xref: /openbmc/linux/drivers/scsi/sd.c (revision 22b6e7f3)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *      sd.c Copyright (C) 1992 Drew Eckhardt
4  *           Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5  *
6  *      Linux scsi disk driver
7  *              Initial versions: Drew Eckhardt
8  *              Subsequent revisions: Eric Youngdale
9  *	Modification history:
10  *       - Drew Eckhardt <drew@colorado.edu> original
11  *       - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12  *         outstanding request, and other enhancements.
13  *         Support loadable low-level scsi drivers.
14  *       - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15  *         eight major numbers.
16  *       - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17  *	 - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18  *	   sd_init and cleanups.
19  *	 - Alex Davis <letmein@erols.com> Fix problem where partition info
20  *	   not being read in sd_open. Fix problem where removable media
21  *	   could be ejected after sd_open.
22  *	 - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23  *	 - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24  *	   <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25  *	   Support 32k/1M disks.
26  *
27  *	Logging policy (needs CONFIG_SCSI_LOGGING defined):
28  *	 - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29  *	 - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30  *	 - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31  *	 - entering other commands: SCSI_LOG_HLQUEUE level 3
32  *	Note: when the logging level is set by the user, it must be greater
33  *	than the level indicated above to trigger output.
34  */
35 
36 #include <linux/module.h>
37 #include <linux/fs.h>
38 #include <linux/kernel.h>
39 #include <linux/mm.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
56 #include <linux/pr.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
60 
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 #include <scsi/scsi_common.h>
71 
72 #include "sd.h"
73 #include "scsi_priv.h"
74 #include "scsi_logging.h"
75 
76 MODULE_AUTHOR("Eric Youngdale");
77 MODULE_DESCRIPTION("SCSI disk (sd) driver");
78 MODULE_LICENSE("GPL");
79 
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
100 
101 #define SD_MINORS	16
102 
103 static void sd_config_discard(struct scsi_disk *, unsigned int);
104 static void sd_config_write_same(struct scsi_disk *);
105 static int  sd_revalidate_disk(struct gendisk *);
106 static void sd_unlock_native_capacity(struct gendisk *disk);
107 static void sd_shutdown(struct device *);
108 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
109 static void scsi_disk_release(struct device *cdev);
110 
111 static DEFINE_IDA(sd_index_ida);
112 
113 static mempool_t *sd_page_pool;
114 static struct lock_class_key sd_bio_compl_lkclass;
115 
116 static const char *sd_cache_types[] = {
117 	"write through", "none", "write back",
118 	"write back, no read (daft)"
119 };
120 
121 static void sd_set_flush_flag(struct scsi_disk *sdkp)
122 {
123 	bool wc = false, fua = false;
124 
125 	if (sdkp->WCE) {
126 		wc = true;
127 		if (sdkp->DPOFUA)
128 			fua = true;
129 	}
130 
131 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
132 }
133 
134 static ssize_t
135 cache_type_store(struct device *dev, struct device_attribute *attr,
136 		 const char *buf, size_t count)
137 {
138 	int ct, rcd, wce, sp;
139 	struct scsi_disk *sdkp = to_scsi_disk(dev);
140 	struct scsi_device *sdp = sdkp->device;
141 	char buffer[64];
142 	char *buffer_data;
143 	struct scsi_mode_data data;
144 	struct scsi_sense_hdr sshdr;
145 	static const char temp[] = "temporary ";
146 	int len;
147 
148 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
149 		/* no cache control on RBC devices; theoretically they
150 		 * can do it, but there's probably so many exceptions
151 		 * it's not worth the risk */
152 		return -EINVAL;
153 
154 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
155 		buf += sizeof(temp) - 1;
156 		sdkp->cache_override = 1;
157 	} else {
158 		sdkp->cache_override = 0;
159 	}
160 
161 	ct = sysfs_match_string(sd_cache_types, buf);
162 	if (ct < 0)
163 		return -EINVAL;
164 
165 	rcd = ct & 0x01 ? 1 : 0;
166 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
167 
168 	if (sdkp->cache_override) {
169 		sdkp->WCE = wce;
170 		sdkp->RCD = rcd;
171 		sd_set_flush_flag(sdkp);
172 		return count;
173 	}
174 
175 	if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
176 			    sdkp->max_retries, &data, NULL))
177 		return -EINVAL;
178 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
179 		  data.block_descriptor_length);
180 	buffer_data = buffer + data.header_length +
181 		data.block_descriptor_length;
182 	buffer_data[2] &= ~0x05;
183 	buffer_data[2] |= wce << 2 | rcd;
184 	sp = buffer_data[0] & 0x80 ? 1 : 0;
185 	buffer_data[0] &= ~0x80;
186 
187 	/*
188 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
189 	 * received mode parameter buffer before doing MODE SELECT.
190 	 */
191 	data.device_specific = 0;
192 
193 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
194 			     sdkp->max_retries, &data, &sshdr)) {
195 		if (scsi_sense_valid(&sshdr))
196 			sd_print_sense_hdr(sdkp, &sshdr);
197 		return -EINVAL;
198 	}
199 	sd_revalidate_disk(sdkp->disk);
200 	return count;
201 }
202 
203 static ssize_t
204 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
205 		       char *buf)
206 {
207 	struct scsi_disk *sdkp = to_scsi_disk(dev);
208 	struct scsi_device *sdp = sdkp->device;
209 
210 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
211 }
212 
213 static ssize_t
214 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
215 			const char *buf, size_t count)
216 {
217 	struct scsi_disk *sdkp = to_scsi_disk(dev);
218 	struct scsi_device *sdp = sdkp->device;
219 	bool v;
220 
221 	if (!capable(CAP_SYS_ADMIN))
222 		return -EACCES;
223 
224 	if (kstrtobool(buf, &v))
225 		return -EINVAL;
226 
227 	sdp->manage_start_stop = v;
228 
229 	return count;
230 }
231 static DEVICE_ATTR_RW(manage_start_stop);
232 
233 static ssize_t
234 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
235 {
236 	struct scsi_disk *sdkp = to_scsi_disk(dev);
237 
238 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
239 }
240 
241 static ssize_t
242 allow_restart_store(struct device *dev, struct device_attribute *attr,
243 		    const char *buf, size_t count)
244 {
245 	bool v;
246 	struct scsi_disk *sdkp = to_scsi_disk(dev);
247 	struct scsi_device *sdp = sdkp->device;
248 
249 	if (!capable(CAP_SYS_ADMIN))
250 		return -EACCES;
251 
252 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
253 		return -EINVAL;
254 
255 	if (kstrtobool(buf, &v))
256 		return -EINVAL;
257 
258 	sdp->allow_restart = v;
259 
260 	return count;
261 }
262 static DEVICE_ATTR_RW(allow_restart);
263 
264 static ssize_t
265 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
266 {
267 	struct scsi_disk *sdkp = to_scsi_disk(dev);
268 	int ct = sdkp->RCD + 2*sdkp->WCE;
269 
270 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
271 }
272 static DEVICE_ATTR_RW(cache_type);
273 
274 static ssize_t
275 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
276 {
277 	struct scsi_disk *sdkp = to_scsi_disk(dev);
278 
279 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
280 }
281 static DEVICE_ATTR_RO(FUA);
282 
283 static ssize_t
284 protection_type_show(struct device *dev, struct device_attribute *attr,
285 		     char *buf)
286 {
287 	struct scsi_disk *sdkp = to_scsi_disk(dev);
288 
289 	return sprintf(buf, "%u\n", sdkp->protection_type);
290 }
291 
292 static ssize_t
293 protection_type_store(struct device *dev, struct device_attribute *attr,
294 		      const char *buf, size_t count)
295 {
296 	struct scsi_disk *sdkp = to_scsi_disk(dev);
297 	unsigned int val;
298 	int err;
299 
300 	if (!capable(CAP_SYS_ADMIN))
301 		return -EACCES;
302 
303 	err = kstrtouint(buf, 10, &val);
304 
305 	if (err)
306 		return err;
307 
308 	if (val <= T10_PI_TYPE3_PROTECTION)
309 		sdkp->protection_type = val;
310 
311 	return count;
312 }
313 static DEVICE_ATTR_RW(protection_type);
314 
315 static ssize_t
316 protection_mode_show(struct device *dev, struct device_attribute *attr,
317 		     char *buf)
318 {
319 	struct scsi_disk *sdkp = to_scsi_disk(dev);
320 	struct scsi_device *sdp = sdkp->device;
321 	unsigned int dif, dix;
322 
323 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
324 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
325 
326 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
327 		dif = 0;
328 		dix = 1;
329 	}
330 
331 	if (!dif && !dix)
332 		return sprintf(buf, "none\n");
333 
334 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
335 }
336 static DEVICE_ATTR_RO(protection_mode);
337 
338 static ssize_t
339 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
340 {
341 	struct scsi_disk *sdkp = to_scsi_disk(dev);
342 
343 	return sprintf(buf, "%u\n", sdkp->ATO);
344 }
345 static DEVICE_ATTR_RO(app_tag_own);
346 
347 static ssize_t
348 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
349 		       char *buf)
350 {
351 	struct scsi_disk *sdkp = to_scsi_disk(dev);
352 
353 	return sprintf(buf, "%u\n", sdkp->lbpme);
354 }
355 static DEVICE_ATTR_RO(thin_provisioning);
356 
357 /* sysfs_match_string() requires dense arrays */
358 static const char *lbp_mode[] = {
359 	[SD_LBP_FULL]		= "full",
360 	[SD_LBP_UNMAP]		= "unmap",
361 	[SD_LBP_WS16]		= "writesame_16",
362 	[SD_LBP_WS10]		= "writesame_10",
363 	[SD_LBP_ZERO]		= "writesame_zero",
364 	[SD_LBP_DISABLE]	= "disabled",
365 };
366 
367 static ssize_t
368 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
369 		       char *buf)
370 {
371 	struct scsi_disk *sdkp = to_scsi_disk(dev);
372 
373 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
374 }
375 
376 static ssize_t
377 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
378 			const char *buf, size_t count)
379 {
380 	struct scsi_disk *sdkp = to_scsi_disk(dev);
381 	struct scsi_device *sdp = sdkp->device;
382 	int mode;
383 
384 	if (!capable(CAP_SYS_ADMIN))
385 		return -EACCES;
386 
387 	if (sd_is_zoned(sdkp)) {
388 		sd_config_discard(sdkp, SD_LBP_DISABLE);
389 		return count;
390 	}
391 
392 	if (sdp->type != TYPE_DISK)
393 		return -EINVAL;
394 
395 	mode = sysfs_match_string(lbp_mode, buf);
396 	if (mode < 0)
397 		return -EINVAL;
398 
399 	sd_config_discard(sdkp, mode);
400 
401 	return count;
402 }
403 static DEVICE_ATTR_RW(provisioning_mode);
404 
405 /* sysfs_match_string() requires dense arrays */
406 static const char *zeroing_mode[] = {
407 	[SD_ZERO_WRITE]		= "write",
408 	[SD_ZERO_WS]		= "writesame",
409 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
410 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
411 };
412 
413 static ssize_t
414 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
415 		  char *buf)
416 {
417 	struct scsi_disk *sdkp = to_scsi_disk(dev);
418 
419 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
420 }
421 
422 static ssize_t
423 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
424 		   const char *buf, size_t count)
425 {
426 	struct scsi_disk *sdkp = to_scsi_disk(dev);
427 	int mode;
428 
429 	if (!capable(CAP_SYS_ADMIN))
430 		return -EACCES;
431 
432 	mode = sysfs_match_string(zeroing_mode, buf);
433 	if (mode < 0)
434 		return -EINVAL;
435 
436 	sdkp->zeroing_mode = mode;
437 
438 	return count;
439 }
440 static DEVICE_ATTR_RW(zeroing_mode);
441 
442 static ssize_t
443 max_medium_access_timeouts_show(struct device *dev,
444 				struct device_attribute *attr, char *buf)
445 {
446 	struct scsi_disk *sdkp = to_scsi_disk(dev);
447 
448 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
449 }
450 
451 static ssize_t
452 max_medium_access_timeouts_store(struct device *dev,
453 				 struct device_attribute *attr, const char *buf,
454 				 size_t count)
455 {
456 	struct scsi_disk *sdkp = to_scsi_disk(dev);
457 	int err;
458 
459 	if (!capable(CAP_SYS_ADMIN))
460 		return -EACCES;
461 
462 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
463 
464 	return err ? err : count;
465 }
466 static DEVICE_ATTR_RW(max_medium_access_timeouts);
467 
468 static ssize_t
469 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
470 			   char *buf)
471 {
472 	struct scsi_disk *sdkp = to_scsi_disk(dev);
473 
474 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
475 }
476 
477 static ssize_t
478 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
479 			    const char *buf, size_t count)
480 {
481 	struct scsi_disk *sdkp = to_scsi_disk(dev);
482 	struct scsi_device *sdp = sdkp->device;
483 	unsigned long max;
484 	int err;
485 
486 	if (!capable(CAP_SYS_ADMIN))
487 		return -EACCES;
488 
489 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
490 		return -EINVAL;
491 
492 	err = kstrtoul(buf, 10, &max);
493 
494 	if (err)
495 		return err;
496 
497 	if (max == 0)
498 		sdp->no_write_same = 1;
499 	else if (max <= SD_MAX_WS16_BLOCKS) {
500 		sdp->no_write_same = 0;
501 		sdkp->max_ws_blocks = max;
502 	}
503 
504 	sd_config_write_same(sdkp);
505 
506 	return count;
507 }
508 static DEVICE_ATTR_RW(max_write_same_blocks);
509 
510 static ssize_t
511 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
512 {
513 	struct scsi_disk *sdkp = to_scsi_disk(dev);
514 
515 	if (sdkp->device->type == TYPE_ZBC)
516 		return sprintf(buf, "host-managed\n");
517 	if (sdkp->zoned == 1)
518 		return sprintf(buf, "host-aware\n");
519 	if (sdkp->zoned == 2)
520 		return sprintf(buf, "drive-managed\n");
521 	return sprintf(buf, "none\n");
522 }
523 static DEVICE_ATTR_RO(zoned_cap);
524 
525 static ssize_t
526 max_retries_store(struct device *dev, struct device_attribute *attr,
527 		  const char *buf, size_t count)
528 {
529 	struct scsi_disk *sdkp = to_scsi_disk(dev);
530 	struct scsi_device *sdev = sdkp->device;
531 	int retries, err;
532 
533 	err = kstrtoint(buf, 10, &retries);
534 	if (err)
535 		return err;
536 
537 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
538 		sdkp->max_retries = retries;
539 		return count;
540 	}
541 
542 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
543 		    SD_MAX_RETRIES);
544 	return -EINVAL;
545 }
546 
547 static ssize_t
548 max_retries_show(struct device *dev, struct device_attribute *attr,
549 		 char *buf)
550 {
551 	struct scsi_disk *sdkp = to_scsi_disk(dev);
552 
553 	return sprintf(buf, "%d\n", sdkp->max_retries);
554 }
555 
556 static DEVICE_ATTR_RW(max_retries);
557 
558 static struct attribute *sd_disk_attrs[] = {
559 	&dev_attr_cache_type.attr,
560 	&dev_attr_FUA.attr,
561 	&dev_attr_allow_restart.attr,
562 	&dev_attr_manage_start_stop.attr,
563 	&dev_attr_protection_type.attr,
564 	&dev_attr_protection_mode.attr,
565 	&dev_attr_app_tag_own.attr,
566 	&dev_attr_thin_provisioning.attr,
567 	&dev_attr_provisioning_mode.attr,
568 	&dev_attr_zeroing_mode.attr,
569 	&dev_attr_max_write_same_blocks.attr,
570 	&dev_attr_max_medium_access_timeouts.attr,
571 	&dev_attr_zoned_cap.attr,
572 	&dev_attr_max_retries.attr,
573 	NULL,
574 };
575 ATTRIBUTE_GROUPS(sd_disk);
576 
577 static struct class sd_disk_class = {
578 	.name		= "scsi_disk",
579 	.dev_release	= scsi_disk_release,
580 	.dev_groups	= sd_disk_groups,
581 };
582 
583 /*
584  * Don't request a new module, as that could deadlock in multipath
585  * environment.
