xref: /linux/drivers/scsi/sd.c (revision 8e1bb4a41aa78d6105e59186af3dcd545fc66e70)
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/bio-integrity.h>
37 #include <linux/module.h>
38 #include <linux/fs.h>
39 #include <linux/kernel.h>
40 #include <linux/mm.h>
41 #include <linux/bio-integrity.h>
42 #include <linux/hdreg.h>
43 #include <linux/errno.h>
44 #include <linux/idr.h>
45 #include <linux/interrupt.h>
46 #include <linux/init.h>
47 #include <linux/blkdev.h>
48 #include <linux/blkpg.h>
49 #include <linux/blk-pm.h>
50 #include <linux/delay.h>
51 #include <linux/rw_hint.h>
52 #include <linux/major.h>
53 #include <linux/mutex.h>
54 #include <linux/string_helpers.h>
55 #include <linux/slab.h>
56 #include <linux/sed-opal.h>
57 #include <linux/pm_runtime.h>
58 #include <linux/pr.h>
59 #include <linux/t10-pi.h>
60 #include <linux/uaccess.h>
61 #include <asm/unaligned.h>
62 
63 #include <scsi/scsi.h>
64 #include <scsi/scsi_cmnd.h>
65 #include <scsi/scsi_dbg.h>
66 #include <scsi/scsi_device.h>
67 #include <scsi/scsi_devinfo.h>
68 #include <scsi/scsi_driver.h>
69 #include <scsi/scsi_eh.h>
70 #include <scsi/scsi_host.h>
71 #include <scsi/scsi_ioctl.h>
72 #include <scsi/scsicam.h>
73 #include <scsi/scsi_common.h>
74 
75 #include "sd.h"
76 #include "scsi_priv.h"
77 #include "scsi_logging.h"
78 
79 MODULE_AUTHOR("Eric Youngdale");
80 MODULE_DESCRIPTION("SCSI disk (sd) driver");
81 MODULE_LICENSE("GPL");
82 
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
96 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
97 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
98 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
100 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
101 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
102 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
103 
104 #define SD_MINORS	16
105 
106 static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
107 		unsigned int mode);
108 static void sd_config_write_same(struct scsi_disk *sdkp,
109 		struct queue_limits *lim);
110 static int  sd_revalidate_disk(struct gendisk *);
111 static void sd_unlock_native_capacity(struct gendisk *disk);
112 static void sd_shutdown(struct device *);
113 static void scsi_disk_release(struct device *cdev);
114 
115 static DEFINE_IDA(sd_index_ida);
116 
117 static mempool_t *sd_page_pool;
118 static struct lock_class_key sd_bio_compl_lkclass;
119 
120 static const char *sd_cache_types[] = {
121 	"write through", "none", "write back",
122 	"write back, no read (daft)"
123 };
124 
125 static void sd_set_flush_flag(struct scsi_disk *sdkp,
126 		struct queue_limits *lim)
127 {
128 	if (sdkp->WCE) {
129 		lim->features |= BLK_FEAT_WRITE_CACHE;
130 		if (sdkp->DPOFUA)
131 			lim->features |= BLK_FEAT_FUA;
132 		else
133 			lim->features &= ~BLK_FEAT_FUA;
134 	} else {
135 		lim->features &= ~(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA);
136 	}
137 }
138 
139 static ssize_t
140 cache_type_store(struct device *dev, struct device_attribute *attr,
141 		 const char *buf, size_t count)
142 {
143 	int ct, rcd, wce, sp;
144 	struct scsi_disk *sdkp = to_scsi_disk(dev);
145 	struct scsi_device *sdp = sdkp->device;
146 	char buffer[64];
147 	char *buffer_data;
148 	struct scsi_mode_data data;
149 	struct scsi_sense_hdr sshdr;
150 	static const char temp[] = "temporary ";
151 	int len, ret;
152 
153 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
154 		/* no cache control on RBC devices; theoretically they
155 		 * can do it, but there's probably so many exceptions
156 		 * it's not worth the risk */
157 		return -EINVAL;
158 
159 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
160 		buf += sizeof(temp) - 1;
161 		sdkp->cache_override = 1;
162 	} else {
163 		sdkp->cache_override = 0;
164 	}
165 
166 	ct = sysfs_match_string(sd_cache_types, buf);
167 	if (ct < 0)
168 		return -EINVAL;
169 
170 	rcd = ct & 0x01 ? 1 : 0;
171 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
172 
173 	if (sdkp->cache_override) {
174 		struct queue_limits lim;
175 
176 		sdkp->WCE = wce;
177 		sdkp->RCD = rcd;
178 
179 		lim = queue_limits_start_update(sdkp->disk->queue);
180 		sd_set_flush_flag(sdkp, &lim);
181 		blk_mq_freeze_queue(sdkp->disk->queue);
182 		ret = queue_limits_commit_update(sdkp->disk->queue, &lim);
183 		blk_mq_unfreeze_queue(sdkp->disk->queue);
184 		if (ret)
185 			return ret;
186 		return count;
187 	}
188 
189 	if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
190 			    sdkp->max_retries, &data, NULL))
191 		return -EINVAL;
192 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
193 		  data.block_descriptor_length);
194 	buffer_data = buffer + data.header_length +
195 		data.block_descriptor_length;
196 	buffer_data[2] &= ~0x05;
197 	buffer_data[2] |= wce << 2 | rcd;
198 	sp = buffer_data[0] & 0x80 ? 1 : 0;
199 	buffer_data[0] &= ~0x80;
200 
201 	/*
202 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
203 	 * received mode parameter buffer before doing MODE SELECT.
204 	 */
205 	data.device_specific = 0;
206 
207 	ret = scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
208 			       sdkp->max_retries, &data, &sshdr);
209 	if (ret) {
210 		if (ret > 0 && scsi_sense_valid(&sshdr))
211 			sd_print_sense_hdr(sdkp, &sshdr);
212 		return -EINVAL;
213 	}
214 	sd_revalidate_disk(sdkp->disk);
215 	return count;
216 }
217 
218 static ssize_t
219 manage_start_stop_show(struct device *dev,
220 		       struct device_attribute *attr, char *buf)
221 {
222 	struct scsi_disk *sdkp = to_scsi_disk(dev);
223 	struct scsi_device *sdp = sdkp->device;
224 
225 	return sysfs_emit(buf, "%u\n",
226 			  sdp->manage_system_start_stop &&
227 			  sdp->manage_runtime_start_stop &&
228 			  sdp->manage_shutdown);
229 }
230 static DEVICE_ATTR_RO(manage_start_stop);
231 
232 static ssize_t
233 manage_system_start_stop_show(struct device *dev,
234 			      struct device_attribute *attr, char *buf)
235 {
236 	struct scsi_disk *sdkp = to_scsi_disk(dev);
237 	struct scsi_device *sdp = sdkp->device;
238 
239 	return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
240 }
241 
242 static ssize_t
243 manage_system_start_stop_store(struct device *dev,
244 			       struct device_attribute *attr,
245 			       const char *buf, size_t count)
246 {
247 	struct scsi_disk *sdkp = to_scsi_disk(dev);
248 	struct scsi_device *sdp = sdkp->device;
249 	bool v;
250 
251 	if (!capable(CAP_SYS_ADMIN))
252 		return -EACCES;
253 
254 	if (kstrtobool(buf, &v))
255 		return -EINVAL;
256 
257 	sdp->manage_system_start_stop = v;
258 
259 	return count;
260 }
261 static DEVICE_ATTR_RW(manage_system_start_stop);
262 
263 static ssize_t
264 manage_runtime_start_stop_show(struct device *dev,
265 			       struct device_attribute *attr, char *buf)
266 {
267 	struct scsi_disk *sdkp = to_scsi_disk(dev);
268 	struct scsi_device *sdp = sdkp->device;
269 
270 	return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
271 }
272 
273 static ssize_t
274 manage_runtime_start_stop_store(struct device *dev,
275 				struct device_attribute *attr,
276 				const char *buf, size_t count)
277 {
278 	struct scsi_disk *sdkp = to_scsi_disk(dev);
279 	struct scsi_device *sdp = sdkp->device;
280 	bool v;
281 
282 	if (!capable(CAP_SYS_ADMIN))
283 		return -EACCES;
284 
285 	if (kstrtobool(buf, &v))
286 		return -EINVAL;
287 
288 	sdp->manage_runtime_start_stop = v;
289 
290 	return count;
291 }
292 static DEVICE_ATTR_RW(manage_runtime_start_stop);
293 
294 static ssize_t manage_shutdown_show(struct device *dev,
295 				    struct device_attribute *attr, char *buf)
296 {
297 	struct scsi_disk *sdkp = to_scsi_disk(dev);
298 	struct scsi_device *sdp = sdkp->device;
299 
300 	return sysfs_emit(buf, "%u\n", sdp->manage_shutdown);
301 }
302 
303 static ssize_t manage_shutdown_store(struct device *dev,
304 				     struct device_attribute *attr,
305 				     const char *buf, size_t count)
306 {
307 	struct scsi_disk *sdkp = to_scsi_disk(dev);
308 	struct scsi_device *sdp = sdkp->device;
309 	bool v;
310 
311 	if (!capable(CAP_SYS_ADMIN))
312 		return -EACCES;
313 
314 	if (kstrtobool(buf, &v))
315 		return -EINVAL;
316 
317 	sdp->manage_shutdown = v;
318 
319 	return count;
320 }
321 static DEVICE_ATTR_RW(manage_shutdown);
322 
323 static ssize_t
324 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
325 {
326 	struct scsi_disk *sdkp = to_scsi_disk(dev);
327 
328 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
329 }
330 
331 static ssize_t
332 allow_restart_store(struct device *dev, struct device_attribute *attr,
333 		    const char *buf, size_t count)
334 {
335 	bool v;
336 	struct scsi_disk *sdkp = to_scsi_disk(dev);
337 	struct scsi_device *sdp = sdkp->device;
338 
339 	if (!capable(CAP_SYS_ADMIN))
340 		return -EACCES;
341 
342 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
343 		return -EINVAL;
344 
345 	if (kstrtobool(buf, &v))
346 		return -EINVAL;
347 
348 	sdp->allow_restart = v;
349 
350 	return count;
351 }
352 static DEVICE_ATTR_RW(allow_restart);
353 
354 static ssize_t
355 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
356 {
357 	struct scsi_disk *sdkp = to_scsi_disk(dev);
358 	int ct = sdkp->RCD + 2*sdkp->WCE;
359 
360 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
361 }
362 static DEVICE_ATTR_RW(cache_type);
363 
364 static ssize_t
365 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
366 {
367 	struct scsi_disk *sdkp = to_scsi_disk(dev);
368 
369 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
370 }
371 static DEVICE_ATTR_RO(FUA);
372 
373 static ssize_t
374 protection_type_show(struct device *dev, struct device_attribute *attr,
375 		     char *buf)
376 {
377 	struct scsi_disk *sdkp = to_scsi_disk(dev);
378 
379 	return sprintf(buf, "%u\n", sdkp->protection_type);
380 }
381 
382 static ssize_t
383 protection_type_store(struct device *dev, struct device_attribute *attr,
384 		      const char *buf, size_t count)
385 {
386 	struct scsi_disk *sdkp = to_scsi_disk(dev);
387 	unsigned int val;
388 	int err;
389 
390 	if (!capable(CAP_SYS_ADMIN))
391 		return -EACCES;
392 
393 	err = kstrtouint(buf, 10, &val);
394 
395 	if (err)
396 		return err;
397 
398 	if (val <= T10_PI_TYPE3_PROTECTION)
399 		sdkp->protection_type = val;
400 
401 	return count;
402 }
403 static DEVICE_ATTR_RW(protection_type);
404 
405 static ssize_t
406 protection_mode_show(struct device *dev, struct device_attribute *attr,
407 		     char *buf)
408 {
409 	struct scsi_disk *sdkp = to_scsi_disk(dev);
410 	struct scsi_device *sdp = sdkp->device;
411 	unsigned int dif, dix;
412 
413 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
414 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
415 
416 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
417 		dif = 0;
418 		dix = 1;
419 	}
420 
421 	if (!dif && !dix)
422 		return sprintf(buf, "none\n");
423 
424 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
425 }
426 static DEVICE_ATTR_RO(protection_mode);
427 
428 static ssize_t
429 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
430 {
431 	struct scsi_disk *sdkp = to_scsi_disk(dev);
432 
433 	return sprintf(buf, "%u\n", sdkp->ATO);
434 }
435 static DEVICE_ATTR_RO(app_tag_own);
436 
437 static ssize_t
438 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
439 		       char *buf)
440 {
441 	struct scsi_disk *sdkp = to_scsi_disk(dev);
442 
443 	return sprintf(buf, "%u\n", sdkp->lbpme);
444 }
445 static DEVICE_ATTR_RO(thin_provisioning);
446 
447 /* sysfs_match_string() requires dense arrays */
448 static const char *lbp_mode[] = {
449 	[SD_LBP_FULL]		= "full",
450 	[SD_LBP_UNMAP]		= "unmap",
451 	[SD_LBP_WS16]		= "writesame_16",
452 	[SD_LBP_WS10]		= "writesame_10",
453 	[SD_LBP_ZERO]		= "writesame_zero",
454 	[SD_LBP_DISABLE]	= "disabled",
455 };
456 
457 static ssize_t
458 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
459 		       char *buf)
460 {
461 	struct scsi_disk *sdkp = to_scsi_disk(dev);
462 
463 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
464 }
465 
466 static ssize_t
467 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
468 			const char *buf, size_t count)
469 {
470 	struct scsi_disk *sdkp = to_scsi_disk(dev);
471 	struct scsi_device *sdp = sdkp->device;
472 	struct queue_limits lim;
473 	int mode, err;
474 
475 	if (!capable(CAP_SYS_ADMIN))
476 		return -EACCES;
477 
478 	if (sdp->type != TYPE_DISK)
479 		return -EINVAL;
480 
481 	mode = sysfs_match_string(lbp_mode, buf);
482 	if (mode < 0)
483 		return -EINVAL;
484 
485 	lim = queue_limits_start_update(sdkp->disk->queue);
486 	sd_config_discard(sdkp, &lim, mode);
487 	blk_mq_freeze_queue(sdkp->disk->queue);
488 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
489 	blk_mq_unfreeze_queue(sdkp->disk->queue);
490 	if (err)
491 		return err;
492 	return count;
493 }
494 static DEVICE_ATTR_RW(provisioning_mode);
495 
496 /* sysfs_match_string() requires dense arrays */
497 static const char *zeroing_mode[] = {
498 	[SD_ZERO_WRITE]		= "write",
499 	[SD_ZERO_WS]		= "writesame",
500 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
501 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
502 };
503 
504 static ssize_t
505 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
506 		  char *buf)
507 {
508 	struct scsi_disk *sdkp = to_scsi_disk(dev);
509 
510 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
511 }
512 
513 static ssize_t
514 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
515 		   const char *buf, size_t count)
516 {
517 	struct scsi_disk *sdkp = to_scsi_disk(dev);
518 	int mode;
519 
520 	if (!capable(CAP_SYS_ADMIN))
521 		return -EACCES;
522 
523 	mode = sysfs_match_string(zeroing_mode, buf);
524 	if (mode < 0)
525 		return -EINVAL;
526 
527 	sdkp->zeroing_mode = mode;
528 
529 	return count;
530 }
531 static DEVICE_ATTR_RW(zeroing_mode);
532 
533 static ssize_t
534 max_medium_access_timeouts_show(struct device *dev,
535 				struct device_attribute *attr, char *buf)
536 {
537 	struct scsi_disk *sdkp = to_scsi_disk(dev);
538 
539 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
540 }
541 
542 static ssize_t
543 max_medium_access_timeouts_store(struct device *dev,
544 				 struct device_attribute *attr, const char *buf,
545 				 size_t count)
546 {
547 	struct scsi_disk *sdkp = to_scsi_disk(dev);
548 	int err;
549 
550 	if (!capable(CAP_SYS_ADMIN))
551 		return -EACCES;
552 
553 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
554 
555 	return err ? err : count;
556 }
557 static DEVICE_ATTR_RW(max_medium_access_timeouts);
558 
559 static ssize_t
560 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
561 			   char *buf)
562 {
563 	struct scsi_disk *sdkp = to_scsi_disk(dev);
564 
565 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
566 }
567 
568 static ssize_t
569 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
570 			    const char *buf, size_t count)
571 {
572 	struct scsi_disk *sdkp = to_scsi_disk(dev);
573 	struct scsi_device *sdp = sdkp->device;
574 	struct queue_limits lim;
575 	unsigned long max;
576 	int err;
577 
578 	if (!capable(CAP_SYS_ADMIN))
579 		return -EACCES;
580 
581 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
582 		return -EINVAL;
583 
584 	err = kstrtoul(buf, 10, &max);
585 
586 	if (err)
587 		return err;
588 
589 	if (max == 0)
590 		sdp->no_write_same = 1;
591 	else if (max <= SD_MAX_WS16_BLOCKS) {
592 		sdp->no_write_same = 0;
593 		sdkp->max_ws_blocks = max;
594 	}
595 
596 	lim = queue_limits_start_update(sdkp->disk->queue);
597 	sd_config_write_same(sdkp, &lim);
598 	blk_mq_freeze_queue(sdkp->disk->queue);
599 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
600 	blk_mq_unfreeze_queue(sdkp->disk->queue);
601 	if (err)
602 		return err;
603 	return count;
604 }
605 static DEVICE_ATTR_RW(max_write_same_blocks);
606 
607 static ssize_t
608 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
609 {
610 	struct scsi_disk *sdkp = to_scsi_disk(dev);
611 
612 	if (sdkp->device->type == TYPE_ZBC)
613 		return sprintf(buf, "host-managed\n");
614 	if (sdkp->zoned == 1)
615 		return sprintf(buf, "host-aware\n");
616 	if (sdkp->zoned == 2)
617 		return sprintf(buf, "drive-managed\n");
618 	return sprintf(buf, "none\n");
619 }
620 static DEVICE_ATTR_RO(zoned_cap);
621 
622 static ssize_t
623 max_retries_store(struct device *dev, struct device_attribute *attr,
624 		  const char *buf, size_t count)
625 {
626 	struct scsi_disk *sdkp = to_scsi_disk(dev);
627 	struct scsi_device *sdev = sdkp->device;
628 	int retries, err;
629 
630 	err = kstrtoint(buf, 10, &retries);
631 	if (err)
632 		return err;
633 
634 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
635 		sdkp->max_retries = retries;
636 		return count;
637 	}
638 
639 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
640 		    SD_MAX_RETRIES);
641 	return -EINVAL;
642 }
643 
644 static ssize_t
645 max_retries_show(struct device *dev, struct device_attribute *attr,
646 		 char *buf)
647 {
648 	struct scsi_disk *sdkp = to_scsi_disk(dev);
649 
650 	return sprintf(buf, "%d\n", sdkp->max_retries);
651 }
652 
653 static DEVICE_ATTR_RW(max_retries);
654 
655 static struct attribute *sd_disk_attrs[] = {
656 	&dev_attr_cache_type.attr,
657 	&dev_attr_FUA.attr,
658 	&dev_attr_allow_restart.attr,
659 	&dev_attr_manage_start_stop.attr,
660 	&dev_attr_manage_system_start_stop.attr,
661 	&dev_attr_manage_runtime_start_stop.attr,
662 	&dev_attr_manage_shutdown.attr,
663 	&dev_attr_protection_type.attr,
664 	&dev_attr_protection_mode.attr,
665 	&dev_attr_app_tag_own.attr,
666 	&dev_attr_thin_provisioning.attr,
667 	&dev_attr_provisioning_mode.attr,
668 	&dev_attr_zeroing_mode.attr,
669 	&dev_attr_max_write_same_blocks.attr,
670 	&dev_attr_max_medium_access_timeouts.attr,
671 	&dev_attr_zoned_cap.attr,
672 	&dev_attr_max_retries.attr,
673 	NULL,
674 };
675 ATTRIBUTE_GROUPS(sd_disk);
676 
677 static struct class sd_disk_class = {
678 	.name		= "scsi_disk",
679 	.dev_release	= scsi_disk_release,
680 	.dev_groups	= sd_disk_groups,
681 };
682 
683 /*
684  * Don't request a new module, as that could deadlock in multipath
685  * environment.
