xref: /linux/drivers/scsi/sd.c (revision 5b026e34120766408e76ba19a0e33a9dc996f9f0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *      sd.c Copyright (C) 1992 Drew Eckhardt
4  *           Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5  *
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.
26  *
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.
34  */
35 
36 #include <linux/module.h>
37 #include <linux/fs.h>
38 #include <linux/kernel.h>
39 #include <linux/mm.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/rw_hint.h>
51 #include <linux/major.h>
52 #include <linux/mutex.h>
53 #include <linux/string_helpers.h>
54 #include <linux/slab.h>
55 #include <linux/sed-opal.h>
56 #include <linux/pm_runtime.h>
57 #include <linux/pr.h>
58 #include <linux/t10-pi.h>
59 #include <linux/uaccess.h>
60 #include <asm/unaligned.h>
61 
62 #include <scsi/scsi.h>
63 #include <scsi/scsi_cmnd.h>
64 #include <scsi/scsi_dbg.h>
65 #include <scsi/scsi_device.h>
66 #include <scsi/scsi_driver.h>
67 #include <scsi/scsi_eh.h>
68 #include <scsi/scsi_host.h>
69 #include <scsi/scsi_ioctl.h>
70 #include <scsi/scsicam.h>
71 #include <scsi/scsi_common.h>
72 
73 #include "sd.h"
74 #include "scsi_priv.h"
75 #include "scsi_logging.h"
76 
77 MODULE_AUTHOR("Eric Youngdale");
78 MODULE_DESCRIPTION("SCSI disk (sd) driver");
79 MODULE_LICENSE("GPL");
80 
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
96 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
100 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
101 
102 #define SD_MINORS	16
103 
104 static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
105 		unsigned int mode);
106 static void sd_config_write_same(struct scsi_disk *sdkp,
107 		struct queue_limits *lim);
108 static int  sd_revalidate_disk(struct gendisk *);
109 static void sd_unlock_native_capacity(struct gendisk *disk);
110 static void sd_shutdown(struct device *);
111 static void scsi_disk_release(struct device *cdev);
112 
113 static DEFINE_IDA(sd_index_ida);
114 
115 static mempool_t *sd_page_pool;
116 static struct lock_class_key sd_bio_compl_lkclass;
117 
118 static const char *sd_cache_types[] = {
119 	"write through", "none", "write back",
120 	"write back, no read (daft)"
121 };
122 
123 static void sd_set_flush_flag(struct scsi_disk *sdkp,
124 		struct queue_limits *lim)
125 {
126 	if (sdkp->WCE) {
127 		lim->features |= BLK_FEAT_WRITE_CACHE;
128 		if (sdkp->DPOFUA)
129 			lim->features |= BLK_FEAT_FUA;
130 		else
131 			lim->features &= ~BLK_FEAT_FUA;
132 	} else {
133 		lim->features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA);
134 	}
135 }
136 
137 static ssize_t
138 cache_type_store(struct device *dev, struct device_attribute *attr,
139 		 const char *buf, size_t count)
140 {
141 	int ct, rcd, wce, sp;
142 	struct scsi_disk *sdkp = to_scsi_disk(dev);
143 	struct scsi_device *sdp = sdkp->device;
144 	char buffer[64];
145 	char *buffer_data;
146 	struct scsi_mode_data data;
147 	struct scsi_sense_hdr sshdr;
148 	static const char temp[] = "temporary ";
149 	int len, ret;
150 
151 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
152 		/* no cache control on RBC devices; theoretically they
153 		 * can do it, but there's probably so many exceptions
154 		 * it's not worth the risk */
155 		return -EINVAL;
156 
157 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
158 		buf += sizeof(temp) - 1;
159 		sdkp->cache_override = 1;
160 	} else {
161 		sdkp->cache_override = 0;
162 	}
163 
164 	ct = sysfs_match_string(sd_cache_types, buf);
165 	if (ct < 0)
166 		return -EINVAL;
167 
168 	rcd = ct & 0x01 ? 1 : 0;
169 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
170 
171 	if (sdkp->cache_override) {
172 		struct queue_limits lim;
173 
174 		sdkp->WCE = wce;
175 		sdkp->RCD = rcd;
176 
177 		lim = queue_limits_start_update(sdkp->disk->queue);
178 		sd_set_flush_flag(sdkp, &lim);
179 		blk_mq_freeze_queue(sdkp->disk->queue);
180 		ret = queue_limits_commit_update(sdkp->disk->queue, &lim);
181 		blk_mq_unfreeze_queue(sdkp->disk->queue);
182 		if (ret)
183 			return ret;
184 		return count;
185 	}
186 
187 	if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
188 			    sdkp->max_retries, &data, NULL))
189 		return -EINVAL;
190 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
191 		  data.block_descriptor_length);
192 	buffer_data = buffer + data.header_length +
193 		data.block_descriptor_length;
194 	buffer_data[2] &= ~0x05;
195 	buffer_data[2] |= wce << 2 | rcd;
196 	sp = buffer_data[0] & 0x80 ? 1 : 0;
197 	buffer_data[0] &= ~0x80;
198 
199 	/*
200 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
201 	 * received mode parameter buffer before doing MODE SELECT.
202 	 */
203 	data.device_specific = 0;
204 
205 	ret = scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
206 			       sdkp->max_retries, &data, &sshdr);
207 	if (ret) {
208 		if (ret > 0 && scsi_sense_valid(&sshdr))
209 			sd_print_sense_hdr(sdkp, &sshdr);
210 		return -EINVAL;
211 	}
212 	sd_revalidate_disk(sdkp->disk);
213 	return count;
214 }
215 
216 static ssize_t
217 manage_start_stop_show(struct device *dev,
218 		       struct device_attribute *attr, char *buf)
219 {
220 	struct scsi_disk *sdkp = to_scsi_disk(dev);
221 	struct scsi_device *sdp = sdkp->device;
222 
223 	return sysfs_emit(buf, "%u\n",
224 			  sdp->manage_system_start_stop &&
225 			  sdp->manage_runtime_start_stop &&
226 			  sdp->manage_shutdown);
227 }
228 static DEVICE_ATTR_RO(manage_start_stop);
229 
230 static ssize_t
231 manage_system_start_stop_show(struct device *dev,
232 			      struct device_attribute *attr, char *buf)
233 {
234 	struct scsi_disk *sdkp = to_scsi_disk(dev);
235 	struct scsi_device *sdp = sdkp->device;
236 
237 	return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
238 }
239 
240 static ssize_t
241 manage_system_start_stop_store(struct device *dev,
242 			       struct device_attribute *attr,
243 			       const char *buf, size_t count)
244 {
245 	struct scsi_disk *sdkp = to_scsi_disk(dev);
246 	struct scsi_device *sdp = sdkp->device;
247 	bool v;
248 
249 	if (!capable(CAP_SYS_ADMIN))
250 		return -EACCES;
251 
252 	if (kstrtobool(buf, &v))
253 		return -EINVAL;
254 
255 	sdp->manage_system_start_stop = v;
256 
257 	return count;
258 }
259 static DEVICE_ATTR_RW(manage_system_start_stop);
260 
261 static ssize_t
262 manage_runtime_start_stop_show(struct device *dev,
263 			       struct device_attribute *attr, char *buf)
264 {
265 	struct scsi_disk *sdkp = to_scsi_disk(dev);
266 	struct scsi_device *sdp = sdkp->device;
267 
268 	return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
269 }
270 
271 static ssize_t
272 manage_runtime_start_stop_store(struct device *dev,
273 				struct device_attribute *attr,
274 				const char *buf, size_t count)
275 {
276 	struct scsi_disk *sdkp = to_scsi_disk(dev);
277 	struct scsi_device *sdp = sdkp->device;
278 	bool v;
279 
280 	if (!capable(CAP_SYS_ADMIN))
281 		return -EACCES;
282 
283 	if (kstrtobool(buf, &v))
284 		return -EINVAL;
285 
286 	sdp->manage_runtime_start_stop = v;
287 
288 	return count;
289 }
290 static DEVICE_ATTR_RW(manage_runtime_start_stop);
291 
292 static ssize_t manage_shutdown_show(struct device *dev,
293 				    struct device_attribute *attr, char *buf)
294 {
295 	struct scsi_disk *sdkp = to_scsi_disk(dev);
296 	struct scsi_device *sdp = sdkp->device;
297 
298 	return sysfs_emit(buf, "%u\n", sdp->manage_shutdown);
299 }
300 
301 static ssize_t manage_shutdown_store(struct device *dev,
302 				     struct device_attribute *attr,
303 				     const char *buf, size_t count)
304 {
305 	struct scsi_disk *sdkp = to_scsi_disk(dev);
306 	struct scsi_device *sdp = sdkp->device;
307 	bool v;
308 
309 	if (!capable(CAP_SYS_ADMIN))
310 		return -EACCES;
311 
312 	if (kstrtobool(buf, &v))
313 		return -EINVAL;
314 
315 	sdp->manage_shutdown = v;
316 
317 	return count;
318 }
319 static DEVICE_ATTR_RW(manage_shutdown);
320 
321 static ssize_t
322 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
323 {
324 	struct scsi_disk *sdkp = to_scsi_disk(dev);
325 
326 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
327 }
328 
329 static ssize_t
330 allow_restart_store(struct device *dev, struct device_attribute *attr,
331 		    const char *buf, size_t count)
332 {
333 	bool v;
334 	struct scsi_disk *sdkp = to_scsi_disk(dev);
335 	struct scsi_device *sdp = sdkp->device;
336 
337 	if (!capable(CAP_SYS_ADMIN))
338 		return -EACCES;
339 
340 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
341 		return -EINVAL;
342 
343 	if (kstrtobool(buf, &v))
344 		return -EINVAL;
345 
346 	sdp->allow_restart = v;
347 
348 	return count;
349 }
350 static DEVICE_ATTR_RW(allow_restart);
351 
352 static ssize_t
353 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
354 {
355 	struct scsi_disk *sdkp = to_scsi_disk(dev);
356 	int ct = sdkp->RCD + 2*sdkp->WCE;
357 
358 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
359 }
360 static DEVICE_ATTR_RW(cache_type);
361 
362 static ssize_t
363 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
364 {
365 	struct scsi_disk *sdkp = to_scsi_disk(dev);
366 
367 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
368 }
369 static DEVICE_ATTR_RO(FUA);
370 
371 static ssize_t
372 protection_type_show(struct device *dev, struct device_attribute *attr,
373 		     char *buf)
374 {
375 	struct scsi_disk *sdkp = to_scsi_disk(dev);
376 
377 	return sprintf(buf, "%u\n", sdkp->protection_type);
378 }
379 
380 static ssize_t
381 protection_type_store(struct device *dev, struct device_attribute *attr,
382 		      const char *buf, size_t count)
383 {
384 	struct scsi_disk *sdkp = to_scsi_disk(dev);
385 	unsigned int val;
386 	int err;
387 
388 	if (!capable(CAP_SYS_ADMIN))
389 		return -EACCES;
390 
391 	err = kstrtouint(buf, 10, &val);
392 
393 	if (err)
394 		return err;
395 
396 	if (val <= T10_PI_TYPE3_PROTECTION)
397 		sdkp->protection_type = val;
398 
399 	return count;
400 }
401 static DEVICE_ATTR_RW(protection_type);
402 
403 static ssize_t
404 protection_mode_show(struct device *dev, struct device_attribute *attr,
405 		     char *buf)
406 {
407 	struct scsi_disk *sdkp = to_scsi_disk(dev);
408 	struct scsi_device *sdp = sdkp->device;
409 	unsigned int dif, dix;
410 
411 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
412 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
413 
414 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
415 		dif = 0;
416 		dix = 1;
417 	}
418 
419 	if (!dif && !dix)
420 		return sprintf(buf, "none\n");
421 
422 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
423 }
424 static DEVICE_ATTR_RO(protection_mode);
425 
426 static ssize_t
427 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
428 {
429 	struct scsi_disk *sdkp = to_scsi_disk(dev);
430 
431 	return sprintf(buf, "%u\n", sdkp->ATO);
432 }
433 static DEVICE_ATTR_RO(app_tag_own);
434 
435 static ssize_t
436 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
437 		       char *buf)
438 {
439 	struct scsi_disk *sdkp = to_scsi_disk(dev);
440 
441 	return sprintf(buf, "%u\n", sdkp->lbpme);
442 }
443 static DEVICE_ATTR_RO(thin_provisioning);
444 
445 /* sysfs_match_string() requires dense arrays */
446 static const char *lbp_mode[] = {
447 	[SD_LBP_FULL]		= "full",
448 	[SD_LBP_UNMAP]		= "unmap",
449 	[SD_LBP_WS16]		= "writesame_16",
450 	[SD_LBP_WS10]		= "writesame_10",
451 	[SD_LBP_ZERO]		= "writesame_zero",
452 	[SD_LBP_DISABLE]	= "disabled",
453 };
454 
455 static ssize_t
456 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
457 		       char *buf)
458 {
459 	struct scsi_disk *sdkp = to_scsi_disk(dev);
460 
461 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
462 }
463 
464 static ssize_t
465 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
466 			const char *buf, size_t count)
467 {
468 	struct scsi_disk *sdkp = to_scsi_disk(dev);
469 	struct scsi_device *sdp = sdkp->device;
470 	struct queue_limits lim;
471 	int mode, err;
472 
473 	if (!capable(CAP_SYS_ADMIN))
474 		return -EACCES;
475 
476 	if (sdp->type != TYPE_DISK)
477 		return -EINVAL;
478 
479 	mode = sysfs_match_string(lbp_mode, buf);
480 	if (mode < 0)
481 		return -EINVAL;
482 
483 	lim = queue_limits_start_update(sdkp->disk->queue);
484 	sd_config_discard(sdkp, &lim, mode);
485 	blk_mq_freeze_queue(sdkp->disk->queue);
486 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
487 	blk_mq_unfreeze_queue(sdkp->disk->queue);
488 	if (err)
489 		return err;
490 	return count;
491 }
492 static DEVICE_ATTR_RW(provisioning_mode);
493 
494 /* sysfs_match_string() requires dense arrays */
495 static const char *zeroing_mode[] = {
496 	[SD_ZERO_WRITE]		= "write",
497 	[SD_ZERO_WS]		= "writesame",
498 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
499 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
500 };
501 
502 static ssize_t
503 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
504 		  char *buf)
505 {
506 	struct scsi_disk *sdkp = to_scsi_disk(dev);
507 
508 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
509 }
510 
511 static ssize_t
512 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
513 		   const char *buf, size_t count)
514 {
515 	struct scsi_disk *sdkp = to_scsi_disk(dev);
516 	int mode;
517 
518 	if (!capable(CAP_SYS_ADMIN))
519 		return -EACCES;
520 
521 	mode = sysfs_match_string(zeroing_mode, buf);
522 	if (mode < 0)
523 		return -EINVAL;
524 
525 	sdkp->zeroing_mode = mode;
526 
527 	return count;
528 }
529 static DEVICE_ATTR_RW(zeroing_mode);
530 
531 static ssize_t
532 max_medium_access_timeouts_show(struct device *dev,
533 				struct device_attribute *attr, char *buf)
534 {
535 	struct scsi_disk *sdkp = to_scsi_disk(dev);
536 
537 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
538 }
539 
540 static ssize_t
541 max_medium_access_timeouts_store(struct device *dev,
542 				 struct device_attribute *attr, const char *buf,
543 				 size_t count)
544 {
545 	struct scsi_disk *sdkp = to_scsi_disk(dev);
546 	int err;
547 
548 	if (!capable(CAP_SYS_ADMIN))
549 		return -EACCES;
550 
551 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
552 
553 	return err ? err : count;
554 }
555 static DEVICE_ATTR_RW(max_medium_access_timeouts);
556 
557 static ssize_t
558 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
559 			   char *buf)
560 {
561 	struct scsi_disk *sdkp = to_scsi_disk(dev);
562 
563 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
564 }
565 
566 static ssize_t
567 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
568 			    const char *buf, size_t count)
569 {
570 	struct scsi_disk *sdkp = to_scsi_disk(dev);
571 	struct scsi_device *sdp = sdkp->device;
572 	struct queue_limits lim;
573 	unsigned long max;
574 	int err;
575 
576 	if (!capable(CAP_SYS_ADMIN))
577 		return -EACCES;
578 
579 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
580 		return -EINVAL;
581 
582 	err = kstrtoul(buf, 10, &max);
583 
584 	if (err)
585 		return err;
586 
587 	if (max == 0)
588 		sdp->no_write_same = 1;
589 	else if (max <= SD_MAX_WS16_BLOCKS) {
590 		sdp->no_write_same = 0;
591 		sdkp->max_ws_blocks = max;
592 	}
593 
594 	lim = queue_limits_start_update(sdkp->disk->queue);
595 	sd_config_write_same(sdkp, &lim);
596 	blk_mq_freeze_queue(sdkp->disk->queue);
597 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
598 	blk_mq_unfreeze_queue(sdkp->disk->queue);
599 	if (err)
600 		return err;
601 	return count;
602 }
603 static DEVICE_ATTR_RW(max_write_same_blocks);
604 
605 static ssize_t
606 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
607 {
608 	struct scsi_disk *sdkp = to_scsi_disk(dev);
609 
610 	if (sdkp->device->type == TYPE_ZBC)
611 		return sprintf(buf, "host-managed\n");
612 	if (sdkp->zoned == 1)
613 		return sprintf(buf, "host-aware\n");
614 	if (sdkp->zoned == 2)
615 		return sprintf(buf, "drive-managed\n");
616 	return sprintf(buf, "none\n");
617 }
618 static DEVICE_ATTR_RO(zoned_cap);
619 
620 static ssize_t
621 max_retries_store(struct device *dev, struct device_attribute *attr,
622 		  const char *buf, size_t count)
623 {
624 	struct scsi_disk *sdkp = to_scsi_disk(dev);
625 	struct scsi_device *sdev = sdkp->device;
626 	int retries, err;
627 
628 	err = kstrtoint(buf, 10, &retries);
629 	if (err)
630 		return err;
631 
632 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
633 		sdkp->max_retries = retries;
634 		return count;
635 	}
636 
637 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
638 		    SD_MAX_RETRIES);
639 	return -EINVAL;
640 }
641 
642 static ssize_t
643 max_retries_show(struct device *dev, struct device_attribute *attr,
644 		 char *buf)
645 {
646 	struct scsi_disk *sdkp = to_scsi_disk(dev);
647 
648 	return sprintf(buf, "%d\n", sdkp->max_retries);
649 }
650 
651 static DEVICE_ATTR_RW(max_retries);
652 
653 static struct attribute *sd_disk_attrs[] = {
654 	&dev_attr_cache_type.attr,
655 	&dev_attr_FUA.attr,
656 	&dev_attr_allow_restart.attr,
657 	&dev_attr_manage_start_stop.attr,
658 	&dev_attr_manage_system_start_stop.attr,
659 	&dev_attr_manage_runtime_start_stop.attr,
660 	&dev_attr_manage_shutdown.attr,
661 	&dev_attr_protection_type.attr,
662 	&dev_attr_protection_mode.attr,
663 	&dev_attr_app_tag_own.attr,
664 	&dev_attr_thin_provisioning.attr,
665 	&dev_attr_provisioning_mode.attr,
666 	&dev_attr_zeroing_mode.attr,
667 	&dev_attr_max_write_same_blocks.attr,
668 	&dev_attr_max_medium_access_timeouts.attr,
669 	&dev_attr_zoned_cap.attr,
670 	&dev_attr_max_retries.attr,
671 	NULL,
672 };
673 ATTRIBUTE_GROUPS(sd_disk);
674 
675 static struct class sd_disk_class = {
676 	.name		= "scsi_disk",
677 	.dev_release	= scsi_disk_release,
678 	.dev_groups	= sd_disk_groups,
679 };
680 
681 /*
682  * Don't request a new module, as that could deadlock in multipath
683  * environment.
