xref: /linux/drivers/scsi/sd.c (revision 1c07425e902cd3137961c3d45b4271bf8a9b8eb9)
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 
71 #include "sd.h"
72 #include "scsi_priv.h"
73 #include "scsi_logging.h"
74 
75 MODULE_AUTHOR("Eric Youngdale");
76 MODULE_DESCRIPTION("SCSI disk (sd) driver");
77 MODULE_LICENSE("GPL");
78 
79 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
95 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
99 
100 #define SD_MINORS	16
101 
102 static void sd_config_discard(struct scsi_disk *, unsigned int);
103 static void sd_config_write_same(struct scsi_disk *);
104 static int  sd_revalidate_disk(struct gendisk *);
105 static void sd_unlock_native_capacity(struct gendisk *disk);
106 static int  sd_probe(struct device *);
107 static int  sd_remove(struct device *);
108 static void sd_shutdown(struct device *);
109 static int sd_suspend_system(struct device *);
110 static int sd_suspend_runtime(struct device *);
111 static int sd_resume_system(struct device *);
112 static int sd_resume_runtime(struct device *);
113 static void sd_rescan(struct device *);
114 static blk_status_t sd_init_command(struct scsi_cmnd *SCpnt);
115 static void sd_uninit_command(struct scsi_cmnd *SCpnt);
116 static int sd_done(struct scsi_cmnd *);
117 static void sd_eh_reset(struct scsi_cmnd *);
118 static int sd_eh_action(struct scsi_cmnd *, int);
119 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
120 static void scsi_disk_release(struct device *cdev);
121 
122 static DEFINE_IDA(sd_index_ida);
123 
124 static struct kmem_cache *sd_cdb_cache;
125 static mempool_t *sd_page_pool;
126 static struct lock_class_key sd_bio_compl_lkclass;
127 
128 static const char *sd_cache_types[] = {
129 	"write through", "none", "write back",
130 	"write back, no read (daft)"
131 };
132 
133 static void sd_set_flush_flag(struct scsi_disk *sdkp)
134 {
135 	bool wc = false, fua = false;
136 
137 	if (sdkp->WCE) {
138 		wc = true;
139 		if (sdkp->DPOFUA)
140 			fua = true;
141 	}
142 
143 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
144 }
145 
146 static ssize_t
147 cache_type_store(struct device *dev, struct device_attribute *attr,
148 		 const char *buf, size_t count)
149 {
150 	int ct, rcd, wce, sp;
151 	struct scsi_disk *sdkp = to_scsi_disk(dev);
152 	struct scsi_device *sdp = sdkp->device;
153 	char buffer[64];
154 	char *buffer_data;
155 	struct scsi_mode_data data;
156 	struct scsi_sense_hdr sshdr;
157 	static const char temp[] = "temporary ";
158 	int len;
159 
160 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
161 		/* no cache control on RBC devices; theoretically they
162 		 * can do it, but there's probably so many exceptions
163 		 * it's not worth the risk */
164 		return -EINVAL;
165 
166 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
167 		buf += sizeof(temp) - 1;
168 		sdkp->cache_override = 1;
169 	} else {
170 		sdkp->cache_override = 0;
171 	}
172 
173 	ct = sysfs_match_string(sd_cache_types, buf);
174 	if (ct < 0)
175 		return -EINVAL;
176 
177 	rcd = ct & 0x01 ? 1 : 0;
178 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
179 
180 	if (sdkp->cache_override) {
181 		sdkp->WCE = wce;
182 		sdkp->RCD = rcd;
183 		sd_set_flush_flag(sdkp);
184 		return count;
185 	}
186 
187 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
188 			    sdkp->max_retries, &data, NULL))
189 		return -EINVAL;
190 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
191 		  data.block_descriptor_length);
192 	buffer_data = buffer + data.header_length +
193 		data.block_descriptor_length;
194 	buffer_data[2] &= ~0x05;
195 	buffer_data[2] |= wce << 2 | rcd;
196 	sp = buffer_data[0] & 0x80 ? 1 : 0;
197 	buffer_data[0] &= ~0x80;
198 
199 	/*
200 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
201 	 * received mode parameter buffer before doing MODE SELECT.
202 	 */
203 	data.device_specific = 0;
204 
205 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
206 			     sdkp->max_retries, &data, &sshdr)) {
207 		if (scsi_sense_valid(&sshdr))
208 			sd_print_sense_hdr(sdkp, &sshdr);
209 		return -EINVAL;
210 	}
211 	sd_revalidate_disk(sdkp->disk);
212 	return count;
213 }
214 
215 static ssize_t
216 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
217 		       char *buf)
218 {
219 	struct scsi_disk *sdkp = to_scsi_disk(dev);
220 	struct scsi_device *sdp = sdkp->device;
221 
222 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
223 }
224 
225 static ssize_t
226 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
227 			const char *buf, size_t count)
228 {
229 	struct scsi_disk *sdkp = to_scsi_disk(dev);
230 	struct scsi_device *sdp = sdkp->device;
231 	bool v;
232 
233 	if (!capable(CAP_SYS_ADMIN))
234 		return -EACCES;
235 
236 	if (kstrtobool(buf, &v))
237 		return -EINVAL;
238 
239 	sdp->manage_start_stop = v;
240 
241 	return count;
242 }
243 static DEVICE_ATTR_RW(manage_start_stop);
244 
245 static ssize_t
246 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
247 {
248 	struct scsi_disk *sdkp = to_scsi_disk(dev);
249 
250 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
251 }
252 
253 static ssize_t
254 allow_restart_store(struct device *dev, struct device_attribute *attr,
255 		    const char *buf, size_t count)
256 {
257 	bool v;
258 	struct scsi_disk *sdkp = to_scsi_disk(dev);
259 	struct scsi_device *sdp = sdkp->device;
260 
261 	if (!capable(CAP_SYS_ADMIN))
262 		return -EACCES;
263 
264 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
265 		return -EINVAL;
266 
267 	if (kstrtobool(buf, &v))
268 		return -EINVAL;
269 
270 	sdp->allow_restart = v;
271 
272 	return count;
273 }
274 static DEVICE_ATTR_RW(allow_restart);
275 
276 static ssize_t
277 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
278 {
279 	struct scsi_disk *sdkp = to_scsi_disk(dev);
280 	int ct = sdkp->RCD + 2*sdkp->WCE;
281 
282 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
283 }
284 static DEVICE_ATTR_RW(cache_type);
285 
286 static ssize_t
287 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
288 {
289 	struct scsi_disk *sdkp = to_scsi_disk(dev);
290 
291 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
292 }
293 static DEVICE_ATTR_RO(FUA);
294 
295 static ssize_t
296 protection_type_show(struct device *dev, struct device_attribute *attr,
297 		     char *buf)
298 {
299 	struct scsi_disk *sdkp = to_scsi_disk(dev);
300 
301 	return sprintf(buf, "%u\n", sdkp->protection_type);
302 }
303 
304 static ssize_t
305 protection_type_store(struct device *dev, struct device_attribute *attr,
306 		      const char *buf, size_t count)
307 {
308 	struct scsi_disk *sdkp = to_scsi_disk(dev);
309 	unsigned int val;
310 	int err;
311 
312 	if (!capable(CAP_SYS_ADMIN))
313 		return -EACCES;
314 
315 	err = kstrtouint(buf, 10, &val);
316 
317 	if (err)
318 		return err;
319 
320 	if (val <= T10_PI_TYPE3_PROTECTION)
321 		sdkp->protection_type = val;
322 
323 	return count;
324 }
325 static DEVICE_ATTR_RW(protection_type);
326 
327 static ssize_t
328 protection_mode_show(struct device *dev, struct device_attribute *attr,
329 		     char *buf)
330 {
331 	struct scsi_disk *sdkp = to_scsi_disk(dev);
332 	struct scsi_device *sdp = sdkp->device;
333 	unsigned int dif, dix;
334 
335 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
336 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
337 
338 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
339 		dif = 0;
340 		dix = 1;
341 	}
342 
343 	if (!dif && !dix)
344 		return sprintf(buf, "none\n");
345 
346 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
347 }
348 static DEVICE_ATTR_RO(protection_mode);
349 
350 static ssize_t
351 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
352 {
353 	struct scsi_disk *sdkp = to_scsi_disk(dev);
354 
355 	return sprintf(buf, "%u\n", sdkp->ATO);
356 }
357 static DEVICE_ATTR_RO(app_tag_own);
358 
359 static ssize_t
360 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
361 		       char *buf)
362 {
363 	struct scsi_disk *sdkp = to_scsi_disk(dev);
364 
365 	return sprintf(buf, "%u\n", sdkp->lbpme);
366 }
367 static DEVICE_ATTR_RO(thin_provisioning);
368 
369 /* sysfs_match_string() requires dense arrays */
370 static const char *lbp_mode[] = {
371 	[SD_LBP_FULL]		= "full",
372 	[SD_LBP_UNMAP]		= "unmap",
373 	[SD_LBP_WS16]		= "writesame_16",
374 	[SD_LBP_WS10]		= "writesame_10",
375 	[SD_LBP_ZERO]		= "writesame_zero",
376 	[SD_LBP_DISABLE]	= "disabled",
377 };
378 
379 static ssize_t
380 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
381 		       char *buf)
382 {
383 	struct scsi_disk *sdkp = to_scsi_disk(dev);
384 
385 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
386 }
387 
388 static ssize_t
389 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
390 			const char *buf, size_t count)
391 {
392 	struct scsi_disk *sdkp = to_scsi_disk(dev);
393 	struct scsi_device *sdp = sdkp->device;
394 	int mode;
395 
396 	if (!capable(CAP_SYS_ADMIN))
397 		return -EACCES;
398 
399 	if (sd_is_zoned(sdkp)) {
400 		sd_config_discard(sdkp, SD_LBP_DISABLE);
401 		return count;
402 	}
403 
404 	if (sdp->type != TYPE_DISK)
405 		return -EINVAL;
406 
407 	mode = sysfs_match_string(lbp_mode, buf);
408 	if (mode < 0)
409 		return -EINVAL;
410 
411 	sd_config_discard(sdkp, mode);
412 
413 	return count;
414 }
415 static DEVICE_ATTR_RW(provisioning_mode);
416 
417 /* sysfs_match_string() requires dense arrays */
418 static const char *zeroing_mode[] = {
419 	[SD_ZERO_WRITE]		= "write",
420 	[SD_ZERO_WS]		= "writesame",
421 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
422 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
423 };
424 
425 static ssize_t
426 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
427 		  char *buf)
428 {
429 	struct scsi_disk *sdkp = to_scsi_disk(dev);
430 
431 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
432 }
433 
434 static ssize_t
435 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
436 		   const char *buf, size_t count)
437 {
438 	struct scsi_disk *sdkp = to_scsi_disk(dev);
439 	int mode;
440 
441 	if (!capable(CAP_SYS_ADMIN))
442 		return -EACCES;
443 
444 	mode = sysfs_match_string(zeroing_mode, buf);
445 	if (mode < 0)
446 		return -EINVAL;
447 
448 	sdkp->zeroing_mode = mode;
449 
450 	return count;
451 }
452 static DEVICE_ATTR_RW(zeroing_mode);
453 
454 static ssize_t
455 max_medium_access_timeouts_show(struct device *dev,
456 				struct device_attribute *attr, char *buf)
457 {
458 	struct scsi_disk *sdkp = to_scsi_disk(dev);
459 
460 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
461 }
462 
463 static ssize_t
464 max_medium_access_timeouts_store(struct device *dev,
465 				 struct device_attribute *attr, const char *buf,
466 				 size_t count)
467 {
468 	struct scsi_disk *sdkp = to_scsi_disk(dev);
469 	int err;
470 
471 	if (!capable(CAP_SYS_ADMIN))
472 		return -EACCES;
473 
474 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
475 
476 	return err ? err : count;
477 }
478 static DEVICE_ATTR_RW(max_medium_access_timeouts);
479 
480 static ssize_t
481 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
482 			   char *buf)
483 {
484 	struct scsi_disk *sdkp = to_scsi_disk(dev);
485 
486 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
487 }
488 
489 static ssize_t
490 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
491 			    const char *buf, size_t count)
492 {
493 	struct scsi_disk *sdkp = to_scsi_disk(dev);
494 	struct scsi_device *sdp = sdkp->device;
495 	unsigned long max;
496 	int err;
497 
498 	if (!capable(CAP_SYS_ADMIN))
499 		return -EACCES;
500 
501 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
502 		return -EINVAL;
503 
504 	err = kstrtoul(buf, 10, &max);
505 
506 	if (err)
507 		return err;
508 
509 	if (max == 0)
510 		sdp->no_write_same = 1;
511 	else if (max <= SD_MAX_WS16_BLOCKS) {
512 		sdp->no_write_same = 0;
513 		sdkp->max_ws_blocks = max;
514 	}
515 
516 	sd_config_write_same(sdkp);
517 
518 	return count;
519 }
520 static DEVICE_ATTR_RW(max_write_same_blocks);
521 
522 static ssize_t
523 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
524 {
525 	struct scsi_disk *sdkp = to_scsi_disk(dev);
526 
527 	if (sdkp->device->type == TYPE_ZBC)
528 		return sprintf(buf, "host-managed\n");
529 	if (sdkp->zoned == 1)
530 		return sprintf(buf, "host-aware\n");
531 	if (sdkp->zoned == 2)
532 		return sprintf(buf, "drive-managed\n");
533 	return sprintf(buf, "none\n");
534 }
535 static DEVICE_ATTR_RO(zoned_cap);
536 
537 static ssize_t
538 max_retries_store(struct device *dev, struct device_attribute *attr,
539 		  const char *buf, size_t count)
540 {
541 	struct scsi_disk *sdkp = to_scsi_disk(dev);
542 	struct scsi_device *sdev = sdkp->device;
543 	int retries, err;
544 
545 	err = kstrtoint(buf, 10, &retries);
546 	if (err)
547 		return err;
548 
549 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
550 		sdkp->max_retries = retries;
551 		return count;
552 	}
553 
554 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
555 		    SD_MAX_RETRIES);
556 	return -EINVAL;
557 }
558 
559 static ssize_t
560 max_retries_show(struct device *dev, struct device_attribute *attr,