586  */
587 static void sd_default_probe(dev_t devt)
588 {
589 }
590 
591 /*
592  * Device no to disk mapping:
593  *
594  *       major         disc2     disc  p1
595  *   |............|.............|....|....| <- dev_t
596  *    31        20 19          8 7  4 3  0
597  *
598  * Inside a major, we have 16k disks, however mapped non-
599  * contiguously. The first 16 disks are for major0, the next
600  * ones with major1, ... Disk 256 is for major0 again, disk 272
601  * for major1, ...
602  * As we stay compatible with our numbering scheme, we can reuse
603  * the well-know SCSI majors 8, 65--71, 136--143.
604  */
605 static int sd_major(int major_idx)
606 {
607 	switch (major_idx) {
608 	case 0:
609 		return SCSI_DISK0_MAJOR;
610 	case 1 ... 7:
611 		return SCSI_DISK1_MAJOR + major_idx - 1;
612 	case 8 ... 15:
613 		return SCSI_DISK8_MAJOR + major_idx - 8;
614 	default:
615 		BUG();
616 		return 0;	/* shut up gcc */
617 	}
618 }
619 
620 #ifdef CONFIG_BLK_SED_OPAL
621 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
622 		size_t len, bool send)
623 {
624 	struct scsi_disk *sdkp = data;
625 	struct scsi_device *sdev = sdkp->device;
626 	u8 cdb[12] = { 0, };
627 	const struct scsi_exec_args exec_args = {
628 		.req_flags = BLK_MQ_REQ_PM,
629 	};
630 	int ret;
631 
632 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
633 	cdb[1] = secp;
634 	put_unaligned_be16(spsp, &cdb[2]);
635 	put_unaligned_be32(len, &cdb[6]);
636 
637 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
638 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
639 			       &exec_args);
640 	return ret <= 0 ? ret : -EIO;
641 }
642 #endif /* CONFIG_BLK_SED_OPAL */
643 
644 /*
645  * Look up the DIX operation based on whether the command is read or
646  * write and whether dix and dif are enabled.
647  */
648 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
649 {
650 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
651 	static const unsigned int ops[] = {	/* wrt dix dif */
652 		SCSI_PROT_NORMAL,		/*  0	0   0  */
653 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
654 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
655 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
656 		SCSI_PROT_NORMAL,		/*  1	0   0  */
657 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
658 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
659 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
660 	};
661 
662 	return ops[write << 2 | dix << 1 | dif];
663 }
664 
665 /*
666  * Returns a mask of the protection flags that are valid for a given DIX
667  * operation.
668  */
669 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
670 {
671 	static const unsigned int flag_mask[] = {
672 		[SCSI_PROT_NORMAL]		= 0,
673 
674 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
675 						  SCSI_PROT_GUARD_CHECK |
676 						  SCSI_PROT_REF_CHECK |
677 						  SCSI_PROT_REF_INCREMENT,
678 
679 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
680 						  SCSI_PROT_IP_CHECKSUM,
681 
682 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
683 						  SCSI_PROT_GUARD_CHECK |
684 						  SCSI_PROT_REF_CHECK |
685 						  SCSI_PROT_REF_INCREMENT |
686 						  SCSI_PROT_IP_CHECKSUM,
687 
688 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
689 						  SCSI_PROT_REF_INCREMENT,
690 
691 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
692 						  SCSI_PROT_REF_CHECK |
693 						  SCSI_PROT_REF_INCREMENT |
694 						  SCSI_PROT_IP_CHECKSUM,
695 
696 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
697 						  SCSI_PROT_GUARD_CHECK |
698 						  SCSI_PROT_REF_CHECK |
699 						  SCSI_PROT_REF_INCREMENT |
700 						  SCSI_PROT_IP_CHECKSUM,
701 	};
702 
703 	return flag_mask[prot_op];
704 }
705 
706 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
707 					   unsigned int dix, unsigned int dif)
708 {
709 	struct request *rq = scsi_cmd_to_rq(scmd);
710 	struct bio *bio = rq->bio;
711 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
712 	unsigned int protect = 0;
713 
714 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
715 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
716 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
717 
718 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
719 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
720 	}
721 
722 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
723 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
724 
725 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
726 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
727 	}
728 
729 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
730 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
731 
732 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
733 			protect = 3 << 5;	/* Disable target PI checking */
734 		else
735 			protect = 1 << 5;	/* Enable target PI checking */
736 	}
737 
738 	scsi_set_prot_op(scmd, prot_op);
739 	scsi_set_prot_type(scmd, dif);
740 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
741 
742 	return protect;
743 }
744 
745 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
746 {
747 	struct request_queue *q = sdkp->disk->queue;
748 	unsigned int logical_block_size = sdkp->device->sector_size;
749 	unsigned int max_blocks = 0;
750 
751 	q->limits.discard_alignment =
752 		sdkp->unmap_alignment * logical_block_size;
753 	q->limits.discard_granularity =
754 		max(sdkp->physical_block_size,
755 		    sdkp->unmap_granularity * logical_block_size);
756 	sdkp->provisioning_mode = mode;
757 
758 	switch (mode) {
759 
760 	case SD_LBP_FULL:
761 	case SD_LBP_DISABLE:
762 		blk_queue_max_discard_sectors(q, 0);
763 		return;
764 
765 	case SD_LBP_UNMAP:
766 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
767 					  (u32)SD_MAX_WS16_BLOCKS);
768 		break;
769 
770 	case SD_LBP_WS16:
771 		if (sdkp->device->unmap_limit_for_ws)
772 			max_blocks = sdkp->max_unmap_blocks;
773 		else
774 			max_blocks = sdkp->max_ws_blocks;
775 
776 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
777 		break;
778 
779 	case SD_LBP_WS10:
780 		if (sdkp->device->unmap_limit_for_ws)
781 			max_blocks = sdkp->max_unmap_blocks;
782 		else
783 			max_blocks = sdkp->max_ws_blocks;
784 
785 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
786 		break;
787 
788 	case SD_LBP_ZERO:
789 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
790 					  (u32)SD_MAX_WS10_BLOCKS);
791 		break;
792 	}
793 
794 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
795 }
796 
797 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
798 {
799 	struct page *page;
800 
801 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
802 	if (!page)
803 		return NULL;
804 	clear_highpage(page);
805 	bvec_set_page(&rq->special_vec, page, data_len, 0);
806 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
807 	return bvec_virt(&rq->special_vec);
808 }
809 
810 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
811 {
812 	struct scsi_device *sdp = cmd->device;
813 	struct request *rq = scsi_cmd_to_rq(cmd);
814 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
815 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
816 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
817 	unsigned int data_len = 24;
818 	char *buf;
819 
820 	buf = sd_set_special_bvec(rq, data_len);
821 	if (!buf)
822 		return BLK_STS_RESOURCE;
823 
824 	cmd->cmd_len = 10;
825 	cmd->cmnd[0] = UNMAP;
826 	cmd->cmnd[8] = 24;
827 
828 	put_unaligned_be16(6 + 16, &buf[0]);
829 	put_unaligned_be16(16, &buf[2]);
830 	put_unaligned_be64(lba, &buf[8]);
831 	put_unaligned_be32(nr_blocks, &buf[16]);
832 
833 	cmd->allowed = sdkp->max_retries;
834 	cmd->transfersize = data_len;
835 	rq->timeout = SD_TIMEOUT;
836 
837 	return scsi_alloc_sgtables(cmd);
838 }
839 
840 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
841 		bool unmap)
842 {
843 	struct scsi_device *sdp = cmd->device;
844 	struct request *rq = scsi_cmd_to_rq(cmd);
845 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
846 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
847 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
848 	u32 data_len = sdp->sector_size;
849 
850 	if (!sd_set_special_bvec(rq, data_len))
851 		return BLK_STS_RESOURCE;
852 
853 	cmd->cmd_len = 16;
854 	cmd->cmnd[0] = WRITE_SAME_16;
855 	if (unmap)
856 		cmd->cmnd[1] = 0x8; /* UNMAP */
857 	put_unaligned_be64(lba, &cmd->cmnd[2]);
858 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
859 
860 	cmd->allowed = sdkp->max_retries;
861 	cmd->transfersize = data_len;
862 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
863 
864 	return scsi_alloc_sgtables(cmd);
865 }
866 
867 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
868 		bool unmap)
869 {
870 	struct scsi_device *sdp = cmd->device;
871 	struct request *rq = scsi_cmd_to_rq(cmd);
872 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
873 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
874 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
875 	u32 data_len = sdp->sector_size;
876 
877 	if (!sd_set_special_bvec(rq, data_len))
878 		return BLK_STS_RESOURCE;
879 
880 	cmd->cmd_len = 10;
881 	cmd->cmnd[0] = WRITE_SAME;
882 	if (unmap)
883 		cmd->cmnd[1] = 0x8; /* UNMAP */
884 	put_unaligned_be32(lba, &cmd->cmnd[2]);
885 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
886 
887 	cmd->allowed = sdkp->max_retries;
888 	cmd->transfersize = data_len;
889 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
890 
891 	return scsi_alloc_sgtables(cmd);
892 }
893 
894 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
895 {
896 	struct request *rq = scsi_cmd_to_rq(cmd);
897 	struct scsi_device *sdp = cmd->device;
898 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
899 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
900 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
901 
902 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
903 		switch (sdkp->zeroing_mode) {
904 		case SD_ZERO_WS16_UNMAP:
905 			return sd_setup_write_same16_cmnd(cmd, true);
906 		case SD_ZERO_WS10_UNMAP:
907 			return sd_setup_write_same10_cmnd(cmd, true);
908 		}
909 	}
910 
911 	if (sdp->no_write_same) {
912 		rq->rq_flags |= RQF_QUIET;
913 		return BLK_STS_TARGET;
914 	}
915 
916 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
917 		return sd_setup_write_same16_cmnd(cmd, false);
918 
919 	return sd_setup_write_same10_cmnd(cmd, false);
920 }
921 
922 static void sd_config_write_same(struct scsi_disk *sdkp)
923 {
924 	struct request_queue *q = sdkp->disk->queue;
925 	unsigned int logical_block_size = sdkp->device->sector_size;
926 
927 	if (sdkp->device->no_write_same) {
928 		sdkp->max_ws_blocks = 0;
929 		goto out;
930 	}
931 
932 	/* Some devices can not handle block counts above 0xffff despite
933 	 * supporting WRITE SAME(16). Consequently we default to 64k
934 	 * blocks per I/O unless the device explicitly advertises a
935 	 * bigger limit.
936 	 */
937 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
938 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
939 						   (u32)SD_MAX_WS16_BLOCKS);
940 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
941 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
942 						   (u32)SD_MAX_WS10_BLOCKS);
943 	else {
944 		sdkp->device->no_write_same = 1;
945 		sdkp->max_ws_blocks = 0;
946 	}
947 
948 	if (sdkp->lbprz && sdkp->lbpws)
949 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
950 	else if (sdkp->lbprz && sdkp->lbpws10)
951 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
952 	else if (sdkp->max_ws_blocks)
953 		sdkp->zeroing_mode = SD_ZERO_WS;
954 	else
955 		sdkp->zeroing_mode = SD_ZERO_WRITE;
956 
957 	if (sdkp->max_ws_blocks &&
958 	    sdkp->physical_block_size > logical_block_size) {
959 		/*
960 		 * Reporting a maximum number of blocks that is not aligned
961 		 * on the device physical size would cause a large write same
962 		 * request to be split into physically unaligned chunks by
963 		 * __blkdev_issue_write_zeroes() even if the caller of this
964 		 * functions took care to align the large request. So make sure
965 		 * the maximum reported is aligned to the device physical block
966 		 * size. This is only an optional optimization for regular
967 		 * disks, but this is mandatory to avoid failure of large write
968 		 * same requests directed at sequential write required zones of
969 		 * host-managed ZBC disks.
970 		 */
971 		sdkp->max_ws_blocks =
972 			round_down(sdkp->max_ws_blocks,
973 				   bytes_to_logical(sdkp->device,
974 						    sdkp->physical_block_size));
975 	}
976 
977 out:
978 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
979 					 (logical_block_size >> 9));
980 }
981 
982 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
983 {
984 	struct request *rq = scsi_cmd_to_rq(cmd);
985 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
986 
987 	/* flush requests don't perform I/O, zero the S/G table */
988 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
989 
990 	if (cmd->device->use_16_for_sync) {
991 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
992 		cmd->cmd_len = 16;
993 	} else {
994 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
995 		cmd->cmd_len = 10;
996 	}
997 	cmd->transfersize = 0;
998 	cmd->allowed = sdkp->max_retries;
999 
1000 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1001 	return BLK_STS_OK;
1002 }
1003 
1004 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1005 				       sector_t lba, unsigned int nr_blocks,
1006 				       unsigned char flags, unsigned int dld)
1007 {
1008 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1009 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1010 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1011 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1012 	cmd->cmnd[10] = flags;
1013 	cmd->cmnd[11] = dld & 0x07;
1014 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1015 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1016 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1017 
1018 	return BLK_STS_OK;
1019 }
1020 
1021 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1022 				       sector_t lba, unsigned int nr_blocks,
1023 				       unsigned char flags, unsigned int dld)
1024 {
1025 	cmd->cmd_len  = 16;
1026 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1027 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1028 	cmd->cmnd[14] = (dld & 0x03) << 6;
1029 	cmd->cmnd[15] = 0;
1030 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1031 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1032 
1033 	return BLK_STS_OK;
1034 }
1035 
1036 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1037 				       sector_t lba, unsigned int nr_blocks,
1038 				       unsigned char flags)
1039 {
1040 	cmd->cmd_len = 10;
1041 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1042 	cmd->cmnd[1] = flags;
1043 	cmd->cmnd[6] = 0;
1044 	cmd->cmnd[9] = 0;
1045 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1046 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1047 
1048 	return BLK_STS_OK;
1049 }
1050 
1051 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1052 				      sector_t lba, unsigned int nr_blocks,
1053 				      unsigned char flags)
1054 {
1055 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1056 	if (WARN_ON_ONCE(nr_blocks == 0))
1057 		return BLK_STS_IOERR;
1058 
1059 	if (unlikely(flags & 0x8)) {
1060 		/*
1061 		 * This happens only if this drive failed 10byte rw
1062 		 * command with ILLEGAL_REQUEST during operation and
1063 		 * thus turned off use_10_for_rw.
