xref: /linux/drivers/scsi/sd.c (revision 5e3b7009f116f684ac6b93d8924506154f3b1f6d)
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 {
125 	bool wc = false, fua = false;
126 
127 	if (sdkp->WCE) {
128 		wc = true;
129 		if (sdkp->DPOFUA)
130 			fua = true;
131 	}
132 
133 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
134 }
135 
136 static ssize_t
137 cache_type_store(struct device *dev, struct device_attribute *attr,
138 		 const char *buf, size_t count)
139 {
140 	int ct, rcd, wce, sp;
141 	struct scsi_disk *sdkp = to_scsi_disk(dev);
142 	struct scsi_device *sdp = sdkp->device;
143 	char buffer[64];
144 	char *buffer_data;
145 	struct scsi_mode_data data;
146 	struct scsi_sense_hdr sshdr;
147 	static const char temp[] = "temporary ";
148 	int len, ret;
149 
150 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
151 		/* no cache control on RBC devices; theoretically they
152 		 * can do it, but there's probably so many exceptions
153 		 * it's not worth the risk */
154 		return -EINVAL;
155 
156 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
157 		buf += sizeof(temp) - 1;
158 		sdkp->cache_override = 1;
159 	} else {
160 		sdkp->cache_override = 0;
161 	}
162 
163 	ct = sysfs_match_string(sd_cache_types, buf);
164 	if (ct < 0)
165 		return -EINVAL;
166 
167 	rcd = ct & 0x01 ? 1 : 0;
168 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
169 
170 	if (sdkp->cache_override) {
171 		sdkp->WCE = wce;
172 		sdkp->RCD = rcd;
173 		sd_set_flush_flag(sdkp);
174 		return count;
175 	}
176 
177 	if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
178 			    sdkp->max_retries, &data, NULL))
179 		return -EINVAL;
180 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
181 		  data.block_descriptor_length);
182 	buffer_data = buffer + data.header_length +
183 		data.block_descriptor_length;
184 	buffer_data[2] &= ~0x05;
185 	buffer_data[2] |= wce << 2 | rcd;
186 	sp = buffer_data[0] & 0x80 ? 1 : 0;
187 	buffer_data[0] &= ~0x80;
188 
189 	/*
190 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
191 	 * received mode parameter buffer before doing MODE SELECT.
192 	 */
193 	data.device_specific = 0;
194 
195 	ret = scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
196 			       sdkp->max_retries, &data, &sshdr);
197 	if (ret) {
198 		if (ret > 0 && scsi_sense_valid(&sshdr))
199 			sd_print_sense_hdr(sdkp, &sshdr);
200 		return -EINVAL;
201 	}
202 	sd_revalidate_disk(sdkp->disk);
203 	return count;
204 }
205 
206 static ssize_t
207 manage_start_stop_show(struct device *dev,
208 		       struct device_attribute *attr, char *buf)
209 {
210 	struct scsi_disk *sdkp = to_scsi_disk(dev);
211 	struct scsi_device *sdp = sdkp->device;
212 
213 	return sysfs_emit(buf, "%u\n",
214 			  sdp->manage_system_start_stop &&
215 			  sdp->manage_runtime_start_stop &&
216 			  sdp->manage_shutdown);
217 }
218 static DEVICE_ATTR_RO(manage_start_stop);
219 
220 static ssize_t
221 manage_system_start_stop_show(struct device *dev,
222 			      struct device_attribute *attr, char *buf)
223 {
224 	struct scsi_disk *sdkp = to_scsi_disk(dev);
225 	struct scsi_device *sdp = sdkp->device;
226 
227 	return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
228 }
229 
230 static ssize_t
231 manage_system_start_stop_store(struct device *dev,
232 			       struct device_attribute *attr,
233 			       const char *buf, size_t count)
234 {
235 	struct scsi_disk *sdkp = to_scsi_disk(dev);
236 	struct scsi_device *sdp = sdkp->device;
237 	bool v;
238 
239 	if (!capable(CAP_SYS_ADMIN))
240 		return -EACCES;
241 
242 	if (kstrtobool(buf, &v))
243 		return -EINVAL;
244 
245 	sdp->manage_system_start_stop = v;
246 
247 	return count;
248 }
249 static DEVICE_ATTR_RW(manage_system_start_stop);
250 
251 static ssize_t
252 manage_runtime_start_stop_show(struct device *dev,
253 			       struct device_attribute *attr, char *buf)
254 {
255 	struct scsi_disk *sdkp = to_scsi_disk(dev);
256 	struct scsi_device *sdp = sdkp->device;
257 
258 	return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
259 }
260 
261 static ssize_t
262 manage_runtime_start_stop_store(struct device *dev,
263 				struct device_attribute *attr,
264 				const char *buf, size_t count)
265 {
266 	struct scsi_disk *sdkp = to_scsi_disk(dev);
267 	struct scsi_device *sdp = sdkp->device;
268 	bool v;
269 
270 	if (!capable(CAP_SYS_ADMIN))
271 		return -EACCES;
272 
273 	if (kstrtobool(buf, &v))
274 		return -EINVAL;
275 
276 	sdp->manage_runtime_start_stop = v;
277 
278 	return count;
279 }
280 static DEVICE_ATTR_RW(manage_runtime_start_stop);
281 
282 static ssize_t manage_shutdown_show(struct device *dev,
283 				    struct device_attribute *attr, char *buf)
284 {
285 	struct scsi_disk *sdkp = to_scsi_disk(dev);
286 	struct scsi_device *sdp = sdkp->device;
287 
288 	return sysfs_emit(buf, "%u\n", sdp->manage_shutdown);
289 }
290 
291 static ssize_t manage_shutdown_store(struct device *dev,
292 				     struct device_attribute *attr,
293 				     const char *buf, size_t count)
294 {
295 	struct scsi_disk *sdkp = to_scsi_disk(dev);
296 	struct scsi_device *sdp = sdkp->device;
297 	bool v;
298 
299 	if (!capable(CAP_SYS_ADMIN))
300 		return -EACCES;
301 
302 	if (kstrtobool(buf, &v))
303 		return -EINVAL;
304 
305 	sdp->manage_shutdown = v;
306 
307 	return count;
308 }
309 static DEVICE_ATTR_RW(manage_shutdown);
310 
311 static ssize_t
312 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
313 {
314 	struct scsi_disk *sdkp = to_scsi_disk(dev);
315 
316 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
317 }
318 
319 static ssize_t
320 allow_restart_store(struct device *dev, struct device_attribute *attr,
321 		    const char *buf, size_t count)
322 {
323 	bool v;
324 	struct scsi_disk *sdkp = to_scsi_disk(dev);
325 	struct scsi_device *sdp = sdkp->device;
326 
327 	if (!capable(CAP_SYS_ADMIN))
328 		return -EACCES;
329 
330 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
331 		return -EINVAL;
332 
333 	if (kstrtobool(buf, &v))
334 		return -EINVAL;
335 
336 	sdp->allow_restart = v;
337 
338 	return count;
339 }
340 static DEVICE_ATTR_RW(allow_restart);
341 
342 static ssize_t
343 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
344 {
345 	struct scsi_disk *sdkp = to_scsi_disk(dev);
346 	int ct = sdkp->RCD + 2*sdkp->WCE;
347 
348 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
349 }
350 static DEVICE_ATTR_RW(cache_type);
351 
352 static ssize_t
353 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
354 {
355 	struct scsi_disk *sdkp = to_scsi_disk(dev);
356 
357 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
358 }
359 static DEVICE_ATTR_RO(FUA);
360 
361 static ssize_t
362 protection_type_show(struct device *dev, struct device_attribute *attr,
363 		     char *buf)
364 {
365 	struct scsi_disk *sdkp = to_scsi_disk(dev);
366 
367 	return sprintf(buf, "%u\n", sdkp->protection_type);
368 }
369 
370 static ssize_t
371 protection_type_store(struct device *dev, struct device_attribute *attr,
372 		      const char *buf, size_t count)
373 {
374 	struct scsi_disk *sdkp = to_scsi_disk(dev);
375 	unsigned int val;
376 	int err;
377 
378 	if (!capable(CAP_SYS_ADMIN))
379 		return -EACCES;
380 
381 	err = kstrtouint(buf, 10, &val);
382 
383 	if (err)
384 		return err;
385 
386 	if (val <= T10_PI_TYPE3_PROTECTION)
387 		sdkp->protection_type = val;
388 
389 	return count;
390 }
391 static DEVICE_ATTR_RW(protection_type);
392 
393 static ssize_t
394 protection_mode_show(struct device *dev, struct device_attribute *attr,
395 		     char *buf)
396 {
397 	struct scsi_disk *sdkp = to_scsi_disk(dev);
398 	struct scsi_device *sdp = sdkp->device;
399 	unsigned int dif, dix;
400 
401 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
402 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
403 
404 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
405 		dif = 0;
406 		dix = 1;
407 	}
408 
409 	if (!dif && !dix)
410 		return sprintf(buf, "none\n");
411 
412 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
413 }
414 static DEVICE_ATTR_RO(protection_mode);
415 
416 static ssize_t
417 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
418 {
419 	struct scsi_disk *sdkp = to_scsi_disk(dev);
420 
421 	return sprintf(buf, "%u\n", sdkp->ATO);
422 }
423 static DEVICE_ATTR_RO(app_tag_own);
424 
425 static ssize_t
426 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
427 		       char *buf)
428 {
429 	struct scsi_disk *sdkp = to_scsi_disk(dev);
430 
431 	return sprintf(buf, "%u\n", sdkp->lbpme);
432 }
433 static DEVICE_ATTR_RO(thin_provisioning);
434 
435 /* sysfs_match_string() requires dense arrays */
436 static const char *lbp_mode[] = {
437 	[SD_LBP_FULL]		= "full",
438 	[SD_LBP_UNMAP]		= "unmap",
439 	[SD_LBP_WS16]		= "writesame_16",
440 	[SD_LBP_WS10]		= "writesame_10",
441 	[SD_LBP_ZERO]		= "writesame_zero",
442 	[SD_LBP_DISABLE]	= "disabled",
443 };
444 
445 static ssize_t
446 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
447 		       char *buf)
448 {
449 	struct scsi_disk *sdkp = to_scsi_disk(dev);
450 
451 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
452 }
453 
454 static ssize_t
455 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
456 			const char *buf, size_t count)
457 {
458 	struct scsi_disk *sdkp = to_scsi_disk(dev);
459 	struct scsi_device *sdp = sdkp->device;
460 	struct queue_limits lim;
461 	int mode, err;
462 
463 	if (!capable(CAP_SYS_ADMIN))
464 		return -EACCES;
465 
466 	if (sdp->type != TYPE_DISK)
467 		return -EINVAL;
468 
469 	/* ignore the provisioning mode for ZBC devices */
470 	if (sd_is_zoned(sdkp))
471 		return count;
472 
473 	mode = sysfs_match_string(lbp_mode, buf);
474 	if (mode < 0)
475 		return -EINVAL;
476 
477 	lim = queue_limits_start_update(sdkp->disk->queue);
478 	sd_config_discard(sdkp, &lim, mode);
479 	blk_mq_freeze_queue(sdkp->disk->queue);
480 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
481 	blk_mq_unfreeze_queue(sdkp->disk->queue);
482 	if (err)
483 		return err;
484 	return count;
485 }
486 static DEVICE_ATTR_RW(provisioning_mode);
487 
488 /* sysfs_match_string() requires dense arrays */
489 static const char *zeroing_mode[] = {
490 	[SD_ZERO_WRITE]		= "write",
491 	[SD_ZERO_WS]		= "writesame",
492 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
493 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
494 };
495 
496 static ssize_t
497 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
498 		  char *buf)
499 {
500 	struct scsi_disk *sdkp = to_scsi_disk(dev);
501 
502 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
503 }
504 
505 static ssize_t
506 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
507 		   const char *buf, size_t count)
508 {
509 	struct scsi_disk *sdkp = to_scsi_disk(dev);
510 	int mode;
511 
512 	if (!capable(CAP_SYS_ADMIN))
513 		return -EACCES;
514 
515 	mode = sysfs_match_string(zeroing_mode, buf);
516 	if (mode < 0)
517 		return -EINVAL;
518 
519 	sdkp->zeroing_mode = mode;
520 
521 	return count;
522 }
523 static DEVICE_ATTR_RW(zeroing_mode);
524 
525 static ssize_t
526 max_medium_access_timeouts_show(struct device *dev,
527 				struct device_attribute *attr, char *buf)
528 {
529 	struct scsi_disk *sdkp = to_scsi_disk(dev);
530 
531 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
532 }
533 
534 static ssize_t
535 max_medium_access_timeouts_store(struct device *dev,
536 				 struct device_attribute *attr, const char *buf,
537 				 size_t count)
538 {
539 	struct scsi_disk *sdkp = to_scsi_disk(dev);
540 	int err;
541 
542 	if (!capable(CAP_SYS_ADMIN))
543 		return -EACCES;
544 
545 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
546 
547 	return err ? err : count;
548 }
549 static DEVICE_ATTR_RW(max_medium_access_timeouts);
550 
551 static ssize_t
552 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
553 			   char *buf)
554 {
555 	struct scsi_disk *sdkp = to_scsi_disk(dev);
556 
557 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
558 }
559 
560 static ssize_t
561 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
562 			    const char *buf, size_t count)
563 {
564 	struct scsi_disk *sdkp = to_scsi_disk(dev);
565 	struct scsi_device *sdp = sdkp->device;
566 	struct queue_limits lim;
567 	unsigned long max;
568 	int err;
569 
570 	if (!capable(CAP_SYS_ADMIN))
571 		return -EACCES;
572 
573 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
574 		return -EINVAL;
575 
576 	err = kstrtoul(buf, 10, &max);
577 
578 	if (err)
579 		return err;
580 
581 	if (max == 0)
582 		sdp->no_write_same = 1;
583 	else if (max <= SD_MAX_WS16_BLOCKS) {
584 		sdp->no_write_same = 0;
585 		sdkp->max_ws_blocks = max;
586 	}
587 
588 	lim = queue_limits_start_update(sdkp->disk->queue);
589 	sd_config_write_same(sdkp, &lim);
590 	blk_mq_freeze_queue(sdkp->disk->queue);
591 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
592 	blk_mq_unfreeze_queue(sdkp->disk->queue);
593 	if (err)
594 		return err;
595 	return count;
596 }
597 static DEVICE_ATTR_RW(max_write_same_blocks);
598 
599 static ssize_t
600 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
601 {
602 	struct scsi_disk *sdkp = to_scsi_disk(dev);
603 
604 	if (sdkp->device->type == TYPE_ZBC)
605 		return sprintf(buf, "host-managed\n");
606 	if (sdkp->zoned == 1)
607 		return sprintf(buf, "host-aware\n");
608 	if (sdkp->zoned == 2)
609 		return sprintf(buf, "drive-managed\n");
610 	return sprintf(buf, "none\n");
611 }
612 static DEVICE_ATTR_RO(zoned_cap);
613 
614 static ssize_t
615 max_retries_store(struct device *dev, struct device_attribute *attr,
616 		  const char *buf, size_t count)
617 {
618 	struct scsi_disk *sdkp = to_scsi_disk(dev);
619 	struct scsi_device *sdev = sdkp->device;
620 	int retries, err;
621 
622 	err = kstrtoint(buf, 10, &retries);
623 	if (err)
624 		return err;
625 
626 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
627 		sdkp->max_retries = retries;
628 		return count;
629 	}
630 
631 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
632 		    SD_MAX_RETRIES);
633 	return -EINVAL;
634 }
635 
636 static ssize_t
637 max_retries_show(struct device *dev, struct device_attribute *attr,
638 		 char *buf)
639 {
640 	struct scsi_disk *sdkp = to_scsi_disk(dev);
641 
642 	return sprintf(buf, "%d\n", sdkp->max_retries);
643 }
644 
645 static DEVICE_ATTR_RW(max_retries);
646 
647 static struct attribute *sd_disk_attrs[] = {
648 	&dev_attr_cache_type.attr,
649 	&dev_attr_FUA.attr,
650 	&dev_attr_allow_restart.attr,
651 	&dev_attr_manage_start_stop.attr,
652 	&dev_attr_manage_system_start_stop.attr,
653 	&dev_attr_manage_runtime_start_stop.attr,
654 	&dev_attr_manage_shutdown.attr,
655 	&dev_attr_protection_type.attr,
656 	&dev_attr_protection_mode.attr,
657 	&dev_attr_app_tag_own.attr,
658 	&dev_attr_thin_provisioning.attr,
659 	&dev_attr_provisioning_mode.attr,
660 	&dev_attr_zeroing_mode.attr,
661 	&dev_attr_max_write_same_blocks.attr,
662 	&dev_attr_max_medium_access_timeouts.attr,
663 	&dev_attr_zoned_cap.attr,
664 	&dev_attr_max_retries.attr,
665 	NULL,
666 };
667 ATTRIBUTE_GROUPS(sd_disk);
668 
669 static struct class sd_disk_class = {
670 	.name		= "scsi_disk",
671 	.dev_release	= scsi_disk_release,
672 	.dev_groups	= sd_disk_groups,
673 };
674 
675 /*
676  * Don't request a new module, as that could deadlock in multipath
677  * environment.