561 		 char *buf)
562 {
563 	struct scsi_disk *sdkp = to_scsi_disk(dev);
564 
565 	return sprintf(buf, "%d\n", sdkp->max_retries);
566 }
567 
568 static DEVICE_ATTR_RW(max_retries);
569 
570 static struct attribute *sd_disk_attrs[] = {
571 	&dev_attr_cache_type.attr,
572 	&dev_attr_FUA.attr,
573 	&dev_attr_allow_restart.attr,
574 	&dev_attr_manage_start_stop.attr,
575 	&dev_attr_protection_type.attr,
576 	&dev_attr_protection_mode.attr,
577 	&dev_attr_app_tag_own.attr,
578 	&dev_attr_thin_provisioning.attr,
579 	&dev_attr_provisioning_mode.attr,
580 	&dev_attr_zeroing_mode.attr,
581 	&dev_attr_max_write_same_blocks.attr,
582 	&dev_attr_max_medium_access_timeouts.attr,
583 	&dev_attr_zoned_cap.attr,
584 	&dev_attr_max_retries.attr,
585 	NULL,
586 };
587 ATTRIBUTE_GROUPS(sd_disk);
588 
589 static struct class sd_disk_class = {
590 	.name		= "scsi_disk",
591 	.owner		= THIS_MODULE,
592 	.dev_release	= scsi_disk_release,
593 	.dev_groups	= sd_disk_groups,
594 };
595 
596 static const struct dev_pm_ops sd_pm_ops = {
597 	.suspend		= sd_suspend_system,
598 	.resume			= sd_resume_system,
599 	.poweroff		= sd_suspend_system,
600 	.restore		= sd_resume_system,
601 	.runtime_suspend	= sd_suspend_runtime,
602 	.runtime_resume		= sd_resume_runtime,
603 };
604 
605 static struct scsi_driver sd_template = {
606 	.gendrv = {
607 		.name		= "sd",
608 		.owner		= THIS_MODULE,
609 		.probe		= sd_probe,
610 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
611 		.remove		= sd_remove,
612 		.shutdown	= sd_shutdown,
613 		.pm		= &sd_pm_ops,
614 	},
615 	.rescan			= sd_rescan,
616 	.init_command		= sd_init_command,
617 	.uninit_command		= sd_uninit_command,
618 	.done			= sd_done,
619 	.eh_action		= sd_eh_action,
620 	.eh_reset		= sd_eh_reset,
621 };
622 
623 /*
624  * Don't request a new module, as that could deadlock in multipath
625  * environment.
626  */
627 static void sd_default_probe(dev_t devt)
628 {
629 }
630 
631 /*
632  * Device no to disk mapping:
633  *
634  *       major         disc2     disc  p1
635  *   |............|.............|....|....| <- dev_t
636  *    31        20 19          8 7  4 3  0
637  *
638  * Inside a major, we have 16k disks, however mapped non-
639  * contiguously. The first 16 disks are for major0, the next
640  * ones with major1, ... Disk 256 is for major0 again, disk 272
641  * for major1, ...
642  * As we stay compatible with our numbering scheme, we can reuse
643  * the well-know SCSI majors 8, 65--71, 136--143.
644  */
645 static int sd_major(int major_idx)
646 {
647 	switch (major_idx) {
648 	case 0:
649 		return SCSI_DISK0_MAJOR;
650 	case 1 ... 7:
651 		return SCSI_DISK1_MAJOR + major_idx - 1;
652 	case 8 ... 15:
653 		return SCSI_DISK8_MAJOR + major_idx - 8;
654 	default:
655 		BUG();
656 		return 0;	/* shut up gcc */
657 	}
658 }
659 
660 #ifdef CONFIG_BLK_SED_OPAL
661 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
662 		size_t len, bool send)
663 {
664 	struct scsi_disk *sdkp = data;
665 	struct scsi_device *sdev = sdkp->device;
666 	u8 cdb[12] = { 0, };
667 	int ret;
668 
669 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
670 	cdb[1] = secp;
671 	put_unaligned_be16(spsp, &cdb[2]);
672 	put_unaligned_be32(len, &cdb[6]);
673 
674 	ret = scsi_execute(sdev, cdb, send ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
675 		buffer, len, NULL, NULL, SD_TIMEOUT, sdkp->max_retries, 0,
676 		RQF_PM, NULL);
677 	return ret <= 0 ? ret : -EIO;
678 }
679 #endif /* CONFIG_BLK_SED_OPAL */
680 
681 /*
682  * Look up the DIX operation based on whether the command is read or
683  * write and whether dix and dif are enabled.
684  */
685 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
686 {
687 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
688 	static const unsigned int ops[] = {	/* wrt dix dif */
689 		SCSI_PROT_NORMAL,		/*  0	0   0  */
690 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
691 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
692 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
693 		SCSI_PROT_NORMAL,		/*  1	0   0  */
694 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
695 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
696 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
697 	};
698 
699 	return ops[write << 2 | dix << 1 | dif];
700 }
701 
702 /*
703  * Returns a mask of the protection flags that are valid for a given DIX
704  * operation.
705  */
706 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
707 {
708 	static const unsigned int flag_mask[] = {
709 		[SCSI_PROT_NORMAL]		= 0,
710 
711 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
712 						  SCSI_PROT_GUARD_CHECK |
713 						  SCSI_PROT_REF_CHECK |
714 						  SCSI_PROT_REF_INCREMENT,
715 
716 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
717 						  SCSI_PROT_IP_CHECKSUM,
718 
719 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
720 						  SCSI_PROT_GUARD_CHECK |
721 						  SCSI_PROT_REF_CHECK |
722 						  SCSI_PROT_REF_INCREMENT |
723 						  SCSI_PROT_IP_CHECKSUM,
724 
725 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
726 						  SCSI_PROT_REF_INCREMENT,
727 
728 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
729 						  SCSI_PROT_REF_CHECK |
730 						  SCSI_PROT_REF_INCREMENT |
731 						  SCSI_PROT_IP_CHECKSUM,
732 
733 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
734 						  SCSI_PROT_GUARD_CHECK |
735 						  SCSI_PROT_REF_CHECK |
736 						  SCSI_PROT_REF_INCREMENT |
737 						  SCSI_PROT_IP_CHECKSUM,
738 	};
739 
740 	return flag_mask[prot_op];
741 }
742 
743 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
744 					   unsigned int dix, unsigned int dif)
745 {
746 	struct request *rq = scsi_cmd_to_rq(scmd);
747 	struct bio *bio = rq->bio;
748 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
749 	unsigned int protect = 0;
750 
751 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
752 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
753 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
754 
755 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
756 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
757 	}
758 
759 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
760 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
761 
762 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
763 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
764 	}
765 
766 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
767 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
768 
769 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
770 			protect = 3 << 5;	/* Disable target PI checking */
771 		else
772 			protect = 1 << 5;	/* Enable target PI checking */
773 	}
774 
775 	scsi_set_prot_op(scmd, prot_op);
776 	scsi_set_prot_type(scmd, dif);
777 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
778 
779 	return protect;
780 }
781 
782 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
783 {
784 	struct request_queue *q = sdkp->disk->queue;
785 	unsigned int logical_block_size = sdkp->device->sector_size;
786 	unsigned int max_blocks = 0;
787 
788 	q->limits.discard_alignment =
789 		sdkp->unmap_alignment * logical_block_size;
790 	q->limits.discard_granularity =
791 		max(sdkp->physical_block_size,
792 		    sdkp->unmap_granularity * logical_block_size);
793 	sdkp->provisioning_mode = mode;
794 
795 	switch (mode) {
796 
797 	case SD_LBP_FULL:
798 	case SD_LBP_DISABLE:
799 		blk_queue_max_discard_sectors(q, 0);
800 		return;
801 
802 	case SD_LBP_UNMAP:
803 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
804 					  (u32)SD_MAX_WS16_BLOCKS);
805 		break;
806 
807 	case SD_LBP_WS16:
808 		if (sdkp->device->unmap_limit_for_ws)
809 			max_blocks = sdkp->max_unmap_blocks;
810 		else
811 			max_blocks = sdkp->max_ws_blocks;
812 
813 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
814 		break;
815 
816 	case SD_LBP_WS10:
817 		if (sdkp->device->unmap_limit_for_ws)
818 			max_blocks = sdkp->max_unmap_blocks;
819 		else
820 			max_blocks = sdkp->max_ws_blocks;
821 
822 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
823 		break;
824 
825 	case SD_LBP_ZERO:
826 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
827 					  (u32)SD_MAX_WS10_BLOCKS);
828 		break;
829 	}
830 
831 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
832 }
833 
834 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
835 {
836 	struct scsi_device *sdp = cmd->device;
837 	struct request *rq = scsi_cmd_to_rq(cmd);
838 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
839 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
840 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
841 	unsigned int data_len = 24;
842 	char *buf;
843 
844 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
845 	if (!rq->special_vec.bv_page)
846 		return BLK_STS_RESOURCE;
847 	clear_highpage(rq->special_vec.bv_page);
848 	rq->special_vec.bv_offset = 0;
849 	rq->special_vec.bv_len = data_len;
850 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
851 
852 	cmd->cmd_len = 10;
853 	cmd->cmnd[0] = UNMAP;
854 	cmd->cmnd[8] = 24;
855 
856 	buf = bvec_virt(&rq->special_vec);
857 	put_unaligned_be16(6 + 16, &buf[0]);
858 	put_unaligned_be16(16, &buf[2]);
859 	put_unaligned_be64(lba, &buf[8]);
860 	put_unaligned_be32(nr_blocks, &buf[16]);
861 
862 	cmd->allowed = sdkp->max_retries;
863 	cmd->transfersize = data_len;
864 	rq->timeout = SD_TIMEOUT;
865 
866 	return scsi_alloc_sgtables(cmd);
867 }
868 
869 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
870 		bool unmap)
871 {
872 	struct scsi_device *sdp = cmd->device;
873 	struct request *rq = scsi_cmd_to_rq(cmd);
874 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
875 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
876 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
877 	u32 data_len = sdp->sector_size;
878 
879 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
880 	if (!rq->special_vec.bv_page)
881 		return BLK_STS_RESOURCE;
882 	clear_highpage(rq->special_vec.bv_page);
883 	rq->special_vec.bv_offset = 0;
884 	rq->special_vec.bv_len = data_len;
885 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
886 
887 	cmd->cmd_len = 16;
888 	cmd->cmnd[0] = WRITE_SAME_16;
889 	if (unmap)
890 		cmd->cmnd[1] = 0x8; /* UNMAP */
891 	put_unaligned_be64(lba, &cmd->cmnd[2]);
892 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
893 
894 	cmd->allowed = sdkp->max_retries;
895 	cmd->transfersize = data_len;
896 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
897 
898 	return scsi_alloc_sgtables(cmd);
899 }
900 
901 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
902 		bool unmap)
903 {
904 	struct scsi_device *sdp = cmd->device;
905 	struct request *rq = scsi_cmd_to_rq(cmd);
906 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
907 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
908 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
909 	u32 data_len = sdp->sector_size;
910 
911 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
912 	if (!rq->special_vec.bv_page)
913 		return BLK_STS_RESOURCE;
914 	clear_highpage(rq->special_vec.bv_page);
915 	rq->special_vec.bv_offset = 0;
916 	rq->special_vec.bv_len = data_len;
917 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
918 
919 	cmd->cmd_len = 10;
920 	cmd->cmnd[0] = WRITE_SAME;
921 	if (unmap)
922 		cmd->cmnd[1] = 0x8; /* UNMAP */
923 	put_unaligned_be32(lba, &cmd->cmnd[2]);
924 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
925 
926 	cmd->allowed = sdkp->max_retries;
927 	cmd->transfersize = data_len;
928 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
929 
930 	return scsi_alloc_sgtables(cmd);
931 }
932 
933 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
934 {
935 	struct request *rq = scsi_cmd_to_rq(cmd);
936 	struct scsi_device *sdp = cmd->device;
937 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
938 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
939 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
940 
941 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
942 		switch (sdkp->zeroing_mode) {
943 		case SD_ZERO_WS16_UNMAP:
944 			return sd_setup_write_same16_cmnd(cmd, true);
945 		case SD_ZERO_WS10_UNMAP:
946 			return sd_setup_write_same10_cmnd(cmd, true);
947 		}
948 	}
949 
950 	if (sdp->no_write_same) {
951 		rq->rq_flags |= RQF_QUIET;
952 		return BLK_STS_TARGET;
953 	}
954 
955 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
956 		return sd_setup_write_same16_cmnd(cmd, false);
957 
958 	return sd_setup_write_same10_cmnd(cmd, false);
959 }
960 
961 static void sd_config_write_same(struct scsi_disk *sdkp)
962 {
963 	struct request_queue *q = sdkp->disk->queue;
964 	unsigned int logical_block_size = sdkp->device->sector_size;
965 
966 	if (sdkp->device->no_write_same) {
967 		sdkp->max_ws_blocks = 0;
968 		goto out;
969 	}
970 
971 	/* Some devices can not handle block counts above 0xffff despite
972 	 * supporting WRITE SAME(16). Consequently we default to 64k
973 	 * blocks per I/O unless the device explicitly advertises a
974 	 * bigger limit.