1064 		 */
1065 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1066 		return BLK_STS_IOERR;
1067 	}
1068 
1069 	cmd->cmd_len = 6;
1070 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1071 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1072 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1073 	cmd->cmnd[3] = lba & 0xff;
1074 	cmd->cmnd[4] = nr_blocks;
1075 	cmd->cmnd[5] = 0;
1076 
1077 	return BLK_STS_OK;
1078 }
1079 
1080 /*
1081  * Check if a command has a duration limit set. If it does, and the target
1082  * device supports CDL and the feature is enabled, return the limit
1083  * descriptor index to use. Return 0 (no limit) otherwise.
1084  */
1085 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1086 {
1087 	struct scsi_device *sdp = sdkp->device;
1088 	int hint;
1089 
1090 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1091 		return 0;
1092 
1093 	/*
1094 	 * Use "no limit" if the request ioprio does not specify a duration
1095 	 * limit hint.
1096 	 */
1097 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1098 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1099 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1100 		return 0;
1101 
1102 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1103 }
1104 
1105 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1106 {
1107 	struct request *rq = scsi_cmd_to_rq(cmd);
1108 	struct scsi_device *sdp = cmd->device;
1109 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1110 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1111 	sector_t threshold;
1112 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1113 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1114 	bool write = rq_data_dir(rq) == WRITE;
1115 	unsigned char protect, fua;
1116 	unsigned int dld;
1117 	blk_status_t ret;
1118 	unsigned int dif;
1119 	bool dix;
1120 
1121 	ret = scsi_alloc_sgtables(cmd);
1122 	if (ret != BLK_STS_OK)
1123 		return ret;
1124 
1125 	ret = BLK_STS_IOERR;
1126 	if (!scsi_device_online(sdp) || sdp->changed) {
1127 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1128 		goto fail;
1129 	}
1130 
1131 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1132 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1133 		goto fail;
1134 	}
1135 
1136 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1137 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1138 		goto fail;
1139 	}
1140 
1141 	/*
1142 	 * Some SD card readers can't handle accesses which touch the
1143 	 * last one or two logical blocks. Split accesses as needed.
1144 	 */
1145 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1146 
1147 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1148 		if (lba < threshold) {
1149 			/* Access up to the threshold but not beyond */
1150 			nr_blocks = threshold - lba;
1151 		} else {
1152 			/* Access only a single logical block */
1153 			nr_blocks = 1;
1154 		}
1155 	}
1156 
1157 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1158 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1159 		if (ret)
1160 			goto fail;
1161 	}
1162 
1163 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1164 	dix = scsi_prot_sg_count(cmd);
1165 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1166 	dld = sd_cdl_dld(sdkp, cmd);
1167 
1168 	if (dif || dix)
1169 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1170 	else
1171 		protect = 0;
1172 
1173 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1174 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1175 					 protect | fua, dld);
1176 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1177 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1178 					 protect | fua, dld);
1179 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1180 		   sdp->use_10_for_rw || protect) {
1181 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1182 					 protect | fua);
1183 	} else {
1184 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1185 					protect | fua);
1186 	}
1187 
1188 	if (unlikely(ret != BLK_STS_OK))
1189 		goto fail;
1190 
1191 	/*
1192 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1193 	 * host adapter, it's safe to assume that we can at least transfer
1194 	 * this many bytes between each connect / disconnect.
1195 	 */
1196 	cmd->transfersize = sdp->sector_size;
1197 	cmd->underflow = nr_blocks << 9;
1198 	cmd->allowed = sdkp->max_retries;
1199 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1200 
1201 	SCSI_LOG_HLQUEUE(1,
1202 			 scmd_printk(KERN_INFO, cmd,
1203 				     "%s: block=%llu, count=%d\n", __func__,
1204 				     (unsigned long long)blk_rq_pos(rq),
1205 				     blk_rq_sectors(rq)));
1206 	SCSI_LOG_HLQUEUE(2,
1207 			 scmd_printk(KERN_INFO, cmd,
1208 				     "%s %d/%u 512 byte blocks.\n",
1209 				     write ? "writing" : "reading", nr_blocks,
1210 				     blk_rq_sectors(rq)));
1211 
1212 	/*
1213 	 * This indicates that the command is ready from our end to be queued.
1214 	 */
1215 	return BLK_STS_OK;
1216 fail:
1217 	scsi_free_sgtables(cmd);
1218 	return ret;
1219 }
1220 
1221 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1222 {
1223 	struct request *rq = scsi_cmd_to_rq(cmd);
1224 
1225 	switch (req_op(rq)) {
1226 	case REQ_OP_DISCARD:
1227 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1228 		case SD_LBP_UNMAP:
1229 			return sd_setup_unmap_cmnd(cmd);
1230 		case SD_LBP_WS16:
1231 			return sd_setup_write_same16_cmnd(cmd, true);
1232 		case SD_LBP_WS10:
1233 			return sd_setup_write_same10_cmnd(cmd, true);
1234 		case SD_LBP_ZERO:
1235 			return sd_setup_write_same10_cmnd(cmd, false);
1236 		default:
1237 			return BLK_STS_TARGET;
1238 		}
1239 	case REQ_OP_WRITE_ZEROES:
1240 		return sd_setup_write_zeroes_cmnd(cmd);
1241 	case REQ_OP_FLUSH:
1242 		return sd_setup_flush_cmnd(cmd);
1243 	case REQ_OP_READ:
1244 	case REQ_OP_WRITE:
1245 	case REQ_OP_ZONE_APPEND:
1246 		return sd_setup_read_write_cmnd(cmd);
1247 	case REQ_OP_ZONE_RESET:
1248 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1249 						   false);
1250 	case REQ_OP_ZONE_RESET_ALL:
1251 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1252 						   true);
1253 	case REQ_OP_ZONE_OPEN:
1254 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1255 	case REQ_OP_ZONE_CLOSE:
1256 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1257 	case REQ_OP_ZONE_FINISH:
1258 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1259 	default:
1260 		WARN_ON_ONCE(1);
1261 		return BLK_STS_NOTSUPP;
1262 	}
1263 }
1264 
1265 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1266 {
1267 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1268 
1269 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1270 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1271 }
1272 
1273 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1274 {
1275 	if (sdkp->device->removable || sdkp->write_prot) {
1276 		if (disk_check_media_change(disk))
1277 			return true;
1278 	}
1279 
1280 	/*
1281 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1282 	 * nothing to do with partitions, BLKRRPART is used to force a full
1283 	 * revalidate after things like a format for historical reasons.
1284 	 */
1285 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1286 }
1287 
1288 /**
1289  *	sd_open - open a scsi disk device
1290  *	@disk: disk to open
1291  *	@mode: open mode
1292  *
1293  *	Returns 0 if successful. Returns a negated errno value in case
1294  *	of error.
1295  *
1296  *	Note: This can be called from a user context (e.g. fsck(1) )
1297  *	or from within the kernel (e.g. as a result of a mount(1) ).
1298  *	In the latter case @inode and @filp carry an abridged amount
1299  *	of information as noted above.
1300  *
1301  *	Locking: called with disk->open_mutex held.
1302  **/
1303 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1304 {
1305 	struct scsi_disk *sdkp = scsi_disk(disk);
1306 	struct scsi_device *sdev = sdkp->device;
1307 	int retval;
1308 
1309 	if (scsi_device_get(sdev))
1310 		return -ENXIO;
1311 
1312 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1313 
1314 	/*
1315 	 * If the device is in error recovery, wait until it is done.
1316 	 * If the device is offline, then disallow any access to it.
1317 	 */
1318 	retval = -ENXIO;
1319 	if (!scsi_block_when_processing_errors(sdev))
1320 		goto error_out;
1321 
1322 	if (sd_need_revalidate(disk, sdkp))
1323 		sd_revalidate_disk(disk);
1324 
1325 	/*
1326 	 * If the drive is empty, just let the open fail.
1327 	 */
1328 	retval = -ENOMEDIUM;
1329 	if (sdev->removable && !sdkp->media_present &&
1330 	    !(mode & BLK_OPEN_NDELAY))
1331 		goto error_out;
1332 
1333 	/*
1334 	 * If the device has the write protect tab set, have the open fail
1335 	 * if the user expects to be able to write to the thing.
1336 	 */
1337 	retval = -EROFS;
1338 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1339 		goto error_out;
1340 
1341 	/*
1342 	 * It is possible that the disk changing stuff resulted in
1343 	 * the device being taken offline.  If this is the case,
1344 	 * report this to the user, and don't pretend that the
1345 	 * open actually succeeded.
1346 	 */
1347 	retval = -ENXIO;
1348 	if (!scsi_device_online(sdev))
1349 		goto error_out;
1350 
1351 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1352 		if (scsi_block_when_processing_errors(sdev))
1353 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1354 	}
1355 
1356 	return 0;
1357 
1358 error_out:
1359 	scsi_device_put(sdev);
1360 	return retval;
1361 }
1362 
1363 /**
1364  *	sd_release - invoked when the (last) close(2) is called on this
1365  *	scsi disk.
1366  *	@disk: disk to release
1367  *
1368  *	Returns 0.
1369  *
1370  *	Note: may block (uninterruptible) if error recovery is underway
1371  *	on this disk.
1372  *
1373  *	Locking: called with disk->open_mutex held.
1374  **/
1375 static void sd_release(struct gendisk *disk)
1376 {
1377 	struct scsi_disk *sdkp = scsi_disk(disk);
1378 	struct scsi_device *sdev = sdkp->device;
1379 
1380 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1381 
1382 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1383 		if (scsi_block_when_processing_errors(sdev))
1384 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1385 	}
1386 
1387 	scsi_device_put(sdev);
1388 }
1389 
1390 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1391 {
1392 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1393 	struct scsi_device *sdp = sdkp->device;
1394 	struct Scsi_Host *host = sdp->host;
1395 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1396 	int diskinfo[4];
1397 
1398 	/* default to most commonly used values */
1399 	diskinfo[0] = 0x40;	/* 1 << 6 */
1400 	diskinfo[1] = 0x20;	/* 1 << 5 */
1401 	diskinfo[2] = capacity >> 11;
1402 
1403 	/* override with calculated, extended default, or driver values */
1404 	if (host->hostt->bios_param)
1405 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1406 	else
1407 		scsicam_bios_param(bdev, capacity, diskinfo);
1408 
1409 	geo->heads = diskinfo[0];
1410 	geo->sectors = diskinfo[1];
1411 	geo->cylinders = diskinfo[2];
1412 	return 0;
1413 }
1414 
1415 /**
1416  *	sd_ioctl - process an ioctl
1417  *	@bdev: target block device
1418  *	@mode: open mode
1419  *	@cmd: ioctl command number
1420  *	@arg: this is third argument given to ioctl(2) system call.
1421  *	Often contains a pointer.
1422  *
1423  *	Returns 0 if successful (some ioctls return positive numbers on
1424  *	success as well). Returns a negated errno value in case of error.
1425  *
1426  *	Note: most ioctls are forward onto the block subsystem or further
1427  *	down in the scsi subsystem.
1428  **/
1429 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1430 		    unsigned int cmd, unsigned long arg)
1431 {
1432 	struct gendisk *disk = bdev->bd_disk;
1433 	struct scsi_disk *sdkp = scsi_disk(disk);
1434 	struct scsi_device *sdp = sdkp->device;
1435 	void __user *p = (void __user *)arg;
1436 	int error;
1437 
1438 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1439 				    "cmd=0x%x\n", disk->disk_name, cmd));
1440 
1441 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1442 		return -ENOIOCTLCMD;
1443 
1444 	/*
1445 	 * If we are in the middle of error recovery, don't let anyone
1446 	 * else try and use this device.  Also, if error recovery fails, it
1447 	 * may try and take the device offline, in which case all further
1448 	 * access to the device is prohibited.
1449 	 */
1450 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1451 			(mode & BLK_OPEN_NDELAY));
1452 	if (error)
1453 		return error;
1454 
1455 	if (is_sed_ioctl(cmd))
1456 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1457 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1458 }
1459 
1460 static void set_media_not_present(struct scsi_disk *sdkp)
1461 {
1462 	if (sdkp->media_present)
1463 		sdkp->device->changed = 1;
1464 
1465 	if (sdkp->device->removable) {
1466 		sdkp->media_present = 0;
1467 		sdkp->capacity = 0;
1468 	}
1469 }
1470 
1471 static int media_not_present(struct scsi_disk *sdkp,
1472 			     struct scsi_sense_hdr *sshdr)
1473 {
1474 	if (!scsi_sense_valid(sshdr))
1475 		return 0;
1476 
1477 	/* not invoked for commands that could return deferred errors */
1478 	switch (sshdr->sense_key) {
1479 	case UNIT_ATTENTION:
1480 	case NOT_READY:
1481 		/* medium not present */
1482 		if (sshdr->asc == 0x3A) {
1483 			set_media_not_present(sdkp);
1484 			return 1;
1485 		}
1486 	}
1487 	return 0;
1488 }
1489 
1490 /**
1491  *	sd_check_events - check media events
1492  *	@disk: kernel device descriptor
1493  *	@clearing: disk events currently being cleared
1494  *
1495  *	Returns mask of DISK_EVENT_*.
1496  *
1497  *	Note: this function is invoked from the block subsystem.
1498  **/
1499 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1500 {
1501 	struct scsi_disk *sdkp = disk->private_data;
1502 	struct scsi_device *sdp;
1503 	int retval;
1504 	bool disk_changed;
1505 
1506 	if (!sdkp)
1507 		return 0;
1508 
1509 	sdp = sdkp->device;
1510 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1511 
1512 	/*
1513 	 * If the device is offline, don't send any commands - just pretend as
1514 	 * if the command failed.  If the device ever comes back online, we
1515 	 * can deal with it then.  It is only because of unrecoverable errors
1516 	 * that we would ever take a device offline in the first place.
1517 	 */
1518 	if (!scsi_device_online(sdp)) {
1519 		set_media_not_present(sdkp);
1520 		goto out;
1521 	}
1522 
1523 	/*
1524 	 * Using TEST_UNIT_READY enables differentiation between drive with
1525 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1526 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1527 	 *
1528 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1529 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1530 	 * sd_revalidate() is called.
1531 	 */
1532 	if (scsi_block_when_processing_errors(sdp)) {
1533 		struct scsi_sense_hdr sshdr = { 0, };
1534 
1535 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1536 					      &sshdr);
1537 
1538 		/* failed to execute TUR, assume media not present */
1539 		if (retval < 0 || host_byte(retval)) {
1540 			set_media_not_present(sdkp);
1541 			goto out;
1542 		}
1543 
1544 		if (media_not_present(sdkp, &sshdr))
1545 			goto out;
1546 	}
1547 
1548 	/*
1549 	 * For removable scsi disk we have to recognise the presence
1550 	 * of a disk in the drive.
1551 	 */
1552 	if (!sdkp->media_present)
1553 		sdp->changed = 1;
1554 	sdkp->media_present = 1;
1555 out:
1556 	/*
1557 	 * sdp->changed is set under the following conditions:
1558 	 *
1559 	 *	Medium present state has changed in either direction.
1560 	 *	Device has indicated UNIT_ATTENTION.
