device
*dev
, struct device_attribute
*attr
,
389 const char *buf
, size_t count
)
391 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
392 struct scsi_device
*sdp
= sdkp
->device
;
394 if (!capable(CAP_SYS_ADMIN
))
397 if (sd_is_zoned(sdkp
)) {
398 sd_config_discard(sdkp
, SD_LBP_DISABLE
);
402 if (sdp
->type
!= TYPE_DISK
)
405 if (!strncmp(buf
, lbp_mode
[SD_LBP_UNMAP
], 20))
406 sd_config_discard(sdkp
, SD_LBP_UNMAP
);
407 else if (!strncmp(buf
, lbp_mode
[SD_LBP_WS16
], 20))
408 sd_config_discard(sdkp
, SD_LBP_WS16
);
409 else if (!strncmp(buf
, lbp_mode
[SD_LBP_WS10
], 20))
410 sd_config_discard(sdkp
, SD_LBP_WS10
);
411 else if (!strncmp(buf
, lbp_mode
[SD_LBP_ZERO
], 20))
412 sd_config_discard(sdkp
, SD_LBP_ZERO
);
413 else if (!strncmp(buf
, lbp_mode
[SD_LBP_DISABLE
], 20))
414 sd_config_discard(sdkp
, SD_LBP_DISABLE
);
420 static DEVICE_ATTR_RW(provisioning_mode
);
423 max_medium_access_timeouts_show(struct device
*dev
,
424 struct device_attribute
*attr
, char *buf
)
426 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
428 return snprintf(buf
, 20, "%u\n", sdkp
->max_medium_access_timeouts
);
432 max_medium_access_timeouts_store(struct device
*dev
,
433 struct device_attribute
*attr
, const char *buf
,
436 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
439 if (!capable(CAP_SYS_ADMIN
))
442 err
= kstrtouint(buf
, 10, &sdkp
->max_medium_access_timeouts
);
444 return err
? err
: count
;
446 static DEVICE_ATTR_RW(max_medium_access_timeouts
);
449 max_write_same_blocks_show(struct device
*dev
, struct device_attribute
*attr
,
452 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
454 return snprintf(buf
, 20, "%u\n", sdkp
->max_ws_blocks
);
458 max_write_same_blocks_store(struct device
*dev
, struct device_attribute
*attr
,
459 const char *buf
, size_t count
)
461 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
462 struct scsi_device
*sdp
= sdkp
->device
;
466 if (!capable(CAP_SYS_ADMIN
))
469 if (sdp
->type
!= TYPE_DISK
&& sdp
->type
!= TYPE_ZBC
)
472 err
= kstrtoul(buf
, 10, &max
);
478 sdp
->no_write_same
= 1;
479 else if (max
<= SD_MAX_WS16_BLOCKS
) {
480 sdp
->no_write_same
= 0;
481 sdkp
->max_ws_blocks
= max
;
484 sd_config_write_same(sdkp
);
488 static DEVICE_ATTR_RW(max_write_same_blocks
);
490 static struct attribute
*sd_disk_attrs
[] = {
491 &dev_attr_cache_type
.attr
,
493 &dev_attr_allow_restart
.attr
,
494 &dev_attr_manage_start_stop
.attr
,
495 &dev_attr_protection_type
.attr
,
496 &dev_attr_protection_mode
.attr
,
497 &dev_attr_app_tag_own
.attr
,
498 &dev_attr_thin_provisioning
.attr
,
499 &dev_attr_provisioning_mode
.attr
,
500 &dev_attr_max_write_same_blocks
.attr
,
501 &dev_attr_max_medium_access_timeouts
.attr
,
504 ATTRIBUTE_GROUPS(sd_disk
);
506 static struct class sd_disk_class
= {
508 .owner
= THIS_MODULE
,
509 .dev_release
= scsi_disk_release
,
510 .dev_groups
= sd_disk_groups
,
513 static const struct dev_pm_ops sd_pm_ops
= {
514 .suspend
= sd_suspend_system
,
516 .poweroff
= sd_suspend_system
,
517 .restore
= sd_resume
,
518 .runtime_suspend
= sd_suspend_runtime
,
519 .runtime_resume
= sd_resume
,
522 static struct scsi_driver sd_template
= {
525 .owner
= THIS_MODULE
,
528 .shutdown
= sd_shutdown
,
532 .init_command
= sd_init_command
,
533 .uninit_command
= sd_uninit_command
,
535 .eh_action
= sd_eh_action
,
536 .eh_reset
= sd_eh_reset
,
540 * Dummy kobj_map->probe function.
541 * The default ->probe function will call modprobe, which is
542 * pointless as this module is already loaded.
544 static struct kobject
*sd_default_probe(dev_t devt
, int *partno
, void *data
)
550 * Device no to disk mapping:
552 * major disc2 disc p1
553 * |............|.............|....|....| <- dev_t
556 * Inside a major, we have 16k disks, however mapped non-
557 * contiguously. The first 16 disks are for major0, the next
558 * ones with major1, ... Disk 256 is for major0 again, disk 272
560 * As we stay compatible with our numbering scheme, we can reuse
561 * the well-know SCSI majors 8, 65--71, 136--143.
563 static int sd_major(int major_idx
)
567 return SCSI_DISK0_MAJOR
;
569 return SCSI_DISK1_MAJOR
+ major_idx
- 1;
571 return SCSI_DISK8_MAJOR
+ major_idx
- 8;
574 return 0; /* shut up gcc */
578 static struct scsi_disk
*scsi_disk_get(struct gendisk
*disk
)
580 struct scsi_disk
*sdkp
= NULL
;
582 mutex_lock(&sd_ref_mutex
);
584 if (disk
->private_data
) {
585 sdkp
= scsi_disk(disk
);
586 if (scsi_device_get(sdkp
->device
) == 0)
587 get_device(&sdkp
->dev
);
591 mutex_unlock(&sd_ref_mutex
);
595 static void scsi_disk_put(struct scsi_disk
*sdkp
)
597 struct scsi_device
*sdev
= sdkp
->device
;
599 mutex_lock(&sd_ref_mutex
);
600 put_device(&sdkp
->dev
);
601 scsi_device_put(sdev
);
602 mutex_unlock(&sd_ref_mutex
);
605 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd
*scmd
,
606 unsigned int dix
, unsigned int dif
)
608 struct bio
*bio
= scmd
->request
->bio
;
609 unsigned int prot_op
= sd_prot_op(rq_data_dir(scmd
->request
), dix
, dif
);
610 unsigned int protect
= 0;
612 if (dix
) { /* DIX Type 0, 1, 2, 3 */
613 if (bio_integrity_flagged(bio
, BIP_IP_CHECKSUM
))
614 scmd
->prot_flags
|= SCSI_PROT_IP_CHECKSUM
;
616 if (bio_integrity_flagged(bio
, BIP_CTRL_NOCHECK
) == false)
617 scmd
->prot_flags
|= SCSI_PROT_GUARD_CHECK
;
620 if (dif
!= T10_PI_TYPE3_PROTECTION
) { /* DIX/DIF Type 0, 1, 2 */
621 scmd
->prot_flags
|= SCSI_PROT_REF_INCREMENT
;
623 if (bio_integrity_flagged(bio
, BIP_CTRL_NOCHECK
) == false)
624 scmd
->prot_flags
|= SCSI_PROT_REF_CHECK
;
627 if (dif
) { /* DIX/DIF Type 1, 2, 3 */
628 scmd
->prot_flags
|= SCSI_PROT_TRANSFER_PI
;
630 if (bio_integrity_flagged(bio
, BIP_DISK_NOCHECK
))
631 protect
= 3 << 5; /* Disable target PI checking */
633 protect
= 1 << 5; /* Enable target PI checking */
636 scsi_set_prot_op(scmd
, prot_op
);
637 scsi_set_prot_type(scmd
, dif
);
638 scmd
->prot_flags
&= sd_prot_flag_mask(prot_op
);
643 static void sd_config_discard(struct scsi_disk
*sdkp
, unsigned int mode
)
645 struct request_queue
*q
= sdkp
->disk
->queue
;
646 unsigned int logical_block_size
= sdkp
->device
->sector_size
;
647 unsigned int max_blocks
= 0;
649 q
->limits
.discard_zeroes_data
= 0;
652 * When LBPRZ is reported, discard alignment and granularity
653 * must be fixed to the logical block size. Otherwise the block
654 * layer will drop misaligned portions of the request which can
655 * lead to data corruption. If LBPRZ is not set, we honor the
659 q
->limits
.discard_alignment
= 0;
660 q
->limits
.discard_granularity
= logical_block_size
;
662 q
->limits
.discard_alignment
= sdkp
->unmap_alignment
*
664 q
->limits
.discard_granularity
=
665 max(sdkp
->physical_block_size
,
666 sdkp
->unmap_granularity
* logical_block_size
);
669 sdkp
->provisioning_mode
= mode
;
674 blk_queue_max_discard_sectors(q
, 0);
675 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD
, q
);
679 max_blocks
= min_not_zero(sdkp
->max_unmap_blocks
,
680 (u32
)SD_MAX_WS16_BLOCKS
);
684 max_blocks
= min_not_zero(sdkp
->max_ws_blocks
,
685 (u32
)SD_MAX_WS16_BLOCKS
);
686 q
->limits
.discard_zeroes_data
= sdkp
->lbprz
;
690 max_blocks
= min_not_zero(sdkp
->max_ws_blocks
,
691 (u32
)SD_MAX_WS10_BLOCKS
);
692 q
->limits
.discard_zeroes_data
= sdkp
->lbprz
;
696 max_blocks
= min_not_zero(sdkp
->max_ws_blocks
,
697 (u32
)SD_MAX_WS10_BLOCKS
);
698 q
->limits
.discard_zeroes_data
= 1;
702 blk_queue_max_discard_sectors(q
, max_blocks
* (logical_block_size
>> 9));
703 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD
, q
);
707 * sd_setup_discard_cmnd - unmap blocks on thinly provisioned device
708 * @cmd: command to prepare
710 * Will set up either UNMAP, WRITE SAME(16) or WRITE SAME(10) depending
711 * on the provisioning mode of the target device.
713 static int sd_setup_discard_cmnd(struct scsi_cmnd
*cmd
)
715 struct request
*rq
= cmd
->request
;
716 struct scsi_device
*sdp
= cmd
->device
;
717 struct scsi_disk
*sdkp
= scsi_disk(rq
->rq_disk
);
718 sector_t sector
= blk_rq_pos(rq
);
719 unsigned int nr_sectors
= blk_rq_sectors(rq
);
725 sector
>>= ilog2(sdp
->sector_size
) - 9;
726 nr_sectors
>>= ilog2(sdp
->sector_size
) - 9;
728 page
= alloc_page(GFP_ATOMIC
| __GFP_ZERO
);
730 return BLKPREP_DEFER
;
732 switch (sdkp
->provisioning_mode
) {
734 buf
= page_address(page
);
737 cmd
->cmnd
[0] = UNMAP
;
740 put_unaligned_be16(6 + 16, &buf
[0]);
741 put_unaligned_be16(16, &buf
[2]);
742 put_unaligned_be64(sector
, &buf
[8]);
743 put_unaligned_be32(nr_sectors
, &buf
[16]);
750 cmd
->cmnd
[0] = WRITE_SAME_16
;
751 cmd
->cmnd
[1] = 0x8; /* UNMAP */
752 put_unaligned_be64(sector
, &cmd
->cmnd
[2]);
753 put_unaligned_be32(nr_sectors
, &cmd
->cmnd
[10]);
755 len
= sdkp
->device
->sector_size
;
761 cmd
->cmnd
[0] = WRITE_SAME
;
762 if (sdkp
->provisioning_mode
== SD_LBP_WS10
)
763 cmd
->cmnd
[1] = 0x8; /* UNMAP */
764 put_unaligned_be32(sector
, &cmd
->cmnd
[2]);
765 put_unaligned_be16(nr_sectors
, &cmd
->cmnd
[7]);
767 len
= sdkp
->device
->sector_size
;
771 ret
= BLKPREP_INVALID
;
775 rq
->timeout
= SD_TIMEOUT
;
777 cmd
->transfersize
= len
;
778 cmd
->allowed
= SD_MAX_RETRIES
;
780 rq
->special_vec
.bv_page
= page
;
781 rq
->special_vec
.bv_offset
= 0;
782 rq
->special_vec
.bv_len
= len
;
784 rq
->rq_flags
|= RQF_SPECIAL_PAYLOAD
;
785 scsi_req(rq
)->resid_len
= len
;
787 ret
= scsi_init_io(cmd
);
789 if (ret
!= BLKPREP_OK
)
794 static void sd_config_write_same(struct scsi_disk
*sdkp
)
796 struct request_queue
*q
= sdkp
->disk
->queue
;
797 unsigned int logical_block_size
= sdkp
->device
->sector_size
;
799 if (sdkp
->device
->no_write_same
) {
800 sdkp
->max_ws_blocks
= 0;
804 /* Some devices can not handle block counts above 0xffff despite
805 * supporting WRITE SAME(16). Consequently we default to 64k
806 * blocks per I/O unless the device explicitly advertises a
809 if (sdkp
->max_ws_blocks
> SD_MAX_WS10_BLOCKS
)
810 sdkp
->max_ws_blocks
= min_not_zero(sdkp
->max_ws_blocks
,
811 (u32
)SD_MAX_WS16_BLOCKS
);
812 else if (sdkp
->ws16
|| sdkp
->ws10
|| sdkp
->device
->no_report_opcodes
)
813 sdkp
->max_ws_blocks
= min_not_zero(sdkp
->max_ws_blocks
,
814 (u32
)SD_MAX_WS10_BLOCKS
);
816 sdkp
->device
->no_write_same
= 1;
817 sdkp
->max_ws_blocks
= 0;
821 blk_queue_max_write_same_sectors(q
, sdkp
->max_ws_blocks
*
822 (logical_block_size
>> 9));
826 * sd_setup_write_same_cmnd - write the same data to multiple blocks
827 * @cmd: command to prepare
829 * Will set up either WRITE SAME(10) or WRITE SAME(16) depending on
830 * the preference indicated by the target device.