686  */
687 static void sd_default_probe(dev_t devt)
688 {
689 }
690 
691 /*
692  * Device no to disk mapping:
693  *
694  *       major         disc2     disc  p1
695  *   |............|.............|....|....| <- dev_t
696  *    31        20 19          8 7  4 3  0
697  *
698  * Inside a major, we have 16k disks, however mapped non-
699  * contiguously. The first 16 disks are for major0, the next
700  * ones with major1, ... Disk 256 is for major0 again, disk 272
701  * for major1, ...
702  * As we stay compatible with our numbering scheme, we can reuse
703  * the well-know SCSI majors 8, 65--71, 136--143.
704  */
705 static int sd_major(int major_idx)
706 {
707 	switch (major_idx) {
708 	case 0:
709 		return SCSI_DISK0_MAJOR;
710 	case 1 ... 7:
711 		return SCSI_DISK1_MAJOR + major_idx - 1;
712 	case 8 ... 15:
713 		return SCSI_DISK8_MAJOR + major_idx - 8;
714 	default:
715 		BUG();
716 		return 0;	/* shut up gcc */
717 	}
718 }
719 
720 #ifdef CONFIG_BLK_SED_OPAL
721 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
722 		size_t len, bool send)
723 {
724 	struct scsi_disk *sdkp = data;
725 	struct scsi_device *sdev = sdkp->device;
726 	u8 cdb[12] = { 0, };
727 	const struct scsi_exec_args exec_args = {
728 		.req_flags = BLK_MQ_REQ_PM,
729 	};
730 	int ret;
731 
732 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
733 	cdb[1] = secp;
734 	put_unaligned_be16(spsp, &cdb[2]);
735 	put_unaligned_be32(len, &cdb[6]);
736 
737 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
738 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
739 			       &exec_args);
740 	return ret <= 0 ? ret : -EIO;
741 }
742 #endif /* CONFIG_BLK_SED_OPAL */
743 
744 /*
745  * Look up the DIX operation based on whether the command is read or
746  * write and whether dix and dif are enabled.
747  */
748 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
749 {
750 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
751 	static const unsigned int ops[] = {	/* wrt dix dif */
752 		SCSI_PROT_NORMAL,		/*  0	0   0  */
753 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
754 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
755 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
756 		SCSI_PROT_NORMAL,		/*  1	0   0  */
757 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
758 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
759 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
760 	};
761 
762 	return ops[write << 2 | dix << 1 | dif];
763 }
764 
765 /*
766  * Returns a mask of the protection flags that are valid for a given DIX
767  * operation.
768  */
769 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
770 {
771 	static const unsigned int flag_mask[] = {
772 		[SCSI_PROT_NORMAL]		= 0,
773 
774 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
775 						  SCSI_PROT_GUARD_CHECK |
776 						  SCSI_PROT_REF_CHECK |
777 						  SCSI_PROT_REF_INCREMENT,
778 
779 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
780 						  SCSI_PROT_IP_CHECKSUM,
781 
782 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
783 						  SCSI_PROT_GUARD_CHECK |
784 						  SCSI_PROT_REF_CHECK |
785 						  SCSI_PROT_REF_INCREMENT |
786 						  SCSI_PROT_IP_CHECKSUM,
787 
788 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
789 						  SCSI_PROT_REF_INCREMENT,
790 
791 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
792 						  SCSI_PROT_REF_CHECK |
793 						  SCSI_PROT_REF_INCREMENT |
794 						  SCSI_PROT_IP_CHECKSUM,
795 
796 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
797 						  SCSI_PROT_GUARD_CHECK |
798 						  SCSI_PROT_REF_CHECK |
799 						  SCSI_PROT_REF_INCREMENT |
800 						  SCSI_PROT_IP_CHECKSUM,
801 	};
802 
803 	return flag_mask[prot_op];
804 }
805 
806 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
807 					   unsigned int dix, unsigned int dif)
808 {
809 	struct request *rq = scsi_cmd_to_rq(scmd);
810 	struct bio *bio = rq->bio;
811 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
812 	unsigned int protect = 0;
813 
814 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
815 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
816 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
817 
818 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
819 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
820 	}
821 
822 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
823 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
824 
825 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
826 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
827 	}
828 
829 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
830 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
831 
832 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
833 			protect = 3 << 5;	/* Disable target PI checking */
834 		else
835 			protect = 1 << 5;	/* Enable target PI checking */
836 	}
837 
838 	scsi_set_prot_op(scmd, prot_op);
839 	scsi_set_prot_type(scmd, dif);
840 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
841 
842 	return protect;
843 }
844 
845 static void sd_disable_discard(struct scsi_disk *sdkp)
846 {
847 	sdkp->provisioning_mode = SD_LBP_DISABLE;
848 	blk_queue_disable_discard(sdkp->disk->queue);
849 }
850 
851 static void sd_config_discard(struct scsi_disk *sdkp, struct queue_limits *lim,
852 		unsigned int mode)
853 {
854 	unsigned int logical_block_size = sdkp->device->sector_size;
855 	unsigned int max_blocks = 0;
856 
857 	lim->discard_alignment = sdkp->unmap_alignment * logical_block_size;
858 	lim->discard_granularity = max(sdkp->physical_block_size,
859 			sdkp->unmap_granularity * logical_block_size);
860 	sdkp->provisioning_mode = mode;
861 
862 	switch (mode) {
863 
864 	case SD_LBP_FULL:
865 	case SD_LBP_DISABLE:
866 		break;
867 
868 	case SD_LBP_UNMAP:
869 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
870 					  (u32)SD_MAX_WS16_BLOCKS);
871 		break;
872 
873 	case SD_LBP_WS16:
874 		if (sdkp->device->unmap_limit_for_ws)
875 			max_blocks = sdkp->max_unmap_blocks;
876 		else
877 			max_blocks = sdkp->max_ws_blocks;
878 
879 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
880 		break;
881 
882 	case SD_LBP_WS10:
883 		if (sdkp->device->unmap_limit_for_ws)
884 			max_blocks = sdkp->max_unmap_blocks;
885 		else
886 			max_blocks = sdkp->max_ws_blocks;
887 
888 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
889 		break;
890 
891 	case SD_LBP_ZERO:
892 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
893 					  (u32)SD_MAX_WS10_BLOCKS);
894 		break;
895 	}
896 
897 	lim->max_hw_discard_sectors = max_blocks *
898 		(logical_block_size >> SECTOR_SHIFT);
899 }
900 
901 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
902 {
903 	struct page *page;
904 
905 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
906 	if (!page)
907 		return NULL;
908 	clear_highpage(page);
909 	bvec_set_page(&rq->special_vec, page, data_len, 0);
910 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
911 	return bvec_virt(&rq->special_vec);
912 }
913 
914 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
915 {
916 	struct scsi_device *sdp = cmd->device;
917 	struct request *rq = scsi_cmd_to_rq(cmd);
918 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
919 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
920 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
921 	unsigned int data_len = 24;
922 	char *buf;
923 
924 	buf = sd_set_special_bvec(rq, data_len);
925 	if (!buf)
926 		return BLK_STS_RESOURCE;
927 
928 	cmd->cmd_len = 10;
929 	cmd->cmnd[0] = UNMAP;
930 	cmd->cmnd[8] = 24;
931 
932 	put_unaligned_be16(6 + 16, &buf[0]);
933 	put_unaligned_be16(16, &buf[2]);
934 	put_unaligned_be64(lba, &buf[8]);
935 	put_unaligned_be32(nr_blocks, &buf[16]);
936 
937 	cmd->allowed = sdkp->max_retries;
938 	cmd->transfersize = data_len;
939 	rq->timeout = SD_TIMEOUT;
940 
941 	return scsi_alloc_sgtables(cmd);
942 }
943 
944 static void sd_config_atomic(struct scsi_disk *sdkp, struct queue_limits *lim)
945 {
946 	unsigned int logical_block_size = sdkp->device->sector_size,
947 		physical_block_size_sectors, max_atomic, unit_min, unit_max;
948 
949 	if ((!sdkp->max_atomic && !sdkp->max_atomic_with_boundary) ||
950 	    sdkp->protection_type == T10_PI_TYPE2_PROTECTION)
951 		return;
952 
953 	physical_block_size_sectors = sdkp->physical_block_size /
954 					sdkp->device->sector_size;
955 
956 	unit_min = rounddown_pow_of_two(sdkp->atomic_granularity ?
957 					sdkp->atomic_granularity :
958 					physical_block_size_sectors);
959 
960 	/*
961 	 * Only use atomic boundary when we have the odd scenario of
962 	 * sdkp->max_atomic == 0, which the spec does permit.
963 	 */
964 	if (sdkp->max_atomic) {
965 		max_atomic = sdkp->max_atomic;
966 		unit_max = rounddown_pow_of_two(sdkp->max_atomic);
967 		sdkp->use_atomic_write_boundary = 0;
968 	} else {
969 		max_atomic = sdkp->max_atomic_with_boundary;
970 		unit_max = rounddown_pow_of_two(sdkp->max_atomic_boundary);
971 		sdkp->use_atomic_write_boundary = 1;
972 	}
973 
974 	/*
975 	 * Ensure compliance with granularity and alignment. For now, keep it
976 	 * simple and just don't support atomic writes for values mismatched
977 	 * with max_{boundary}atomic, physical block size, and
978 	 * atomic_granularity itself.
979 	 *
980 	 * We're really being distrustful by checking unit_max also...
981 	 */
982 	if (sdkp->atomic_granularity > 1) {
983 		if (unit_min > 1 && unit_min % sdkp->atomic_granularity)
984 			return;
985 		if (unit_max > 1 && unit_max % sdkp->atomic_granularity)
986 			return;
987 	}
988 
989 	if (sdkp->atomic_alignment > 1) {
990 		if (unit_min > 1 && unit_min % sdkp->atomic_alignment)
991 			return;
992 		if (unit_max > 1 && unit_max % sdkp->atomic_alignment)
993 			return;
994 	}
995 
996 	lim->atomic_write_hw_max = max_atomic * logical_block_size;
997 	lim->atomic_write_hw_boundary = 0;
998 	lim->atomic_write_hw_unit_min = unit_min * logical_block_size;
999 	lim->atomic_write_hw_unit_max = unit_max * logical_block_size;
1000 }
1001 
1002 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
1003 		bool unmap)
1004 {
1005 	struct scsi_device *sdp = cmd->device;
1006 	struct request *rq = scsi_cmd_to_rq(cmd);
1007 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1008 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1009 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1010 	u32 data_len = sdp->sector_size;
1011 
1012 	if (!sd_set_special_bvec(rq, data_len))
1013 		return BLK_STS_RESOURCE;
1014 
1015 	cmd->cmd_len = 16;
1016 	cmd->cmnd[0] = WRITE_SAME_16;
1017 	if (unmap)
1018 		cmd->cmnd[1] = 0x8; /* UNMAP */
1019 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1020 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1021 
1022 	cmd->allowed = sdkp->max_retries;
1023 	cmd->transfersize = data_len;
1024 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
1025 
1026 	return scsi_alloc_sgtables(cmd);
1027 }
1028 
1029 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
1030 		bool unmap)
1031 {
1032 	struct scsi_device *sdp = cmd->device;
1033 	struct request *rq = scsi_cmd_to_rq(cmd);
1034 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1035 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1036 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1037 	u32 data_len = sdp->sector_size;
1038 
1039 	if (!sd_set_special_bvec(rq, data_len))
1040 		return BLK_STS_RESOURCE;
1041 
1042 	cmd->cmd_len = 10;
1043 	cmd->cmnd[0] = WRITE_SAME;
1044 	if (unmap)
1045 		cmd->cmnd[1] = 0x8; /* UNMAP */
1046 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1047 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1048 
1049 	cmd->allowed = sdkp->max_retries;
1050 	cmd->transfersize = data_len;
1051 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
1052 
1053 	return scsi_alloc_sgtables(cmd);
1054 }
1055 
1056 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
1057 {
1058 	struct request *rq = scsi_cmd_to_rq(cmd);
1059 	struct scsi_device *sdp = cmd->device;
1060 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1061 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1062 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1063 
1064 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
1065 		switch (sdkp->zeroing_mode) {
1066 		case SD_ZERO_WS16_UNMAP:
1067 			return sd_setup_write_same16_cmnd(cmd, true);
1068 		case SD_ZERO_WS10_UNMAP:
1069 			return sd_setup_write_same10_cmnd(cmd, true);
1070 		}
1071 	}
1072 
1073 	if (sdp->no_write_same) {
1074 		rq->rq_flags |= RQF_QUIET;
1075 		return BLK_STS_TARGET;
1076 	}
1077 
1078 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
1079 		return sd_setup_write_same16_cmnd(cmd, false);
1080 
1081 	return sd_setup_write_same10_cmnd(cmd, false);
1082 }
1083 
1084 static void sd_disable_write_same(struct scsi_disk *sdkp)
1085 {
1086 	sdkp->device->no_write_same = 1;
1087 	sdkp->max_ws_blocks = 0;
1088 	blk_queue_disable_write_zeroes(sdkp->disk->queue);
1089 }
1090 
1091 static void sd_config_write_same(struct scsi_disk *sdkp,
1092 		struct queue_limits *lim)
1093 {
1094 	unsigned int logical_block_size = sdkp->device->sector_size;
1095 
1096 	if (sdkp->device->no_write_same) {
1097 		sdkp->max_ws_blocks = 0;
1098 		goto out;
1099 	}
1100 
1101 	/* Some devices can not handle block counts above 0xffff despite
1102 	 * supporting WRITE SAME(16). Consequently we default to 64k
1103 	 * blocks per I/O unless the device explicitly advertises a
1104 	 * bigger limit.
1105 	 */
1106 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
1107 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1108 						   (u32)SD_MAX_WS16_BLOCKS);
1109 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
1110 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1111 						   (u32)SD_MAX_WS10_BLOCKS);
1112 	else {
1113 		sdkp->device->no_write_same = 1;
1114 		sdkp->max_ws_blocks = 0;
1115 	}
1116 
1117 	if (sdkp->lbprz && sdkp->lbpws)
1118 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
1119 	else if (sdkp->lbprz && sdkp->lbpws10)
1120 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
1121 	else if (sdkp->max_ws_blocks)
1122 		sdkp->zeroing_mode = SD_ZERO_WS;
1123 	else
1124 		sdkp->zeroing_mode = SD_ZERO_WRITE;
1125 
1126 	if (sdkp->max_ws_blocks &&
1127 	    sdkp->physical_block_size > logical_block_size) {
1128 		/*
1129 		 * Reporting a maximum number of blocks that is not aligned
1130 		 * on the device physical size would cause a large write same
1131 		 * request to be split into physically unaligned chunks by
1132 		 * __blkdev_issue_write_zeroes() even if the caller of this
1133 		 * functions took care to align the large request. So make sure
1134 		 * the maximum reported is aligned to the device physical block
1135 		 * size. This is only an optional optimization for regular
1136 		 * disks, but this is mandatory to avoid failure of large write
1137 		 * same requests directed at sequential write required zones of
1138 		 * host-managed ZBC disks.