678  */
679 static void sd_default_probe(dev_t devt)
680 {
681 }
682 
683 /*
684  * Device no to disk mapping:
685  *
686  *       major         disc2     disc  p1
687  *   |............|.............|....|....| <- dev_t
688  *    31        20 19          8 7  4 3  0
689  *
690  * Inside a major, we have 16k disks, however mapped non-
691  * contiguously. The first 16 disks are for major0, the next
692  * ones with major1, ... Disk 256 is for major0 again, disk 272
693  * for major1, ...
694  * As we stay compatible with our numbering scheme, we can reuse
695  * the well-know SCSI majors 8, 65--71, 136--143.
696  */
697 static int sd_major(int major_idx)
698 {
699 	switch (major_idx) {
700 	case 0:
701 		return SCSI_DISK0_MAJOR;
702 	case 1 ... 7:
703 		return SCSI_DISK1_MAJOR + major_idx - 1;
704 	case 8 ... 15:
705 		return SCSI_DISK8_MAJOR + major_idx - 8;
706 	default:
707 		BUG();
708 		return 0;	/* shut up gcc */
709 	}
710 }
711 
712 #ifdef CONFIG_BLK_SED_OPAL
713 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
714 		size_t len, bool send)
715 {
716 	struct scsi_disk *sdkp = data;
717 	struct scsi_device *sdev = sdkp->device;
718 	u8 cdb[12] = { 0, };
719 	const struct scsi_exec_args exec_args = {
720 		.req_flags = BLK_MQ_REQ_PM,
721 	};
722 	int ret;
723 
724 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
725 	cdb[1] = secp;
726 	put_unaligned_be16(spsp, &cdb[2]);
727 	put_unaligned_be32(len, &cdb[6]);
728 
729 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
730 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
731 			       &exec_args);
732 	return ret <= 0 ? ret : -EIO;
733 }
734 #endif /* CONFIG_BLK_SED_OPAL */
735 
736 /*
737  * Look up the DIX operation based on whether the command is read or
738  * write and whether dix and dif are enabled.
739  */
740 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
741 {
742 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
743 	static const unsigned int ops[] = {	/* wrt dix dif */
744 		SCSI_PROT_NORMAL,		/*  0	0   0  */
745 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
746 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
747 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
748 		SCSI_PROT_NORMAL,		/*  1	0   0  */
749 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
750 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
751 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
752 	};
753 
754 	return ops[write << 2 | dix << 1 | dif];
755 }
756 
757 /*
758  * Returns a mask of the protection flags that are valid for a given DIX
759  * operation.
760  */
761 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
762 {
763 	static const unsigned int flag_mask[] = {
764 		[SCSI_PROT_NORMAL]		= 0,
765 
766 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
767 						  SCSI_PROT_GUARD_CHECK |
768 						  SCSI_PROT_REF_CHECK |
769 						  SCSI_PROT_REF_INCREMENT,
770 
771 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
772 						  SCSI_PROT_IP_CHECKSUM,
773 
774 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
775 						  SCSI_PROT_GUARD_CHECK |
776 						  SCSI_PROT_REF_CHECK |
777 						  SCSI_PROT_REF_INCREMENT |
778 						  SCSI_PROT_IP_CHECKSUM,
779 
780 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
781 						  SCSI_PROT_REF_INCREMENT,
782 
783 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
784 						  SCSI_PROT_REF_CHECK |
785 						  SCSI_PROT_REF_INCREMENT |
786 						  SCSI_PROT_IP_CHECKSUM,
787 
788 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
789 						  SCSI_PROT_GUARD_CHECK |
790 						  SCSI_PROT_REF_CHECK |
791 						  SCSI_PROT_REF_INCREMENT |
792 						  SCSI_PROT_IP_CHECKSUM,
793 	};
794 
795 	return flag_mask[prot_op];
796 }
797 
798 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
799 					   unsigned int dix, unsigned int dif)
800 {
801 	struct request *rq = scsi_cmd_to_rq(scmd);
802 	struct bio *bio = rq->bio;
803 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
804 	unsigned int protect = 0;
805 
806 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
807 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
808 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
809 
810 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
811 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
812 	}
813 
814 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
815 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
816 
817 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
818 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
819 	}
820 
821 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
822 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
823 
824 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
825 			protect = 3 << 5;	/* Disable target PI checking */
826 		else
827 			protect = 1 << 5;	/* Enable target PI checking */
828 	}
829 
830 	scsi_set_prot_op(scmd, prot_op);
831 	scsi_set_prot_type(scmd, dif);
832 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
833 
834 	return protect;
835 }
836 
837 static void sd_disable_discard(struct scsi_disk *sdkp)
838 {
839 	sdkp->provisioning_mode = SD_LBP_DISABLE;
840 	blk_queue_disable_discard(sdkp->disk->queue);
841 }
842 
843 static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
844 		unsigned int mode)
845 {
846 	unsigned int logical_block_size = sdkp->device->sector_size;
847 	unsigned int max_blocks = 0;
848 
849 	lim->discard_alignment = sdkp->unmap_alignment * logical_block_size;
850 	lim->discard_granularity = max(sdkp->physical_block_size,
851 			sdkp->unmap_granularity * logical_block_size);
852 	sdkp->provisioning_mode = mode;
853 
854 	switch (mode) {
855 
856 	case SD_LBP_FULL:
857 	case SD_LBP_DISABLE:
858 		break;
859 
860 	case SD_LBP_UNMAP:
861 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
862 					  (u32)SD_MAX_WS16_BLOCKS);
863 		break;
864 
865 	case SD_LBP_WS16:
866 		if (sdkp->device->unmap_limit_for_ws)
867 			max_blocks = sdkp->max_unmap_blocks;
868 		else
869 			max_blocks = sdkp->max_ws_blocks;
870 
871 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
872 		break;
873 
874 	case SD_LBP_WS10:
875 		if (sdkp->device->unmap_limit_for_ws)
876 			max_blocks = sdkp->max_unmap_blocks;
877 		else
878 			max_blocks = sdkp->max_ws_blocks;
879 
880 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
881 		break;
882 
883 	case SD_LBP_ZERO:
884 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
885 					  (u32)SD_MAX_WS10_BLOCKS);
886 		break;
887 	}
888 
889 	lim->max_hw_discard_sectors = max_blocks *
890 		(logical_block_size >> SECTOR_SHIFT);
891 }
892 
893 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
894 {
895 	struct page *page;
896 
897 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
898 	if (!page)
899 		return NULL;
900 	clear_highpage(page);
901 	bvec_set_page(&rq->special_vec, page, data_len, 0);
902 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
903 	return bvec_virt(&rq->special_vec);
904 }
905 
906 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
907 {
908 	struct scsi_device *sdp = cmd->device;
909 	struct request *rq = scsi_cmd_to_rq(cmd);
910 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
911 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
912 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
913 	unsigned int data_len = 24;
914 	char *buf;
915 
916 	buf = sd_set_special_bvec(rq, data_len);
917 	if (!buf)
918 		return BLK_STS_RESOURCE;
919 
920 	cmd->cmd_len = 10;
921 	cmd->cmnd[0] = UNMAP;
922 	cmd->cmnd[8] = 24;
923 
924 	put_unaligned_be16(6 + 16, &buf[0]);
925 	put_unaligned_be16(16, &buf[2]);
926 	put_unaligned_be64(lba, &buf[8]);
927 	put_unaligned_be32(nr_blocks, &buf[16]);
928 
929 	cmd->allowed = sdkp->max_retries;
930 	cmd->transfersize = data_len;
931 	rq->timeout = SD_TIMEOUT;
932 
933 	return scsi_alloc_sgtables(cmd);
934 }
935 
936 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
937 		bool unmap)
938 {
939 	struct scsi_device *sdp = cmd->device;
940 	struct request *rq = scsi_cmd_to_rq(cmd);
941 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
942 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
943 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
944 	u32 data_len = sdp->sector_size;
945 
946 	if (!sd_set_special_bvec(rq, data_len))
947 		return BLK_STS_RESOURCE;
948 
949 	cmd->cmd_len = 16;
950 	cmd->cmnd[0] = WRITE_SAME_16;
951 	if (unmap)
952 		cmd->cmnd[1] = 0x8; /* UNMAP */
953 	put_unaligned_be64(lba, &cmd->cmnd[2]);
954 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
955 
956 	cmd->allowed = sdkp->max_retries;
957 	cmd->transfersize = data_len;
958 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
959 
960 	return scsi_alloc_sgtables(cmd);
961 }
962 
963 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
964 		bool unmap)
965 {
966 	struct scsi_device *sdp = cmd->device;
967 	struct request *rq = scsi_cmd_to_rq(cmd);
968 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
969 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
970 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
971 	u32 data_len = sdp->sector_size;
972 
973 	if (!sd_set_special_bvec(rq, data_len))
974 		return BLK_STS_RESOURCE;
975 
976 	cmd->cmd_len = 10;
977 	cmd->cmnd[0] = WRITE_SAME;
978 	if (unmap)
979 		cmd->cmnd[1] = 0x8; /* UNMAP */
980 	put_unaligned_be32(lba, &cmd->cmnd[2]);
981 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
982 
983 	cmd->allowed = sdkp->max_retries;
984 	cmd->transfersize = data_len;
985 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
986 
987 	return scsi_alloc_sgtables(cmd);
988 }
989 
990 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
991 {
992 	struct request *rq = scsi_cmd_to_rq(cmd);
993 	struct scsi_device *sdp = cmd->device;
994 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
995 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
996 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
997 
998 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
999 		switch (sdkp->zeroing_mode) {
1000 		case SD_ZERO_WS16_UNMAP:
1001 			return sd_setup_write_same16_cmnd(cmd, true);
1002 		case SD_ZERO_WS10_UNMAP:
1003 			return sd_setup_write_same10_cmnd(cmd, true);
1004 		}
1005 	}
1006 
1007 	if (sdp->no_write_same) {
1008 		rq->rq_flags |= RQF_QUIET;
1009 		return BLK_STS_TARGET;
1010 	}
1011 
1012 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
1013 		return sd_setup_write_same16_cmnd(cmd, false);
1014 
1015 	return sd_setup_write_same10_cmnd(cmd, false);
1016 }
1017 
1018 static void sd_disable_write_same(struct scsi_disk *sdkp)
1019 {
1020 	sdkp->device->no_write_same = 1;
1021 	sdkp->max_ws_blocks = 0;
1022 	blk_queue_disable_write_zeroes(sdkp->disk->queue);
1023 }
1024 
1025 static void sd_config_write_same(struct scsi_disk *sdkp,
1026 		struct queue_limits *lim)
1027 {
1028 	unsigned int logical_block_size = sdkp->device->sector_size;
1029 
1030 	if (sdkp->device->no_write_same) {
1031 		sdkp->max_ws_blocks = 0;
1032 		goto out;
1033 	}
1034 
1035 	/* Some devices can not handle block counts above 0xffff despite
1036 	 * supporting WRITE SAME(16). Consequently we default to 64k
1037 	 * blocks per I/O unless the device explicitly advertises a
1038 	 * bigger limit.
1039 	 */
1040 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
1041 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1042 						   (u32)SD_MAX_WS16_BLOCKS);
1043 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
1044 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1045 						   (u32)SD_MAX_WS10_BLOCKS);
1046 	else {
1047 		sdkp->device->no_write_same = 1;
1048 		sdkp->max_ws_blocks = 0;
1049 	}
1050 
1051 	if (sdkp->lbprz && sdkp->lbpws)
1052 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
1053 	else if (sdkp->lbprz && sdkp->lbpws10)
1054 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
1055 	else if (sdkp->max_ws_blocks)
1056 		sdkp->zeroing_mode = SD_ZERO_WS;
1057 	else
1058 		sdkp->zeroing_mode = SD_ZERO_WRITE;
1059 
1060 	if (sdkp->max_ws_blocks &&
1061 	    sdkp->physical_block_size > logical_block_size) {
1062 		/*
1063 		 * Reporting a maximum number of blocks that is not aligned
1064 		 * on the device physical size would cause a large write same
1065 		 * request to be split into physically unaligned chunks by
1066 		 * __blkdev_issue_write_zeroes() even if the caller of this
1067 		 * functions took care to align the large request. So make sure
1068 		 * the maximum reported is aligned to the device physical block
1069 		 * size. This is only an optional optimization for regular
1070 		 * disks, but this is mandatory to avoid failure of large write
1071 		 * same requests directed at sequential write required zones of
1072 		 * host-managed ZBC disks.
1073 		 */
1074 		sdkp->max_ws_blocks =
1075 			round_down(sdkp->max_ws_blocks,
1076 				   bytes_to_logical(sdkp->device,
1077 						    sdkp->physical_block_size));
1078 	}
1079 
1080 out:
1081 	lim->max_write_zeroes_sectors =
1082 		sdkp->max_ws_blocks * (logical_block_size >> SECTOR_SHIFT);
1083 }
1084 
1085 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1086 {
1087 	struct request *rq = scsi_cmd_to_rq(cmd);
1088 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1089 
1090 	/* flush requests don't perform I/O, zero the S/G table */
1091 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1092 
1093 	if (cmd->device->use_16_for_sync) {
1094 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1095 		cmd->cmd_len = 16;
1096 	} else {
1097 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1098 		cmd->cmd_len = 10;
1099 	}
1100 	cmd->transfersize = 0;
1101 	cmd->allowed = sdkp->max_retries;
1102 
1103 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1104 	return BLK_STS_OK;
1105 }
1106 
1107 /**
1108  * sd_group_number() - Compute the GROUP NUMBER field
1109  * @cmd: SCSI command for which to compute the value of the six-bit GROUP NUMBER
1110  *	field.
1111  *
1112  * From SBC-5 r05 (https://www.t10.org/cgi-bin/ac.pl?t=f&f=sbc5r05.pdf):
1113  * 0: no relative lifetime.
1114  * 1: shortest relative lifetime.
1115  * 2: second shortest relative lifetime.
1116  * 3 - 0x3d: intermediate relative lifetimes.
1117  * 0x3e: second longest relative lifetime.
1118  * 0x3f: longest relative lifetime.
1119  */
1120 static u8 sd_group_number(struct scsi_cmnd *cmd)
1121 {
1122 	const struct request *rq = scsi_cmd_to_rq(cmd);
1123 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1124 
1125 	if (!sdkp->rscs)
1126 		return 0;
1127 
1128 	return min3((u32)rq->write_hint, (u32)sdkp->permanent_stream_count,
1129 		    0x3fu);
1130 }
1131 
1132 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1133 				       sector_t lba, unsigned int nr_blocks,
1134 				       unsigned char flags, unsigned int dld)
1135 {
1136 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1137 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1138 	cmd->cmnd[6]  = sd_group_number(cmd);
1139 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1140 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1141 	cmd->cmnd[10] = flags;
1142 	cmd->cmnd[11] = dld & 0x07;
1143 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1144 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1145 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1146 
1147 	return BLK_STS_OK;
1148 }
1149 
1150 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1151 				       sector_t lba, unsigned int nr_blocks,
1152 				       unsigned char flags, unsigned int dld)
1153 {
1154 	cmd->cmd_len  = 16;
1155 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1156 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1157 	cmd->cmnd[14] = ((dld & 0x03) << 6) | sd_group_number(cmd);
1158 	cmd->cmnd[15] = 0;
1159 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1160 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1161 
1162 	return BLK_STS_OK;
1163 }
1164 
1165 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1166 				       sector_t lba, unsigned int nr_blocks,
1167 				       unsigned char flags)
1168 {
1169 	cmd->cmd_len = 10;
1170 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1171 	cmd->cmnd[1] = flags;
1172 	cmd->cmnd[6] = sd_group_number(cmd);
1173 	cmd->cmnd[9] = 0;
1174 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1175 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1176 
1177 	return BLK_STS_OK;
1178 }
1179 
1180 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1181 				      sector_t lba, unsigned int nr_blocks,
1182 				      unsigned char flags)
1183 {
1184 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1185 	if (WARN_ON_ONCE(nr_blocks == 0))
1186 		return BLK_STS_IOERR;
1187 
1188 	if (unlikely(flags & 0x8)) {
1189 		/*
1190 		 * This happens only if this drive failed 10byte rw
1191 		 * command with ILLEGAL_REQUEST during operation and
1192 		 * thus turned off use_10_for_rw.