1561 	 */
1562 	disk_changed = sdp->changed;
1563 	sdp->changed = 0;
1564 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1565 }
1566 
1567 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1568 {
1569 	int retries, res;
1570 	struct scsi_device *sdp = sdkp->device;
1571 	const int timeout = sdp->request_queue->rq_timeout
1572 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1573 	struct scsi_sense_hdr my_sshdr;
1574 	const struct scsi_exec_args exec_args = {
1575 		.req_flags = BLK_MQ_REQ_PM,
1576 		/* caller might not be interested in sense, but we need it */
1577 		.sshdr = sshdr ? : &my_sshdr,
1578 	};
1579 
1580 	if (!scsi_device_online(sdp))
1581 		return -ENODEV;
1582 
1583 	sshdr = exec_args.sshdr;
1584 
1585 	for (retries = 3; retries > 0; --retries) {
1586 		unsigned char cmd[16] = { 0 };
1587 
1588 		if (sdp->use_16_for_sync)
1589 			cmd[0] = SYNCHRONIZE_CACHE_16;
1590 		else
1591 			cmd[0] = SYNCHRONIZE_CACHE;
1592 		/*
1593 		 * Leave the rest of the command zero to indicate
1594 		 * flush everything.
1595 		 */
1596 		res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1597 				       timeout, sdkp->max_retries, &exec_args);
1598 		if (res == 0)
1599 			break;
1600 	}
1601 
1602 	if (res) {
1603 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1604 
1605 		if (res < 0)
1606 			return res;
1607 
1608 		if (scsi_status_is_check_condition(res) &&
1609 		    scsi_sense_valid(sshdr)) {
1610 			sd_print_sense_hdr(sdkp, sshdr);
1611 
1612 			/* we need to evaluate the error return  */
1613 			if (sshdr->asc == 0x3a ||	/* medium not present */
1614 			    sshdr->asc == 0x20 ||	/* invalid command */
1615 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1616 				/* this is no error here */
1617 				return 0;
1618 		}
1619 
1620 		switch (host_byte(res)) {
1621 		/* ignore errors due to racing a disconnection */
1622 		case DID_BAD_TARGET:
1623 		case DID_NO_CONNECT:
1624 			return 0;
1625 		/* signal the upper layer it might try again */
1626 		case DID_BUS_BUSY:
1627 		case DID_IMM_RETRY:
1628 		case DID_REQUEUE:
1629 		case DID_SOFT_ERROR:
1630 			return -EBUSY;
1631 		default:
1632 			return -EIO;
1633 		}
1634 	}
1635 	return 0;
1636 }
1637 
1638 static void sd_rescan(struct device *dev)
1639 {
1640 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1641 
1642 	sd_revalidate_disk(sdkp->disk);
1643 }
1644 
1645 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1646 		enum blk_unique_id type)
1647 {
1648 	struct scsi_device *sdev = scsi_disk(disk)->device;
1649 	const struct scsi_vpd *vpd;
1650 	const unsigned char *d;
1651 	int ret = -ENXIO, len;
1652 
1653 	rcu_read_lock();
1654 	vpd = rcu_dereference(sdev->vpd_pg83);
1655 	if (!vpd)
1656 		goto out_unlock;
1657 
1658 	ret = -EINVAL;
1659 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1660 		/* we only care about designators with LU association */
1661 		if (((d[1] >> 4) & 0x3) != 0x00)
1662 			continue;
1663 		if ((d[1] & 0xf) != type)
1664 			continue;
1665 
1666 		/*
1667 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1668 		 * keep looking as one with more entropy might still show up.
1669 		 */
1670 		len = d[3];
1671 		if (len != 8 && len != 12 && len != 16)
1672 			continue;
1673 		ret = len;
1674 		memcpy(id, d + 4, len);
1675 		if (len == 16)
1676 			break;
1677 	}
1678 out_unlock:
1679 	rcu_read_unlock();
1680 	return ret;
1681 }
1682 
1683 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1684 {
1685 	switch (host_byte(result)) {
1686 	case DID_TRANSPORT_MARGINAL:
1687 	case DID_TRANSPORT_DISRUPTED:
1688 	case DID_BUS_BUSY:
1689 		return PR_STS_RETRY_PATH_FAILURE;
1690 	case DID_NO_CONNECT:
1691 		return PR_STS_PATH_FAILED;
1692 	case DID_TRANSPORT_FAILFAST:
1693 		return PR_STS_PATH_FAST_FAILED;
1694 	}
1695 
1696 	switch (status_byte(result)) {
1697 	case SAM_STAT_RESERVATION_CONFLICT:
1698 		return PR_STS_RESERVATION_CONFLICT;
1699 	case SAM_STAT_CHECK_CONDITION:
1700 		if (!scsi_sense_valid(sshdr))
1701 			return PR_STS_IOERR;
1702 
1703 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1704 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1705 			return -EINVAL;
1706 
1707 		fallthrough;
1708 	default:
1709 		return PR_STS_IOERR;
1710 	}
1711 }
1712 
1713 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1714 			    unsigned char *data, int data_len)
1715 {
1716 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1717 	struct scsi_device *sdev = sdkp->device;
1718 	struct scsi_sense_hdr sshdr;
1719 	u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1720 	const struct scsi_exec_args exec_args = {
1721 		.sshdr = &sshdr,
1722 	};
1723 	int result;
1724 
1725 	put_unaligned_be16(data_len, &cmd[7]);
1726 
1727 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1728 				  SD_TIMEOUT, sdkp->max_retries, &exec_args);
1729 	if (scsi_status_is_check_condition(result) &&
1730 	    scsi_sense_valid(&sshdr)) {
1731 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1732 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1733 	}
1734 
1735 	if (result <= 0)
1736 		return result;
1737 
1738 	return sd_scsi_to_pr_err(&sshdr, result);
1739 }
1740 
1741 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1742 {
1743 	int result, i, data_offset, num_copy_keys;
1744 	u32 num_keys = keys_info->num_keys;
1745 	int data_len = num_keys * 8 + 8;
1746 	u8 *data;
1747 
1748 	data = kzalloc(data_len, GFP_KERNEL);
1749 	if (!data)
1750 		return -ENOMEM;
1751 
1752 	result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1753 	if (result)
1754 		goto free_data;
1755 
1756 	keys_info->generation = get_unaligned_be32(&data[0]);
1757 	keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1758 
1759 	data_offset = 8;
1760 	num_copy_keys = min(num_keys, keys_info->num_keys);
1761 
1762 	for (i = 0; i < num_copy_keys; i++) {
1763 		keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1764 		data_offset += 8;
1765 	}
1766 
1767 free_data:
1768 	kfree(data);
1769 	return result;
1770 }
1771 
1772 static int sd_pr_read_reservation(struct block_device *bdev,
1773 				  struct pr_held_reservation *rsv)
1774 {
1775 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1776 	struct scsi_device *sdev = sdkp->device;
1777 	u8 data[24] = { };
1778 	int result, len;
1779 
1780 	result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1781 	if (result)
1782 		return result;
1783 
1784 	len = get_unaligned_be32(&data[4]);
1785 	if (!len)
1786 		return 0;
1787 
1788 	/* Make sure we have at least the key and type */
1789 	if (len < 14) {
1790 		sdev_printk(KERN_INFO, sdev,
1791 			    "READ RESERVATION failed due to short return buffer of %d bytes\n",
1792 			    len);
1793 		return -EINVAL;
1794 	}
1795 
1796 	rsv->generation = get_unaligned_be32(&data[0]);
1797 	rsv->key = get_unaligned_be64(&data[8]);
1798 	rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1799 	return 0;
1800 }
1801 
1802 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1803 			     u64 sa_key, enum scsi_pr_type type, u8 flags)
1804 {
1805 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1806 	struct scsi_device *sdev = sdkp->device;
1807 	struct scsi_sense_hdr sshdr;
1808 	const struct scsi_exec_args exec_args = {
1809 		.sshdr = &sshdr,
1810 	};
1811 	int result;
1812 	u8 cmd[16] = { 0, };
1813 	u8 data[24] = { 0, };
1814 
1815 	cmd[0] = PERSISTENT_RESERVE_OUT;
1816 	cmd[1] = sa;
1817 	cmd[2] = type;
1818 	put_unaligned_be32(sizeof(data), &cmd[5]);
1819 
1820 	put_unaligned_be64(key, &data[0]);
1821 	put_unaligned_be64(sa_key, &data[8]);
1822 	data[20] = flags;
1823 
1824 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1825 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1826 				  &exec_args);
1827 
1828 	if (scsi_status_is_check_condition(result) &&
1829 	    scsi_sense_valid(&sshdr)) {
1830 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1831 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1832 	}
1833 
1834 	if (result <= 0)
1835 		return result;
1836 
1837 	return sd_scsi_to_pr_err(&sshdr, result);
1838 }
1839 
1840 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1841 		u32 flags)
1842 {
1843 	if (flags & ~PR_FL_IGNORE_KEY)
1844 		return -EOPNOTSUPP;
1845 	return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1846 			old_key, new_key, 0,
1847 			(1 << 0) /* APTPL */);
1848 }
1849 
1850 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1851 		u32 flags)
1852 {
1853 	if (flags)
1854 		return -EOPNOTSUPP;
1855 	return sd_pr_out_command(bdev, 0x01, key, 0,
1856 				 block_pr_type_to_scsi(type), 0);
1857 }
1858 
1859 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1860 {
1861 	return sd_pr_out_command(bdev, 0x02, key, 0,
1862 				 block_pr_type_to_scsi(type), 0);
1863 }
1864 
1865 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1866 		enum pr_type type, bool abort)
1867 {
1868 	return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1869 				 block_pr_type_to_scsi(type), 0);
1870 }
1871 
1872 static int sd_pr_clear(struct block_device *bdev, u64 key)
1873 {
1874 	return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
1875 }
1876 
1877 static const struct pr_ops sd_pr_ops = {
1878 	.pr_register	= sd_pr_register,
1879 	.pr_reserve	= sd_pr_reserve,
1880 	.pr_release	= sd_pr_release,
1881 	.pr_preempt	= sd_pr_preempt,
1882 	.pr_clear	= sd_pr_clear,
1883 	.pr_read_keys	= sd_pr_read_keys,
1884 	.pr_read_reservation = sd_pr_read_reservation,
1885 };
1886 
1887 static void scsi_disk_free_disk(struct gendisk *disk)
1888 {
1889 	struct scsi_disk *sdkp = scsi_disk(disk);
1890 
1891 	put_device(&sdkp->disk_dev);
1892 }
1893 
1894 static const struct block_device_operations sd_fops = {
1895 	.owner			= THIS_MODULE,
1896 	.open			= sd_open,
1897 	.release		= sd_release,
1898 	.ioctl			= sd_ioctl,
1899 	.getgeo			= sd_getgeo,
1900 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1901 	.check_events		= sd_check_events,
1902 	.unlock_native_capacity	= sd_unlock_native_capacity,
1903 	.report_zones		= sd_zbc_report_zones,
1904 	.get_unique_id		= sd_get_unique_id,
1905 	.free_disk		= scsi_disk_free_disk,
1906 	.pr_ops			= &sd_pr_ops,
1907 };
1908 
1909 /**
1910  *	sd_eh_reset - reset error handling callback
1911  *	@scmd:		sd-issued command that has failed
1912  *
1913  *	This function is called by the SCSI midlayer before starting
1914  *	SCSI EH. When counting medium access failures we have to be
1915  *	careful to register it only only once per device and SCSI EH run;
1916  *	there might be several timed out commands which will cause the
1917  *	'max_medium_access_timeouts' counter to trigger after the first
1918  *	SCSI EH run already and set the device to offline.
1919  *	So this function resets the internal counter before starting SCSI EH.
1920  **/
1921 static void sd_eh_reset(struct scsi_cmnd *scmd)
1922 {
1923 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1924 
1925 	/* New SCSI EH run, reset gate variable */
1926 	sdkp->ignore_medium_access_errors = false;
1927 }
1928 
1929 /**
1930  *	sd_eh_action - error handling callback
1931  *	@scmd:		sd-issued command that has failed
1932  *	@eh_disp:	The recovery disposition suggested by the midlayer
1933  *
1934  *	This function is called by the SCSI midlayer upon completion of an
1935  *	error test command (currently TEST UNIT READY). The result of sending
1936  *	the eh command is passed in eh_disp.  We're looking for devices that
1937  *	fail medium access commands but are OK with non access commands like
1938  *	test unit ready (so wrongly see the device as having a successful
1939  *	recovery)
1940  **/
1941 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1942 {
1943 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1944 	struct scsi_device *sdev = scmd->device;
1945 
1946 	if (!scsi_device_online(sdev) ||
1947 	    !scsi_medium_access_command(scmd) ||
1948 	    host_byte(scmd->result) != DID_TIME_OUT ||
1949 	    eh_disp != SUCCESS)
1950 		return eh_disp;
1951 
1952 	/*
1953 	 * The device has timed out executing a medium access command.
1954 	 * However, the TEST UNIT READY command sent during error
1955 	 * handling completed successfully. Either the device is in the
1956 	 * process of recovering or has it suffered an internal failure
1957 	 * that prevents access to the storage medium.
1958 	 */
1959 	if (!sdkp->ignore_medium_access_errors) {
1960 		sdkp->medium_access_timed_out++;
1961 		sdkp->ignore_medium_access_errors = true;
1962 	}
1963 
1964 	/*
1965 	 * If the device keeps failing read/write commands but TEST UNIT
1966 	 * READY always completes successfully we assume that medium
1967 	 * access is no longer possible and take the device offline.
1968 	 */
1969 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1970 		scmd_printk(KERN_ERR, scmd,
1971 			    "Medium access timeout failure. Offlining disk!\n");
1972 		mutex_lock(&sdev->state_mutex);
1973 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1974 		mutex_unlock(&sdev->state_mutex);
1975 
1976 		return SUCCESS;
1977 	}
1978 
1979 	return eh_disp;
1980 }
1981 
1982 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1983 {
1984 	struct request *req = scsi_cmd_to_rq(scmd);
1985 	struct scsi_device *sdev = scmd->device;
1986 	unsigned int transferred, good_bytes;
1987 	u64 start_lba, end_lba, bad_lba;
1988 
1989 	/*
1990 	 * Some commands have a payload smaller than the device logical
1991 	 * block size (e.g. INQUIRY on a 4K disk).
1992 	 */
1993 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1994 		return 0;
1995 
1996 	/* Check if we have a 'bad_lba' information */
1997 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1998 				     SCSI_SENSE_BUFFERSIZE,
1999 				     &bad_lba))
2000 		return 0;
2001 
2002 	/*
2003 	 * If the bad lba was reported incorrectly, we have no idea where
2004 	 * the error is.
2005 	 */
2006 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2007 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2008 	if (bad_lba < start_lba || bad_lba >= end_lba)
2009 		return 0;
2010 
2011 	/*
2012 	 * resid is optional but mostly filled in.  When it's unused,
2013 	 * its value is zero, so we assume the whole buffer transferred
2014 	 */
2015 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2016 
2017 	/* This computation should always be done in terms of the
2018 	 * resolution of the device's medium.
2019 	 */
2020 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2021 
2022 	return min(good_bytes, transferred);
2023 }
2024 
2025 /**
2026  *	sd_done - bottom half handler: called when the lower level
2027  *	driver has completed (successfully or otherwise) a scsi command.
2028  *	@SCpnt: mid-level's per command structure.
2029  *
2030  *	Note: potentially run from within an ISR. Must not block.