684  */
685 static void sd_default_probe(dev_t devt)
686 {
687 }
688 
689 /*
690  * Device no to disk mapping:
691  *
692  *       major         disc2     disc  p1
693  *   |............|.............|....|....| <- dev_t
694  *    31        20 19          8 7  4 3  0
695  *
696  * Inside a major, we have 16k disks, however mapped non-
697  * contiguously. The first 16 disks are for major0, the next
698  * ones with major1, ... Disk 256 is for major0 again, disk 272
699  * for major1, ...
700  * As we stay compatible with our numbering scheme, we can reuse
701  * the well-know SCSI majors 8, 65--71, 136--143.
702  */
703 static int sd_major(int major_idx)
704 {
705 	switch (major_idx) {
706 	case 0:
707 		return SCSI_DISK0_MAJOR;
708 	case 1 ... 7:
709 		return SCSI_DISK1_MAJOR + major_idx - 1;
710 	case 8 ... 15:
711 		return SCSI_DISK8_MAJOR + major_idx - 8;
712 	default:
713 		BUG();
714 		return 0;	/* shut up gcc */
715 	}
716 }
717 
718 #ifdef CONFIG_BLK_SED_OPAL
719 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
720 		size_t len, bool send)
721 {
722 	struct scsi_disk *sdkp = data;
723 	struct scsi_device *sdev = sdkp->device;
724 	u8 cdb[12] = { 0, };
725 	const struct scsi_exec_args exec_args = {
726 		.req_flags = BLK_MQ_REQ_PM,
727 	};
728 	int ret;
729 
730 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
731 	cdb[1] = secp;
732 	put_unaligned_be16(spsp, &cdb[2]);
733 	put_unaligned_be32(len, &cdb[6]);
734 
735 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
736 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
737 			       &exec_args);
738 	return ret <= 0 ? ret : -EIO;
739 }
740 #endif /* CONFIG_BLK_SED_OPAL */
741 
742 /*
743  * Look up the DIX operation based on whether the command is read or
744  * write and whether dix and dif are enabled.
745  */
746 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
747 {
748 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
749 	static const unsigned int ops[] = {	/* wrt dix dif */
750 		SCSI_PROT_NORMAL,		/*  0	0   0  */
751 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
752 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
753 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
754 		SCSI_PROT_NORMAL,		/*  1	0   0  */
755 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
756 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
757 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
758 	};
759 
760 	return ops[write << 2 | dix << 1 | dif];
761 }
762 
763 /*
764  * Returns a mask of the protection flags that are valid for a given DIX
765  * operation.
766  */
767 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
768 {
769 	static const unsigned int flag_mask[] = {
770 		[SCSI_PROT_NORMAL]		= 0,
771 
772 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
773 						  SCSI_PROT_GUARD_CHECK |
774 						  SCSI_PROT_REF_CHECK |
775 						  SCSI_PROT_REF_INCREMENT,
776 
777 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
778 						  SCSI_PROT_IP_CHECKSUM,
779 
780 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
781 						  SCSI_PROT_GUARD_CHECK |
782 						  SCSI_PROT_REF_CHECK |
783 						  SCSI_PROT_REF_INCREMENT |
784 						  SCSI_PROT_IP_CHECKSUM,
785 
786 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
787 						  SCSI_PROT_REF_INCREMENT,
788 
789 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
790 						  SCSI_PROT_REF_CHECK |
791 						  SCSI_PROT_REF_INCREMENT |
792 						  SCSI_PROT_IP_CHECKSUM,
793 
794 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
795 						  SCSI_PROT_GUARD_CHECK |
796 						  SCSI_PROT_REF_CHECK |
797 						  SCSI_PROT_REF_INCREMENT |
798 						  SCSI_PROT_IP_CHECKSUM,
799 	};
800 
801 	return flag_mask[prot_op];
802 }
803 
804 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
805 					   unsigned int dix, unsigned int dif)
806 {
807 	struct request *rq = scsi_cmd_to_rq(scmd);
808 	struct bio *bio = rq->bio;
809 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
810 	unsigned int protect = 0;
811 
812 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
813 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
814 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
815 
816 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
817 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
818 	}
819 
820 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
821 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
822 
823 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
824 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
825 	}
826 
827 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
828 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
829 
830 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
831 			protect = 3 << 5;	/* Disable target PI checking */
832 		else
833 			protect = 1 << 5;	/* Enable target PI checking */
834 	}
835 
836 	scsi_set_prot_op(scmd, prot_op);
837 	scsi_set_prot_type(scmd, dif);
838 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
839 
840 	return protect;
841 }
842 
843 static void sd_disable_discard(struct scsi_disk *sdkp)
844 {
845 	sdkp->provisioning_mode = SD_LBP_DISABLE;
846 	blk_queue_disable_discard(sdkp->disk->queue);
847 }
848 
849 static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
850 		unsigned int mode)
851 {
852 	unsigned int logical_block_size = sdkp->device->sector_size;
853 	unsigned int max_blocks = 0;
854 
855 	lim->discard_alignment = sdkp->unmap_alignment * logical_block_size;
856 	lim->discard_granularity = max(sdkp->physical_block_size,
857 			sdkp->unmap_granularity * logical_block_size);
858 	sdkp->provisioning_mode = mode;
859 
860 	switch (mode) {
861 
862 	case SD_LBP_FULL:
863 	case SD_LBP_DISABLE:
864 		break;
865 
866 	case SD_LBP_UNMAP:
867 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
868 					  (u32)SD_MAX_WS16_BLOCKS);
869 		break;
870 
871 	case SD_LBP_WS16:
872 		if (sdkp->device->unmap_limit_for_ws)
873 			max_blocks = sdkp->max_unmap_blocks;
874 		else
875 			max_blocks = sdkp->max_ws_blocks;
876 
877 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
878 		break;
879 
880 	case SD_LBP_WS10:
881 		if (sdkp->device->unmap_limit_for_ws)
882 			max_blocks = sdkp->max_unmap_blocks;
883 		else
884 			max_blocks = sdkp->max_ws_blocks;
885 
886 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
887 		break;
888 
889 	case SD_LBP_ZERO:
890 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
891 					  (u32)SD_MAX_WS10_BLOCKS);
892 		break;
893 	}
894 
895 	lim->max_hw_discard_sectors = max_blocks *
896 		(logical_block_size >> SECTOR_SHIFT);
897 }
898 
899 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
900 {
901 	struct page *page;
902 
903 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
904 	if (!page)
905 		return NULL;
906 	clear_highpage(page);
907 	bvec_set_page(&rq->special_vec, page, data_len, 0);
908 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
909 	return bvec_virt(&rq->special_vec);
910 }
911 
912 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
913 {
914 	struct scsi_device *sdp = cmd->device;
915 	struct request *rq = scsi_cmd_to_rq(cmd);
916 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
917 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
918 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
919 	unsigned int data_len = 24;
920 	char *buf;
921 
922 	buf = sd_set_special_bvec(rq, data_len);
923 	if (!buf)
924 		return BLK_STS_RESOURCE;
925 
926 	cmd->cmd_len = 10;
927 	cmd->cmnd[0] = UNMAP;
928 	cmd->cmnd[8] = 24;
929 
930 	put_unaligned_be16(6 + 16, &buf[0]);
931 	put_unaligned_be16(16, &buf[2]);
932 	put_unaligned_be64(lba, &buf[8]);
933 	put_unaligned_be32(nr_blocks, &buf[16]);
934 
935 	cmd->allowed = sdkp->max_retries;
936 	cmd->transfersize = data_len;
937 	rq->timeout = SD_TIMEOUT;
938 
939 	return scsi_alloc_sgtables(cmd);
940 }
941 
942 static void sd_config_atomic(struct scsi_disk *sdkp, struct queue_limits *lim)
943 {
944 	unsigned int logical_block_size = sdkp->device->sector_size,
945 		physical_block_size_sectors, max_atomic, unit_min, unit_max;
946 
947 	if ((!sdkp->max_atomic && !sdkp->max_atomic_with_boundary) ||
948 	    sdkp->protection_type == T10_PI_TYPE2_PROTECTION)
949 		return;
950 
951 	physical_block_size_sectors = sdkp->physical_block_size /
952 					sdkp->device->sector_size;
953 
954 	unit_min = rounddown_pow_of_two(sdkp->atomic_granularity ?
955 					sdkp->atomic_granularity :
956 					physical_block_size_sectors);
957 
958 	/*
959 	 * Only use atomic boundary when we have the odd scenario of
960 	 * sdkp->max_atomic == 0, which the spec does permit.
961 	 */
962 	if (sdkp->max_atomic) {
963 		max_atomic = sdkp->max_atomic;
964 		unit_max = rounddown_pow_of_two(sdkp->max_atomic);
965 		sdkp->use_atomic_write_boundary = 0;
966 	} else {
967 		max_atomic = sdkp->max_atomic_with_boundary;
968 		unit_max = rounddown_pow_of_two(sdkp->max_atomic_boundary);
969 		sdkp->use_atomic_write_boundary = 1;
970 	}
971 
972 	/*
973 	 * Ensure compliance with granularity and alignment. For now, keep it
974 	 * simple and just don't support atomic writes for values mismatched
975 	 * with max_{boundary}atomic, physical block size, and
976 	 * atomic_granularity itself.
977 	 *
978 	 * We're really being distrustful by checking unit_max also...
979 	 */
980 	if (sdkp->atomic_granularity > 1) {
981 		if (unit_min > 1 && unit_min % sdkp->atomic_granularity)
982 			return;
983 		if (unit_max > 1 && unit_max % sdkp->atomic_granularity)
984 			return;
985 	}
986 
987 	if (sdkp->atomic_alignment > 1) {
988 		if (unit_min > 1 && unit_min % sdkp->atomic_alignment)
989 			return;
990 		if (unit_max > 1 && unit_max % sdkp->atomic_alignment)
991 			return;
992 	}
993 
994 	lim->atomic_write_hw_max = max_atomic * logical_block_size;
995 	lim->atomic_write_hw_boundary = 0;
996 	lim->atomic_write_hw_unit_min = unit_min * logical_block_size;
997 	lim->atomic_write_hw_unit_max = unit_max * logical_block_size;
998 }
999 
1000 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
1001 		bool unmap)
1002 {
1003 	struct scsi_device *sdp = cmd->device;
1004 	struct request *rq = scsi_cmd_to_rq(cmd);
1005 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1006 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1007 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1008 	u32 data_len = sdp->sector_size;
1009 
1010 	if (!sd_set_special_bvec(rq, data_len))
1011 		return BLK_STS_RESOURCE;
1012 
1013 	cmd->cmd_len = 16;
1014 	cmd->cmnd[0] = WRITE_SAME_16;
1015 	if (unmap)
1016 		cmd->cmnd[1] = 0x8; /* UNMAP */
1017 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1018 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1019 
1020 	cmd->allowed = sdkp->max_retries;
1021 	cmd->transfersize = data_len;
1022 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
1023 
1024 	return scsi_alloc_sgtables(cmd);
1025 }
1026 
1027 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
1028 		bool unmap)
1029 {
1030 	struct scsi_device *sdp = cmd->device;
1031 	struct request *rq = scsi_cmd_to_rq(cmd);
1032 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1033 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1034 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1035 	u32 data_len = sdp->sector_size;
1036 
1037 	if (!sd_set_special_bvec(rq, data_len))
1038 		return BLK_STS_RESOURCE;
1039 
1040 	cmd->cmd_len = 10;
1041 	cmd->cmnd[0] = WRITE_SAME;
1042 	if (unmap)
1043 		cmd->cmnd[1] = 0x8; /* UNMAP */
1044 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1045 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1046 
1047 	cmd->allowed = sdkp->max_retries;
1048 	cmd->transfersize = data_len;
1049 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
1050 
1051 	return scsi_alloc_sgtables(cmd);
1052 }
1053 
1054 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
1055 {
1056 	struct request *rq = scsi_cmd_to_rq(cmd);
1057 	struct scsi_device *sdp = cmd->device;
1058 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1059 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1060 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1061 
1062 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
1063 		switch (sdkp->zeroing_mode) {
1064 		case SD_ZERO_WS16_UNMAP:
1065 			return sd_setup_write_same16_cmnd(cmd, true);
1066 		case SD_ZERO_WS10_UNMAP:
1067 			return sd_setup_write_same10_cmnd(cmd, true);
1068 		}
1069 	}
1070 
1071 	if (sdp->no_write_same) {
1072 		rq->rq_flags |= RQF_QUIET;
1073 		return BLK_STS_TARGET;
1074 	}
1075 
1076 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
1077 		return sd_setup_write_same16_cmnd(cmd, false);
1078 
1079 	return sd_setup_write_same10_cmnd(cmd, false);
1080 }
1081 
1082 static void sd_disable_write_same(struct scsi_disk *sdkp)
1083 {
1084 	sdkp->device->no_write_same = 1;
1085 	sdkp->max_ws_blocks = 0;
1086 	blk_queue_disable_write_zeroes(sdkp->disk->queue);
1087 }
1088 
1089 static void sd_config_write_same(struct scsi_disk *sdkp,
1090 		struct queue_limits *lim)
1091 {
1092 	unsigned int logical_block_size = sdkp->device->sector_size;
1093 
1094 	if (sdkp->device->no_write_same) {
1095 		sdkp->max_ws_blocks = 0;
1096 		goto out;
1097 	}
1098 
1099 	/* Some devices can not handle block counts above 0xffff despite
1100 	 * supporting WRITE SAME(16). Consequently we default to 64k
1101 	 * blocks per I/O unless the device explicitly advertises a
1102 	 * bigger limit.
1103 	 */
1104 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
1105 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1106 						   (u32)SD_MAX_WS16_BLOCKS);
1107 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
1108 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1109 						   (u32)SD_MAX_WS10_BLOCKS);
1110 	else {
1111 		sdkp->device->no_write_same = 1;
1112 		sdkp->max_ws_blocks = 0;
1113 	}
1114 
1115 	if (sdkp->lbprz && sdkp->lbpws)
1116 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
1117 	else if (sdkp->lbprz && sdkp->lbpws10)
1118 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
1119 	else if (sdkp->max_ws_blocks)
1120 		sdkp->zeroing_mode = SD_ZERO_WS;
1121 	else
1122 		sdkp->zeroing_mode = SD_ZERO_WRITE;
1123 
1124 	if (sdkp->max_ws_blocks &&
1125 	    sdkp->physical_block_size > logical_block_size) {
1126 		/*
1127 		 * Reporting a maximum number of blocks that is not aligned
1128 		 * on the device physical size would cause a large write same
1129 		 * request to be split into physically unaligned chunks by
1130 		 * __blkdev_issue_write_zeroes() even if the caller of this
1131 		 * functions took care to align the large request. So make sure
1132 		 * the maximum reported is aligned to the device physical block
1133 		 * size. This is only an optional optimization for regular
1134 		 * disks, but this is mandatory to avoid failure of large write
1135 		 * same requests directed at sequential write required zones of
1136 		 * host-managed ZBC disks.