1193 		 */
1194 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1195 		return BLK_STS_IOERR;
1196 	}
1197 
1198 	cmd->cmd_len = 6;
1199 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1200 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1201 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1202 	cmd->cmnd[3] = lba & 0xff;
1203 	cmd->cmnd[4] = nr_blocks;
1204 	cmd->cmnd[5] = 0;
1205 
1206 	return BLK_STS_OK;
1207 }
1208 
1209 /*
1210  * Check if a command has a duration limit set. If it does, and the target
1211  * device supports CDL and the feature is enabled, return the limit
1212  * descriptor index to use. Return 0 (no limit) otherwise.
1213  */
1214 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1215 {
1216 	struct scsi_device *sdp = sdkp->device;
1217 	int hint;
1218 
1219 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1220 		return 0;
1221 
1222 	/*
1223 	 * Use "no limit" if the request ioprio does not specify a duration
1224 	 * limit hint.
1225 	 */
1226 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1227 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1228 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1229 		return 0;
1230 
1231 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1232 }
1233 
1234 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1235 {
1236 	struct request *rq = scsi_cmd_to_rq(cmd);
1237 	struct scsi_device *sdp = cmd->device;
1238 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1239 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1240 	sector_t threshold;
1241 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1242 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1243 	bool write = rq_data_dir(rq) == WRITE;
1244 	unsigned char protect, fua;
1245 	unsigned int dld;
1246 	blk_status_t ret;
1247 	unsigned int dif;
1248 	bool dix;
1249 
1250 	ret = scsi_alloc_sgtables(cmd);
1251 	if (ret != BLK_STS_OK)
1252 		return ret;
1253 
1254 	ret = BLK_STS_IOERR;
1255 	if (!scsi_device_online(sdp) || sdp->changed) {
1256 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1257 		goto fail;
1258 	}
1259 
1260 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1261 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1262 		goto fail;
1263 	}
1264 
1265 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1266 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1267 		goto fail;
1268 	}
1269 
1270 	/*
1271 	 * Some SD card readers can't handle accesses which touch the
1272 	 * last one or two logical blocks. Split accesses as needed.
1273 	 */
1274 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1275 
1276 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1277 		if (lba < threshold) {
1278 			/* Access up to the threshold but not beyond */
1279 			nr_blocks = threshold - lba;
1280 		} else {
1281 			/* Access only a single logical block */
1282 			nr_blocks = 1;
1283 		}
1284 	}
1285 
1286 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1287 	dix = scsi_prot_sg_count(cmd);
1288 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1289 	dld = sd_cdl_dld(sdkp, cmd);
1290 
1291 	if (dif || dix)
1292 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1293 	else
1294 		protect = 0;
1295 
1296 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1297 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1298 					 protect | fua, dld);
1299 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1300 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1301 					 protect | fua, dld);
1302 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1303 		   sdp->use_10_for_rw || protect || rq->write_hint) {
1304 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1305 					 protect | fua);
1306 	} else {
1307 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1308 					protect | fua);
1309 	}
1310 
1311 	if (unlikely(ret != BLK_STS_OK))
1312 		goto fail;
1313 
1314 	/*
1315 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1316 	 * host adapter, it's safe to assume that we can at least transfer
1317 	 * this many bytes between each connect / disconnect.
1318 	 */
1319 	cmd->transfersize = sdp->sector_size;
1320 	cmd->underflow = nr_blocks << 9;
1321 	cmd->allowed = sdkp->max_retries;
1322 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1323 
1324 	SCSI_LOG_HLQUEUE(1,
1325 			 scmd_printk(KERN_INFO, cmd,
1326 				     "%s: block=%llu, count=%d\n", __func__,
1327 				     (unsigned long long)blk_rq_pos(rq),
1328 				     blk_rq_sectors(rq)));
1329 	SCSI_LOG_HLQUEUE(2,
1330 			 scmd_printk(KERN_INFO, cmd,
1331 				     "%s %d/%u 512 byte blocks.\n",
1332 				     write ? "writing" : "reading", nr_blocks,
1333 				     blk_rq_sectors(rq)));
1334 
1335 	/*
1336 	 * This indicates that the command is ready from our end to be queued.
1337 	 */
1338 	return BLK_STS_OK;
1339 fail:
1340 	scsi_free_sgtables(cmd);
1341 	return ret;
1342 }
1343 
1344 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1345 {
1346 	struct request *rq = scsi_cmd_to_rq(cmd);
1347 
1348 	switch (req_op(rq)) {
1349 	case REQ_OP_DISCARD:
1350 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1351 		case SD_LBP_UNMAP:
1352 			return sd_setup_unmap_cmnd(cmd);
1353 		case SD_LBP_WS16:
1354 			return sd_setup_write_same16_cmnd(cmd, true);
1355 		case SD_LBP_WS10:
1356 			return sd_setup_write_same10_cmnd(cmd, true);
1357 		case SD_LBP_ZERO:
1358 			return sd_setup_write_same10_cmnd(cmd, false);
1359 		default:
1360 			return BLK_STS_TARGET;
1361 		}
1362 	case REQ_OP_WRITE_ZEROES:
1363 		return sd_setup_write_zeroes_cmnd(cmd);
1364 	case REQ_OP_FLUSH:
1365 		return sd_setup_flush_cmnd(cmd);
1366 	case REQ_OP_READ:
1367 	case REQ_OP_WRITE:
1368 		return sd_setup_read_write_cmnd(cmd);
1369 	case REQ_OP_ZONE_RESET:
1370 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1371 						   false);
1372 	case REQ_OP_ZONE_RESET_ALL:
1373 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1374 						   true);
1375 	case REQ_OP_ZONE_OPEN:
1376 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1377 	case REQ_OP_ZONE_CLOSE:
1378 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1379 	case REQ_OP_ZONE_FINISH:
1380 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1381 	default:
1382 		WARN_ON_ONCE(1);
1383 		return BLK_STS_NOTSUPP;
1384 	}
1385 }
1386 
1387 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1388 {
1389 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1390 
1391 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1392 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1393 }
1394 
1395 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1396 {
1397 	if (sdkp->device->removable || sdkp->write_prot) {
1398 		if (disk_check_media_change(disk))
1399 			return true;
1400 	}
1401 
1402 	/*
1403 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1404 	 * nothing to do with partitions, BLKRRPART is used to force a full
1405 	 * revalidate after things like a format for historical reasons.
1406 	 */
1407 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1408 }
1409 
1410 /**
1411  *	sd_open - open a scsi disk device
1412  *	@disk: disk to open
1413  *	@mode: open mode
1414  *
1415  *	Returns 0 if successful. Returns a negated errno value in case
1416  *	of error.
1417  *
1418  *	Note: This can be called from a user context (e.g. fsck(1) )
1419  *	or from within the kernel (e.g. as a result of a mount(1) ).
1420  *	In the latter case @inode and @filp carry an abridged amount
1421  *	of information as noted above.
1422  *
1423  *	Locking: called with disk->open_mutex held.
1424  **/
1425 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1426 {
1427 	struct scsi_disk *sdkp = scsi_disk(disk);
1428 	struct scsi_device *sdev = sdkp->device;
1429 	int retval;
1430 
1431 	if (scsi_device_get(sdev))
1432 		return -ENXIO;
1433 
1434 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1435 
1436 	/*
1437 	 * If the device is in error recovery, wait until it is done.
1438 	 * If the device is offline, then disallow any access to it.
1439 	 */
1440 	retval = -ENXIO;
1441 	if (!scsi_block_when_processing_errors(sdev))
1442 		goto error_out;
1443 
1444 	if (sd_need_revalidate(disk, sdkp))
1445 		sd_revalidate_disk(disk);
1446 
1447 	/*
1448 	 * If the drive is empty, just let the open fail.
1449 	 */
1450 	retval = -ENOMEDIUM;
1451 	if (sdev->removable && !sdkp->media_present &&
1452 	    !(mode & BLK_OPEN_NDELAY))
1453 		goto error_out;
1454 
1455 	/*
1456 	 * If the device has the write protect tab set, have the open fail
1457 	 * if the user expects to be able to write to the thing.
1458 	 */
1459 	retval = -EROFS;
1460 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1461 		goto error_out;
1462 
1463 	/*
1464 	 * It is possible that the disk changing stuff resulted in
1465 	 * the device being taken offline.  If this is the case,
1466 	 * report this to the user, and don't pretend that the
1467 	 * open actually succeeded.
1468 	 */
1469 	retval = -ENXIO;
1470 	if (!scsi_device_online(sdev))
1471 		goto error_out;
1472 
1473 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1474 		if (scsi_block_when_processing_errors(sdev))
1475 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1476 	}
1477 
1478 	return 0;
1479 
1480 error_out:
1481 	scsi_device_put(sdev);
1482 	return retval;
1483 }
1484 
1485 /**
1486  *	sd_release - invoked when the (last) close(2) is called on this
1487  *	scsi disk.
1488  *	@disk: disk to release
1489  *
1490  *	Returns 0.
1491  *
1492  *	Note: may block (uninterruptible) if error recovery is underway
1493  *	on this disk.
1494  *
1495  *	Locking: called with disk->open_mutex held.
1496  **/
1497 static void sd_release(struct gendisk *disk)
1498 {
1499 	struct scsi_disk *sdkp = scsi_disk(disk);
1500 	struct scsi_device *sdev = sdkp->device;
1501 
1502 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1503 
1504 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1505 		if (scsi_block_when_processing_errors(sdev))
1506 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1507 	}
1508 
1509 	scsi_device_put(sdev);
1510 }
1511 
1512 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1513 {
1514 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1515 	struct scsi_device *sdp = sdkp->device;
1516 	struct Scsi_Host *host = sdp->host;
1517 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1518 	int diskinfo[4];
1519 
1520 	/* default to most commonly used values */
1521 	diskinfo[0] = 0x40;	/* 1 << 6 */
1522 	diskinfo[1] = 0x20;	/* 1 << 5 */
1523 	diskinfo[2] = capacity >> 11;
1524 
1525 	/* override with calculated, extended default, or driver values */
1526 	if (host->hostt->bios_param)
1527 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1528 	else
1529 		scsicam_bios_param(bdev, capacity, diskinfo);
1530 
1531 	geo->heads = diskinfo[0];
1532 	geo->sectors = diskinfo[1];
1533 	geo->cylinders = diskinfo[2];
1534 	return 0;
1535 }
1536 
1537 /**
1538  *	sd_ioctl - process an ioctl
1539  *	@bdev: target block device
1540  *	@mode: open mode
1541  *	@cmd: ioctl command number
1542  *	@arg: this is third argument given to ioctl(2) system call.
1543  *	Often contains a pointer.
1544  *
1545  *	Returns 0 if successful (some ioctls return positive numbers on
1546  *	success as well). Returns a negated errno value in case of error.
1547  *
1548  *	Note: most ioctls are forward onto the block subsystem or further
1549  *	down in the scsi subsystem.
1550  **/
1551 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1552 		    unsigned int cmd, unsigned long arg)
1553 {
1554 	struct gendisk *disk = bdev->bd_disk;
1555 	struct scsi_disk *sdkp = scsi_disk(disk);
1556 	struct scsi_device *sdp = sdkp->device;
1557 	void __user *p = (void __user *)arg;
1558 	int error;
1559 
1560 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1561 				    "cmd=0x%x\n", disk->disk_name, cmd));
1562 
1563 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1564 		return -ENOIOCTLCMD;
1565 
1566 	/*
1567 	 * If we are in the middle of error recovery, don't let anyone
1568 	 * else try and use this device.  Also, if error recovery fails, it
1569 	 * may try and take the device offline, in which case all further
1570 	 * access to the device is prohibited.
1571 	 */
1572 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1573 			(mode & BLK_OPEN_NDELAY));
1574 	if (error)
1575 		return error;
1576 
1577 	if (is_sed_ioctl(cmd))
1578 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1579 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1580 }
1581 
1582 static void set_media_not_present(struct scsi_disk *sdkp)
1583 {
1584 	if (sdkp->media_present)
1585 		sdkp->device->changed = 1;
1586 
1587 	if (sdkp->device->removable) {
1588 		sdkp->media_present = 0;
1589 		sdkp->capacity = 0;
1590 	}
1591 }
1592 
1593 static int media_not_present(struct scsi_disk *sdkp,
1594 			     struct scsi_sense_hdr *sshdr)
1595 {
1596 	if (!scsi_sense_valid(sshdr))
1597 		return 0;
1598 
1599 	/* not invoked for commands that could return deferred errors */
1600 	switch (sshdr->sense_key) {
1601 	case UNIT_ATTENTION:
1602 	case NOT_READY:
1603 		/* medium not present */
1604 		if (sshdr->asc == 0x3A) {
1605 			set_media_not_present(sdkp);
1606 			return 1;
1607 		}
1608 	}
1609 	return 0;
1610 }
1611 
1612 /**
1613  *	sd_check_events - check media events
1614  *	@disk: kernel device descriptor
1615  *	@clearing: disk events currently being cleared
1616  *
1617  *	Returns mask of DISK_EVENT_*.
1618  *
1619  *	Note: this function is invoked from the block subsystem.
1620  **/
1621 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1622 {
1623 	struct scsi_disk *sdkp = disk->private_data;
1624 	struct scsi_device *sdp;
1625 	int retval;
1626 	bool disk_changed;
1627 
1628 	if (!sdkp)
1629 		return 0;
1630 
1631 	sdp = sdkp->device;
1632 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1633 
1634 	/*
1635 	 * If the device is offline, don't send any commands - just pretend as
1636 	 * if the command failed.  If the device ever comes back online, we
1637 	 * can deal with it then.  It is only because of unrecoverable errors
1638 	 * that we would ever take a device offline in the first place.
1639 	 */
1640 	if (!scsi_device_online(sdp)) {
1641 		set_media_not_present(sdkp);
1642 		goto out;
1643 	}
1644 
1645 	/*
1646 	 * Using TEST_UNIT_READY enables differentiation between drive with
1647 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1648 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1649 	 *
1650 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1651 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1652 	 * sd_revalidate() is called.
1653 	 */
1654 	if (scsi_block_when_processing_errors(sdp)) {
1655 		struct scsi_sense_hdr sshdr = { 0, };
1656 
1657 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1658 					      &sshdr);
1659 
1660 		/* failed to execute TUR, assume media not present */
1661 		if (retval < 0 || host_byte(retval)) {
1662 			set_media_not_present(sdkp);
1663 			goto out;
1664 		}
1665 
1666 		if (media_not_present(sdkp, &sshdr))
1667 			goto out;
1668 	}
1669 
1670 	/*
1671 	 * For removable scsi disk we have to recognise the presence
1672 	 * of a disk in the drive.
1673 	 */
1674 	if (!sdkp->media_present)
1675 		sdp->changed = 1;
1676 	sdkp->media_present = 1;
1677 out:
1678 	/*
1679 	 * sdp->changed is set under the following conditions:
1680 	 *
1681 	 *	Medium present state has changed in either direction.
1682 	 *	Device has indicated UNIT_ATTENTION.
1683 	 */
1684 	disk_changed = sdp->changed;
1685 	sdp->changed = 0;
1686 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1687 }
1688 
1689 static int sd_sync_cache(struct scsi_disk *sdkp)
1690 {
1691 	int res;
1692 	struct scsi_device *sdp = sdkp->device;
1693 	const int timeout = sdp->request_queue->rq_timeout
1694 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1695 	/* Leave the rest of the command zero to indicate flush everything. */
1696 	const unsigned char cmd[16] = { sdp->use_16_for_sync ?
1697 				SYNCHRONIZE_CACHE_16 : SYNCHRONIZE_CACHE };
1698 	struct scsi_sense_hdr sshdr;
1699 	struct scsi_failure failure_defs[] = {
1700 		{
1701 			.allowed = 3,
1702 			.result = SCMD_FAILURE_RESULT_ANY,
1703 		},
1704 		{}
1705 	};
1706 	struct scsi_failures failures = {
1707 		.failure_definitions = failure_defs,
1708 	};
1709 	const struct scsi_exec_args exec_args = {
1710 		.req_flags = BLK_MQ_REQ_PM,
1711 		.sshdr = &sshdr,
1712 		.failures = &failures,
1713 	};
1714 
1715 	if (!scsi_device_online(sdp))
1716 		return -ENODEV;
1717 
1718 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, timeout,
1719 			       sdkp->max_retries, &exec_args);
1720 	if (res) {
1721 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1722 
1723 		if (res < 0)
1724 			return res;
1725 
1726 		if (scsi_status_is_check_condition(res) &&
1727 		    scsi_sense_valid(&sshdr)) {
1728 			sd_print_sense_hdr(sdkp, &sshdr);
1729 
1730 			/* we need to evaluate the error return  */
1731 			if (sshdr.asc == 0x3a ||	/* medium not present */
1732 			    sshdr.asc == 0x20 ||	/* invalid command */
1733 			    (sshdr.asc == 0x74 && sshdr.ascq == 0x71))	/* drive is password locked */
1734 				/* this is no error here */
1735 				return 0;
1736 			/*
1737 			 * This drive doesn't support sync and there's not much
1738 			 * we can do because this is called during shutdown
1739 			 * or suspend so just return success so those operations
1740 			 * can proceed.