975 	 */
976 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
977 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
978 						   (u32)SD_MAX_WS16_BLOCKS);
979 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
980 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
981 						   (u32)SD_MAX_WS10_BLOCKS);
982 	else {
983 		sdkp->device->no_write_same = 1;
984 		sdkp->max_ws_blocks = 0;
985 	}
986 
987 	if (sdkp->lbprz && sdkp->lbpws)
988 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
989 	else if (sdkp->lbprz && sdkp->lbpws10)
990 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
991 	else if (sdkp->max_ws_blocks)
992 		sdkp->zeroing_mode = SD_ZERO_WS;
993 	else
994 		sdkp->zeroing_mode = SD_ZERO_WRITE;
995 
996 	if (sdkp->max_ws_blocks &&
997 	    sdkp->physical_block_size > logical_block_size) {
998 		/*
999 		 * Reporting a maximum number of blocks that is not aligned
1000 		 * on the device physical size would cause a large write same
1001 		 * request to be split into physically unaligned chunks by
1002 		 * __blkdev_issue_write_zeroes() even if the caller of this
1003 		 * functions took care to align the large request. So make sure
1004 		 * the maximum reported is aligned to the device physical block
1005 		 * size. This is only an optional optimization for regular
1006 		 * disks, but this is mandatory to avoid failure of large write
1007 		 * same requests directed at sequential write required zones of
1008 		 * host-managed ZBC disks.
1009 		 */
1010 		sdkp->max_ws_blocks =
1011 			round_down(sdkp->max_ws_blocks,
1012 				   bytes_to_logical(sdkp->device,
1013 						    sdkp->physical_block_size));
1014 	}
1015 
1016 out:
1017 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1018 					 (logical_block_size >> 9));
1019 }
1020 
1021 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1022 {
1023 	struct request *rq = scsi_cmd_to_rq(cmd);
1024 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1025 
1026 	/* flush requests don't perform I/O, zero the S/G table */
1027 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1028 
1029 	if (cmd->device->use_16_for_sync) {
1030 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1031 		cmd->cmd_len = 16;
1032 	} else {
1033 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1034 		cmd->cmd_len = 10;
1035 	}
1036 	cmd->transfersize = 0;
1037 	cmd->allowed = sdkp->max_retries;
1038 
1039 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1040 	return BLK_STS_OK;
1041 }
1042 
1043 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1044 				       sector_t lba, unsigned int nr_blocks,
1045 				       unsigned char flags)
1046 {
1047 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1048 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1049 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1050 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1051 	cmd->cmnd[10] = flags;
1052 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1053 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1054 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1055 
1056 	return BLK_STS_OK;
1057 }
1058 
1059 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1060 				       sector_t lba, unsigned int nr_blocks,
1061 				       unsigned char flags)
1062 {
1063 	cmd->cmd_len  = 16;
1064 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1065 	cmd->cmnd[1]  = flags;
1066 	cmd->cmnd[14] = 0;
1067 	cmd->cmnd[15] = 0;
1068 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1069 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1070 
1071 	return BLK_STS_OK;
1072 }
1073 
1074 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1075 				       sector_t lba, unsigned int nr_blocks,
1076 				       unsigned char flags)
1077 {
1078 	cmd->cmd_len = 10;
1079 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1080 	cmd->cmnd[1] = flags;
1081 	cmd->cmnd[6] = 0;
1082 	cmd->cmnd[9] = 0;
1083 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1084 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1085 
1086 	return BLK_STS_OK;
1087 }
1088 
1089 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1090 				      sector_t lba, unsigned int nr_blocks,
1091 				      unsigned char flags)
1092 {
1093 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1094 	if (WARN_ON_ONCE(nr_blocks == 0))
1095 		return BLK_STS_IOERR;
1096 
1097 	if (unlikely(flags & 0x8)) {
1098 		/*
1099 		 * This happens only if this drive failed 10byte rw
1100 		 * command with ILLEGAL_REQUEST during operation and
1101 		 * thus turned off use_10_for_rw.
1102 		 */
1103 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1104 		return BLK_STS_IOERR;
1105 	}
1106 
1107 	cmd->cmd_len = 6;
1108 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1109 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1110 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1111 	cmd->cmnd[3] = lba & 0xff;
1112 	cmd->cmnd[4] = nr_blocks;
1113 	cmd->cmnd[5] = 0;
1114 
1115 	return BLK_STS_OK;
1116 }
1117 
1118 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1119 {
1120 	struct request *rq = scsi_cmd_to_rq(cmd);
1121 	struct scsi_device *sdp = cmd->device;
1122 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1123 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1124 	sector_t threshold;
1125 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1126 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1127 	bool write = rq_data_dir(rq) == WRITE;
1128 	unsigned char protect, fua;
1129 	blk_status_t ret;
1130 	unsigned int dif;
1131 	bool dix;
1132 
1133 	ret = scsi_alloc_sgtables(cmd);
1134 	if (ret != BLK_STS_OK)
1135 		return ret;
1136 
1137 	ret = BLK_STS_IOERR;
1138 	if (!scsi_device_online(sdp) || sdp->changed) {
1139 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1140 		goto fail;
1141 	}
1142 
1143 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1144 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1145 		goto fail;
1146 	}
1147 
1148 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1149 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1150 		goto fail;
1151 	}
1152 
1153 	/*
1154 	 * Some SD card readers can't handle accesses which touch the
1155 	 * last one or two logical blocks. Split accesses as needed.
1156 	 */
1157 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1158 
1159 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1160 		if (lba < threshold) {
1161 			/* Access up to the threshold but not beyond */
1162 			nr_blocks = threshold - lba;
1163 		} else {
1164 			/* Access only a single logical block */
1165 			nr_blocks = 1;
1166 		}
1167 	}
1168 
1169 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1170 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1171 		if (ret)
1172 			goto fail;
1173 	}
1174 
1175 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1176 	dix = scsi_prot_sg_count(cmd);
1177 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1178 
1179 	if (dif || dix)
1180 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1181 	else
1182 		protect = 0;
1183 
1184 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1185 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1186 					 protect | fua);
1187 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1188 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1189 					 protect | fua);
1190 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1191 		   sdp->use_10_for_rw || protect) {
1192 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1193 					 protect | fua);
1194 	} else {
1195 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1196 					protect | fua);
1197 	}
1198 
1199 	if (unlikely(ret != BLK_STS_OK))
1200 		goto fail;
1201 
1202 	/*
1203 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1204 	 * host adapter, it's safe to assume that we can at least transfer
1205 	 * this many bytes between each connect / disconnect.
1206 	 */
1207 	cmd->transfersize = sdp->sector_size;
1208 	cmd->underflow = nr_blocks << 9;
1209 	cmd->allowed = sdkp->max_retries;
1210 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1211 
1212 	SCSI_LOG_HLQUEUE(1,
1213 			 scmd_printk(KERN_INFO, cmd,
1214 				     "%s: block=%llu, count=%d\n", __func__,
1215 				     (unsigned long long)blk_rq_pos(rq),
1216 				     blk_rq_sectors(rq)));
1217 	SCSI_LOG_HLQUEUE(2,
1218 			 scmd_printk(KERN_INFO, cmd,
1219 				     "%s %d/%u 512 byte blocks.\n",
1220 				     write ? "writing" : "reading", nr_blocks,
1221 				     blk_rq_sectors(rq)));
1222 
1223 	/*
1224 	 * This indicates that the command is ready from our end to be queued.
1225 	 */
1226 	return BLK_STS_OK;
1227 fail:
1228 	scsi_free_sgtables(cmd);
1229 	return ret;
1230 }
1231 
1232 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1233 {
1234 	struct request *rq = scsi_cmd_to_rq(cmd);
1235 
1236 	switch (req_op(rq)) {
1237 	case REQ_OP_DISCARD:
1238 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1239 		case SD_LBP_UNMAP:
1240 			return sd_setup_unmap_cmnd(cmd);
1241 		case SD_LBP_WS16:
1242 			return sd_setup_write_same16_cmnd(cmd, true);
1243 		case SD_LBP_WS10:
1244 			return sd_setup_write_same10_cmnd(cmd, true);
1245 		case SD_LBP_ZERO:
1246 			return sd_setup_write_same10_cmnd(cmd, false);
1247 		default:
1248 			return BLK_STS_TARGET;
1249 		}
1250 	case REQ_OP_WRITE_ZEROES:
1251 		return sd_setup_write_zeroes_cmnd(cmd);
1252 	case REQ_OP_FLUSH:
1253 		return sd_setup_flush_cmnd(cmd);
1254 	case REQ_OP_READ:
1255 	case REQ_OP_WRITE:
1256 	case REQ_OP_ZONE_APPEND:
1257 		return sd_setup_read_write_cmnd(cmd);
1258 	case REQ_OP_ZONE_RESET:
1259 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1260 						   false);
1261 	case REQ_OP_ZONE_RESET_ALL:
1262 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1263 						   true);
1264 	case REQ_OP_ZONE_OPEN:
1265 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1266 	case REQ_OP_ZONE_CLOSE:
1267 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1268 	case REQ_OP_ZONE_FINISH:
1269 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1270 	default:
1271 		WARN_ON_ONCE(1);
1272 		return BLK_STS_NOTSUPP;
1273 	}
1274 }
1275 
1276 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1277 {
1278 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1279 
1280 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1281 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1282 }
1283 
1284 static bool sd_need_revalidate(struct block_device *bdev,
1285 		struct scsi_disk *sdkp)
1286 {
1287 	if (sdkp->device->removable || sdkp->write_prot) {
1288 		if (bdev_check_media_change(bdev))
1289 			return true;
1290 	}
1291 
1292 	/*
1293 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1294 	 * nothing to do with partitions, BLKRRPART is used to force a full
1295 	 * revalidate after things like a format for historical reasons.