2031  **/
2032 static int sd_done(struct scsi_cmnd *SCpnt)
2033 {
2034 	int result = SCpnt->result;
2035 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2036 	unsigned int sector_size = SCpnt->device->sector_size;
2037 	unsigned int resid;
2038 	struct scsi_sense_hdr sshdr;
2039 	struct request *req = scsi_cmd_to_rq(SCpnt);
2040 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2041 	int sense_valid = 0;
2042 	int sense_deferred = 0;
2043 
2044 	switch (req_op(req)) {
2045 	case REQ_OP_DISCARD:
2046 	case REQ_OP_WRITE_ZEROES:
2047 	case REQ_OP_ZONE_RESET:
2048 	case REQ_OP_ZONE_RESET_ALL:
2049 	case REQ_OP_ZONE_OPEN:
2050 	case REQ_OP_ZONE_CLOSE:
2051 	case REQ_OP_ZONE_FINISH:
2052 		if (!result) {
2053 			good_bytes = blk_rq_bytes(req);
2054 			scsi_set_resid(SCpnt, 0);
2055 		} else {
2056 			good_bytes = 0;
2057 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2058 		}
2059 		break;
2060 	default:
2061 		/*
2062 		 * In case of bogus fw or device, we could end up having
2063 		 * an unaligned partial completion. Check this here and force
2064 		 * alignment.
2065 		 */
2066 		resid = scsi_get_resid(SCpnt);
2067 		if (resid & (sector_size - 1)) {
2068 			sd_printk(KERN_INFO, sdkp,
2069 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2070 				resid, sector_size);
2071 			scsi_print_command(SCpnt);
2072 			resid = min(scsi_bufflen(SCpnt),
2073 				    round_up(resid, sector_size));
2074 			scsi_set_resid(SCpnt, resid);
2075 		}
2076 	}
2077 
2078 	if (result) {
2079 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2080 		if (sense_valid)
2081 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2082 	}
2083 	sdkp->medium_access_timed_out = 0;
2084 
2085 	if (!scsi_status_is_check_condition(result) &&
2086 	    (!sense_valid || sense_deferred))
2087 		goto out;
2088 
2089 	switch (sshdr.sense_key) {
2090 	case HARDWARE_ERROR:
2091 	case MEDIUM_ERROR:
2092 		good_bytes = sd_completed_bytes(SCpnt);
2093 		break;
2094 	case RECOVERED_ERROR:
2095 		good_bytes = scsi_bufflen(SCpnt);
2096 		break;
2097 	case NO_SENSE:
2098 		/* This indicates a false check condition, so ignore it.  An
2099 		 * unknown amount of data was transferred so treat it as an
2100 		 * error.
2101 		 */
2102 		SCpnt->result = 0;
2103 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2104 		break;
2105 	case ABORTED_COMMAND:
2106 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2107 			good_bytes = sd_completed_bytes(SCpnt);
2108 		break;
2109 	case ILLEGAL_REQUEST:
2110 		switch (sshdr.asc) {
2111 		case 0x10:	/* DIX: Host detected corruption */
2112 			good_bytes = sd_completed_bytes(SCpnt);
2113 			break;
2114 		case 0x20:	/* INVALID COMMAND OPCODE */
2115 		case 0x24:	/* INVALID FIELD IN CDB */
2116 			switch (SCpnt->cmnd[0]) {
2117 			case UNMAP:
2118 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2119 				break;
2120 			case WRITE_SAME_16:
2121 			case WRITE_SAME:
2122 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2123 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2124 				} else {
2125 					sdkp->device->no_write_same = 1;
2126 					sd_config_write_same(sdkp);
2127 					req->rq_flags |= RQF_QUIET;
2128 				}
2129 				break;
2130 			}
2131 		}
2132 		break;
2133 	default:
2134 		break;
2135 	}
2136 
2137  out:
2138 	if (sd_is_zoned(sdkp))
2139 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2140 
2141 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2142 					   "sd_done: completed %d of %d bytes\n",
2143 					   good_bytes, scsi_bufflen(SCpnt)));
2144 
2145 	return good_bytes;
2146 }
2147 
2148 /*
2149  * spinup disk - called only in sd_revalidate_disk()
2150  */
2151 static void
2152 sd_spinup_disk(struct scsi_disk *sdkp)
2153 {
2154 	unsigned char cmd[10];
2155 	unsigned long spintime_expire = 0;
2156 	int retries, spintime;
2157 	unsigned int the_result;
2158 	struct scsi_sense_hdr sshdr;
2159 	const struct scsi_exec_args exec_args = {
2160 		.sshdr = &sshdr,
2161 	};
2162 	int sense_valid = 0;
2163 
2164 	spintime = 0;
2165 
2166 	/* Spin up drives, as required.  Only do this at boot time */
2167 	/* Spinup needs to be done for module loads too. */
2168 	do {
2169 		retries = 0;
2170 
2171 		do {
2172 			bool media_was_present = sdkp->media_present;
2173 
2174 			cmd[0] = TEST_UNIT_READY;
2175 			memset((void *) &cmd[1], 0, 9);
2176 
2177 			the_result = scsi_execute_cmd(sdkp->device, cmd,
2178 						      REQ_OP_DRV_IN, NULL, 0,
2179 						      SD_TIMEOUT,
2180 						      sdkp->max_retries,
2181 						      &exec_args);
2182 
2183 			/*
2184 			 * If the drive has indicated to us that it
2185 			 * doesn't have any media in it, don't bother
2186 			 * with any more polling.
2187 			 */
2188 			if (media_not_present(sdkp, &sshdr)) {
2189 				if (media_was_present)
2190 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2191 				return;
2192 			}
2193 
2194 			if (the_result)
2195 				sense_valid = scsi_sense_valid(&sshdr);
2196 			retries++;
2197 		} while (retries < 3 &&
2198 			 (!scsi_status_is_good(the_result) ||
2199 			  (scsi_status_is_check_condition(the_result) &&
2200 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2201 
2202 		if (!scsi_status_is_check_condition(the_result)) {
2203 			/* no sense, TUR either succeeded or failed
2204 			 * with a status error */
2205 			if(!spintime && !scsi_status_is_good(the_result)) {
2206 				sd_print_result(sdkp, "Test Unit Ready failed",
2207 						the_result);
2208 			}
2209 			break;
2210 		}
2211 
2212 		/*
2213 		 * The device does not want the automatic start to be issued.
2214 		 */
2215 		if (sdkp->device->no_start_on_add)
2216 			break;
2217 
2218 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2219 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2220 				break;	/* manual intervention required */
2221 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2222 				break;	/* standby */
2223 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2224 				break;	/* unavailable */
2225 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2226 				break;	/* sanitize in progress */
2227 			/*
2228 			 * Issue command to spin up drive when not ready
2229 			 */
2230 			if (!spintime) {
2231 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2232 				cmd[0] = START_STOP;
2233 				cmd[1] = 1;	/* Return immediately */
2234 				memset((void *) &cmd[2], 0, 8);
2235 				cmd[4] = 1;	/* Start spin cycle */
2236 				if (sdkp->device->start_stop_pwr_cond)
2237 					cmd[4] |= 1 << 4;
2238 				scsi_execute_cmd(sdkp->device, cmd,
2239 						 REQ_OP_DRV_IN, NULL, 0,
2240 						 SD_TIMEOUT, sdkp->max_retries,
2241 						 &exec_args);
2242 				spintime_expire = jiffies + 100 * HZ;
2243 				spintime = 1;
2244 			}
2245 			/* Wait 1 second for next try */
2246 			msleep(1000);
2247 			printk(KERN_CONT ".");
2248 
2249 		/*
2250 		 * Wait for USB flash devices with slow firmware.
2251 		 * Yes, this sense key/ASC combination shouldn't
2252 		 * occur here.  It's characteristic of these devices.
2253 		 */
2254 		} else if (sense_valid &&
2255 				sshdr.sense_key == UNIT_ATTENTION &&
2256 				sshdr.asc == 0x28) {
2257 			if (!spintime) {
2258 				spintime_expire = jiffies + 5 * HZ;
2259 				spintime = 1;
2260 			}
2261 			/* Wait 1 second for next try */
2262 			msleep(1000);
2263 		} else {
2264 			/* we don't understand the sense code, so it's
2265 			 * probably pointless to loop */
2266 			if(!spintime) {
2267 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2268 				sd_print_sense_hdr(sdkp, &sshdr);
2269 			}
2270 			break;
2271 		}
2272 
2273 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2274 
2275 	if (spintime) {
2276 		if (scsi_status_is_good(the_result))
2277 			printk(KERN_CONT "ready\n");
2278 		else
2279 			printk(KERN_CONT "not responding...\n");
2280 	}
2281 }
2282 
2283 /*
2284  * Determine whether disk supports Data Integrity Field.
2285  */
2286 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2287 {
2288 	struct scsi_device *sdp = sdkp->device;
2289 	u8 type;
2290 
2291 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2292 		sdkp->protection_type = 0;
2293 		return 0;
2294 	}
2295 
2296 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2297 
2298 	if (type > T10_PI_TYPE3_PROTECTION) {
2299 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2300 			  " protection type %u. Disabling disk!\n",
2301 			  type);
2302 		sdkp->protection_type = 0;
2303 		return -ENODEV;
2304 	}
2305 
2306 	sdkp->protection_type = type;
2307 
2308 	return 0;
2309 }
2310 
2311 static void sd_config_protection(struct scsi_disk *sdkp)
2312 {
2313 	struct scsi_device *sdp = sdkp->device;
2314 
2315 	sd_dif_config_host(sdkp);
2316 
2317 	if (!sdkp->protection_type)
2318 		return;
2319 
2320 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2321 		sd_first_printk(KERN_NOTICE, sdkp,
2322 				"Disabling DIF Type %u protection\n",
2323 				sdkp->protection_type);
2324 		sdkp->protection_type = 0;
2325 	}
2326 
2327 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2328 			sdkp->protection_type);
2329 }
2330 
2331 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2332 			struct scsi_sense_hdr *sshdr, int sense_valid,
2333 			int the_result)
2334 {
2335 	if (sense_valid)
2336 		sd_print_sense_hdr(sdkp, sshdr);
2337 	else
2338 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2339 
2340 	/*
2341 	 * Set dirty bit for removable devices if not ready -
2342 	 * sometimes drives will not report this properly.
2343 	 */
2344 	if (sdp->removable &&
2345 	    sense_valid && sshdr->sense_key == NOT_READY)
2346 		set_media_not_present(sdkp);
2347 
2348 	/*
2349 	 * We used to set media_present to 0 here to indicate no media
2350 	 * in the drive, but some drives fail read capacity even with
2351 	 * media present, so we can't do that.
2352 	 */
2353 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2354 }
2355 
2356 #define RC16_LEN 32
2357 #if RC16_LEN > SD_BUF_SIZE
2358 #error RC16_LEN must not be more than SD_BUF_SIZE
2359 #endif
2360 
2361 #define READ_CAPACITY_RETRIES_ON_RESET	10
2362 
2363 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2364 						unsigned char *buffer)
2365 {
2366 	unsigned char cmd[16];
2367 	struct scsi_sense_hdr sshdr;
2368 	const struct scsi_exec_args exec_args = {
2369 		.sshdr = &sshdr,
2370 	};
2371 	int sense_valid = 0;
2372 	int the_result;
2373 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2374 	unsigned int alignment;
2375 	unsigned long long lba;
2376 	unsigned sector_size;
2377 
2378 	if (sdp->no_read_capacity_16)
2379 		return -EINVAL;
2380 
2381 	do {
2382 		memset(cmd, 0, 16);
2383 		cmd[0] = SERVICE_ACTION_IN_16;
2384 		cmd[1] = SAI_READ_CAPACITY_16;
2385 		cmd[13] = RC16_LEN;
2386 		memset(buffer, 0, RC16_LEN);
2387 
2388 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2389 					      buffer, RC16_LEN, SD_TIMEOUT,
2390 					      sdkp->max_retries, &exec_args);
2391 
2392 		if (media_not_present(sdkp, &sshdr))
2393 			return -ENODEV;
2394 
2395 		if (the_result > 0) {
2396 			sense_valid = scsi_sense_valid(&sshdr);
2397 			if (sense_valid &&
2398 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2399 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2400 			    sshdr.ascq == 0x00)
2401 				/* Invalid Command Operation Code or
2402 				 * Invalid Field in CDB, just retry
2403 				 * silently with RC10 */
2404 				return -EINVAL;
2405 			if (sense_valid &&
2406 			    sshdr.sense_key == UNIT_ATTENTION &&
2407 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2408 				/* Device reset might occur several times,
2409 				 * give it one more chance */
2410 				if (--reset_retries > 0)
2411 					continue;
2412 		}
2413 		retries--;
2414 
2415 	} while (the_result && retries);
2416 
2417 	if (the_result) {
2418 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2419 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2420 		return -EINVAL;
2421 	}
2422 
2423 	sector_size = get_unaligned_be32(&buffer[8]);
2424 	lba = get_unaligned_be64(&buffer[0]);
2425 
2426 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2427 		sdkp->capacity = 0;
2428 		return -ENODEV;
2429 	}
2430 
2431 	/* Logical blocks per physical block exponent */
2432 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2433 
2434 	/* RC basis */
2435 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2436 
2437 	/* Lowest aligned logical block */
2438 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2439 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2440 	if (alignment && sdkp->first_scan)
2441 		sd_printk(KERN_NOTICE, sdkp,
2442 			  "physical block alignment offset: %u\n", alignment);
2443 
2444 	if (buffer[14] & 0x80) { /* LBPME */
2445 		sdkp->lbpme = 1;
2446 
2447 		if (buffer[14] & 0x40) /* LBPRZ */
2448 			sdkp->lbprz = 1;
2449 
2450 		sd_config_discard(sdkp, SD_LBP_WS16);
2451 	}
2452 
2453 	sdkp->capacity = lba + 1;
2454 	return sector_size;
2455 }
2456 
2457 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2458 						unsigned char *buffer)
2459 {
2460 	unsigned char cmd[16];
2461 	struct scsi_sense_hdr sshdr;
2462 	const struct scsi_exec_args exec_args = {
2463 		.sshdr = &sshdr,
2464 	};
2465 	int sense_valid = 0;
2466 	int the_result;
2467 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2468 	sector_t lba;
2469 	unsigned sector_size;
2470 
2471 	do {
2472 		cmd[0] = READ_CAPACITY;
2473 		memset(&cmd[1], 0, 9);
2474 		memset(buffer, 0, 8);
2475 
2476 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2477 					      8, SD_TIMEOUT, sdkp->max_retries,
2478 					      &exec_args);
2479 
2480 		if (media_not_present(sdkp, &sshdr))
2481 			return -ENODEV;
2482 
2483 		if (the_result > 0) {
2484 			sense_valid = scsi_sense_valid(&sshdr);
2485 			if (sense_valid &&
2486 			    sshdr.sense_key == UNIT_ATTENTION &&
2487 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2488 				/* Device reset might occur several times,
2489 				 * give it one more chance */
2490 				if (--reset_retries > 0)
2491 					continue;
2492 		}
2493 		retries--;
2494 
2495 	} while (the_result && retries);
2496 
2497 	if (the_result) {
2498 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2499 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2500 		return -EINVAL;
2501 	}
2502 
2503 	sector_size = get_unaligned_be32(&buffer[4]);
2504 	lba = get_unaligned_be32(&buffer[0]);
2505 
2506 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2507 		/* Some buggy (usb cardreader) devices return an lba of
2508 		   0xffffffff when the want to report a size of 0 (with
2509 		   which they really mean no media is present) */
2510 		sdkp->capacity = 0;
2511 		sdkp->physical_block_size = sector_size;
2512 		return sector_size;
2513 	}
2514 
2515 	sdkp->capacity = lba + 1;
2516 	sdkp->physical_block_size = sector_size;
2517 	return sector_size;
2518 }
2519 
2520 static int sd_try_rc16_first(struct scsi_device *sdp)
2521 {
2522 	if (sdp->host->max_cmd_len < 16)
2523 		return 0;
2524 	if (sdp->try_rc_10_first)
2525 		return 0;
2526 	if (sdp->scsi_level > SCSI_SPC_2)
2527 		return 1;
2528 	if (scsi_device_protection(sdp))
2529 		return 1;
2530 	return 0;
2531 }
2532 
2533 /*
2534  * read disk capacity
2535  */
2536 static void
2537 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2538 {
2539 	int sector_size;
2540 	struct scsi_device *sdp = sdkp->device;
2541 
2542 	if (sd_try_rc16_first(sdp)) {
2543 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2544 		if (sector_size == -EOVERFLOW)
2545 			goto got_data;
2546 		if (sector_size == -ENODEV)
2547 			return;
2548 		if (sector_size < 0)
2549 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2550 		if (sector_size < 0)
2551 			return;
2552 	} else {
2553 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2554 		if (sector_size == -EOVERFLOW)
2555 			goto got_data;
2556 		if (sector_size < 0)
2557 			return;
2558 		if ((sizeof(sdkp->capacity) > 4) &&
2559 		    (sdkp->capacity > 0xffffffffULL)) {
2560 			int old_sector_size = sector_size;
2561 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2562 					"Trying to use READ CAPACITY(16).\n");
2563 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2564 			if (sector_size < 0) {
2565 				sd_printk(KERN_NOTICE, sdkp,
2566 					"Using 0xffffffff as device size\n");
2567 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2568 				sector_size = old_sector_size;
2569 				goto got_data;
2570 			}
2571 			/* Remember that READ CAPACITY(16) succeeded */
2572 			sdp->try_rc_10_first = 0;
2573 		}
2574 	}
2575 
2576 	/* Some devices are known to return the total number of blocks,
2577 	 * not the highest block number.  Some devices have versions
2578 	 * which do this and others which do not.  Some devices we might
2579 	 * suspect of doing this but we don't know for certain.