832 static int sd_setup_write_same_cmnd(struct scsi_cmnd
*cmd
)
834 struct request
*rq
= cmd
->request
;
835 struct scsi_device
*sdp
= cmd
->device
;
836 struct scsi_disk
*sdkp
= scsi_disk(rq
->rq_disk
);
837 struct bio
*bio
= rq
->bio
;
838 sector_t sector
= blk_rq_pos(rq
);
839 unsigned int nr_sectors
= blk_rq_sectors(rq
);
840 unsigned int nr_bytes
= blk_rq_bytes(rq
);
843 if (sdkp
->device
->no_write_same
)
844 return BLKPREP_INVALID
;
846 BUG_ON(bio_offset(bio
) || bio_iovec(bio
).bv_len
!= sdp
->sector_size
);
848 if (sd_is_zoned(sdkp
)) {
849 ret
= sd_zbc_write_lock_zone(cmd
);
850 if (ret
!= BLKPREP_OK
)
854 sector
>>= ilog2(sdp
->sector_size
) - 9;
855 nr_sectors
>>= ilog2(sdp
->sector_size
) - 9;
857 rq
->timeout
= SD_WRITE_SAME_TIMEOUT
;
859 if (sdkp
->ws16
|| sector
> 0xffffffff || nr_sectors
> 0xffff) {
861 cmd
->cmnd
[0] = WRITE_SAME_16
;
862 put_unaligned_be64(sector
, &cmd
->cmnd
[2]);
863 put_unaligned_be32(nr_sectors
, &cmd
->cmnd
[10]);
866 cmd
->cmnd
[0] = WRITE_SAME
;
867 put_unaligned_be32(sector
, &cmd
->cmnd
[2]);
868 put_unaligned_be16(nr_sectors
, &cmd
->cmnd
[7]);
871 cmd
->transfersize
= sdp
->sector_size
;
872 cmd
->allowed
= SD_MAX_RETRIES
;
875 * For WRITE SAME the data transferred via the DATA OUT buffer is
876 * different from the amount of data actually written to the target.
878 * We set up __data_len to the amount of data transferred via the
879 * DATA OUT buffer so that blk_rq_map_sg sets up the proper S/G list
880 * to transfer a single sector of data first, but then reset it to
881 * the amount of data to be written right after so that the I/O path
882 * knows how much to actually write.
884 rq
->__data_len
= sdp
->sector_size
;
885 ret
= scsi_init_io(cmd
);
886 rq
->__data_len
= nr_bytes
;
890 static int sd_setup_flush_cmnd(struct scsi_cmnd
*cmd
)
892 struct request
*rq
= cmd
->request
;
894 /* flush requests don't perform I/O, zero the S/G table */
895 memset(&cmd
->sdb
, 0, sizeof(cmd
->sdb
));
897 cmd
->cmnd
[0] = SYNCHRONIZE_CACHE
;
899 cmd
->transfersize
= 0;
900 cmd
->allowed
= SD_MAX_RETRIES
;
902 rq
->timeout
= rq
->q
->rq_timeout
* SD_FLUSH_TIMEOUT_MULTIPLIER
;
906 static int sd_setup_read_write_cmnd(struct scsi_cmnd
*SCpnt
)
908 struct request
*rq
= SCpnt
->request
;
909 struct scsi_device
*sdp
= SCpnt
->device
;
910 struct gendisk
*disk
= rq
->rq_disk
;
911 struct scsi_disk
*sdkp
= scsi_disk(disk
);
912 sector_t block
= blk_rq_pos(rq
);
914 unsigned int this_count
= blk_rq_sectors(rq
);
915 unsigned int dif
, dix
;
916 bool zoned_write
= sd_is_zoned(sdkp
) && rq_data_dir(rq
) == WRITE
;
918 unsigned char protect
;
921 ret
= sd_zbc_write_lock_zone(SCpnt
);
922 if (ret
!= BLKPREP_OK
)
926 ret
= scsi_init_io(SCpnt
);
927 if (ret
!= BLKPREP_OK
)
931 /* from here on until we're complete, any goto out
932 * is used for a killable error condition */
936 scmd_printk(KERN_INFO
, SCpnt
,
937 "%s: block=%llu, count=%d\n",
938 __func__
, (unsigned long long)block
, this_count
));
940 if (!sdp
|| !scsi_device_online(sdp
) ||
941 block
+ blk_rq_sectors(rq
) > get_capacity(disk
)) {
942 SCSI_LOG_HLQUEUE(2, scmd_printk(KERN_INFO
, SCpnt
,
943 "Finishing %u sectors\n",
944 blk_rq_sectors(rq
)));
945 SCSI_LOG_HLQUEUE(2, scmd_printk(KERN_INFO
, SCpnt
,
946 "Retry with 0x%p\n", SCpnt
));
952 * quietly refuse to do anything to a changed disc until
953 * the changed bit has been reset
955 /* printk("SCSI disk has been changed or is not present. Prohibiting further I/O.\n"); */
960 * Some SD card readers can't handle multi-sector accesses which touch
961 * the last one or two hardware sectors. Split accesses as needed.
963 threshold
= get_capacity(disk
) - SD_LAST_BUGGY_SECTORS
*
964 (sdp
->sector_size
/ 512);
966 if (unlikely(sdp
->last_sector_bug
&& block
+ this_count
> threshold
)) {
967 if (block
< threshold
) {
968 /* Access up to the threshold but not beyond */
969 this_count
= threshold
- block
;
971 /* Access only a single hardware sector */
972 this_count
= sdp
->sector_size
/ 512;
976 SCSI_LOG_HLQUEUE(2, scmd_printk(KERN_INFO
, SCpnt
, "block=%llu\n",
977 (unsigned long long)block
));
980 * If we have a 1K hardware sectorsize, prevent access to single
981 * 512 byte sectors. In theory we could handle this - in fact
982 * the scsi cdrom driver must be able to handle this because
983 * we typically use 1K blocksizes, and cdroms typically have
984 * 2K hardware sectorsizes. Of course, things are simpler
985 * with the cdrom, since it is read-only. For performance
986 * reasons, the filesystems should be able to handle this
987 * and not force the scsi disk driver to use bounce buffers
990 if (sdp
->sector_size
== 1024) {
991 if ((block
& 1) || (blk_rq_sectors(rq
) & 1)) {
992 scmd_printk(KERN_ERR
, SCpnt
,
993 "Bad block number requested\n");
997 this_count
= this_count
>> 1;
1000 if (sdp
->sector_size
== 2048) {
1001 if ((block
& 3) || (blk_rq_sectors(rq
) & 3)) {
1002 scmd_printk(KERN_ERR
, SCpnt
,
1003 "Bad block number requested\n");
1007 this_count
= this_count
>> 2;
1010 if (sdp
->sector_size
== 4096) {
1011 if ((block
& 7) || (blk_rq_sectors(rq
) & 7)) {
1012 scmd_printk(KERN_ERR
, SCpnt
,
1013 "Bad block number requested\n");
1017 this_count
= this_count
>> 3;
1020 if (rq_data_dir(rq
) == WRITE
) {
1021 SCpnt
->cmnd
[0] = WRITE_6
;
1023 if (blk_integrity_rq(rq
))
1024 sd_dif_prepare(SCpnt
);
1026 } else if (rq_data_dir(rq
) == READ
) {
1027 SCpnt
->cmnd
[0] = READ_6
;
1029 scmd_printk(KERN_ERR
, SCpnt
, "Unknown command %d\n", req_op(rq
));
1033 SCSI_LOG_HLQUEUE(2, scmd_printk(KERN_INFO
, SCpnt
,
1034 "%s %d/%u 512 byte blocks.\n",
1035 (rq_data_dir(rq
) == WRITE
) ?
1036 "writing" : "reading", this_count
,
1037 blk_rq_sectors(rq
)));
1039 dix
= scsi_prot_sg_count(SCpnt
);
1040 dif
= scsi_host_dif_capable(SCpnt
->device
->host
, sdkp
->protection_type
);
1043 protect
= sd_setup_protect_cmnd(SCpnt
, dix
, dif
);
1047 if (protect
&& sdkp
->protection_type
== T10_PI_TYPE2_PROTECTION
) {
1048 SCpnt
->cmnd
= mempool_alloc(sd_cdb_pool
, GFP_ATOMIC
);
1050 if (unlikely(SCpnt
->cmnd
== NULL
)) {
1051 ret
= BLKPREP_DEFER
;
1055 SCpnt
->cmd_len
= SD_EXT_CDB_SIZE
;
1056 memset(SCpnt
->cmnd
, 0, SCpnt
->cmd_len
);
1057 SCpnt
->cmnd
[0] = VARIABLE_LENGTH_CMD
;
1058 SCpnt
->cmnd
[7] = 0x18;
1059 SCpnt
->cmnd
[9] = (rq_data_dir(rq
) == READ
) ? READ_32
: WRITE_32
;
1060 SCpnt
->cmnd
[10] = protect
| ((rq
->cmd_flags
& REQ_FUA
) ? 0x8 : 0);
1063 SCpnt
->cmnd
[12] = sizeof(block
) > 4 ? (unsigned char) (block
>> 56) & 0xff : 0;
1064 SCpnt
->cmnd
[13] = sizeof(block
) > 4 ? (unsigned char) (block
>> 48) & 0xff : 0;
1065 SCpnt
->cmnd
[14] = sizeof(block
) > 4 ? (unsigned char) (block
>> 40) & 0xff : 0;
1066 SCpnt
->cmnd
[15] = sizeof(block
) > 4 ? (unsigned char) (block
>> 32) & 0xff : 0;
1067 SCpnt
->cmnd
[16] = (unsigned char) (block
>> 24) & 0xff;
1068 SCpnt
->cmnd
[17] = (unsigned char) (block
>> 16) & 0xff;
1069 SCpnt
->cmnd
[18] = (unsigned char) (block
>> 8) & 0xff;
1070 SCpnt
->cmnd
[19] = (unsigned char) block
& 0xff;
1072 /* Expected Indirect LBA */
1073 SCpnt
->cmnd
[20] = (unsigned char) (block
>> 24) & 0xff;
1074 SCpnt
->cmnd
[21] = (unsigned char) (block
>> 16) & 0xff;
1075 SCpnt
->cmnd
[22] = (unsigned char) (block
>> 8) & 0xff;
1076 SCpnt
->cmnd
[23] = (unsigned char) block
& 0xff;
1078 /* Transfer length */
1079 SCpnt
->cmnd
[28] = (unsigned char) (this_count
>> 24) & 0xff;
1080 SCpnt
->cmnd
[29] = (unsigned char) (this_count
>> 16) & 0xff;
1081 SCpnt
->cmnd
[30] = (unsigned char) (this_count
>> 8) & 0xff;
1082 SCpnt
->cmnd
[31] = (unsigned char) this_count
& 0xff;
1083 } else if (sdp
->use_16_for_rw
|| (this_count
> 0xffff)) {
1084 SCpnt
->cmnd
[0] += READ_16
- READ_6
;
1085 SCpnt
->cmnd
[1] = protect
| ((rq
->cmd_flags
& REQ_FUA
) ? 0x8 : 0);
1086 SCpnt
->cmnd
[2] = sizeof(block
) > 4 ? (unsigned char) (block
>> 56) & 0xff : 0;
1087 SCpnt
->cmnd
[3] = sizeof(block
) > 4 ? (unsigned char) (block
>> 48) & 0xff : 0;
1088 SCpnt
->cmnd
[4] = sizeof(block
) > 4 ? (unsigned char) (block
>> 40) & 0xff : 0;
1089 SCpnt
->cmnd
[5] = sizeof(block
) > 4 ? (unsigned char) (block
>> 32) & 0xff : 0;
1090 SCpnt
->cmnd
[6] = (unsigned char) (block
>> 24) & 0xff;
1091 SCpnt
->cmnd
[7] = (unsigned char) (block
>> 16) & 0xff;
1092 SCpnt
->cmnd
[8] = (unsigned char) (block
>> 8) & 0xff;
1093 SCpnt
->cmnd
[9] = (unsigned char) block
& 0xff;
1094 SCpnt
->cmnd
[10] = (unsigned char) (this_count
>> 24) & 0xff;
1095 SCpnt
->cmnd
[11] = (unsigned char) (this_count
>> 16) & 0xff;
1096 SCpnt
->cmnd
[12] = (unsigned char) (this_count
>> 8) & 0xff;
1097 SCpnt
->cmnd
[13] = (unsigned char) this_count
& 0xff;
1098 SCpnt
->cmnd
[14] = SCpnt
->cmnd
[15] = 0;
1099 } else if ((this_count
> 0xff) || (block
> 0x1fffff) ||
1100 scsi_device_protection(SCpnt
->device
) ||
1101 SCpnt
->device
->use_10_for_rw
) {
1102 SCpnt
->cmnd
[0] += READ_10
- READ_6
;
1103 SCpnt
->cmnd
[1] = protect
| ((rq
->cmd_flags
& REQ_FUA
) ? 0x8 : 0);
1104 SCpnt
->cmnd
[2] = (unsigned char) (block
>> 24) & 0xff;
1105 SCpnt
->cmnd
[3] = (unsigned char) (block
>> 16) & 0xff;
1106 SCpnt
->cmnd
[4] = (unsigned char) (block
>> 8) & 0xff;
1107 SCpnt
->cmnd
[5] = (unsigned char) block
& 0xff;
1108 SCpnt
->cmnd
[6] = SCpnt
->cmnd
[9] = 0;
1109 SCpnt
->cmnd
[7] = (unsigned char) (this_count
>> 8) & 0xff;
1110 SCpnt
->cmnd
[8] = (unsigned char) this_count
& 0xff;
1112 if (unlikely(rq
->cmd_flags
& REQ_FUA
)) {
1114 * This happens only if this drive failed
1115 * 10byte rw command with ILLEGAL_REQUEST
1116 * during operation and thus turned off
1119 scmd_printk(KERN_ERR
, SCpnt
,
1120 "FUA write on READ/WRITE(6) drive\n");
1124 SCpnt
->cmnd
[1] |= (unsigned char) ((block
>> 16) & 0x1f);
1125 SCpnt
->cmnd
[2] = (unsigned char) ((block
>> 8) & 0xff);
1126 SCpnt
->cmnd
[3] = (unsigned char) block
& 0xff;
1127 SCpnt
->cmnd
[4] = (unsigned char) this_count
;
1130 SCpnt
->sdb
.length
= this_count
* sdp
->sector_size
;
1133 * We shouldn't disconnect in the middle of a sector, so with a dumb
1134 * host adapter, it's safe to assume that we can at least transfer
1135 * this many bytes between each connect / disconnect.