1139 		 */
1140 		sdkp->max_ws_blocks =
1141 			round_down(sdkp->max_ws_blocks,
1142 				   bytes_to_logical(sdkp->device,
1143 						    sdkp->physical_block_size));
1144 	}
1145 
1146 out:
1147 	lim->max_write_zeroes_sectors =
1148 		sdkp->max_ws_blocks * (logical_block_size >> SECTOR_SHIFT);
1149 }
1150 
1151 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1152 {
1153 	struct request *rq = scsi_cmd_to_rq(cmd);
1154 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1155 
1156 	/* flush requests don't perform I/O, zero the S/G table */
1157 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1158 
1159 	if (cmd->device->use_16_for_sync) {
1160 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1161 		cmd->cmd_len = 16;
1162 	} else {
1163 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1164 		cmd->cmd_len = 10;
1165 	}
1166 	cmd->transfersize = 0;
1167 	cmd->allowed = sdkp->max_retries;
1168 
1169 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1170 	return BLK_STS_OK;
1171 }
1172 
1173 /**
1174  * sd_group_number() - Compute the GROUP NUMBER field
1175  * @cmd: SCSI command for which to compute the value of the six-bit GROUP NUMBER
1176  *	field.
1177  *
1178  * From SBC-5 r05 (https://www.t10.org/cgi-bin/ac.pl?t=f&f=sbc5r05.pdf):
1179  * 0: no relative lifetime.
1180  * 1: shortest relative lifetime.
1181  * 2: second shortest relative lifetime.
1182  * 3 - 0x3d: intermediate relative lifetimes.
1183  * 0x3e: second longest relative lifetime.
1184  * 0x3f: longest relative lifetime.
1185  */
1186 static u8 sd_group_number(struct scsi_cmnd *cmd)
1187 {
1188 	const struct request *rq = scsi_cmd_to_rq(cmd);
1189 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1190 
1191 	if (!sdkp->rscs)
1192 		return 0;
1193 
1194 	return min3((u32)rq->write_hint, (u32)sdkp->permanent_stream_count,
1195 		    0x3fu);
1196 }
1197 
1198 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1199 				       sector_t lba, unsigned int nr_blocks,
1200 				       unsigned char flags, unsigned int dld)
1201 {
1202 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1203 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1204 	cmd->cmnd[6]  = sd_group_number(cmd);
1205 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1206 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1207 	cmd->cmnd[10] = flags;
1208 	cmd->cmnd[11] = dld & 0x07;
1209 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1210 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1211 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1212 
1213 	return BLK_STS_OK;
1214 }
1215 
1216 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1217 				       sector_t lba, unsigned int nr_blocks,
1218 				       unsigned char flags, unsigned int dld)
1219 {
1220 	cmd->cmd_len  = 16;
1221 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1222 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1223 	cmd->cmnd[14] = ((dld & 0x03) << 6) | sd_group_number(cmd);
1224 	cmd->cmnd[15] = 0;
1225 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1226 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1227 
1228 	return BLK_STS_OK;
1229 }
1230 
1231 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1232 				       sector_t lba, unsigned int nr_blocks,
1233 				       unsigned char flags)
1234 {
1235 	cmd->cmd_len = 10;
1236 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1237 	cmd->cmnd[1] = flags;
1238 	cmd->cmnd[6] = sd_group_number(cmd);
1239 	cmd->cmnd[9] = 0;
1240 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1241 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1242 
1243 	return BLK_STS_OK;
1244 }
1245 
1246 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1247 				      sector_t lba, unsigned int nr_blocks,
1248 				      unsigned char flags)
1249 {
1250 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1251 	if (WARN_ON_ONCE(nr_blocks == 0))
1252 		return BLK_STS_IOERR;
1253 
1254 	if (unlikely(flags & 0x8)) {
1255 		/*
1256 		 * This happens only if this drive failed 10byte rw
1257 		 * command with ILLEGAL_REQUEST during operation and
1258 		 * thus turned off use_10_for_rw.
1259 		 */
1260 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1261 		return BLK_STS_IOERR;
1262 	}
1263 
1264 	cmd->cmd_len = 6;
1265 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1266 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1267 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1268 	cmd->cmnd[3] = lba & 0xff;
1269 	cmd->cmnd[4] = nr_blocks;
1270 	cmd->cmnd[5] = 0;
1271 
1272 	return BLK_STS_OK;
1273 }
1274 
1275 /*
1276  * Check if a command has a duration limit set. If it does, and the target
1277  * device supports CDL and the feature is enabled, return the limit
1278  * descriptor index to use. Return 0 (no limit) otherwise.
1279  */
1280 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1281 {
1282 	struct scsi_device *sdp = sdkp->device;
1283 	int hint;
1284 
1285 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1286 		return 0;
1287 
1288 	/*
1289 	 * Use "no limit" if the request ioprio does not specify a duration
1290 	 * limit hint.
1291 	 */
1292 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1293 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1294 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1295 		return 0;
1296 
1297 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1298 }
1299 
1300 static blk_status_t sd_setup_atomic_cmnd(struct scsi_cmnd *cmd,
1301 					sector_t lba, unsigned int nr_blocks,
1302 					bool boundary, unsigned char flags)
1303 {
1304 	cmd->cmd_len  = 16;
1305 	cmd->cmnd[0]  = WRITE_ATOMIC_16;
1306 	cmd->cmnd[1]  = flags;
1307 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1308 	put_unaligned_be16(nr_blocks, &cmd->cmnd[12]);
1309 	if (boundary)
1310 		put_unaligned_be16(nr_blocks, &cmd->cmnd[10]);
1311 	else
1312 		put_unaligned_be16(0, &cmd->cmnd[10]);
1313 	put_unaligned_be16(nr_blocks, &cmd->cmnd[12]);
1314 	cmd->cmnd[14] = 0;
1315 	cmd->cmnd[15] = 0;
1316 
1317 	return BLK_STS_OK;
1318 }
1319 
1320 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1321 {
1322 	struct request *rq = scsi_cmd_to_rq(cmd);
1323 	struct scsi_device *sdp = cmd->device;
1324 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1325 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1326 	sector_t threshold;
1327 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1328 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1329 	bool write = rq_data_dir(rq) == WRITE;
1330 	unsigned char protect, fua;
1331 	unsigned int dld;
1332 	blk_status_t ret;
1333 	unsigned int dif;
1334 	bool dix;
1335 
1336 	ret = scsi_alloc_sgtables(cmd);
1337 	if (ret != BLK_STS_OK)
1338 		return ret;
1339 
1340 	ret = BLK_STS_IOERR;
1341 	if (!scsi_device_online(sdp) || sdp->changed) {
1342 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1343 		goto fail;
1344 	}
1345 
1346 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1347 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1348 		goto fail;
1349 	}
1350 
1351 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1352 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1353 		goto fail;
1354 	}
1355 
1356 	/*
1357 	 * Some SD card readers can't handle accesses which touch the
1358 	 * last one or two logical blocks. Split accesses as needed.
1359 	 */
1360 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1361 
1362 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1363 		if (lba < threshold) {
1364 			/* Access up to the threshold but not beyond */
1365 			nr_blocks = threshold - lba;
1366 		} else {
1367 			/* Access only a single logical block */
1368 			nr_blocks = 1;
1369 		}
1370 	}
1371 
1372 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1373 	dix = scsi_prot_sg_count(cmd);
1374 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1375 	dld = sd_cdl_dld(sdkp, cmd);
1376 
1377 	if (dif || dix)
1378 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1379 	else
1380 		protect = 0;
1381 
1382 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1383 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1384 					 protect | fua, dld);
1385 	} else if (rq->cmd_flags & REQ_ATOMIC && write) {
1386 		ret = sd_setup_atomic_cmnd(cmd, lba, nr_blocks,
1387 				sdkp->use_atomic_write_boundary,
1388 				protect | fua);
1389 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1390 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1391 					 protect | fua, dld);
1392 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1393 		   sdp->use_10_for_rw || protect || rq->write_hint) {
1394 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1395 					 protect | fua);
1396 	} else {
1397 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1398 					protect | fua);
1399 	}
1400 
1401 	if (unlikely(ret != BLK_STS_OK))
1402 		goto fail;
1403 
1404 	/*
1405 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1406 	 * host adapter, it's safe to assume that we can at least transfer
1407 	 * this many bytes between each connect / disconnect.
1408 	 */
1409 	cmd->transfersize = sdp->sector_size;
1410 	cmd->underflow = nr_blocks << 9;
1411 	cmd->allowed = sdkp->max_retries;
1412 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1413 
1414 	SCSI_LOG_HLQUEUE(1,
1415 			 scmd_printk(KERN_INFO, cmd,
1416 				     "%s: block=%llu, count=%d\n", __func__,
1417 				     (unsigned long long)blk_rq_pos(rq),
1418 				     blk_rq_sectors(rq)));
1419 	SCSI_LOG_HLQUEUE(2,
1420 			 scmd_printk(KERN_INFO, cmd,
1421 				     "%s %d/%u 512 byte blocks.\n",
1422 				     write ? "writing" : "reading", nr_blocks,
1423 				     blk_rq_sectors(rq)));
1424 
1425 	/*
1426 	 * This indicates that the command is ready from our end to be queued.
1427 	 */
1428 	return BLK_STS_OK;
1429 fail:
1430 	scsi_free_sgtables(cmd);
1431 	return ret;
1432 }
1433 
1434 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1435 {
1436 	struct request *rq = scsi_cmd_to_rq(cmd);
1437 
1438 	switch (req_op(rq)) {
1439 	case REQ_OP_DISCARD:
1440 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1441 		case SD_LBP_UNMAP:
1442 			return sd_setup_unmap_cmnd(cmd);
1443 		case SD_LBP_WS16:
1444 			return sd_setup_write_same16_cmnd(cmd, true);
1445 		case SD_LBP_WS10:
1446 			return sd_setup_write_same10_cmnd(cmd, true);
1447 		case SD_LBP_ZERO:
1448 			return sd_setup_write_same10_cmnd(cmd, false);
1449 		default:
1450 			return BLK_STS_TARGET;
1451 		}
1452 	case REQ_OP_WRITE_ZEROES:
1453 		return sd_setup_write_zeroes_cmnd(cmd);
1454 	case REQ_OP_FLUSH:
1455 		return sd_setup_flush_cmnd(cmd);
1456 	case REQ_OP_READ:
1457 	case REQ_OP_WRITE:
1458 		return sd_setup_read_write_cmnd(cmd);
1459 	case REQ_OP_ZONE_RESET:
1460 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1461 						   false);
1462 	case REQ_OP_ZONE_RESET_ALL:
1463 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1464 						   true);
1465 	case REQ_OP_ZONE_OPEN:
1466 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1467 	case REQ_OP_ZONE_CLOSE:
1468 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1469 	case REQ_OP_ZONE_FINISH:
1470 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1471 	default:
1472 		WARN_ON_ONCE(1);
1473 		return BLK_STS_NOTSUPP;
1474 	}
1475 }
1476 
1477 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1478 {
1479 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1480 
1481 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1482 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1483 }
1484 
1485 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1486 {
1487 	if (sdkp->device->removable || sdkp->write_prot) {
1488 		if (disk_check_media_change(disk))
1489 			return true;
1490 	}
1491 
1492 	/*
1493 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1494 	 * nothing to do with partitions, BLKRRPART is used to force a full
1495 	 * revalidate after things like a format for historical reasons.
1496 	 */
1497 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1498 }
1499 
1500 /**
1501  *	sd_open - open a scsi disk device
1502  *	@disk: disk to open
1503  *	@mode: open mode
1504  *
1505  *	Returns 0 if successful. Returns a negated errno value in case
1506  *	of error.
1507  *
1508  *	Note: This can be called from a user context (e.g. fsck(1) )
1509  *	or from within the kernel (e.g. as a result of a mount(1) ).
1510  *	In the latter case @inode and @filp carry an abridged amount
1511  *	of information as noted above.
1512  *
1513  *	Locking: called with disk->open_mutex held.
1514  **/
1515 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1516 {
1517 	struct scsi_disk *sdkp = scsi_disk(disk);
1518 	struct scsi_device *sdev = sdkp->device;
1519 	int retval;
1520 
1521 	if (scsi_device_get(sdev))
1522 		return -ENXIO;
1523 
1524 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1525 
1526 	/*
1527 	 * If the device is in error recovery, wait until it is done.
1528 	 * If the device is offline, then disallow any access to it.
1529 	 */
1530 	retval = -ENXIO;
1531 	if (!scsi_block_when_processing_errors(sdev))
1532 		goto error_out;
1533 
1534 	if (sd_need_revalidate(disk, sdkp))
1535 		sd_revalidate_disk(disk);
1536 
1537 	/*
1538 	 * If the drive is empty, just let the open fail.
1539 	 */
1540 	retval = -ENOMEDIUM;
1541 	if (sdev->removable && !sdkp->media_present &&
1542 	    !(mode & BLK_OPEN_NDELAY))
1543 		goto error_out;
1544 
1545 	/*
1546 	 * If the device has the write protect tab set, have the open fail
1547 	 * if the user expects to be able to write to the thing.
1548 	 */
1549 	retval = -EROFS;
1550 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1551 		goto error_out;
1552 
1553 	/*
1554 	 * It is possible that the disk changing stuff resulted in
1555 	 * the device being taken offline.  If this is the case,
1556 	 * report this to the user, and don't pretend that the
1557 	 * open actually succeeded.
1558 	 */
1559 	retval = -ENXIO;
1560 	if (!scsi_device_online(sdev))
1561 		goto error_out;
1562 
1563 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1564 		if (scsi_block_when_processing_errors(sdev))
1565 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1566 	}
1567 
1568 	return 0;
1569 
1570 error_out:
1571 	scsi_device_put(sdev);
1572 	return retval;
1573 }
1574 
1575 /**
1576  *	sd_release - invoked when the (last) close(2) is called on this
1577  *	scsi disk.
1578  *	@disk: disk to release
1579  *
1580  *	Returns 0.
1581  *
1582  *	Note: may block (uninterruptible) if error recovery is underway
1583  *	on this disk.
1584  *
1585  *	Locking: called with disk->open_mutex held.
1586  **/
1587 static void sd_release(struct gendisk *disk)
1588 {
1589 	struct scsi_disk *sdkp = scsi_disk(disk);
1590 	struct scsi_device *sdev = sdkp->device;
1591 
1592 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1593 
1594 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1595 		if (scsi_block_when_processing_errors(sdev))
1596 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1597 	}
1598 
1599 	scsi_device_put(sdev);
1600 }
1601 
1602 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1603 {
1604 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1605 	struct scsi_device *sdp = sdkp->device;
1606 	struct Scsi_Host *host = sdp->host;
1607 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1608 	int diskinfo[4];
1609 
1610 	/* default to most commonly used values */
1611 	diskinfo[0] = 0x40;	/* 1 << 6 */
1612 	diskinfo[1] = 0x20;	/* 1 << 5 */
1613 	diskinfo[2] = capacity >> 11;
1614 
1615 	/* override with calculated, extended default, or driver values */
1616 	if (host->hostt->bios_param)
1617 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1618 	else
1619 		scsicam_bios_param(bdev, capacity, diskinfo);
1620 
1621 	geo->heads = diskinfo[0];
1622 	geo->sectors = diskinfo[1];
1623 	geo->cylinders = diskinfo[2];
1624 	return 0;
1625 }
1626 
1627 /**
1628  *	sd_ioctl - process an ioctl
1629  *	@bdev: target block device
1630  *	@mode: open mode
1631  *	@cmd: ioctl command number
1632  *	@arg: this is third argument given to ioctl(2) system call.
1633  *	Often contains a pointer.
1634  *
1635  *	Returns 0 if successful (some ioctls return positive numbers on
1636  *	success as well). Returns a negated errno value in case of error.
1637  *
1638  *	Note: most ioctls are forward onto the block subsystem or further
1639  *	down in the scsi subsystem.
1640  **/
1641 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1642 		    unsigned int cmd, unsigned long arg)
1643 {
1644 	struct gendisk *disk = bdev->bd_disk;
1645 	struct scsi_disk *sdkp = scsi_disk(disk);
1646 	struct scsi_device *sdp = sdkp->device;
1647 	void __user *p = (void __user *)arg;
1648 	int error;
1649 
1650 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1651 				    "cmd=0x%x\n", disk->disk_name, cmd));
1652 
1653 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1654 		return -ENOIOCTLCMD;
1655 
1656 	/*
1657 	 * If we are in the middle of error recovery, don't let anyone
1658 	 * else try and use this device.  Also, if error recovery fails, it
1659 	 * may try and take the device offline, in which case all further
1660 	 * access to the device is prohibited.
1661 	 */
1662 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1663 			(mode & BLK_OPEN_NDELAY));
1664 	if (error)
1665 		return error;
1666 
1667 	if (is_sed_ioctl(cmd))
1668 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1669 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1670 }
1671 
1672 static void set_media_not_present(struct scsi_disk *sdkp)
1673 {
1674 	if (sdkp->media_present)
1675 		sdkp->device->changed = 1;
1676 
1677 	if (sdkp->device->removable) {
1678 		sdkp->media_present = 0;
1679 		sdkp->capacity = 0;
1680 	}
1681 }
1682 
1683 static int media_not_present(struct scsi_disk *sdkp,
1684 			     struct scsi_sense_hdr *sshdr)
1685 {
1686 	if (!scsi_sense_valid(sshdr))
1687 		return 0;
1688 
1689 	/* not invoked for commands that could return deferred errors */
1690 	switch (sshdr->sense_key) {
1691 	case UNIT_ATTENTION:
1692 	case NOT_READY:
1693 		/* medium not present */
1694 		if (sshdr->asc == 0x3A) {
1695 			set_media_not_present(sdkp);
1696 			return 1;
1697 		}
1698 	}
1699 	return 0;
1700 }
1701 
1702 /**
1703  *	sd_check_events - check media events
1704  *	@disk: kernel device descriptor
1705  *	@clearing: disk events currently being cleared
1706  *
1707  *	Returns mask of DISK_EVENT_*.
1708  *
1709  *	Note: this function is invoked from the block subsystem.