1296 	 */
1297 	return test_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1298 }
1299 
1300 /**
1301  *	sd_open - open a scsi disk device
1302  *	@bdev: Block device of the scsi disk to open
1303  *	@mode: FMODE_* mask
1304  *
1305  *	Returns 0 if successful. Returns a negated errno value in case
1306  *	of error.
1307  *
1308  *	Note: This can be called from a user context (e.g. fsck(1) )
1309  *	or from within the kernel (e.g. as a result of a mount(1) ).
1310  *	In the latter case @inode and @filp carry an abridged amount
1311  *	of information as noted above.
1312  *
1313  *	Locking: called with bdev->bd_disk->open_mutex held.
1314  **/
1315 static int sd_open(struct block_device *bdev, fmode_t mode)
1316 {
1317 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1318 	struct scsi_device *sdev = sdkp->device;
1319 	int retval;
1320 
1321 	if (scsi_device_get(sdev))
1322 		return -ENXIO;
1323 
1324 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1325 
1326 	/*
1327 	 * If the device is in error recovery, wait until it is done.
1328 	 * If the device is offline, then disallow any access to it.
1329 	 */
1330 	retval = -ENXIO;
1331 	if (!scsi_block_when_processing_errors(sdev))
1332 		goto error_out;
1333 
1334 	if (sd_need_revalidate(bdev, sdkp))
1335 		sd_revalidate_disk(bdev->bd_disk);
1336 
1337 	/*
1338 	 * If the drive is empty, just let the open fail.
1339 	 */
1340 	retval = -ENOMEDIUM;
1341 	if (sdev->removable && !sdkp->media_present && !(mode & FMODE_NDELAY))
1342 		goto error_out;
1343 
1344 	/*
1345 	 * If the device has the write protect tab set, have the open fail
1346 	 * if the user expects to be able to write to the thing.
1347 	 */
1348 	retval = -EROFS;
1349 	if (sdkp->write_prot && (mode & FMODE_WRITE))
1350 		goto error_out;
1351 
1352 	/*
1353 	 * It is possible that the disk changing stuff resulted in
1354 	 * the device being taken offline.  If this is the case,
1355 	 * report this to the user, and don't pretend that the
1356 	 * open actually succeeded.
1357 	 */
1358 	retval = -ENXIO;
1359 	if (!scsi_device_online(sdev))
1360 		goto error_out;
1361 
1362 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1363 		if (scsi_block_when_processing_errors(sdev))
1364 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1365 	}
1366 
1367 	return 0;
1368 
1369 error_out:
1370 	scsi_device_put(sdev);
1371 	return retval;
1372 }
1373 
1374 /**
1375  *	sd_release - invoked when the (last) close(2) is called on this
1376  *	scsi disk.
1377  *	@disk: disk to release
1378  *	@mode: FMODE_* mask
1379  *
1380  *	Returns 0.
1381  *
1382  *	Note: may block (uninterruptible) if error recovery is underway
1383  *	on this disk.
1384  *
1385  *	Locking: called with bdev->bd_disk->open_mutex held.
1386  **/
1387 static void sd_release(struct gendisk *disk, fmode_t mode)
1388 {
1389 	struct scsi_disk *sdkp = scsi_disk(disk);
1390 	struct scsi_device *sdev = sdkp->device;
1391 
1392 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1393 
1394 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1395 		if (scsi_block_when_processing_errors(sdev))
1396 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1397 	}
1398 
1399 	scsi_device_put(sdev);
1400 }
1401 
1402 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1403 {
1404 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1405 	struct scsi_device *sdp = sdkp->device;
1406 	struct Scsi_Host *host = sdp->host;
1407 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1408 	int diskinfo[4];
1409 
1410 	/* default to most commonly used values */
1411 	diskinfo[0] = 0x40;	/* 1 << 6 */
1412 	diskinfo[1] = 0x20;	/* 1 << 5 */
1413 	diskinfo[2] = capacity >> 11;
1414 
1415 	/* override with calculated, extended default, or driver values */
1416 	if (host->hostt->bios_param)
1417 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1418 	else
1419 		scsicam_bios_param(bdev, capacity, diskinfo);
1420 
1421 	geo->heads = diskinfo[0];
1422 	geo->sectors = diskinfo[1];
1423 	geo->cylinders = diskinfo[2];
1424 	return 0;
1425 }
1426 
1427 /**
1428  *	sd_ioctl - process an ioctl
1429  *	@bdev: target block device
1430  *	@mode: FMODE_* mask
1431  *	@cmd: ioctl command number
1432  *	@arg: this is third argument given to ioctl(2) system call.
1433  *	Often contains a pointer.
1434  *
1435  *	Returns 0 if successful (some ioctls return positive numbers on
1436  *	success as well). Returns a negated errno value in case of error.
1437  *
1438  *	Note: most ioctls are forward onto the block subsystem or further
1439  *	down in the scsi subsystem.
1440  **/
1441 static int sd_ioctl(struct block_device *bdev, fmode_t mode,
1442 		    unsigned int cmd, unsigned long arg)
1443 {
1444 	struct gendisk *disk = bdev->bd_disk;
1445 	struct scsi_disk *sdkp = scsi_disk(disk);
1446 	struct scsi_device *sdp = sdkp->device;
1447 	void __user *p = (void __user *)arg;
1448 	int error;
1449 
1450 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1451 				    "cmd=0x%x\n", disk->disk_name, cmd));
1452 
1453 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1454 		return -ENOIOCTLCMD;
1455 
1456 	/*
1457 	 * If we are in the middle of error recovery, don't let anyone
1458 	 * else try and use this device.  Also, if error recovery fails, it
1459 	 * may try and take the device offline, in which case all further
1460 	 * access to the device is prohibited.
1461 	 */
1462 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1463 			(mode & FMODE_NDELAY) != 0);
1464 	if (error)
1465 		return error;
1466 
1467 	if (is_sed_ioctl(cmd))
1468 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1469 	return scsi_ioctl(sdp, mode, cmd, p);
1470 }
1471 
1472 static void set_media_not_present(struct scsi_disk *sdkp)
1473 {
1474 	if (sdkp->media_present)
1475 		sdkp->device->changed = 1;
1476 
1477 	if (sdkp->device->removable) {
1478 		sdkp->media_present = 0;
1479 		sdkp->capacity = 0;
1480 	}
1481 }
1482 
1483 static int media_not_present(struct scsi_disk *sdkp,
1484 			     struct scsi_sense_hdr *sshdr)
1485 {
1486 	if (!scsi_sense_valid(sshdr))
1487 		return 0;
1488 
1489 	/* not invoked for commands that could return deferred errors */
1490 	switch (sshdr->sense_key) {
1491 	case UNIT_ATTENTION:
1492 	case NOT_READY:
1493 		/* medium not present */
1494 		if (sshdr->asc == 0x3A) {
1495 			set_media_not_present(sdkp);
1496 			return 1;
1497 		}
1498 	}
1499 	return 0;
1500 }
1501 
1502 /**
1503  *	sd_check_events - check media events
1504  *	@disk: kernel device descriptor
1505  *	@clearing: disk events currently being cleared
1506  *
1507  *	Returns mask of DISK_EVENT_*.
1508  *
1509  *	Note: this function is invoked from the block subsystem.
1510  **/
1511 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1512 {
1513 	struct scsi_disk *sdkp = disk->private_data;
1514 	struct scsi_device *sdp;
1515 	int retval;
1516 	bool disk_changed;
1517 
1518 	if (!sdkp)
1519 		return 0;
1520 
1521 	sdp = sdkp->device;
1522 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1523 
1524 	/*
1525 	 * If the device is offline, don't send any commands - just pretend as
1526 	 * if the command failed.  If the device ever comes back online, we
1527 	 * can deal with it then.  It is only because of unrecoverable errors
1528 	 * that we would ever take a device offline in the first place.
1529 	 */
1530 	if (!scsi_device_online(sdp)) {
1531 		set_media_not_present(sdkp);
1532 		goto out;
1533 	}
1534 
1535 	/*
1536 	 * Using TEST_UNIT_READY enables differentiation between drive with
1537 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1538 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1539 	 *
1540 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1541 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1542 	 * sd_revalidate() is called.
1543 	 */
1544 	if (scsi_block_when_processing_errors(sdp)) {
1545 		struct scsi_sense_hdr sshdr = { 0, };
1546 
1547 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1548 					      &sshdr);
1549 
1550 		/* failed to execute TUR, assume media not present */
1551 		if (retval < 0 || host_byte(retval)) {
1552 			set_media_not_present(sdkp);
1553 			goto out;
1554 		}
1555 
1556 		if (media_not_present(sdkp, &sshdr))
1557 			goto out;
1558 	}
1559 
1560 	/*
1561 	 * For removable scsi disk we have to recognise the presence
1562 	 * of a disk in the drive.
1563 	 */
1564 	if (!sdkp->media_present)
1565 		sdp->changed = 1;
1566 	sdkp->media_present = 1;
1567 out:
1568 	/*
1569 	 * sdp->changed is set under the following conditions:
1570 	 *
1571 	 *	Medium present state has changed in either direction.
1572 	 *	Device has indicated UNIT_ATTENTION.
1573 	 */
1574 	disk_changed = sdp->changed;
1575 	sdp->changed = 0;
1576 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1577 }
1578 
1579 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1580 {
1581 	int retries, res;
1582 	struct scsi_device *sdp = sdkp->device;
1583 	const int timeout = sdp->request_queue->rq_timeout
1584 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1585 	struct scsi_sense_hdr my_sshdr;
1586 
1587 	if (!scsi_device_online(sdp))
1588 		return -ENODEV;
1589 
1590 	/* caller might not be interested in sense, but we need it */
1591 	if (!sshdr)
1592 		sshdr = &my_sshdr;
1593 
1594 	for (retries = 3; retries > 0; --retries) {
1595 		unsigned char cmd[16] = { 0 };
1596 
1597 		if (sdp->use_16_for_sync)
1598 			cmd[0] = SYNCHRONIZE_CACHE_16;
1599 		else
1600 			cmd[0] = SYNCHRONIZE_CACHE;
1601 		/*
1602 		 * Leave the rest of the command zero to indicate
1603 		 * flush everything.