1741 			 */
1742 			if (sshdr.sense_key == ILLEGAL_REQUEST)
1743 				return 0;
1744 		}
1745 
1746 		switch (host_byte(res)) {
1747 		/* ignore errors due to racing a disconnection */
1748 		case DID_BAD_TARGET:
1749 		case DID_NO_CONNECT:
1750 			return 0;
1751 		/* signal the upper layer it might try again */
1752 		case DID_BUS_BUSY:
1753 		case DID_IMM_RETRY:
1754 		case DID_REQUEUE:
1755 		case DID_SOFT_ERROR:
1756 			return -EBUSY;
1757 		default:
1758 			return -EIO;
1759 		}
1760 	}
1761 	return 0;
1762 }
1763 
1764 static void sd_rescan(struct device *dev)
1765 {
1766 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1767 
1768 	sd_revalidate_disk(sdkp->disk);
1769 }
1770 
1771 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1772 		enum blk_unique_id type)
1773 {
1774 	struct scsi_device *sdev = scsi_disk(disk)->device;
1775 	const struct scsi_vpd *vpd;
1776 	const unsigned char *d;
1777 	int ret = -ENXIO, len;
1778 
1779 	rcu_read_lock();
1780 	vpd = rcu_dereference(sdev->vpd_pg83);
1781 	if (!vpd)
1782 		goto out_unlock;
1783 
1784 	ret = -EINVAL;
1785 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1786 		/* we only care about designators with LU association */
1787 		if (((d[1] >> 4) & 0x3) != 0x00)
1788 			continue;
1789 		if ((d[1] & 0xf) != type)
1790 			continue;
1791 
1792 		/*
1793 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1794 		 * keep looking as one with more entropy might still show up.
1795 		 */
1796 		len = d[3];
1797 		if (len != 8 && len != 12 && len != 16)
1798 			continue;
1799 		ret = len;
1800 		memcpy(id, d + 4, len);
1801 		if (len == 16)
1802 			break;
1803 	}
1804 out_unlock:
1805 	rcu_read_unlock();
1806 	return ret;
1807 }
1808 
1809 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1810 {
1811 	switch (host_byte(result)) {
1812 	case DID_TRANSPORT_MARGINAL:
1813 	case DID_TRANSPORT_DISRUPTED:
1814 	case DID_BUS_BUSY:
1815 		return PR_STS_RETRY_PATH_FAILURE;
1816 	case DID_NO_CONNECT:
1817 		return PR_STS_PATH_FAILED;
1818 	case DID_TRANSPORT_FAILFAST:
1819 		return PR_STS_PATH_FAST_FAILED;
1820 	}
1821 
1822 	switch (status_byte(result)) {
1823 	case SAM_STAT_RESERVATION_CONFLICT:
1824 		return PR_STS_RESERVATION_CONFLICT;
1825 	case SAM_STAT_CHECK_CONDITION:
1826 		if (!scsi_sense_valid(sshdr))
1827 			return PR_STS_IOERR;
1828 
1829 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1830 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1831 			return -EINVAL;
1832 
1833 		fallthrough;
1834 	default:
1835 		return PR_STS_IOERR;
1836 	}
1837 }
1838 
1839 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1840 			    unsigned char *data, int data_len)
1841 {
1842 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1843 	struct scsi_device *sdev = sdkp->device;
1844 	struct scsi_sense_hdr sshdr;
1845 	u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1846 	struct scsi_failure failure_defs[] = {
1847 		{
1848 			.sense = UNIT_ATTENTION,
1849 			.asc = SCMD_FAILURE_ASC_ANY,
1850 			.ascq = SCMD_FAILURE_ASCQ_ANY,
1851 			.allowed = 5,
1852 			.result = SAM_STAT_CHECK_CONDITION,
1853 		},
1854 		{}
1855 	};
1856 	struct scsi_failures failures = {
1857 		.failure_definitions = failure_defs,
1858 	};
1859 	const struct scsi_exec_args exec_args = {
1860 		.sshdr = &sshdr,
1861 		.failures = &failures,
1862 	};
1863 	int result;
1864 
1865 	put_unaligned_be16(data_len, &cmd[7]);
1866 
1867 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1868 				  SD_TIMEOUT, sdkp->max_retries, &exec_args);
1869 	if (scsi_status_is_check_condition(result) &&
1870 	    scsi_sense_valid(&sshdr)) {
1871 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1872 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1873 	}
1874 
1875 	if (result <= 0)
1876 		return result;
1877 
1878 	return sd_scsi_to_pr_err(&sshdr, result);
1879 }
1880 
1881 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1882 {
1883 	int result, i, data_offset, num_copy_keys;
1884 	u32 num_keys = keys_info->num_keys;
1885 	int data_len = num_keys * 8 + 8;
1886 	u8 *data;
1887 
1888 	data = kzalloc(data_len, GFP_KERNEL);
1889 	if (!data)
1890 		return -ENOMEM;
1891 
1892 	result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1893 	if (result)
1894 		goto free_data;
1895 
1896 	keys_info->generation = get_unaligned_be32(&data[0]);
1897 	keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1898 
1899 	data_offset = 8;
1900 	num_copy_keys = min(num_keys, keys_info->num_keys);
1901 
1902 	for (i = 0; i < num_copy_keys; i++) {
1903 		keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1904 		data_offset += 8;
1905 	}
1906 
1907 free_data:
1908 	kfree(data);
1909 	return result;
1910 }
1911 
1912 static int sd_pr_read_reservation(struct block_device *bdev,
1913 				  struct pr_held_reservation *rsv)
1914 {
1915 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1916 	struct scsi_device *sdev = sdkp->device;
1917 	u8 data[24] = { };
1918 	int result, len;
1919 
1920 	result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1921 	if (result)
1922 		return result;
1923 
1924 	len = get_unaligned_be32(&data[4]);
1925 	if (!len)
1926 		return 0;
1927 
1928 	/* Make sure we have at least the key and type */
1929 	if (len < 14) {
1930 		sdev_printk(KERN_INFO, sdev,
1931 			    "READ RESERVATION failed due to short return buffer of %d bytes\n",
1932 			    len);
1933 		return -EINVAL;
1934 	}
1935 
1936 	rsv->generation = get_unaligned_be32(&data[0]);
1937 	rsv->key = get_unaligned_be64(&data[8]);
1938 	rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1939 	return 0;
1940 }
1941 
1942 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1943 			     u64 sa_key, enum scsi_pr_type type, u8 flags)
1944 {
1945 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1946 	struct scsi_device *sdev = sdkp->device;
1947 	struct scsi_sense_hdr sshdr;
1948 	struct scsi_failure failure_defs[] = {
1949 		{
1950 			.sense = UNIT_ATTENTION,
1951 			.asc = SCMD_FAILURE_ASC_ANY,
1952 			.ascq = SCMD_FAILURE_ASCQ_ANY,
1953 			.allowed = 5,
1954 			.result = SAM_STAT_CHECK_CONDITION,
1955 		},
1956 		{}
1957 	};
1958 	struct scsi_failures failures = {
1959 		.failure_definitions = failure_defs,
1960 	};
1961 	const struct scsi_exec_args exec_args = {
1962 		.sshdr = &sshdr,
1963 		.failures = &failures,
1964 	};
1965 	int result;
1966 	u8 cmd[16] = { 0, };
1967 	u8 data[24] = { 0, };
1968 
1969 	cmd[0] = PERSISTENT_RESERVE_OUT;
1970 	cmd[1] = sa;
1971 	cmd[2] = type;
1972 	put_unaligned_be32(sizeof(data), &cmd[5]);
1973 
1974 	put_unaligned_be64(key, &data[0]);
1975 	put_unaligned_be64(sa_key, &data[8]);
1976 	data[20] = flags;
1977 
1978 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1979 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1980 				  &exec_args);
1981 
1982 	if (scsi_status_is_check_condition(result) &&
1983 	    scsi_sense_valid(&sshdr)) {
1984 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1985 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1986 	}
1987 
1988 	if (result <= 0)
1989 		return result;
1990 
1991 	return sd_scsi_to_pr_err(&sshdr, result);
1992 }
1993 
1994 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1995 		u32 flags)
1996 {
1997 	if (flags & ~PR_FL_IGNORE_KEY)
1998 		return -EOPNOTSUPP;
1999 	return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
2000 			old_key, new_key, 0,
2001 			(1 << 0) /* APTPL */);
2002 }
2003 
2004 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2005 		u32 flags)
2006 {
2007 	if (flags)
2008 		return -EOPNOTSUPP;
2009 	return sd_pr_out_command(bdev, 0x01, key, 0,
2010 				 block_pr_type_to_scsi(type), 0);
2011 }
2012 
2013 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
2014 {
2015 	return sd_pr_out_command(bdev, 0x02, key, 0,
2016 				 block_pr_type_to_scsi(type), 0);
2017 }
2018 
2019 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2020 		enum pr_type type, bool abort)
2021 {
2022 	return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
2023 				 block_pr_type_to_scsi(type), 0);
2024 }
2025 
2026 static int sd_pr_clear(struct block_device *bdev, u64 key)
2027 {
2028 	return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
2029 }
2030 
2031 static const struct pr_ops sd_pr_ops = {
2032 	.pr_register	= sd_pr_register,
2033 	.pr_reserve	= sd_pr_reserve,
2034 	.pr_release	= sd_pr_release,
2035 	.pr_preempt	= sd_pr_preempt,
2036 	.pr_clear	= sd_pr_clear,
2037 	.pr_read_keys	= sd_pr_read_keys,
2038 	.pr_read_reservation = sd_pr_read_reservation,
2039 };
2040 
2041 static void scsi_disk_free_disk(struct gendisk *disk)
2042 {
2043 	struct scsi_disk *sdkp = scsi_disk(disk);
2044 
2045 	put_device(&sdkp->disk_dev);
2046 }
2047 
2048 static const struct block_device_operations sd_fops = {
2049 	.owner			= THIS_MODULE,
2050 	.open			= sd_open,
2051 	.release		= sd_release,
2052 	.ioctl			= sd_ioctl,
2053 	.getgeo			= sd_getgeo,
2054 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
2055 	.check_events		= sd_check_events,
2056 	.unlock_native_capacity	= sd_unlock_native_capacity,
2057 	.report_zones		= sd_zbc_report_zones,
2058 	.get_unique_id		= sd_get_unique_id,
2059 	.free_disk		= scsi_disk_free_disk,
2060 	.pr_ops			= &sd_pr_ops,
2061 };
2062 
2063 /**
2064  *	sd_eh_reset - reset error handling callback
2065  *	@scmd:		sd-issued command that has failed
2066  *
2067  *	This function is called by the SCSI midlayer before starting
2068  *	SCSI EH. When counting medium access failures we have to be
2069  *	careful to register it only only once per device and SCSI EH run;
2070  *	there might be several timed out commands which will cause the
2071  *	'max_medium_access_timeouts' counter to trigger after the first
2072  *	SCSI EH run already and set the device to offline.
2073  *	So this function resets the internal counter before starting SCSI EH.
2074  **/
2075 static void sd_eh_reset(struct scsi_cmnd *scmd)
2076 {
2077 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2078 
2079 	/* New SCSI EH run, reset gate variable */
2080 	sdkp->ignore_medium_access_errors = false;
2081 }
2082 
2083 /**
2084  *	sd_eh_action - error handling callback
2085  *	@scmd:		sd-issued command that has failed
2086  *	@eh_disp:	The recovery disposition suggested by the midlayer
2087  *
2088  *	This function is called by the SCSI midlayer upon completion of an
2089  *	error test command (currently TEST UNIT READY). The result of sending
2090  *	the eh command is passed in eh_disp.  We're looking for devices that
2091  *	fail medium access commands but are OK with non access commands like
2092  *	test unit ready (so wrongly see the device as having a successful
2093  *	recovery)
2094  **/
2095 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
2096 {
2097 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2098 	struct scsi_device *sdev = scmd->device;
2099 
2100 	if (!scsi_device_online(sdev) ||
2101 	    !scsi_medium_access_command(scmd) ||
2102 	    host_byte(scmd->result) != DID_TIME_OUT ||
2103 	    eh_disp != SUCCESS)
2104 		return eh_disp;
2105 
2106 	/*
2107 	 * The device has timed out executing a medium access command.
2108 	 * However, the TEST UNIT READY command sent during error
2109 	 * handling completed successfully. Either the device is in the
2110 	 * process of recovering or has it suffered an internal failure
2111 	 * that prevents access to the storage medium.
2112 	 */
2113 	if (!sdkp->ignore_medium_access_errors) {
2114 		sdkp->medium_access_timed_out++;
2115 		sdkp->ignore_medium_access_errors = true;
2116 	}
2117 
2118 	/*
2119 	 * If the device keeps failing read/write commands but TEST UNIT
2120 	 * READY always completes successfully we assume that medium
2121 	 * access is no longer possible and take the device offline.
2122 	 */
2123 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2124 		scmd_printk(KERN_ERR, scmd,
2125 			    "Medium access timeout failure. Offlining disk!\n");
2126 		mutex_lock(&sdev->state_mutex);
2127 		scsi_device_set_state(sdev, SDEV_OFFLINE);
2128 		mutex_unlock(&sdev->state_mutex);
2129 
2130 		return SUCCESS;
2131 	}
2132 
2133 	return eh_disp;
2134 }
2135 
2136 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2137 {
2138 	struct request *req = scsi_cmd_to_rq(scmd);
2139 	struct scsi_device *sdev = scmd->device;
2140 	unsigned int transferred, good_bytes;
2141 	u64 start_lba, end_lba, bad_lba;
2142 
2143 	/*
2144 	 * Some commands have a payload smaller than the device logical
2145 	 * block size (e.g. INQUIRY on a 4K disk).
2146 	 */
2147 	if (scsi_bufflen(scmd) <= sdev->sector_size)
2148 		return 0;
2149 
2150 	/* Check if we have a 'bad_lba' information */
2151 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2152 				     SCSI_SENSE_BUFFERSIZE,
2153 				     &bad_lba))
2154 		return 0;
2155 
2156 	/*
2157 	 * If the bad lba was reported incorrectly, we have no idea where
2158 	 * the error is.
2159 	 */
2160 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2161 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2162 	if (bad_lba < start_lba || bad_lba >= end_lba)
2163 		return 0;
2164 
2165 	/*
2166 	 * resid is optional but mostly filled in.  When it's unused,
2167 	 * its value is zero, so we assume the whole buffer transferred
2168 	 */
2169 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2170 
2171 	/* This computation should always be done in terms of the
2172 	 * resolution of the device's medium.
2173 	 */
2174 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2175 
2176 	return min(good_bytes, transferred);
2177 }
2178 
2179 /**
2180  *	sd_done - bottom half handler: called when the lower level
2181  *	driver has completed (successfully or otherwise) a scsi command.
2182  *	@SCpnt: mid-level's per command structure.
2183  *
2184  *	Note: potentially run from within an ISR. Must not block.
2185  **/
2186 static int sd_done(struct scsi_cmnd *SCpnt)
2187 {
2188 	int result = SCpnt->result;
2189 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2190 	unsigned int sector_size = SCpnt->device->sector_size;
2191 	unsigned int resid;
2192 	struct scsi_sense_hdr sshdr;
2193 	struct request *req = scsi_cmd_to_rq(SCpnt);
2194 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2195 	int sense_valid = 0;
2196 	int sense_deferred = 0;
2197 
2198 	switch (req_op(req)) {
2199 	case REQ_OP_DISCARD:
2200 	case REQ_OP_WRITE_ZEROES:
2201 	case REQ_OP_ZONE_RESET:
2202 	case REQ_OP_ZONE_RESET_ALL:
2203 	case REQ_OP_ZONE_OPEN:
2204 	case REQ_OP_ZONE_CLOSE:
2205 	case REQ_OP_ZONE_FINISH:
2206 		if (!result) {
2207 			good_bytes = blk_rq_bytes(req);
2208 			scsi_set_resid(SCpnt, 0);
2209 		} else {
2210 			good_bytes = 0;
2211 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2212 		}
2213 		break;
2214 	default:
2215 		/*
2216 		 * In case of bogus fw or device, we could end up having
2217 		 * an unaligned partial completion. Check this here and force
2218 		 * alignment.
2219 		 */
2220 		resid = scsi_get_resid(SCpnt);
2221 		if (resid & (sector_size - 1)) {
2222 			sd_printk(KERN_INFO, sdkp,
2223 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2224 				resid, sector_size);
2225 			scsi_print_command(SCpnt);
2226 			resid = min(scsi_bufflen(SCpnt),
2227 				    round_up(resid, sector_size));
2228 			scsi_set_resid(SCpnt, resid);
2229 		}
2230 	}
2231 
2232 	if (result) {
2233 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2234 		if (sense_valid)
2235 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2236 	}
2237 	sdkp->medium_access_timed_out = 0;
2238 
2239 	if (!scsi_status_is_check_condition(result) &&
2240 	    (!sense_valid || sense_deferred))
2241 		goto out;
2242 
2243 	switch (sshdr.sense_key) {
2244 	case HARDWARE_ERROR:
2245 	case MEDIUM_ERROR:
2246 		good_bytes = sd_completed_bytes(SCpnt);
2247 		break;
2248 	case RECOVERED_ERROR:
2249 		good_bytes = scsi_bufflen(SCpnt);
2250 		break;
2251 	case NO_SENSE:
2252 		/* This indicates a false check condition, so ignore it.  An
2253 		 * unknown amount of data was transferred so treat it as an
2254 		 * error.