1710  **/
1711 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1712 {
1713 	struct scsi_disk *sdkp = disk->private_data;
1714 	struct scsi_device *sdp;
1715 	int retval;
1716 	bool disk_changed;
1717 
1718 	if (!sdkp)
1719 		return 0;
1720 
1721 	sdp = sdkp->device;
1722 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1723 
1724 	/*
1725 	 * If the device is offline, don't send any commands - just pretend as
1726 	 * if the command failed.  If the device ever comes back online, we
1727 	 * can deal with it then.  It is only because of unrecoverable errors
1728 	 * that we would ever take a device offline in the first place.
1729 	 */
1730 	if (!scsi_device_online(sdp)) {
1731 		set_media_not_present(sdkp);
1732 		goto out;
1733 	}
1734 
1735 	/*
1736 	 * Using TEST_UNIT_READY enables differentiation between drive with
1737 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1738 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1739 	 *
1740 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1741 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1742 	 * sd_revalidate() is called.
1743 	 */
1744 	if (scsi_block_when_processing_errors(sdp)) {
1745 		struct scsi_sense_hdr sshdr = { 0, };
1746 
1747 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1748 					      &sshdr);
1749 
1750 		/* failed to execute TUR, assume media not present */
1751 		if (retval < 0 || host_byte(retval)) {
1752 			set_media_not_present(sdkp);
1753 			goto out;
1754 		}
1755 
1756 		if (media_not_present(sdkp, &sshdr))
1757 			goto out;
1758 	}
1759 
1760 	/*
1761 	 * For removable scsi disk we have to recognise the presence
1762 	 * of a disk in the drive.
1763 	 */
1764 	if (!sdkp->media_present)
1765 		sdp->changed = 1;
1766 	sdkp->media_present = 1;
1767 out:
1768 	/*
1769 	 * sdp->changed is set under the following conditions:
1770 	 *
1771 	 *	Medium present state has changed in either direction.
1772 	 *	Device has indicated UNIT_ATTENTION.
1773 	 */
1774 	disk_changed = sdp->changed;
1775 	sdp->changed = 0;
1776 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1777 }
1778 
1779 static int sd_sync_cache(struct scsi_disk *sdkp)
1780 {
1781 	int res;
1782 	struct scsi_device *sdp = sdkp->device;
1783 	const int timeout = sdp->request_queue->rq_timeout
1784 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1785 	/* Leave the rest of the command zero to indicate flush everything. */
1786 	const unsigned char cmd[16] = { sdp->use_16_for_sync ?
1787 				SYNCHRONIZE_CACHE_16 : SYNCHRONIZE_CACHE };
1788 	struct scsi_sense_hdr sshdr;
1789 	struct scsi_failure failure_defs[] = {
1790 		{
1791 			.allowed = 3,
1792 			.result = SCMD_FAILURE_RESULT_ANY,
1793 		},
1794 		{}
1795 	};
1796 	struct scsi_failures failures = {
1797 		.failure_definitions = failure_defs,
1798 	};
1799 	const struct scsi_exec_args exec_args = {
1800 		.req_flags = BLK_MQ_REQ_PM,
1801 		.sshdr = &sshdr,
1802 		.failures = &failures,
1803 	};
1804 
1805 	if (!scsi_device_online(sdp))
1806 		return -ENODEV;
1807 
1808 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, timeout,
1809 			       sdkp->max_retries, &exec_args);
1810 	if (res) {
1811 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1812 
1813 		if (res < 0)
1814 			return res;
1815 
1816 		if (scsi_status_is_check_condition(res) &&
1817 		    scsi_sense_valid(&sshdr)) {
1818 			sd_print_sense_hdr(sdkp, &sshdr);
1819 
1820 			/* we need to evaluate the error return  */
1821 			if (sshdr.asc == 0x3a ||	/* medium not present */
1822 			    sshdr.asc == 0x20 ||	/* invalid command */
1823 			    (sshdr.asc == 0x74 && sshdr.ascq == 0x71))	/* drive is password locked */
1824 				/* this is no error here */
1825 				return 0;
1826 			/*
1827 			 * This drive doesn't support sync and there's not much
1828 			 * we can do because this is called during shutdown
1829 			 * or suspend so just return success so those operations
1830 			 * can proceed.
1831 			 */
1832 			if (sshdr.sense_key == ILLEGAL_REQUEST)
1833 				return 0;
1834 		}
1835 
1836 		switch (host_byte(res)) {
1837 		/* ignore errors due to racing a disconnection */
1838 		case DID_BAD_TARGET:
1839 		case DID_NO_CONNECT:
1840 			return 0;
1841 		/* signal the upper layer it might try again */
1842 		case DID_BUS_BUSY:
1843 		case DID_IMM_RETRY:
1844 		case DID_REQUEUE:
1845 		case DID_SOFT_ERROR:
1846 			return -EBUSY;
1847 		default:
1848 			return -EIO;
1849 		}
1850 	}
1851 	return 0;
1852 }
1853 
1854 static void sd_rescan(struct device *dev)
1855 {
1856 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1857 
1858 	sd_revalidate_disk(sdkp->disk);
1859 }
1860 
1861 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1862 		enum blk_unique_id type)
1863 {
1864 	struct scsi_device *sdev = scsi_disk(disk)->device;
1865 	const struct scsi_vpd *vpd;
1866 	const unsigned char *d;
1867 	int ret = -ENXIO, len;
1868 
1869 	rcu_read_lock();
1870 	vpd = rcu_dereference(sdev->vpd_pg83);
1871 	if (!vpd)
1872 		goto out_unlock;
1873 
1874 	ret = -EINVAL;
1875 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1876 		/* we only care about designators with LU association */
1877 		if (((d[1] >> 4) & 0x3) != 0x00)
1878 			continue;
1879 		if ((d[1] & 0xf) != type)
1880 			continue;
1881 
1882 		/*
1883 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1884 		 * keep looking as one with more entropy might still show up.
1885 		 */
1886 		len = d[3];
1887 		if (len != 8 && len != 12 && len != 16)
1888 			continue;
1889 		ret = len;
1890 		memcpy(id, d + 4, len);
1891 		if (len == 16)
1892 			break;
1893 	}
1894 out_unlock:
1895 	rcu_read_unlock();
1896 	return ret;
1897 }
1898 
1899 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1900 {
1901 	switch (host_byte(result)) {
1902 	case DID_TRANSPORT_MARGINAL:
1903 	case DID_TRANSPORT_DISRUPTED:
1904 	case DID_BUS_BUSY:
1905 		return PR_STS_RETRY_PATH_FAILURE;
1906 	case DID_NO_CONNECT:
1907 		return PR_STS_PATH_FAILED;
1908 	case DID_TRANSPORT_FAILFAST:
1909 		return PR_STS_PATH_FAST_FAILED;
1910 	}
1911 
1912 	switch (status_byte(result)) {
1913 	case SAM_STAT_RESERVATION_CONFLICT:
1914 		return PR_STS_RESERVATION_CONFLICT;
1915 	case SAM_STAT_CHECK_CONDITION:
1916 		if (!scsi_sense_valid(sshdr))
1917 			return PR_STS_IOERR;
1918 
1919 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1920 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1921 			return -EINVAL;
1922 
1923 		fallthrough;
1924 	default:
1925 		return PR_STS_IOERR;
1926 	}
1927 }
1928 
1929 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1930 			    unsigned char *data, int data_len)
1931 {
1932 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1933 	struct scsi_device *sdev = sdkp->device;
1934 	struct scsi_sense_hdr sshdr;
1935 	u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1936 	struct scsi_failure failure_defs[] = {
1937 		{
1938 			.sense = UNIT_ATTENTION,
1939 			.asc = SCMD_FAILURE_ASC_ANY,
1940 			.ascq = SCMD_FAILURE_ASCQ_ANY,
1941 			.allowed = 5,
1942 			.result = SAM_STAT_CHECK_CONDITION,
1943 		},
1944 		{}
1945 	};
1946 	struct scsi_failures failures = {
1947 		.failure_definitions = failure_defs,
1948 	};
1949 	const struct scsi_exec_args exec_args = {
1950 		.sshdr = &sshdr,
1951 		.failures = &failures,
1952 	};
1953 	int result;
1954 
1955 	put_unaligned_be16(data_len, &cmd[7]);
1956 
1957 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1958 				  SD_TIMEOUT, sdkp->max_retries, &exec_args);
1959 	if (scsi_status_is_check_condition(result) &&
1960 	    scsi_sense_valid(&sshdr)) {
1961 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1962 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1963 	}
1964 
1965 	if (result <= 0)
1966 		return result;
1967 
1968 	return sd_scsi_to_pr_err(&sshdr, result);
1969 }
1970 
1971 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1972 {
1973 	int result, i, data_offset, num_copy_keys;
1974 	u32 num_keys = keys_info->num_keys;
1975 	int data_len = num_keys * 8 + 8;
1976 	u8 *data;
1977 
1978 	data = kzalloc(data_len, GFP_KERNEL);
1979 	if (!data)
1980 		return -ENOMEM;
1981 
1982 	result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1983 	if (result)
1984 		goto free_data;
1985 
1986 	keys_info->generation = get_unaligned_be32(&data[0]);
1987 	keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1988 
1989 	data_offset = 8;
1990 	num_copy_keys = min(num_keys, keys_info->num_keys);
1991 
1992 	for (i = 0; i < num_copy_keys; i++) {
1993 		keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1994 		data_offset += 8;
1995 	}
1996 
1997 free_data:
1998 	kfree(data);
1999 	return result;
2000 }
2001 
2002 static int sd_pr_read_reservation(struct block_device *bdev,
2003 				  struct pr_held_reservation *rsv)
2004 {
2005 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
2006 	struct scsi_device *sdev = sdkp->device;
2007 	u8 data[24] = { };
2008 	int result, len;
2009 
2010 	result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
2011 	if (result)
2012 		return result;
2013 
2014 	len = get_unaligned_be32(&data[4]);
2015 	if (!len)
2016 		return 0;
2017 
2018 	/* Make sure we have at least the key and type */
2019 	if (len < 14) {
2020 		sdev_printk(KERN_INFO, sdev,
2021 			    "READ RESERVATION failed due to short return buffer of %d bytes\n",
2022 			    len);
2023 		return -EINVAL;
2024 	}
2025 
2026 	rsv->generation = get_unaligned_be32(&data[0]);
2027 	rsv->key = get_unaligned_be64(&data[8]);
2028 	rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
2029 	return 0;
2030 }
2031 
2032 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
2033 			     u64 sa_key, enum scsi_pr_type type, u8 flags)
2034 {
2035 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
2036 	struct scsi_device *sdev = sdkp->device;
2037 	struct scsi_sense_hdr sshdr;
2038 	struct scsi_failure failure_defs[] = {
2039 		{
2040 			.sense = UNIT_ATTENTION,
2041 			.asc = SCMD_FAILURE_ASC_ANY,
2042 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2043 			.allowed = 5,
2044 			.result = SAM_STAT_CHECK_CONDITION,
2045 		},
2046 		{}
2047 	};
2048 	struct scsi_failures failures = {
2049 		.failure_definitions = failure_defs,
2050 	};
2051 	const struct scsi_exec_args exec_args = {
2052 		.sshdr = &sshdr,
2053 		.failures = &failures,
2054 	};
2055 	int result;
2056 	u8 cmd[16] = { 0, };
2057 	u8 data[24] = { 0, };
2058 
2059 	cmd[0] = PERSISTENT_RESERVE_OUT;
2060 	cmd[1] = sa;
2061 	cmd[2] = type;
2062 	put_unaligned_be32(sizeof(data), &cmd[5]);
2063 
2064 	put_unaligned_be64(key, &data[0]);
2065 	put_unaligned_be64(sa_key, &data[8]);
2066 	data[20] = flags;
2067 
2068 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
2069 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
2070 				  &exec_args);
2071 
2072 	if (scsi_status_is_check_condition(result) &&
2073 	    scsi_sense_valid(&sshdr)) {
2074 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
2075 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
2076 	}
2077 
2078 	if (result <= 0)
2079 		return result;
2080 
2081 	return sd_scsi_to_pr_err(&sshdr, result);
2082 }
2083 
2084 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
2085 		u32 flags)
2086 {
2087 	if (flags & ~PR_FL_IGNORE_KEY)
2088 		return -EOPNOTSUPP;
2089 	return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
2090 			old_key, new_key, 0,
2091 			(1 << 0) /* APTPL */);
2092 }
2093 
2094 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
2095 		u32 flags)
2096 {
2097 	if (flags)
2098 		return -EOPNOTSUPP;
2099 	return sd_pr_out_command(bdev, 0x01, key, 0,
2100 				 block_pr_type_to_scsi(type), 0);
2101 }
2102 
2103 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
2104 {
2105 	return sd_pr_out_command(bdev, 0x02, key, 0,
2106 				 block_pr_type_to_scsi(type), 0);
2107 }
2108 
2109 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
2110 		enum pr_type type, bool abort)
2111 {
2112 	return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
2113 				 block_pr_type_to_scsi(type), 0);
2114 }
2115 
2116 static int sd_pr_clear(struct block_device *bdev, u64 key)
2117 {
2118 	return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
2119 }
2120 
2121 static const struct pr_ops sd_pr_ops = {
2122 	.pr_register	= sd_pr_register,
2123 	.pr_reserve	= sd_pr_reserve,
2124 	.pr_release	= sd_pr_release,
2125 	.pr_preempt	= sd_pr_preempt,
2126 	.pr_clear	= sd_pr_clear,
2127 	.pr_read_keys	= sd_pr_read_keys,
2128 	.pr_read_reservation = sd_pr_read_reservation,
2129 };
2130 
2131 static void scsi_disk_free_disk(struct gendisk *disk)
2132 {
2133 	struct scsi_disk *sdkp = scsi_disk(disk);
2134 
2135 	put_device(&sdkp->disk_dev);
2136 }
2137 
2138 static const struct block_device_operations sd_fops = {
2139 	.owner			= THIS_MODULE,
2140 	.open			= sd_open,
2141 	.release		= sd_release,
2142 	.ioctl			= sd_ioctl,
2143 	.getgeo			= sd_getgeo,
2144 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
2145 	.check_events		= sd_check_events,
2146 	.unlock_native_capacity	= sd_unlock_native_capacity,
2147 	.report_zones		= sd_zbc_report_zones,
2148 	.get_unique_id		= sd_get_unique_id,
2149 	.free_disk		= scsi_disk_free_disk,
2150 	.pr_ops			= &sd_pr_ops,
2151 };
2152 
2153 /**
2154  *	sd_eh_reset - reset error handling callback
2155  *	@scmd:		sd-issued command that has failed
2156  *
2157  *	This function is called by the SCSI midlayer before starting
2158  *	SCSI EH. When counting medium access failures we have to be
2159  *	careful to register it only only once per device and SCSI EH run;
2160  *	there might be several timed out commands which will cause the
2161  *	'max_medium_access_timeouts' counter to trigger after the first
2162  *	SCSI EH run already and set the device to offline.
2163  *	So this function resets the internal counter before starting SCSI EH.
2164  **/
2165 static void sd_eh_reset(struct scsi_cmnd *scmd)
2166 {
2167 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2168 
2169 	/* New SCSI EH run, reset gate variable */
2170 	sdkp->ignore_medium_access_errors = false;
2171 }
2172 
2173 /**
2174  *	sd_eh_action - error handling callback
2175  *	@scmd:		sd-issued command that has failed
2176  *	@eh_disp:	The recovery disposition suggested by the midlayer
2177  *
2178  *	This function is called by the SCSI midlayer upon completion of an
2179  *	error test command (currently TEST UNIT READY). The result of sending
2180  *	the eh command is passed in eh_disp.  We're looking for devices that
2181  *	fail medium access commands but are OK with non access commands like
2182  *	test unit ready (so wrongly see the device as having a successful
2183  *	recovery)
2184  **/
2185 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
2186 {
2187 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2188 	struct scsi_device *sdev = scmd->device;
2189 
2190 	if (!scsi_device_online(sdev) ||
2191 	    !scsi_medium_access_command(scmd) ||
2192 	    host_byte(scmd->result) != DID_TIME_OUT ||
2193 	    eh_disp != SUCCESS)
2194 		return eh_disp;
2195 
2196 	/*
2197 	 * The device has timed out executing a medium access command.
2198 	 * However, the TEST UNIT READY command sent during error
2199 	 * handling completed successfully. Either the device is in the
2200 	 * process of recovering or has it suffered an internal failure
2201 	 * that prevents access to the storage medium.
2202 	 */
2203 	if (!sdkp->ignore_medium_access_errors) {
2204 		sdkp->medium_access_timed_out++;
2205 		sdkp->ignore_medium_access_errors = true;
2206 	}
2207 
2208 	/*
2209 	 * If the device keeps failing read/write commands but TEST UNIT
2210 	 * READY always completes successfully we assume that medium
2211 	 * access is no longer possible and take the device offline.
2212 	 */
2213 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2214 		scmd_printk(KERN_ERR, scmd,
2215 			    "Medium access timeout failure. Offlining disk!\n");
2216 		mutex_lock(&sdev->state_mutex);
2217 		scsi_device_set_state(sdev, SDEV_OFFLINE);
2218 		mutex_unlock(&sdev->state_mutex);
2219 
2220 		return SUCCESS;
2221 	}
2222 
2223 	return eh_disp;
2224 }
2225 
2226 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2227 {
2228 	struct request *req = scsi_cmd_to_rq(scmd);
2229 	struct scsi_device *sdev = scmd->device;
2230 	unsigned int transferred, good_bytes;
2231 	u64 start_lba, end_lba, bad_lba;
2232 
2233 	/*
2234 	 * Some commands have a payload smaller than the device logical
2235 	 * block size (e.g. INQUIRY on a 4K disk).
2236 	 */
2237 	if (scsi_bufflen(scmd) <= sdev->sector_size)
2238 		return 0;
2239 
2240 	/* Check if we have a 'bad_lba' information */
2241 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2242 				     SCSI_SENSE_BUFFERSIZE,
2243 				     &bad_lba))
2244 		return 0;
2245 
2246 	/*
2247 	 * If the bad lba was reported incorrectly, we have no idea where
2248 	 * the error is.