1604 		 */
1605 		res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, sshdr,
1606 				timeout, sdkp->max_retries, 0, RQF_PM, NULL);
1607 		if (res == 0)
1608 			break;
1609 	}
1610 
1611 	if (res) {
1612 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1613 
1614 		if (res < 0)
1615 			return res;
1616 
1617 		if (scsi_status_is_check_condition(res) &&
1618 		    scsi_sense_valid(sshdr)) {
1619 			sd_print_sense_hdr(sdkp, sshdr);
1620 
1621 			/* we need to evaluate the error return  */
1622 			if (sshdr->asc == 0x3a ||	/* medium not present */
1623 			    sshdr->asc == 0x20 ||	/* invalid command */
1624 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1625 				/* this is no error here */
1626 				return 0;
1627 		}
1628 
1629 		switch (host_byte(res)) {
1630 		/* ignore errors due to racing a disconnection */
1631 		case DID_BAD_TARGET:
1632 		case DID_NO_CONNECT:
1633 			return 0;
1634 		/* signal the upper layer it might try again */
1635 		case DID_BUS_BUSY:
1636 		case DID_IMM_RETRY:
1637 		case DID_REQUEUE:
1638 		case DID_SOFT_ERROR:
1639 			return -EBUSY;
1640 		default:
1641 			return -EIO;
1642 		}
1643 	}
1644 	return 0;
1645 }
1646 
1647 static void sd_rescan(struct device *dev)
1648 {
1649 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1650 
1651 	sd_revalidate_disk(sdkp->disk);
1652 }
1653 
1654 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1655 		enum blk_unique_id type)
1656 {
1657 	struct scsi_device *sdev = scsi_disk(disk)->device;
1658 	const struct scsi_vpd *vpd;
1659 	const unsigned char *d;
1660 	int ret = -ENXIO, len;
1661 
1662 	rcu_read_lock();
1663 	vpd = rcu_dereference(sdev->vpd_pg83);
1664 	if (!vpd)
1665 		goto out_unlock;
1666 
1667 	ret = -EINVAL;
1668 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1669 		/* we only care about designators with LU association */
1670 		if (((d[1] >> 4) & 0x3) != 0x00)
1671 			continue;
1672 		if ((d[1] & 0xf) != type)
1673 			continue;
1674 
1675 		/*
1676 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1677 		 * keep looking as one with more entropy might still show up.
1678 		 */
1679 		len = d[3];
1680 		if (len != 8 && len != 12 && len != 16)
1681 			continue;
1682 		ret = len;
1683 		memcpy(id, d + 4, len);
1684 		if (len == 16)
1685 			break;
1686 	}
1687 out_unlock:
1688 	rcu_read_unlock();
1689 	return ret;
1690 }
1691 
1692 static char sd_pr_type(enum pr_type type)
1693 {
1694 	switch (type) {
1695 	case PR_WRITE_EXCLUSIVE:
1696 		return 0x01;
1697 	case PR_EXCLUSIVE_ACCESS:
1698 		return 0x03;
1699 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1700 		return 0x05;
1701 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1702 		return 0x06;
1703 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1704 		return 0x07;
1705 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1706 		return 0x08;
1707 	default:
1708 		return 0;
1709 	}
1710 };
1711 
1712 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1713 {
1714 	switch (host_byte(result)) {
1715 	case DID_TRANSPORT_MARGINAL:
1716 	case DID_TRANSPORT_DISRUPTED:
1717 	case DID_BUS_BUSY:
1718 		return PR_STS_RETRY_PATH_FAILURE;
1719 	case DID_NO_CONNECT:
1720 		return PR_STS_PATH_FAILED;
1721 	case DID_TRANSPORT_FAILFAST:
1722 		return PR_STS_PATH_FAST_FAILED;
1723 	}
1724 
1725 	switch (status_byte(result)) {
1726 	case SAM_STAT_RESERVATION_CONFLICT:
1727 		return PR_STS_RESERVATION_CONFLICT;
1728 	case SAM_STAT_CHECK_CONDITION:
1729 		if (!scsi_sense_valid(sshdr))
1730 			return PR_STS_IOERR;
1731 
1732 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1733 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1734 			return -EINVAL;
1735 
1736 		fallthrough;
1737 	default:
1738 		return PR_STS_IOERR;
1739 	}
1740 }
1741 
1742 static int sd_pr_command(struct block_device *bdev, u8 sa,
1743 		u64 key, u64 sa_key, u8 type, u8 flags)
1744 {
1745 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1746 	struct scsi_device *sdev = sdkp->device;
1747 	struct scsi_sense_hdr sshdr;
1748 	int result;
1749 	u8 cmd[16] = { 0, };
1750 	u8 data[24] = { 0, };
1751 
1752 	cmd[0] = PERSISTENT_RESERVE_OUT;
1753 	cmd[1] = sa;
1754 	cmd[2] = type;
1755 	put_unaligned_be32(sizeof(data), &cmd[5]);
1756 
1757 	put_unaligned_be64(key, &data[0]);
1758 	put_unaligned_be64(sa_key, &data[8]);
1759 	data[20] = flags;
1760 
1761 	result = scsi_execute_req(sdev, cmd, DMA_TO_DEVICE, &data, sizeof(data),
1762 			&sshdr, SD_TIMEOUT, sdkp->max_retries, NULL);
1763 
1764 	if (scsi_status_is_check_condition(result) &&
1765 	    scsi_sense_valid(&sshdr)) {
1766 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1767 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1768 	}
1769 
1770 	if (result <= 0)
1771 		return result;
1772 
1773 	return sd_scsi_to_pr_err(&sshdr, result);
1774 }
1775 
1776 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1777 		u32 flags)
1778 {
1779 	if (flags & ~PR_FL_IGNORE_KEY)
1780 		return -EOPNOTSUPP;
1781 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1782 			old_key, new_key, 0,
1783 			(1 << 0) /* APTPL */);
1784 }
1785 
1786 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1787 		u32 flags)
1788 {
1789 	if (flags)
1790 		return -EOPNOTSUPP;
1791 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1792 }
1793 
1794 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1795 {
1796 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1797 }
1798 
1799 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1800 		enum pr_type type, bool abort)
1801 {
1802 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1803 			     sd_pr_type(type), 0);
1804 }
1805 
1806 static int sd_pr_clear(struct block_device *bdev, u64 key)
1807 {
1808 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1809 }
1810 
1811 static const struct pr_ops sd_pr_ops = {
1812 	.pr_register	= sd_pr_register,
1813 	.pr_reserve	= sd_pr_reserve,
1814 	.pr_release	= sd_pr_release,
1815 	.pr_preempt	= sd_pr_preempt,
1816 	.pr_clear	= sd_pr_clear,
1817 };
1818 
1819 static void scsi_disk_free_disk(struct gendisk *disk)
1820 {
1821 	struct scsi_disk *sdkp = scsi_disk(disk);
1822 
1823 	put_device(&sdkp->disk_dev);
1824 }
1825 
1826 static const struct block_device_operations sd_fops = {
1827 	.owner			= THIS_MODULE,
1828 	.open			= sd_open,
1829 	.release		= sd_release,
1830 	.ioctl			= sd_ioctl,
1831 	.getgeo			= sd_getgeo,
1832 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1833 	.check_events		= sd_check_events,
1834 	.unlock_native_capacity	= sd_unlock_native_capacity,
1835 	.report_zones		= sd_zbc_report_zones,
1836 	.get_unique_id		= sd_get_unique_id,
1837 	.free_disk		= scsi_disk_free_disk,
1838 	.pr_ops			= &sd_pr_ops,
1839 };
1840 
1841 /**
1842  *	sd_eh_reset - reset error handling callback
1843  *	@scmd:		sd-issued command that has failed
1844  *
1845  *	This function is called by the SCSI midlayer before starting
1846  *	SCSI EH. When counting medium access failures we have to be
1847  *	careful to register it only only once per device and SCSI EH run;
1848  *	there might be several timed out commands which will cause the
1849  *	'max_medium_access_timeouts' counter to trigger after the first
1850  *	SCSI EH run already and set the device to offline.
1851  *	So this function resets the internal counter before starting SCSI EH.
1852  **/
1853 static void sd_eh_reset(struct scsi_cmnd *scmd)
1854 {
1855 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1856 
1857 	/* New SCSI EH run, reset gate variable */
1858 	sdkp->ignore_medium_access_errors = false;
1859 }
1860 
1861 /**
1862  *	sd_eh_action - error handling callback
1863  *	@scmd:		sd-issued command that has failed
1864  *	@eh_disp:	The recovery disposition suggested by the midlayer
1865  *
1866  *	This function is called by the SCSI midlayer upon completion of an
1867  *	error test command (currently TEST UNIT READY). The result of sending
1868  *	the eh command is passed in eh_disp.  We're looking for devices that
1869  *	fail medium access commands but are OK with non access commands like
1870  *	test unit ready (so wrongly see the device as having a successful
1871  *	recovery)
1872  **/
1873 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1874 {
1875 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1876 	struct scsi_device *sdev = scmd->device;
1877 
1878 	if (!scsi_device_online(sdev) ||
1879 	    !scsi_medium_access_command(scmd) ||
1880 	    host_byte(scmd->result) != DID_TIME_OUT ||
1881 	    eh_disp != SUCCESS)
1882 		return eh_disp;
1883 
1884 	/*
1885 	 * The device has timed out executing a medium access command.
1886 	 * However, the TEST UNIT READY command sent during error
1887 	 * handling completed successfully. Either the device is in the
1888 	 * process of recovering or has it suffered an internal failure
1889 	 * that prevents access to the storage medium.
1890 	 */
1891 	if (!sdkp->ignore_medium_access_errors) {
1892 		sdkp->medium_access_timed_out++;
1893 		sdkp->ignore_medium_access_errors = true;
1894 	}
1895 
1896 	/*
1897 	 * If the device keeps failing read/write commands but TEST UNIT
1898 	 * READY always completes successfully we assume that medium
1899 	 * access is no longer possible and take the device offline.
1900 	 */
1901 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1902 		scmd_printk(KERN_ERR, scmd,
1903 			    "Medium access timeout failure. Offlining disk!\n");
1904 		mutex_lock(&sdev->state_mutex);
1905 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1906 		mutex_unlock(&sdev->state_mutex);
1907 
1908 		return SUCCESS;
1909 	}
1910 
1911 	return eh_disp;
1912 }
1913 
1914 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1915 {
1916 	struct request *req = scsi_cmd_to_rq(scmd);
1917 	struct scsi_device *sdev = scmd->device;
1918 	unsigned int transferred, good_bytes;
1919 	u64 start_lba, end_lba, bad_lba;
1920 
1921 	/*
1922 	 * Some commands have a payload smaller than the device logical
1923 	 * block size (e.g. INQUIRY on a 4K disk).
1924 	 */
1925 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1926 		return 0;
1927 
1928 	/* Check if we have a 'bad_lba' information */
1929 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1930 				     SCSI_SENSE_BUFFERSIZE,
1931 				     &bad_lba))
1932 		return 0;
1933 
1934 	/*
1935 	 * If the bad lba was reported incorrectly, we have no idea where
1936 	 * the error is.
1937 	 */
1938 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1939 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1940 	if (bad_lba < start_lba || bad_lba >= end_lba)
1941 		return 0;
1942 
1943 	/*
1944 	 * resid is optional but mostly filled in.  When it's unused,
1945 	 * its value is zero, so we assume the whole buffer transferred
1946 	 */
1947 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1948 
1949 	/* This computation should always be done in terms of the
1950 	 * resolution of the device's medium.
1951 	 */
1952 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1953 
1954 	return min(good_bytes, transferred);
1955 }
1956 
1957 /**
1958  *	sd_done - bottom half handler: called when the lower level
1959  *	driver has completed (successfully or otherwise) a scsi command.
1960  *	@SCpnt: mid-level's per command structure.
1961  *
1962  *	Note: potentially run from within an ISR. Must not block.
1963  **/
1964 static int sd_done(struct scsi_cmnd *SCpnt)
1965 {
1966 	int result = SCpnt->result;
1967 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1968 	unsigned int sector_size = SCpnt->device->sector_size;
1969 	unsigned int resid;
1970 	struct scsi_sense_hdr sshdr;
1971 	struct request *req = scsi_cmd_to_rq(SCpnt);
1972 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1973 	int sense_valid = 0;
1974 	int sense_deferred = 0;
1975 
1976 	switch (req_op(req)) {
1977 	case REQ_OP_DISCARD:
1978 	case REQ_OP_WRITE_ZEROES:
1979 	case REQ_OP_ZONE_RESET:
1980 	case REQ_OP_ZONE_RESET_ALL:
1981 	case REQ_OP_ZONE_OPEN:
1982 	case REQ_OP_ZONE_CLOSE:
1983 	case REQ_OP_ZONE_FINISH:
1984 		if (!result) {
1985 			good_bytes = blk_rq_bytes(req);
1986 			scsi_set_resid(SCpnt, 0);
1987 		} else {
1988 			good_bytes = 0;
1989 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1990 		}
1991 		break;
1992 	default:
1993 		/*
1994 		 * In case of bogus fw or device, we could end up having
1995 		 * an unaligned partial completion. Check this here and force
1996 		 * alignment.