1137 SCpnt
->transfersize
= sdp
->sector_size
;
1138 SCpnt
->underflow
= this_count
<< 9;
1139 SCpnt
->allowed
= SD_MAX_RETRIES
;
1142 * This indicates that the command is ready from our end to be
1147 if (zoned_write
&& ret
!= BLKPREP_OK
)
1148 sd_zbc_write_unlock_zone(SCpnt
);
1153 static int sd_init_command(struct scsi_cmnd
*cmd
)
1155 struct request
*rq
= cmd
->request
;
1157 switch (req_op(rq
)) {
1158 case REQ_OP_DISCARD
:
1159 return sd_setup_discard_cmnd(cmd
);
1160 case REQ_OP_WRITE_SAME
:
1161 return sd_setup_write_same_cmnd(cmd
);
1163 return sd_setup_flush_cmnd(cmd
);
1166 return sd_setup_read_write_cmnd(cmd
);
1167 case REQ_OP_ZONE_REPORT
:
1168 return sd_zbc_setup_report_cmnd(cmd
);
1169 case REQ_OP_ZONE_RESET
:
1170 return sd_zbc_setup_reset_cmnd(cmd
);
1176 static void sd_uninit_command(struct scsi_cmnd
*SCpnt
)
1178 struct request
*rq
= SCpnt
->request
;
1180 if (rq
->rq_flags
& RQF_SPECIAL_PAYLOAD
)
1181 __free_page(rq
->special_vec
.bv_page
);
1183 if (SCpnt
->cmnd
!= scsi_req(rq
)->cmd
) {
1184 mempool_free(SCpnt
->cmnd
, sd_cdb_pool
);
1191 * sd_open - open a scsi disk device
1192 * @bdev: Block device of the scsi disk to open
1193 * @mode: FMODE_* mask
1195 * Returns 0 if successful. Returns a negated errno value in case
1198 * Note: This can be called from a user context (e.g. fsck(1) )
1199 * or from within the kernel (e.g. as a result of a mount(1) ).
1200 * In the latter case @inode and @filp carry an abridged amount
1201 * of information as noted above.
1203 * Locking: called with bdev->bd_mutex held.
1205 static int sd_open(struct block_device
*bdev
, fmode_t mode
)
1207 struct scsi_disk
*sdkp
= scsi_disk_get(bdev
->bd_disk
);
1208 struct scsi_device
*sdev
;
1214 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO
, sdkp
, "sd_open\n"));
1216 sdev
= sdkp
->device
;
1219 * If the device is in error recovery, wait until it is done.
1220 * If the device is offline, then disallow any access to it.
1223 if (!scsi_block_when_processing_errors(sdev
))
1226 if (sdev
->removable
|| sdkp
->write_prot
)
1227 check_disk_change(bdev
);
1230 * If the drive is empty, just let the open fail.
1232 retval
= -ENOMEDIUM
;
1233 if (sdev
->removable
&& !sdkp
->media_present
&& !(mode
& FMODE_NDELAY
))
1237 * If the device has the write protect tab set, have the open fail
1238 * if the user expects to be able to write to the thing.
1241 if (sdkp
->write_prot
&& (mode
& FMODE_WRITE
))
1245 * It is possible that the disk changing stuff resulted in
1246 * the device being taken offline. If this is the case,
1247 * report this to the user, and don't pretend that the
1248 * open actually succeeded.
1251 if (!scsi_device_online(sdev
))
1254 if ((atomic_inc_return(&sdkp
->openers
) == 1) && sdev
->removable
) {
1255 if (scsi_block_when_processing_errors(sdev
))
1256 scsi_set_medium_removal(sdev
, SCSI_REMOVAL_PREVENT
);
1262 scsi_disk_put(sdkp
);
1267 * sd_release - invoked when the (last) close(2) is called on this
1269 * @disk: disk to release
1270 * @mode: FMODE_* mask
1274 * Note: may block (uninterruptible) if error recovery is underway
1277 * Locking: called with bdev->bd_mutex held.
1279 static void sd_release(struct gendisk
*disk
, fmode_t mode
)
1281 struct scsi_disk
*sdkp
= scsi_disk(disk
);
1282 struct scsi_device
*sdev
= sdkp
->device
;
1284 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO
, sdkp
, "sd_release\n"));
1286 if (atomic_dec_return(&sdkp
->openers
) == 0 && sdev
->removable
) {
1287 if (scsi_block_when_processing_errors(sdev
))
1288 scsi_set_medium_removal(sdev
, SCSI_REMOVAL_ALLOW
);
1292 * XXX and what if there are packets in flight and this close()
1293 * XXX is followed by a "rmmod sd_mod"?
1296 scsi_disk_put(sdkp
);
1299 static int sd_getgeo(struct block_device
*bdev
, struct hd_geometry
*geo
)
1301 struct scsi_disk
*sdkp
= scsi_disk(bdev
->bd_disk
);
1302 struct scsi_device
*sdp
= sdkp
->device
;
1303 struct Scsi_Host
*host
= sdp
->host
;
1304 sector_t capacity
= logical_to_sectors(sdp
, sdkp
->capacity
);
1307 /* default to most commonly used values */
1308 diskinfo
[0] = 0x40; /* 1 << 6 */
1309 diskinfo
[1] = 0x20; /* 1 << 5 */
1310 diskinfo
[2] = capacity
>> 11;
1312 /* override with calculated, extended default, or driver values */
1313 if (host
->hostt
->bios_param
)
1314 host
->hostt
->bios_param(sdp
, bdev
, capacity
, diskinfo
);
1316 scsicam_bios_param(bdev
, capacity
, diskinfo
);
1318 geo
->heads
= diskinfo
[0];
1319 geo
->sectors
= diskinfo
[1];
1320 geo
->cylinders
= diskinfo
[2];
1325 * sd_ioctl - process an ioctl
1326 * @bdev: target block device
1327 * @mode: FMODE_* mask
1328 * @cmd: ioctl command number
1329 * @arg: this is third argument given to ioctl(2) system call.
1330 * Often contains a pointer.
1332 * Returns 0 if successful (some ioctls return positive numbers on
1333 * success as well). Returns a negated errno value in case of error.
1335 * Note: most ioctls are forward onto the block subsystem or further
1336 * down in the scsi subsystem.
1338 static int sd_ioctl(struct block_device
*bdev
, fmode_t mode
,
1339 unsigned int cmd
, unsigned long arg
)
1341 struct gendisk
*disk
= bdev
->bd_disk
;
1342 struct scsi_disk
*sdkp
= scsi_disk(disk
);
1343 struct scsi_device
*sdp
= sdkp
->device
;
1344 void __user
*p
= (void __user
*)arg
;
1347 SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO
, sdkp
, "sd_ioctl: disk=%s, "
1348 "cmd=0x%x\n", disk
->disk_name
, cmd
));
1350 error
= scsi_verify_blk_ioctl(bdev
, cmd
);
1355 * If we are in the middle of error recovery, don't let anyone
1356 * else try and use this device. Also, if error recovery fails, it
1357 * may try and take the device offline, in which case all further
1358 * access to the device is prohibited.
1360 error
= scsi_ioctl_block_when_processing_errors(sdp
, cmd
,
1361 (mode
& FMODE_NDELAY
) != 0);
1366 * Send SCSI addressing ioctls directly to mid level, send other
1367 * ioctls to block level and then onto mid level if they can't be
1371 case SCSI_IOCTL_GET_IDLUN
:
1372 case SCSI_IOCTL_GET_BUS_NUMBER
:
1373 error
= scsi_ioctl(sdp
, cmd
, p
);
1376 error
= scsi_cmd_blk_ioctl(bdev
, mode
, cmd
, p
);
1377 if (error
!= -ENOTTY
)
1379 error
= scsi_ioctl(sdp
, cmd
, p
);
1386 static void set_media_not_present(struct scsi_disk
*sdkp
)
1388 if (sdkp
->media_present
)
1389 sdkp
->device
->changed
= 1;
1391 if (sdkp
->device
->removable
) {
1392 sdkp
->media_present
= 0;
1397 static int media_not_present(struct scsi_disk
*sdkp
,
1398 struct scsi_sense_hdr
*sshdr
)
1400 if (!scsi_sense_valid(sshdr
))
1403 /* not invoked for commands that could return deferred errors */
1404 switch (sshdr
->sense_key
) {
1405 case UNIT_ATTENTION
:
1407 /* medium not present */
1408 if (sshdr
->asc
== 0x3A) {
1409 set_media_not_present(sdkp
);
1417 * sd_check_events - check media events
1418 * @disk: kernel device descriptor
1419 * @clearing: disk events currently being cleared
1421 * Returns mask of DISK_EVENT_*.
1423 * Note: this function is invoked from the block subsystem.
1425 static unsigned int sd_check_events(struct gendisk
*disk
, unsigned int clearing
)
1427 struct scsi_disk
*sdkp
= scsi_disk_get(disk
);
1428 struct scsi_device
*sdp
;
1435 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO
, sdkp
, "sd_check_events\n"));
1438 * If the device is offline, don't send any commands - just pretend as
1439 * if the command failed. If the device ever comes back online, we
1440 * can deal with it then. It is only because of unrecoverable errors
1441 * that we would ever take a device offline in the first place.
1443 if (!scsi_device_online(sdp
)) {
1444 set_media_not_present(sdkp
);
1449 * Using TEST_UNIT_READY enables differentiation between drive with
1450 * no cartridge loaded - NOT READY, drive with changed cartridge -
1451 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1453 * Drives that auto spin down. eg iomega jaz 1G, will be started
1454 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1455 * sd_revalidate() is called.
1457 if (scsi_block_when_processing_errors(sdp
)) {
1458 struct scsi_sense_hdr sshdr
= { 0, };
1460 retval
= scsi_test_unit_ready(sdp
, SD_TIMEOUT
, SD_MAX_RETRIES
,
1463 /* failed to execute TUR, assume media not present */
1464 if (host_byte(retval
)) {
1465 set_media_not_present(sdkp
);
1469 if (media_not_present(sdkp
, &sshdr
))
1474 * For removable scsi disk we have to recognise the presence
1475 * of a disk in the drive.
1477 if (!sdkp
->media_present
)
1479 sdkp
->media_present
= 1;
1482 * sdp->changed is set under the following conditions:
1484 * Medium present state has changed in either direction.
1485 * Device has indicated UNIT_ATTENTION.