1137 		 */
1138 		sdkp->max_ws_blocks =
1139 			round_down(sdkp->max_ws_blocks,
1140 				   bytes_to_logical(sdkp->device,
1141 						    sdkp->physical_block_size));
1142 	}
1143 
1144 out:
1145 	lim->max_write_zeroes_sectors =
1146 		sdkp->max_ws_blocks * (logical_block_size >> SECTOR_SHIFT);
1147 }
1148 
1149 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1150 {
1151 	struct request *rq = scsi_cmd_to_rq(cmd);
1152 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1153 
1154 	/* flush requests don't perform I/O, zero the S/G table */
1155 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1156 
1157 	if (cmd->device->use_16_for_sync) {
1158 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1159 		cmd->cmd_len = 16;
1160 	} else {
1161 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1162 		cmd->cmd_len = 10;
1163 	}
1164 	cmd->transfersize = 0;
1165 	cmd->allowed = sdkp->max_retries;
1166 
1167 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1168 	return BLK_STS_OK;
1169 }
1170 
1171 /**
1172  * sd_group_number() - Compute the GROUP NUMBER field
1173  * @cmd: SCSI command for which to compute the value of the six-bit GROUP NUMBER
1174  *	field.
1175  *
1176  * From SBC-5 r05 (https://www.t10.org/cgi-bin/ac.pl?t=f&f=sbc5r05.pdf):
1177  * 0: no relative lifetime.
1178  * 1: shortest relative lifetime.
1179  * 2: second shortest relative lifetime.
1180  * 3 - 0x3d: intermediate relative lifetimes.
1181  * 0x3e: second longest relative lifetime.
1182  * 0x3f: longest relative lifetime.
1183  */
1184 static u8 sd_group_number(struct scsi_cmnd *cmd)
1185 {
1186 	const struct request *rq = scsi_cmd_to_rq(cmd);
1187 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1188 
1189 	if (!sdkp->rscs)
1190 		return 0;
1191 
1192 	return min3((u32)rq->write_hint, (u32)sdkp->permanent_stream_count,
1193 		    0x3fu);
1194 }
1195 
1196 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1197 				       sector_t lba, unsigned int nr_blocks,
1198 				       unsigned char flags, unsigned int dld)
1199 {
1200 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1201 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1202 	cmd->cmnd[6]  = sd_group_number(cmd);
1203 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1204 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1205 	cmd->cmnd[10] = flags;
1206 	cmd->cmnd[11] = dld & 0x07;
1207 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1208 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1209 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1210 
1211 	return BLK_STS_OK;
1212 }
1213 
1214 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1215 				       sector_t lba, unsigned int nr_blocks,
1216 				       unsigned char flags, unsigned int dld)
1217 {
1218 	cmd->cmd_len  = 16;
1219 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1220 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1221 	cmd->cmnd[14] = ((dld & 0x03) << 6) | sd_group_number(cmd);
1222 	cmd->cmnd[15] = 0;
1223 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1224 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1225 
1226 	return BLK_STS_OK;
1227 }
1228 
1229 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1230 				       sector_t lba, unsigned int nr_blocks,
1231 				       unsigned char flags)
1232 {
1233 	cmd->cmd_len = 10;
1234 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1235 	cmd->cmnd[1] = flags;
1236 	cmd->cmnd[6] = sd_group_number(cmd);
1237 	cmd->cmnd[9] = 0;
1238 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1239 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1240 
1241 	return BLK_STS_OK;
1242 }
1243 
1244 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1245 				      sector_t lba, unsigned int nr_blocks,
1246 				      unsigned char flags)
1247 {
1248 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1249 	if (WARN_ON_ONCE(nr_blocks == 0))
1250 		return BLK_STS_IOERR;
1251 
1252 	if (unlikely(flags & 0x8)) {
1253 		/*
1254 		 * This happens only if this drive failed 10byte rw
1255 		 * command with ILLEGAL_REQUEST during operation and
1256 		 * thus turned off use_10_for_rw.
1257 		 */
1258 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1259 		return BLK_STS_IOERR;
1260 	}
1261 
1262 	cmd->cmd_len = 6;
1263 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1264 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1265 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1266 	cmd->cmnd[3] = lba & 0xff;
1267 	cmd->cmnd[4] = nr_blocks;
1268 	cmd->cmnd[5] = 0;
1269 
1270 	return BLK_STS_OK;
1271 }
1272 
1273 /*
1274  * Check if a command has a duration limit set. If it does, and the target
1275  * device supports CDL and the feature is enabled, return the limit
1276  * descriptor index to use. Return 0 (no limit) otherwise.
1277  */
1278 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1279 {
1280 	struct scsi_device *sdp = sdkp->device;
1281 	int hint;
1282 
1283 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1284 		return 0;
1285 
1286 	/*
1287 	 * Use "no limit" if the request ioprio does not specify a duration
1288 	 * limit hint.
1289 	 */
1290 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1291 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1292 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1293 		return 0;
1294 
1295 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1296 }
1297 
1298 static blk_status_t sd_setup_atomic_cmnd(struct scsi_cmnd *cmd,
1299 					sector_t lba, unsigned int nr_blocks,
1300 					bool boundary, unsigned char flags)
1301 {
1302 	cmd->cmd_len  = 16;
1303 	cmd->cmnd[0]  = WRITE_ATOMIC_16;
1304 	cmd->cmnd[1]  = flags;
1305 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1306 	put_unaligned_be16(nr_blocks, &cmd->cmnd[12]);
1307 	if (boundary)
1308 		put_unaligned_be16(nr_blocks, &cmd->cmnd[10]);
1309 	else
1310 		put_unaligned_be16(0, &cmd->cmnd[10]);
1311 	put_unaligned_be16(nr_blocks, &cmd->cmnd[12]);
1312 	cmd->cmnd[14] = 0;
1313 	cmd->cmnd[15] = 0;
1314 
1315 	return BLK_STS_OK;
1316 }
1317 
1318 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1319 {
1320 	struct request *rq = scsi_cmd_to_rq(cmd);
1321 	struct scsi_device *sdp = cmd->device;
1322 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1323 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1324 	sector_t threshold;
1325 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1326 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1327 	bool write = rq_data_dir(rq) == WRITE;
1328 	unsigned char protect, fua;
1329 	unsigned int dld;
1330 	blk_status_t ret;
1331 	unsigned int dif;
1332 	bool dix;
1333 
1334 	ret = scsi_alloc_sgtables(cmd);
1335 	if (ret != BLK_STS_OK)
1336 		return ret;
1337 
1338 	ret = BLK_STS_IOERR;
1339 	if (!scsi_device_online(sdp) || sdp->changed) {
1340 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1341 		goto fail;
1342 	}
1343 
1344 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1345 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1346 		goto fail;
1347 	}
1348 
1349 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1350 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1351 		goto fail;
1352 	}
1353 
1354 	/*
1355 	 * Some SD card readers can't handle accesses which touch the
1356 	 * last one or two logical blocks. Split accesses as needed.
1357 	 */
1358 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1359 
1360 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1361 		if (lba < threshold) {
1362 			/* Access up to the threshold but not beyond */
1363 			nr_blocks = threshold - lba;
1364 		} else {
1365 			/* Access only a single logical block */
1366 			nr_blocks = 1;
1367 		}
1368 	}
1369 
1370 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1371 	dix = scsi_prot_sg_count(cmd);
1372 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1373 	dld = sd_cdl_dld(sdkp, cmd);
1374 
1375 	if (dif || dix)
1376 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1377 	else
1378 		protect = 0;
1379 
1380 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1381 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1382 					 protect | fua, dld);
1383 	} else if (rq->cmd_flags & REQ_ATOMIC && write) {
1384 		ret = sd_setup_atomic_cmnd(cmd, lba, nr_blocks,
1385 				sdkp->use_atomic_write_boundary,
1386 				protect | fua);
1387 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1388 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1389 					 protect | fua, dld);
1390 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1391 		   sdp->use_10_for_rw || protect || rq->write_hint) {
1392 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1393 					 protect | fua);
1394 	} else {
1395 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1396 					protect | fua);
1397 	}
1398 
1399 	if (unlikely(ret != BLK_STS_OK))
1400 		goto fail;
1401 
1402 	/*
1403 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1404 	 * host adapter, it's safe to assume that we can at least transfer
1405 	 * this many bytes between each connect / disconnect.
1406 	 */
1407 	cmd->transfersize = sdp->sector_size;
1408 	cmd->underflow = nr_blocks << 9;
1409 	cmd->allowed = sdkp->max_retries;
1410 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1411 
1412 	SCSI_LOG_HLQUEUE(1,
1413 			 scmd_printk(KERN_INFO, cmd,
1414 				     "%s: block=%llu, count=%d\n", __func__,
1415 				     (unsigned long long)blk_rq_pos(rq),
1416 				     blk_rq_sectors(rq)));
1417 	SCSI_LOG_HLQUEUE(2,
1418 			 scmd_printk(KERN_INFO, cmd,
1419 				     "%s %d/%u 512 byte blocks.\n",
1420 				     write ? "writing" : "reading", nr_blocks,
1421 				     blk_rq_sectors(rq)));
1422 
1423 	/*
1424 	 * This indicates that the command is ready from our end to be queued.
1425 	 */
1426 	return BLK_STS_OK;
1427 fail:
1428 	scsi_free_sgtables(cmd);
1429 	return ret;
1430 }
1431 
1432 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1433 {
1434 	struct request *rq = scsi_cmd_to_rq(cmd);
1435 
1436 	switch (req_op(rq)) {
1437 	case REQ_OP_DISCARD:
1438 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1439 		case SD_LBP_UNMAP:
1440 			return sd_setup_unmap_cmnd(cmd);
1441 		case SD_LBP_WS16:
1442 			return sd_setup_write_same16_cmnd(cmd, true);
1443 		case SD_LBP_WS10:
1444 			return sd_setup_write_same10_cmnd(cmd, true);
1445 		case SD_LBP_ZERO:
1446 			return sd_setup_write_same10_cmnd(cmd, false);
1447 		default:
1448 			return BLK_STS_TARGET;
1449 		}
1450 	case REQ_OP_WRITE_ZEROES:
1451 		return sd_setup_write_zeroes_cmnd(cmd);
1452 	case REQ_OP_FLUSH:
1453 		return sd_setup_flush_cmnd(cmd);
1454 	case REQ_OP_READ:
1455 	case REQ_OP_WRITE:
1456 		return sd_setup_read_write_cmnd(cmd);
1457 	case REQ_OP_ZONE_RESET:
1458 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1459 						   false);
1460 	case REQ_OP_ZONE_RESET_ALL:
1461 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1462 						   true);
1463 	case REQ_OP_ZONE_OPEN:
1464 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1465 	case REQ_OP_ZONE_CLOSE:
1466 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1467 	case REQ_OP_ZONE_FINISH:
1468 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1469 	default:
1470 		WARN_ON_ONCE(1);
1471 		return BLK_STS_NOTSUPP;
1472 	}
1473 }
1474 
1475 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1476 {
1477 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1478 
1479 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1480 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1481 }
1482 
1483 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1484 {
1485 	if (sdkp->device->removable || sdkp->write_prot) {
1486 		if (disk_check_media_change(disk))
1487 			return true;
1488 	}
1489 
1490 	/*
1491 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1492 	 * nothing to do with partitions, BLKRRPART is used to force a full
1493 	 * revalidate after things like a format for historical reasons.
1494 	 */
1495 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1496 }
1497 
1498 /**
1499  *	sd_open - open a scsi disk device
1500  *	@disk: disk to open
1501  *	@mode: open mode
1502  *
1503  *	Returns 0 if successful. Returns a negated errno value in case
1504  *	of error.
1505  *
1506  *	Note: This can be called from a user context (e.g. fsck(1) )
1507  *	or from within the kernel (e.g. as a result of a mount(1) ).
1508  *	In the latter case @inode and @filp carry an abridged amount
1509  *	of information as noted above.
1510  *
1511  *	Locking: called with disk->open_mutex held.
1512  **/
1513 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1514 {
1515 	struct scsi_disk *sdkp = scsi_disk(disk);
1516 	struct scsi_device *sdev = sdkp->device;
1517 	int retval;
1518 
1519 	if (scsi_device_get(sdev))
1520 		return -ENXIO;
1521 
1522 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1523 
1524 	/*
1525 	 * If the device is in error recovery, wait until it is done.
1526 	 * If the device is offline, then disallow any access to it.
1527 	 */
1528 	retval = -ENXIO;
1529 	if (!scsi_block_when_processing_errors(sdev))
1530 		goto error_out;
1531 
1532 	if (sd_need_revalidate(disk, sdkp))
1533 		sd_revalidate_disk(disk);
1534 
1535 	/*
1536 	 * If the drive is empty, just let the open fail.
1537 	 */
1538 	retval = -ENOMEDIUM;
1539 	if (sdev->removable && !sdkp->media_present &&
1540 	    !(mode & BLK_OPEN_NDELAY))
1541 		goto error_out;
1542 
1543 	/*
1544 	 * If the device has the write protect tab set, have the open fail
1545 	 * if the user expects to be able to write to the thing.
1546 	 */
1547 	retval = -EROFS;
1548 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1549 		goto error_out;
1550 
1551 	/*
1552 	 * It is possible that the disk changing stuff resulted in
1553 	 * the device being taken offline.  If this is the case,
1554 	 * report this to the user, and don't pretend that the
1555 	 * open actually succeeded.
1556 	 */
1557 	retval = -ENXIO;
1558 	if (!scsi_device_online(sdev))
1559 		goto error_out;
1560 
1561 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1562 		if (scsi_block_when_processing_errors(sdev))
1563 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1564 	}
1565 
1566 	return 0;
1567 
1568 error_out:
1569 	scsi_device_put(sdev);
1570 	return retval;
1571 }
1572 
1573 /**
1574  *	sd_release - invoked when the (last) close(2) is called on this
1575  *	scsi disk.
1576  *	@disk: disk to release
1577  *
1578  *	Returns 0.
1579  *
1580  *	Note: may block (uninterruptible) if error recovery is underway
1581  *	on this disk.
1582  *
1583  *	Locking: called with disk->open_mutex held.
1584  **/
1585 static void sd_release(struct gendisk *disk)
1586 {
1587 	struct scsi_disk *sdkp = scsi_disk(disk);
1588 	struct scsi_device *sdev = sdkp->device;
1589 
1590 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1591 
1592 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1593 		if (scsi_block_when_processing_errors(sdev))
1594 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1595 	}
1596 
1597 	scsi_device_put(sdev);
1598 }
1599 
1600 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1601 {
1602 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1603 	struct scsi_device *sdp = sdkp->device;
1604 	struct Scsi_Host *host = sdp->host;
1605 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1606 	int diskinfo[4];
1607 
1608 	/* default to most commonly used values */
1609 	diskinfo[0] = 0x40;	/* 1 << 6 */
1610 	diskinfo[1] = 0x20;	/* 1 << 5 */
1611 	diskinfo[2] = capacity >> 11;
1612 
1613 	/* override with calculated, extended default, or driver values */
1614 	if (host->hostt->bios_param)
1615 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1616 	else
1617 		scsicam_bios_param(bdev, capacity, diskinfo);
1618 
1619 	geo->heads = diskinfo[0];
1620 	geo->sectors = diskinfo[1];
1621 	geo->cylinders = diskinfo[2];
1622 	return 0;
1623 }
1624 
1625 /**
1626  *	sd_ioctl - process an ioctl
1627  *	@bdev: target block device
1628  *	@mode: open mode
1629  *	@cmd: ioctl command number
1630  *	@arg: this is third argument given to ioctl(2) system call.
1631  *	Often contains a pointer.
1632  *
1633  *	Returns 0 if successful (some ioctls return positive numbers on
1634  *	success as well). Returns a negated errno value in case of error.
1635  *
1636  *	Note: most ioctls are forward onto the block subsystem or further
1637  *	down in the scsi subsystem.
1638  **/
1639 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1640 		    unsigned int cmd, unsigned long arg)
1641 {
1642 	struct gendisk *disk = bdev->bd_disk;
1643 	struct scsi_disk *sdkp = scsi_disk(disk);
1644 	struct scsi_device *sdp = sdkp->device;
1645 	void __user *p = (void __user *)arg;
1646 	int error;
1647 
1648 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1649 				    "cmd=0x%x\n", disk->disk_name, cmd));
1650 
1651 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1652 		return -ENOIOCTLCMD;
1653 
1654 	/*
1655 	 * If we are in the middle of error recovery, don't let anyone
1656 	 * else try and use this device.  Also, if error recovery fails, it
1657 	 * may try and take the device offline, in which case all further
1658 	 * access to the device is prohibited.
1659 	 */
1660 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1661 			(mode & BLK_OPEN_NDELAY));
1662 	if (error)
1663 		return error;
1664 
1665 	if (is_sed_ioctl(cmd))
1666 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1667 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1668 }
1669 
1670 static void set_media_not_present(struct scsi_disk *sdkp)
1671 {
1672 	if (sdkp->media_present)
1673 		sdkp->device->changed = 1;
1674 
1675 	if (sdkp->device->removable) {
1676 		sdkp->media_present = 0;
1677 		sdkp->capacity = 0;
1678 	}
1679 }
1680 
1681 static int media_not_present(struct scsi_disk *sdkp,
1682 			     struct scsi_sense_hdr *sshdr)
1683 {
1684 	if (!scsi_sense_valid(sshdr))
1685 		return 0;
1686 
1687 	/* not invoked for commands that could return deferred errors */
1688 	switch (sshdr->sense_key) {
1689 	case UNIT_ATTENTION:
1690 	case NOT_READY:
1691 		/* medium not present */
1692 		if (sshdr->asc == 0x3A) {
1693 			set_media_not_present(sdkp);
1694 			return 1;
1695 		}
1696 	}
1697 	return 0;
1698 }
1699 
1700 /**
1701  *	sd_check_events - check media events
1702  *	@disk: kernel device descriptor
1703  *	@clearing: disk events currently being cleared
1704  *
1705  *	Returns mask of DISK_EVENT_*.
1706  *
1707  *	Note: this function is invoked from the block subsystem.