2249 	 */
2250 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2251 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2252 	if (bad_lba < start_lba || bad_lba >= end_lba)
2253 		return 0;
2254 
2255 	/*
2256 	 * resid is optional but mostly filled in.  When it's unused,
2257 	 * its value is zero, so we assume the whole buffer transferred
2258 	 */
2259 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2260 
2261 	/* This computation should always be done in terms of the
2262 	 * resolution of the device's medium.
2263 	 */
2264 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2265 
2266 	return min(good_bytes, transferred);
2267 }
2268 
2269 /**
2270  *	sd_done - bottom half handler: called when the lower level
2271  *	driver has completed (successfully or otherwise) a scsi command.
2272  *	@SCpnt: mid-level's per command structure.
2273  *
2274  *	Note: potentially run from within an ISR. Must not block.
2275  **/
2276 static int sd_done(struct scsi_cmnd *SCpnt)
2277 {
2278 	int result = SCpnt->result;
2279 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2280 	unsigned int sector_size = SCpnt->device->sector_size;
2281 	unsigned int resid;
2282 	struct scsi_sense_hdr sshdr;
2283 	struct request *req = scsi_cmd_to_rq(SCpnt);
2284 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2285 	int sense_valid = 0;
2286 	int sense_deferred = 0;
2287 
2288 	switch (req_op(req)) {
2289 	case REQ_OP_DISCARD:
2290 	case REQ_OP_WRITE_ZEROES:
2291 	case REQ_OP_ZONE_RESET:
2292 	case REQ_OP_ZONE_RESET_ALL:
2293 	case REQ_OP_ZONE_OPEN:
2294 	case REQ_OP_ZONE_CLOSE:
2295 	case REQ_OP_ZONE_FINISH:
2296 		if (!result) {
2297 			good_bytes = blk_rq_bytes(req);
2298 			scsi_set_resid(SCpnt, 0);
2299 		} else {
2300 			good_bytes = 0;
2301 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2302 		}
2303 		break;
2304 	default:
2305 		/*
2306 		 * In case of bogus fw or device, we could end up having
2307 		 * an unaligned partial completion. Check this here and force
2308 		 * alignment.
2309 		 */
2310 		resid = scsi_get_resid(SCpnt);
2311 		if (resid & (sector_size - 1)) {
2312 			sd_printk(KERN_INFO, sdkp,
2313 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2314 				resid, sector_size);
2315 			scsi_print_command(SCpnt);
2316 			resid = min(scsi_bufflen(SCpnt),
2317 				    round_up(resid, sector_size));
2318 			scsi_set_resid(SCpnt, resid);
2319 		}
2320 	}
2321 
2322 	if (result) {
2323 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2324 		if (sense_valid)
2325 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2326 	}
2327 	sdkp->medium_access_timed_out = 0;
2328 
2329 	if (!scsi_status_is_check_condition(result) &&
2330 	    (!sense_valid || sense_deferred))
2331 		goto out;
2332 
2333 	switch (sshdr.sense_key) {
2334 	case HARDWARE_ERROR:
2335 	case MEDIUM_ERROR:
2336 		good_bytes = sd_completed_bytes(SCpnt);
2337 		break;
2338 	case RECOVERED_ERROR:
2339 		good_bytes = scsi_bufflen(SCpnt);
2340 		break;
2341 	case NO_SENSE:
2342 		/* This indicates a false check condition, so ignore it.  An
2343 		 * unknown amount of data was transferred so treat it as an
2344 		 * error.
2345 		 */
2346 		SCpnt->result = 0;
2347 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2348 		break;
2349 	case ABORTED_COMMAND:
2350 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2351 			good_bytes = sd_completed_bytes(SCpnt);
2352 		break;
2353 	case ILLEGAL_REQUEST:
2354 		switch (sshdr.asc) {
2355 		case 0x10:	/* DIX: Host detected corruption */
2356 			good_bytes = sd_completed_bytes(SCpnt);
2357 			break;
2358 		case 0x20:	/* INVALID COMMAND OPCODE */
2359 		case 0x24:	/* INVALID FIELD IN CDB */
2360 			switch (SCpnt->cmnd[0]) {
2361 			case UNMAP:
2362 				sd_disable_discard(sdkp);
2363 				break;
2364 			case WRITE_SAME_16:
2365 			case WRITE_SAME:
2366 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2367 					sd_disable_discard(sdkp);
2368 				} else {
2369 					sd_disable_write_same(sdkp);
2370 					req->rq_flags |= RQF_QUIET;
2371 				}
2372 				break;
2373 			}
2374 		}
2375 		break;
2376 	default:
2377 		break;
2378 	}
2379 
2380  out:
2381 	if (sdkp->device->type == TYPE_ZBC)
2382 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2383 
2384 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2385 					   "sd_done: completed %d of %d bytes\n",
2386 					   good_bytes, scsi_bufflen(SCpnt)));
2387 
2388 	return good_bytes;
2389 }
2390 
2391 /*
2392  * spinup disk - called only in sd_revalidate_disk()
2393  */
2394 static void
2395 sd_spinup_disk(struct scsi_disk *sdkp)
2396 {
2397 	static const u8 cmd[10] = { TEST_UNIT_READY };
2398 	unsigned long spintime_expire = 0;
2399 	int spintime, sense_valid = 0;
2400 	unsigned int the_result;
2401 	struct scsi_sense_hdr sshdr;
2402 	struct scsi_failure failure_defs[] = {
2403 		/* Do not retry Medium Not Present */
2404 		{
2405 			.sense = UNIT_ATTENTION,
2406 			.asc = 0x3A,
2407 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2408 			.result = SAM_STAT_CHECK_CONDITION,
2409 		},
2410 		{
2411 			.sense = NOT_READY,
2412 			.asc = 0x3A,
2413 			.ascq = SCMD_FAILURE_ASCQ_ANY,
2414 			.result = SAM_STAT_CHECK_CONDITION,
2415 		},
2416 		/* Retry when scsi_status_is_good would return false 3 times */
2417 		{
2418 			.result = SCMD_FAILURE_STAT_ANY,
2419 			.allowed = 3,
2420 		},
2421 		{}
2422 	};
2423 	struct scsi_failures failures = {
2424 		.failure_definitions = failure_defs,
2425 	};
2426 	const struct scsi_exec_args exec_args = {
2427 		.sshdr = &sshdr,
2428 		.failures = &failures,
2429 	};
2430 
2431 	spintime = 0;
2432 
2433 	/* Spin up drives, as required.  Only do this at boot time */
2434 	/* Spinup needs to be done for module loads too. */
2435 	do {
2436 		bool media_was_present = sdkp->media_present;
2437 
2438 		scsi_failures_reset_retries(&failures);
2439 
2440 		the_result = scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN,
2441 					      NULL, 0, SD_TIMEOUT,
2442 					      sdkp->max_retries, &exec_args);
2443 
2444 
2445 		if (the_result > 0) {
2446 			/*
2447 			 * If the drive has indicated to us that it doesn't
2448 			 * have any media in it, don't bother with any more
2449 			 * polling.
2450 			 */
2451 			if (media_not_present(sdkp, &sshdr)) {
2452 				if (media_was_present)
2453 					sd_printk(KERN_NOTICE, sdkp,
2454 						  "Media removed, stopped polling\n");
2455 				return;
2456 			}
2457 			sense_valid = scsi_sense_valid(&sshdr);
2458 		}
2459 
2460 		if (!scsi_status_is_check_condition(the_result)) {
2461 			/* no sense, TUR either succeeded or failed
2462 			 * with a status error */
2463 			if(!spintime && !scsi_status_is_good(the_result)) {
2464 				sd_print_result(sdkp, "Test Unit Ready failed",
2465 						the_result);
2466 			}
2467 			break;
2468 		}
2469 
2470 		/*
2471 		 * The device does not want the automatic start to be issued.
2472 		 */
2473 		if (sdkp->device->no_start_on_add)
2474 			break;
2475 
2476 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2477 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2478 				break;	/* manual intervention required */
2479 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2480 				break;	/* standby */
2481 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2482 				break;	/* unavailable */
2483 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2484 				break;	/* sanitize in progress */
2485 			if (sshdr.asc == 4 && sshdr.ascq == 0x24)
2486 				break;	/* depopulation in progress */
2487 			if (sshdr.asc == 4 && sshdr.ascq == 0x25)
2488 				break;	/* depopulation restoration in progress */
2489 			/*
2490 			 * Issue command to spin up drive when not ready
2491 			 */
2492 			if (!spintime) {
2493 				/* Return immediately and start spin cycle */
2494 				const u8 start_cmd[10] = {
2495 					[0] = START_STOP,
2496 					[1] = 1,
2497 					[4] = sdkp->device->start_stop_pwr_cond ?
2498 						0x11 : 1,
2499 				};
2500 
2501 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2502 				scsi_execute_cmd(sdkp->device, start_cmd,
2503 						 REQ_OP_DRV_IN, NULL, 0,
2504 						 SD_TIMEOUT, sdkp->max_retries,
2505 						 &exec_args);
2506 				spintime_expire = jiffies + 100 * HZ;
2507 				spintime = 1;
2508 			}
2509 			/* Wait 1 second for next try */
2510 			msleep(1000);
2511 			printk(KERN_CONT ".");
2512 
2513 		/*
2514 		 * Wait for USB flash devices with slow firmware.
2515 		 * Yes, this sense key/ASC combination shouldn't
2516 		 * occur here.  It's characteristic of these devices.
2517 		 */
2518 		} else if (sense_valid &&
2519 				sshdr.sense_key == UNIT_ATTENTION &&
2520 				sshdr.asc == 0x28) {
2521 			if (!spintime) {
2522 				spintime_expire = jiffies + 5 * HZ;
2523 				spintime = 1;
2524 			}
2525 			/* Wait 1 second for next try */
2526 			msleep(1000);
2527 		} else {
2528 			/* we don't understand the sense code, so it's
2529 			 * probably pointless to loop */
2530 			if(!spintime) {
2531 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2532 				sd_print_sense_hdr(sdkp, &sshdr);
2533 			}
2534 			break;
2535 		}
2536 
2537 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2538 
2539 	if (spintime) {
2540 		if (scsi_status_is_good(the_result))
2541 			printk(KERN_CONT "ready\n");
2542 		else
2543 			printk(KERN_CONT "not responding...\n");
2544 	}
2545 }
2546 
2547 /*
2548  * Determine whether disk supports Data Integrity Field.
2549  */
2550 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2551 {
2552 	struct scsi_device *sdp = sdkp->device;
2553 	u8 type;
2554 
2555 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2556 		sdkp->protection_type = 0;
2557 		return 0;
2558 	}
2559 
2560 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2561 
2562 	if (type > T10_PI_TYPE3_PROTECTION) {
2563 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2564 			  " protection type %u. Disabling disk!\n",
2565 			  type);
2566 		sdkp->protection_type = 0;
2567 		return -ENODEV;
2568 	}
2569 
2570 	sdkp->protection_type = type;
2571 
2572 	return 0;
2573 }
2574 
2575 static void sd_config_protection(struct scsi_disk *sdkp,
2576 		struct queue_limits *lim)
2577 {
2578 	struct scsi_device *sdp = sdkp->device;
2579 
2580 	if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY))
2581 		sd_dif_config_host(sdkp, lim);
2582 
2583 	if (!sdkp->protection_type)
2584 		return;
2585 
2586 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2587 		sd_first_printk(KERN_NOTICE, sdkp,
2588 				"Disabling DIF Type %u protection\n",
2589 				sdkp->protection_type);
2590 		sdkp->protection_type = 0;
2591 	}
2592 
2593 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2594 			sdkp->protection_type);
2595 }
2596 
2597 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2598 			struct scsi_sense_hdr *sshdr, int sense_valid,
2599 			int the_result)
2600 {
2601 	if (sense_valid)
2602 		sd_print_sense_hdr(sdkp, sshdr);
2603 	else
2604 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2605 
2606 	/*
2607 	 * Set dirty bit for removable devices if not ready -
2608 	 * sometimes drives will not report this properly.
2609 	 */
2610 	if (sdp->removable &&
2611 	    sense_valid && sshdr->sense_key == NOT_READY)
2612 		set_media_not_present(sdkp);
2613 
2614 	/*
2615 	 * We used to set media_present to 0 here to indicate no media
2616 	 * in the drive, but some drives fail read capacity even with
2617 	 * media present, so we can't do that.
2618 	 */
2619 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2620 }
2621 
2622 #define RC16_LEN 32
2623 #if RC16_LEN > SD_BUF_SIZE
2624 #error RC16_LEN must not be more than SD_BUF_SIZE
2625 #endif
2626 
2627 #define READ_CAPACITY_RETRIES_ON_RESET	10
2628 
2629 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2630 		struct queue_limits *lim, unsigned char *buffer)
2631 {
2632 	unsigned char cmd[16];
2633 	struct scsi_sense_hdr sshdr;
2634 	const struct scsi_exec_args exec_args = {
2635 		.sshdr = &sshdr,
2636 	};
2637 	int sense_valid = 0;
2638 	int the_result;
2639 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2640 	unsigned int alignment;
2641 	unsigned long long lba;
2642 	unsigned sector_size;
2643 
2644 	if (sdp->no_read_capacity_16)
2645 		return -EINVAL;
2646 
2647 	do {
2648 		memset(cmd, 0, 16);
2649 		cmd[0] = SERVICE_ACTION_IN_16;
2650 		cmd[1] = SAI_READ_CAPACITY_16;
2651 		cmd[13] = RC16_LEN;
2652 		memset(buffer, 0, RC16_LEN);
2653 
2654 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2655 					      buffer, RC16_LEN, SD_TIMEOUT,
2656 					      sdkp->max_retries, &exec_args);
2657 		if (the_result > 0) {
2658 			if (media_not_present(sdkp, &sshdr))
2659 				return -ENODEV;
2660 
2661 			sense_valid = scsi_sense_valid(&sshdr);
2662 			if (sense_valid &&
2663 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2664 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2665 			    sshdr.ascq == 0x00)
2666 				/* Invalid Command Operation Code or
2667 				 * Invalid Field in CDB, just retry
2668 				 * silently with RC10 */
2669 				return -EINVAL;
2670 			if (sense_valid &&
2671 			    sshdr.sense_key == UNIT_ATTENTION &&
2672 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2673 				/* Device reset might occur several times,
2674 				 * give it one more chance */
2675 				if (--reset_retries > 0)
2676 					continue;
2677 		}
2678 		retries--;
2679 
2680 	} while (the_result && retries);
2681 
2682 	if (the_result) {
2683 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2684 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2685 		return -EINVAL;
2686 	}
2687 
2688 	sector_size = get_unaligned_be32(&buffer[8]);
2689 	lba = get_unaligned_be64(&buffer[0]);
2690 
2691 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2692 		sdkp->capacity = 0;
2693 		return -ENODEV;
2694 	}
2695 
2696 	/* Logical blocks per physical block exponent */
2697 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2698 
2699 	/* RC basis */
2700 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2701 
2702 	/* Lowest aligned logical block */
2703 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2704 	lim->alignment_offset = alignment;
2705 	if (alignment && sdkp->first_scan)
2706 		sd_printk(KERN_NOTICE, sdkp,
2707 			  "physical block alignment offset: %u\n", alignment);
2708 
2709 	if (buffer[14] & 0x80) { /* LBPME */
2710 		sdkp->lbpme = 1;
2711 
2712 		if (buffer[14] & 0x40) /* LBPRZ */
2713 			sdkp->lbprz = 1;
2714 
2715 		sd_config_discard(sdkp, lim, SD_LBP_WS16);
2716 	}
2717 
2718 	sdkp->capacity = lba + 1;
2719 	return sector_size;
2720 }
2721 
2722 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2723 						unsigned char *buffer)
2724 {
2725 	static const u8 cmd[10] = { READ_CAPACITY };
2726 	struct scsi_sense_hdr sshdr;
2727 	struct scsi_failure failure_defs[] = {
2728 		/* Do not retry Medium Not Present */
2729 		{
2730 			.sense = UNIT_ATTENTION,
2731 			.asc = 0x3A,
2732 			.result = SAM_STAT_CHECK_CONDITION,
2733 		},
2734 		{
2735 			.sense = NOT_READY,
2736 			.asc = 0x3A,
2737 			.result = SAM_STAT_CHECK_CONDITION,
2738 		},
2739 		 /* Device reset might occur several times so retry a lot */
2740 		{
2741 			.sense = UNIT_ATTENTION,
2742 			.asc = 0x29,
2743 			.allowed = READ_CAPACITY_RETRIES_ON_RESET,
2744 			.result = SAM_STAT_CHECK_CONDITION,
2745 		},
2746 		/* Any other error not listed above retry 3 times */
2747 		{
2748 			.result = SCMD_FAILURE_RESULT_ANY,
2749 			.allowed = 3,
2750 		},
2751 		{}
2752 	};
2753 	struct scsi_failures failures = {
2754 		.failure_definitions = failure_defs,
2755 	};
2756 	const struct scsi_exec_args exec_args = {
2757 		.sshdr = &sshdr,
2758 		.failures = &failures,
2759 	};
2760 	int sense_valid = 0;
2761 	int the_result;
2762 	sector_t lba;
2763 	unsigned sector_size;
2764 
2765 	memset(buffer, 0, 8);
2766 
2767 	the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2768 				      8, SD_TIMEOUT, sdkp->max_retries,
2769 				      &exec_args);
2770 
2771 	if (the_result > 0) {
2772 		sense_valid = scsi_sense_valid(&sshdr);
2773 
2774 		if (media_not_present(sdkp, &sshdr))
2775 			return -ENODEV;
2776 	}
2777 
2778 	if (the_result) {
2779 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2780 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2781 		return -EINVAL;
2782 	}
2783 
2784 	sector_size = get_unaligned_be32(&buffer[4]);
2785 	lba = get_unaligned_be32(&buffer[0]);
2786 
2787 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2788 		/* Some buggy (usb cardreader) devices return an lba of
2789 		   0xffffffff when the want to report a size of 0 (with
2790 		   which they really mean no media is present) */
2791 		sdkp->capacity = 0;
2792 		sdkp->physical_block_size = sector_size;
2793 		return sector_size;
2794 	}
2795 
2796 	sdkp->capacity = lba + 1;
2797 	sdkp->physical_block_size = sector_size;
2798 	return sector_size;
2799 }
2800 
2801 static int sd_try_rc16_first(struct scsi_device *sdp)
2802 {
2803 	if (sdp->host->max_cmd_len < 16)
2804 		return 0;
2805 	if (sdp->try_rc_10_first)
2806 		return 0;
2807 	if (sdp->scsi_level > SCSI_SPC_2)
2808 		return 1;
2809 	if (scsi_device_protection(sdp))
2810 		return 1;
2811 	return 0;
2812 }
2813 
2814 /*
2815  * read disk capacity
2816  */
2817 static void
2818 sd_read_capacity(struct scsi_disk *sdkp, struct queue_limits *lim,
2819 		unsigned char *buffer)
2820 {
2821 	int sector_size;
2822 	struct scsi_device *sdp = sdkp->device;
2823 
2824 	if (sd_try_rc16_first(sdp)) {
2825 		sector_size = read_capacity_16(sdkp, sdp, lim, buffer);
2826 		if (sector_size == -EOVERFLOW)
2827 			goto got_data;
2828 		if (sector_size == -ENODEV)
2829 			return;
2830 		if (sector_size < 0)
2831 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2832 		if (sector_size < 0)
2833 			return;
2834 	} else {
2835 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2836 		if (sector_size == -EOVERFLOW)
2837 			goto got_data;
2838 		if (sector_size < 0)
2839 			return;
2840 		if ((sizeof(sdkp->capacity) > 4) &&
2841 		    (sdkp->capacity > 0xffffffffULL)) {
2842 			int old_sector_size = sector_size;
2843 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2844 					"Trying to use READ CAPACITY(16).\n");
2845 			sector_size = read_capacity_16(sdkp, sdp, lim, buffer);
2846 			if (sector_size < 0) {
2847 				sd_printk(KERN_NOTICE, sdkp,
2848 					"Using 0xffffffff as device size\n");
2849 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2850 				sector_size = old_sector_size;
2851 				goto got_data;
2852 			}
2853 			/* Remember that READ CAPACITY(16) succeeded */
2854 			sdp->try_rc_10_first = 0;
2855 		}
2856 	}
2857 
2858 	/* Some devices are known to return the total number of blocks,
2859 	 * not the highest block number.  Some devices have versions
2860 	 * which do this and others which do not.  Some devices we might
2861 	 * suspect of doing this but we don't know for certain.