1487 retval
= sdp
->changed
? DISK_EVENT_MEDIA_CHANGE
: 0;
1489 scsi_disk_put(sdkp
);
1493 static int sd_sync_cache(struct scsi_disk
*sdkp
)
1496 struct scsi_device
*sdp
= sdkp
->device
;
1497 const int timeout
= sdp
->request_queue
->rq_timeout
1498 * SD_FLUSH_TIMEOUT_MULTIPLIER
;
1499 struct scsi_sense_hdr sshdr
;
1501 if (!scsi_device_online(sdp
))
1504 for (retries
= 3; retries
> 0; --retries
) {
1505 unsigned char cmd
[10] = { 0 };
1507 cmd
[0] = SYNCHRONIZE_CACHE
;
1509 * Leave the rest of the command zero to indicate
1512 res
= scsi_execute(sdp
, cmd
, DMA_NONE
, NULL
, 0, NULL
, &sshdr
,
1513 timeout
, SD_MAX_RETRIES
, 0, RQF_PM
, NULL
);
1519 sd_print_result(sdkp
, "Synchronize Cache(10) failed", res
);
1521 if (driver_byte(res
) & DRIVER_SENSE
)
1522 sd_print_sense_hdr(sdkp
, &sshdr
);
1523 /* we need to evaluate the error return */
1524 if (scsi_sense_valid(&sshdr
) &&
1525 (sshdr
.asc
== 0x3a || /* medium not present */
1526 sshdr
.asc
== 0x20)) /* invalid command */
1527 /* this is no error here */
1530 switch (host_byte(res
)) {
1531 /* ignore errors due to racing a disconnection */
1532 case DID_BAD_TARGET
:
1533 case DID_NO_CONNECT
:
1535 /* signal the upper layer it might try again */
1539 case DID_SOFT_ERROR
:
1548 static void sd_rescan(struct device
*dev
)
1550 struct scsi_disk
*sdkp
= dev_get_drvdata(dev
);
1552 revalidate_disk(sdkp
->disk
);
1556 #ifdef CONFIG_COMPAT
1558 * This gets directly called from VFS. When the ioctl
1559 * is not recognized we go back to the other translation paths.
1561 static int sd_compat_ioctl(struct block_device
*bdev
, fmode_t mode
,
1562 unsigned int cmd
, unsigned long arg
)
1564 struct scsi_device
*sdev
= scsi_disk(bdev
->bd_disk
)->device
;
1567 error
= scsi_ioctl_block_when_processing_errors(sdev
, cmd
,
1568 (mode
& FMODE_NDELAY
) != 0);
1573 * Let the static ioctl translation table take care of it.
1575 if (!sdev
->host
->hostt
->compat_ioctl
)
1576 return -ENOIOCTLCMD
;
1577 return sdev
->host
->hostt
->compat_ioctl(sdev
, cmd
, (void __user
*)arg
);
1581 static char sd_pr_type(enum pr_type type
)
1584 case PR_WRITE_EXCLUSIVE
:
1586 case PR_EXCLUSIVE_ACCESS
:
1588 case PR_WRITE_EXCLUSIVE_REG_ONLY
:
1590 case PR_EXCLUSIVE_ACCESS_REG_ONLY
:
1592 case PR_WRITE_EXCLUSIVE_ALL_REGS
:
1594 case PR_EXCLUSIVE_ACCESS_ALL_REGS
:
1601 static int sd_pr_command(struct block_device
*bdev
, u8 sa
,
1602 u64 key
, u64 sa_key
, u8 type
, u8 flags
)
1604 struct scsi_device
*sdev
= scsi_disk(bdev
->bd_disk
)->device
;
1605 struct scsi_sense_hdr sshdr
;
1607 u8 cmd
[16] = { 0, };
1608 u8 data
[24] = { 0, };
1610 cmd
[0] = PERSISTENT_RESERVE_OUT
;
1613 put_unaligned_be32(sizeof(data
), &cmd
[5]);
1615 put_unaligned_be64(key
, &data
[0]);
1616 put_unaligned_be64(sa_key
, &data
[8]);
1619 result
= scsi_execute_req(sdev
, cmd
, DMA_TO_DEVICE
, &data
, sizeof(data
),
1620 &sshdr
, SD_TIMEOUT
, SD_MAX_RETRIES
, NULL
);
1622 if ((driver_byte(result
) & DRIVER_SENSE
) &&
1623 (scsi_sense_valid(&sshdr
))) {
1624 sdev_printk(KERN_INFO
, sdev
, "PR command failed: %d\n", result
);
1625 scsi_print_sense_hdr(sdev
, NULL
, &sshdr
);
1631 static int sd_pr_register(struct block_device
*bdev
, u64 old_key
, u64 new_key
,
1634 if (flags
& ~PR_FL_IGNORE_KEY
)
1636 return sd_pr_command(bdev
, (flags
& PR_FL_IGNORE_KEY
) ? 0x06 : 0x00,
1637 old_key
, new_key
, 0,
1638 (1 << 0) /* APTPL */);
1641 static int sd_pr_reserve(struct block_device
*bdev
, u64 key
, enum pr_type type
,
1646 return sd_pr_command(bdev
, 0x01, key
, 0, sd_pr_type(type
), 0);
1649 static int sd_pr_release(struct block_device
*bdev
, u64 key
, enum pr_type type
)
1651 return sd_pr_command(bdev
, 0x02, key
, 0, sd_pr_type(type
), 0);
1654 static int sd_pr_preempt(struct block_device
*bdev
, u64 old_key
, u64 new_key
,
1655 enum pr_type type
, bool abort
)
1657 return sd_pr_command(bdev
, abort
? 0x05 : 0x04, old_key
, new_key
,
1658 sd_pr_type(type
), 0);
1661 static int sd_pr_clear(struct block_device
*bdev
, u64 key
)
1663 return sd_pr_command(bdev
, 0x03, key
, 0, 0, 0);
1666 static const struct pr_ops sd_pr_ops
= {
1667 .pr_register
= sd_pr_register
,
1668 .pr_reserve
= sd_pr_reserve
,
1669 .pr_release
= sd_pr_release
,
1670 .pr_preempt
= sd_pr_preempt
,
1671 .pr_clear
= sd_pr_clear
,
1674 static const struct block_device_operations sd_fops
= {
1675 .owner
= THIS_MODULE
,
1677 .release
= sd_release
,
1679 .getgeo
= sd_getgeo
,
1680 #ifdef CONFIG_COMPAT
1681 .compat_ioctl
= sd_compat_ioctl
,
1683 .check_events
= sd_check_events
,
1684 .revalidate_disk
= sd_revalidate_disk
,
1685 .unlock_native_capacity
= sd_unlock_native_capacity
,
1686 .pr_ops
= &sd_pr_ops
,
1690 * sd_eh_reset - reset error handling callback
1691 * @scmd: sd-issued command that has failed
1693 * This function is called by the SCSI midlayer before starting
1694 * SCSI EH. When counting medium access failures we have to be
1695 * careful to register it only only once per device and SCSI EH run;
1696 * there might be several timed out commands which will cause the
1697 * 'max_medium_access_timeouts' counter to trigger after the first
1698 * SCSI EH run already and set the device to offline.
1699 * So this function resets the internal counter before starting SCSI EH.
1701 static void sd_eh_reset(struct scsi_cmnd
*scmd
)
1703 struct scsi_disk
*sdkp
= scsi_disk(scmd
->request
->rq_disk
);
1705 /* New SCSI EH run, reset gate variable */
1706 sdkp
->ignore_medium_access_errors
= false;
1710 * sd_eh_action - error handling callback
1711 * @scmd: sd-issued command that has failed
1712 * @eh_disp: The recovery disposition suggested by the midlayer
1714 * This function is called by the SCSI midlayer upon completion of an
1715 * error test command (currently TEST UNIT READY). The result of sending
1716 * the eh command is passed in eh_disp. We're looking for devices that
1717 * fail medium access commands but are OK with non access commands like
1718 * test unit ready (so wrongly see the device as having a successful
1721 static int sd_eh_action(struct scsi_cmnd
*scmd
, int eh_disp
)
1723 struct scsi_disk
*sdkp
= scsi_disk(scmd
->request
->rq_disk
);
1725 if (!scsi_device_online(scmd
->device
) ||
1726 !scsi_medium_access_command(scmd
) ||
1727 host_byte(scmd
->result
) != DID_TIME_OUT
||
1732 * The device has timed out executing a medium access command.
1733 * However, the TEST UNIT READY command sent during error
1734 * handling completed successfully. Either the device is in the
1735 * process of recovering or has it suffered an internal failure
1736 * that prevents access to the storage medium.
1738 if (!sdkp
->ignore_medium_access_errors
) {
1739 sdkp
->medium_access_timed_out
++;
1740 sdkp
->ignore_medium_access_errors
= true;
1744 * If the device keeps failing read/write commands but TEST UNIT
1745 * READY always completes successfully we assume that medium
1746 * access is no longer possible and take the device offline.
1748 if (sdkp
->medium_access_timed_out
>= sdkp
->max_medium_access_timeouts
) {
1749 scmd_printk(KERN_ERR
, scmd
,
1750 "Medium access timeout failure. Offlining disk!\n");
1751 scsi_device_set_state(scmd
->device
, SDEV_OFFLINE
);
1759 static unsigned int sd_completed_bytes(struct scsi_cmnd
*scmd
)
1761 struct request
*req
= scmd
->request
;
1762 struct scsi_device
*sdev
= scmd
->device
;
1763 unsigned int transferred
, good_bytes
;
1764 u64 start_lba
, end_lba
, bad_lba
;
1767 * Some commands have a payload smaller than the device logical
1768 * block size (e.g. INQUIRY on a 4K disk).
1770 if (scsi_bufflen(scmd
) <= sdev
->sector_size
)
1773 /* Check if we have a 'bad_lba' information */
1774 if (!scsi_get_sense_info_fld(scmd
->sense_buffer
,
1775 SCSI_SENSE_BUFFERSIZE
,
1780 * If the bad lba was reported incorrectly, we have no idea where
1783 start_lba
= sectors_to_logical(sdev
, blk_rq_pos(req
));
1784 end_lba
= start_lba
+ bytes_to_logical(sdev
, scsi_bufflen(scmd
));
1785 if (bad_lba
< start_lba
|| bad_lba
>= end_lba
)
1789 * resid is optional but mostly filled in. When it's unused,
1790 * its value is zero, so we assume the whole buffer transferred
1792 transferred
= scsi_bufflen(scmd
) - scsi_get_resid(scmd
);
1794 /* This computation should always be done in terms of the
1795 * resolution of the device's medium.
1797 good_bytes
= logical_to_bytes(sdev
, bad_lba
- start_lba
);
1799 return min(good_bytes
, transferred
);
1803 * sd_done - bottom half handler: called when the lower level
1804 * driver has completed (successfully or otherwise) a scsi command.
1805 * @SCpnt: mid-level's per command structure.
1807 * Note: potentially run from within an ISR. Must not block.
1809 static int sd_done(struct scsi_cmnd
*SCpnt
)
1811 int result
= SCpnt
->result
;
1812 unsigned int good_bytes
= result
? 0 : scsi_bufflen(SCpnt
);
1813 struct scsi_sense_hdr sshdr
;
1814 struct scsi_disk
*sdkp
= scsi_disk(SCpnt
->request
->rq_disk
);
1815 struct request
*req
= SCpnt
->request
;
1816 int sense_valid
= 0;
1817 int sense_deferred
= 0;
1819 switch (req_op(req
)) {
1820 case REQ_OP_DISCARD
:
1821 case REQ_OP_WRITE_SAME
:
1822 case REQ_OP_ZONE_RESET
:
1824 good_bytes
= blk_rq_bytes(req
);
1825 scsi_set_resid(SCpnt
, 0);
1828 scsi_set_resid(SCpnt
, blk_rq_bytes(req
));
1831 case REQ_OP_ZONE_REPORT
:
1833 good_bytes
= scsi_bufflen(SCpnt
)
1834 - scsi_get_resid(SCpnt
);
1835 scsi_set_resid(SCpnt
, 0);
1838 scsi_set_resid(SCpnt
, blk_rq_bytes(req
));
1844 sense_valid
= scsi_command_normalize_sense(SCpnt
, &sshdr
);
1846 sense_deferred
= scsi_sense_is_deferred(&sshdr
);
1848 sdkp
->medium_access_timed_out
= 0;
1850 if (driver_byte(result
) != DRIVER_SENSE
&&
1851 (!sense_valid
|| sense_deferred
))
1854 switch (sshdr
.sense_key
) {
1855 case HARDWARE_ERROR
:
1857 good_bytes
= sd_completed_bytes(SCpnt
);
1859 case RECOVERED_ERROR
:
1860 good_bytes
= scsi_bufflen(SCpnt
);
1863 /* This indicates a false check condition, so ignore it. An
1864 * unknown amount of data was transferred so treat it as an
1868 memset(SCpnt
->sense_buffer
, 0, SCSI_SENSE_BUFFERSIZE
);
1870 case ABORTED_COMMAND
:
1871 if (sshdr
.asc
== 0x10) /* DIF: Target detected corruption */
1872 good_bytes
= sd_completed_bytes(SCpnt
);
1874 case ILLEGAL_REQUEST
:
1875 switch (sshdr
.asc
) {
1876 case 0x10: /* DIX: Host detected corruption */
1877 good_bytes
= sd_completed_bytes(SCpnt
);
1879 case 0x20: /* INVALID COMMAND OPCODE */
1880 case 0x24: /* INVALID FIELD IN CDB */
1881 switch (SCpnt
->cmnd
[0]) {
1883 sd_config_discard(sdkp
, SD_LBP_DISABLE
);
1887 if (SCpnt
->cmnd
[1] & 8) { /* UNMAP */
1888 sd_config_discard(sdkp
, SD_LBP_DISABLE
);
1890 sdkp
->device
->no_write_same
= 1;
1891 sd_config_write_same(sdkp
);
1892 req
->__data_len
= blk_rq_bytes(req
);
1893 req
->rq_flags
|= RQF_QUIET
;
1904 if (sd_is_zoned(sdkp
))
1905 sd_zbc_complete(SCpnt
, good_bytes
, &sshdr
);
1907 SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO
, SCpnt
,
1908 "sd_done: completed %d of %d bytes\n",
1909 good_bytes
, scsi_bufflen(SCpnt
)));
1911 if (rq_data_dir(SCpnt
->request
) == READ
&& scsi_prot_sg_count(SCpnt
))
1912 sd_dif_complete(SCpnt
, good_bytes
);
1918 * spinup disk - called only in sd_revalidate_disk()
1921 sd_spinup_disk(struct scsi_disk
*sdkp
)
1923 unsigned char cmd
[10];
1924 unsigned long spintime_expire
= 0;
1925 int retries
, spintime
;
1926 unsigned int the_result
;
1927 struct scsi_sense_hdr sshdr
;
1928 int sense_valid
= 0;
1932 /* Spin up drives, as required. Only do this at boot time */
1933 /* Spinup needs to be done for module loads too. */
1938 cmd
[0] = TEST_UNIT_READY
;
1939 memset((void *) &cmd
[1], 0, 9);
1941 the_result
= scsi_execute_req(sdkp
->device
, cmd
,
1944 SD_MAX_RETRIES
, NULL
);
1947 * If the drive has indicated to us that it
1948 * doesn't have any media in it, don't bother
1949 * with any more polling.