1997 		 */
1998 		resid = scsi_get_resid(SCpnt);
1999 		if (resid & (sector_size - 1)) {
2000 			sd_printk(KERN_INFO, sdkp,
2001 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2002 				resid, sector_size);
2003 			scsi_print_command(SCpnt);
2004 			resid = min(scsi_bufflen(SCpnt),
2005 				    round_up(resid, sector_size));
2006 			scsi_set_resid(SCpnt, resid);
2007 		}
2008 	}
2009 
2010 	if (result) {
2011 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2012 		if (sense_valid)
2013 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2014 	}
2015 	sdkp->medium_access_timed_out = 0;
2016 
2017 	if (!scsi_status_is_check_condition(result) &&
2018 	    (!sense_valid || sense_deferred))
2019 		goto out;
2020 
2021 	switch (sshdr.sense_key) {
2022 	case HARDWARE_ERROR:
2023 	case MEDIUM_ERROR:
2024 		good_bytes = sd_completed_bytes(SCpnt);
2025 		break;
2026 	case RECOVERED_ERROR:
2027 		good_bytes = scsi_bufflen(SCpnt);
2028 		break;
2029 	case NO_SENSE:
2030 		/* This indicates a false check condition, so ignore it.  An
2031 		 * unknown amount of data was transferred so treat it as an
2032 		 * error.
2033 		 */
2034 		SCpnt->result = 0;
2035 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2036 		break;
2037 	case ABORTED_COMMAND:
2038 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2039 			good_bytes = sd_completed_bytes(SCpnt);
2040 		break;
2041 	case ILLEGAL_REQUEST:
2042 		switch (sshdr.asc) {
2043 		case 0x10:	/* DIX: Host detected corruption */
2044 			good_bytes = sd_completed_bytes(SCpnt);
2045 			break;
2046 		case 0x20:	/* INVALID COMMAND OPCODE */
2047 		case 0x24:	/* INVALID FIELD IN CDB */
2048 			switch (SCpnt->cmnd[0]) {
2049 			case UNMAP:
2050 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2051 				break;
2052 			case WRITE_SAME_16:
2053 			case WRITE_SAME:
2054 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2055 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2056 				} else {
2057 					sdkp->device->no_write_same = 1;
2058 					sd_config_write_same(sdkp);
2059 					req->rq_flags |= RQF_QUIET;
2060 				}
2061 				break;
2062 			}
2063 		}
2064 		break;
2065 	default:
2066 		break;
2067 	}
2068 
2069  out:
2070 	if (sd_is_zoned(sdkp))
2071 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2072 
2073 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2074 					   "sd_done: completed %d of %d bytes\n",
2075 					   good_bytes, scsi_bufflen(SCpnt)));
2076 
2077 	return good_bytes;
2078 }
2079 
2080 /*
2081  * spinup disk - called only in sd_revalidate_disk()
2082  */
2083 static void
2084 sd_spinup_disk(struct scsi_disk *sdkp)
2085 {
2086 	unsigned char cmd[10];
2087 	unsigned long spintime_expire = 0;
2088 	int retries, spintime;
2089 	unsigned int the_result;
2090 	struct scsi_sense_hdr sshdr;
2091 	int sense_valid = 0;
2092 
2093 	spintime = 0;
2094 
2095 	/* Spin up drives, as required.  Only do this at boot time */
2096 	/* Spinup needs to be done for module loads too. */
2097 	do {
2098 		retries = 0;
2099 
2100 		do {
2101 			bool media_was_present = sdkp->media_present;
2102 
2103 			cmd[0] = TEST_UNIT_READY;
2104 			memset((void *) &cmd[1], 0, 9);
2105 
2106 			the_result = scsi_execute_req(sdkp->device, cmd,
2107 						      DMA_NONE, NULL, 0,
2108 						      &sshdr, SD_TIMEOUT,
2109 						      sdkp->max_retries, NULL);
2110 
2111 			/*
2112 			 * If the drive has indicated to us that it
2113 			 * doesn't have any media in it, don't bother
2114 			 * with any more polling.
2115 			 */
2116 			if (media_not_present(sdkp, &sshdr)) {
2117 				if (media_was_present)
2118 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2119 				return;
2120 			}
2121 
2122 			if (the_result)
2123 				sense_valid = scsi_sense_valid(&sshdr);
2124 			retries++;
2125 		} while (retries < 3 &&
2126 			 (!scsi_status_is_good(the_result) ||
2127 			  (scsi_status_is_check_condition(the_result) &&
2128 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2129 
2130 		if (!scsi_status_is_check_condition(the_result)) {
2131 			/* no sense, TUR either succeeded or failed
2132 			 * with a status error */
2133 			if(!spintime && !scsi_status_is_good(the_result)) {
2134 				sd_print_result(sdkp, "Test Unit Ready failed",
2135 						the_result);
2136 			}
2137 			break;
2138 		}
2139 
2140 		/*
2141 		 * The device does not want the automatic start to be issued.
2142 		 */
2143 		if (sdkp->device->no_start_on_add)
2144 			break;
2145 
2146 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2147 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2148 				break;	/* manual intervention required */
2149 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2150 				break;	/* standby */
2151 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2152 				break;	/* unavailable */
2153 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2154 				break;	/* sanitize in progress */
2155 			/*
2156 			 * Issue command to spin up drive when not ready
2157 			 */
2158 			if (!spintime) {
2159 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2160 				cmd[0] = START_STOP;
2161 				cmd[1] = 1;	/* Return immediately */
2162 				memset((void *) &cmd[2], 0, 8);
2163 				cmd[4] = 1;	/* Start spin cycle */
2164 				if (sdkp->device->start_stop_pwr_cond)
2165 					cmd[4] |= 1 << 4;
2166 				scsi_execute_req(sdkp->device, cmd, DMA_NONE,
2167 						 NULL, 0, &sshdr,
2168 						 SD_TIMEOUT, sdkp->max_retries,
2169 						 NULL);
2170 				spintime_expire = jiffies + 100 * HZ;
2171 				spintime = 1;
2172 			}
2173 			/* Wait 1 second for next try */
2174 			msleep(1000);
2175 			printk(KERN_CONT ".");
2176 
2177 		/*
2178 		 * Wait for USB flash devices with slow firmware.
2179 		 * Yes, this sense key/ASC combination shouldn't
2180 		 * occur here.  It's characteristic of these devices.
2181 		 */
2182 		} else if (sense_valid &&
2183 				sshdr.sense_key == UNIT_ATTENTION &&
2184 				sshdr.asc == 0x28) {
2185 			if (!spintime) {
2186 				spintime_expire = jiffies + 5 * HZ;
2187 				spintime = 1;
2188 			}
2189 			/* Wait 1 second for next try */
2190 			msleep(1000);
2191 		} else {
2192 			/* we don't understand the sense code, so it's
2193 			 * probably pointless to loop */
2194 			if(!spintime) {
2195 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2196 				sd_print_sense_hdr(sdkp, &sshdr);
2197 			}
2198 			break;
2199 		}
2200 
2201 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2202 
2203 	if (spintime) {
2204 		if (scsi_status_is_good(the_result))
2205 			printk(KERN_CONT "ready\n");
2206 		else
2207 			printk(KERN_CONT "not responding...\n");
2208 	}
2209 }
2210 
2211 /*
2212  * Determine whether disk supports Data Integrity Field.
2213  */
2214 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2215 {
2216 	struct scsi_device *sdp = sdkp->device;
2217 	u8 type;
2218 
2219 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2220 		sdkp->protection_type = 0;
2221 		return 0;
2222 	}
2223 
2224 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2225 
2226 	if (type > T10_PI_TYPE3_PROTECTION) {
2227 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2228 			  " protection type %u. Disabling disk!\n",
2229 			  type);
2230 		sdkp->protection_type = 0;
2231 		return -ENODEV;
2232 	}
2233 
2234 	sdkp->protection_type = type;
2235 
2236 	return 0;
2237 }
2238 
2239 static void sd_config_protection(struct scsi_disk *sdkp)
2240 {
2241 	struct scsi_device *sdp = sdkp->device;
2242 
2243 	if (!sdkp->first_scan)
2244 		return;
2245 
2246 	sd_dif_config_host(sdkp);
2247 
2248 	if (!sdkp->protection_type)
2249 		return;
2250 
2251 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2252 		sd_printk(KERN_NOTICE, sdkp,
2253 			  "Disabling DIF Type %u protection\n",
2254 			  sdkp->protection_type);
2255 		sdkp->protection_type = 0;
2256 	}
2257 
2258 	sd_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2259 		  sdkp->protection_type);
2260 }
2261 
2262 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2263 			struct scsi_sense_hdr *sshdr, int sense_valid,
2264 			int the_result)
2265 {
2266 	if (sense_valid)
2267 		sd_print_sense_hdr(sdkp, sshdr);
2268 	else
2269 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2270 
2271 	/*
2272 	 * Set dirty bit for removable devices if not ready -
2273 	 * sometimes drives will not report this properly.
2274 	 */
2275 	if (sdp->removable &&
2276 	    sense_valid && sshdr->sense_key == NOT_READY)
2277 		set_media_not_present(sdkp);
2278 
2279 	/*
2280 	 * We used to set media_present to 0 here to indicate no media
2281 	 * in the drive, but some drives fail read capacity even with
2282 	 * media present, so we can't do that.