2862 	 *
2863 	 * If we know the reported capacity is wrong, decrement it.  If
2864 	 * we can only guess, then assume the number of blocks is even
2865 	 * (usually true but not always) and err on the side of lowering
2866 	 * the capacity.
2867 	 */
2868 	if (sdp->fix_capacity ||
2869 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2870 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2871 				"from its reported value: %llu\n",
2872 				(unsigned long long) sdkp->capacity);
2873 		--sdkp->capacity;
2874 	}
2875 
2876 got_data:
2877 	if (sector_size == 0) {
2878 		sector_size = 512;
2879 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2880 			  "assuming 512.\n");
2881 	}
2882 
2883 	if (sector_size != 512 &&
2884 	    sector_size != 1024 &&
2885 	    sector_size != 2048 &&
2886 	    sector_size != 4096) {
2887 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2888 			  sector_size);
2889 		/*
2890 		 * The user might want to re-format the drive with
2891 		 * a supported sectorsize.  Once this happens, it
2892 		 * would be relatively trivial to set the thing up.
2893 		 * For this reason, we leave the thing in the table.
2894 		 */
2895 		sdkp->capacity = 0;
2896 		/*
2897 		 * set a bogus sector size so the normal read/write
2898 		 * logic in the block layer will eventually refuse any
2899 		 * request on this device without tripping over power
2900 		 * of two sector size assumptions
2901 		 */
2902 		sector_size = 512;
2903 	}
2904 	lim->logical_block_size = sector_size;
2905 	lim->physical_block_size = sdkp->physical_block_size;
2906 	sdkp->device->sector_size = sector_size;
2907 
2908 	if (sdkp->capacity > 0xffffffff)
2909 		sdp->use_16_for_rw = 1;
2910 
2911 }
2912 
2913 /*
2914  * Print disk capacity
2915  */
2916 static void
2917 sd_print_capacity(struct scsi_disk *sdkp,
2918 		  sector_t old_capacity)
2919 {
2920 	int sector_size = sdkp->device->sector_size;
2921 	char cap_str_2[10], cap_str_10[10];
2922 
2923 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2924 		return;
2925 
2926 	string_get_size(sdkp->capacity, sector_size,
2927 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2928 	string_get_size(sdkp->capacity, sector_size,
2929 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2930 
2931 	sd_printk(KERN_NOTICE, sdkp,
2932 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2933 		  (unsigned long long)sdkp->capacity,
2934 		  sector_size, cap_str_10, cap_str_2);
2935 
2936 	if (sdkp->physical_block_size != sector_size)
2937 		sd_printk(KERN_NOTICE, sdkp,
2938 			  "%u-byte physical blocks\n",
2939 			  sdkp->physical_block_size);
2940 }
2941 
2942 /* called with buffer of length 512 */
2943 static inline int
2944 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2945 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2946 		 struct scsi_sense_hdr *sshdr)
2947 {
2948 	/*
2949 	 * If we must use MODE SENSE(10), make sure that the buffer length
2950 	 * is at least 8 bytes so that the mode sense header fits.
2951 	 */
2952 	if (sdkp->device->use_10_for_ms && len < 8)
2953 		len = 8;
2954 
2955 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2956 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2957 }
2958 
2959 /*
2960  * read write protect setting, if possible - called only in sd_revalidate_disk()
2961  * called with buffer of length SD_BUF_SIZE
2962  */
2963 static void
2964 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2965 {
2966 	int res;
2967 	struct scsi_device *sdp = sdkp->device;
2968 	struct scsi_mode_data data;
2969 	int old_wp = sdkp->write_prot;
2970 
2971 	set_disk_ro(sdkp->disk, 0);
2972 	if (sdp->skip_ms_page_3f) {
2973 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2974 		return;
2975 	}
2976 
2977 	if (sdp->use_192_bytes_for_3f) {
2978 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2979 	} else {
2980 		/*
2981 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2982 		 * We have to start carefully: some devices hang if we ask
2983 		 * for more than is available.
2984 		 */
2985 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2986 
2987 		/*
2988 		 * Second attempt: ask for page 0 When only page 0 is
2989 		 * implemented, a request for page 3F may return Sense Key
2990 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2991 		 * CDB.
2992 		 */
2993 		if (res < 0)
2994 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2995 
2996 		/*
2997 		 * Third attempt: ask 255 bytes, as we did earlier.
2998 		 */
2999 		if (res < 0)
3000 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
3001 					       &data, NULL);
3002 	}
3003 
3004 	if (res < 0) {
3005 		sd_first_printk(KERN_WARNING, sdkp,
3006 			  "Test WP failed, assume Write Enabled\n");
3007 	} else {
3008 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
3009 		set_disk_ro(sdkp->disk, sdkp->write_prot);
3010 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
3011 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
3012 				  sdkp->write_prot ? "on" : "off");
3013 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
3014 		}
3015 	}
3016 }
3017 
3018 /*
3019  * sd_read_cache_type - called only from sd_revalidate_disk()
3020  * called with buffer of length SD_BUF_SIZE
3021  */
3022 static void
3023 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
3024 {
3025 	int len = 0, res;
3026 	struct scsi_device *sdp = sdkp->device;
3027 
3028 	int dbd;
3029 	int modepage;
3030 	int first_len;
3031 	struct scsi_mode_data data;
3032 	struct scsi_sense_hdr sshdr;
3033 	int old_wce = sdkp->WCE;
3034 	int old_rcd = sdkp->RCD;
3035 	int old_dpofua = sdkp->DPOFUA;
3036 
3037 
3038 	if (sdkp->cache_override)
3039 		return;
3040 
3041 	first_len = 4;
3042 	if (sdp->skip_ms_page_8) {
3043 		if (sdp->type == TYPE_RBC)
3044 			goto defaults;
3045 		else {
3046 			if (sdp->skip_ms_page_3f)
3047 				goto defaults;
3048 			modepage = 0x3F;
3049 			if (sdp->use_192_bytes_for_3f)
3050 				first_len = 192;
3051 			dbd = 0;
3052 		}
3053 	} else if (sdp->type == TYPE_RBC) {
3054 		modepage = 6;
3055 		dbd = 8;
3056 	} else {
3057 		modepage = 8;
3058 		dbd = 0;
3059 	}
3060 
3061 	/* cautiously ask */
3062 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
3063 			&data, &sshdr);
3064 
3065 	if (res < 0)
3066 		goto bad_sense;
3067 
3068 	if (!data.header_length) {
3069 		modepage = 6;
3070 		first_len = 0;
3071 		sd_first_printk(KERN_ERR, sdkp,
3072 				"Missing header in MODE_SENSE response\n");
3073 	}
3074 
3075 	/* that went OK, now ask for the proper length */
3076 	len = data.length;
3077 
3078 	/*
3079 	 * We're only interested in the first three bytes, actually.
3080 	 * But the data cache page is defined for the first 20.
3081 	 */
3082 	if (len < 3)
3083 		goto bad_sense;
3084 	else if (len > SD_BUF_SIZE) {
3085 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
3086 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
3087 		len = SD_BUF_SIZE;
3088 	}
3089 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
3090 		len = 192;
3091 
3092 	/* Get the data */
3093 	if (len > first_len)
3094 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
3095 				&data, &sshdr);
3096 
3097 	if (!res) {
3098 		int offset = data.header_length + data.block_descriptor_length;
3099 
3100 		while (offset < len) {
3101 			u8 page_code = buffer[offset] & 0x3F;
3102 			u8 spf       = buffer[offset] & 0x40;
3103 
3104 			if (page_code == 8 || page_code == 6) {
3105 				/* We're interested only in the first 3 bytes.
3106 				 */
3107 				if (len - offset <= 2) {
3108 					sd_first_printk(KERN_ERR, sdkp,
3109 						"Incomplete mode parameter "
3110 							"data\n");
3111 					goto defaults;
3112 				} else {
3113 					modepage = page_code;
3114 					goto Page_found;
3115 				}
3116 			} else {
3117 				/* Go to the next page */
3118 				if (spf && len - offset > 3)
3119 					offset += 4 + (buffer[offset+2] << 8) +
3120 						buffer[offset+3];
3121 				else if (!spf && len - offset > 1)
3122 					offset += 2 + buffer[offset+1];
3123 				else {
3124 					sd_first_printk(KERN_ERR, sdkp,
3125 							"Incomplete mode "
3126 							"parameter data\n");
3127 					goto defaults;
3128 				}
3129 			}
3130 		}
3131 
3132 		sd_first_printk(KERN_WARNING, sdkp,
3133 				"No Caching mode page found\n");
3134 		goto defaults;
3135 
3136 	Page_found:
3137 		if (modepage == 8) {
3138 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
3139 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
3140 		} else {
3141 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
3142 			sdkp->RCD = 0;
3143 		}
3144 
3145 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
3146 		if (sdp->broken_fua) {
3147 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
3148 			sdkp->DPOFUA = 0;
3149 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
3150 			   !sdkp->device->use_16_for_rw) {
3151 			sd_first_printk(KERN_NOTICE, sdkp,
3152 				  "Uses READ/WRITE(6), disabling FUA\n");
3153 			sdkp->DPOFUA = 0;
3154 		}
3155 
3156 		/* No cache flush allowed for write protected devices */
3157 		if (sdkp->WCE && sdkp->write_prot)
3158 			sdkp->WCE = 0;
3159 
3160 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
3161 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
3162 			sd_printk(KERN_NOTICE, sdkp,
3163 				  "Write cache: %s, read cache: %s, %s\n",
3164 				  sdkp->WCE ? "enabled" : "disabled",
3165 				  sdkp->RCD ? "disabled" : "enabled",
3166 				  sdkp->DPOFUA ? "supports DPO and FUA"
3167 				  : "doesn't support DPO or FUA");
3168 
3169 		return;
3170 	}
3171 
3172 bad_sense:
3173 	if (res == -EIO && scsi_sense_valid(&sshdr) &&
3174 	    sshdr.sense_key == ILLEGAL_REQUEST &&
3175 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
3176 		/* Invalid field in CDB */
3177 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
3178 	else
3179 		sd_first_printk(KERN_ERR, sdkp,
3180 				"Asking for cache data failed\n");
3181 
3182 defaults:
3183 	if (sdp->wce_default_on) {
3184 		sd_first_printk(KERN_NOTICE, sdkp,
3185 				"Assuming drive cache: write back\n");
3186 		sdkp->WCE = 1;
3187 	} else {
3188 		sd_first_printk(KERN_WARNING, sdkp,
3189 				"Assuming drive cache: write through\n");
3190 		sdkp->WCE = 0;
3191 	}
3192 	sdkp->RCD = 0;
3193 	sdkp->DPOFUA = 0;
3194 }
3195 
3196 static bool sd_is_perm_stream(struct scsi_disk *sdkp, unsigned int stream_id)
3197 {
3198 	u8 cdb[16] = { SERVICE_ACTION_IN_16, SAI_GET_STREAM_STATUS };
3199 	struct {
3200 		struct scsi_stream_status_header h;
3201 		struct scsi_stream_status s;
3202 	} buf;
3203 	struct scsi_device *sdev = sdkp->device;
3204 	struct scsi_sense_hdr sshdr;
3205 	const struct scsi_exec_args exec_args = {
3206 		.sshdr = &sshdr,
3207 	};
3208 	int res;
3209 
3210 	put_unaligned_be16(stream_id, &cdb[4]);
3211 	put_unaligned_be32(sizeof(buf), &cdb[10]);
3212 
3213 	res = scsi_execute_cmd(sdev, cdb, REQ_OP_DRV_IN, &buf, sizeof(buf),
3214 			       SD_TIMEOUT, sdkp->max_retries, &exec_args);
3215 	if (res < 0)
3216 		return false;
3217 	if (scsi_status_is_check_condition(res) && scsi_sense_valid(&sshdr))
3218 		sd_print_sense_hdr(sdkp, &sshdr);
3219 	if (res)
3220 		return false;
3221 	if (get_unaligned_be32(&buf.h.len) < sizeof(struct scsi_stream_status))
3222 		return false;
3223 	return buf.h.stream_status[0].perm;
3224 }
3225 
3226 static void sd_read_io_hints(struct scsi_disk *sdkp, unsigned char *buffer)
3227 {
3228 	struct scsi_device *sdp = sdkp->device;
3229 	const struct scsi_io_group_descriptor *desc, *start, *end;
3230 	u16 permanent_stream_count_old;
3231 	struct scsi_sense_hdr sshdr;
3232 	struct scsi_mode_data data;
3233 	int res;
3234 
3235 	if (sdp->sdev_bflags & BLIST_SKIP_IO_HINTS)
3236 		return;
3237 
3238 	res = scsi_mode_sense(sdp, /*dbd=*/0x8, /*modepage=*/0x0a,
3239 			      /*subpage=*/0x05, buffer, SD_BUF_SIZE, SD_TIMEOUT,
3240 			      sdkp->max_retries, &data, &sshdr);
3241 	if (res < 0)
3242 		return;
3243 	start = (void *)buffer + data.header_length + 16;
3244 	end = (void *)buffer + ALIGN_DOWN(data.header_length + data.length,
3245 					  sizeof(*end));
3246 	/*
3247 	 * From "SBC-5 Constrained Streams with Data Lifetimes": Device severs
3248 	 * should assign the lowest numbered stream identifiers to permanent
3249 	 * streams.
3250 	 */
3251 	for (desc = start; desc < end; desc++)
3252 		if (!desc->st_enble || !sd_is_perm_stream(sdkp, desc - start))
3253 			break;
3254 	permanent_stream_count_old = sdkp->permanent_stream_count;
3255 	sdkp->permanent_stream_count = desc - start;
3256 	if (sdkp->rscs && sdkp->permanent_stream_count < 2)
3257 		sd_printk(KERN_INFO, sdkp,
3258 			  "Unexpected: RSCS has been set and the permanent stream count is %u\n",
3259 			  sdkp->permanent_stream_count);
3260 	else if (sdkp->permanent_stream_count != permanent_stream_count_old)
3261 		sd_printk(KERN_INFO, sdkp, "permanent stream count = %d\n",
3262 			  sdkp->permanent_stream_count);
3263 }
3264 
3265 /*
3266  * The ATO bit indicates whether the DIF application tag is available
3267  * for use by the operating system.