1951 if (media_not_present(sdkp
, &sshdr
))
1955 sense_valid
= scsi_sense_valid(&sshdr
);
1957 } while (retries
< 3 &&
1958 (!scsi_status_is_good(the_result
) ||
1959 ((driver_byte(the_result
) & DRIVER_SENSE
) &&
1960 sense_valid
&& sshdr
.sense_key
== UNIT_ATTENTION
)));
1962 if ((driver_byte(the_result
) & DRIVER_SENSE
) == 0) {
1963 /* no sense, TUR either succeeded or failed
1964 * with a status error */
1965 if(!spintime
&& !scsi_status_is_good(the_result
)) {
1966 sd_print_result(sdkp
, "Test Unit Ready failed",
1973 * The device does not want the automatic start to be issued.
1975 if (sdkp
->device
->no_start_on_add
)
1978 if (sense_valid
&& sshdr
.sense_key
== NOT_READY
) {
1979 if (sshdr
.asc
== 4 && sshdr
.ascq
== 3)
1980 break; /* manual intervention required */
1981 if (sshdr
.asc
== 4 && sshdr
.ascq
== 0xb)
1982 break; /* standby */
1983 if (sshdr
.asc
== 4 && sshdr
.ascq
== 0xc)
1984 break; /* unavailable */
1986 * Issue command to spin up drive when not ready
1989 sd_printk(KERN_NOTICE
, sdkp
, "Spinning up disk...");
1990 cmd
[0] = START_STOP
;
1991 cmd
[1] = 1; /* Return immediately */
1992 memset((void *) &cmd
[2], 0, 8);
1993 cmd
[4] = 1; /* Start spin cycle */
1994 if (sdkp
->device
->start_stop_pwr_cond
)
1996 scsi_execute_req(sdkp
->device
, cmd
, DMA_NONE
,
1998 SD_TIMEOUT
, SD_MAX_RETRIES
,
2000 spintime_expire
= jiffies
+ 100 * HZ
;
2003 /* Wait 1 second for next try */
2008 * Wait for USB flash devices with slow firmware.
2009 * Yes, this sense key/ASC combination shouldn't
2010 * occur here. It's characteristic of these devices.
2012 } else if (sense_valid
&&
2013 sshdr
.sense_key
== UNIT_ATTENTION
&&
2014 sshdr
.asc
== 0x28) {
2016 spintime_expire
= jiffies
+ 5 * HZ
;
2019 /* Wait 1 second for next try */
2022 /* we don't understand the sense code, so it's
2023 * probably pointless to loop */
2025 sd_printk(KERN_NOTICE
, sdkp
, "Unit Not Ready\n");
2026 sd_print_sense_hdr(sdkp
, &sshdr
);
2031 } while (spintime
&& time_before_eq(jiffies
, spintime_expire
));
2034 if (scsi_status_is_good(the_result
))
2037 printk("not responding...\n");
2042 * Determine whether disk supports Data Integrity Field.
2044 static int sd_read_protection_type(struct scsi_disk
*sdkp
, unsigned char *buffer
)
2046 struct scsi_device
*sdp
= sdkp
->device
;
2050 if (scsi_device_protection(sdp
) == 0 || (buffer
[12] & 1) == 0)
2053 type
= ((buffer
[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2055 if (type
> T10_PI_TYPE3_PROTECTION
)
2057 else if (scsi_host_dif_capable(sdp
->host
, type
))
2060 if (sdkp
->first_scan
|| type
!= sdkp
->protection_type
)
2063 sd_printk(KERN_ERR
, sdkp
, "formatted with unsupported" \
2064 " protection type %u. Disabling disk!\n",
2068 sd_printk(KERN_NOTICE
, sdkp
,
2069 "Enabling DIF Type %u protection\n", type
);
2072 sd_printk(KERN_NOTICE
, sdkp
,
2073 "Disabling DIF Type %u protection\n", type
);
2077 sdkp
->protection_type
= type
;
2082 static void read_capacity_error(struct scsi_disk
*sdkp
, struct scsi_device
*sdp
,
2083 struct scsi_sense_hdr
*sshdr
, int sense_valid
,
2086 if (driver_byte(the_result
) & DRIVER_SENSE
)
2087 sd_print_sense_hdr(sdkp
, sshdr
);
2089 sd_printk(KERN_NOTICE
, sdkp
, "Sense not available.\n");
2092 * Set dirty bit for removable devices if not ready -
2093 * sometimes drives will not report this properly.
2095 if (sdp
->removable
&&
2096 sense_valid
&& sshdr
->sense_key
== NOT_READY
)
2097 set_media_not_present(sdkp
);
2100 * We used to set media_present to 0 here to indicate no media
2101 * in the drive, but some drives fail read capacity even with
2102 * media present, so we can't do that.
2104 sdkp
->capacity
= 0; /* unknown mapped to zero - as usual */
2108 #if RC16_LEN > SD_BUF_SIZE
2109 #error RC16_LEN must not be more than SD_BUF_SIZE
2112 #define READ_CAPACITY_RETRIES_ON_RESET 10
2114 static int read_capacity_16(struct scsi_disk
*sdkp
, struct scsi_device
*sdp
,
2115 unsigned char *buffer
)
2117 unsigned char cmd
[16];
2118 struct scsi_sense_hdr sshdr
;
2119 int sense_valid
= 0;
2121 int retries
= 3, reset_retries
= READ_CAPACITY_RETRIES_ON_RESET
;
2122 unsigned int alignment
;
2123 unsigned long long lba
;
2124 unsigned sector_size
;
2126 if (sdp
->no_read_capacity_16
)
2131 cmd
[0] = SERVICE_ACTION_IN_16
;
2132 cmd
[1] = SAI_READ_CAPACITY_16
;
2134 memset(buffer
, 0, RC16_LEN
);
2136 the_result
= scsi_execute_req(sdp
, cmd
, DMA_FROM_DEVICE
,
2137 buffer
, RC16_LEN
, &sshdr
,
2138 SD_TIMEOUT
, SD_MAX_RETRIES
, NULL
);
2140 if (media_not_present(sdkp
, &sshdr
))
2144 sense_valid
= scsi_sense_valid(&sshdr
);
2146 sshdr
.sense_key
== ILLEGAL_REQUEST
&&
2147 (sshdr
.asc
== 0x20 || sshdr
.asc
== 0x24) &&
2149 /* Invalid Command Operation Code or
2150 * Invalid Field in CDB, just retry
2151 * silently with RC10 */
2154 sshdr
.sense_key
== UNIT_ATTENTION
&&
2155 sshdr
.asc
== 0x29 && sshdr
.ascq
== 0x00)
2156 /* Device reset might occur several times,
2157 * give it one more chance */
2158 if (--reset_retries
> 0)
2163 } while (the_result
&& retries
);
2166 sd_print_result(sdkp
, "Read Capacity(16) failed", the_result
);
2167 read_capacity_error(sdkp
, sdp
, &sshdr
, sense_valid
, the_result
);
2171 sector_size
= get_unaligned_be32(&buffer
[8]);
2172 lba
= get_unaligned_be64(&buffer
[0]);
2174 if (sd_read_protection_type(sdkp
, buffer
) < 0) {
2179 if ((sizeof(sdkp
->capacity
) == 4) && (lba
>= 0xffffffffULL
)) {
2180 sd_printk(KERN_ERR
, sdkp
, "Too big for this kernel. Use a "
2181 "kernel compiled with support for large block "
2187 /* Logical blocks per physical block exponent */
2188 sdkp
->physical_block_size
= (1 << (buffer
[13] & 0xf)) * sector_size
;
2191 sdkp
->rc_basis
= (buffer
[12] >> 4) & 0x3;
2193 /* Lowest aligned logical block */
2194 alignment
= ((buffer
[14] & 0x3f) << 8 | buffer
[15]) * sector_size
;
2195 blk_queue_alignment_offset(sdp
->request_queue
, alignment
);
2196 if (alignment
&& sdkp
->first_scan
)
2197 sd_printk(KERN_NOTICE
, sdkp
,
2198 "physical block alignment offset: %u\n", alignment
);
2200 if (buffer
[14] & 0x80) { /* LBPME */
2203 if (buffer
[14] & 0x40) /* LBPRZ */
2206 sd_config_discard(sdkp
, SD_LBP_WS16
);
2209 sdkp
->capacity
= lba
+ 1;
2213 static int read_capacity_10(struct scsi_disk
*sdkp
, struct scsi_device
*sdp
,
2214 unsigned char *buffer
)
2216 unsigned char cmd
[16];
2217 struct scsi_sense_hdr sshdr
;
2218 int sense_valid
= 0;
2220 int retries
= 3, reset_retries
= READ_CAPACITY_RETRIES_ON_RESET
;
2222 unsigned sector_size
;
2225 cmd
[0] = READ_CAPACITY
;
2226 memset(&cmd
[1], 0, 9);
2227 memset(buffer
, 0, 8);
2229 the_result
= scsi_execute_req(sdp
, cmd
, DMA_FROM_DEVICE
,
2231 SD_TIMEOUT
, SD_MAX_RETRIES
, NULL
);
2233 if (media_not_present(sdkp
, &sshdr
))
2237 sense_valid
= scsi_sense_valid(&sshdr
);
2239 sshdr
.sense_key
== UNIT_ATTENTION
&&
2240 sshdr
.asc
== 0x29 && sshdr
.ascq
== 0x00)
2241 /* Device reset might occur several times,
2242 * give it one more chance */
2243 if (--reset_retries
> 0)
2248 } while (the_result
&& retries
);
2251 sd_print_result(sdkp
, "Read Capacity(10) failed", the_result
);
2252 read_capacity_error(sdkp
, sdp
, &sshdr
, sense_valid
, the_result
);
2256 sector_size
= get_unaligned_be32(&buffer
[4]);
2257 lba
= get_unaligned_be32(&buffer
[0]);
2259 if (sdp
->no_read_capacity_16
&& (lba
== 0xffffffff)) {
2260 /* Some buggy (usb cardreader) devices return an lba of
2261 0xffffffff when the want to report a size of 0 (with
2262 which they really mean no media is present) */
2264 sdkp
->physical_block_size
= sector_size
;
2268 if ((sizeof(sdkp
->capacity
) == 4) && (lba
== 0xffffffff)) {
2269 sd_printk(KERN_ERR
, sdkp
, "Too big for this kernel. Use a "
2270 "kernel compiled with support for large block "
2276 sdkp
->capacity
= lba
+ 1;
2277 sdkp
->physical_block_size
= sector_size
;
2281 static int sd_try_rc16_first(struct scsi_device
*sdp
)
2283 if (sdp
->host
->max_cmd_len
< 16)
2285 if (sdp
->try_rc_10_first
)
2287 if (sdp
->scsi_level
> SCSI_SPC_2
)
2289 if (scsi_device_protection(sdp
))
2295 * read disk capacity
2298 sd_read_capacity(struct scsi_disk
*sdkp
, unsigned char *buffer
)
2301 struct scsi_device
*sdp
= sdkp
->device
;
2303 if (sd_try_rc16_first(sdp
)) {
2304 sector_size
= read_capacity_16(sdkp
, sdp
, buffer
);
2305 if (sector_size
== -EOVERFLOW
)
2307 if (sector_size
== -ENODEV
)
2309 if (sector_size
< 0)
2310 sector_size
= read_capacity_10(sdkp
, sdp
, buffer
);
2311 if (sector_size
< 0)
2314 sector_size
= read_capacity_10(sdkp
, sdp
, buffer
);
2315 if (sector_size
== -EOVERFLOW
)
2317 if (sector_size
< 0)
2319 if ((sizeof(sdkp
->capacity
) > 4) &&
2320 (sdkp
->capacity
> 0xffffffffULL
)) {
2321 int old_sector_size
= sector_size
;
2322 sd_printk(KERN_NOTICE
, sdkp
, "Very big device. "
2323 "Trying to use READ CAPACITY(16).\n");
2324 sector_size
= read_capacity_16(sdkp
, sdp
, buffer
);
2325 if (sector_size
< 0) {
2326 sd_printk(KERN_NOTICE
, sdkp
,
2327 "Using 0xffffffff as device size\n");
2328 sdkp
->capacity
= 1 + (sector_t
) 0xffffffff;
2329 sector_size
= old_sector_size
;
2335 /* Some devices are known to return the total number of blocks,
2336 * not the highest block number. Some devices have versions
2337 * which do this and others which do not. Some devices we might
2338 * suspect of doing this but we don't know for certain.