1708  **/
1709 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1710 {
1711 	struct scsi_disk *sdkp = disk->private_data;
1712 	struct scsi_device *sdp;
1713 	int retval;
1714 	bool disk_changed;
1715 
1716 	if (!sdkp)
1717 		return 0;
1718 
1719 	sdp = sdkp->device;
1720 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1721 
1722 	/*
1723 	 * If the device is offline, don't send any commands - just pretend as
1724 	 * if the command failed.  If the device ever comes back online, we
1725 	 * can deal with it then.  It is only because of unrecoverable errors
1726 	 * that we would ever take a device offline in the first place.
1727 	 */
1728 	if (!scsi_device_online(sdp)) {
1729 		set_media_not_present(sdkp);
1730 		goto out;
1731 	}
1732 
1733 	/*
1734 	 * Using TEST_UNIT_READY enables differentiation between drive with
1735 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1736 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1737 	 *
1738 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1739 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1740 	 * sd_revalidate() is called.
1741 	 */
1742 	if (scsi_block_when_processing_errors(sdp)) {
1743 		struct scsi_sense_hdr sshdr = { 0, };
1744 
1745 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1746 					      &sshdr);
1747 
1748 		/* failed to execute TUR, assume media not present */
1749 		if (retval < 0 || host_byte(retval)) {
1750 			set_media_not_present(sdkp);
1751 			goto out;
1752 		}
1753 
1754 		if (media_not_present(sdkp, &sshdr))
1755 			goto out;
1756 	}
1757 
1758 	/*
1759 	 * For removable scsi disk we have to recognise the presence
1760 	 * of a disk in the drive.
1761 	 */
1762 	if (!sdkp->media_present)
1763 		sdp->changed = 1;
1764 	sdkp->media_present = 1;
1765 out:
1766 	/*
1767 	 * sdp->changed is set under the following conditions:
1768 	 *
1769 	 *	Medium present state has changed in either direction.
1770 	 *	Device has indicated UNIT_ATTENTION.
1771 	 */
1772 	disk_changed = sdp->changed;
1773 	sdp->changed = 0;
1774 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1775 }
1776 
1777 static int sd_sync_cache(struct scsi_disk *sdkp)
1778 {
1779 	int res;
1780 	struct scsi_device *sdp = sdkp->device;
1781 	const int timeout = sdp->request_queue->rq_timeout
1782 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1783 	/* Leave the rest of the command zero to indicate flush everything. */
1784 	const unsigned char cmd[16] = { sdp->use_16_for_sync ?
1785 				SYNCHRONIZE_CACHE_16 : SYNCHRONIZE_CACHE };
1786 	struct scsi_sense_hdr sshdr;
1787 	struct scsi_failure failure_defs[] = {
1788 		{
1789 			.allowed = 3,
1790 			.result = SCMD_FAILURE_RESULT_ANY,
1791 		},
1792 		{}
1793 	};
1794 	struct scsi_failures failures = {
1795 		.failure_definitions = failure_defs,
1796 	};
1797 	const struct scsi_exec_args exec_args = {
1798 		.req_flags = BLK_MQ_REQ_PM,
1799 		.sshdr = &sshdr,
1800 		.failures = &failures,
1801 	};
1802 
1803 	if (!scsi_device_online(sdp))
1804 		return -ENODEV;
1805 
1806 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, timeout,
1807 			       sdkp->max_retries, &exec_args);
1808 	if (res) {
1809 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1810 
1811 		if (res < 0)
1812 			return res;
1813 
1814 		if (scsi_status_is_check_condition(res) &&
1815 		    scsi_sense_valid(&sshdr)) {
1816 			sd_print_sense_hdr(sdkp, &sshdr);
1817 
1818 			/* we need to evaluate the error return  */
1819 			if (sshdr.asc == 0x3a ||	/* medium not present */
1820 			    sshdr.asc == 0x20 ||	/* invalid command */
1821 			    (sshdr.asc == 0x74 && sshdr.ascq == 0x71))	/* drive is password locked */
1822 				/* this is no error here */
1823 				return 0;
1824 			/*
1825 			 * This drive doesn't support sync and there's not much
1826 			 * we can do because this is called during shutdown
1827 			 * or suspend so just return success so those operations
1828 			 * can proceed.
1829 			 */
1830 			if (sshdr.sense_key == ILLEGAL_REQUEST)
1831 				return 0;
1832 		}
1833 
1834 		switch (host_byte(res)) {
1835 		/* ignore errors due to racing a disconnection */
1836 		case DID_BAD_TARGET:
1837 		case DID_NO_CONNECT:
1838 			return 0;
1839 		/* signal the upper layer it might try again */
1840 		case DID_BUS_BUSY:
1841 		case DID_IMM_RETRY:
1842 		case DID_REQUEUE:
1843 		case DID_SOFT_ERROR:
1844 			return -EBUSY;
1845 		default:
1846 			return -EIO;
1847 		}
1848 	}
1849 	return 0;
1850 }
1851 
1852 static void sd_rescan(struct device *dev)
1853 {
1854 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1855 
1856 	sd_revalidate_disk(sdkp->disk);
1857 }
1858 
1859 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1860 		enum blk_unique_id type)
1861 {
1862 	struct scsi_device *sdev = scsi_disk(disk)->device;
1863 	const struct scsi_vpd *vpd;
1864 	const unsigned char *d;
1865 	int ret = -ENXIO, len;
1866 
1867 	rcu_read_lock();
1868 	vpd = rcu_dereference(sdev->vpd_pg83);
1869 	if (!vpd)
1870 		goto out_unlock;
1871 
1872 	ret = -EINVAL;
1873 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1874 		/* we only care about designators with LU association */
1875 		if (((d[1] >> 4) & 0x3) != 0x00)
1876 			continue;
1877 		if ((d[1] & 0xf) != type)
1878 			continue;
1879 
1880 		/*
1881 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1882 		 * keep looking as one with more entropy might still show up.
1883 		 */
1884 		len = d[3];
1885 		if (len != 8 && len != 12 && len != 16)
1886 			continue;
1887 		ret = len;
1888 		memcpy(id, d + 4, len);
1889 		if (len == 16)
1890 			break;
1891 	}
1892 out_unlock:
1893 	rcu_read_unlock();
1894 	return ret;
1895 }
1896 
1897 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1898 {
1899 	switch (host_byte(result)) {
1900 	case DID_TRANSPORT_MARGINAL:
1901 	case DID_TRANSPORT_DISRUPTED:
1902 	case DID_BUS_BUSY:
1903 		return PR_STS_RETRY_PATH_FAILURE;
1904 	case DID_NO_CONNECT:
1905 		return PR_STS_PATH_FAILED;
1906 	case DID_TRANSPORT_FAILFAST:
1907 		return PR_STS_PATH_FAST_FAILED;
1908 	}
1909 
1910 	switch (status_byte(result)) {
1911 	case SAM_STAT_RESERVATION_CONFLICT:
1912 		return PR_STS_RESERVATION_CONFLICT;
1913 	case SAM_STAT_CHECK_CONDITION:
1914 		if (!scsi_sense_valid(sshdr))
1915 			return PR_STS_IOERR;
1916 
1917 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1918 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1919 			return -EINVAL;
1920 
1921 		fallthrough;
1922 	default:
1923 		return PR_STS_IOERR;
1924 	}
1925 }
1926 
1927 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1928 			    unsigned char *data, int data_len)
1929 {
1930 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1931 	struct scsi_device *sdev = sdkp->device;
1932 	struct scsi_sense_hdr sshdr;
1933 	u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1934 	struct scsi_failure failure_defs[] = {
1935 		{
1936 			.sense = UNIT_ATTENTION,
1937 			.asc = SCMD_FAILURE_ASC_ANY,
1938 			.ascq = SCMD_FAILURE_ASCQ_ANY,
1939 			.allowed = 5,
1940 			.result = SAM_STAT_CHECK_CONDITION,
1941 		},
1942 		{}
1943 	};
1944 	struct scsi_failures failures = {
1945 		.failure_definitions = failure_defs,
1946 	};
1947 	const struct scsi_exec_args exec_args = {
1948 		.sshdr = &sshdr,
1949 		.failures = &failures,
1950 	};
1951 	int result;
1952 
1953 	put_unaligned_be16(data_len, &cmd[7]);
1954 
1955 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1956 				  SD_TIMEOUT, sdkp->max_retries, &exec_args);
1957 	if (scsi_status_is_check_condition(result) &&
1958 	    scsi_sense_valid(&sshdr)) {
1959 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1960 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1961 	}
1962 
1963 	if (result <= 0)
1964 		return result;
1965 
1966 	return sd_scsi_to_pr_err(&sshdr, result);
1967 }
1968 
1969 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1970 {
1971 	int result, i, data_offset, num_copy_keys;
1972 	u32 num_keys = keys_info->num_keys;
1973 	int data_len = num_keys * 8 + 8;
1974 	u8 *data;
1975 
1976 	data = kzalloc(data_len, GFP_KERNEL);
1977 	if (!data)
1978 		return -ENOMEM;
1979 
1980 	result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1981 	if (result)
1982 		goto free_data;
1983 
1984 	keys_info->generation = get_unaligned_be32(&data[0]);
1985 	keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1986 
1987 	data_offset = 8;
1988 	num_copy_keys = min(num_keys, keys_info->num_keys);
1989 
1990 	for (i = 0; i < num_copy_keys; i++) {
1991 		keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1992 		data_offset += 8;
1993 	}
1994 
1995 free_data:
1996 	kfree(data);
1997 	return result;
1998 }
1999 
2000 static int sd_pr_read_reservation(struct block_device *bdev,
2001 				  struct pr_held_reservation *rsv)
2002 {
2003 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
2004 	struct scsi_device *sdev = sdkp->device;
2005 	u8 data[24] = { };
2006 	int result, len;
2007 
2008 	result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
2009 	if (result)
2010 		return result;
2011 
2012 	len = get_unaligned_be32(&data[4]);
2013 	if (!len)
2014 		return 0;
2015 
2016 	/* Make sure we have at least the key and type */
2017 	if (len < 14) {
2018 		sdev_printk(KERN_INFO, sdev,
2019 			    "READ RESERVATION failed due to short return buffer of %d bytes\n",
2020 			    len);
2021 		return -EINVAL;
2022 	}
2023 
2024 	rsv->generation = get_unaligned_be32(&data[0]);
2025 	rsv->key = get_unaligned_be64(&data[8]);
2026 	rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
2027 	return 0;
2028 }
2029 
2030 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
2031 			     u64 sa_key, enum scsi_pr_type type, u8 flags)
2032 {
2033 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
2034 	struct scsi_device *sdev = sdkp->device;
2035 	struct scsi_sense_hdr sshdr;
2036 	struct scsi_failure failure_defs[] = {
2037 		{
2038 			.sense = UNIT_ATTENTION,
2039 			.asc = SCMD_FAILURE_ASC_ANY,
2040 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2041 			.allowed = 5,
2042 			.result = SAM_STAT_CHECK_CONDITION,
2043 		},
2044 		{}
2045 	};
2046 	struct scsi_failures failures = {
2047 		.failure_definitions = failure_defs,
2048 	};
2049 	const struct scsi_exec_args exec_args = {
2050 		.sshdr = &sshdr,
2051 		.failures = &failures,
2052 	};
2053 	int result;
2054 	u8 cmd[16] = { 0, };
2055 	u8 data[24] = { 0, };
2056 
2057 	cmd[0] = PERSISTENT_RESERVE_OUT;
2058 	cmd[1] = sa;
2059 	cmd[2] = type;
2060 	put_unaligned_be32(sizeof(data), &cmd[5]);
2061 
2062 	put_unaligned_be64(key, &data[0]);
2063 	put_unaligned_be64(sa_key, &data[8]);
2064 	data[20] = flags;
2065 
2066 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
2067 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
2068 				  &exec_args);
2069 
2070 	if (scsi_status_is_check_condition(result) &&
2071 	    scsi_sense_valid(&sshdr)) {
2072 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
2073 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
2074 	}
2075 
2076 	if (result <= 0)
2077 		return result;
2078 
2079 	return sd_scsi_to_pr_err(&sshdr, result);
2080 }
2081 
2082 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
2083 		u32 flags)
2084 {
2085 	if (flags & ~PR_FL_IGNORE_KEY)
2086 		return -EOPNOTSUPP;
2087 	return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
2088 			old_key, new_key, 0,
2089 			(1 << 0) /* APTPL */);
2090 }
2091 
2092 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2093 		u32 flags)
2094 {
2095 	if (flags)
2096 		return -EOPNOTSUPP;
2097 	return sd_pr_out_command(bdev, 0x01, key, 0,
2098 				 block_pr_type_to_scsi(type), 0);
2099 }
2100 
2101 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
2102 {
2103 	return sd_pr_out_command(bdev, 0x02, key, 0,
2104 				 block_pr_type_to_scsi(type), 0);
2105 }
2106 
2107 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2108 		enum pr_type type, bool abort)
2109 {
2110 	return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
2111 				 block_pr_type_to_scsi(type), 0);
2112 }
2113 
2114 static int sd_pr_clear(struct block_device *bdev, u64 key)
2115 {
2116 	return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
2117 }
2118 
2119 static const struct pr_ops sd_pr_ops = {
2120 	.pr_register	= sd_pr_register,
2121 	.pr_reserve	= sd_pr_reserve,
2122 	.pr_release	= sd_pr_release,
2123 	.pr_preempt	= sd_pr_preempt,
2124 	.pr_clear	= sd_pr_clear,
2125 	.pr_read_keys	= sd_pr_read_keys,
2126 	.pr_read_reservation = sd_pr_read_reservation,
2127 };
2128 
2129 static void scsi_disk_free_disk(struct gendisk *disk)
2130 {
2131 	struct scsi_disk *sdkp = scsi_disk(disk);
2132 
2133 	put_device(&sdkp->disk_dev);
2134 }
2135 
2136 static const struct block_device_operations sd_fops = {
2137 	.owner			= THIS_MODULE,
2138 	.open			= sd_open,
2139 	.release		= sd_release,
2140 	.ioctl			= sd_ioctl,
2141 	.getgeo			= sd_getgeo,
2142 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
2143 	.check_events		= sd_check_events,
2144 	.unlock_native_capacity	= sd_unlock_native_capacity,
2145 	.report_zones		= sd_zbc_report_zones,
2146 	.get_unique_id		= sd_get_unique_id,
2147 	.free_disk		= scsi_disk_free_disk,
2148 	.pr_ops			= &sd_pr_ops,
2149 };
2150 
2151 /**
2152  *	sd_eh_reset - reset error handling callback
2153  *	@scmd:		sd-issued command that has failed
2154  *
2155  *	This function is called by the SCSI midlayer before starting
2156  *	SCSI EH. When counting medium access failures we have to be
2157  *	careful to register it only only once per device and SCSI EH run;
2158  *	there might be several timed out commands which will cause the
2159  *	'max_medium_access_timeouts' counter to trigger after the first
2160  *	SCSI EH run already and set the device to offline.
2161  *	So this function resets the internal counter before starting SCSI EH.
2162  **/
2163 static void sd_eh_reset(struct scsi_cmnd *scmd)
2164 {
2165 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2166 
2167 	/* New SCSI EH run, reset gate variable */
2168 	sdkp->ignore_medium_access_errors = false;
2169 }
2170 
2171 /**
2172  *	sd_eh_action - error handling callback
2173  *	@scmd:		sd-issued command that has failed
2174  *	@eh_disp:	The recovery disposition suggested by the midlayer
2175  *
2176  *	This function is called by the SCSI midlayer upon completion of an
2177  *	error test command (currently TEST UNIT READY). The result of sending
2178  *	the eh command is passed in eh_disp.  We're looking for devices that
2179  *	fail medium access commands but are OK with non access commands like
2180  *	test unit ready (so wrongly see the device as having a successful
2181  *	recovery)
2182  **/
2183 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
2184 {
2185 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2186 	struct scsi_device *sdev = scmd->device;
2187 
2188 	if (!scsi_device_online(sdev) ||
2189 	    !scsi_medium_access_command(scmd) ||
2190 	    host_byte(scmd->result) != DID_TIME_OUT ||
2191 	    eh_disp != SUCCESS)
2192 		return eh_disp;
2193 
2194 	/*
2195 	 * The device has timed out executing a medium access command.
2196 	 * However, the TEST UNIT READY command sent during error
2197 	 * handling completed successfully. Either the device is in the
2198 	 * process of recovering or has it suffered an internal failure
2199 	 * that prevents access to the storage medium.
2200 	 */
2201 	if (!sdkp->ignore_medium_access_errors) {
2202 		sdkp->medium_access_timed_out++;
2203 		sdkp->ignore_medium_access_errors = true;
2204 	}
2205 
2206 	/*
2207 	 * If the device keeps failing read/write commands but TEST UNIT
2208 	 * READY always completes successfully we assume that medium
2209 	 * access is no longer possible and take the device offline.
2210 	 */
2211 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2212 		scmd_printk(KERN_ERR, scmd,
2213 			    "Medium access timeout failure. Offlining disk!\n");
2214 		mutex_lock(&sdev->state_mutex);
2215 		scsi_device_set_state(sdev, SDEV_OFFLINE);
2216 		mutex_unlock(&sdev->state_mutex);
2217 
2218 		return SUCCESS;
2219 	}
2220 
2221 	return eh_disp;
2222 }
2223 
2224 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2225 {
2226 	struct request *req = scsi_cmd_to_rq(scmd);
2227 	struct scsi_device *sdev = scmd->device;
2228 	unsigned int transferred, good_bytes;
2229 	u64 start_lba, end_lba, bad_lba;
2230 
2231 	/*
2232 	 * Some commands have a payload smaller than the device logical
2233 	 * block size (e.g. INQUIRY on a 4K disk).