2580 	 *
2581 	 * If we know the reported capacity is wrong, decrement it.  If
2582 	 * we can only guess, then assume the number of blocks is even
2583 	 * (usually true but not always) and err on the side of lowering
2584 	 * the capacity.
2585 	 */
2586 	if (sdp->fix_capacity ||
2587 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2588 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2589 				"from its reported value: %llu\n",
2590 				(unsigned long long) sdkp->capacity);
2591 		--sdkp->capacity;
2592 	}
2593 
2594 got_data:
2595 	if (sector_size == 0) {
2596 		sector_size = 512;
2597 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2598 			  "assuming 512.\n");
2599 	}
2600 
2601 	if (sector_size != 512 &&
2602 	    sector_size != 1024 &&
2603 	    sector_size != 2048 &&
2604 	    sector_size != 4096) {
2605 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2606 			  sector_size);
2607 		/*
2608 		 * The user might want to re-format the drive with
2609 		 * a supported sectorsize.  Once this happens, it
2610 		 * would be relatively trivial to set the thing up.
2611 		 * For this reason, we leave the thing in the table.
2612 		 */
2613 		sdkp->capacity = 0;
2614 		/*
2615 		 * set a bogus sector size so the normal read/write
2616 		 * logic in the block layer will eventually refuse any
2617 		 * request on this device without tripping over power
2618 		 * of two sector size assumptions
2619 		 */
2620 		sector_size = 512;
2621 	}
2622 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2623 	blk_queue_physical_block_size(sdp->request_queue,
2624 				      sdkp->physical_block_size);
2625 	sdkp->device->sector_size = sector_size;
2626 
2627 	if (sdkp->capacity > 0xffffffff)
2628 		sdp->use_16_for_rw = 1;
2629 
2630 }
2631 
2632 /*
2633  * Print disk capacity
2634  */
2635 static void
2636 sd_print_capacity(struct scsi_disk *sdkp,
2637 		  sector_t old_capacity)
2638 {
2639 	int sector_size = sdkp->device->sector_size;
2640 	char cap_str_2[10], cap_str_10[10];
2641 
2642 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2643 		return;
2644 
2645 	string_get_size(sdkp->capacity, sector_size,
2646 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2647 	string_get_size(sdkp->capacity, sector_size,
2648 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2649 
2650 	sd_printk(KERN_NOTICE, sdkp,
2651 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2652 		  (unsigned long long)sdkp->capacity,
2653 		  sector_size, cap_str_10, cap_str_2);
2654 
2655 	if (sdkp->physical_block_size != sector_size)
2656 		sd_printk(KERN_NOTICE, sdkp,
2657 			  "%u-byte physical blocks\n",
2658 			  sdkp->physical_block_size);
2659 }
2660 
2661 /* called with buffer of length 512 */
2662 static inline int
2663 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2664 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2665 		 struct scsi_sense_hdr *sshdr)
2666 {
2667 	/*
2668 	 * If we must use MODE SENSE(10), make sure that the buffer length
2669 	 * is at least 8 bytes so that the mode sense header fits.
2670 	 */
2671 	if (sdkp->device->use_10_for_ms && len < 8)
2672 		len = 8;
2673 
2674 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2675 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2676 }
2677 
2678 /*
2679  * read write protect setting, if possible - called only in sd_revalidate_disk()
2680  * called with buffer of length SD_BUF_SIZE
2681  */
2682 static void
2683 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2684 {
2685 	int res;
2686 	struct scsi_device *sdp = sdkp->device;
2687 	struct scsi_mode_data data;
2688 	int old_wp = sdkp->write_prot;
2689 
2690 	set_disk_ro(sdkp->disk, 0);
2691 	if (sdp->skip_ms_page_3f) {
2692 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2693 		return;
2694 	}
2695 
2696 	if (sdp->use_192_bytes_for_3f) {
2697 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2698 	} else {
2699 		/*
2700 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2701 		 * We have to start carefully: some devices hang if we ask
2702 		 * for more than is available.
2703 		 */
2704 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2705 
2706 		/*
2707 		 * Second attempt: ask for page 0 When only page 0 is
2708 		 * implemented, a request for page 3F may return Sense Key
2709 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2710 		 * CDB.
2711 		 */
2712 		if (res < 0)
2713 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2714 
2715 		/*
2716 		 * Third attempt: ask 255 bytes, as we did earlier.
2717 		 */
2718 		if (res < 0)
2719 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2720 					       &data, NULL);
2721 	}
2722 
2723 	if (res < 0) {
2724 		sd_first_printk(KERN_WARNING, sdkp,
2725 			  "Test WP failed, assume Write Enabled\n");
2726 	} else {
2727 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2728 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2729 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2730 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2731 				  sdkp->write_prot ? "on" : "off");
2732 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2733 		}
2734 	}
2735 }
2736 
2737 /*
2738  * sd_read_cache_type - called only from sd_revalidate_disk()
2739  * called with buffer of length SD_BUF_SIZE
2740  */
2741 static void
2742 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2743 {
2744 	int len = 0, res;
2745 	struct scsi_device *sdp = sdkp->device;
2746 
2747 	int dbd;
2748 	int modepage;
2749 	int first_len;
2750 	struct scsi_mode_data data;
2751 	struct scsi_sense_hdr sshdr;
2752 	int old_wce = sdkp->WCE;
2753 	int old_rcd = sdkp->RCD;
2754 	int old_dpofua = sdkp->DPOFUA;
2755 
2756 
2757 	if (sdkp->cache_override)
2758 		return;
2759 
2760 	first_len = 4;
2761 	if (sdp->skip_ms_page_8) {
2762 		if (sdp->type == TYPE_RBC)
2763 			goto defaults;
2764 		else {
2765 			if (sdp->skip_ms_page_3f)
2766 				goto defaults;
2767 			modepage = 0x3F;
2768 			if (sdp->use_192_bytes_for_3f)
2769 				first_len = 192;
2770 			dbd = 0;
2771 		}
2772 	} else if (sdp->type == TYPE_RBC) {
2773 		modepage = 6;
2774 		dbd = 8;
2775 	} else {
2776 		modepage = 8;
2777 		dbd = 0;
2778 	}
2779 
2780 	/* cautiously ask */
2781 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2782 			&data, &sshdr);
2783 
2784 	if (res < 0)
2785 		goto bad_sense;
2786 
2787 	if (!data.header_length) {
2788 		modepage = 6;
2789 		first_len = 0;
2790 		sd_first_printk(KERN_ERR, sdkp,
2791 				"Missing header in MODE_SENSE response\n");
2792 	}
2793 
2794 	/* that went OK, now ask for the proper length */
2795 	len = data.length;
2796 
2797 	/*
2798 	 * We're only interested in the first three bytes, actually.
2799 	 * But the data cache page is defined for the first 20.
2800 	 */
2801 	if (len < 3)
2802 		goto bad_sense;
2803 	else if (len > SD_BUF_SIZE) {
2804 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2805 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2806 		len = SD_BUF_SIZE;
2807 	}
2808 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2809 		len = 192;
2810 
2811 	/* Get the data */
2812 	if (len > first_len)
2813 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2814 				&data, &sshdr);
2815 
2816 	if (!res) {
2817 		int offset = data.header_length + data.block_descriptor_length;
2818 
2819 		while (offset < len) {
2820 			u8 page_code = buffer[offset] & 0x3F;
2821 			u8 spf       = buffer[offset] & 0x40;
2822 
2823 			if (page_code == 8 || page_code == 6) {
2824 				/* We're interested only in the first 3 bytes.
2825 				 */
2826 				if (len - offset <= 2) {
2827 					sd_first_printk(KERN_ERR, sdkp,
2828 						"Incomplete mode parameter "
2829 							"data\n");
2830 					goto defaults;
2831 				} else {
2832 					modepage = page_code;
2833 					goto Page_found;
2834 				}
2835 			} else {
2836 				/* Go to the next page */
2837 				if (spf && len - offset > 3)
2838 					offset += 4 + (buffer[offset+2] << 8) +
2839 						buffer[offset+3];
2840 				else if (!spf && len - offset > 1)
2841 					offset += 2 + buffer[offset+1];
2842 				else {
2843 					sd_first_printk(KERN_ERR, sdkp,
2844 							"Incomplete mode "
2845 							"parameter data\n");
2846 					goto defaults;
2847 				}
2848 			}
2849 		}
2850 
2851 		sd_first_printk(KERN_WARNING, sdkp,
2852 				"No Caching mode page found\n");
2853 		goto defaults;
2854 
2855 	Page_found:
2856 		if (modepage == 8) {
2857 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2858 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2859 		} else {
2860 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2861 			sdkp->RCD = 0;
2862 		}
2863 
2864 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2865 		if (sdp->broken_fua) {
2866 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2867 			sdkp->DPOFUA = 0;
2868 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2869 			   !sdkp->device->use_16_for_rw) {
2870 			sd_first_printk(KERN_NOTICE, sdkp,
2871 				  "Uses READ/WRITE(6), disabling FUA\n");
2872 			sdkp->DPOFUA = 0;
2873 		}
2874 
2875 		/* No cache flush allowed for write protected devices */
2876 		if (sdkp->WCE && sdkp->write_prot)
2877 			sdkp->WCE = 0;
2878 
2879 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2880 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2881 			sd_printk(KERN_NOTICE, sdkp,
2882 				  "Write cache: %s, read cache: %s, %s\n",
2883 				  sdkp->WCE ? "enabled" : "disabled",
2884 				  sdkp->RCD ? "disabled" : "enabled",
2885 				  sdkp->DPOFUA ? "supports DPO and FUA"
2886 				  : "doesn't support DPO or FUA");
2887 
2888 		return;
2889 	}
2890 
2891 bad_sense:
2892 	if (scsi_sense_valid(&sshdr) &&
2893 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2894 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2895 		/* Invalid field in CDB */
2896 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2897 	else
2898 		sd_first_printk(KERN_ERR, sdkp,
2899 				"Asking for cache data failed\n");
2900 
2901 defaults:
2902 	if (sdp->wce_default_on) {
2903 		sd_first_printk(KERN_NOTICE, sdkp,
2904 				"Assuming drive cache: write back\n");
2905 		sdkp->WCE = 1;
2906 	} else {
2907 		sd_first_printk(KERN_WARNING, sdkp,
2908 				"Assuming drive cache: write through\n");
2909 		sdkp->WCE = 0;
2910 	}
2911 	sdkp->RCD = 0;
2912 	sdkp->DPOFUA = 0;
2913 }
2914 
2915 /*
2916  * The ATO bit indicates whether the DIF application tag is available
2917  * for use by the operating system.
2918  */
2919 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2920 {
2921 	int res, offset;
2922 	struct scsi_device *sdp = sdkp->device;
2923 	struct scsi_mode_data data;
2924 	struct scsi_sense_hdr sshdr;
2925 
2926 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2927 		return;
2928 
2929 	if (sdkp->protection_type == 0)
2930 		return;
2931 
2932 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
2933 			      sdkp->max_retries, &data, &sshdr);
2934 
2935 	if (res < 0 || !data.header_length ||
2936 	    data.length < 6) {
2937 		sd_first_printk(KERN_WARNING, sdkp,
2938 			  "getting Control mode page failed, assume no ATO\n");
2939 
2940 		if (scsi_sense_valid(&sshdr))
2941 			sd_print_sense_hdr(sdkp, &sshdr);
2942 
2943 		return;
2944 	}
2945 
2946 	offset = data.header_length + data.block_descriptor_length;
2947 
2948 	if ((buffer[offset] & 0x3f) != 0x0a) {
2949 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2950 		return;
2951 	}
2952 
2953 	if ((buffer[offset + 5] & 0x80) == 0)
2954 		return;
2955 
2956 	sdkp->ATO = 1;
2957 
2958 	return;
2959 }
2960 
2961 /**
2962  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2963  * @sdkp: disk to query
2964  */
2965 static void sd_read_block_limits(struct scsi_disk *sdkp)
2966 {
2967 	struct scsi_vpd *vpd;
2968 
2969 	rcu_read_lock();
2970 
2971 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2972 	if (!vpd || vpd->len < 16)
2973 		goto out;
2974 
2975 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2976 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2977 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2978 
2979 	if (vpd->len >= 64) {
2980 		unsigned int lba_count, desc_count;
2981 
2982 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2983 
2984 		if (!sdkp->lbpme)
2985 			goto out;
2986 
2987 		lba_count = get_unaligned_be32(&vpd->data[20]);
2988 		desc_count = get_unaligned_be32(&vpd->data[24]);
2989 
2990 		if (lba_count && desc_count)
2991 			sdkp->max_unmap_blocks = lba_count;
2992 
2993 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2994 
2995 		if (vpd->data[32] & 0x80)
2996 			sdkp->unmap_alignment =
2997 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2998 
2999 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
3000 
3001 			if (sdkp->max_unmap_blocks)
3002 				sd_config_discard(sdkp, SD_LBP_UNMAP);
3003 			else
3004 				sd_config_discard(sdkp, SD_LBP_WS16);
3005 
3006 		} else {	/* LBP VPD page tells us what to use */
3007 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
3008 				sd_config_discard(sdkp, SD_LBP_UNMAP);
3009 			else if (sdkp->lbpws)
3010 				sd_config_discard(sdkp, SD_LBP_WS16);
3011 			else if (sdkp->lbpws10)
3012 				sd_config_discard(sdkp, SD_LBP_WS10);
3013 			else
3014 				sd_config_discard(sdkp, SD_LBP_DISABLE);
3015 		}
3016 	}
3017 
3018  out:
3019 	rcu_read_unlock();
3020 }
3021 
3022 /**
3023  * sd_read_block_characteristics - Query block dev. characteristics
3024  * @sdkp: disk to query
3025  */
3026 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
3027 {
3028 	struct request_queue *q = sdkp->disk->queue;
3029 	struct scsi_vpd *vpd;
3030 	u16 rot;
3031 	u8 zoned;
3032 
3033 	rcu_read_lock();
3034 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3035 
3036 	if (!vpd || vpd->len < 8) {
3037 		rcu_read_unlock();
3038 	        return;
3039 	}
3040 
3041 	rot = get_unaligned_be16(&vpd->data[4]);
3042 	zoned = (vpd->data[8] >> 4) & 3;
3043 	rcu_read_unlock();
3044 
3045 	if (rot == 1) {
3046 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3047 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3048 	}
3049 
3050 	if (sdkp->device->type == TYPE_ZBC) {
3051 		/*
3052 		 * Host-managed: Per ZBC and ZAC specifications, writes in
3053 		 * sequential write required zones of host-managed devices must
3054 		 * be aligned to the device physical block size.