2340 * If we know the reported capacity is wrong, decrement it. If
2341 * we can only guess, then assume the number of blocks is even
2342 * (usually true but not always) and err on the side of lowering
2345 if (sdp
->fix_capacity
||
2346 (sdp
->guess_capacity
&& (sdkp
->capacity
& 0x01))) {
2347 sd_printk(KERN_INFO
, sdkp
, "Adjusting the sector count "
2348 "from its reported value: %llu\n",
2349 (unsigned long long) sdkp
->capacity
);
2354 if (sector_size
== 0) {
2356 sd_printk(KERN_NOTICE
, sdkp
, "Sector size 0 reported, "
2360 if (sector_size
!= 512 &&
2361 sector_size
!= 1024 &&
2362 sector_size
!= 2048 &&
2363 sector_size
!= 4096) {
2364 sd_printk(KERN_NOTICE
, sdkp
, "Unsupported sector size %d.\n",
2367 * The user might want to re-format the drive with
2368 * a supported sectorsize. Once this happens, it
2369 * would be relatively trivial to set the thing up.
2370 * For this reason, we leave the thing in the table.
2374 * set a bogus sector size so the normal read/write
2375 * logic in the block layer will eventually refuse any
2376 * request on this device without tripping over power
2377 * of two sector size assumptions
2381 blk_queue_logical_block_size(sdp
->request_queue
, sector_size
);
2382 blk_queue_physical_block_size(sdp
->request_queue
,
2383 sdkp
->physical_block_size
);
2384 sdkp
->device
->sector_size
= sector_size
;
2386 if (sdkp
->capacity
> 0xffffffff)
2387 sdp
->use_16_for_rw
= 1;
2392 * Print disk capacity
2395 sd_print_capacity(struct scsi_disk
*sdkp
,
2396 sector_t old_capacity
)
2398 int sector_size
= sdkp
->device
->sector_size
;
2399 char cap_str_2
[10], cap_str_10
[10];
2401 string_get_size(sdkp
->capacity
, sector_size
,
2402 STRING_UNITS_2
, cap_str_2
, sizeof(cap_str_2
));
2403 string_get_size(sdkp
->capacity
, sector_size
,
2404 STRING_UNITS_10
, cap_str_10
,
2405 sizeof(cap_str_10
));
2407 if (sdkp
->first_scan
|| old_capacity
!= sdkp
->capacity
) {
2408 sd_printk(KERN_NOTICE
, sdkp
,
2409 "%llu %d-byte logical blocks: (%s/%s)\n",
2410 (unsigned long long)sdkp
->capacity
,
2411 sector_size
, cap_str_10
, cap_str_2
);
2413 if (sdkp
->physical_block_size
!= sector_size
)
2414 sd_printk(KERN_NOTICE
, sdkp
,
2415 "%u-byte physical blocks\n",
2416 sdkp
->physical_block_size
);
2418 sd_zbc_print_zones(sdkp
);
2422 /* called with buffer of length 512 */
2424 sd_do_mode_sense(struct scsi_device
*sdp
, int dbd
, int modepage
,
2425 unsigned char *buffer
, int len
, struct scsi_mode_data
*data
,
2426 struct scsi_sense_hdr
*sshdr
)
2428 return scsi_mode_sense(sdp
, dbd
, modepage
, buffer
, len
,
2429 SD_TIMEOUT
, SD_MAX_RETRIES
, data
,
2434 * read write protect setting, if possible - called only in sd_revalidate_disk()
2435 * called with buffer of length SD_BUF_SIZE
2438 sd_read_write_protect_flag(struct scsi_disk
*sdkp
, unsigned char *buffer
)
2441 struct scsi_device
*sdp
= sdkp
->device
;
2442 struct scsi_mode_data data
;
2443 int old_wp
= sdkp
->write_prot
;
2445 set_disk_ro(sdkp
->disk
, 0);
2446 if (sdp
->skip_ms_page_3f
) {
2447 sd_first_printk(KERN_NOTICE
, sdkp
, "Assuming Write Enabled\n");
2451 if (sdp
->use_192_bytes_for_3f
) {
2452 res
= sd_do_mode_sense(sdp
, 0, 0x3F, buffer
, 192, &data
, NULL
);
2455 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2456 * We have to start carefully: some devices hang if we ask
2457 * for more than is available.
2459 res
= sd_do_mode_sense(sdp
, 0, 0x3F, buffer
, 4, &data
, NULL
);
2462 * Second attempt: ask for page 0 When only page 0 is
2463 * implemented, a request for page 3F may return Sense Key
2464 * 5: Illegal Request, Sense Code 24: Invalid field in
2467 if (!scsi_status_is_good(res
))
2468 res
= sd_do_mode_sense(sdp
, 0, 0, buffer
, 4, &data
, NULL
);
2471 * Third attempt: ask 255 bytes, as we did earlier.
2473 if (!scsi_status_is_good(res
))
2474 res
= sd_do_mode_sense(sdp
, 0, 0x3F, buffer
, 255,
2478 if (!scsi_status_is_good(res
)) {
2479 sd_first_printk(KERN_WARNING
, sdkp
,
2480 "Test WP failed, assume Write Enabled\n");
2482 sdkp
->write_prot
= ((data
.device_specific
& 0x80) != 0);
2483 set_disk_ro(sdkp
->disk
, sdkp
->write_prot
);
2484 if (sdkp
->first_scan
|| old_wp
!= sdkp
->write_prot
) {
2485 sd_printk(KERN_NOTICE
, sdkp
, "Write Protect is %s\n",
2486 sdkp
->write_prot
? "on" : "off");
2487 sd_printk(KERN_DEBUG
, sdkp
, "Mode Sense: %4ph\n", buffer
);
2493 * sd_read_cache_type - called only from sd_revalidate_disk()
2494 * called with buffer of length SD_BUF_SIZE
2497 sd_read_cache_type(struct scsi_disk
*sdkp
, unsigned char *buffer
)
2500 struct scsi_device
*sdp
= sdkp
->device
;
2505 struct scsi_mode_data data
;
2506 struct scsi_sense_hdr sshdr
;
2507 int old_wce
= sdkp
->WCE
;
2508 int old_rcd
= sdkp
->RCD
;
2509 int old_dpofua
= sdkp
->DPOFUA
;
2512 if (sdkp
->cache_override
)
2516 if (sdp
->skip_ms_page_8
) {
2517 if (sdp
->type
== TYPE_RBC
)
2520 if (sdp
->skip_ms_page_3f
)
2523 if (sdp
->use_192_bytes_for_3f
)
2527 } else if (sdp
->type
== TYPE_RBC
) {
2535 /* cautiously ask */
2536 res
= sd_do_mode_sense(sdp
, dbd
, modepage
, buffer
, first_len
,
2539 if (!scsi_status_is_good(res
))
2542 if (!data
.header_length
) {
2545 sd_first_printk(KERN_ERR
, sdkp
,
2546 "Missing header in MODE_SENSE response\n");
2549 /* that went OK, now ask for the proper length */
2553 * We're only interested in the first three bytes, actually.
2554 * But the data cache page is defined for the first 20.
2558 else if (len
> SD_BUF_SIZE
) {
2559 sd_first_printk(KERN_NOTICE
, sdkp
, "Truncating mode parameter "
2560 "data from %d to %d bytes\n", len
, SD_BUF_SIZE
);
2563 if (modepage
== 0x3F && sdp
->use_192_bytes_for_3f
)
2567 if (len
> first_len
)
2568 res
= sd_do_mode_sense(sdp
, dbd
, modepage
, buffer
, len
,
2571 if (scsi_status_is_good(res
)) {
2572 int offset
= data
.header_length
+ data
.block_descriptor_length
;
2574 while (offset
< len
) {
2575 u8 page_code
= buffer
[offset
] & 0x3F;
2576 u8 spf
= buffer
[offset
] & 0x40;
2578 if (page_code
== 8 || page_code
== 6) {
2579 /* We're interested only in the first 3 bytes.
2581 if (len
- offset
<= 2) {
2582 sd_first_printk(KERN_ERR
, sdkp
,
2583 "Incomplete mode parameter "
2587 modepage
= page_code
;
2591 /* Go to the next page */
2592 if (spf
&& len
- offset
> 3)
2593 offset
+= 4 + (buffer
[offset
+2] << 8) +
2595 else if (!spf
&& len
- offset
> 1)
2596 offset
+= 2 + buffer
[offset
+1];
2598 sd_first_printk(KERN_ERR
, sdkp
,
2600 "parameter data\n");
2606 sd_first_printk(KERN_ERR
, sdkp
, "No Caching mode page found\n");
2610 if (modepage
== 8) {
2611 sdkp
->WCE
= ((buffer
[offset
+ 2] & 0x04) != 0);
2612 sdkp
->RCD
= ((buffer
[offset
+ 2] & 0x01) != 0);
2614 sdkp
->WCE
= ((buffer
[offset
+ 2] & 0x01) == 0);
2618 sdkp
->DPOFUA
= (data
.device_specific
& 0x10) != 0;
2619 if (sdp
->broken_fua
) {
2620 sd_first_printk(KERN_NOTICE
, sdkp
, "Disabling FUA\n");
2622 } else if (sdkp
->DPOFUA
&& !sdkp
->device
->use_10_for_rw
&&
2623 !sdkp
->device
->use_16_for_rw
) {
2624 sd_first_printk(KERN_NOTICE
, sdkp
,
2625 "Uses READ/WRITE(6), disabling FUA\n");
2629 /* No cache flush allowed for write protected devices */
2630 if (sdkp
->WCE
&& sdkp
->write_prot
)
2633 if (sdkp
->first_scan
|| old_wce
!= sdkp
->WCE
||
2634 old_rcd
!= sdkp
->RCD
|| old_dpofua
!= sdkp
->DPOFUA
)
2635 sd_printk(KERN_NOTICE
, sdkp
,
2636 "Write cache: %s, read cache: %s, %s\n",
2637 sdkp
->WCE
? "enabled" : "disabled",
2638 sdkp
->RCD
? "disabled" : "enabled",
2639 sdkp
->DPOFUA
? "supports DPO and FUA"
2640 : "doesn't support DPO or FUA");
2646 if (scsi_sense_valid(&sshdr
) &&
2647 sshdr
.sense_key
== ILLEGAL_REQUEST
&&
2648 sshdr
.asc
== 0x24 && sshdr
.ascq
== 0x0)
2649 /* Invalid field in CDB */
2650 sd_first_printk(KERN_NOTICE
, sdkp
, "Cache data unavailable\n");
2652 sd_first_printk(KERN_ERR
, sdkp
,
2653 "Asking for cache data failed\n");
2656 if (sdp
->wce_default_on
) {
2657 sd_first_printk(KERN_NOTICE
, sdkp
,
2658 "Assuming drive cache: write back\n");
2661 sd_first_printk(KERN_ERR
, sdkp
,
2662 "Assuming drive cache: write through\n");
2670 * The ATO bit indicates whether the DIF application tag is available
2671 * for use by the operating system.
2673 static void sd_read_app_tag_own(struct scsi_disk
*sdkp
, unsigned char *buffer
)
2676 struct scsi_device
*sdp
= sdkp
->device
;
2677 struct scsi_mode_data data
;
2678 struct scsi_sense_hdr sshdr
;
2680 if (sdp
->type
!= TYPE_DISK
&& sdp
->type
!= TYPE_ZBC
)
2683 if (sdkp
->protection_type
== 0)
2686 res
= scsi_mode_sense(sdp
, 1, 0x0a, buffer
, 36, SD_TIMEOUT
,
2687 SD_MAX_RETRIES
, &data
, &sshdr
);
2689 if (!scsi_status_is_good(res
) || !data
.header_length
||
2691 sd_first_printk(KERN_WARNING
, sdkp
,
2692 "getting Control mode page failed, assume no ATO\n");
2694 if (scsi_sense_valid(&sshdr
))
2695 sd_print_sense_hdr(sdkp
, &sshdr
);
2700 offset
= data
.header_length
+ data
.block_descriptor_length
;
2702 if ((buffer
[offset
] & 0x3f) != 0x0a) {
2703 sd_first_printk(KERN_ERR
, sdkp
, "ATO Got wrong page\n");
2707 if ((buffer
[offset
+ 5] & 0x80) == 0)
2716 * sd_read_block_limits - Query disk device for preferred I/O sizes.