2234 	 */
2235 	if (scsi_bufflen(scmd) <= sdev->sector_size)
2236 		return 0;
2237 
2238 	/* Check if we have a 'bad_lba' information */
2239 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2240 				     SCSI_SENSE_BUFFERSIZE,
2241 				     &bad_lba))
2242 		return 0;
2243 
2244 	/*
2245 	 * If the bad lba was reported incorrectly, we have no idea where
2246 	 * the error is.
2247 	 */
2248 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2249 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2250 	if (bad_lba < start_lba || bad_lba >= end_lba)
2251 		return 0;
2252 
2253 	/*
2254 	 * resid is optional but mostly filled in.  When it's unused,
2255 	 * its value is zero, so we assume the whole buffer transferred
2256 	 */
2257 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2258 
2259 	/* This computation should always be done in terms of the
2260 	 * resolution of the device's medium.
2261 	 */
2262 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2263 
2264 	return min(good_bytes, transferred);
2265 }
2266 
2267 /**
2268  *	sd_done - bottom half handler: called when the lower level
2269  *	driver has completed (successfully or otherwise) a scsi command.
2270  *	@SCpnt: mid-level's per command structure.
2271  *
2272  *	Note: potentially run from within an ISR. Must not block.
2273  **/
2274 static int sd_done(struct scsi_cmnd *SCpnt)
2275 {
2276 	int result = SCpnt->result;
2277 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2278 	unsigned int sector_size = SCpnt->device->sector_size;
2279 	unsigned int resid;
2280 	struct scsi_sense_hdr sshdr;
2281 	struct request *req = scsi_cmd_to_rq(SCpnt);
2282 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2283 	int sense_valid = 0;
2284 	int sense_deferred = 0;
2285 
2286 	switch (req_op(req)) {
2287 	case REQ_OP_DISCARD:
2288 	case REQ_OP_WRITE_ZEROES:
2289 	case REQ_OP_ZONE_RESET:
2290 	case REQ_OP_ZONE_RESET_ALL:
2291 	case REQ_OP_ZONE_OPEN:
2292 	case REQ_OP_ZONE_CLOSE:
2293 	case REQ_OP_ZONE_FINISH:
2294 		if (!result) {
2295 			good_bytes = blk_rq_bytes(req);
2296 			scsi_set_resid(SCpnt, 0);
2297 		} else {
2298 			good_bytes = 0;
2299 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2300 		}
2301 		break;
2302 	default:
2303 		/*
2304 		 * In case of bogus fw or device, we could end up having
2305 		 * an unaligned partial completion. Check this here and force
2306 		 * alignment.
2307 		 */
2308 		resid = scsi_get_resid(SCpnt);
2309 		if (resid & (sector_size - 1)) {
2310 			sd_printk(KERN_INFO, sdkp,
2311 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2312 				resid, sector_size);
2313 			scsi_print_command(SCpnt);
2314 			resid = min(scsi_bufflen(SCpnt),
2315 				    round_up(resid, sector_size));
2316 			scsi_set_resid(SCpnt, resid);
2317 		}
2318 	}
2319 
2320 	if (result) {
2321 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2322 		if (sense_valid)
2323 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2324 	}
2325 	sdkp->medium_access_timed_out = 0;
2326 
2327 	if (!scsi_status_is_check_condition(result) &&
2328 	    (!sense_valid || sense_deferred))
2329 		goto out;
2330 
2331 	switch (sshdr.sense_key) {
2332 	case HARDWARE_ERROR:
2333 	case MEDIUM_ERROR:
2334 		good_bytes = sd_completed_bytes(SCpnt);
2335 		break;
2336 	case RECOVERED_ERROR:
2337 		good_bytes = scsi_bufflen(SCpnt);
2338 		break;
2339 	case NO_SENSE:
2340 		/* This indicates a false check condition, so ignore it.  An
2341 		 * unknown amount of data was transferred so treat it as an
2342 		 * error.
2343 		 */
2344 		SCpnt->result = 0;
2345 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2346 		break;
2347 	case ABORTED_COMMAND:
2348 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2349 			good_bytes = sd_completed_bytes(SCpnt);
2350 		break;
2351 	case ILLEGAL_REQUEST:
2352 		switch (sshdr.asc) {
2353 		case 0x10:	/* DIX: Host detected corruption */
2354 			good_bytes = sd_completed_bytes(SCpnt);
2355 			break;
2356 		case 0x20:	/* INVALID COMMAND OPCODE */
2357 		case 0x24:	/* INVALID FIELD IN CDB */
2358 			switch (SCpnt->cmnd[0]) {
2359 			case UNMAP:
2360 				sd_disable_discard(sdkp);
2361 				break;
2362 			case WRITE_SAME_16:
2363 			case WRITE_SAME:
2364 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2365 					sd_disable_discard(sdkp);
2366 				} else {
2367 					sd_disable_write_same(sdkp);
2368 					req->rq_flags |= RQF_QUIET;
2369 				}
2370 				break;
2371 			}
2372 		}
2373 		break;
2374 	default:
2375 		break;
2376 	}
2377 
2378  out:
2379 	if (sdkp->device->type == TYPE_ZBC)
2380 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2381 
2382 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2383 					   "sd_done: completed %d of %d bytes\n",
2384 					   good_bytes, scsi_bufflen(SCpnt)));
2385 
2386 	return good_bytes;
2387 }
2388 
2389 /*
2390  * spinup disk - called only in sd_revalidate_disk()
2391  */
2392 static void
2393 sd_spinup_disk(struct scsi_disk *sdkp)
2394 {
2395 	static const u8 cmd[10] = { TEST_UNIT_READY };
2396 	unsigned long spintime_expire = 0;
2397 	int spintime, sense_valid = 0;
2398 	unsigned int the_result;
2399 	struct scsi_sense_hdr sshdr;
2400 	struct scsi_failure failure_defs[] = {
2401 		/* Do not retry Medium Not Present */
2402 		{
2403 			.sense = UNIT_ATTENTION,
2404 			.asc = 0x3A,
2405 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2406 			.result = SAM_STAT_CHECK_CONDITION,
2407 		},
2408 		{
2409 			.sense = NOT_READY,
2410 			.asc = 0x3A,
2411 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2412 			.result = SAM_STAT_CHECK_CONDITION,
2413 		},
2414 		/* Retry when scsi_status_is_good would return false 3 times */
2415 		{
2416 			.result = SCMD_FAILURE_STAT_ANY,
2417 			.allowed = 3,
2418 		},
2419 		{}
2420 	};
2421 	struct scsi_failures failures = {
2422 		.failure_definitions = failure_defs,
2423 	};
2424 	const struct scsi_exec_args exec_args = {
2425 		.sshdr = &sshdr,
2426 		.failures = &failures,
2427 	};
2428 
2429 	spintime = 0;
2430 
2431 	/* Spin up drives, as required.  Only do this at boot time */
2432 	/* Spinup needs to be done for module loads too. */
2433 	do {
2434 		bool media_was_present = sdkp->media_present;
2435 
2436 		scsi_failures_reset_retries(&failures);
2437 
2438 		the_result = scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN,
2439 					      NULL, 0, SD_TIMEOUT,
2440 					      sdkp->max_retries, &exec_args);
2441 
2442 
2443 		if (the_result > 0) {
2444 			/*
2445 			 * If the drive has indicated to us that it doesn't
2446 			 * have any media in it, don't bother with any more
2447 			 * polling.
2448 			 */
2449 			if (media_not_present(sdkp, &sshdr)) {
2450 				if (media_was_present)
2451 					sd_printk(KERN_NOTICE, sdkp,
2452 						  "Media removed, stopped polling\n");
2453 				return;
2454 			}
2455 			sense_valid = scsi_sense_valid(&sshdr);
2456 		}
2457 
2458 		if (!scsi_status_is_check_condition(the_result)) {
2459 			/* no sense, TUR either succeeded or failed
2460 			 * with a status error */
2461 			if(!spintime && !scsi_status_is_good(the_result)) {
2462 				sd_print_result(sdkp, "Test Unit Ready failed",
2463 						the_result);
2464 			}
2465 			break;
2466 		}
2467 
2468 		/*
2469 		 * The device does not want the automatic start to be issued.
2470 		 */
2471 		if (sdkp->device->no_start_on_add)
2472 			break;
2473 
2474 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2475 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2476 				break;	/* manual intervention required */
2477 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2478 				break;	/* standby */
2479 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2480 				break;	/* unavailable */
2481 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2482 				break;	/* sanitize in progress */
2483 			if (sshdr.asc == 4 && sshdr.ascq == 0x24)
2484 				break;	/* depopulation in progress */
2485 			if (sshdr.asc == 4 && sshdr.ascq == 0x25)
2486 				break;	/* depopulation restoration in progress */
2487 			/*
2488 			 * Issue command to spin up drive when not ready
2489 			 */
2490 			if (!spintime) {
2491 				/* Return immediately and start spin cycle */
2492 				const u8 start_cmd[10] = {
2493 					[0] = START_STOP,
2494 					[1] = 1,
2495 					[4] = sdkp->device->start_stop_pwr_cond ?
2496 						0x11 : 1,
2497 				};
2498 
2499 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2500 				scsi_execute_cmd(sdkp->device, start_cmd,
2501 						 REQ_OP_DRV_IN, NULL, 0,
2502 						 SD_TIMEOUT, sdkp->max_retries,
2503 						 &exec_args);
2504 				spintime_expire = jiffies + 100 * HZ;
2505 				spintime = 1;
2506 			}
2507 			/* Wait 1 second for next try */
2508 			msleep(1000);
2509 			printk(KERN_CONT ".");
2510 
2511 		/*
2512 		 * Wait for USB flash devices with slow firmware.
2513 		 * Yes, this sense key/ASC combination shouldn't
2514 		 * occur here.  It's characteristic of these devices.
2515 		 */
2516 		} else if (sense_valid &&
2517 				sshdr.sense_key == UNIT_ATTENTION &&
2518 				sshdr.asc == 0x28) {
2519 			if (!spintime) {
2520 				spintime_expire = jiffies + 5 * HZ;
2521 				spintime = 1;
2522 			}
2523 			/* Wait 1 second for next try */
2524 			msleep(1000);
2525 		} else {
2526 			/* we don't understand the sense code, so it's
2527 			 * probably pointless to loop */
2528 			if(!spintime) {
2529 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2530 				sd_print_sense_hdr(sdkp, &sshdr);
2531 			}
2532 			break;
2533 		}
2534 
2535 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2536 
2537 	if (spintime) {
2538 		if (scsi_status_is_good(the_result))
2539 			printk(KERN_CONT "ready\n");
2540 		else
2541 			printk(KERN_CONT "not responding...\n");
2542 	}
2543 }
2544 
2545 /*
2546  * Determine whether disk supports Data Integrity Field.
2547  */
2548 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2549 {
2550 	struct scsi_device *sdp = sdkp->device;
2551 	u8 type;
2552 
2553 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2554 		sdkp->protection_type = 0;
2555 		return 0;
2556 	}
2557 
2558 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2559 
2560 	if (type > T10_PI_TYPE3_PROTECTION) {
2561 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2562 			  " protection type %u. Disabling disk!\n",
2563 			  type);
2564 		sdkp->protection_type = 0;
2565 		return -ENODEV;
2566 	}
2567 
2568 	sdkp->protection_type = type;
2569 
2570 	return 0;
2571 }
2572 
2573 static void sd_config_protection(struct scsi_disk *sdkp,
2574 		struct queue_limits *lim)
2575 {
2576 	struct scsi_device *sdp = sdkp->device;
2577 
2578 	if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY))
2579 		sd_dif_config_host(sdkp, lim);
2580 
2581 	if (!sdkp->protection_type)
2582 		return;
2583 
2584 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2585 		sd_first_printk(KERN_NOTICE, sdkp,
2586 				"Disabling DIF Type %u protection\n",
2587 				sdkp->protection_type);
2588 		sdkp->protection_type = 0;
2589 	}
2590 
2591 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2592 			sdkp->protection_type);
2593 }
2594 
2595 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2596 			struct scsi_sense_hdr *sshdr, int sense_valid,
2597 			int the_result)
2598 {
2599 	if (sense_valid)
2600 		sd_print_sense_hdr(sdkp, sshdr);
2601 	else
2602 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2603 
2604 	/*
2605 	 * Set dirty bit for removable devices if not ready -
2606 	 * sometimes drives will not report this properly.
2607 	 */
2608 	if (sdp->removable &&
2609 	    sense_valid && sshdr->sense_key == NOT_READY)
2610 		set_media_not_present(sdkp);
2611 
2612 	/*
2613 	 * We used to set media_present to 0 here to indicate no media
2614 	 * in the drive, but some drives fail read capacity even with
2615 	 * media present, so we can't do that.
2616 	 */
2617 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2618 }
2619 
2620 #define RC16_LEN 32
2621 #if RC16_LEN > SD_BUF_SIZE
2622 #error RC16_LEN must not be more than SD_BUF_SIZE
2623 #endif
2624 
2625 #define READ_CAPACITY_RETRIES_ON_RESET	10
2626 
2627 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2628 		struct queue_limits *lim, unsigned char *buffer)
2629 {
2630 	unsigned char cmd[16];
2631 	struct scsi_sense_hdr sshdr;
2632 	const struct scsi_exec_args exec_args = {
2633 		.sshdr = &sshdr,
2634 	};
2635 	int sense_valid = 0;
2636 	int the_result;
2637 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2638 	unsigned int alignment;
2639 	unsigned long long lba;
2640 	unsigned sector_size;
2641 
2642 	if (sdp->no_read_capacity_16)
2643 		return -EINVAL;
2644 
2645 	do {
2646 		memset(cmd, 0, 16);
2647 		cmd[0] = SERVICE_ACTION_IN_16;
2648 		cmd[1] = SAI_READ_CAPACITY_16;
2649 		cmd[13] = RC16_LEN;
2650 		memset(buffer, 0, RC16_LEN);
2651 
2652 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2653 					      buffer, RC16_LEN, SD_TIMEOUT,
2654 					      sdkp->max_retries, &exec_args);
2655 		if (the_result > 0) {
2656 			if (media_not_present(sdkp, &sshdr))
2657 				return -ENODEV;
2658 
2659 			sense_valid = scsi_sense_valid(&sshdr);
2660 			if (sense_valid &&
2661 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2662 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2663 			    sshdr.ascq == 0x00)
2664 				/* Invalid Command Operation Code or
2665 				 * Invalid Field in CDB, just retry
2666 				 * silently with RC10 */
2667 				return -EINVAL;
2668 			if (sense_valid &&
2669 			    sshdr.sense_key == UNIT_ATTENTION &&
2670 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2671 				/* Device reset might occur several times,
2672 				 * give it one more chance */
2673 				if (--reset_retries > 0)
2674 					continue;
2675 		}
2676 		retries--;
2677 
2678 	} while (the_result && retries);
2679 
2680 	if (the_result) {
2681 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2682 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2683 		return -EINVAL;
2684 	}
2685 
2686 	sector_size = get_unaligned_be32(&buffer[8]);
2687 	lba = get_unaligned_be64(&buffer[0]);
2688 
2689 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2690 		sdkp->capacity = 0;
2691 		return -ENODEV;
2692 	}
2693 
2694 	/* Logical blocks per physical block exponent */
2695 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2696 
2697 	/* RC basis */
2698 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2699 
2700 	/* Lowest aligned logical block */
2701 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2702 	lim->alignment_offset = alignment;
2703 	if (alignment && sdkp->first_scan)
2704 		sd_printk(KERN_NOTICE, sdkp,
2705 			  "physical block alignment offset: %u\n", alignment);
2706 
2707 	if (buffer[14] & 0x80) { /* LBPME */
2708 		sdkp->lbpme = 1;
2709 
2710 		if (buffer[14] & 0x40) /* LBPRZ */
2711 			sdkp->lbprz = 1;
2712 
2713 		sd_config_discard(sdkp, lim, SD_LBP_WS16);
2714 	}
2715 
2716 	sdkp->capacity = lba + 1;
2717 	return sector_size;
2718 }
2719 
2720 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2721 						unsigned char *buffer)
2722 {
2723 	static const u8 cmd[10] = { READ_CAPACITY };
2724 	struct scsi_sense_hdr sshdr;
2725 	struct scsi_failure failure_defs[] = {
2726 		/* Do not retry Medium Not Present */
2727 		{
2728 			.sense = UNIT_ATTENTION,
2729 			.asc = 0x3A,
2730 			.result = SAM_STAT_CHECK_CONDITION,
2731 		},
2732 		{
2733 			.sense = NOT_READY,
2734 			.asc = 0x3A,
2735 			.result = SAM_STAT_CHECK_CONDITION,
2736 		},
2737 		 /* Device reset might occur several times so retry a lot */
2738 		{
2739 			.sense = UNIT_ATTENTION,
2740 			.asc = 0x29,
2741 			.allowed = READ_CAPACITY_RETRIES_ON_RESET,
2742 			.result = SAM_STAT_CHECK_CONDITION,
2743 		},
2744 		/* Any other error not listed above retry 3 times */
2745 		{
2746 			.result = SCMD_FAILURE_RESULT_ANY,
2747 			.allowed = 3,
2748 		},
2749 		{}
2750 	};
2751 	struct scsi_failures failures = {
2752 		.failure_definitions = failure_defs,
2753 	};
2754 	const struct scsi_exec_args exec_args = {
2755 		.sshdr = &sshdr,
2756 		.failures = &failures,
2757 	};
2758 	int sense_valid = 0;
2759 	int the_result;
2760 	sector_t lba;
2761 	unsigned sector_size;
2762 
2763 	memset(buffer, 0, 8);
2764 
2765 	the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2766 				      8, SD_TIMEOUT, sdkp->max_retries,
2767 				      &exec_args);
2768 
2769 	if (the_result > 0) {
2770 		sense_valid = scsi_sense_valid(&sshdr);
2771 
2772 		if (media_not_present(sdkp, &sshdr))
2773 			return -ENODEV;
2774 	}
2775 
2776 	if (the_result) {
2777 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2778 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2779 		return -EINVAL;
2780 	}
2781 
2782 	sector_size = get_unaligned_be32(&buffer[4]);
2783 	lba = get_unaligned_be32(&buffer[0]);
2784 
2785 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2786 		/* Some buggy (usb cardreader) devices return an lba of
2787 		   0xffffffff when the want to report a size of 0 (with
2788 		   which they really mean no media is present) */
2789 		sdkp->capacity = 0;
2790 		sdkp->physical_block_size = sector_size;
2791 		return sector_size;
2792 	}
2793 
2794 	sdkp->capacity = lba + 1;
2795 	sdkp->physical_block_size = sector_size;
2796 	return sector_size;
2797 }
2798 
2799 static int sd_try_rc16_first(struct scsi_device *sdp)
2800 {
2801 	if (sdp->host->max_cmd_len < 16)
2802 		return 0;
2803 	if (sdp->try_rc_10_first)
2804 		return 0;
2805 	if (sdp->scsi_level > SCSI_SPC_2)
2806 		return 1;
2807 	if (scsi_device_protection(sdp))
2808 		return 1;
2809 	return 0;
2810 }
2811 
2812 /*
2813  * read disk capacity
2814  */
2815 static void
2816 sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim,
2817 		unsigned char *buffer)
2818 {
2819 	int sector_size;
2820 	struct scsi_device *sdp = sdkp->device;
2821 
2822 	if (sd_try_rc16_first(sdp)) {
2823 		sector_size = read_capacity_16(sdkp, sdp, lim, buffer);
2824 		if (sector_size == -EOVERFLOW)
2825 			goto got_data;
2826 		if (sector_size == -ENODEV)
2827 			return;
2828 		if (sector_size < 0)
2829 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2830 		if (sector_size < 0)
2831 			return;
2832 	} else {
2833 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2834 		if (sector_size == -EOVERFLOW)
2835 			goto got_data;
2836 		if (sector_size < 0)
2837 			return;
2838 		if ((sizeof(sdkp->capacity) > 4) &&
2839 		    (sdkp->capacity > 0xffffffffULL)) {
2840 			int old_sector_size = sector_size;
2841 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2842 					"Trying to use READ CAPACITY(16).\n");
2843 			sector_size = read_capacity_16(sdkp, sdp, lim, buffer);
2844 			if (sector_size < 0) {
2845 				sd_printk(KERN_NOTICE, sdkp,
2846 					"Using 0xffffffff as device size\n");
2847 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2848 				sector_size = old_sector_size;
2849 				goto got_data;
2850 			}
2851 			/* Remember that READ CAPACITY(16) succeeded */
2852 			sdp->try_rc_10_first = 0;
2853 		}
2854 	}
2855 
2856 	/* Some devices are known to return the total number of blocks,
2857 	 * not the highest block number.  Some devices have versions
2858 	 * which do this and others which do not.  Some devices we might
2859 	 * suspect of doing this but we don't know for certain.