2255 		 */
2256 		SCpnt->result = 0;
2257 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2258 		break;
2259 	case ABORTED_COMMAND:
2260 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2261 			good_bytes = sd_completed_bytes(SCpnt);
2262 		break;
2263 	case ILLEGAL_REQUEST:
2264 		switch (sshdr.asc) {
2265 		case 0x10:	/* DIX: Host detected corruption */
2266 			good_bytes = sd_completed_bytes(SCpnt);
2267 			break;
2268 		case 0x20:	/* INVALID COMMAND OPCODE */
2269 		case 0x24:	/* INVALID FIELD IN CDB */
2270 			switch (SCpnt->cmnd[0]) {
2271 			case UNMAP:
2272 				sd_disable_discard(sdkp);
2273 				break;
2274 			case WRITE_SAME_16:
2275 			case WRITE_SAME:
2276 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2277 					sd_disable_discard(sdkp);
2278 				} else {
2279 					sd_disable_write_same(sdkp);
2280 					req->rq_flags |= RQF_QUIET;
2281 				}
2282 				break;
2283 			}
2284 		}
2285 		break;
2286 	default:
2287 		break;
2288 	}
2289 
2290  out:
2291 	if (sd_is_zoned(sdkp))
2292 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2293 
2294 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2295 					   "sd_done: completed %d of %d bytes\n",
2296 					   good_bytes, scsi_bufflen(SCpnt)));
2297 
2298 	return good_bytes;
2299 }
2300 
2301 /*
2302  * spinup disk - called only in sd_revalidate_disk()
2303  */
2304 static void
2305 sd_spinup_disk(struct scsi_disk *sdkp)
2306 {
2307 	static const u8 cmd[10] = { TEST_UNIT_READY };
2308 	unsigned long spintime_expire = 0;
2309 	int spintime, sense_valid = 0;
2310 	unsigned int the_result;
2311 	struct scsi_sense_hdr sshdr;
2312 	struct scsi_failure failure_defs[] = {
2313 		/* Do not retry Medium Not Present */
2314 		{
2315 			.sense = UNIT_ATTENTION,
2316 			.asc = 0x3A,
2317 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2318 			.result = SAM_STAT_CHECK_CONDITION,
2319 		},
2320 		{
2321 			.sense = NOT_READY,
2322 			.asc = 0x3A,
2323 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2324 			.result = SAM_STAT_CHECK_CONDITION,
2325 		},
2326 		/* Retry when scsi_status_is_good would return false 3 times */
2327 		{
2328 			.result = SCMD_FAILURE_STAT_ANY,
2329 			.allowed = 3,
2330 		},
2331 		{}
2332 	};
2333 	struct scsi_failures failures = {
2334 		.failure_definitions = failure_defs,
2335 	};
2336 	const struct scsi_exec_args exec_args = {
2337 		.sshdr = &sshdr,
2338 		.failures = &failures,
2339 	};
2340 
2341 	spintime = 0;
2342 
2343 	/* Spin up drives, as required.  Only do this at boot time */
2344 	/* Spinup needs to be done for module loads too. */
2345 	do {
2346 		bool media_was_present = sdkp->media_present;
2347 
2348 		scsi_failures_reset_retries(&failures);
2349 
2350 		the_result = scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN,
2351 					      NULL, 0, SD_TIMEOUT,
2352 					      sdkp->max_retries, &exec_args);
2353 
2354 
2355 		if (the_result > 0) {
2356 			/*
2357 			 * If the drive has indicated to us that it doesn't
2358 			 * have any media in it, don't bother with any more
2359 			 * polling.
2360 			 */
2361 			if (media_not_present(sdkp, &sshdr)) {
2362 				if (media_was_present)
2363 					sd_printk(KERN_NOTICE, sdkp,
2364 						  "Media removed, stopped polling\n");
2365 				return;
2366 			}
2367 			sense_valid = scsi_sense_valid(&sshdr);
2368 		}
2369 
2370 		if (!scsi_status_is_check_condition(the_result)) {
2371 			/* no sense, TUR either succeeded or failed
2372 			 * with a status error */
2373 			if(!spintime && !scsi_status_is_good(the_result)) {
2374 				sd_print_result(sdkp, "Test Unit Ready failed",
2375 						the_result);
2376 			}
2377 			break;
2378 		}
2379 
2380 		/*
2381 		 * The device does not want the automatic start to be issued.
2382 		 */
2383 		if (sdkp->device->no_start_on_add)
2384 			break;
2385 
2386 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2387 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2388 				break;	/* manual intervention required */
2389 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2390 				break;	/* standby */
2391 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2392 				break;	/* unavailable */
2393 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2394 				break;	/* sanitize in progress */
2395 			if (sshdr.asc == 4 && sshdr.ascq == 0x24)
2396 				break;	/* depopulation in progress */
2397 			if (sshdr.asc == 4 && sshdr.ascq == 0x25)
2398 				break;	/* depopulation restoration in progress */
2399 			/*
2400 			 * Issue command to spin up drive when not ready
2401 			 */
2402 			if (!spintime) {
2403 				/* Return immediately and start spin cycle */
2404 				const u8 start_cmd[10] = {
2405 					[0] = START_STOP,
2406 					[1] = 1,
2407 					[4] = sdkp->device->start_stop_pwr_cond ?
2408 						0x11 : 1,
2409 				};
2410 
2411 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2412 				scsi_execute_cmd(sdkp->device, start_cmd,
2413 						 REQ_OP_DRV_IN, NULL, 0,
2414 						 SD_TIMEOUT, sdkp->max_retries,
2415 						 &exec_args);
2416 				spintime_expire = jiffies + 100 * HZ;
2417 				spintime = 1;
2418 			}
2419 			/* Wait 1 second for next try */
2420 			msleep(1000);
2421 			printk(KERN_CONT ".");
2422 
2423 		/*
2424 		 * Wait for USB flash devices with slow firmware.
2425 		 * Yes, this sense key/ASC combination shouldn't
2426 		 * occur here.  It's characteristic of these devices.
2427 		 */
2428 		} else if (sense_valid &&
2429 				sshdr.sense_key == UNIT_ATTENTION &&
2430 				sshdr.asc == 0x28) {
2431 			if (!spintime) {
2432 				spintime_expire = jiffies + 5 * HZ;
2433 				spintime = 1;
2434 			}
2435 			/* Wait 1 second for next try */
2436 			msleep(1000);
2437 		} else {
2438 			/* we don't understand the sense code, so it's
2439 			 * probably pointless to loop */
2440 			if(!spintime) {
2441 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2442 				sd_print_sense_hdr(sdkp, &sshdr);
2443 			}
2444 			break;
2445 		}
2446 
2447 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2448 
2449 	if (spintime) {
2450 		if (scsi_status_is_good(the_result))
2451 			printk(KERN_CONT "ready\n");
2452 		else
2453 			printk(KERN_CONT "not responding...\n");
2454 	}
2455 }
2456 
2457 /*
2458  * Determine whether disk supports Data Integrity Field.
2459  */
2460 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2461 {
2462 	struct scsi_device *sdp = sdkp->device;
2463 	u8 type;
2464 
2465 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2466 		sdkp->protection_type = 0;
2467 		return 0;
2468 	}
2469 
2470 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2471 
2472 	if (type > T10_PI_TYPE3_PROTECTION) {
2473 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2474 			  " protection type %u. Disabling disk!\n",
2475 			  type);
2476 		sdkp->protection_type = 0;
2477 		return -ENODEV;
2478 	}
2479 
2480 	sdkp->protection_type = type;
2481 
2482 	return 0;
2483 }
2484 
2485 static void sd_config_protection(struct scsi_disk *sdkp,
2486 		struct queue_limits *lim)
2487 {
2488 	struct scsi_device *sdp = sdkp->device;
2489 
2490 	if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY))
2491 		sd_dif_config_host(sdkp, lim);
2492 
2493 	if (!sdkp->protection_type)
2494 		return;
2495 
2496 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2497 		sd_first_printk(KERN_NOTICE, sdkp,
2498 				"Disabling DIF Type %u protection\n",
2499 				sdkp->protection_type);
2500 		sdkp->protection_type = 0;
2501 	}
2502 
2503 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2504 			sdkp->protection_type);
2505 }
2506 
2507 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2508 			struct scsi_sense_hdr *sshdr, int sense_valid,
2509 			int the_result)
2510 {
2511 	if (sense_valid)
2512 		sd_print_sense_hdr(sdkp, sshdr);
2513 	else
2514 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2515 
2516 	/*
2517 	 * Set dirty bit for removable devices if not ready -
2518 	 * sometimes drives will not report this properly.
2519 	 */
2520 	if (sdp->removable &&
2521 	    sense_valid && sshdr->sense_key == NOT_READY)
2522 		set_media_not_present(sdkp);
2523 
2524 	/*
2525 	 * We used to set media_present to 0 here to indicate no media
2526 	 * in the drive, but some drives fail read capacity even with
2527 	 * media present, so we can't do that.
2528 	 */
2529 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2530 }
2531 
2532 #define RC16_LEN 32
2533 #if RC16_LEN > SD_BUF_SIZE
2534 #error RC16_LEN must not be more than SD_BUF_SIZE
2535 #endif
2536 
2537 #define READ_CAPACITY_RETRIES_ON_RESET	10
2538 
2539 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2540 		struct queue_limits *lim, unsigned char *buffer)
2541 {
2542 	unsigned char cmd[16];
2543 	struct scsi_sense_hdr sshdr;
2544 	const struct scsi_exec_args exec_args = {
2545 		.sshdr = &sshdr,
2546 	};
2547 	int sense_valid = 0;
2548 	int the_result;
2549 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2550 	unsigned int alignment;
2551 	unsigned long long lba;
2552 	unsigned sector_size;
2553 
2554 	if (sdp->no_read_capacity_16)
2555 		return -EINVAL;
2556 
2557 	do {
2558 		memset(cmd, 0, 16);
2559 		cmd[0] = SERVICE_ACTION_IN_16;
2560 		cmd[1] = SAI_READ_CAPACITY_16;
2561 		cmd[13] = RC16_LEN;
2562 		memset(buffer, 0, RC16_LEN);
2563 
2564 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2565 					      buffer, RC16_LEN, SD_TIMEOUT,
2566 					      sdkp->max_retries, &exec_args);
2567 		if (the_result > 0) {
2568 			if (media_not_present(sdkp, &sshdr))
2569 				return -ENODEV;
2570 
2571 			sense_valid = scsi_sense_valid(&sshdr);
2572 			if (sense_valid &&
2573 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2574 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2575 			    sshdr.ascq == 0x00)
2576 				/* Invalid Command Operation Code or
2577 				 * Invalid Field in CDB, just retry
2578 				 * silently with RC10 */
2579 				return -EINVAL;
2580 			if (sense_valid &&
2581 			    sshdr.sense_key == UNIT_ATTENTION &&
2582 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2583 				/* Device reset might occur several times,
2584 				 * give it one more chance */
2585 				if (--reset_retries > 0)
2586 					continue;
2587 		}
2588 		retries--;
2589 
2590 	} while (the_result && retries);
2591 
2592 	if (the_result) {
2593 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2594 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2595 		return -EINVAL;
2596 	}
2597 
2598 	sector_size = get_unaligned_be32(&buffer[8]);
2599 	lba = get_unaligned_be64(&buffer[0]);
2600 
2601 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2602 		sdkp->capacity = 0;
2603 		return -ENODEV;
2604 	}
2605 
2606 	/* Logical blocks per physical block exponent */
2607 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2608 
2609 	/* RC basis */
2610 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2611 
2612 	/* Lowest aligned logical block */
2613 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2614 	lim->alignment_offset = alignment;
2615 	if (alignment && sdkp->first_scan)
2616 		sd_printk(KERN_NOTICE, sdkp,
2617 			  "physical block alignment offset: %u\n", alignment);
2618 
2619 	if (buffer[14] & 0x80) { /* LBPME */
2620 		sdkp->lbpme = 1;
2621 
2622 		if (buffer[14] & 0x40) /* LBPRZ */
2623 			sdkp->lbprz = 1;
2624 
2625 		sd_config_discard(sdkp, lim, SD_LBP_WS16);
2626 	}
2627 
2628 	sdkp->capacity = lba + 1;
2629 	return sector_size;
2630 }
2631 
2632 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2633 						unsigned char *buffer)
2634 {
2635 	static const u8 cmd[10] = { READ_CAPACITY };
2636 	struct scsi_sense_hdr sshdr;
2637 	struct scsi_failure failure_defs[] = {
2638 		/* Do not retry Medium Not Present */
2639 		{
2640 			.sense = UNIT_ATTENTION,
2641 			.asc = 0x3A,
2642 			.result = SAM_STAT_CHECK_CONDITION,
2643 		},
2644 		{
2645 			.sense = NOT_READY,
2646 			.asc = 0x3A,
2647 			.result = SAM_STAT_CHECK_CONDITION,
2648 		},
2649 		 /* Device reset might occur several times so retry a lot */
2650 		{
2651 			.sense = UNIT_ATTENTION,
2652 			.asc = 0x29,
2653 			.allowed = READ_CAPACITY_RETRIES_ON_RESET,
2654 			.result = SAM_STAT_CHECK_CONDITION,
2655 		},
2656 		/* Any other error not listed above retry 3 times */
2657 		{
2658 			.result = SCMD_FAILURE_RESULT_ANY,
2659 			.allowed = 3,
2660 		},
2661 		{}
2662 	};
2663 	struct scsi_failures failures = {
2664 		.failure_definitions = failure_defs,
2665 	};
2666 	const struct scsi_exec_args exec_args = {
2667 		.sshdr = &sshdr,
2668 		.failures = &failures,
2669 	};
2670 	int sense_valid = 0;
2671 	int the_result;
2672 	sector_t lba;
2673 	unsigned sector_size;
2674 
2675 	memset(buffer, 0, 8);
2676 
2677 	the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2678 				      8, SD_TIMEOUT, sdkp->max_retries,
2679 				      &exec_args);
2680 
2681 	if (the_result > 0) {
2682 		sense_valid = scsi_sense_valid(&sshdr);
2683 
2684 		if (media_not_present(sdkp, &sshdr))
2685 			return -ENODEV;
2686 	}
2687 
2688 	if (the_result) {
2689 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2690 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2691 		return -EINVAL;
2692 	}
2693 
2694 	sector_size = get_unaligned_be32(&buffer[4]);
2695 	lba = get_unaligned_be32(&buffer[0]);
2696 
2697 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2698 		/* Some buggy (usb cardreader) devices return an lba of
2699 		   0xffffffff when the want to report a size of 0 (with
2700 		   which they really mean no media is present) */
2701 		sdkp->capacity = 0;
2702 		sdkp->physical_block_size = sector_size;
2703 		return sector_size;
2704 	}
2705 
2706 	sdkp->capacity = lba + 1;
2707 	sdkp->physical_block_size = sector_size;
2708 	return sector_size;
2709 }
2710 
2711 static int sd_try_rc16_first(struct scsi_device *sdp)
2712 {
2713 	if (sdp->host->max_cmd_len < 16)
2714 		return 0;
2715 	if (sdp->try_rc_10_first)
2716 		return 0;
2717 	if (sdp->scsi_level > SCSI_SPC_2)
2718 		return 1;
2719 	if (scsi_device_protection(sdp))
2720 		return 1;
2721 	return 0;
2722 }
2723 
2724 /*
2725  * read disk capacity
2726  */
2727 static void
2728 sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim,
2729 		unsigned char *buffer)
2730 {
2731 	int sector_size;
2732 	struct scsi_device *sdp = sdkp->device;
2733 
2734 	if (sd_try_rc16_first(sdp)) {
2735 		sector_size = read_capacity_16(sdkp, sdp, lim, buffer);
2736 		if (sector_size == -EOVERFLOW)
2737 			goto got_data;
2738 		if (sector_size == -ENODEV)
2739 			return;
2740 		if (sector_size < 0)
2741 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2742 		if (sector_size < 0)
2743 			return;
2744 	} else {
2745 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2746 		if (sector_size == -EOVERFLOW)
2747 			goto got_data;
2748 		if (sector_size < 0)
2749 			return;
2750 		if ((sizeof(sdkp->capacity) > 4) &&
2751 		    (sdkp->capacity > 0xffffffffULL)) {
2752 			int old_sector_size = sector_size;
2753 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2754 					"Trying to use READ CAPACITY(16).\n");
2755 			sector_size = read_capacity_16(sdkp, sdp, lim, buffer);
2756 			if (sector_size < 0) {
2757 				sd_printk(KERN_NOTICE, sdkp,
2758 					"Using 0xffffffff as device size\n");
2759 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2760 				sector_size = old_sector_size;
2761 				goto got_data;
2762 			}
2763 			/* Remember that READ CAPACITY(16) succeeded */
2764 			sdp->try_rc_10_first = 0;
2765 		}
2766 	}
2767 
2768 	/* Some devices are known to return the total number of blocks,
2769 	 * not the highest block number.  Some devices have versions
2770 	 * which do this and others which do not.  Some devices we might
2771 	 * suspect of doing this but we don't know for certain.