2717 * @sdkp: disk to query
2719 static void sd_read_block_limits(struct scsi_disk
*sdkp
)
2721 unsigned int sector_sz
= sdkp
->device
->sector_size
;
2722 const int vpd_len
= 64;
2723 unsigned char *buffer
= kmalloc(vpd_len
, GFP_KERNEL
);
2726 /* Block Limits VPD */
2727 scsi_get_vpd_page(sdkp
->device
, 0xb0, buffer
, vpd_len
))
2730 blk_queue_io_min(sdkp
->disk
->queue
,
2731 get_unaligned_be16(&buffer
[6]) * sector_sz
);
2733 sdkp
->max_xfer_blocks
= get_unaligned_be32(&buffer
[8]);
2734 sdkp
->opt_xfer_blocks
= get_unaligned_be32(&buffer
[12]);
2736 if (buffer
[3] == 0x3c) {
2737 unsigned int lba_count
, desc_count
;
2739 sdkp
->max_ws_blocks
= (u32
)get_unaligned_be64(&buffer
[36]);
2744 lba_count
= get_unaligned_be32(&buffer
[20]);
2745 desc_count
= get_unaligned_be32(&buffer
[24]);
2747 if (lba_count
&& desc_count
)
2748 sdkp
->max_unmap_blocks
= lba_count
;
2750 sdkp
->unmap_granularity
= get_unaligned_be32(&buffer
[28]);
2752 if (buffer
[32] & 0x80)
2753 sdkp
->unmap_alignment
=
2754 get_unaligned_be32(&buffer
[32]) & ~(1 << 31);
2756 if (!sdkp
->lbpvpd
) { /* LBP VPD page not provided */
2758 if (sdkp
->max_unmap_blocks
)
2759 sd_config_discard(sdkp
, SD_LBP_UNMAP
);
2761 sd_config_discard(sdkp
, SD_LBP_WS16
);
2763 } else { /* LBP VPD page tells us what to use */
2764 if (sdkp
->lbpu
&& sdkp
->max_unmap_blocks
&& !sdkp
->lbprz
)
2765 sd_config_discard(sdkp
, SD_LBP_UNMAP
);
2766 else if (sdkp
->lbpws
)
2767 sd_config_discard(sdkp
, SD_LBP_WS16
);
2768 else if (sdkp
->lbpws10
)
2769 sd_config_discard(sdkp
, SD_LBP_WS10
);
2770 else if (sdkp
->lbpu
&& sdkp
->max_unmap_blocks
)
2771 sd_config_discard(sdkp
, SD_LBP_UNMAP
);
2773 sd_config_discard(sdkp
, SD_LBP_DISABLE
);
2782 * sd_read_block_characteristics - Query block dev. characteristics
2783 * @sdkp: disk to query
2785 static void sd_read_block_characteristics(struct scsi_disk
*sdkp
)
2787 struct request_queue
*q
= sdkp
->disk
->queue
;
2788 unsigned char *buffer
;
2790 const int vpd_len
= 64;
2792 buffer
= kmalloc(vpd_len
, GFP_KERNEL
);
2795 /* Block Device Characteristics VPD */
2796 scsi_get_vpd_page(sdkp
->device
, 0xb1, buffer
, vpd_len
))
2799 rot
= get_unaligned_be16(&buffer
[4]);
2802 queue_flag_set_unlocked(QUEUE_FLAG_NONROT
, q
);
2803 queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM
, q
);
2806 if (sdkp
->device
->type
== TYPE_ZBC
) {
2808 q
->limits
.zoned
= BLK_ZONED_HM
;
2810 sdkp
->zoned
= (buffer
[8] >> 4) & 3;
2811 if (sdkp
->zoned
== 1)
2813 q
->limits
.zoned
= BLK_ZONED_HA
;
2816 * Treat drive-managed devices as
2817 * regular block devices.
2819 q
->limits
.zoned
= BLK_ZONED_NONE
;
2821 if (blk_queue_is_zoned(q
) && sdkp
->first_scan
)
2822 sd_printk(KERN_NOTICE
, sdkp
, "Host-%s zoned block device\n",
2823 q
->limits
.zoned
== BLK_ZONED_HM
? "managed" : "aware");
2830 * sd_read_block_provisioning - Query provisioning VPD page
2831 * @sdkp: disk to query
2833 static void sd_read_block_provisioning(struct scsi_disk
*sdkp
)
2835 unsigned char *buffer
;
2836 const int vpd_len
= 8;
2838 if (sdkp
->lbpme
== 0)
2841 buffer
= kmalloc(vpd_len
, GFP_KERNEL
);
2843 if (!buffer
|| scsi_get_vpd_page(sdkp
->device
, 0xb2, buffer
, vpd_len
))
2847 sdkp
->lbpu
= (buffer
[5] >> 7) & 1; /* UNMAP */
2848 sdkp
->lbpws
= (buffer
[5] >> 6) & 1; /* WRITE SAME(16) with UNMAP */
2849 sdkp
->lbpws10
= (buffer
[5] >> 5) & 1; /* WRITE SAME(10) with UNMAP */
2855 static void sd_read_write_same(struct scsi_disk
*sdkp
, unsigned char *buffer
)
2857 struct scsi_device
*sdev
= sdkp
->device
;
2859 if (sdev
->host
->no_write_same
) {
2860 sdev
->no_write_same
= 1;
2865 if (scsi_report_opcode(sdev
, buffer
, SD_BUF_SIZE
, INQUIRY
) < 0) {
2866 /* too large values might cause issues with arcmsr */
2867 int vpd_buf_len
= 64;
2869 sdev
->no_report_opcodes
= 1;
2871 /* Disable WRITE SAME if REPORT SUPPORTED OPERATION
2872 * CODES is unsupported and the device has an ATA
2873 * Information VPD page (SAT).
2875 if (!scsi_get_vpd_page(sdev
, 0x89, buffer
, vpd_buf_len
))
2876 sdev
->no_write_same
= 1;
2879 if (scsi_report_opcode(sdev
, buffer
, SD_BUF_SIZE
, WRITE_SAME_16
) == 1)
2882 if (scsi_report_opcode(sdev
, buffer
, SD_BUF_SIZE
, WRITE_SAME
) == 1)
2887 * sd_revalidate_disk - called the first time a new disk is seen,
2888 * performs disk spin up, read_capacity, etc.
2889 * @disk: struct gendisk we care about
2891 static int sd_revalidate_disk(struct gendisk
*disk
)
2893 struct scsi_disk
*sdkp
= scsi_disk(disk
);
2894 struct scsi_device
*sdp
= sdkp
->device
;
2895 struct request_queue
*q
= sdkp
->disk
->queue
;
2896 sector_t old_capacity
= sdkp
->capacity
;
2897 unsigned char *buffer
;
2898 unsigned int dev_max
, rw_max
;
2900 SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO
, sdkp
,
2901 "sd_revalidate_disk\n"));
2904 * If the device is offline, don't try and read capacity or any
2905 * of the other niceties.
2907 if (!scsi_device_online(sdp
))
2910 buffer
= kmalloc(SD_BUF_SIZE
, GFP_KERNEL
);
2912 sd_printk(KERN_WARNING
, sdkp
, "sd_revalidate_disk: Memory "
2913 "allocation failure.\n");
2917 sd_spinup_disk(sdkp
);
2920 * Without media there is no reason to ask; moreover, some devices
2921 * react badly if we do.
2923 if (sdkp
->media_present
) {
2924 sd_read_capacity(sdkp
, buffer
);
2926 if (scsi_device_supports_vpd(sdp
)) {
2927 sd_read_block_provisioning(sdkp
);
2928 sd_read_block_limits(sdkp
);
2929 sd_read_block_characteristics(sdkp
);
2930 sd_zbc_read_zones(sdkp
, buffer
);
2933 sd_print_capacity(sdkp
, old_capacity
);
2935 sd_read_write_protect_flag(sdkp
, buffer
);
2936 sd_read_cache_type(sdkp
, buffer
);
2937 sd_read_app_tag_own(sdkp
, buffer
);
2938 sd_read_write_same(sdkp
, buffer
);
2941 sdkp
->first_scan
= 0;
2944 * We now have all cache related info, determine how we deal
2945 * with flush requests.
2947 sd_set_flush_flag(sdkp
);
2949 /* Initial block count limit based on CDB TRANSFER LENGTH field size. */
2950 dev_max
= sdp
->use_16_for_rw
? SD_MAX_XFER_BLOCKS
: SD_DEF_XFER_BLOCKS
;
2952 /* Some devices report a maximum block count for READ/WRITE requests. */
2953 dev_max
= min_not_zero(dev_max
, sdkp
->max_xfer_blocks
);
2954 q
->limits
.max_dev_sectors
= logical_to_sectors(sdp
, dev_max
);
2957 * Use the device's preferred I/O size for reads and writes
2958 * unless the reported value is unreasonably small, large, or
2961 if (sdkp
->opt_xfer_blocks
&&
2962 sdkp
->opt_xfer_blocks
<= dev_max
&&
2963 sdkp
->opt_xfer_blocks
<= SD_DEF_XFER_BLOCKS
&&
2964 logical_to_bytes(sdp
, sdkp
->opt_xfer_blocks
) >= PAGE_SIZE
) {
2965 q
->limits
.io_opt
= logical_to_bytes(sdp
, sdkp
->opt_xfer_blocks
);
2966 rw_max
= logical_to_sectors(sdp
, sdkp
->opt_xfer_blocks
);
2968 rw_max
= BLK_DEF_MAX_SECTORS
;
2970 /* Combine with controller limits */
2971 q
->limits
.max_sectors
= min(rw_max
, queue_max_hw_sectors(q
));
2973 set_capacity(disk
, logical_to_sectors(sdp
, sdkp
->capacity
));
2974 sd_config_write_same(sdkp
);
2982 * sd_unlock_native_capacity - unlock native capacity
2983 * @disk: struct gendisk to set capacity for
2985 * Block layer calls this function if it detects that partitions
2986 * on @disk reach beyond the end of the device. If the SCSI host
2987 * implements ->unlock_native_capacity() method, it's invoked to
2988 * give it a chance to adjust the device capacity.
2991 * Defined by block layer. Might sleep.
2993 static void sd_unlock_native_capacity(struct gendisk
*disk
)
2995 struct scsi_device
*sdev
= scsi_disk(disk
)->device
;
2997 if (sdev
->host
->hostt
->unlock_native_capacity
)
2998 sdev
->host
->hostt
->unlock_native_capacity(sdev
);
3002 * sd_format_disk_name - format disk name
3003 * @prefix: name prefix - ie. "sd" for SCSI disks
3004 * @index: index of the disk to format name for
3005 * @buf: output buffer
3006 * @buflen: length of the output buffer
3008 * SCSI disk names starts at sda. The 26th device is sdz and the
3009 * 27th is sdaa. The last one for two lettered suffix is sdzz
3010 * which is followed by sdaaa.
3012 * This is basically 26 base counting with one extra 'nil' entry
3013 * at the beginning from the second digit on and can be
3014 * determined using similar method as 26 base conversion with the
3015 * index shifted -1 after each digit is computed.
3021 * 0 on success, -errno on failure.
3023 static int sd_format_disk_name(char *prefix
, int index
, char *buf
, int buflen
)
3025 const int base
= 'z' - 'a' + 1;
3026 char *begin
= buf
+ strlen(prefix
);
3027 char *end
= buf
+ buflen
;
3037 *--p
= 'a' + (index
% unit
);
3038 index
= (index
/ unit
) - 1;
3039 } while (index
>= 0);
3041 memmove(begin
, p
, end
- p
);
3042 memcpy(buf
, prefix
, strlen(prefix
));
3048 * The asynchronous part of sd_probe
3050 static void sd_probe_async(void *data
, async_cookie_t cookie
)
3052 struct scsi_disk
*sdkp
= data
;
3053 struct scsi_device
*sdp
;
3060 index
= sdkp
->index
;
3061 dev
= &sdp
->sdev_gendev
;
3063 gd
->major
= sd_major((index
& 0xf0) >> 4);
3064 gd
->first_minor
= ((index
& 0xf) << 4) | (index
& 0xfff00);
3065 gd
->minors
= SD_MINORS
;
3067 gd
->fops
= &sd_fops
;
3068 gd
->private_data
= &sdkp
->driver
;
3069 gd
->queue
= sdkp
->device
->request_queue
;
3071 /* defaults, until the device tells us otherwise */
3072 sdp
->sector_size
= 512;
3074 sdkp
->media_present
= 1;
3075 sdkp
->write_prot
= 0;
3076 sdkp
->cache_override
= 0;
3080 sdkp
->first_scan
= 1;
3081 sdkp
->max_medium_access_timeouts
= SD_MAX_MEDIUM_TIMEOUTS
;
3083 sd_revalidate_disk(gd
);
3085 gd
->flags
= GENHD_FL_EXT_DEVT
;
3086 if (sdp
->removable
) {
3087 gd
->flags
|= GENHD_FL_REMOVABLE
;
3088 gd
->events
|= DISK_EVENT_MEDIA_CHANGE
;
3091 blk_pm_runtime_init(sdp
->request_queue
, dev
);
3092 device_add_disk(dev
, gd
);
3094 sd_dif_config_host(sdkp
);
3096 sd_revalidate_disk(gd
);
3098 sd_printk(KERN_NOTICE
, sdkp
, "Attached SCSI %sdisk\n",
3099 sdp
->removable
? "removable " : "");
3100 scsi_autopm_put_device(sdp
);
3101 put_device(&sdkp
->dev
);
3105 * sd_probe - called during driver initialization and whenever a
3106 * new scsi device is attached to the system. It is called once
3107 * for each scsi device (not just disks) present.