2860 	 *
2861 	 * If we know the reported capacity is wrong, decrement it.  If
2862 	 * we can only guess, then assume the number of blocks is even
2863 	 * (usually true but not always) and err on the side of lowering
2864 	 * the capacity.
2865 	 */
2866 	if (sdp->fix_capacity ||
2867 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2868 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2869 				"from its reported value: %llu\n",
2870 				(unsigned long long) sdkp->capacity);
2871 		--sdkp->capacity;
2872 	}
2873 
2874 got_data:
2875 	if (sector_size == 0) {
2876 		sector_size = 512;
2877 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2878 			  "assuming 512.\n");
2879 	}
2880 
2881 	if (sector_size != 512 &&
2882 	    sector_size != 1024 &&
2883 	    sector_size != 2048 &&
2884 	    sector_size != 4096) {
2885 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2886 			  sector_size);
2887 		/*
2888 		 * The user might want to re-format the drive with
2889 		 * a supported sectorsize.  Once this happens, it
2890 		 * would be relatively trivial to set the thing up.
2891 		 * For this reason, we leave the thing in the table.
2892 		 */
2893 		sdkp->capacity = 0;
2894 		/*
2895 		 * set a bogus sector size so the normal read/write
2896 		 * logic in the block layer will eventually refuse any
2897 		 * request on this device without tripping over power
2898 		 * of two sector size assumptions
2899 		 */
2900 		sector_size = 512;
2901 	}
2902 	lim->logical_block_size = sector_size;
2903 	lim->physical_block_size = sdkp->physical_block_size;
2904 	sdkp->device->sector_size = sector_size;
2905 
2906 	if (sdkp->capacity > 0xffffffff)
2907 		sdp->use_16_for_rw = 1;
2908 
2909 }
2910 
2911 /*
2912  * Print disk capacity
2913  */
2914 static void
2915 sd_print_capacity(struct scsi_disk *sdkp,
2916 		  sector_t old_capacity)
2917 {
2918 	int sector_size = sdkp->device->sector_size;
2919 	char cap_str_2[10], cap_str_10[10];
2920 
2921 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2922 		return;
2923 
2924 	string_get_size(sdkp->capacity, sector_size,
2925 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2926 	string_get_size(sdkp->capacity, sector_size,
2927 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2928 
2929 	sd_printk(KERN_NOTICE, sdkp,
2930 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2931 		  (unsigned long long)sdkp->capacity,
2932 		  sector_size, cap_str_10, cap_str_2);
2933 
2934 	if (sdkp->physical_block_size != sector_size)
2935 		sd_printk(KERN_NOTICE, sdkp,
2936 			  "%u-byte physical blocks\n",
2937 			  sdkp->physical_block_size);
2938 }
2939 
2940 /* called with buffer of length 512 */
2941 static inline int
2942 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2943 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2944 		 struct scsi_sense_hdr *sshdr)
2945 {
2946 	/*
2947 	 * If we must use MODE SENSE(10), make sure that the buffer length
2948 	 * is at least 8 bytes so that the mode sense header fits.
2949 	 */
2950 	if (sdkp->device->use_10_for_ms && len < 8)
2951 		len = 8;
2952 
2953 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2954 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2955 }
2956 
2957 /*
2958  * read write protect setting, if possible - called only in sd_revalidate_disk()
2959  * called with buffer of length SD_BUF_SIZE
2960  */
2961 static void
2962 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2963 {
2964 	int res;
2965 	struct scsi_device *sdp = sdkp->device;
2966 	struct scsi_mode_data data;
2967 	int old_wp = sdkp->write_prot;
2968 
2969 	set_disk_ro(sdkp->disk, 0);
2970 	if (sdp->skip_ms_page_3f) {
2971 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2972 		return;
2973 	}
2974 
2975 	if (sdp->use_192_bytes_for_3f) {
2976 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2977 	} else {
2978 		/*
2979 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2980 		 * We have to start carefully: some devices hang if we ask
2981 		 * for more than is available.
2982 		 */
2983 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2984 
2985 		/*
2986 		 * Second attempt: ask for page 0 When only page 0 is
2987 		 * implemented, a request for page 3F may return Sense Key
2988 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2989 		 * CDB.
2990 		 */
2991 		if (res < 0)
2992 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2993 
2994 		/*
2995 		 * Third attempt: ask 255 bytes, as we did earlier.
2996 		 */
2997 		if (res < 0)
2998 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2999 					       &data, NULL);
3000 	}
3001 
3002 	if (res < 0) {
3003 		sd_first_printk(KERN_WARNING, sdkp,
3004 			  "Test WP failed, assume Write Enabled\n");
3005 	} else {
3006 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
3007 		set_disk_ro(sdkp->disk, sdkp->write_prot);
3008 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
3009 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
3010 				  sdkp->write_prot ? "on" : "off");
3011 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
3012 		}
3013 	}
3014 }
3015 
3016 /*
3017  * sd_read_cache_type - called only from sd_revalidate_disk()
3018  * called with buffer of length SD_BUF_SIZE
3019  */
3020 static void
3021 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
3022 {
3023 	int len = 0, res;
3024 	struct scsi_device *sdp = sdkp->device;
3025 
3026 	int dbd;
3027 	int modepage;
3028 	int first_len;
3029 	struct scsi_mode_data data;
3030 	struct scsi_sense_hdr sshdr;
3031 	int old_wce = sdkp->WCE;
3032 	int old_rcd = sdkp->RCD;
3033 	int old_dpofua = sdkp->DPOFUA;
3034 
3035 
3036 	if (sdkp->cache_override)
3037 		return;
3038 
3039 	first_len = 4;
3040 	if (sdp->skip_ms_page_8) {
3041 		if (sdp->type == TYPE_RBC)
3042 			goto defaults;
3043 		else {
3044 			if (sdp->skip_ms_page_3f)
3045 				goto defaults;
3046 			modepage = 0x3F;
3047 			if (sdp->use_192_bytes_for_3f)
3048 				first_len = 192;
3049 			dbd = 0;
3050 		}
3051 	} else if (sdp->type == TYPE_RBC) {
3052 		modepage = 6;
3053 		dbd = 8;
3054 	} else {
3055 		modepage = 8;
3056 		dbd = 0;
3057 	}
3058 
3059 	/* cautiously ask */
3060 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
3061 			&data, &sshdr);
3062 
3063 	if (res < 0)
3064 		goto bad_sense;
3065 
3066 	if (!data.header_length) {
3067 		modepage = 6;
3068 		first_len = 0;
3069 		sd_first_printk(KERN_ERR, sdkp,
3070 				"Missing header in MODE_SENSE response\n");
3071 	}
3072 
3073 	/* that went OK, now ask for the proper length */
3074 	len = data.length;
3075 
3076 	/*
3077 	 * We're only interested in the first three bytes, actually.
3078 	 * But the data cache page is defined for the first 20.
3079 	 */
3080 	if (len < 3)
3081 		goto bad_sense;
3082 	else if (len > SD_BUF_SIZE) {
3083 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
3084 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
3085 		len = SD_BUF_SIZE;
3086 	}
3087 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
3088 		len = 192;
3089 
3090 	/* Get the data */
3091 	if (len > first_len)
3092 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
3093 				&data, &sshdr);
3094 
3095 	if (!res) {
3096 		int offset = data.header_length + data.block_descriptor_length;
3097 
3098 		while (offset < len) {
3099 			u8 page_code = buffer[offset] & 0x3F;
3100 			u8 spf       = buffer[offset] & 0x40;
3101 
3102 			if (page_code == 8 || page_code == 6) {
3103 				/* We're interested only in the first 3 bytes.
3104 				 */
3105 				if (len - offset <= 2) {
3106 					sd_first_printk(KERN_ERR, sdkp,
3107 						"Incomplete mode parameter "
3108 							"data\n");
3109 					goto defaults;
3110 				} else {
3111 					modepage = page_code;
3112 					goto Page_found;
3113 				}
3114 			} else {
3115 				/* Go to the next page */
3116 				if (spf && len - offset > 3)
3117 					offset += 4 + (buffer[offset+2] << 8) +
3118 						buffer[offset+3];
3119 				else if (!spf && len - offset > 1)
3120 					offset += 2 + buffer[offset+1];
3121 				else {
3122 					sd_first_printk(KERN_ERR, sdkp,
3123 							"Incomplete mode "
3124 							"parameter data\n");
3125 					goto defaults;
3126 				}
3127 			}
3128 		}
3129 
3130 		sd_first_printk(KERN_WARNING, sdkp,
3131 				"No Caching mode page found\n");
3132 		goto defaults;
3133 
3134 	Page_found:
3135 		if (modepage == 8) {
3136 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
3137 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
3138 		} else {
3139 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
3140 			sdkp->RCD = 0;
3141 		}
3142 
3143 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
3144 		if (sdp->broken_fua) {
3145 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
3146 			sdkp->DPOFUA = 0;
3147 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
3148 			   !sdkp->device->use_16_for_rw) {
3149 			sd_first_printk(KERN_NOTICE, sdkp,
3150 				  "Uses READ/WRITE(6), disabling FUA\n");
3151 			sdkp->DPOFUA = 0;
3152 		}
3153 
3154 		/* No cache flush allowed for write protected devices */
3155 		if (sdkp->WCE && sdkp->write_prot)
3156 			sdkp->WCE = 0;
3157 
3158 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
3159 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
3160 			sd_printk(KERN_NOTICE, sdkp,
3161 				  "Write cache: %s, read cache: %s, %s\n",
3162 				  sdkp->WCE ? "enabled" : "disabled",
3163 				  sdkp->RCD ? "disabled" : "enabled",
3164 				  sdkp->DPOFUA ? "supports DPO and FUA"
3165 				  : "doesn't support DPO or FUA");
3166 
3167 		return;
3168 	}
3169 
3170 bad_sense:
3171 	if (res == -EIO && scsi_sense_valid(&sshdr) &&
3172 	    sshdr.sense_key == ILLEGAL_REQUEST &&
3173 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
3174 		/* Invalid field in CDB */
3175 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
3176 	else
3177 		sd_first_printk(KERN_ERR, sdkp,
3178 				"Asking for cache data failed\n");
3179 
3180 defaults:
3181 	if (sdp->wce_default_on) {
3182 		sd_first_printk(KERN_NOTICE, sdkp,
3183 				"Assuming drive cache: write back\n");
3184 		sdkp->WCE = 1;
3185 	} else {
3186 		sd_first_printk(KERN_WARNING, sdkp,
3187 				"Assuming drive cache: write through\n");
3188 		sdkp->WCE = 0;
3189 	}
3190 	sdkp->RCD = 0;
3191 	sdkp->DPOFUA = 0;
3192 }
3193 
3194 static bool sd_is_perm_stream(struct scsi_disk *sdkp, unsigned int stream_id)
3195 {
3196 	u8 cdb[16] = { SERVICE_ACTION_IN_16, SAI_GET_STREAM_STATUS };
3197 	struct {
3198 		struct scsi_stream_status_header h;
3199 		struct scsi_stream_status s;
3200 	} buf;
3201 	struct scsi_device *sdev = sdkp->device;
3202 	struct scsi_sense_hdr sshdr;
3203 	const struct scsi_exec_args exec_args = {
3204 		.sshdr = &sshdr,
3205 	};
3206 	int res;
3207 
3208 	put_unaligned_be16(stream_id, &cdb[4]);
3209 	put_unaligned_be32(sizeof(buf), &cdb[10]);
3210 
3211 	res = scsi_execute_cmd(sdev, cdb, REQ_OP_DRV_IN, &buf, sizeof(buf),
3212 			       SD_TIMEOUT, sdkp->max_retries, &exec_args);
3213 	if (res < 0)
3214 		return false;
3215 	if (scsi_status_is_check_condition(res) && scsi_sense_valid(&sshdr))
3216 		sd_print_sense_hdr(sdkp, &sshdr);
3217 	if (res)
3218 		return false;
3219 	if (get_unaligned_be32(&buf.h.len) < sizeof(struct scsi_stream_status))
3220 		return false;
3221 	return buf.h.stream_status[0].perm;
3222 }
3223 
3224 static void sd_read_io_hints(struct scsi_disk *sdkp, unsigned char *buffer)
3225 {
3226 	struct scsi_device *sdp = sdkp->device;
3227 	const struct scsi_io_group_descriptor *desc, *start, *end;
3228 	u16 permanent_stream_count_old;
3229 	struct scsi_sense_hdr sshdr;
3230 	struct scsi_mode_data data;
3231 	int res;
3232 
3233 	res = scsi_mode_sense(sdp, /*dbd=*/0x8, /*modepage=*/0x0a,
3234 			      /*subpage=*/0x05, buffer, SD_BUF_SIZE, SD_TIMEOUT,
3235 			      sdkp->max_retries, &data, &sshdr);
3236 	if (res < 0)
3237 		return;
3238 	start = (void *)buffer + data.header_length + 16;
3239 	end = (void *)buffer + ALIGN_DOWN(data.header_length + data.length,
3240 					  sizeof(*end));
3241 	/*
3242 	 * From "SBC-5 Constrained Streams with Data Lifetimes": Device severs
3243 	 * should assign the lowest numbered stream identifiers to permanent
3244 	 * streams.
3245 	 */
3246 	for (desc = start; desc < end; desc++)
3247 		if (!desc->st_enble || !sd_is_perm_stream(sdkp, desc - start))
3248 			break;
3249 	permanent_stream_count_old = sdkp->permanent_stream_count;
3250 	sdkp->permanent_stream_count = desc - start;
3251 	if (sdkp->rscs && sdkp->permanent_stream_count < 2)
3252 		sd_printk(KERN_INFO, sdkp,
3253 			  "Unexpected: RSCS has been set and the permanent stream count is %u\n",
3254 			  sdkp->permanent_stream_count);
3255 	else if (sdkp->permanent_stream_count != permanent_stream_count_old)
3256 		sd_printk(KERN_INFO, sdkp, "permanent stream count = %d\n",
3257 			  sdkp->permanent_stream_count);
3258 }
3259 
3260 /*
3261  * The ATO bit indicates whether the DIF application tag is available
3262  * for use by the operating system.