2772 	 *
2773 	 * If we know the reported capacity is wrong, decrement it.  If
2774 	 * we can only guess, then assume the number of blocks is even
2775 	 * (usually true but not always) and err on the side of lowering
2776 	 * the capacity.
2777 	 */
2778 	if (sdp->fix_capacity ||
2779 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2780 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2781 				"from its reported value: %llu\n",
2782 				(unsigned long long) sdkp->capacity);
2783 		--sdkp->capacity;
2784 	}
2785 
2786 got_data:
2787 	if (sector_size == 0) {
2788 		sector_size = 512;
2789 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2790 			  "assuming 512.\n");
2791 	}
2792 
2793 	if (sector_size != 512 &&
2794 	    sector_size != 1024 &&
2795 	    sector_size != 2048 &&
2796 	    sector_size != 4096) {
2797 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2798 			  sector_size);
2799 		/*
2800 		 * The user might want to re-format the drive with
2801 		 * a supported sectorsize.  Once this happens, it
2802 		 * would be relatively trivial to set the thing up.
2803 		 * For this reason, we leave the thing in the table.
2804 		 */
2805 		sdkp->capacity = 0;
2806 		/*
2807 		 * set a bogus sector size so the normal read/write
2808 		 * logic in the block layer will eventually refuse any
2809 		 * request on this device without tripping over power
2810 		 * of two sector size assumptions
2811 		 */
2812 		sector_size = 512;
2813 	}
2814 	lim->logical_block_size = sector_size;
2815 	lim->physical_block_size = sdkp->physical_block_size;
2816 	sdkp->device->sector_size = sector_size;
2817 
2818 	if (sdkp->capacity > 0xffffffff)
2819 		sdp->use_16_for_rw = 1;
2820 
2821 }
2822 
2823 /*
2824  * Print disk capacity
2825  */
2826 static void
2827 sd_print_capacity(struct scsi_disk *sdkp,
2828 		  sector_t old_capacity)
2829 {
2830 	int sector_size = sdkp->device->sector_size;
2831 	char cap_str_2[10], cap_str_10[10];
2832 
2833 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2834 		return;
2835 
2836 	string_get_size(sdkp->capacity, sector_size,
2837 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2838 	string_get_size(sdkp->capacity, sector_size,
2839 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2840 
2841 	sd_printk(KERN_NOTICE, sdkp,
2842 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2843 		  (unsigned long long)sdkp->capacity,
2844 		  sector_size, cap_str_10, cap_str_2);
2845 
2846 	if (sdkp->physical_block_size != sector_size)
2847 		sd_printk(KERN_NOTICE, sdkp,
2848 			  "%u-byte physical blocks\n",
2849 			  sdkp->physical_block_size);
2850 }
2851 
2852 /* called with buffer of length 512 */
2853 static inline int
2854 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2855 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2856 		 struct scsi_sense_hdr *sshdr)
2857 {
2858 	/*
2859 	 * If we must use MODE SENSE(10), make sure that the buffer length
2860 	 * is at least 8 bytes so that the mode sense header fits.
2861 	 */
2862 	if (sdkp->device->use_10_for_ms && len < 8)
2863 		len = 8;
2864 
2865 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2866 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2867 }
2868 
2869 /*
2870  * read write protect setting, if possible - called only in sd_revalidate_disk()
2871  * called with buffer of length SD_BUF_SIZE
2872  */
2873 static void
2874 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2875 {
2876 	int res;
2877 	struct scsi_device *sdp = sdkp->device;
2878 	struct scsi_mode_data data;
2879 	int old_wp = sdkp->write_prot;
2880 
2881 	set_disk_ro(sdkp->disk, 0);
2882 	if (sdp->skip_ms_page_3f) {
2883 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2884 		return;
2885 	}
2886 
2887 	if (sdp->use_192_bytes_for_3f) {
2888 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2889 	} else {
2890 		/*
2891 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2892 		 * We have to start carefully: some devices hang if we ask
2893 		 * for more than is available.
2894 		 */
2895 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2896 
2897 		/*
2898 		 * Second attempt: ask for page 0 When only page 0 is
2899 		 * implemented, a request for page 3F may return Sense Key
2900 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2901 		 * CDB.
2902 		 */
2903 		if (res < 0)
2904 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2905 
2906 		/*
2907 		 * Third attempt: ask 255 bytes, as we did earlier.
2908 		 */
2909 		if (res < 0)
2910 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2911 					       &data, NULL);
2912 	}
2913 
2914 	if (res < 0) {
2915 		sd_first_printk(KERN_WARNING, sdkp,
2916 			  "Test WP failed, assume Write Enabled\n");
2917 	} else {
2918 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2919 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2920 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2921 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2922 				  sdkp->write_prot ? "on" : "off");
2923 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2924 		}
2925 	}
2926 }
2927 
2928 /*
2929  * sd_read_cache_type - called only from sd_revalidate_disk()
2930  * called with buffer of length SD_BUF_SIZE
2931  */
2932 static void
2933 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2934 {
2935 	int len = 0, res;
2936 	struct scsi_device *sdp = sdkp->device;
2937 
2938 	int dbd;
2939 	int modepage;
2940 	int first_len;
2941 	struct scsi_mode_data data;
2942 	struct scsi_sense_hdr sshdr;
2943 	int old_wce = sdkp->WCE;
2944 	int old_rcd = sdkp->RCD;
2945 	int old_dpofua = sdkp->DPOFUA;
2946 
2947 
2948 	if (sdkp->cache_override)
2949 		return;
2950 
2951 	first_len = 4;
2952 	if (sdp->skip_ms_page_8) {
2953 		if (sdp->type == TYPE_RBC)
2954 			goto defaults;
2955 		else {
2956 			if (sdp->skip_ms_page_3f)
2957 				goto defaults;
2958 			modepage = 0x3F;
2959 			if (sdp->use_192_bytes_for_3f)
2960 				first_len = 192;
2961 			dbd = 0;
2962 		}
2963 	} else if (sdp->type == TYPE_RBC) {
2964 		modepage = 6;
2965 		dbd = 8;
2966 	} else {
2967 		modepage = 8;
2968 		dbd = 0;
2969 	}
2970 
2971 	/* cautiously ask */
2972 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2973 			&data, &sshdr);
2974 
2975 	if (res < 0)
2976 		goto bad_sense;
2977 
2978 	if (!data.header_length) {
2979 		modepage = 6;
2980 		first_len = 0;
2981 		sd_first_printk(KERN_ERR, sdkp,
2982 				"Missing header in MODE_SENSE response\n");
2983 	}
2984 
2985 	/* that went OK, now ask for the proper length */
2986 	len = data.length;
2987 
2988 	/*
2989 	 * We're only interested in the first three bytes, actually.
2990 	 * But the data cache page is defined for the first 20.
2991 	 */
2992 	if (len < 3)
2993 		goto bad_sense;
2994 	else if (len > SD_BUF_SIZE) {
2995 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2996 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2997 		len = SD_BUF_SIZE;
2998 	}
2999 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
3000 		len = 192;
3001 
3002 	/* Get the data */
3003 	if (len > first_len)
3004 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
3005 				&data, &sshdr);
3006 
3007 	if (!res) {
3008 		int offset = data.header_length + data.block_descriptor_length;
3009 
3010 		while (offset < len) {
3011 			u8 page_code = buffer[offset] & 0x3F;
3012 			u8 spf       = buffer[offset] & 0x40;
3013 
3014 			if (page_code == 8 || page_code == 6) {
3015 				/* We're interested only in the first 3 bytes.
3016 				 */
3017 				if (len - offset <= 2) {
3018 					sd_first_printk(KERN_ERR, sdkp,
3019 						"Incomplete mode parameter "
3020 							"data\n");
3021 					goto defaults;
3022 				} else {
3023 					modepage = page_code;
3024 					goto Page_found;
3025 				}
3026 			} else {
3027 				/* Go to the next page */
3028 				if (spf && len - offset > 3)
3029 					offset += 4 + (buffer[offset+2] << 8) +
3030 						buffer[offset+3];
3031 				else if (!spf && len - offset > 1)
3032 					offset += 2 + buffer[offset+1];
3033 				else {
3034 					sd_first_printk(KERN_ERR, sdkp,
3035 							"Incomplete mode "
3036 							"parameter data\n");
3037 					goto defaults;
3038 				}
3039 			}
3040 		}
3041 
3042 		sd_first_printk(KERN_WARNING, sdkp,
3043 				"No Caching mode page found\n");
3044 		goto defaults;
3045 
3046 	Page_found:
3047 		if (modepage == 8) {
3048 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
3049 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
3050 		} else {
3051 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
3052 			sdkp->RCD = 0;
3053 		}
3054 
3055 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
3056 		if (sdp->broken_fua) {
3057 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
3058 			sdkp->DPOFUA = 0;
3059 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
3060 			   !sdkp->device->use_16_for_rw) {
3061 			sd_first_printk(KERN_NOTICE, sdkp,
3062 				  "Uses READ/WRITE(6), disabling FUA\n");
3063 			sdkp->DPOFUA = 0;
3064 		}
3065 
3066 		/* No cache flush allowed for write protected devices */
3067 		if (sdkp->WCE && sdkp->write_prot)
3068 			sdkp->WCE = 0;
3069 
3070 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
3071 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
3072 			sd_printk(KERN_NOTICE, sdkp,
3073 				  "Write cache: %s, read cache: %s, %s\n",
3074 				  sdkp->WCE ? "enabled" : "disabled",
3075 				  sdkp->RCD ? "disabled" : "enabled",
3076 				  sdkp->DPOFUA ? "supports DPO and FUA"
3077 				  : "doesn't support DPO or FUA");
3078 
3079 		return;
3080 	}
3081 
3082 bad_sense:
3083 	if (res == -EIO && scsi_sense_valid(&sshdr) &&
3084 	    sshdr.sense_key == ILLEGAL_REQUEST &&
3085 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
3086 		/* Invalid field in CDB */
3087 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
3088 	else
3089 		sd_first_printk(KERN_ERR, sdkp,
3090 				"Asking for cache data failed\n");
3091 
3092 defaults:
3093 	if (sdp->wce_default_on) {
3094 		sd_first_printk(KERN_NOTICE, sdkp,
3095 				"Assuming drive cache: write back\n");
3096 		sdkp->WCE = 1;
3097 	} else {
3098 		sd_first_printk(KERN_WARNING, sdkp,
3099 				"Assuming drive cache: write through\n");
3100 		sdkp->WCE = 0;
3101 	}
3102 	sdkp->RCD = 0;
3103 	sdkp->DPOFUA = 0;
3104 }
3105 
3106 static bool sd_is_perm_stream(struct scsi_disk *sdkp, unsigned int stream_id)
3107 {
3108 	u8 cdb[16] = { SERVICE_ACTION_IN_16, SAI_GET_STREAM_STATUS };
3109 	struct {
3110 		struct scsi_stream_status_header h;
3111 		struct scsi_stream_status s;
3112 	} buf;
3113 	struct scsi_device *sdev = sdkp->device;
3114 	struct scsi_sense_hdr sshdr;
3115 	const struct scsi_exec_args exec_args = {
3116 		.sshdr = &sshdr,
3117 	};
3118 	int res;
3119 
3120 	put_unaligned_be16(stream_id, &cdb[4]);
3121 	put_unaligned_be32(sizeof(buf), &cdb[10]);
3122 
3123 	res = scsi_execute_cmd(sdev, cdb, REQ_OP_DRV_IN, &buf, sizeof(buf),
3124 			       SD_TIMEOUT, sdkp->max_retries, &exec_args);
3125 	if (res < 0)
3126 		return false;
3127 	if (scsi_status_is_check_condition(res) && scsi_sense_valid(&sshdr))
3128 		sd_print_sense_hdr(sdkp, &sshdr);
3129 	if (res)
3130 		return false;
3131 	if (get_unaligned_be32(&buf.h.len) < sizeof(struct scsi_stream_status))
3132 		return false;
3133 	return buf.h.stream_status[0].perm;
3134 }
3135 
3136 static void sd_read_io_hints(struct scsi_disk *sdkp, unsigned char *buffer)
3137 {
3138 	struct scsi_device *sdp = sdkp->device;
3139 	const struct scsi_io_group_descriptor *desc, *start, *end;
3140 	u16 permanent_stream_count_old;
3141 	struct scsi_sense_hdr sshdr;
3142 	struct scsi_mode_data data;
3143 	int res;
3144 
3145 	res = scsi_mode_sense(sdp, /*dbd=*/0x8, /*modepage=*/0x0a,
3146 			      /*subpage=*/0x05, buffer, SD_BUF_SIZE, SD_TIMEOUT,
3147 			      sdkp->max_retries, &data, &sshdr);
3148 	if (res < 0)
3149 		return;
3150 	start = (void *)buffer + data.header_length + 16;
3151 	end = (void *)buffer + ALIGN_DOWN(data.header_length + data.length,
3152 					  sizeof(*end));
3153 	/*
3154 	 * From "SBC-5 Constrained Streams with Data Lifetimes": Device severs
3155 	 * should assign the lowest numbered stream identifiers to permanent
3156 	 * streams.
3157 	 */
3158 	for (desc = start; desc < end; desc++)
3159 		if (!desc->st_enble || !sd_is_perm_stream(sdkp, desc - start))
3160 			break;
3161 	permanent_stream_count_old = sdkp->permanent_stream_count;
3162 	sdkp->permanent_stream_count = desc - start;
3163 	if (sdkp->rscs && sdkp->permanent_stream_count < 2)
3164 		sd_printk(KERN_INFO, sdkp,
3165 			  "Unexpected: RSCS has been set and the permanent stream count is %u\n",
3166 			  sdkp->permanent_stream_count);
3167 	else if (sdkp->permanent_stream_count != permanent_stream_count_old)
3168 		sd_printk(KERN_INFO, sdkp, "permanent stream count = %d\n",
3169 			  sdkp->permanent_stream_count);
3170 }
3171 
3172 /*
3173  * The ATO bit indicates whether the DIF application tag is available
3174  * for use by the operating system.
3175  */
3176 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
3177 {
3178 	int res, offset;
3179 	struct scsi_device *sdp = sdkp->device;
3180 	struct scsi_mode_data data;
3181 	struct scsi_sense_hdr sshdr;
3182 
3183 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
3184 		return;
3185 
3186 	if (sdkp->protection_type == 0)
3187 		return;
3188 
3189 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
3190 			      sdkp->max_retries, &data, &sshdr);
3191 
3192 	if (res < 0 || !data.header_length ||
3193 	    data.length < 6) {
3194 		sd_first_printk(KERN_WARNING, sdkp,
3195 			  "getting Control mode page failed, assume no ATO\n");
3196 
3197 		if (res == -EIO && scsi_sense_valid(&sshdr))
3198 			sd_print_sense_hdr(sdkp, &sshdr);
3199 
3200 		return;
3201 	}
3202 
3203 	offset = data.header_length + data.block_descriptor_length;
3204 
3205 	if ((buffer[offset] & 0x3f) != 0x0a) {
3206 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
3207 		return;
3208 	}
3209 
3210 	if ((buffer[offset + 5] & 0x80) == 0)
3211 		return;
3212 
3213 	sdkp->ATO = 1;
3214 
3215 	return;
3216 }
3217 
3218 static unsigned int sd_discard_mode(struct scsi_disk *sdkp)
3219 {
3220 	if (!sdkp->lbpvpd) {
3221 		/* LBP VPD page not provided */
3222 		if (sdkp->max_unmap_blocks)
3223 			return SD_LBP_UNMAP;
3224 		return SD_LBP_WS16;
3225 	}
3226 
3227 	/* LBP VPD page tells us what to use */
3228 	if (sdkp->lbpu && sdkp->max_unmap_blocks)
3229 		return SD_LBP_UNMAP;
3230 	if (sdkp->lbpws)
3231 		return SD_LBP_WS16;
3232 	if (sdkp->lbpws10)
3233 		return SD_LBP_WS10;
3234 	return SD_LBP_DISABLE;
3235 }
3236 
3237 /*
3238  * Query disk device for preferred I/O sizes.