3268  */
3269 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
3270 {
3271 	int res, offset;
3272 	struct scsi_device *sdp = sdkp->device;
3273 	struct scsi_mode_data data;
3274 	struct scsi_sense_hdr sshdr;
3275 
3276 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
3277 		return;
3278 
3279 	if (sdkp->protection_type == 0)
3280 		return;
3281 
3282 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
3283 			      sdkp->max_retries, &data, &sshdr);
3284 
3285 	if (res < 0 || !data.header_length ||
3286 	    data.length < 6) {
3287 		sd_first_printk(KERN_WARNING, sdkp,
3288 			  "getting Control mode page failed, assume no ATO\n");
3289 
3290 		if (res == -EIO && scsi_sense_valid(&sshdr))
3291 			sd_print_sense_hdr(sdkp, &sshdr);
3292 
3293 		return;
3294 	}
3295 
3296 	offset = data.header_length + data.block_descriptor_length;
3297 
3298 	if ((buffer[offset] & 0x3f) != 0x0a) {
3299 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
3300 		return;
3301 	}
3302 
3303 	if ((buffer[offset + 5] & 0x80) == 0)
3304 		return;
3305 
3306 	sdkp->ATO = 1;
3307 
3308 	return;
3309 }
3310 
3311 static unsigned int sd_discard_mode(struct scsi_disk *sdkp)
3312 {
3313 	if (!sdkp->lbpvpd) {
3314 		/* LBP VPD page not provided */
3315 		if (sdkp->max_unmap_blocks)
3316 			return SD_LBP_UNMAP;
3317 		return SD_LBP_WS16;
3318 	}
3319 
3320 	/* LBP VPD page tells us what to use */
3321 	if (sdkp->lbpu && sdkp->max_unmap_blocks)
3322 		return SD_LBP_UNMAP;
3323 	if (sdkp->lbpws)
3324 		return SD_LBP_WS16;
3325 	if (sdkp->lbpws10)
3326 		return SD_LBP_WS10;
3327 	return SD_LBP_DISABLE;
3328 }
3329 
3330 /*
3331  * Query disk device for preferred I/O sizes.
3332  */
3333 static void sd_read_block_limits(struct scsi_disk *sdkp,
3334 		struct queue_limits *lim)
3335 {
3336 	struct scsi_vpd *vpd;
3337 
3338 	rcu_read_lock();
3339 
3340 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3341 	if (!vpd || vpd->len < 16)
3342 		goto out;
3343 
3344 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3345 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3346 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3347 
3348 	if (vpd->len >= 64) {
3349 		unsigned int lba_count, desc_count;
3350 
3351 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3352 
3353 		if (!sdkp->lbpme)
3354 			goto config_atomic;
3355 
3356 		lba_count = get_unaligned_be32(&vpd->data[20]);
3357 		desc_count = get_unaligned_be32(&vpd->data[24]);
3358 
3359 		if (lba_count && desc_count)
3360 			sdkp->max_unmap_blocks = lba_count;
3361 
3362 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3363 
3364 		if (vpd->data[32] & 0x80)
3365 			sdkp->unmap_alignment =
3366 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3367 
3368 		sd_config_discard(sdkp, lim, sd_discard_mode(sdkp));
3369 
3370 config_atomic:
3371 		sdkp->max_atomic = get_unaligned_be32(&vpd->data[44]);
3372 		sdkp->atomic_alignment = get_unaligned_be32(&vpd->data[48]);
3373 		sdkp->atomic_granularity = get_unaligned_be32(&vpd->data[52]);
3374 		sdkp->max_atomic_with_boundary = get_unaligned_be32(&vpd->data[56]);
3375 		sdkp->max_atomic_boundary = get_unaligned_be32(&vpd->data[60]);
3376 
3377 		sd_config_atomic(sdkp, lim);
3378 	}
3379 
3380  out:
3381 	rcu_read_unlock();
3382 }
3383 
3384 /* Parse the Block Limits Extension VPD page (0xb7) */
3385 static void sd_read_block_limits_ext(struct scsi_disk *sdkp)
3386 {
3387 	struct scsi_vpd *vpd;
3388 
3389 	rcu_read_lock();
3390 	vpd = rcu_dereference(sdkp->device->vpd_pgb7);
3391 	if (vpd && vpd->len >= 2)
3392 		sdkp->rscs = vpd->data[5] & 1;
3393 	rcu_read_unlock();
3394 }
3395 
3396 /* Query block device characteristics */
3397 static void sd_read_block_characteristics(struct scsi_disk *sdkp,
3398 		struct queue_limits *lim)
3399 {
3400 	struct scsi_vpd *vpd;
3401 	u16 rot;
3402 
3403 	rcu_read_lock();
3404 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3405 
3406 	if (!vpd || vpd->len < 8) {
3407 		rcu_read_unlock();
3408 	        return;
3409 	}
3410 
3411 	rot = get_unaligned_be16(&vpd->data[4]);
3412 	sdkp->zoned = (vpd->data[8] >> 4) & 3;
3413 	rcu_read_unlock();
3414 
3415 	if (rot == 1)
3416 		lim->features &= ~(BLK_FEAT_ROTATIONAL | BLK_FEAT_ADD_RANDOM);
3417 
3418 	if (!sdkp->first_scan)
3419 		return;
3420 
3421 	if (sdkp->device->type == TYPE_ZBC)
3422 		sd_printk(KERN_NOTICE, sdkp, "Host-managed zoned block device\n");
3423 	else if (sdkp->zoned == 1)
3424 		sd_printk(KERN_NOTICE, sdkp, "Host-aware SMR disk used as regular disk\n");
3425 	else if (sdkp->zoned == 2)
3426 		sd_printk(KERN_NOTICE, sdkp, "Drive-managed SMR disk\n");
3427 }
3428 
3429 /**
3430  * sd_read_block_provisioning - Query provisioning VPD page
3431  * @sdkp: disk to query
3432  */
3433 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3434 {
3435 	struct scsi_vpd *vpd;
3436 
3437 	if (sdkp->lbpme == 0)
3438 		return;
3439 
3440 	rcu_read_lock();
3441 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3442 
3443 	if (!vpd || vpd->len < 8) {
3444 		rcu_read_unlock();
3445 		return;
3446 	}
3447 
3448 	sdkp->lbpvpd	= 1;
3449 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3450 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3451 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3452 	rcu_read_unlock();
3453 }
3454 
3455 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3456 {
3457 	struct scsi_device *sdev = sdkp->device;
3458 
3459 	if (sdev->host->no_write_same) {
3460 		sdev->no_write_same = 1;
3461 
3462 		return;
3463 	}
3464 
3465 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3466 		struct scsi_vpd *vpd;
3467 
3468 		sdev->no_report_opcodes = 1;
3469 
3470 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3471 		 * CODES is unsupported and the device has an ATA
3472 		 * Information VPD page (SAT).
3473 		 */
3474 		rcu_read_lock();
3475 		vpd = rcu_dereference(sdev->vpd_pg89);
3476 		if (vpd)
3477 			sdev->no_write_same = 1;
3478 		rcu_read_unlock();
3479 	}
3480 
3481 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3482 		sdkp->ws16 = 1;
3483 
3484 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3485 		sdkp->ws10 = 1;
3486 }
3487 
3488 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3489 {
3490 	struct scsi_device *sdev = sdkp->device;
3491 
3492 	if (!sdev->security_supported)
3493 		return;
3494 
3495 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3496 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3497 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3498 			SECURITY_PROTOCOL_OUT, 0) == 1)
3499 		sdkp->security = 1;
3500 }
3501 
3502 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3503 {
3504 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3505 }
3506 
3507 /**
3508  * sd_read_cpr - Query concurrent positioning ranges
3509  * @sdkp:	disk to query
3510  */
3511 static void sd_read_cpr(struct scsi_disk *sdkp)
3512 {
3513 	struct blk_independent_access_ranges *iars = NULL;
3514 	unsigned char *buffer = NULL;
3515 	unsigned int nr_cpr = 0;
3516 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3517 	u8 *desc;
3518 
3519 	/*
3520 	 * We need to have the capacity set first for the block layer to be
3521 	 * able to check the ranges.
3522 	 */
3523 	if (sdkp->first_scan)
3524 		return;
3525 
3526 	if (!sdkp->capacity)
3527 		goto out;
3528 
3529 	/*
3530 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3531 	 * leading to a maximum page size of 64 + 256*32 bytes.
3532 	 */
3533 	buf_len = 64 + 256*32;
3534 	buffer = kmalloc(buf_len, GFP_KERNEL);
3535 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3536 		goto out;
3537 
3538 	/* We must have at least a 64B header and one 32B range descriptor */
3539 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3540 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3541 		sd_printk(KERN_ERR, sdkp,
3542 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3543 		goto out;
3544 	}
3545 
3546 	nr_cpr = (vpd_len - 64) / 32;
3547 	if (nr_cpr == 1) {
3548 		nr_cpr = 0;
3549 		goto out;
3550 	}
3551 
3552 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3553 	if (!iars) {
3554 		nr_cpr = 0;
3555 		goto out;
3556 	}
3557 
3558 	desc = &buffer[64];
3559 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3560 		if (desc[0] != i) {
3561 			sd_printk(KERN_ERR, sdkp,
3562 				"Invalid Concurrent Positioning Range number\n");
3563 			nr_cpr = 0;
3564 			break;
3565 		}
3566 
3567 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3568 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3569 	}
3570 
3571 out:
3572 	disk_set_independent_access_ranges(sdkp->disk, iars);
3573 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3574 		sd_printk(KERN_NOTICE, sdkp,
3575 			  "%u concurrent positioning ranges\n", nr_cpr);
3576 		sdkp->nr_actuators = nr_cpr;
3577 	}
3578 
3579 	kfree(buffer);
3580 }
3581 
3582 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3583 {
3584 	struct scsi_device *sdp = sdkp->device;
3585 	unsigned int min_xfer_bytes =
3586 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3587 
3588 	if (sdkp->min_xfer_blocks == 0)
3589 		return false;
3590 
3591 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3592 		sd_first_printk(KERN_WARNING, sdkp,
3593 				"Preferred minimum I/O size %u bytes not a " \
3594 				"multiple of physical block size (%u bytes)\n",
3595 				min_xfer_bytes, sdkp->physical_block_size);
3596 		sdkp->min_xfer_blocks = 0;
3597 		return false;
3598 	}
3599 
3600 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3601 			min_xfer_bytes);
3602 	return true;
3603 }
3604 
3605 /*
3606  * Determine the device's preferred I/O size for reads and writes
3607  * unless the reported value is unreasonably small, large, not a
3608  * multiple of the physical block size, or simply garbage.
3609  */
3610 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3611 				      unsigned int dev_max)
3612 {
3613 	struct scsi_device *sdp = sdkp->device;
3614 	unsigned int opt_xfer_bytes =
3615 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3616 	unsigned int min_xfer_bytes =
3617 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3618 
3619 	if (sdkp->opt_xfer_blocks == 0)
3620 		return false;
3621 
3622 	if (sdkp->opt_xfer_blocks > dev_max) {
3623 		sd_first_printk(KERN_WARNING, sdkp,
3624 				"Optimal transfer size %u logical blocks " \
3625 				"> dev_max (%u logical blocks)\n",
3626 				sdkp->opt_xfer_blocks, dev_max);
3627 		return false;
3628 	}
3629 
3630 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3631 		sd_first_printk(KERN_WARNING, sdkp,
3632 				"Optimal transfer size %u logical blocks " \
3633 				"> sd driver limit (%u logical blocks)\n",
3634 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3635 		return false;
3636 	}
3637 
3638 	if (opt_xfer_bytes < PAGE_SIZE) {
3639 		sd_first_printk(KERN_WARNING, sdkp,
3640 				"Optimal transfer size %u bytes < " \
3641 				"PAGE_SIZE (%u bytes)\n",
3642 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3643 		return false;
3644 	}
3645 
3646 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3647 		sd_first_printk(KERN_WARNING, sdkp,
3648 				"Optimal transfer size %u bytes not a " \
3649 				"multiple of preferred minimum block " \
3650 				"size (%u bytes)\n",
3651 				opt_xfer_bytes, min_xfer_bytes);
3652 		return false;
3653 	}
3654 
3655 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3656 		sd_first_printk(KERN_WARNING, sdkp,
3657 				"Optimal transfer size %u bytes not a " \
3658 				"multiple of physical block size (%u bytes)\n",
3659 				opt_xfer_bytes, sdkp->physical_block_size);
3660 		return false;
3661 	}
3662 
3663 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3664 			opt_xfer_bytes);
3665 	return true;
3666 }
3667 
3668 static void sd_read_block_zero(struct scsi_disk *sdkp)
3669 {
3670 	struct scsi_device *sdev = sdkp->device;
3671 	unsigned int buf_len = sdev->sector_size;
3672 	u8 *buffer, cmd[16] = { };
3673 
3674 	buffer = kmalloc(buf_len, GFP_KERNEL);
3675 	if (!buffer)
3676 		return;
3677 
3678 	if (sdev->use_16_for_rw) {
3679 		cmd[0] = READ_16;
3680 		put_unaligned_be64(0, &cmd[2]); /* Logical block address 0 */
3681 		put_unaligned_be32(1, &cmd[10]);/* Transfer 1 logical block */
3682 	} else {
3683 		cmd[0] = READ_10;
3684 		put_unaligned_be32(0, &cmd[2]); /* Logical block address 0 */
3685 		put_unaligned_be16(1, &cmd[7]);	/* Transfer 1 logical block */
3686 	}
3687 
3688 	scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN, buffer, buf_len,
3689 			 SD_TIMEOUT, sdkp->max_retries, NULL);
3690 	kfree(buffer);
3691 }
3692 
3693 /**
3694  *	sd_revalidate_disk - called the first time a new disk is seen,
3695  *	performs disk spin up, read_capacity, etc.
3696  *	@disk: struct gendisk we care about
3697  **/
3698 static int sd_revalidate_disk(struct gendisk *disk)
3699 {
3700 	struct scsi_disk *sdkp = scsi_disk(disk);
3701 	struct scsi_device *sdp = sdkp->device;
3702 	sector_t old_capacity = sdkp->capacity;
3703 	struct queue_limits lim;
3704 	unsigned char *buffer;
3705 	unsigned int dev_max;
3706 	int err;
3707 
3708 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3709 				      "sd_revalidate_disk\n"));
3710 
3711 	/*
3712 	 * If the device is offline, don't try and read capacity or any
3713 	 * of the other niceties.
3714 	 */
3715 	if (!scsi_device_online(sdp))
3716 		goto out;
3717 
3718 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3719 	if (!buffer) {
3720 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3721 			  "allocation failure.\n");
3722 		goto out;
3723 	}
3724 
3725 	sd_spinup_disk(sdkp);
3726 
3727 	lim = queue_limits_start_update(sdkp->disk->queue);
3728 
3729 	/*
3730 	 * Without media there is no reason to ask; moreover, some devices
3731 	 * react badly if we do.
3732 	 */
3733 	if (sdkp->media_present) {
3734 		sd_read_capacity(sdkp, &lim, buffer);
3735 		/*
3736 		 * Some USB/UAS devices return generic values for mode pages
3737 		 * until the media has been accessed. Trigger a READ operation
3738 		 * to force the device to populate mode pages.
3739 		 */
3740 		if (sdp->read_before_ms)
3741 			sd_read_block_zero(sdkp);
3742 		/*
3743 		 * set the default to rotational.  All non-rotational devices
3744 		 * support the block characteristics VPD page, which will
3745 		 * cause this to be updated correctly and any device which
3746 		 * doesn't support it should be treated as rotational.
3747 		 */
3748 		lim.features |= (BLK_FEAT_ROTATIONAL | BLK_FEAT_ADD_RANDOM);
3749 
3750 		if (scsi_device_supports_vpd(sdp)) {
3751 			sd_read_block_provisioning(sdkp);
3752 			sd_read_block_limits(sdkp, &lim);
3753 			sd_read_block_limits_ext(sdkp);
3754 			sd_read_block_characteristics(sdkp, &lim);
3755 			sd_zbc_read_zones(sdkp, &lim, buffer);
3756 			sd_read_cpr(sdkp);
3757 		}
3758 
3759 		sd_print_capacity(sdkp, old_capacity);
3760 
3761 		sd_read_write_protect_flag(sdkp, buffer);
3762 		sd_read_cache_type(sdkp, buffer);
3763 		sd_read_io_hints(sdkp, buffer);
3764 		sd_read_app_tag_own(sdkp, buffer);
3765 		sd_read_write_same(sdkp, buffer);
3766 		sd_read_security(sdkp, buffer);
3767 		sd_config_protection(sdkp, &lim);
3768 	}
3769 
3770 	/*
3771 	 * We now have all cache related info, determine how we deal
3772 	 * with flush requests.
3773 	 */
3774 	sd_set_flush_flag(sdkp, &lim);
3775 
3776 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3777 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3778 
3779 	/* Some devices report a maximum block count for READ/WRITE requests. */
3780 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3781 	lim.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3782 
3783 	if (sd_validate_min_xfer_size(sdkp))
3784 		lim.io_min = logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3785 	else
3786 		lim.io_min = 0;
3787 
3788 	/*
3789 	 * Limit default to SCSI host optimal sector limit if set. There may be
3790 	 * an impact on performance for when the size of a request exceeds this
3791 	 * host limit.
3792 	 */
3793 	lim.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT;
3794 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3795 		lim.io_opt = min_not_zero(lim.io_opt,
3796 				logical_to_bytes(sdp, sdkp->opt_xfer_blocks));
3797 	}
3798 
3799 	sdkp->first_scan = 0;
3800 
3801 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3802 	sd_config_write_same(sdkp, &lim);
3803 	kfree(buffer);
3804 
3805 	blk_mq_freeze_queue(sdkp->disk->queue);
3806 	err = queue_limits_commit_update(sdkp->disk->queue, &lim);
3807 	blk_mq_unfreeze_queue(sdkp->disk->queue);
3808 	if (err)
3809 		return err;
3810 
3811 	/*
3812 	 * For a zoned drive, revalidating the zones can be done only once
3813 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3814 	 * capacity to 0.
3815 	 */
3816 	if (sd_zbc_revalidate_zones(sdkp))
3817 		set_capacity_and_notify(disk, 0);
3818 
3819  out:
3820 	return 0;
3821 }
3822 
3823 /**
3824  *	sd_unlock_native_capacity - unlock native capacity
3825  *	@disk: struct gendisk to set capacity for
3826  *
3827  *	Block layer calls this function if it detects that partitions
3828  *	on @disk reach beyond the end of the device.  If the SCSI host
3829  *	implements ->unlock_native_capacity() method, it's invoked to
3830  *	give it a chance to adjust the device capacity.