3055 		 */
3056 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
3057 		blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3058 	} else {
3059 		sdkp->zoned = zoned;
3060 		if (sdkp->zoned == 1) {
3061 			/* Host-aware */
3062 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
3063 		} else {
3064 			/* Regular disk or drive managed disk */
3065 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3066 		}
3067 	}
3068 
3069 	if (!sdkp->first_scan)
3070 		return;
3071 
3072 	if (blk_queue_is_zoned(q)) {
3073 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3074 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3075 	} else {
3076 		if (sdkp->zoned == 1)
3077 			sd_printk(KERN_NOTICE, sdkp,
3078 				  "Host-aware SMR disk used as regular disk\n");
3079 		else if (sdkp->zoned == 2)
3080 			sd_printk(KERN_NOTICE, sdkp,
3081 				  "Drive-managed SMR disk\n");
3082 	}
3083 }
3084 
3085 /**
3086  * sd_read_block_provisioning - Query provisioning VPD page
3087  * @sdkp: disk to query
3088  */
3089 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3090 {
3091 	struct scsi_vpd *vpd;
3092 
3093 	if (sdkp->lbpme == 0)
3094 		return;
3095 
3096 	rcu_read_lock();
3097 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3098 
3099 	if (!vpd || vpd->len < 8) {
3100 		rcu_read_unlock();
3101 		return;
3102 	}
3103 
3104 	sdkp->lbpvpd	= 1;
3105 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3106 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3107 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3108 	rcu_read_unlock();
3109 }
3110 
3111 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3112 {
3113 	struct scsi_device *sdev = sdkp->device;
3114 
3115 	if (sdev->host->no_write_same) {
3116 		sdev->no_write_same = 1;
3117 
3118 		return;
3119 	}
3120 
3121 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3122 		struct scsi_vpd *vpd;
3123 
3124 		sdev->no_report_opcodes = 1;
3125 
3126 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3127 		 * CODES is unsupported and the device has an ATA
3128 		 * Information VPD page (SAT).
3129 		 */
3130 		rcu_read_lock();
3131 		vpd = rcu_dereference(sdev->vpd_pg89);
3132 		if (vpd)
3133 			sdev->no_write_same = 1;
3134 		rcu_read_unlock();
3135 	}
3136 
3137 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3138 		sdkp->ws16 = 1;
3139 
3140 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3141 		sdkp->ws10 = 1;
3142 }
3143 
3144 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3145 {
3146 	struct scsi_device *sdev = sdkp->device;
3147 
3148 	if (!sdev->security_supported)
3149 		return;
3150 
3151 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3152 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3153 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3154 			SECURITY_PROTOCOL_OUT, 0) == 1)
3155 		sdkp->security = 1;
3156 }
3157 
3158 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3159 {
3160 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3161 }
3162 
3163 /**
3164  * sd_read_cpr - Query concurrent positioning ranges
3165  * @sdkp:	disk to query
3166  */
3167 static void sd_read_cpr(struct scsi_disk *sdkp)
3168 {
3169 	struct blk_independent_access_ranges *iars = NULL;
3170 	unsigned char *buffer = NULL;
3171 	unsigned int nr_cpr = 0;
3172 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3173 	u8 *desc;
3174 
3175 	/*
3176 	 * We need to have the capacity set first for the block layer to be
3177 	 * able to check the ranges.
3178 	 */
3179 	if (sdkp->first_scan)
3180 		return;
3181 
3182 	if (!sdkp->capacity)
3183 		goto out;
3184 
3185 	/*
3186 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3187 	 * leading to a maximum page size of 64 + 256*32 bytes.
3188 	 */
3189 	buf_len = 64 + 256*32;
3190 	buffer = kmalloc(buf_len, GFP_KERNEL);
3191 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3192 		goto out;
3193 
3194 	/* We must have at least a 64B header and one 32B range descriptor */
3195 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3196 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3197 		sd_printk(KERN_ERR, sdkp,
3198 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3199 		goto out;
3200 	}
3201 
3202 	nr_cpr = (vpd_len - 64) / 32;
3203 	if (nr_cpr == 1) {
3204 		nr_cpr = 0;
3205 		goto out;
3206 	}
3207 
3208 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3209 	if (!iars) {
3210 		nr_cpr = 0;
3211 		goto out;
3212 	}
3213 
3214 	desc = &buffer[64];
3215 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3216 		if (desc[0] != i) {
3217 			sd_printk(KERN_ERR, sdkp,
3218 				"Invalid Concurrent Positioning Range number\n");
3219 			nr_cpr = 0;
3220 			break;
3221 		}
3222 
3223 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3224 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3225 	}
3226 
3227 out:
3228 	disk_set_independent_access_ranges(sdkp->disk, iars);
3229 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3230 		sd_printk(KERN_NOTICE, sdkp,
3231 			  "%u concurrent positioning ranges\n", nr_cpr);
3232 		sdkp->nr_actuators = nr_cpr;
3233 	}
3234 
3235 	kfree(buffer);
3236 }
3237 
3238 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3239 {
3240 	struct scsi_device *sdp = sdkp->device;
3241 	unsigned int min_xfer_bytes =
3242 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3243 
3244 	if (sdkp->min_xfer_blocks == 0)
3245 		return false;
3246 
3247 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3248 		sd_first_printk(KERN_WARNING, sdkp,
3249 				"Preferred minimum I/O size %u bytes not a " \
3250 				"multiple of physical block size (%u bytes)\n",
3251 				min_xfer_bytes, sdkp->physical_block_size);
3252 		sdkp->min_xfer_blocks = 0;
3253 		return false;
3254 	}
3255 
3256 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3257 			min_xfer_bytes);
3258 	return true;
3259 }
3260 
3261 /*
3262  * Determine the device's preferred I/O size for reads and writes
3263  * unless the reported value is unreasonably small, large, not a
3264  * multiple of the physical block size, or simply garbage.
3265  */
3266 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3267 				      unsigned int dev_max)
3268 {
3269 	struct scsi_device *sdp = sdkp->device;
3270 	unsigned int opt_xfer_bytes =
3271 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3272 	unsigned int min_xfer_bytes =
3273 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3274 
3275 	if (sdkp->opt_xfer_blocks == 0)
3276 		return false;
3277 
3278 	if (sdkp->opt_xfer_blocks > dev_max) {
3279 		sd_first_printk(KERN_WARNING, sdkp,
3280 				"Optimal transfer size %u logical blocks " \
3281 				"> dev_max (%u logical blocks)\n",
3282 				sdkp->opt_xfer_blocks, dev_max);
3283 		return false;
3284 	}
3285 
3286 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3287 		sd_first_printk(KERN_WARNING, sdkp,
3288 				"Optimal transfer size %u logical blocks " \
3289 				"> sd driver limit (%u logical blocks)\n",
3290 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3291 		return false;
3292 	}
3293 
3294 	if (opt_xfer_bytes < PAGE_SIZE) {
3295 		sd_first_printk(KERN_WARNING, sdkp,
3296 				"Optimal transfer size %u bytes < " \
3297 				"PAGE_SIZE (%u bytes)\n",
3298 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3299 		return false;
3300 	}
3301 
3302 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3303 		sd_first_printk(KERN_WARNING, sdkp,
3304 				"Optimal transfer size %u bytes not a " \
3305 				"multiple of preferred minimum block " \
3306 				"size (%u bytes)\n",
3307 				opt_xfer_bytes, min_xfer_bytes);
3308 		return false;
3309 	}
3310 
3311 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3312 		sd_first_printk(KERN_WARNING, sdkp,
3313 				"Optimal transfer size %u bytes not a " \
3314 				"multiple of physical block size (%u bytes)\n",
3315 				opt_xfer_bytes, sdkp->physical_block_size);
3316 		return false;
3317 	}
3318 
3319 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3320 			opt_xfer_bytes);
3321 	return true;
3322 }
3323 
3324 /**
3325  *	sd_revalidate_disk - called the first time a new disk is seen,
3326  *	performs disk spin up, read_capacity, etc.
3327  *	@disk: struct gendisk we care about
3328  **/
3329 static int sd_revalidate_disk(struct gendisk *disk)
3330 {
3331 	struct scsi_disk *sdkp = scsi_disk(disk);
3332 	struct scsi_device *sdp = sdkp->device;
3333 	struct request_queue *q = sdkp->disk->queue;
3334 	sector_t old_capacity = sdkp->capacity;
3335 	unsigned char *buffer;
3336 	unsigned int dev_max, rw_max;
3337 
3338 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3339 				      "sd_revalidate_disk\n"));
3340 
3341 	/*
3342 	 * If the device is offline, don't try and read capacity or any
3343 	 * of the other niceties.
3344 	 */
3345 	if (!scsi_device_online(sdp))
3346 		goto out;
3347 
3348 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3349 	if (!buffer) {
3350 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3351 			  "allocation failure.\n");
3352 		goto out;
3353 	}
3354 
3355 	sd_spinup_disk(sdkp);
3356 
3357 	/*
3358 	 * Without media there is no reason to ask; moreover, some devices
3359 	 * react badly if we do.
3360 	 */
3361 	if (sdkp->media_present) {
3362 		sd_read_capacity(sdkp, buffer);
3363 
3364 		/*
3365 		 * set the default to rotational.  All non-rotational devices
3366 		 * support the block characteristics VPD page, which will
3367 		 * cause this to be updated correctly and any device which
3368 		 * doesn't support it should be treated as rotational.
3369 		 */
3370 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3371 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3372 
3373 		if (scsi_device_supports_vpd(sdp)) {
3374 			sd_read_block_provisioning(sdkp);
3375 			sd_read_block_limits(sdkp);
3376 			sd_read_block_characteristics(sdkp);
3377 			sd_zbc_read_zones(sdkp, buffer);
3378 			sd_read_cpr(sdkp);
3379 		}
3380 
3381 		sd_print_capacity(sdkp, old_capacity);
3382 
3383 		sd_read_write_protect_flag(sdkp, buffer);
3384 		sd_read_cache_type(sdkp, buffer);
3385 		sd_read_app_tag_own(sdkp, buffer);
3386 		sd_read_write_same(sdkp, buffer);
3387 		sd_read_security(sdkp, buffer);
3388 		sd_config_protection(sdkp);
3389 	}
3390 
3391 	/*
3392 	 * We now have all cache related info, determine how we deal
3393 	 * with flush requests.
3394 	 */
3395 	sd_set_flush_flag(sdkp);
3396 
3397 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3398 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3399 
3400 	/* Some devices report a maximum block count for READ/WRITE requests. */
3401 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3402 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3403 
3404 	if (sd_validate_min_xfer_size(sdkp))
3405 		blk_queue_io_min(sdkp->disk->queue,
3406 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3407 	else
3408 		blk_queue_io_min(sdkp->disk->queue, 0);
3409 
3410 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3411 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3412 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3413 	} else {
3414 		q->limits.io_opt = 0;
3415 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3416 				      (sector_t)BLK_DEF_MAX_SECTORS);
3417 	}
3418 
3419 	/*
3420 	 * Limit default to SCSI host optimal sector limit if set. There may be
3421 	 * an impact on performance for when the size of a request exceeds this
3422 	 * host limit.
3423 	 */
3424 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3425 
3426 	/* Do not exceed controller limit */
3427 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3428 
3429 	/*
3430 	 * Only update max_sectors if previously unset or if the current value
3431 	 * exceeds the capabilities of the hardware.
3432 	 */
3433 	if (sdkp->first_scan ||
3434 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3435 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3436 		q->limits.max_sectors = rw_max;
3437 
3438 	sdkp->first_scan = 0;
3439 
3440 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3441 	sd_config_write_same(sdkp);
3442 	kfree(buffer);
3443 
3444 	/*
3445 	 * For a zoned drive, revalidating the zones can be done only once
3446 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3447 	 * capacity to 0.
3448 	 */
3449 	if (sd_zbc_revalidate_zones(sdkp))
3450 		set_capacity_and_notify(disk, 0);
3451 
3452  out:
3453 	return 0;
3454 }
3455 
3456 /**
3457  *	sd_unlock_native_capacity - unlock native capacity
3458  *	@disk: struct gendisk to set capacity for
3459  *
3460  *	Block layer calls this function if it detects that partitions
3461  *	on @disk reach beyond the end of the device.  If the SCSI host
3462  *	implements ->unlock_native_capacity() method, it's invoked to
3463  *	give it a chance to adjust the device capacity.
3464  *
3465  *	CONTEXT:
3466  *	Defined by block layer.  Might sleep.
3467  */
3468 static void sd_unlock_native_capacity(struct gendisk *disk)
3469 {
3470 	struct scsi_device *sdev = scsi_disk(disk)->device;
3471 
3472 	if (sdev->host->hostt->unlock_native_capacity)
3473 		sdev->host->hostt->unlock_native_capacity(sdev);
3474 }
3475 
3476 /**
3477  *	sd_format_disk_name - format disk name
3478  *	@prefix: name prefix - ie. "sd" for SCSI disks
3479  *	@index: index of the disk to format name for
3480  *	@buf: output buffer
3481  *	@buflen: length of the output buffer
3482  *
3483  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3484  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3485  *	which is followed by sdaaa.
3486  *
3487  *	This is basically 26 base counting with one extra 'nil' entry
3488  *	at the beginning from the second digit on and can be
3489  *	determined using similar method as 26 base conversion with the
3490  *	index shifted -1 after each digit is computed.
3491  *
3492  *	CONTEXT:
3493  *	Don't care.
3494  *
3495  *	RETURNS:
3496  *	0 on success, -errno on failure.