3239  */
3240 static void sd_read_block_limits(struct scsi_disk *sdkp,
3241 		struct queue_limits *lim)
3242 {
3243 	struct scsi_vpd *vpd;
3244 
3245 	rcu_read_lock();
3246 
3247 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3248 	if (!vpd || vpd->len < 16)
3249 		goto out;
3250 
3251 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3252 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3253 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3254 
3255 	if (vpd->len >= 64) {
3256 		unsigned int lba_count, desc_count;
3257 
3258 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3259 
3260 		if (!sdkp->lbpme)
3261 			goto out;
3262 
3263 		lba_count = get_unaligned_be32(&vpd->data[20]);
3264 		desc_count = get_unaligned_be32(&vpd->data[24]);
3265 
3266 		if (lba_count && desc_count)
3267 			sdkp->max_unmap_blocks = lba_count;
3268 
3269 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3270 
3271 		if (vpd->data[32] & 0x80)
3272 			sdkp->unmap_alignment =
3273 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3274 
3275 		sd_config_discard(sdkp, lim, sd_discard_mode(sdkp));
3276 	}
3277 
3278  out:
3279 	rcu_read_unlock();
3280 }
3281 
3282 /* Parse the Block Limits Extension VPD page (0xb7) */
3283 static void sd_read_block_limits_ext(struct scsi_disk *sdkp)
3284 {
3285 	struct scsi_vpd *vpd;
3286 
3287 	rcu_read_lock();
3288 	vpd = rcu_dereference(sdkp->device->vpd_pgb7);
3289 	if (vpd && vpd->len >= 2)
3290 		sdkp->rscs = vpd->data[5] & 1;
3291 	rcu_read_unlock();
3292 }
3293 
3294 /* Query block device characteristics */
3295 static void sd_read_block_characteristics(struct scsi_disk *sdkp,
3296 		struct queue_limits *lim)
3297 {
3298 	struct request_queue *q = sdkp->disk->queue;
3299 	struct scsi_vpd *vpd;
3300 	u16 rot;
3301 
3302 	rcu_read_lock();
3303 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3304 
3305 	if (!vpd || vpd->len < 8) {
3306 		rcu_read_unlock();
3307 	        return;
3308 	}
3309 
3310 	rot = get_unaligned_be16(&vpd->data[4]);
3311 	sdkp->zoned = (vpd->data[8] >> 4) & 3;
3312 	rcu_read_unlock();
3313 
3314 	if (rot == 1) {
3315 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3316 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3317 	}
3318 
3319 
3320 #ifdef CONFIG_BLK_DEV_ZONED /* sd_probe rejects ZBD devices early otherwise */
3321 	if (sdkp->device->type == TYPE_ZBC) {
3322 		lim->zoned = true;
3323 
3324 		/*
3325 		 * Per ZBC and ZAC specifications, writes in sequential write
3326 		 * required zones of host-managed devices must be aligned to
3327 		 * the device physical block size.
3328 		 */
3329 		lim->zone_write_granularity = sdkp->physical_block_size;
3330 	} else {
3331 		/*
3332 		 * Host-aware devices are treated as conventional.
3333 		 */
3334 		lim->zoned = false;
3335 	}
3336 #endif /* CONFIG_BLK_DEV_ZONED */
3337 
3338 	if (!sdkp->first_scan)
3339 		return;
3340 
3341 	if (lim->zoned)
3342 		sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n");
3343 	else if (sdkp->zoned == 1)
3344 		sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n");
3345 	else if (sdkp->zoned == 2)
3346 		sd_printk(KERN_NOTICE, sdkp, "Drive-managed SMR disk\n");
3347 }
3348 
3349 /**
3350  * sd_read_block_provisioning - Query provisioning VPD page
3351  * @sdkp: disk to query
3352  */
3353 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3354 {
3355 	struct scsi_vpd *vpd;
3356 
3357 	if (sdkp->lbpme == 0)
3358 		return;
3359 
3360 	rcu_read_lock();
3361 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3362 
3363 	if (!vpd || vpd->len < 8) {
3364 		rcu_read_unlock();
3365 		return;
3366 	}
3367 
3368 	sdkp->lbpvpd	= 1;
3369 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3370 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3371 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3372 	rcu_read_unlock();
3373 }
3374 
3375 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3376 {
3377 	struct scsi_device *sdev = sdkp->device;
3378 
3379 	if (sdev->host->no_write_same) {
3380 		sdev->no_write_same = 1;
3381 
3382 		return;
3383 	}
3384 
3385 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3386 		struct scsi_vpd *vpd;
3387 
3388 		sdev->no_report_opcodes = 1;
3389 
3390 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3391 		 * CODES is unsupported and the device has an ATA
3392 		 * Information VPD page (SAT).
3393 		 */
3394 		rcu_read_lock();
3395 		vpd = rcu_dereference(sdev->vpd_pg89);
3396 		if (vpd)
3397 			sdev->no_write_same = 1;
3398 		rcu_read_unlock();
3399 	}
3400 
3401 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3402 		sdkp->ws16 = 1;
3403 
3404 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3405 		sdkp->ws10 = 1;
3406 }
3407 
3408 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3409 {
3410 	struct scsi_device *sdev = sdkp->device;
3411 
3412 	if (!sdev->security_supported)
3413 		return;
3414 
3415 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3416 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3417 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3418 			SECURITY_PROTOCOL_OUT, 0) == 1)
3419 		sdkp->security = 1;
3420 }
3421 
3422 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3423 {
3424 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3425 }
3426 
3427 /**
3428  * sd_read_cpr - Query concurrent positioning ranges
3429  * @sdkp:	disk to query
3430  */
3431 static void sd_read_cpr(struct scsi_disk *sdkp)
3432 {
3433 	struct blk_independent_access_ranges *iars = NULL;
3434 	unsigned char *buffer = NULL;
3435 	unsigned int nr_cpr = 0;
3436 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3437 	u8 *desc;
3438 
3439 	/*
3440 	 * We need to have the capacity set first for the block layer to be
3441 	 * able to check the ranges.
3442 	 */
3443 	if (sdkp->first_scan)
3444 		return;
3445 
3446 	if (!sdkp->capacity)
3447 		goto out;
3448 
3449 	/*
3450 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3451 	 * leading to a maximum page size of 64 + 256*32 bytes.
3452 	 */
3453 	buf_len = 64 + 256*32;
3454 	buffer = kmalloc(buf_len, GFP_KERNEL);
3455 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3456 		goto out;
3457 
3458 	/* We must have at least a 64B header and one 32B range descriptor */
3459 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3460 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3461 		sd_printk(KERN_ERR, sdkp,
3462 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3463 		goto out;
3464 	}
3465 
3466 	nr_cpr = (vpd_len - 64) / 32;
3467 	if (nr_cpr == 1) {
3468 		nr_cpr = 0;
3469 		goto out;
3470 	}
3471 
3472 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3473 	if (!iars) {
3474 		nr_cpr = 0;
3475 		goto out;
3476 	}
3477 
3478 	desc = &buffer[64];
3479 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3480 		if (desc[0] != i) {
3481 			sd_printk(KERN_ERR, sdkp,
3482 				"Invalid Concurrent Positioning Range number\n");
3483 			nr_cpr = 0;
3484 			break;
3485 		}
3486 
3487 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3488 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3489 	}
3490 
3491 out:
3492 	disk_set_independent_access_ranges(sdkp->disk, iars);
3493 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3494 		sd_printk(KERN_NOTICE, sdkp,
3495 			  "%u concurrent positioning ranges\n", nr_cpr);
3496 		sdkp->nr_actuators = nr_cpr;
3497 	}
3498 
3499 	kfree(buffer);
3500 }
3501 
3502 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3503 {
3504 	struct scsi_device *sdp = sdkp->device;
3505 	unsigned int min_xfer_bytes =
3506 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3507 
3508 	if (sdkp->min_xfer_blocks == 0)
3509 		return false;
3510 
3511 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3512 		sd_first_printk(KERN_WARNING, sdkp,
3513 				"Preferred minimum I/O size %u bytes not a " \
3514 				"multiple of physical block size (%u bytes)\n",
3515 				min_xfer_bytes, sdkp->physical_block_size);
3516 		sdkp->min_xfer_blocks = 0;
3517 		return false;
3518 	}
3519 
3520 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3521 			min_xfer_bytes);
3522 	return true;
3523 }
3524 
3525 /*
3526  * Determine the device's preferred I/O size for reads and writes
3527  * unless the reported value is unreasonably small, large, not a
3528  * multiple of the physical block size, or simply garbage.
3529  */
3530 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3531 				      unsigned int dev_max)
3532 {
3533 	struct scsi_device *sdp = sdkp->device;
3534 	unsigned int opt_xfer_bytes =
3535 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3536 	unsigned int min_xfer_bytes =
3537 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3538 
3539 	if (sdkp->opt_xfer_blocks == 0)
3540 		return false;
3541 
3542 	if (sdkp->opt_xfer_blocks > dev_max) {
3543 		sd_first_printk(KERN_WARNING, sdkp,
3544 				"Optimal transfer size %u logical blocks " \
3545 				"> dev_max (%u logical blocks)\n",
3546 				sdkp->opt_xfer_blocks, dev_max);
3547 		return false;
3548 	}
3549 
3550 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3551 		sd_first_printk(KERN_WARNING, sdkp,
3552 				"Optimal transfer size %u logical blocks " \
3553 				"> sd driver limit (%u logical blocks)\n",
3554 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3555 		return false;
3556 	}
3557 
3558 	if (opt_xfer_bytes < PAGE_SIZE) {
3559 		sd_first_printk(KERN_WARNING, sdkp,
3560 				"Optimal transfer size %u bytes < " \
3561 				"PAGE_SIZE (%u bytes)\n",
3562 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3563 		return false;
3564 	}
3565 
3566 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3567 		sd_first_printk(KERN_WARNING, sdkp,
3568 				"Optimal transfer size %u bytes not a " \
3569 				"multiple of preferred minimum block " \
3570 				"size (%u bytes)\n",
3571 				opt_xfer_bytes, min_xfer_bytes);
3572 		return false;
3573 	}
3574 
3575 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3576 		sd_first_printk(KERN_WARNING, sdkp,
3577 				"Optimal transfer size %u bytes not a " \
3578 				"multiple of physical block size (%u bytes)\n",
3579 				opt_xfer_bytes, sdkp->physical_block_size);
3580 		return false;
3581 	}
3582 
3583 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3584 			opt_xfer_bytes);
3585 	return true;
3586 }
3587 
3588 static void sd_read_block_zero(struct scsi_disk *sdkp)
3589 {
3590 	unsigned int buf_len = sdkp->device->sector_size;
3591 	char *buffer, cmd[10] = { };
3592 
3593 	buffer = kmalloc(buf_len, GFP_KERNEL);
3594 	if (!buffer)
3595 		return;
3596 
3597 	cmd[0] = READ_10;
3598 	put_unaligned_be32(0, &cmd[2]); /* Logical block address 0 */
3599 	put_unaligned_be16(1, &cmd[7]);	/* Transfer 1 logical block */
3600 
3601 	scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN, buffer, buf_len,
3602 			 SD_TIMEOUT, sdkp->max_retries, NULL);
3603 	kfree(buffer);
3604 }
3605 
3606 /**
3607  *	sd_revalidate_disk - called the first time a new disk is seen,
3608  *	performs disk spin up, read_capacity, etc.
3609  *	@disk: struct gendisk we care about
3610  **/
3611 static int sd_revalidate_disk(struct gendisk *disk)
3612 {
3613 	struct scsi_disk *sdkp = scsi_disk(disk);
3614 	struct scsi_device *sdp = sdkp->device;
3615 	struct request_queue *q = sdkp->disk->queue;
3616 	sector_t old_capacity = sdkp->capacity;
3617 	struct queue_limits lim;
3618 	unsigned char *buffer;
3619 	unsigned int dev_max;
3620 	int err;
3621 
3622 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3623 				      "sd_revalidate_disk\n"));
3624 
3625 	/*
3626 	 * If the device is offline, don't try and read capacity or any
3627 	 * of the other niceties.
3628 	 */
3629 	if (!scsi_device_online(sdp))
3630 		goto out;
3631 
3632 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3633 	if (!buffer) {
3634 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3635 			  "allocation failure.\n");
3636 		goto out;
3637 	}
3638 
3639 	sd_spinup_disk(sdkp);
3640 
3641 	lim = queue_limits_start_update(sdkp->disk->queue);
3642 
3643 	/*
3644 	 * Without media there is no reason to ask; moreover, some devices
3645 	 * react badly if we do.
3646 	 */
3647 	if (sdkp->media_present) {
3648 		sd_read_capacity(sdkp, &lim, buffer);
3649 		/*
3650 		 * Some USB/UAS devices return generic values for mode pages
3651 		 * until the media has been accessed. Trigger a READ operation
3652 		 * to force the device to populate mode pages.
3653 		 */
3654 		if (sdp->read_before_ms)
3655 			sd_read_block_zero(sdkp);
3656 		/*
3657 		 * set the default to rotational.  All non-rotational devices
3658 		 * support the block characteristics VPD page, which will
3659 		 * cause this to be updated correctly and any device which
3660 		 * doesn't support it should be treated as rotational.
3661 		 */
3662 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3663 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3664 
3665 		if (scsi_device_supports_vpd(sdp)) {
3666 			sd_read_block_provisioning(sdkp);
3667 			sd_read_block_limits(sdkp, &lim);
3668 			sd_read_block_limits_ext(sdkp);
3669 			sd_read_block_characteristics(sdkp, &lim);
3670 			sd_zbc_read_zones(sdkp, &lim, buffer);
3671 			sd_read_cpr(sdkp);
3672 		}
3673 
3674 		sd_print_capacity(sdkp, old_capacity);
3675 
3676 		sd_read_write_protect_flag(sdkp, buffer);
3677 		sd_read_cache_type(sdkp, buffer);
3678 		sd_read_io_hints(sdkp, buffer);
3679 		sd_read_app_tag_own(sdkp, buffer);
3680 		sd_read_write_same(sdkp, buffer);
3681 		sd_read_security(sdkp, buffer);
3682 		sd_config_protection(sdkp, &lim);
3683 	}
3684 
3685 	/*
3686 	 * We now have all cache related info, determine how we deal
3687 	 * with flush requests.
3688 	 */
3689 	sd_set_flush_flag(sdkp);
3690 
3691 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3692 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3693 
3694 	/* Some devices report a maximum block count for READ/WRITE requests. */
3695 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3696 	lim.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3697 
3698 	if (sd_validate_min_xfer_size(sdkp))
3699 		lim.io_min = logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3700 	else
3701 		lim.io_min = 0;
3702 
3703 	/*
3704 	 * Limit default to SCSI host optimal sector limit if set. There may be
3705 	 * an impact on performance for when the size of a request exceeds this
3706 	 * host limit.
3707 	 */
3708 	lim.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT;
3709 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3710 		lim.io_opt = min_not_zero(lim.io_opt,
3711 				logical_to_bytes(sdp, sdkp->opt_xfer_blocks));
3712 	}
3713 
3714 	sdkp->first_scan = 0;
3715 
3716 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3717 	sd_config_write_same(sdkp, &lim);
3718 	kfree(buffer);
3719 
3720 	blk_mq_freeze_queue(sdkp->disk->queue);
3721 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
3722 	blk_mq_unfreeze_queue(sdkp->disk->queue);
3723 	if (err)
3724 		return err;
3725 
3726 	/*
3727 	 * For a zoned drive, revalidating the zones can be done only once
3728 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3729 	 * capacity to 0.
3730 	 */
3731 	if (sd_zbc_revalidate_zones(sdkp))
3732 		set_capacity_and_notify(disk, 0);
3733 
3734  out:
3735 	return 0;
3736 }
3737 
3738 /**
3739  *	sd_unlock_native_capacity - unlock native capacity
3740  *	@disk: struct gendisk to set capacity for
3741  *
3742  *	Block layer calls this function if it detects that partitions
3743  *	on @disk reach beyond the end of the device.  If the SCSI host
3744  *	implements ->unlock_native_capacity() method, it's invoked to
3745  *	give it a chance to adjust the device capacity.
3746  *
3747  *	CONTEXT:
3748  *	Defined by block layer.  Might sleep.
3749  */
3750 static void sd_unlock_native_capacity(struct gendisk *disk)
3751 {
3752 	struct scsi_device *sdev = scsi_disk(disk)->device;
3753 
3754 	if (sdev->host->hostt->unlock_native_capacity)
3755 		sdev->host->hostt->unlock_native_capacity(sdev);
3756 }
3757 
3758 /**
3759  *	sd_format_disk_name - format disk name
3760  *	@prefix: name prefix - ie. "sd" for SCSI disks
3761  *	@index: index of the disk to format name for
3762  *	@buf: output buffer
3763  *	@buflen: length of the output buffer
3764  *
3765  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3766  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3767  *	which is followed by sdaaa.