2283 	 */
2284 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2285 }
2286 
2287 #define RC16_LEN 32
2288 #if RC16_LEN > SD_BUF_SIZE
2289 #error RC16_LEN must not be more than SD_BUF_SIZE
2290 #endif
2291 
2292 #define READ_CAPACITY_RETRIES_ON_RESET	10
2293 
2294 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2295 						unsigned char *buffer)
2296 {
2297 	unsigned char cmd[16];
2298 	struct scsi_sense_hdr sshdr;
2299 	int sense_valid = 0;
2300 	int the_result;
2301 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2302 	unsigned int alignment;
2303 	unsigned long long lba;
2304 	unsigned sector_size;
2305 
2306 	if (sdp->no_read_capacity_16)
2307 		return -EINVAL;
2308 
2309 	do {
2310 		memset(cmd, 0, 16);
2311 		cmd[0] = SERVICE_ACTION_IN_16;
2312 		cmd[1] = SAI_READ_CAPACITY_16;
2313 		cmd[13] = RC16_LEN;
2314 		memset(buffer, 0, RC16_LEN);
2315 
2316 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2317 					buffer, RC16_LEN, &sshdr,
2318 					SD_TIMEOUT, sdkp->max_retries, NULL);
2319 
2320 		if (media_not_present(sdkp, &sshdr))
2321 			return -ENODEV;
2322 
2323 		if (the_result > 0) {
2324 			sense_valid = scsi_sense_valid(&sshdr);
2325 			if (sense_valid &&
2326 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2327 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2328 			    sshdr.ascq == 0x00)
2329 				/* Invalid Command Operation Code or
2330 				 * Invalid Field in CDB, just retry
2331 				 * silently with RC10 */
2332 				return -EINVAL;
2333 			if (sense_valid &&
2334 			    sshdr.sense_key == UNIT_ATTENTION &&
2335 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2336 				/* Device reset might occur several times,
2337 				 * give it one more chance */
2338 				if (--reset_retries > 0)
2339 					continue;
2340 		}
2341 		retries--;
2342 
2343 	} while (the_result && retries);
2344 
2345 	if (the_result) {
2346 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2347 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2348 		return -EINVAL;
2349 	}
2350 
2351 	sector_size = get_unaligned_be32(&buffer[8]);
2352 	lba = get_unaligned_be64(&buffer[0]);
2353 
2354 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2355 		sdkp->capacity = 0;
2356 		return -ENODEV;
2357 	}
2358 
2359 	/* Logical blocks per physical block exponent */
2360 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2361 
2362 	/* RC basis */
2363 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2364 
2365 	/* Lowest aligned logical block */
2366 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2367 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2368 	if (alignment && sdkp->first_scan)
2369 		sd_printk(KERN_NOTICE, sdkp,
2370 			  "physical block alignment offset: %u\n", alignment);
2371 
2372 	if (buffer[14] & 0x80) { /* LBPME */
2373 		sdkp->lbpme = 1;
2374 
2375 		if (buffer[14] & 0x40) /* LBPRZ */
2376 			sdkp->lbprz = 1;
2377 
2378 		sd_config_discard(sdkp, SD_LBP_WS16);
2379 	}
2380 
2381 	sdkp->capacity = lba + 1;
2382 	return sector_size;
2383 }
2384 
2385 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2386 						unsigned char *buffer)
2387 {
2388 	unsigned char cmd[16];
2389 	struct scsi_sense_hdr sshdr;
2390 	int sense_valid = 0;
2391 	int the_result;
2392 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2393 	sector_t lba;
2394 	unsigned sector_size;
2395 
2396 	do {
2397 		cmd[0] = READ_CAPACITY;
2398 		memset(&cmd[1], 0, 9);
2399 		memset(buffer, 0, 8);
2400 
2401 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2402 					buffer, 8, &sshdr,
2403 					SD_TIMEOUT, sdkp->max_retries, NULL);
2404 
2405 		if (media_not_present(sdkp, &sshdr))
2406 			return -ENODEV;
2407 
2408 		if (the_result > 0) {
2409 			sense_valid = scsi_sense_valid(&sshdr);
2410 			if (sense_valid &&
2411 			    sshdr.sense_key == UNIT_ATTENTION &&
2412 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2413 				/* Device reset might occur several times,
2414 				 * give it one more chance */
2415 				if (--reset_retries > 0)
2416 					continue;
2417 		}
2418 		retries--;
2419 
2420 	} while (the_result && retries);
2421 
2422 	if (the_result) {
2423 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2424 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2425 		return -EINVAL;
2426 	}
2427 
2428 	sector_size = get_unaligned_be32(&buffer[4]);
2429 	lba = get_unaligned_be32(&buffer[0]);
2430 
2431 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2432 		/* Some buggy (usb cardreader) devices return an lba of
2433 		   0xffffffff when the want to report a size of 0 (with
2434 		   which they really mean no media is present) */
2435 		sdkp->capacity = 0;
2436 		sdkp->physical_block_size = sector_size;
2437 		return sector_size;
2438 	}
2439 
2440 	sdkp->capacity = lba + 1;
2441 	sdkp->physical_block_size = sector_size;
2442 	return sector_size;
2443 }
2444 
2445 static int sd_try_rc16_first(struct scsi_device *sdp)
2446 {
2447 	if (sdp->host->max_cmd_len < 16)
2448 		return 0;
2449 	if (sdp->try_rc_10_first)
2450 		return 0;
2451 	if (sdp->scsi_level > SCSI_SPC_2)
2452 		return 1;
2453 	if (scsi_device_protection(sdp))
2454 		return 1;
2455 	return 0;
2456 }
2457 
2458 /*
2459  * read disk capacity
2460  */
2461 static void
2462 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2463 {
2464 	int sector_size;
2465 	struct scsi_device *sdp = sdkp->device;
2466 
2467 	if (sd_try_rc16_first(sdp)) {
2468 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2469 		if (sector_size == -EOVERFLOW)
2470 			goto got_data;
2471 		if (sector_size == -ENODEV)
2472 			return;
2473 		if (sector_size < 0)
2474 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2475 		if (sector_size < 0)
2476 			return;
2477 	} else {
2478 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2479 		if (sector_size == -EOVERFLOW)
2480 			goto got_data;
2481 		if (sector_size < 0)
2482 			return;
2483 		if ((sizeof(sdkp->capacity) > 4) &&
2484 		    (sdkp->capacity > 0xffffffffULL)) {
2485 			int old_sector_size = sector_size;
2486 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2487 					"Trying to use READ CAPACITY(16).\n");
2488 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2489 			if (sector_size < 0) {
2490 				sd_printk(KERN_NOTICE, sdkp,
2491 					"Using 0xffffffff as device size\n");
2492 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2493 				sector_size = old_sector_size;
2494 				goto got_data;
2495 			}
2496 			/* Remember that READ CAPACITY(16) succeeded */
2497 			sdp->try_rc_10_first = 0;
2498 		}
2499 	}
2500 
2501 	/* Some devices are known to return the total number of blocks,
2502 	 * not the highest block number.  Some devices have versions
2503 	 * which do this and others which do not.  Some devices we might
2504 	 * suspect of doing this but we don't know for certain.
2505 	 *
2506 	 * If we know the reported capacity is wrong, decrement it.  If
2507 	 * we can only guess, then assume the number of blocks is even
2508 	 * (usually true but not always) and err on the side of lowering
2509 	 * the capacity.
2510 	 */
2511 	if (sdp->fix_capacity ||
2512 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2513 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2514 				"from its reported value: %llu\n",
2515 				(unsigned long long) sdkp->capacity);
2516 		--sdkp->capacity;
2517 	}
2518 
2519 got_data:
2520 	if (sector_size == 0) {
2521 		sector_size = 512;
2522 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2523 			  "assuming 512.\n");
2524 	}
2525 
2526 	if (sector_size != 512 &&
2527 	    sector_size != 1024 &&
2528 	    sector_size != 2048 &&
2529 	    sector_size != 4096) {
2530 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2531 			  sector_size);
2532 		/*
2533 		 * The user might want to re-format the drive with
2534 		 * a supported sectorsize.  Once this happens, it
2535 		 * would be relatively trivial to set the thing up.
2536 		 * For this reason, we leave the thing in the table.
2537 		 */
2538 		sdkp->capacity = 0;
2539 		/*
2540 		 * set a bogus sector size so the normal read/write
2541 		 * logic in the block layer will eventually refuse any
2542 		 * request on this device without tripping over power
2543 		 * of two sector size assumptions
2544 		 */
2545 		sector_size = 512;
2546 	}
2547 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2548 	blk_queue_physical_block_size(sdp->request_queue,
2549 				      sdkp->physical_block_size);
2550 	sdkp->device->sector_size = sector_size;
2551 
2552 	if (sdkp->capacity > 0xffffffff)
2553 		sdp->use_16_for_rw = 1;
2554 
2555 }
2556 
2557 /*
2558  * Print disk capacity
2559  */
2560 static void
2561 sd_print_capacity(struct scsi_disk *sdkp,
2562 		  sector_t old_capacity)
2563 {
2564 	int sector_size = sdkp->device->sector_size;
2565 	char cap_str_2[10], cap_str_10[10];
2566 
2567 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2568 		return;
2569 
2570 	string_get_size(sdkp->capacity, sector_size,
2571 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2572 	string_get_size(sdkp->capacity, sector_size,
2573 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2574 
2575 	sd_printk(KERN_NOTICE, sdkp,
2576 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2577 		  (unsigned long long)sdkp->capacity,
2578 		  sector_size, cap_str_10, cap_str_2);
2579 
2580 	if (sdkp->physical_block_size != sector_size)
2581 		sd_printk(KERN_NOTICE, sdkp,
2582 			  "%u-byte physical blocks\n",
2583 			  sdkp->physical_block_size);
2584 }
2585 
2586 /* called with buffer of length 512 */
2587 static inline int
2588 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2589 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2590 		 struct scsi_sense_hdr *sshdr)
2591 {
2592 	/*
2593 	 * If we must use MODE SENSE(10), make sure that the buffer length
2594 	 * is at least 8 bytes so that the mode sense header fits.
2595 	 */
2596 	if (sdkp->device->use_10_for_ms && len < 8)
2597 		len = 8;
2598 
2599 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2600 			       SD_TIMEOUT, sdkp->max_retries, data,
2601 			       sshdr);
2602 }
2603 
2604 /*
2605  * read write protect setting, if possible - called only in sd_revalidate_disk()
2606  * called with buffer of length SD_BUF_SIZE
2607  */
2608 static void
2609 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2610 {
2611 	int res;
2612 	struct scsi_device *sdp = sdkp->device;
2613 	struct scsi_mode_data data;
2614 	int old_wp = sdkp->write_prot;
2615 
2616 	set_disk_ro(sdkp->disk, 0);
2617 	if (sdp->skip_ms_page_3f) {
2618 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2619 		return;
2620 	}
2621 
2622 	if (sdp->use_192_bytes_for_3f) {
2623 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2624 	} else {
2625 		/*
2626 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2627 		 * We have to start carefully: some devices hang if we ask
2628 		 * for more than is available.
2629 		 */
2630 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2631 
2632 		/*
2633 		 * Second attempt: ask for page 0 When only page 0 is
2634 		 * implemented, a request for page 3F may return Sense Key
2635 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2636 		 * CDB.
2637 		 */
2638 		if (res < 0)
2639 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2640 
2641 		/*
2642 		 * Third attempt: ask 255 bytes, as we did earlier.
2643 		 */
2644 		if (res < 0)
2645 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2646 					       &data, NULL);
2647 	}
2648 
2649 	if (res < 0) {
2650 		sd_first_printk(KERN_WARNING, sdkp,
2651 			  "Test WP failed, assume Write Enabled\n");
2652 	} else {
2653 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2654 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2655 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2656 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2657 				  sdkp->write_prot ? "on" : "off");
2658 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2659 		}
2660 	}
2661 }
2662 
2663 /*
2664  * sd_read_cache_type - called only from sd_revalidate_disk()
2665  * called with buffer of length SD_BUF_SIZE
2666  */
2667 static void
2668 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2669 {
2670 	int len = 0, res;
2671 	struct scsi_device *sdp = sdkp->device;
2672 
2673 	int dbd;
2674 	int modepage;
2675 	int first_len;
2676 	struct scsi_mode_data data;
2677 	struct scsi_sense_hdr sshdr;
2678 	int old_wce = sdkp->WCE;
2679 	int old_rcd = sdkp->RCD;
2680 	int old_dpofua = sdkp->DPOFUA;
2681 
2682 
2683 	if (sdkp->cache_override)
2684 		return;
2685 
2686 	first_len = 4;
2687 	if (sdp->skip_ms_page_8) {
2688 		if (sdp->type == TYPE_RBC)
2689 			goto defaults;
2690 		else {
2691 			if (sdp->skip_ms_page_3f)
2692 				goto defaults;
2693 			modepage = 0x3F;
2694 			if (sdp->use_192_bytes_for_3f)
2695 				first_len = 192;
2696 			dbd = 0;
2697 		}
2698 	} else if (sdp->type == TYPE_RBC) {
2699 		modepage = 6;
2700 		dbd = 8;
2701 	} else {
2702 		modepage = 8;
2703 		dbd = 0;
2704 	}
2705 
2706 	/* cautiously ask */
2707 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2708 			&data, &sshdr);
2709 
2710 	if (res < 0)
2711 		goto bad_sense;
2712 
2713 	if (!data.header_length) {
2714 		modepage = 6;
2715 		first_len = 0;
2716 		sd_first_printk(KERN_ERR, sdkp,
2717 				"Missing header in MODE_SENSE response\n");
2718 	}
2719 
2720 	/* that went OK, now ask for the proper length */
2721 	len = data.length;
2722 
2723 	/*
2724 	 * We're only interested in the first three bytes, actually.
2725 	 * But the data cache page is defined for the first 20.
2726 	 */
2727 	if (len < 3)
2728 		goto bad_sense;
2729 	else if (len > SD_BUF_SIZE) {
2730 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2731 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2732 		len = SD_BUF_SIZE;
2733 	}
2734 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2735 		len = 192;
2736 
2737 	/* Get the data */
2738 	if (len > first_len)
2739 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2740 				&data, &sshdr);
2741 
2742 	if (!res) {
2743 		int offset = data.header_length + data.block_descriptor_length;
2744 
2745 		while (offset < len) {
2746 			u8 page_code = buffer[offset] & 0x3F;
2747 			u8 spf       = buffer[offset] & 0x40;
2748 
2749 			if (page_code == 8 || page_code == 6) {
2750 				/* We're interested only in the first 3 bytes.