3831  *
3832  *	CONTEXT:
3833  *	Defined by block layer.  Might sleep.
3834  */
3835 static void sd_unlock_native_capacity(struct gendisk *disk)
3836 {
3837 	struct scsi_device *sdev = scsi_disk(disk)->device;
3838 
3839 	if (sdev->host->hostt->unlock_native_capacity)
3840 		sdev->host->hostt->unlock_native_capacity(sdev);
3841 }
3842 
3843 /**
3844  *	sd_format_disk_name - format disk name
3845  *	@prefix: name prefix - ie. "sd" for SCSI disks
3846  *	@index: index of the disk to format name for
3847  *	@buf: output buffer
3848  *	@buflen: length of the output buffer
3849  *
3850  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3851  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3852  *	which is followed by sdaaa.
3853  *
3854  *	This is basically 26 base counting with one extra 'nil' entry
3855  *	at the beginning from the second digit on and can be
3856  *	determined using similar method as 26 base conversion with the
3857  *	index shifted -1 after each digit is computed.
3858  *
3859  *	CONTEXT:
3860  *	Don't care.
3861  *
3862  *	RETURNS:
3863  *	0 on success, -errno on failure.
3864  */
3865 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3866 {
3867 	const int base = 'z' - 'a' + 1;
3868 	char *begin = buf + strlen(prefix);
3869 	char *end = buf + buflen;
3870 	char *p;
3871 	int unit;
3872 
3873 	p = end - 1;
3874 	*p = '\0';
3875 	unit = base;
3876 	do {
3877 		if (p == begin)
3878 			return -EINVAL;
3879 		*--p = 'a' + (index % unit);
3880 		index = (index / unit) - 1;
3881 	} while (index >= 0);
3882 
3883 	memmove(begin, p, end - p);
3884 	memcpy(buf, prefix, strlen(prefix));
3885 
3886 	return 0;
3887 }
3888 
3889 /**
3890  *	sd_probe - called during driver initialization and whenever a
3891  *	new scsi device is attached to the system. It is called once
3892  *	for each scsi device (not just disks) present.
3893  *	@dev: pointer to device object
3894  *
3895  *	Returns 0 if successful (or not interested in this scsi device
3896  *	(e.g. scanner)); 1 when there is an error.
3897  *
3898  *	Note: this function is invoked from the scsi mid-level.
3899  *	This function sets up the mapping between a given
3900  *	<host,channel,id,lun> (found in sdp) and new device name
3901  *	(e.g. /dev/sda). More precisely it is the block device major
3902  *	and minor number that is chosen here.
3903  *
3904  *	Assume sd_probe is not re-entrant (for time being)
3905  *	Also think about sd_probe() and sd_remove() running coincidentally.
3906  **/
3907 static int sd_probe(struct device *dev)
3908 {
3909 	struct scsi_device *sdp = to_scsi_device(dev);
3910 	struct scsi_disk *sdkp;
3911 	struct gendisk *gd;
3912 	int index;
3913 	int error;
3914 
3915 	scsi_autopm_get_device(sdp);
3916 	error = -ENODEV;
3917 	if (sdp->type != TYPE_DISK &&
3918 	    sdp->type != TYPE_ZBC &&
3919 	    sdp->type != TYPE_MOD &&
3920 	    sdp->type != TYPE_RBC)
3921 		goto out;
3922 
3923 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3924 		sdev_printk(KERN_WARNING, sdp,
3925 			    "Unsupported ZBC host-managed device.\n");
3926 		goto out;
3927 	}
3928 
3929 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3930 					"sd_probe\n"));
3931 
3932 	error = -ENOMEM;
3933 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3934 	if (!sdkp)
3935 		goto out;
3936 
3937 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3938 					 &sd_bio_compl_lkclass);
3939 	if (!gd)
3940 		goto out_free;
3941 
3942 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3943 	if (index < 0) {
3944 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3945 		goto out_put;
3946 	}
3947 
3948 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3949 	if (error) {
3950 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3951 		goto out_free_index;
3952 	}
3953 
3954 	sdkp->device = sdp;
3955 	sdkp->disk = gd;
3956 	sdkp->index = index;
3957 	sdkp->max_retries = SD_MAX_RETRIES;
3958 	atomic_set(&sdkp->openers, 0);
3959 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3960 
3961 	if (!sdp->request_queue->rq_timeout) {
3962 		if (sdp->type != TYPE_MOD)
3963 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3964 		else
3965 			blk_queue_rq_timeout(sdp->request_queue,
3966 					     SD_MOD_TIMEOUT);
3967 	}
3968 
3969 	device_initialize(&sdkp->disk_dev);
3970 	sdkp->disk_dev.parent = get_device(dev);
3971 	sdkp->disk_dev.class = &sd_disk_class;
3972 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3973 
3974 	error = device_add(&sdkp->disk_dev);
3975 	if (error) {
3976 		put_device(&sdkp->disk_dev);
3977 		goto out;
3978 	}
3979 
3980 	dev_set_drvdata(dev, sdkp);
3981 
3982 	gd->major = sd_major((index & 0xf0) >> 4);
3983 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3984 	gd->minors = SD_MINORS;
3985 
3986 	gd->fops = &sd_fops;
3987 	gd->private_data = sdkp;
3988 
3989 	/* defaults, until the device tells us otherwise */
3990 	sdp->sector_size = 512;
3991 	sdkp->capacity = 0;
3992 	sdkp->media_present = 1;
3993 	sdkp->write_prot = 0;
3994 	sdkp->cache_override = 0;
3995 	sdkp->WCE = 0;
3996 	sdkp->RCD = 0;
3997 	sdkp->ATO = 0;
3998 	sdkp->first_scan = 1;
3999 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
4000 
4001 	sd_revalidate_disk(gd);
4002 
4003 	if (sdp->removable) {
4004 		gd->flags |= GENHD_FL_REMOVABLE;
4005 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
4006 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
4007 	}
4008 
4009 	blk_pm_runtime_init(sdp->request_queue, dev);
4010 	if (sdp->rpm_autosuspend) {
4011 		pm_runtime_set_autosuspend_delay(dev,
4012 			sdp->host->rpm_autosuspend_delay);
4013 	}
4014 
4015 	error = device_add_disk(dev, gd, NULL);
4016 	if (error) {
4017 		device_unregister(&sdkp->disk_dev);
4018 		put_disk(gd);
4019 		goto out;
4020 	}
4021 
4022 	if (sdkp->security) {
4023 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
4024 		if (sdkp->opal_dev)
4025 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
4026 	}
4027 
4028 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
4029 		  sdp->removable ? "removable " : "");
4030 	scsi_autopm_put_device(sdp);
4031 
4032 	return 0;
4033 
4034  out_free_index:
4035 	ida_free(&sd_index_ida, index);
4036  out_put:
4037 	put_disk(gd);
4038  out_free:
4039 	kfree(sdkp);
4040  out:
4041 	scsi_autopm_put_device(sdp);
4042 	return error;
4043 }
4044 
4045 /**
4046  *	sd_remove - called whenever a scsi disk (previously recognized by
4047  *	sd_probe) is detached from the system. It is called (potentially
4048  *	multiple times) during sd module unload.
4049  *	@dev: pointer to device object
4050  *
4051  *	Note: this function is invoked from the scsi mid-level.
4052  *	This function potentially frees up a device name (e.g. /dev/sdc)
4053  *	that could be re-used by a subsequent sd_probe().
4054  *	This function is not called when the built-in sd driver is "exit-ed".
4055  **/
4056 static int sd_remove(struct device *dev)
4057 {
4058 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4059 
4060 	scsi_autopm_get_device(sdkp->device);
4061 
4062 	device_del(&sdkp->disk_dev);
4063 	del_gendisk(sdkp->disk);
4064 	if (!sdkp->suspended)
4065 		sd_shutdown(dev);
4066 
4067 	put_disk(sdkp->disk);
4068 	return 0;
4069 }
4070 
4071 static void scsi_disk_release(struct device *dev)
4072 {
4073 	struct scsi_disk *sdkp = to_scsi_disk(dev);
4074 
4075 	ida_free(&sd_index_ida, sdkp->index);
4076 	put_device(&sdkp->device->sdev_gendev);
4077 	free_opal_dev(sdkp->opal_dev);
4078 
4079 	kfree(sdkp);
4080 }
4081 
4082 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
4083 {
4084 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
4085 	struct scsi_sense_hdr sshdr;
4086 	const struct scsi_exec_args exec_args = {
4087 		.sshdr = &sshdr,
4088 		.req_flags = BLK_MQ_REQ_PM,
4089 	};
4090 	struct scsi_device *sdp = sdkp->device;
4091 	int res;
4092 
4093 	if (start)
4094 		cmd[4] |= 1;	/* START */
4095 
4096 	if (sdp->start_stop_pwr_cond)
4097 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
4098 
4099 	if (!scsi_device_online(sdp))
4100 		return -ENODEV;
4101 
4102 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
4103 			       sdkp->max_retries, &exec_args);
4104 	if (res) {
4105 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
4106 		if (res > 0 && scsi_sense_valid(&sshdr)) {
4107 			sd_print_sense_hdr(sdkp, &sshdr);
4108 			/* 0x3a is medium not present */
4109 			if (sshdr.asc == 0x3a)
4110 				res = 0;
4111 		}
4112 	}
4113 
4114 	/* SCSI error codes must not go to the generic layer */
4115 	if (res)
4116 		return -EIO;
4117 
4118 	return 0;
4119 }
4120 
4121 /*
4122  * Send a SYNCHRONIZE CACHE instruction down to the device through
4123  * the normal SCSI command structure.  Wait for the command to
4124  * complete.
4125  */
4126 static void sd_shutdown(struct device *dev)
4127 {
4128 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4129 
4130 	if (!sdkp)
4131 		return;         /* this can happen */
4132 
4133 	if (pm_runtime_suspended(dev))
4134 		return;
4135 
4136 	if (sdkp->WCE && sdkp->media_present) {
4137 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4138 		sd_sync_cache(sdkp);
4139 	}
4140 
4141 	if ((system_state != SYSTEM_RESTART &&
4142 	     sdkp->device->manage_system_start_stop) ||
4143 	    (system_state == SYSTEM_POWER_OFF &&
4144 	     sdkp->device->manage_shutdown)) {
4145 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4146 		sd_start_stop_device(sdkp, 0);
4147 	}
4148 }
4149 
4150 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
4151 {
4152 	return (sdev->manage_system_start_stop && !runtime) ||
4153 		(sdev->manage_runtime_start_stop && runtime);
4154 }
4155 
4156 static int sd_suspend_common(struct device *dev, bool runtime)
4157 {
4158 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4159 	int ret = 0;
4160 
4161 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
4162 		return 0;
4163 
4164 	if (sdkp->WCE && sdkp->media_present) {
4165 		if (!sdkp->device->silence_suspend)
4166 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
4167 		ret = sd_sync_cache(sdkp);
4168 		/* ignore OFFLINE device */
4169 		if (ret == -ENODEV)
4170 			return 0;
4171 
4172 		if (ret)
4173 			return ret;
4174 	}
4175 
4176 	if (sd_do_start_stop(sdkp->device, runtime)) {
4177 		if (!sdkp->device->silence_suspend)
4178 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
4179 		/* an error is not worth aborting a system sleep */
4180 		ret = sd_start_stop_device(sdkp, 0);
4181 		if (!runtime)
4182 			ret = 0;
4183 	}
4184 
4185 	if (!ret)
4186 		sdkp->suspended = true;
4187 
4188 	return ret;
4189 }
4190 
4191 static int sd_suspend_system(struct device *dev)
4192 {
4193 	if (pm_runtime_suspended(dev))
4194 		return 0;
4195 
4196 	return sd_suspend_common(dev, false);
4197 }
4198 
4199 static int sd_suspend_runtime(struct device *dev)
4200 {
4201 	return sd_suspend_common(dev, true);
4202 }
4203 
4204 static int sd_resume(struct device *dev)
4205 {
4206 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4207 
4208 	if (opal_unlock_from_suspend(sdkp->opal_dev)) {
4209 		sd_printk(KERN_NOTICE, sdkp, "OPAL unlock failed\n");
4210 		return -EIO;
4211 	}
4212 
4213 	return 0;
4214 }
4215 
4216 static int sd_resume_common(struct device *dev, bool runtime)
4217 {
4218 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4219 	int ret;
4220 
4221 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4222 		return 0;
4223 
4224 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
4225 
4226 	if (!sd_do_start_stop(sdkp->device, runtime)) {
4227 		sdkp->suspended = false;
4228 		return 0;
4229 	}
4230 
4231 	ret = sd_start_stop_device(sdkp, 1);
4232 	if (!ret) {
4233 		sd_resume(dev);
4234 		sdkp->suspended = false;
4235 	}
4236 
4237 	return ret;
4238 }
4239 
4240 static int sd_resume_system(struct device *dev)
4241 {
4242 	if (pm_runtime_suspended(dev)) {
4243 		struct scsi_disk *sdkp = dev_get_drvdata(dev);
4244 		struct scsi_device *sdp = sdkp ? sdkp->device : NULL;
4245 
4246 		if (sdp && sdp->force_runtime_start_on_system_start)
4247 			pm_request_resume(dev);
4248 
4249 		return 0;
4250 	}
4251 
4252 	return sd_resume_common(dev, false);
4253 }
4254 
4255 static int sd_resume_runtime(struct device *dev)
4256 {
4257 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4258 	struct scsi_device *sdp;
4259 
4260 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4261 		return 0;
4262 
4263 	sdp = sdkp->device;
4264 
4265 	if (sdp->ignore_media_change) {
4266 		/* clear the device's sense data */
4267 		static const u8 cmd[10] = { REQUEST_SENSE };
4268 		const struct scsi_exec_args exec_args = {
4269 			.req_flags = BLK_MQ_REQ_PM,
4270 		};
4271 
4272 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
4273 				     sdp->request_queue->rq_timeout, 1,
4274 				     &exec_args))
4275 			sd_printk(KERN_NOTICE, sdkp,
4276 				  "Failed to clear sense data\n");
4277 	}
4278 
4279 	return sd_resume_common(dev, true);
4280 }
4281 
4282 static const struct dev_pm_ops sd_pm_ops = {
4283 	.suspend		= sd_suspend_system,
4284 	.resume			= sd_resume_system,
4285 	.poweroff		= sd_suspend_system,
4286 	.restore		= sd_resume_system,
4287 	.runtime_suspend	= sd_suspend_runtime,
4288 	.runtime_resume		= sd_resume_runtime,
4289 };
4290 
4291 static struct scsi_driver sd_template = {
4292 	.gendrv = {
4293 		.name		= "sd",
4294 		.probe		= sd_probe,
4295 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
4296 		.remove		= sd_remove,
4297 		.shutdown	= sd_shutdown,
4298 		.pm		= &sd_pm_ops,
4299 	},
4300 	.rescan			= sd_rescan,
4301 	.resume			= sd_resume,
4302 	.init_command		= sd_init_command,
4303 	.uninit_command		= sd_uninit_command,
4304 	.done			= sd_done,
4305 	.eh_action		= sd_eh_action,
4306 	.eh_reset		= sd_eh_reset,
4307 };
4308 
4309 /**
4310  *	init_sd - entry point for this driver (both when built in or when
4311  *	a module).
4312  *
4313  *	Note: this function registers this driver with the scsi mid-level.
4314  **/
4315 static int __init init_sd(void)
4316 {
4317 	int majors = 0, i, err;
4318 
4319 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
4320 
4321 	for (i = 0; i < SD_MAJORS; i++) {
4322 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
4323 			continue;
4324 		majors++;
4325 	}
4326 
4327 	if (!majors)
4328 		return -ENODEV;
4329 
4330 	err = class_register(&sd_disk_class);
4331 	if (err)
4332 		goto err_out;
4333 
4334 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4335 	if (!sd_page_pool) {
4336 		printk(KERN_ERR "sd: can't init discard page pool\n");
4337 		err = -ENOMEM;
4338 		goto err_out_class;
4339 	}
4340 
4341 	err = scsi_register_driver(&sd_template.gendrv);
4342 	if (err)
4343 		goto err_out_driver;
4344 
4345 	return 0;
4346 
4347 err_out_driver:
4348 	mempool_destroy(sd_page_pool);
4349 err_out_class:
4350 	class_unregister(&sd_disk_class);
4351 err_out:
4352 	for (i = 0; i < SD_MAJORS; i++)
4353 		unregister_blkdev(sd_major(i), "sd");
4354 	return err;
4355 }
4356 
4357 /**
4358  *	exit_sd - exit point for this driver (when it is a module).
4359  *
4360  *	Note: this function unregisters this driver from the scsi mid-level.
4361  **/
4362 static void __exit exit_sd(void)
4363 {
4364 	int i;
4365 
4366 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4367 
4368 	scsi_unregister_driver(&sd_template.gendrv);
4369 	mempool_destroy(sd_page_pool);
4370 
4371 	class_unregister(&sd_disk_class);
4372 
4373 	for (i = 0; i < SD_MAJORS; i++)
4374 		unregister_blkdev(sd_major(i), "sd");
4375 }
4376 
4377 module_init(init_sd);
4378 module_exit(exit_sd);
4379 
4380 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4381 {
4382 	scsi_print_sense_hdr(sdkp->device,
4383 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4384 }
4385 
4386 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4387 {
4388 	const char *hb_string = scsi_hostbyte_string(result);
4389 
4390 	if (hb_string)
4391 		sd_printk(KERN_INFO, sdkp,
4392 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4393 			  hb_string ? hb_string : "invalid",
4394 			  "DRIVER_OK");
4395 	else
4396 		sd_printk(KERN_INFO, sdkp,
4397 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4398 			  msg, host_byte(result), "DRIVER_OK");
4399 }
4400