3108 * @dev: pointer to device object
3110 * Returns 0 if successful (or not interested in this scsi device
3111 * (e.g. scanner)); 1 when there is an error.
3113 * Note: this function is invoked from the scsi mid-level.
3114 * This function sets up the mapping between a given
3115 * <host,channel,id,lun> (found in sdp) and new device name
3116 * (e.g. /dev/sda). More precisely it is the block device major
3117 * and minor number that is chosen here.
3119 * Assume sd_probe is not re-entrant (for time being)
3120 * Also think about sd_probe() and sd_remove() running coincidentally.
3122 static int sd_probe(struct device
*dev
)
3124 struct scsi_device
*sdp
= to_scsi_device(dev
);
3125 struct scsi_disk
*sdkp
;
3130 scsi_autopm_get_device(sdp
);
3132 if (sdp
->type
!= TYPE_DISK
&&
3133 sdp
->type
!= TYPE_ZBC
&&
3134 sdp
->type
!= TYPE_MOD
&&
3135 sdp
->type
!= TYPE_RBC
)
3138 #ifndef CONFIG_BLK_DEV_ZONED
3139 if (sdp
->type
== TYPE_ZBC
)
3142 SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO
, sdp
,
3146 sdkp
= kzalloc(sizeof(*sdkp
), GFP_KERNEL
);
3150 gd
= alloc_disk(SD_MINORS
);
3155 if (!ida_pre_get(&sd_index_ida
, GFP_KERNEL
))
3158 spin_lock(&sd_index_lock
);
3159 error
= ida_get_new(&sd_index_ida
, &index
);
3160 spin_unlock(&sd_index_lock
);
3161 } while (error
== -EAGAIN
);
3164 sdev_printk(KERN_WARNING
, sdp
, "sd_probe: memory exhausted.\n");
3168 error
= sd_format_disk_name("sd", index
, gd
->disk_name
, DISK_NAME_LEN
);
3170 sdev_printk(KERN_WARNING
, sdp
, "SCSI disk (sd) name length exceeded.\n");
3171 goto out_free_index
;
3175 sdkp
->driver
= &sd_template
;
3177 sdkp
->index
= index
;
3178 atomic_set(&sdkp
->openers
, 0);
3179 atomic_set(&sdkp
->device
->ioerr_cnt
, 0);
3181 if (!sdp
->request_queue
->rq_timeout
) {
3182 if (sdp
->type
!= TYPE_MOD
)
3183 blk_queue_rq_timeout(sdp
->request_queue
, SD_TIMEOUT
);
3185 blk_queue_rq_timeout(sdp
->request_queue
,
3189 device_initialize(&sdkp
->dev
);
3190 sdkp
->dev
.parent
= dev
;
3191 sdkp
->dev
.class = &sd_disk_class
;
3192 dev_set_name(&sdkp
->dev
, "%s", dev_name(dev
));
3194 error
= device_add(&sdkp
->dev
);
3196 goto out_free_index
;
3199 dev_set_drvdata(dev
, sdkp
);
3201 get_device(&sdkp
->dev
); /* prevent release before async_schedule */
3202 async_schedule_domain(sd_probe_async
, sdkp
, &scsi_sd_probe_domain
);
3207 spin_lock(&sd_index_lock
);
3208 ida_remove(&sd_index_ida
, index
);
3209 spin_unlock(&sd_index_lock
);
3215 scsi_autopm_put_device(sdp
);
3220 * sd_remove - called whenever a scsi disk (previously recognized by
3221 * sd_probe) is detached from the system. It is called (potentially
3222 * multiple times) during sd module unload.
3223 * @dev: pointer to device object
3225 * Note: this function is invoked from the scsi mid-level.
3226 * This function potentially frees up a device name (e.g. /dev/sdc)
3227 * that could be re-used by a subsequent sd_probe().
3228 * This function is not called when the built-in sd driver is "exit-ed".
3230 static int sd_remove(struct device
*dev
)
3232 struct scsi_disk
*sdkp
;
3235 sdkp
= dev_get_drvdata(dev
);
3236 devt
= disk_devt(sdkp
->disk
);
3237 scsi_autopm_get_device(sdkp
->device
);
3239 async_synchronize_full_domain(&scsi_sd_pm_domain
);
3240 async_synchronize_full_domain(&scsi_sd_probe_domain
);
3241 device_del(&sdkp
->dev
);
3242 del_gendisk(sdkp
->disk
);
3245 sd_zbc_remove(sdkp
);
3247 blk_register_region(devt
, SD_MINORS
, NULL
,
3248 sd_default_probe
, NULL
, NULL
);
3250 mutex_lock(&sd_ref_mutex
);
3251 dev_set_drvdata(dev
, NULL
);
3252 put_device(&sdkp
->dev
);
3253 mutex_unlock(&sd_ref_mutex
);
3259 * scsi_disk_release - Called to free the scsi_disk structure
3260 * @dev: pointer to embedded class device
3262 * sd_ref_mutex must be held entering this routine. Because it is
3263 * called on last put, you should always use the scsi_disk_get()
3264 * scsi_disk_put() helpers which manipulate the semaphore directly
3265 * and never do a direct put_device.
3267 static void scsi_disk_release(struct device
*dev
)
3269 struct scsi_disk
*sdkp
= to_scsi_disk(dev
);
3270 struct gendisk
*disk
= sdkp
->disk
;
3272 spin_lock(&sd_index_lock
);
3273 ida_remove(&sd_index_ida
, sdkp
->index
);
3274 spin_unlock(&sd_index_lock
);
3276 disk
->private_data
= NULL
;
3278 put_device(&sdkp
->device
->sdev_gendev
);
3283 static int sd_start_stop_device(struct scsi_disk
*sdkp
, int start
)
3285 unsigned char cmd
[6] = { START_STOP
}; /* START_VALID */
3286 struct scsi_sense_hdr sshdr
;
3287 struct scsi_device
*sdp
= sdkp
->device
;
3291 cmd
[4] |= 1; /* START */
3293 if (sdp
->start_stop_pwr_cond
)
3294 cmd
[4] |= start
? 1 << 4 : 3 << 4; /* Active or Standby */
3296 if (!scsi_device_online(sdp
))
3299 res
= scsi_execute(sdp
, cmd
, DMA_NONE
, NULL
, 0, NULL
, &sshdr
,
3300 SD_TIMEOUT
, SD_MAX_RETRIES
, 0, RQF_PM
, NULL
);
3302 sd_print_result(sdkp
, "Start/Stop Unit failed", res
);
3303 if (driver_byte(res
) & DRIVER_SENSE
)
3304 sd_print_sense_hdr(sdkp
, &sshdr
);
3305 if (scsi_sense_valid(&sshdr
) &&
3306 /* 0x3a is medium not present */
3311 /* SCSI error codes must not go to the generic layer */
3319 * Send a SYNCHRONIZE CACHE instruction down to the device through
3320 * the normal SCSI command structure. Wait for the command to
3323 static void sd_shutdown(struct device
*dev
)
3325 struct scsi_disk
*sdkp
= dev_get_drvdata(dev
);
3328 return; /* this can happen */
3330 if (pm_runtime_suspended(dev
))
3333 if (sdkp
->WCE
&& sdkp
->media_present
) {
3334 sd_printk(KERN_NOTICE
, sdkp
, "Synchronizing SCSI cache\n");
3335 sd_sync_cache(sdkp
);
3338 if (system_state
!= SYSTEM_RESTART
&& sdkp
->device
->manage_start_stop
) {
3339 sd_printk(KERN_NOTICE
, sdkp
, "Stopping disk\n");
3340 sd_start_stop_device(sdkp
, 0);
3344 static int sd_suspend_common(struct device
*dev
, bool ignore_stop_errors
)
3346 struct scsi_disk
*sdkp
= dev_get_drvdata(dev
);
3349 if (!sdkp
) /* E.g.: runtime suspend following sd_remove() */
3352 if (sdkp
->WCE
&& sdkp
->media_present
) {
3353 sd_printk(KERN_NOTICE
, sdkp
, "Synchronizing SCSI cache\n");
3354 ret
= sd_sync_cache(sdkp
);
3356 /* ignore OFFLINE device */
3363 if (sdkp
->device
->manage_start_stop
) {
3364 sd_printk(KERN_NOTICE
, sdkp
, "Stopping disk\n");
3365 /* an error is not worth aborting a system sleep */
3366 ret
= sd_start_stop_device(sdkp
, 0);
3367 if (ignore_stop_errors
)
3375 static int sd_suspend_system(struct device
*dev
)
3377 return sd_suspend_common(dev
, true);
3380 static int sd_suspend_runtime(struct device
*dev
)
3382 return sd_suspend_common(dev
, false);
3385 static int sd_resume(struct device
*dev
)
3387 struct scsi_disk
*sdkp
= dev_get_drvdata(dev
);
3389 if (!sdkp
) /* E.g.: runtime resume at the start of sd_probe() */
3392 if (!sdkp
->device
->manage_start_stop
)
3395 sd_printk(KERN_NOTICE
, sdkp
, "Starting disk\n");
3396 return sd_start_stop_device(sdkp
, 1);
3400 * init_sd - entry point for this driver (both when built in or when
3403 * Note: this function registers this driver with the scsi mid-level.
3405 static int __init
init_sd(void)
3407 int majors
= 0, i
, err
;
3409 SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3411 for (i
= 0; i
< SD_MAJORS
; i
++) {
3412 if (register_blkdev(sd_major(i
), "sd") != 0)
3415 blk_register_region(sd_major(i
), SD_MINORS
, NULL
,
3416 sd_default_probe
, NULL
, NULL
);
3422 err
= class_register(&sd_disk_class
);
3426 sd_cdb_cache
= kmem_cache_create("sd_ext_cdb", SD_EXT_CDB_SIZE
,
3428 if (!sd_cdb_cache
) {
3429 printk(KERN_ERR
"sd: can't init extended cdb cache\n");
3434 sd_cdb_pool
= mempool_create_slab_pool(SD_MEMPOOL_SIZE
, sd_cdb_cache
);
3436 printk(KERN_ERR
"sd: can't init extended cdb pool\n");
3441 err
= scsi_register_driver(&sd_template
.gendrv
);
3443 goto err_out_driver
;
3448 mempool_destroy(sd_cdb_pool
);
3451 kmem_cache_destroy(sd_cdb_cache
);
3454 class_unregister(&sd_disk_class
);
3456 for (i
= 0; i
< SD_MAJORS
; i
++)
3457 unregister_blkdev(sd_major(i
), "sd");
3462 * exit_sd - exit point for this driver (when it is a module).
3464 * Note: this function unregisters this driver from the scsi mid-level.
3466 static void __exit
exit_sd(void)
3470 SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3472 scsi_unregister_driver(&sd_template
.gendrv
);
3473 mempool_destroy(sd_cdb_pool
);
3474 kmem_cache_destroy(sd_cdb_cache
);
3476 class_unregister(&sd_disk_class
);
3478 for (i
= 0; i
< SD_MAJORS
; i
++) {
3479 blk_unregister_region(sd_major(i
), SD_MINORS
);
3480 unregister_blkdev(sd_major(i
), "sd");
3484 module_init(init_sd
);
3485 module_exit(exit_sd
);
3487 static void sd_print_sense_hdr(struct scsi_disk
*sdkp
,
3488 struct scsi_sense_hdr
*sshdr
)
3490 scsi_print_sense_hdr(sdkp
->device
,
3491 sdkp
->disk
? sdkp
->disk
->disk_name
: NULL
, sshdr
);
3494 static void sd_print_result(const struct scsi_disk
*sdkp
, const char *msg
,
3497 const char *hb_string
= scsi_hostbyte_string(result
);
3498 const char *db_string
= scsi_driverbyte_string(result
);
3500 if (hb_string
|| db_string
)
3501 sd_printk(KERN_INFO
, sdkp
,
3502 "%s: Result: hostbyte=%s driverbyte=%s\n", msg
,
3503 hb_string
? hb_string
: "invalid",
3504 db_string
? db_string
: "invalid");
3506 sd_printk(KERN_INFO
, sdkp
,
3507 "%s: Result: hostbyte=0x%02x driverbyte=0x%02x\n",
3508 msg
, host_byte(result
), driver_byte(result
));