3263  */
3264 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
3265 {
3266 	int res, offset;
3267 	struct scsi_device *sdp = sdkp->device;
3268 	struct scsi_mode_data data;
3269 	struct scsi_sense_hdr sshdr;
3270 
3271 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
3272 		return;
3273 
3274 	if (sdkp->protection_type == 0)
3275 		return;
3276 
3277 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
3278 			      sdkp->max_retries, &data, &sshdr);
3279 
3280 	if (res < 0 || !data.header_length ||
3281 	    data.length < 6) {
3282 		sd_first_printk(KERN_WARNING, sdkp,
3283 			  "getting Control mode page failed, assume no ATO\n");
3284 
3285 		if (res == -EIO && scsi_sense_valid(&sshdr))
3286 			sd_print_sense_hdr(sdkp, &sshdr);
3287 
3288 		return;
3289 	}
3290 
3291 	offset = data.header_length + data.block_descriptor_length;
3292 
3293 	if ((buffer[offset] & 0x3f) != 0x0a) {
3294 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
3295 		return;
3296 	}
3297 
3298 	if ((buffer[offset + 5] & 0x80) == 0)
3299 		return;
3300 
3301 	sdkp->ATO = 1;
3302 
3303 	return;
3304 }
3305 
3306 static unsigned int sd_discard_mode(struct scsi_disk *sdkp)
3307 {
3308 	if (!sdkp->lbpvpd) {
3309 		/* LBP VPD page not provided */
3310 		if (sdkp->max_unmap_blocks)
3311 			return SD_LBP_UNMAP;
3312 		return SD_LBP_WS16;
3313 	}
3314 
3315 	/* LBP VPD page tells us what to use */
3316 	if (sdkp->lbpu && sdkp->max_unmap_blocks)
3317 		return SD_LBP_UNMAP;
3318 	if (sdkp->lbpws)
3319 		return SD_LBP_WS16;
3320 	if (sdkp->lbpws10)
3321 		return SD_LBP_WS10;
3322 	return SD_LBP_DISABLE;
3323 }
3324 
3325 /*
3326  * Query disk device for preferred I/O sizes.
3327  */
3328 static void sd_read_block_limits(struct scsi_disk *sdkp,
3329 		struct queue_limits *lim)
3330 {
3331 	struct scsi_vpd *vpd;
3332 
3333 	rcu_read_lock();
3334 
3335 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3336 	if (!vpd || vpd->len < 16)
3337 		goto out;
3338 
3339 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3340 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3341 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3342 
3343 	if (vpd->len >= 64) {
3344 		unsigned int lba_count, desc_count;
3345 
3346 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3347 
3348 		if (!sdkp->lbpme)
3349 			goto config_atomic;
3350 
3351 		lba_count = get_unaligned_be32(&vpd->data[20]);
3352 		desc_count = get_unaligned_be32(&vpd->data[24]);
3353 
3354 		if (lba_count && desc_count)
3355 			sdkp->max_unmap_blocks = lba_count;
3356 
3357 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3358 
3359 		if (vpd->data[32] & 0x80)
3360 			sdkp->unmap_alignment =
3361 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3362 
3363 		sd_config_discard(sdkp, lim, sd_discard_mode(sdkp));
3364 
3365 config_atomic:
3366 		sdkp->max_atomic = get_unaligned_be32(&vpd->data[44]);
3367 		sdkp->atomic_alignment = get_unaligned_be32(&vpd->data[48]);
3368 		sdkp->atomic_granularity = get_unaligned_be32(&vpd->data[52]);
3369 		sdkp->max_atomic_with_boundary = get_unaligned_be32(&vpd->data[56]);
3370 		sdkp->max_atomic_boundary = get_unaligned_be32(&vpd->data[60]);
3371 
3372 		sd_config_atomic(sdkp, lim);
3373 	}
3374 
3375  out:
3376 	rcu_read_unlock();
3377 }
3378 
3379 /* Parse the Block Limits Extension VPD page (0xb7) */
3380 static void sd_read_block_limits_ext(struct scsi_disk *sdkp)
3381 {
3382 	struct scsi_vpd *vpd;
3383 
3384 	rcu_read_lock();
3385 	vpd = rcu_dereference(sdkp->device->vpd_pgb7);
3386 	if (vpd && vpd->len >= 2)
3387 		sdkp->rscs = vpd->data[5] & 1;
3388 	rcu_read_unlock();
3389 }
3390 
3391 /* Query block device characteristics */
3392 static void sd_read_block_characteristics(struct scsi_disk *sdkp,
3393 		struct queue_limits *lim)
3394 {
3395 	struct scsi_vpd *vpd;
3396 	u16 rot;
3397 
3398 	rcu_read_lock();
3399 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3400 
3401 	if (!vpd || vpd->len < 8) {
3402 		rcu_read_unlock();
3403 	        return;
3404 	}
3405 
3406 	rot = get_unaligned_be16(&vpd->data[4]);
3407 	sdkp->zoned = (vpd->data[8] >> 4) & 3;
3408 	rcu_read_unlock();
3409 
3410 	if (rot == 1)
3411 		lim->features &= ~(BLK_FEAT_ROTATIONAL | BLK_FEAT_ADD_RANDOM);
3412 
3413 	if (!sdkp->first_scan)
3414 		return;
3415 
3416 	if (sdkp->device->type == TYPE_ZBC)
3417 		sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n");
3418 	else if (sdkp->zoned == 1)
3419 		sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n");
3420 	else if (sdkp->zoned == 2)
3421 		sd_printk(KERN_NOTICE, sdkp, "Drive-managed SMR disk\n");
3422 }
3423 
3424 /**
3425  * sd_read_block_provisioning - Query provisioning VPD page
3426  * @sdkp: disk to query
3427  */
3428 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3429 {
3430 	struct scsi_vpd *vpd;
3431 
3432 	if (sdkp->lbpme == 0)
3433 		return;
3434 
3435 	rcu_read_lock();
3436 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3437 
3438 	if (!vpd || vpd->len < 8) {
3439 		rcu_read_unlock();
3440 		return;
3441 	}
3442 
3443 	sdkp->lbpvpd	= 1;
3444 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3445 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3446 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3447 	rcu_read_unlock();
3448 }
3449 
3450 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3451 {
3452 	struct scsi_device *sdev = sdkp->device;
3453 
3454 	if (sdev->host->no_write_same) {
3455 		sdev->no_write_same = 1;
3456 
3457 		return;
3458 	}
3459 
3460 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3461 		struct scsi_vpd *vpd;
3462 
3463 		sdev->no_report_opcodes = 1;
3464 
3465 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3466 		 * CODES is unsupported and the device has an ATA
3467 		 * Information VPD page (SAT).
3468 		 */
3469 		rcu_read_lock();
3470 		vpd = rcu_dereference(sdev->vpd_pg89);
3471 		if (vpd)
3472 			sdev->no_write_same = 1;
3473 		rcu_read_unlock();
3474 	}
3475 
3476 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3477 		sdkp->ws16 = 1;
3478 
3479 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3480 		sdkp->ws10 = 1;
3481 }
3482 
3483 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3484 {
3485 	struct scsi_device *sdev = sdkp->device;
3486 
3487 	if (!sdev->security_supported)
3488 		return;
3489 
3490 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3491 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3492 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3493 			SECURITY_PROTOCOL_OUT, 0) == 1)
3494 		sdkp->security = 1;
3495 }
3496 
3497 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3498 {
3499 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3500 }
3501 
3502 /**
3503  * sd_read_cpr - Query concurrent positioning ranges
3504  * @sdkp:	disk to query
3505  */
3506 static void sd_read_cpr(struct scsi_disk *sdkp)
3507 {
3508 	struct blk_independent_access_ranges *iars = NULL;
3509 	unsigned char *buffer = NULL;
3510 	unsigned int nr_cpr = 0;
3511 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3512 	u8 *desc;
3513 
3514 	/*
3515 	 * We need to have the capacity set first for the block layer to be
3516 	 * able to check the ranges.
3517 	 */
3518 	if (sdkp->first_scan)
3519 		return;
3520 
3521 	if (!sdkp->capacity)
3522 		goto out;
3523 
3524 	/*
3525 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3526 	 * leading to a maximum page size of 64 + 256*32 bytes.
3527 	 */
3528 	buf_len = 64 + 256*32;
3529 	buffer = kmalloc(buf_len, GFP_KERNEL);
3530 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3531 		goto out;
3532 
3533 	/* We must have at least a 64B header and one 32B range descriptor */
3534 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3535 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3536 		sd_printk(KERN_ERR, sdkp,
3537 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3538 		goto out;
3539 	}
3540 
3541 	nr_cpr = (vpd_len - 64) / 32;
3542 	if (nr_cpr == 1) {
3543 		nr_cpr = 0;
3544 		goto out;
3545 	}
3546 
3547 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3548 	if (!iars) {
3549 		nr_cpr = 0;
3550 		goto out;
3551 	}
3552 
3553 	desc = &buffer[64];
3554 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3555 		if (desc[0] != i) {
3556 			sd_printk(KERN_ERR, sdkp,
3557 				"Invalid Concurrent Positioning Range number\n");
3558 			nr_cpr = 0;
3559 			break;
3560 		}
3561 
3562 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3563 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3564 	}
3565 
3566 out:
3567 	disk_set_independent_access_ranges(sdkp->disk, iars);
3568 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3569 		sd_printk(KERN_NOTICE, sdkp,
3570 			  "%u concurrent positioning ranges\n", nr_cpr);
3571 		sdkp->nr_actuators = nr_cpr;
3572 	}
3573 
3574 	kfree(buffer);
3575 }
3576 
3577 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3578 {
3579 	struct scsi_device *sdp = sdkp->device;
3580 	unsigned int min_xfer_bytes =
3581 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3582 
3583 	if (sdkp->min_xfer_blocks == 0)
3584 		return false;
3585 
3586 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3587 		sd_first_printk(KERN_WARNING, sdkp,
3588 				"Preferred minimum I/O size %u bytes not a " \
3589 				"multiple of physical block size (%u bytes)\n",
3590 				min_xfer_bytes, sdkp->physical_block_size);
3591 		sdkp->min_xfer_blocks = 0;
3592 		return false;
3593 	}
3594 
3595 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3596 			min_xfer_bytes);
3597 	return true;
3598 }
3599 
3600 /*
3601  * Determine the device's preferred I/O size for reads and writes
3602  * unless the reported value is unreasonably small, large, not a
3603  * multiple of the physical block size, or simply garbage.
3604  */
3605 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3606 				      unsigned int dev_max)
3607 {
3608 	struct scsi_device *sdp = sdkp->device;
3609 	unsigned int opt_xfer_bytes =
3610 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3611 	unsigned int min_xfer_bytes =
3612 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3613 
3614 	if (sdkp->opt_xfer_blocks == 0)
3615 		return false;
3616 
3617 	if (sdkp->opt_xfer_blocks > dev_max) {
3618 		sd_first_printk(KERN_WARNING, sdkp,
3619 				"Optimal transfer size %u logical blocks " \
3620 				"> dev_max (%u logical blocks)\n",
3621 				sdkp->opt_xfer_blocks, dev_max);
3622 		return false;
3623 	}
3624 
3625 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3626 		sd_first_printk(KERN_WARNING, sdkp,
3627 				"Optimal transfer size %u logical blocks " \
3628 				"> sd driver limit (%u logical blocks)\n",
3629 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3630 		return false;
3631 	}
3632 
3633 	if (opt_xfer_bytes < PAGE_SIZE) {
3634 		sd_first_printk(KERN_WARNING, sdkp,
3635 				"Optimal transfer size %u bytes < " \
3636 				"PAGE_SIZE (%u bytes)\n",
3637 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3638 		return false;
3639 	}
3640 
3641 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3642 		sd_first_printk(KERN_WARNING, sdkp,
3643 				"Optimal transfer size %u bytes not a " \
3644 				"multiple of preferred minimum block " \
3645 				"size (%u bytes)\n",
3646 				opt_xfer_bytes, min_xfer_bytes);
3647 		return false;
3648 	}
3649 
3650 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3651 		sd_first_printk(KERN_WARNING, sdkp,
3652 				"Optimal transfer size %u bytes not a " \
3653 				"multiple of physical block size (%u bytes)\n",
3654 				opt_xfer_bytes, sdkp->physical_block_size);
3655 		return false;
3656 	}
3657 
3658 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3659 			opt_xfer_bytes);
3660 	return true;
3661 }
3662 
3663 static void sd_read_block_zero(struct scsi_disk *sdkp)
3664 {
3665 	unsigned int buf_len = sdkp->device->sector_size;
3666 	char *buffer, cmd[10] = { };
3667 
3668 	buffer = kmalloc(buf_len, GFP_KERNEL);
3669 	if (!buffer)
3670 		return;
3671 
3672 	cmd[0] = READ_10;
3673 	put_unaligned_be32(0, &cmd[2]); /* Logical block address 0 */
3674 	put_unaligned_be16(1, &cmd[7]);	/* Transfer 1 logical block */
3675 
3676 	scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN, buffer, buf_len,
3677 			 SD_TIMEOUT, sdkp->max_retries, NULL);
3678 	kfree(buffer);
3679 }
3680 
3681 /**
3682  *	sd_revalidate_disk - called the first time a new disk is seen,
3683  *	performs disk spin up, read_capacity, etc.
3684  *	@disk: struct gendisk we care about
3685  **/
3686 static int sd_revalidate_disk(struct gendisk *disk)
3687 {
3688 	struct scsi_disk *sdkp = scsi_disk(disk);
3689 	struct scsi_device *sdp = sdkp->device;
3690 	sector_t old_capacity = sdkp->capacity;
3691 	struct queue_limits lim;
3692 	unsigned char *buffer;
3693 	unsigned int dev_max;
3694 	int err;
3695 
3696 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3697 				      "sd_revalidate_disk\n"));
3698 
3699 	/*
3700 	 * If the device is offline, don't try and read capacity or any
3701 	 * of the other niceties.
3702 	 */
3703 	if (!scsi_device_online(sdp))
3704 		goto out;
3705 
3706 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3707 	if (!buffer) {
3708 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3709 			  "allocation failure.\n");
3710 		goto out;
3711 	}
3712 
3713 	sd_spinup_disk(sdkp);
3714 
3715 	lim = queue_limits_start_update(sdkp->disk->queue);
3716 
3717 	/*
3718 	 * Without media there is no reason to ask; moreover, some devices
3719 	 * react badly if we do.
3720 	 */
3721 	if (sdkp->media_present) {
3722 		sd_read_capacity(sdkp, &lim, buffer);
3723 		/*
3724 		 * Some USB/UAS devices return generic values for mode pages
3725 		 * until the media has been accessed. Trigger a READ operation
3726 		 * to force the device to populate mode pages.
3727 		 */
3728 		if (sdp->read_before_ms)
3729 			sd_read_block_zero(sdkp);
3730 		/*
3731 		 * set the default to rotational.  All non-rotational devices
3732 		 * support the block characteristics VPD page, which will
3733 		 * cause this to be updated correctly and any device which
3734 		 * doesn't support it should be treated as rotational.
3735 		 */
3736 		lim.features |= (BLK_FEAT_ROTATIONAL | BLK_FEAT_ADD_RANDOM);
3737 
3738 		if (scsi_device_supports_vpd(sdp)) {
3739 			sd_read_block_provisioning(sdkp);
3740 			sd_read_block_limits(sdkp, &lim);
3741 			sd_read_block_limits_ext(sdkp);
3742 			sd_read_block_characteristics(sdkp, &lim);
3743 			sd_zbc_read_zones(sdkp, &lim, buffer);
3744 			sd_read_cpr(sdkp);
3745 		}
3746 
3747 		sd_print_capacity(sdkp, old_capacity);
3748 
3749 		sd_read_write_protect_flag(sdkp, buffer);
3750 		sd_read_cache_type(sdkp, buffer);
3751 		sd_read_io_hints(sdkp, buffer);
3752 		sd_read_app_tag_own(sdkp, buffer);
3753 		sd_read_write_same(sdkp, buffer);
3754 		sd_read_security(sdkp, buffer);
3755 		sd_config_protection(sdkp, &lim);
3756 	}
3757 
3758 	/*
3759 	 * We now have all cache related info, determine how we deal
3760 	 * with flush requests.
3761 	 */
3762 	sd_set_flush_flag(sdkp, &lim);
3763 
3764 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3765 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3766 
3767 	/* Some devices report a maximum block count for READ/WRITE requests. */
3768 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3769 	lim.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3770 
3771 	if (sd_validate_min_xfer_size(sdkp))
3772 		lim.io_min = logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3773 	else
3774 		lim.io_min = 0;
3775 
3776 	/*
3777 	 * Limit default to SCSI host optimal sector limit if set. There may be
3778 	 * an impact on performance for when the size of a request exceeds this
3779 	 * host limit.
3780 	 */
3781 	lim.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT;
3782 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3783 		lim.io_opt = min_not_zero(lim.io_opt,
3784 				logical_to_bytes(sdp, sdkp->opt_xfer_blocks));
3785 	}
3786 
3787 	sdkp->first_scan = 0;
3788 
3789 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3790 	sd_config_write_same(sdkp, &lim);
3791 	kfree(buffer);
3792 
3793 	blk_mq_freeze_queue(sdkp->disk->queue);
3794 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
3795 	blk_mq_unfreeze_queue(sdkp->disk->queue);
3796 	if (err)
3797 		return err;
3798 
3799 	/*
3800 	 * For a zoned drive, revalidating the zones can be done only once
3801 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3802 	 * capacity to 0.
3803 	 */
3804 	if (sd_zbc_revalidate_zones(sdkp))
3805 		set_capacity_and_notify(disk, 0);
3806 
3807  out:
3808 	return 0;
3809 }
3810 
3811 /**
3812  *	sd_unlock_native_capacity - unlock native capacity
3813  *	@disk: struct gendisk to set capacity for
3814  *
3815  *	Block layer calls this function if it detects that partitions
3816  *	on @disk reach beyond the end of the device.  If the SCSI host
3817  *	implements ->unlock_native_capacity() method, it's invoked to
3818  *	give it a chance to adjust the device capacity.