3768  *
3769  *	This is basically 26 base counting with one extra 'nil' entry
3770  *	at the beginning from the second digit on and can be
3771  *	determined using similar method as 26 base conversion with the
3772  *	index shifted -1 after each digit is computed.
3773  *
3774  *	CONTEXT:
3775  *	Don't care.
3776  *
3777  *	RETURNS:
3778  *	0 on success, -errno on failure.
3779  */
3780 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3781 {
3782 	const int base = 'z' - 'a' + 1;
3783 	char *begin = buf + strlen(prefix);
3784 	char *end = buf + buflen;
3785 	char *p;
3786 	int unit;
3787 
3788 	p = end - 1;
3789 	*p = '\0';
3790 	unit = base;
3791 	do {
3792 		if (p == begin)
3793 			return -EINVAL;
3794 		*--p = 'a' + (index % unit);
3795 		index = (index / unit) - 1;
3796 	} while (index >= 0);
3797 
3798 	memmove(begin, p, end - p);
3799 	memcpy(buf, prefix, strlen(prefix));
3800 
3801 	return 0;
3802 }
3803 
3804 /**
3805  *	sd_probe - called during driver initialization and whenever a
3806  *	new scsi device is attached to the system. It is called once
3807  *	for each scsi device (not just disks) present.
3808  *	@dev: pointer to device object
3809  *
3810  *	Returns 0 if successful (or not interested in this scsi device
3811  *	(e.g. scanner)); 1 when there is an error.
3812  *
3813  *	Note: this function is invoked from the scsi mid-level.
3814  *	This function sets up the mapping between a given
3815  *	<host,channel,id,lun> (found in sdp) and new device name
3816  *	(e.g. /dev/sda). More precisely it is the block device major
3817  *	and minor number that is chosen here.
3818  *
3819  *	Assume sd_probe is not re-entrant (for time being)
3820  *	Also think about sd_probe() and sd_remove() running coincidentally.
3821  **/
3822 static int sd_probe(struct device *dev)
3823 {
3824 	struct scsi_device *sdp = to_scsi_device(dev);
3825 	struct scsi_disk *sdkp;
3826 	struct gendisk *gd;
3827 	int index;
3828 	int error;
3829 
3830 	scsi_autopm_get_device(sdp);
3831 	error = -ENODEV;
3832 	if (sdp->type != TYPE_DISK &&
3833 	    sdp->type != TYPE_ZBC &&
3834 	    sdp->type != TYPE_MOD &&
3835 	    sdp->type != TYPE_RBC)
3836 		goto out;
3837 
3838 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3839 		sdev_printk(KERN_WARNING, sdp,
3840 			    "Unsupported ZBC host-managed device.\n");
3841 		goto out;
3842 	}
3843 
3844 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3845 					"sd_probe\n"));
3846 
3847 	error = -ENOMEM;
3848 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3849 	if (!sdkp)
3850 		goto out;
3851 
3852 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3853 					 &sd_bio_compl_lkclass);
3854 	if (!gd)
3855 		goto out_free;
3856 
3857 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3858 	if (index < 0) {
3859 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3860 		goto out_put;
3861 	}
3862 
3863 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3864 	if (error) {
3865 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3866 		goto out_free_index;
3867 	}
3868 
3869 	sdkp->device = sdp;
3870 	sdkp->disk = gd;
3871 	sdkp->index = index;
3872 	sdkp->max_retries = SD_MAX_RETRIES;
3873 	atomic_set(&sdkp->openers, 0);
3874 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3875 
3876 	if (!sdp->request_queue->rq_timeout) {
3877 		if (sdp->type != TYPE_MOD)
3878 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3879 		else
3880 			blk_queue_rq_timeout(sdp->request_queue,
3881 					     SD_MOD_TIMEOUT);
3882 	}
3883 
3884 	device_initialize(&sdkp->disk_dev);
3885 	sdkp->disk_dev.parent = get_device(dev);
3886 	sdkp->disk_dev.class = &sd_disk_class;
3887 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3888 
3889 	error = device_add(&sdkp->disk_dev);
3890 	if (error) {
3891 		put_device(&sdkp->disk_dev);
3892 		goto out;
3893 	}
3894 
3895 	dev_set_drvdata(dev, sdkp);
3896 
3897 	gd->major = sd_major((index & 0xf0) >> 4);
3898 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3899 	gd->minors = SD_MINORS;
3900 
3901 	gd->fops = &sd_fops;
3902 	gd->private_data = sdkp;
3903 
3904 	/* defaults, until the device tells us otherwise */
3905 	sdp->sector_size = 512;
3906 	sdkp->capacity = 0;
3907 	sdkp->media_present = 1;
3908 	sdkp->write_prot = 0;
3909 	sdkp->cache_override = 0;
3910 	sdkp->WCE = 0;
3911 	sdkp->RCD = 0;
3912 	sdkp->ATO = 0;
3913 	sdkp->first_scan = 1;
3914 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3915 
3916 	sd_revalidate_disk(gd);
3917 
3918 	if (sdp->removable) {
3919 		gd->flags |= GENHD_FL_REMOVABLE;
3920 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3921 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3922 	}
3923 
3924 	blk_pm_runtime_init(sdp->request_queue, dev);
3925 	if (sdp->rpm_autosuspend) {
3926 		pm_runtime_set_autosuspend_delay(dev,
3927 			sdp->host->rpm_autosuspend_delay);
3928 	}
3929 
3930 	error = device_add_disk(dev, gd, NULL);
3931 	if (error) {
3932 		device_unregister(&sdkp->disk_dev);
3933 		put_disk(gd);
3934 		goto out;
3935 	}
3936 
3937 	if (sdkp->security) {
3938 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3939 		if (sdkp->opal_dev)
3940 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3941 	}
3942 
3943 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3944 		  sdp->removable ? "removable " : "");
3945 	scsi_autopm_put_device(sdp);
3946 
3947 	return 0;
3948 
3949  out_free_index:
3950 	ida_free(&sd_index_ida, index);
3951  out_put:
3952 	put_disk(gd);
3953  out_free:
3954 	kfree(sdkp);
3955  out:
3956 	scsi_autopm_put_device(sdp);
3957 	return error;
3958 }
3959 
3960 /**
3961  *	sd_remove - called whenever a scsi disk (previously recognized by
3962  *	sd_probe) is detached from the system. It is called (potentially
3963  *	multiple times) during sd module unload.
3964  *	@dev: pointer to device object
3965  *
3966  *	Note: this function is invoked from the scsi mid-level.
3967  *	This function potentially frees up a device name (e.g. /dev/sdc)
3968  *	that could be re-used by a subsequent sd_probe().
3969  *	This function is not called when the built-in sd driver is "exit-ed".
3970  **/
3971 static int sd_remove(struct device *dev)
3972 {
3973 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3974 
3975 	scsi_autopm_get_device(sdkp->device);
3976 
3977 	device_del(&sdkp->disk_dev);
3978 	del_gendisk(sdkp->disk);
3979 	if (!sdkp->suspended)
3980 		sd_shutdown(dev);
3981 
3982 	put_disk(sdkp->disk);
3983 	return 0;
3984 }
3985 
3986 static void scsi_disk_release(struct device *dev)
3987 {
3988 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3989 
3990 	ida_free(&sd_index_ida, sdkp->index);
3991 	put_device(&sdkp->device->sdev_gendev);
3992 	free_opal_dev(sdkp->opal_dev);
3993 
3994 	kfree(sdkp);
3995 }
3996 
3997 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3998 {
3999 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
4000 	struct scsi_sense_hdr sshdr;
4001 	const struct scsi_exec_args exec_args = {
4002 		.sshdr = &sshdr,
4003 		.req_flags = BLK_MQ_REQ_PM,
4004 	};
4005 	struct scsi_device *sdp = sdkp->device;
4006 	int res;
4007 
4008 	if (start)
4009 		cmd[4] |= 1;	/* START */
4010 
4011 	if (sdp->start_stop_pwr_cond)
4012 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
4013 
4014 	if (!scsi_device_online(sdp))
4015 		return -ENODEV;
4016 
4017 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
4018 			       sdkp->max_retries, &exec_args);
4019 	if (res) {
4020 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
4021 		if (res > 0 && scsi_sense_valid(&sshdr)) {
4022 			sd_print_sense_hdr(sdkp, &sshdr);
4023 			/* 0x3a is medium not present */
4024 			if (sshdr.asc == 0x3a)
4025 				res = 0;
4026 		}
4027 	}
4028 
4029 	/* SCSI error codes must not go to the generic layer */
4030 	if (res)
4031 		return -EIO;
4032 
4033 	return 0;
4034 }
4035 
4036 /*
4037  * Send a SYNCHRONIZE CACHE instruction down to the device through
4038  * the normal SCSI command structure.  Wait for the command to
4039  * complete.
4040  */
4041 static void sd_shutdown(struct device *dev)
4042 {
4043 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4044 
4045 	if (!sdkp)
4046 		return;         /* this can happen */
4047 
4048 	if (pm_runtime_suspended(dev))
4049 		return;
4050 
4051 	if (sdkp->WCE && sdkp->media_present) {
4052 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4053 		sd_sync_cache(sdkp);
4054 	}
4055 
4056 	if ((system_state != SYSTEM_RESTART &&
4057 	     sdkp->device->manage_system_start_stop) ||
4058 	    (system_state == SYSTEM_POWER_OFF &&
4059 	     sdkp->device->manage_shutdown)) {
4060 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4061 		sd_start_stop_device(sdkp, 0);
4062 	}
4063 }
4064 
4065 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
4066 {
4067 	return (sdev->manage_system_start_stop && !runtime) ||
4068 		(sdev->manage_runtime_start_stop && runtime);
4069 }
4070 
4071 static int sd_suspend_common(struct device *dev, bool runtime)
4072 {
4073 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4074 	int ret = 0;
4075 
4076 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
4077 		return 0;
4078 
4079 	if (sdkp->WCE && sdkp->media_present) {
4080 		if (!sdkp->device->silence_suspend)
4081 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4082 		ret = sd_sync_cache(sdkp);
4083 		/* ignore OFFLINE device */
4084 		if (ret == -ENODEV)
4085 			return 0;
4086 
4087 		if (ret)
4088 			return ret;
4089 	}
4090 
4091 	if (sd_do_start_stop(sdkp->device, runtime)) {
4092 		if (!sdkp->device->silence_suspend)
4093 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4094 		/* an error is not worth aborting a system sleep */
4095 		ret = sd_start_stop_device(sdkp, 0);
4096 		if (!runtime)
4097 			ret = 0;
4098 	}
4099 
4100 	if (!ret)
4101 		sdkp->suspended = true;
4102 
4103 	return ret;
4104 }
4105 
4106 static int sd_suspend_system(struct device *dev)
4107 {
4108 	if (pm_runtime_suspended(dev))
4109 		return 0;
4110 
4111 	return sd_suspend_common(dev, false);
4112 }
4113 
4114 static int sd_suspend_runtime(struct device *dev)
4115 {
4116 	return sd_suspend_common(dev, true);
4117 }
4118 
4119 static int sd_resume(struct device *dev)
4120 {
4121 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4122 
4123 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4124 
4125 	if (opal_unlock_from_suspend(sdkp->opal_dev)) {
4126 		sd_printk(KERN_NOTICE, sdkp, "OPAL unlock failed\n");
4127 		return -EIO;
4128 	}
4129 
4130 	return 0;
4131 }
4132 
4133 static int sd_resume_common(struct device *dev, bool runtime)
4134 {
4135 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4136 	int ret;
4137 
4138 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4139 		return 0;
4140 
4141 	if (!sd_do_start_stop(sdkp->device, runtime)) {
4142 		sdkp->suspended = false;
4143 		return 0;
4144 	}
4145 
4146 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4147 	ret = sd_start_stop_device(sdkp, 1);
4148 	if (!ret) {
4149 		sd_resume(dev);
4150 		sdkp->suspended = false;
4151 	}
4152 
4153 	return ret;
4154 }
4155 
4156 static int sd_resume_system(struct device *dev)
4157 {
4158 	if (pm_runtime_suspended(dev)) {
4159 		struct scsi_disk *sdkp = dev_get_drvdata(dev);
4160 		struct scsi_device *sdp = sdkp ? sdkp->device : NULL;
4161 
4162 		if (sdp && sdp->force_runtime_start_on_system_start)
4163 			pm_request_resume(dev);
4164 
4165 		return 0;
4166 	}
4167 
4168 	return sd_resume_common(dev, false);
4169 }
4170 
4171 static int sd_resume_runtime(struct device *dev)
4172 {
4173 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4174 	struct scsi_device *sdp;
4175 
4176 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4177 		return 0;
4178 
4179 	sdp = sdkp->device;
4180 
4181 	if (sdp->ignore_media_change) {
4182 		/* clear the device's sense data */
4183 		static const u8 cmd[10] = { REQUEST_SENSE };
4184 		const struct scsi_exec_args exec_args = {
4185 			.req_flags = BLK_MQ_REQ_PM,
4186 		};
4187 
4188 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
4189 				     sdp->request_queue->rq_timeout, 1,
4190 				     &exec_args))
4191 			sd_printk(KERN_NOTICE, sdkp,
4192 				  "Failed to clear sense data\n");
4193 	}
4194 
4195 	return sd_resume_common(dev, true);
4196 }
4197 
4198 static const struct dev_pm_ops sd_pm_ops = {
4199 	.suspend		= sd_suspend_system,
4200 	.resume			= sd_resume_system,
4201 	.poweroff		= sd_suspend_system,
4202 	.restore		= sd_resume_system,
4203 	.runtime_suspend	= sd_suspend_runtime,
4204 	.runtime_resume		= sd_resume_runtime,
4205 };
4206 
4207 static struct scsi_driver sd_template = {
4208 	.gendrv = {
4209 		.name		= "sd",
4210 		.probe		= sd_probe,
4211 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
4212 		.remove		= sd_remove,
4213 		.shutdown	= sd_shutdown,
4214 		.pm		= &sd_pm_ops,
4215 	},
4216 	.rescan			= sd_rescan,
4217 	.resume			= sd_resume,
4218 	.init_command		= sd_init_command,
4219 	.uninit_command		= sd_uninit_command,
4220 	.done			= sd_done,
4221 	.eh_action		= sd_eh_action,
4222 	.eh_reset		= sd_eh_reset,
4223 };
4224 
4225 /**
4226  *	init_sd - entry point for this driver (both when built in or when
4227  *	a module).
4228  *
4229  *	Note: this function registers this driver with the scsi mid-level.
4230  **/
4231 static int __init init_sd(void)
4232 {
4233 	int majors = 0, i, err;
4234 
4235 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
4236 
4237 	for (i = 0; i < SD_MAJORS; i++) {
4238 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
4239 			continue;
4240 		majors++;
4241 	}
4242 
4243 	if (!majors)
4244 		return -ENODEV;
4245 
4246 	err = class_register(&sd_disk_class);
4247 	if (err)
4248 		goto err_out;
4249 
4250 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4251 	if (!sd_page_pool) {
4252 		printk(KERN_ERR "sd: can't init discard page pool\n");
4253 		err = -ENOMEM;
4254 		goto err_out_class;
4255 	}
4256 
4257 	err = scsi_register_driver(&sd_template.gendrv);
4258 	if (err)
4259 		goto err_out_driver;
4260 
4261 	return 0;
4262 
4263 err_out_driver:
4264 	mempool_destroy(sd_page_pool);
4265 err_out_class:
4266 	class_unregister(&sd_disk_class);
4267 err_out:
4268 	for (i = 0; i < SD_MAJORS; i++)
4269 		unregister_blkdev(sd_major(i), "sd");
4270 	return err;
4271 }
4272 
4273 /**
4274  *	exit_sd - exit point for this driver (when it is a module).
4275  *
4276  *	Note: this function unregisters this driver from the scsi mid-level.
4277  **/
4278 static void __exit exit_sd(void)
4279 {
4280 	int i;
4281 
4282 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4283 
4284 	scsi_unregister_driver(&sd_template.gendrv);
4285 	mempool_destroy(sd_page_pool);
4286 
4287 	class_unregister(&sd_disk_class);
4288 
4289 	for (i = 0; i < SD_MAJORS; i++)
4290 		unregister_blkdev(sd_major(i), "sd");
4291 }
4292 
4293 module_init(init_sd);
4294 module_exit(exit_sd);
4295 
4296 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4297 {
4298 	scsi_print_sense_hdr(sdkp->device,
4299 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4300 }
4301 
4302 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4303 {
4304 	const char *hb_string = scsi_hostbyte_string(result);
4305 
4306 	if (hb_string)
4307 		sd_printk(KERN_INFO, sdkp,
4308 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4309 			  hb_string ? hb_string : "invalid",
4310 			  "DRIVER_OK");
4311 	else
4312 		sd_printk(KERN_INFO, sdkp,
4313 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4314 			  msg, host_byte(result), "DRIVER_OK");
4315 }
4316