3497  */
3498 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3499 {
3500 	const int base = 'z' - 'a' + 1;
3501 	char *begin = buf + strlen(prefix);
3502 	char *end = buf + buflen;
3503 	char *p;
3504 	int unit;
3505 
3506 	p = end - 1;
3507 	*p = '\0';
3508 	unit = base;
3509 	do {
3510 		if (p == begin)
3511 			return -EINVAL;
3512 		*--p = 'a' + (index % unit);
3513 		index = (index / unit) - 1;
3514 	} while (index >= 0);
3515 
3516 	memmove(begin, p, end - p);
3517 	memcpy(buf, prefix, strlen(prefix));
3518 
3519 	return 0;
3520 }
3521 
3522 /**
3523  *	sd_probe - called during driver initialization and whenever a
3524  *	new scsi device is attached to the system. It is called once
3525  *	for each scsi device (not just disks) present.
3526  *	@dev: pointer to device object
3527  *
3528  *	Returns 0 if successful (or not interested in this scsi device
3529  *	(e.g. scanner)); 1 when there is an error.
3530  *
3531  *	Note: this function is invoked from the scsi mid-level.
3532  *	This function sets up the mapping between a given
3533  *	<host,channel,id,lun> (found in sdp) and new device name
3534  *	(e.g. /dev/sda). More precisely it is the block device major
3535  *	and minor number that is chosen here.
3536  *
3537  *	Assume sd_probe is not re-entrant (for time being)
3538  *	Also think about sd_probe() and sd_remove() running coincidentally.
3539  **/
3540 static int sd_probe(struct device *dev)
3541 {
3542 	struct scsi_device *sdp = to_scsi_device(dev);
3543 	struct scsi_disk *sdkp;
3544 	struct gendisk *gd;
3545 	int index;
3546 	int error;
3547 
3548 	scsi_autopm_get_device(sdp);
3549 	error = -ENODEV;
3550 	if (sdp->type != TYPE_DISK &&
3551 	    sdp->type != TYPE_ZBC &&
3552 	    sdp->type != TYPE_MOD &&
3553 	    sdp->type != TYPE_RBC)
3554 		goto out;
3555 
3556 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3557 		sdev_printk(KERN_WARNING, sdp,
3558 			    "Unsupported ZBC host-managed device.\n");
3559 		goto out;
3560 	}
3561 
3562 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3563 					"sd_probe\n"));
3564 
3565 	error = -ENOMEM;
3566 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3567 	if (!sdkp)
3568 		goto out;
3569 
3570 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3571 					 &sd_bio_compl_lkclass);
3572 	if (!gd)
3573 		goto out_free;
3574 
3575 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3576 	if (index < 0) {
3577 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3578 		goto out_put;
3579 	}
3580 
3581 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3582 	if (error) {
3583 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3584 		goto out_free_index;
3585 	}
3586 
3587 	sdkp->device = sdp;
3588 	sdkp->disk = gd;
3589 	sdkp->index = index;
3590 	sdkp->max_retries = SD_MAX_RETRIES;
3591 	atomic_set(&sdkp->openers, 0);
3592 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3593 
3594 	if (!sdp->request_queue->rq_timeout) {
3595 		if (sdp->type != TYPE_MOD)
3596 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3597 		else
3598 			blk_queue_rq_timeout(sdp->request_queue,
3599 					     SD_MOD_TIMEOUT);
3600 	}
3601 
3602 	device_initialize(&sdkp->disk_dev);
3603 	sdkp->disk_dev.parent = get_device(dev);
3604 	sdkp->disk_dev.class = &sd_disk_class;
3605 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3606 
3607 	error = device_add(&sdkp->disk_dev);
3608 	if (error) {
3609 		put_device(&sdkp->disk_dev);
3610 		goto out;
3611 	}
3612 
3613 	dev_set_drvdata(dev, sdkp);
3614 
3615 	gd->major = sd_major((index & 0xf0) >> 4);
3616 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3617 	gd->minors = SD_MINORS;
3618 
3619 	gd->fops = &sd_fops;
3620 	gd->private_data = sdkp;
3621 
3622 	/* defaults, until the device tells us otherwise */
3623 	sdp->sector_size = 512;
3624 	sdkp->capacity = 0;
3625 	sdkp->media_present = 1;
3626 	sdkp->write_prot = 0;
3627 	sdkp->cache_override = 0;
3628 	sdkp->WCE = 0;
3629 	sdkp->RCD = 0;
3630 	sdkp->ATO = 0;
3631 	sdkp->first_scan = 1;
3632 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3633 
3634 	sd_revalidate_disk(gd);
3635 
3636 	if (sdp->removable) {
3637 		gd->flags |= GENHD_FL_REMOVABLE;
3638 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3639 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3640 	}
3641 
3642 	blk_pm_runtime_init(sdp->request_queue, dev);
3643 	if (sdp->rpm_autosuspend) {
3644 		pm_runtime_set_autosuspend_delay(dev,
3645 			sdp->host->hostt->rpm_autosuspend_delay);
3646 	}
3647 
3648 	error = device_add_disk(dev, gd, NULL);
3649 	if (error) {
3650 		put_device(&sdkp->disk_dev);
3651 		put_disk(gd);
3652 		goto out;
3653 	}
3654 
3655 	if (sdkp->security) {
3656 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3657 		if (sdkp->opal_dev)
3658 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3659 	}
3660 
3661 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3662 		  sdp->removable ? "removable " : "");
3663 	scsi_autopm_put_device(sdp);
3664 
3665 	return 0;
3666 
3667  out_free_index:
3668 	ida_free(&sd_index_ida, index);
3669  out_put:
3670 	put_disk(gd);
3671  out_free:
3672 	kfree(sdkp);
3673  out:
3674 	scsi_autopm_put_device(sdp);
3675 	return error;
3676 }
3677 
3678 /**
3679  *	sd_remove - called whenever a scsi disk (previously recognized by
3680  *	sd_probe) is detached from the system. It is called (potentially
3681  *	multiple times) during sd module unload.
3682  *	@dev: pointer to device object
3683  *
3684  *	Note: this function is invoked from the scsi mid-level.
3685  *	This function potentially frees up a device name (e.g. /dev/sdc)
3686  *	that could be re-used by a subsequent sd_probe().
3687  *	This function is not called when the built-in sd driver is "exit-ed".
3688  **/
3689 static int sd_remove(struct device *dev)
3690 {
3691 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3692 
3693 	scsi_autopm_get_device(sdkp->device);
3694 
3695 	device_del(&sdkp->disk_dev);
3696 	del_gendisk(sdkp->disk);
3697 	sd_shutdown(dev);
3698 
3699 	put_disk(sdkp->disk);
3700 	return 0;
3701 }
3702 
3703 static void scsi_disk_release(struct device *dev)
3704 {
3705 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3706 
3707 	ida_free(&sd_index_ida, sdkp->index);
3708 	sd_zbc_free_zone_info(sdkp);
3709 	put_device(&sdkp->device->sdev_gendev);
3710 	free_opal_dev(sdkp->opal_dev);
3711 
3712 	kfree(sdkp);
3713 }
3714 
3715 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3716 {
3717 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3718 	struct scsi_sense_hdr sshdr;
3719 	const struct scsi_exec_args exec_args = {
3720 		.sshdr = &sshdr,
3721 		.req_flags = BLK_MQ_REQ_PM,
3722 	};
3723 	struct scsi_device *sdp = sdkp->device;
3724 	int res;
3725 
3726 	if (start)
3727 		cmd[4] |= 1;	/* START */
3728 
3729 	if (sdp->start_stop_pwr_cond)
3730 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3731 
3732 	if (!scsi_device_online(sdp))
3733 		return -ENODEV;
3734 
3735 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3736 			       sdkp->max_retries, &exec_args);
3737 	if (res) {
3738 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3739 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3740 			sd_print_sense_hdr(sdkp, &sshdr);
3741 			/* 0x3a is medium not present */
3742 			if (sshdr.asc == 0x3a)
3743 				res = 0;
3744 		}
3745 	}
3746 
3747 	/* SCSI error codes must not go to the generic layer */
3748 	if (res)
3749 		return -EIO;
3750 
3751 	return 0;
3752 }
3753 
3754 /*
3755  * Send a SYNCHRONIZE CACHE instruction down to the device through
3756  * the normal SCSI command structure.  Wait for the command to
3757  * complete.
3758  */
3759 static void sd_shutdown(struct device *dev)
3760 {
3761 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3762 
3763 	if (!sdkp)
3764 		return;         /* this can happen */
3765 
3766 	if (pm_runtime_suspended(dev))
3767 		return;
3768 
3769 	if (sdkp->WCE && sdkp->media_present) {
3770 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3771 		sd_sync_cache(sdkp, NULL);
3772 	}
3773 
3774 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3775 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3776 		sd_start_stop_device(sdkp, 0);
3777 	}
3778 }
3779 
3780 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3781 {
3782 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3783 	struct scsi_sense_hdr sshdr;
3784 	int ret = 0;
3785 
3786 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3787 		return 0;
3788 
3789 	if (sdkp->WCE && sdkp->media_present) {
3790 		if (!sdkp->device->silence_suspend)
3791 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3792 		ret = sd_sync_cache(sdkp, &sshdr);
3793 
3794 		if (ret) {
3795 			/* ignore OFFLINE device */
3796 			if (ret == -ENODEV)
3797 				return 0;
3798 
3799 			if (!scsi_sense_valid(&sshdr) ||
3800 			    sshdr.sense_key != ILLEGAL_REQUEST)
3801 				return ret;
3802 
3803 			/*
3804 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3805 			 * doesn't support sync. There's not much to do and
3806 			 * suspend shouldn't fail.
3807 			 */
3808 			ret = 0;
3809 		}
3810 	}
3811 
3812 	if (sdkp->device->manage_start_stop) {
3813 		if (!sdkp->device->silence_suspend)
3814 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3815 		/* an error is not worth aborting a system sleep */
3816 		ret = sd_start_stop_device(sdkp, 0);
3817 		if (ignore_stop_errors)
3818 			ret = 0;
3819 	}
3820 
3821 	return ret;
3822 }
3823 
3824 static int sd_suspend_system(struct device *dev)
3825 {
3826 	if (pm_runtime_suspended(dev))
3827 		return 0;
3828 
3829 	return sd_suspend_common(dev, true);
3830 }
3831 
3832 static int sd_suspend_runtime(struct device *dev)
3833 {
3834 	return sd_suspend_common(dev, false);
3835 }
3836 
3837 static int sd_resume(struct device *dev)
3838 {
3839 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3840 	int ret = 0;
3841 
3842 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3843 		return 0;
3844 
3845 	if (!sdkp->device->manage_start_stop)
3846 		return 0;
3847 
3848 	if (!sdkp->device->no_start_on_resume) {
3849 		sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3850 		ret = sd_start_stop_device(sdkp, 1);
3851 	}
3852 
3853 	if (!ret)
3854 		opal_unlock_from_suspend(sdkp->opal_dev);
3855 	return ret;
3856 }
3857 
3858 static int sd_resume_system(struct device *dev)
3859 {
3860 	if (pm_runtime_suspended(dev))
3861 		return 0;
3862 
3863 	return sd_resume(dev);
3864 }
3865 
3866 static int sd_resume_runtime(struct device *dev)
3867 {
3868 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3869 	struct scsi_device *sdp;
3870 
3871 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3872 		return 0;
3873 
3874 	sdp = sdkp->device;
3875 
3876 	if (sdp->ignore_media_change) {
3877 		/* clear the device's sense data */
3878 		static const u8 cmd[10] = { REQUEST_SENSE };
3879 		const struct scsi_exec_args exec_args = {
3880 			.req_flags = BLK_MQ_REQ_PM,
3881 		};
3882 
3883 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
3884 				     sdp->request_queue->rq_timeout, 1,
3885 				     &exec_args))
3886 			sd_printk(KERN_NOTICE, sdkp,
3887 				  "Failed to clear sense data\n");
3888 	}
3889 
3890 	return sd_resume(dev);
3891 }
3892 
3893 static const struct dev_pm_ops sd_pm_ops = {
3894 	.suspend		= sd_suspend_system,
3895 	.resume			= sd_resume_system,
3896 	.poweroff		= sd_suspend_system,
3897 	.restore		= sd_resume_system,
3898 	.runtime_suspend	= sd_suspend_runtime,
3899 	.runtime_resume		= sd_resume_runtime,
3900 };
3901 
3902 static struct scsi_driver sd_template = {
3903 	.gendrv = {
3904 		.name		= "sd",
3905 		.owner		= THIS_MODULE,
3906 		.probe		= sd_probe,
3907 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
3908 		.remove		= sd_remove,
3909 		.shutdown	= sd_shutdown,
3910 		.pm		= &sd_pm_ops,
3911 	},
3912 	.rescan			= sd_rescan,
3913 	.init_command		= sd_init_command,
3914 	.uninit_command		= sd_uninit_command,
3915 	.done			= sd_done,
3916 	.eh_action		= sd_eh_action,
3917 	.eh_reset		= sd_eh_reset,
3918 };
3919 
3920 /**
3921  *	init_sd - entry point for this driver (both when built in or when
3922  *	a module).
3923  *
3924  *	Note: this function registers this driver with the scsi mid-level.
3925  **/
3926 static int __init init_sd(void)
3927 {
3928 	int majors = 0, i, err;
3929 
3930 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3931 
3932 	for (i = 0; i < SD_MAJORS; i++) {
3933 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3934 			continue;
3935 		majors++;
3936 	}
3937 
3938 	if (!majors)
3939 		return -ENODEV;
3940 
3941 	err = class_register(&sd_disk_class);
3942 	if (err)
3943 		goto err_out;
3944 
3945 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3946 	if (!sd_page_pool) {
3947 		printk(KERN_ERR "sd: can't init discard page pool\n");
3948 		err = -ENOMEM;
3949 		goto err_out_class;
3950 	}
3951 
3952 	err = scsi_register_driver(&sd_template.gendrv);
3953 	if (err)
3954 		goto err_out_driver;
3955 
3956 	return 0;
3957 
3958 err_out_driver:
3959 	mempool_destroy(sd_page_pool);
3960 err_out_class:
3961 	class_unregister(&sd_disk_class);
3962 err_out:
3963 	for (i = 0; i < SD_MAJORS; i++)
3964 		unregister_blkdev(sd_major(i), "sd");
3965 	return err;
3966 }
3967 
3968 /**
3969  *	exit_sd - exit point for this driver (when it is a module).
3970  *
3971  *	Note: this function unregisters this driver from the scsi mid-level.
3972  **/
3973 static void __exit exit_sd(void)
3974 {
3975 	int i;
3976 
3977 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3978 
3979 	scsi_unregister_driver(&sd_template.gendrv);
3980 	mempool_destroy(sd_page_pool);
3981 
3982 	class_unregister(&sd_disk_class);
3983 
3984 	for (i = 0; i < SD_MAJORS; i++)
3985 		unregister_blkdev(sd_major(i), "sd");
3986 }
3987 
3988 module_init(init_sd);
3989 module_exit(exit_sd);
3990 
3991 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3992 {
3993 	scsi_print_sense_hdr(sdkp->device,
3994 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3995 }
3996 
3997 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3998 {
3999 	const char *hb_string = scsi_hostbyte_string(result);
4000 
4001 	if (hb_string)
4002 		sd_printk(KERN_INFO, sdkp,
4003 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4004 			  hb_string ? hb_string : "invalid",
4005 			  "DRIVER_OK");
4006 	else
4007 		sd_printk(KERN_INFO, sdkp,
4008 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4009 			  msg, host_byte(result), "DRIVER_OK");
4010 }
4011