3819  *
3820  *	CONTEXT:
3821  *	Defined by block layer.  Might sleep.
3822  */
3823 static void sd_unlock_native_capacity(struct gendisk *disk)
3824 {
3825 	struct scsi_device *sdev = scsi_disk(disk)->device;
3826 
3827 	if (sdev->host->hostt->unlock_native_capacity)
3828 		sdev->host->hostt->unlock_native_capacity(sdev);
3829 }
3830 
3831 /**
3832  *	sd_format_disk_name - format disk name
3833  *	@prefix: name prefix - ie. "sd" for SCSI disks
3834  *	@index: index of the disk to format name for
3835  *	@buf: output buffer
3836  *	@buflen: length of the output buffer
3837  *
3838  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3839  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3840  *	which is followed by sdaaa.
3841  *
3842  *	This is basically 26 base counting with one extra 'nil' entry
3843  *	at the beginning from the second digit on and can be
3844  *	determined using similar method as 26 base conversion with the
3845  *	index shifted -1 after each digit is computed.
3846  *
3847  *	CONTEXT:
3848  *	Don't care.
3849  *
3850  *	RETURNS:
3851  *	0 on success, -errno on failure.
3852  */
3853 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3854 {
3855 	const int base = 'z' - 'a' + 1;
3856 	char *begin = buf + strlen(prefix);
3857 	char *end = buf + buflen;
3858 	char *p;
3859 	int unit;
3860 
3861 	p = end - 1;
3862 	*p = '\0';
3863 	unit = base;
3864 	do {
3865 		if (p == begin)
3866 			return -EINVAL;
3867 		*--p = 'a' + (index % unit);
3868 		index = (index / unit) - 1;
3869 	} while (index >= 0);
3870 
3871 	memmove(begin, p, end - p);
3872 	memcpy(buf, prefix, strlen(prefix));
3873 
3874 	return 0;
3875 }
3876 
3877 /**
3878  *	sd_probe - called during driver initialization and whenever a
3879  *	new scsi device is attached to the system. It is called once
3880  *	for each scsi device (not just disks) present.
3881  *	@dev: pointer to device object
3882  *
3883  *	Returns 0 if successful (or not interested in this scsi device
3884  *	(e.g. scanner)); 1 when there is an error.
3885  *
3886  *	Note: this function is invoked from the scsi mid-level.
3887  *	This function sets up the mapping between a given
3888  *	<host,channel,id,lun> (found in sdp) and new device name
3889  *	(e.g. /dev/sda). More precisely it is the block device major
3890  *	and minor number that is chosen here.
3891  *
3892  *	Assume sd_probe is not re-entrant (for time being)
3893  *	Also think about sd_probe() and sd_remove() running coincidentally.
3894  **/
3895 static int sd_probe(struct device *dev)
3896 {
3897 	struct scsi_device *sdp = to_scsi_device(dev);
3898 	struct scsi_disk *sdkp;
3899 	struct gendisk *gd;
3900 	int index;
3901 	int error;
3902 
3903 	scsi_autopm_get_device(sdp);
3904 	error = -ENODEV;
3905 	if (sdp->type != TYPE_DISK &&
3906 	    sdp->type != TYPE_ZBC &&
3907 	    sdp->type != TYPE_MOD &&
3908 	    sdp->type != TYPE_RBC)
3909 		goto out;
3910 
3911 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3912 		sdev_printk(KERN_WARNING, sdp,
3913 			    "Unsupported ZBC host-managed device.\n");
3914 		goto out;
3915 	}
3916 
3917 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3918 					"sd_probe\n"));
3919 
3920 	error = -ENOMEM;
3921 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3922 	if (!sdkp)
3923 		goto out;
3924 
3925 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3926 					 &sd_bio_compl_lkclass);
3927 	if (!gd)
3928 		goto out_free;
3929 
3930 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3931 	if (index < 0) {
3932 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3933 		goto out_put;
3934 	}
3935 
3936 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3937 	if (error) {
3938 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3939 		goto out_free_index;
3940 	}
3941 
3942 	sdkp->device = sdp;
3943 	sdkp->disk = gd;
3944 	sdkp->index = index;
3945 	sdkp->max_retries = SD_MAX_RETRIES;
3946 	atomic_set(&sdkp->openers, 0);
3947 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3948 
3949 	if (!sdp->request_queue->rq_timeout) {
3950 		if (sdp->type != TYPE_MOD)
3951 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3952 		else
3953 			blk_queue_rq_timeout(sdp->request_queue,
3954 					     SD_MOD_TIMEOUT);
3955 	}
3956 
3957 	device_initialize(&sdkp->disk_dev);
3958 	sdkp->disk_dev.parent = get_device(dev);
3959 	sdkp->disk_dev.class = &sd_disk_class;
3960 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3961 
3962 	error = device_add(&sdkp->disk_dev);
3963 	if (error) {
3964 		put_device(&sdkp->disk_dev);
3965 		goto out;
3966 	}
3967 
3968 	dev_set_drvdata(dev, sdkp);
3969 
3970 	gd->major = sd_major((index & 0xf0) >> 4);
3971 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3972 	gd->minors = SD_MINORS;
3973 
3974 	gd->fops = &sd_fops;
3975 	gd->private_data = sdkp;
3976 
3977 	/* defaults, until the device tells us otherwise */
3978 	sdp->sector_size = 512;
3979 	sdkp->capacity = 0;
3980 	sdkp->media_present = 1;
3981 	sdkp->write_prot = 0;
3982 	sdkp->cache_override = 0;
3983 	sdkp->WCE = 0;
3984 	sdkp->RCD = 0;
3985 	sdkp->ATO = 0;
3986 	sdkp->first_scan = 1;
3987 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3988 
3989 	sd_revalidate_disk(gd);
3990 
3991 	if (sdp->removable) {
3992 		gd->flags |= GENHD_FL_REMOVABLE;
3993 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3994 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3995 	}
3996 
3997 	blk_pm_runtime_init(sdp->request_queue, dev);
3998 	if (sdp->rpm_autosuspend) {
3999 		pm_runtime_set_autosuspend_delay(dev,
4000 			sdp->host->rpm_autosuspend_delay);
4001 	}
4002 
4003 	error = device_add_disk(dev, gd, NULL);
4004 	if (error) {
4005 		device_unregister(&sdkp->disk_dev);
4006 		put_disk(gd);
4007 		goto out;
4008 	}
4009 
4010 	if (sdkp->security) {
4011 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
4012 		if (sdkp->opal_dev)
4013 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
4014 	}
4015 
4016 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
4017 		  sdp->removable ? "removable " : "");
4018 	scsi_autopm_put_device(sdp);
4019 
4020 	return 0;
4021 
4022  out_free_index:
4023 	ida_free(&sd_index_ida, index);
4024  out_put:
4025 	put_disk(gd);
4026  out_free:
4027 	kfree(sdkp);
4028  out:
4029 	scsi_autopm_put_device(sdp);
4030 	return error;
4031 }
4032 
4033 /**
4034  *	sd_remove - called whenever a scsi disk (previously recognized by
4035  *	sd_probe) is detached from the system. It is called (potentially
4036  *	multiple times) during sd module unload.
4037  *	@dev: pointer to device object
4038  *
4039  *	Note: this function is invoked from the scsi mid-level.
4040  *	This function potentially frees up a device name (e.g. /dev/sdc)
4041  *	that could be re-used by a subsequent sd_probe().
4042  *	This function is not called when the built-in sd driver is "exit-ed".
4043  **/
4044 static int sd_remove(struct device *dev)
4045 {
4046 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4047 
4048 	scsi_autopm_get_device(sdkp->device);
4049 
4050 	device_del(&sdkp->disk_dev);
4051 	del_gendisk(sdkp->disk);
4052 	if (!sdkp->suspended)
4053 		sd_shutdown(dev);
4054 
4055 	put_disk(sdkp->disk);
4056 	return 0;
4057 }
4058 
4059 static void scsi_disk_release(struct device *dev)
4060 {
4061 	struct scsi_disk *sdkp = to_scsi_disk(dev);
4062 
4063 	ida_free(&sd_index_ida, sdkp->index);
4064 	put_device(&sdkp->device->sdev_gendev);
4065 	free_opal_dev(sdkp->opal_dev);
4066 
4067 	kfree(sdkp);
4068 }
4069 
4070 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
4071 {
4072 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
4073 	struct scsi_sense_hdr sshdr;
4074 	const struct scsi_exec_args exec_args = {
4075 		.sshdr = &sshdr,
4076 		.req_flags = BLK_MQ_REQ_PM,
4077 	};
4078 	struct scsi_device *sdp = sdkp->device;
4079 	int res;
4080 
4081 	if (start)
4082 		cmd[4] |= 1;	/* START */
4083 
4084 	if (sdp->start_stop_pwr_cond)
4085 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
4086 
4087 	if (!scsi_device_online(sdp))
4088 		return -ENODEV;
4089 
4090 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
4091 			       sdkp->max_retries, &exec_args);
4092 	if (res) {
4093 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
4094 		if (res > 0 && scsi_sense_valid(&sshdr)) {
4095 			sd_print_sense_hdr(sdkp, &sshdr);
4096 			/* 0x3a is medium not present */
4097 			if (sshdr.asc == 0x3a)
4098 				res = 0;
4099 		}
4100 	}
4101 
4102 	/* SCSI error codes must not go to the generic layer */
4103 	if (res)
4104 		return -EIO;
4105 
4106 	return 0;
4107 }
4108 
4109 /*
4110  * Send a SYNCHRONIZE CACHE instruction down to the device through
4111  * the normal SCSI command structure.  Wait for the command to
4112  * complete.
4113  */
4114 static void sd_shutdown(struct device *dev)
4115 {
4116 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4117 
4118 	if (!sdkp)
4119 		return;         /* this can happen */
4120 
4121 	if (pm_runtime_suspended(dev))
4122 		return;
4123 
4124 	if (sdkp->WCE && sdkp->media_present) {
4125 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4126 		sd_sync_cache(sdkp);
4127 	}
4128 
4129 	if ((system_state != SYSTEM_RESTART &&
4130 	     sdkp->device->manage_system_start_stop) ||
4131 	    (system_state == SYSTEM_POWER_OFF &&
4132 	     sdkp->device->manage_shutdown)) {
4133 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4134 		sd_start_stop_device(sdkp, 0);
4135 	}
4136 }
4137 
4138 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
4139 {
4140 	return (sdev->manage_system_start_stop && !runtime) ||
4141 		(sdev->manage_runtime_start_stop && runtime);
4142 }
4143 
4144 static int sd_suspend_common(struct device *dev, bool runtime)
4145 {
4146 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4147 	int ret = 0;
4148 
4149 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
4150 		return 0;
4151 
4152 	if (sdkp->WCE && sdkp->media_present) {
4153 		if (!sdkp->device->silence_suspend)
4154 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4155 		ret = sd_sync_cache(sdkp);
4156 		/* ignore OFFLINE device */
4157 		if (ret == -ENODEV)
4158 			return 0;
4159 
4160 		if (ret)
4161 			return ret;
4162 	}
4163 
4164 	if (sd_do_start_stop(sdkp->device, runtime)) {
4165 		if (!sdkp->device->silence_suspend)
4166 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4167 		/* an error is not worth aborting a system sleep */
4168 		ret = sd_start_stop_device(sdkp, 0);
4169 		if (!runtime)
4170 			ret = 0;
4171 	}
4172 
4173 	if (!ret)
4174 		sdkp->suspended = true;
4175 
4176 	return ret;
4177 }
4178 
4179 static int sd_suspend_system(struct device *dev)
4180 {
4181 	if (pm_runtime_suspended(dev))
4182 		return 0;
4183 
4184 	return sd_suspend_common(dev, false);
4185 }
4186 
4187 static int sd_suspend_runtime(struct device *dev)
4188 {
4189 	return sd_suspend_common(dev, true);
4190 }
4191 
4192 static int sd_resume(struct device *dev)
4193 {
4194 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4195 
4196 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4197 
4198 	if (opal_unlock_from_suspend(sdkp->opal_dev)) {
4199 		sd_printk(KERN_NOTICE, sdkp, "OPAL unlock failed\n");
4200 		return -EIO;
4201 	}
4202 
4203 	return 0;
4204 }
4205 
4206 static int sd_resume_common(struct device *dev, bool runtime)
4207 {
4208 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4209 	int ret;
4210 
4211 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4212 		return 0;
4213 
4214 	if (!sd_do_start_stop(sdkp->device, runtime)) {
4215 		sdkp->suspended = false;
4216 		return 0;
4217 	}
4218 
4219 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4220 	ret = sd_start_stop_device(sdkp, 1);
4221 	if (!ret) {
4222 		sd_resume(dev);
4223 		sdkp->suspended = false;
4224 	}
4225 
4226 	return ret;
4227 }
4228 
4229 static int sd_resume_system(struct device *dev)
4230 {
4231 	if (pm_runtime_suspended(dev)) {
4232 		struct scsi_disk *sdkp = dev_get_drvdata(dev);
4233 		struct scsi_device *sdp = sdkp ? sdkp->device : NULL;
4234 
4235 		if (sdp && sdp->force_runtime_start_on_system_start)
4236 			pm_request_resume(dev);
4237 
4238 		return 0;
4239 	}
4240 
4241 	return sd_resume_common(dev, false);
4242 }
4243 
4244 static int sd_resume_runtime(struct device *dev)
4245 {
4246 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4247 	struct scsi_device *sdp;
4248 
4249 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4250 		return 0;
4251 
4252 	sdp = sdkp->device;
4253 
4254 	if (sdp->ignore_media_change) {
4255 		/* clear the device's sense data */
4256 		static const u8 cmd[10] = { REQUEST_SENSE };
4257 		const struct scsi_exec_args exec_args = {
4258 			.req_flags = BLK_MQ_REQ_PM,
4259 		};
4260 
4261 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
4262 				     sdp->request_queue->rq_timeout, 1,
4263 				     &exec_args))
4264 			sd_printk(KERN_NOTICE, sdkp,
4265 				  "Failed to clear sense data\n");
4266 	}
4267 
4268 	return sd_resume_common(dev, true);
4269 }
4270 
4271 static const struct dev_pm_ops sd_pm_ops = {
4272 	.suspend		= sd_suspend_system,
4273 	.resume			= sd_resume_system,
4274 	.poweroff		= sd_suspend_system,
4275 	.restore		= sd_resume_system,
4276 	.runtime_suspend	= sd_suspend_runtime,
4277 	.runtime_resume		= sd_resume_runtime,
4278 };
4279 
4280 static struct scsi_driver sd_template = {
4281 	.gendrv = {
4282 		.name		= "sd",
4283 		.probe		= sd_probe,
4284 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
4285 		.remove		= sd_remove,
4286 		.shutdown	= sd_shutdown,
4287 		.pm		= &sd_pm_ops,
4288 	},
4289 	.rescan			= sd_rescan,
4290 	.resume			= sd_resume,
4291 	.init_command		= sd_init_command,
4292 	.uninit_command		= sd_uninit_command,
4293 	.done			= sd_done,
4294 	.eh_action		= sd_eh_action,
4295 	.eh_reset		= sd_eh_reset,
4296 };
4297 
4298 /**
4299  *	init_sd - entry point for this driver (both when built in or when
4300  *	a module).
4301  *
4302  *	Note: this function registers this driver with the scsi mid-level.
4303  **/
4304 static int __init init_sd(void)
4305 {
4306 	int majors = 0, i, err;
4307 
4308 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
4309 
4310 	for (i = 0; i < SD_MAJORS; i++) {
4311 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
4312 			continue;
4313 		majors++;
4314 	}
4315 
4316 	if (!majors)
4317 		return -ENODEV;
4318 
4319 	err = class_register(&sd_disk_class);
4320 	if (err)
4321 		goto err_out;
4322 
4323 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4324 	if (!sd_page_pool) {
4325 		printk(KERN_ERR "sd: can't init discard page pool\n");
4326 		err = -ENOMEM;
4327 		goto err_out_class;
4328 	}
4329 
4330 	err = scsi_register_driver(&sd_template.gendrv);
4331 	if (err)
4332 		goto err_out_driver;
4333 
4334 	return 0;
4335 
4336 err_out_driver:
4337 	mempool_destroy(sd_page_pool);
4338 err_out_class:
4339 	class_unregister(&sd_disk_class);
4340 err_out:
4341 	for (i = 0; i < SD_MAJORS; i++)
4342 		unregister_blkdev(sd_major(i), "sd");
4343 	return err;
4344 }
4345 
4346 /**
4347  *	exit_sd - exit point for this driver (when it is a module).
4348  *
4349  *	Note: this function unregisters this driver from the scsi mid-level.
4350  **/
4351 static void __exit exit_sd(void)
4352 {
4353 	int i;
4354 
4355 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4356 
4357 	scsi_unregister_driver(&sd_template.gendrv);
4358 	mempool_destroy(sd_page_pool);
4359 
4360 	class_unregister(&sd_disk_class);
4361 
4362 	for (i = 0; i < SD_MAJORS; i++)
4363 		unregister_blkdev(sd_major(i), "sd");
4364 }
4365 
4366 module_init(init_sd);
4367 module_exit(exit_sd);
4368 
4369 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4370 {
4371 	scsi_print_sense_hdr(sdkp->device,
4372 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4373 }
4374 
4375 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4376 {
4377 	const char *hb_string = scsi_hostbyte_string(result);
4378 
4379 	if (hb_string)
4380 		sd_printk(KERN_INFO, sdkp,
4381 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4382 			  hb_string ? hb_string : "invalid",
4383 			  "DRIVER_OK");
4384 	else
4385 		sd_printk(KERN_INFO, sdkp,
4386 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4387 			  msg, host_byte(result), "DRIVER_OK");
4388 }
4389