2751 				 */
2752 				if (len - offset <= 2) {
2753 					sd_first_printk(KERN_ERR, sdkp,
2754 						"Incomplete mode parameter "
2755 							"data\n");
2756 					goto defaults;
2757 				} else {
2758 					modepage = page_code;
2759 					goto Page_found;
2760 				}
2761 			} else {
2762 				/* Go to the next page */
2763 				if (spf && len - offset > 3)
2764 					offset += 4 + (buffer[offset+2] << 8) +
2765 						buffer[offset+3];
2766 				else if (!spf && len - offset > 1)
2767 					offset += 2 + buffer[offset+1];
2768 				else {
2769 					sd_first_printk(KERN_ERR, sdkp,
2770 							"Incomplete mode "
2771 							"parameter data\n");
2772 					goto defaults;
2773 				}
2774 			}
2775 		}
2776 
2777 		sd_first_printk(KERN_WARNING, sdkp,
2778 				"No Caching mode page found\n");
2779 		goto defaults;
2780 
2781 	Page_found:
2782 		if (modepage == 8) {
2783 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2784 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2785 		} else {
2786 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2787 			sdkp->RCD = 0;
2788 		}
2789 
2790 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2791 		if (sdp->broken_fua) {
2792 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2793 			sdkp->DPOFUA = 0;
2794 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2795 			   !sdkp->device->use_16_for_rw) {
2796 			sd_first_printk(KERN_NOTICE, sdkp,
2797 				  "Uses READ/WRITE(6), disabling FUA\n");
2798 			sdkp->DPOFUA = 0;
2799 		}
2800 
2801 		/* No cache flush allowed for write protected devices */
2802 		if (sdkp->WCE && sdkp->write_prot)
2803 			sdkp->WCE = 0;
2804 
2805 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2806 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2807 			sd_printk(KERN_NOTICE, sdkp,
2808 				  "Write cache: %s, read cache: %s, %s\n",
2809 				  sdkp->WCE ? "enabled" : "disabled",
2810 				  sdkp->RCD ? "disabled" : "enabled",
2811 				  sdkp->DPOFUA ? "supports DPO and FUA"
2812 				  : "doesn't support DPO or FUA");
2813 
2814 		return;
2815 	}
2816 
2817 bad_sense:
2818 	if (scsi_sense_valid(&sshdr) &&
2819 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2820 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2821 		/* Invalid field in CDB */
2822 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2823 	else
2824 		sd_first_printk(KERN_ERR, sdkp,
2825 				"Asking for cache data failed\n");
2826 
2827 defaults:
2828 	if (sdp->wce_default_on) {
2829 		sd_first_printk(KERN_NOTICE, sdkp,
2830 				"Assuming drive cache: write back\n");
2831 		sdkp->WCE = 1;
2832 	} else {
2833 		sd_first_printk(KERN_WARNING, sdkp,
2834 				"Assuming drive cache: write through\n");
2835 		sdkp->WCE = 0;
2836 	}
2837 	sdkp->RCD = 0;
2838 	sdkp->DPOFUA = 0;
2839 }
2840 
2841 /*
2842  * The ATO bit indicates whether the DIF application tag is available
2843  * for use by the operating system.
2844  */
2845 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2846 {
2847 	int res, offset;
2848 	struct scsi_device *sdp = sdkp->device;
2849 	struct scsi_mode_data data;
2850 	struct scsi_sense_hdr sshdr;
2851 
2852 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2853 		return;
2854 
2855 	if (sdkp->protection_type == 0)
2856 		return;
2857 
2858 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2859 			      sdkp->max_retries, &data, &sshdr);
2860 
2861 	if (res < 0 || !data.header_length ||
2862 	    data.length < 6) {
2863 		sd_first_printk(KERN_WARNING, sdkp,
2864 			  "getting Control mode page failed, assume no ATO\n");
2865 
2866 		if (scsi_sense_valid(&sshdr))
2867 			sd_print_sense_hdr(sdkp, &sshdr);
2868 
2869 		return;
2870 	}
2871 
2872 	offset = data.header_length + data.block_descriptor_length;
2873 
2874 	if ((buffer[offset] & 0x3f) != 0x0a) {
2875 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2876 		return;
2877 	}
2878 
2879 	if ((buffer[offset + 5] & 0x80) == 0)
2880 		return;
2881 
2882 	sdkp->ATO = 1;
2883 
2884 	return;
2885 }
2886 
2887 /**
2888  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2889  * @sdkp: disk to query
2890  */
2891 static void sd_read_block_limits(struct scsi_disk *sdkp)
2892 {
2893 	struct scsi_vpd *vpd;
2894 
2895 	rcu_read_lock();
2896 
2897 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2898 	if (!vpd || vpd->len < 16)
2899 		goto out;
2900 
2901 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2902 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2903 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2904 
2905 	if (vpd->len >= 64) {
2906 		unsigned int lba_count, desc_count;
2907 
2908 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2909 
2910 		if (!sdkp->lbpme)
2911 			goto out;
2912 
2913 		lba_count = get_unaligned_be32(&vpd->data[20]);
2914 		desc_count = get_unaligned_be32(&vpd->data[24]);
2915 
2916 		if (lba_count && desc_count)
2917 			sdkp->max_unmap_blocks = lba_count;
2918 
2919 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2920 
2921 		if (vpd->data[32] & 0x80)
2922 			sdkp->unmap_alignment =
2923 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2924 
2925 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2926 
2927 			if (sdkp->max_unmap_blocks)
2928 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2929 			else
2930 				sd_config_discard(sdkp, SD_LBP_WS16);
2931 
2932 		} else {	/* LBP VPD page tells us what to use */
2933 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2934 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2935 			else if (sdkp->lbpws)
2936 				sd_config_discard(sdkp, SD_LBP_WS16);
2937 			else if (sdkp->lbpws10)
2938 				sd_config_discard(sdkp, SD_LBP_WS10);
2939 			else
2940 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2941 		}
2942 	}
2943 
2944  out:
2945 	rcu_read_unlock();
2946 }
2947 
2948 /**
2949  * sd_read_block_characteristics - Query block dev. characteristics
2950  * @sdkp: disk to query
2951  */
2952 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2953 {
2954 	struct request_queue *q = sdkp->disk->queue;
2955 	struct scsi_vpd *vpd;
2956 	u16 rot;
2957 	u8 zoned;
2958 
2959 	rcu_read_lock();
2960 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
2961 
2962 	if (!vpd || vpd->len < 8) {
2963 		rcu_read_unlock();
2964 	        return;
2965 	}
2966 
2967 	rot = get_unaligned_be16(&vpd->data[4]);
2968 	zoned = (vpd->data[8] >> 4) & 3;
2969 	rcu_read_unlock();
2970 
2971 	if (rot == 1) {
2972 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2973 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2974 	}
2975 
2976 	if (sdkp->device->type == TYPE_ZBC) {
2977 		/* Host-managed */
2978 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
2979 	} else {
2980 		sdkp->zoned = zoned;
2981 		if (sdkp->zoned == 1) {
2982 			/* Host-aware */
2983 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
2984 		} else {
2985 			/* Regular disk or drive managed disk */
2986 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
2987 		}
2988 	}
2989 
2990 	if (!sdkp->first_scan)
2991 		return;
2992 
2993 	if (blk_queue_is_zoned(q)) {
2994 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
2995 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
2996 	} else {
2997 		if (sdkp->zoned == 1)
2998 			sd_printk(KERN_NOTICE, sdkp,
2999 				  "Host-aware SMR disk used as regular disk\n");
3000 		else if (sdkp->zoned == 2)
3001 			sd_printk(KERN_NOTICE, sdkp,
3002 				  "Drive-managed SMR disk\n");
3003 	}
3004 }
3005 
3006 /**
3007  * sd_read_block_provisioning - Query provisioning VPD page
3008  * @sdkp: disk to query
3009  */
3010 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3011 {
3012 	struct scsi_vpd *vpd;
3013 
3014 	if (sdkp->lbpme == 0)
3015 		return;
3016 
3017 	rcu_read_lock();
3018 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3019 
3020 	if (!vpd || vpd->len < 8) {
3021 		rcu_read_unlock();
3022 		return;
3023 	}
3024 
3025 	sdkp->lbpvpd	= 1;
3026 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3027 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3028 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3029 	rcu_read_unlock();
3030 }
3031 
3032 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3033 {
3034 	struct scsi_device *sdev = sdkp->device;
3035 
3036 	if (sdev->host->no_write_same) {
3037 		sdev->no_write_same = 1;
3038 
3039 		return;
3040 	}
3041 
3042 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
3043 		struct scsi_vpd *vpd;
3044 
3045 		sdev->no_report_opcodes = 1;
3046 
3047 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3048 		 * CODES is unsupported and the device has an ATA
3049 		 * Information VPD page (SAT).
3050 		 */
3051 		rcu_read_lock();
3052 		vpd = rcu_dereference(sdev->vpd_pg89);
3053 		if (vpd)
3054 			sdev->no_write_same = 1;
3055 		rcu_read_unlock();
3056 	}
3057 
3058 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3059 		sdkp->ws16 = 1;
3060 
3061 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3062 		sdkp->ws10 = 1;
3063 }
3064 
3065 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3066 {
3067 	struct scsi_device *sdev = sdkp->device;
3068 
3069 	if (!sdev->security_supported)
3070 		return;
3071 
3072 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3073 			SECURITY_PROTOCOL_IN) == 1 &&
3074 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3075 			SECURITY_PROTOCOL_OUT) == 1)
3076 		sdkp->security = 1;
3077 }
3078 
3079 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3080 {
3081 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3082 }
3083 
3084 /**
3085  * sd_read_cpr - Query concurrent positioning ranges
3086  * @sdkp:	disk to query
3087  */
3088 static void sd_read_cpr(struct scsi_disk *sdkp)
3089 {
3090 	struct blk_independent_access_ranges *iars = NULL;
3091 	unsigned char *buffer = NULL;
3092 	unsigned int nr_cpr = 0;
3093 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3094 	u8 *desc;
3095 
3096 	/*
3097 	 * We need to have the capacity set first for the block layer to be
3098 	 * able to check the ranges.
3099 	 */
3100 	if (sdkp->first_scan)
3101 		return;
3102 
3103 	if (!sdkp->capacity)
3104 		goto out;
3105 
3106 	/*
3107 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3108 	 * leading to a maximum page size of 64 + 256*32 bytes.
3109 	 */
3110 	buf_len = 64 + 256*32;
3111 	buffer = kmalloc(buf_len, GFP_KERNEL);
3112 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3113 		goto out;
3114 
3115 	/* We must have at least a 64B header and one 32B range descriptor */
3116 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3117 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3118 		sd_printk(KERN_ERR, sdkp,
3119 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3120 		goto out;
3121 	}
3122 
3123 	nr_cpr = (vpd_len - 64) / 32;
3124 	if (nr_cpr == 1) {
3125 		nr_cpr = 0;
3126 		goto out;
3127 	}
3128 
3129 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3130 	if (!iars) {
3131 		nr_cpr = 0;
3132 		goto out;
3133 	}
3134 
3135 	desc = &buffer[64];
3136 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3137 		if (desc[0] != i) {
3138 			sd_printk(KERN_ERR, sdkp,
3139 				"Invalid Concurrent Positioning Range number\n");
3140 			nr_cpr = 0;
3141 			break;
3142 		}
3143 
3144 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3145 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3146 	}
3147 
3148 out:
3149 	disk_set_independent_access_ranges(sdkp->disk, iars);
3150 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3151 		sd_printk(KERN_NOTICE, sdkp,
3152 			  "%u concurrent positioning ranges\n", nr_cpr);
3153 		sdkp->nr_actuators = nr_cpr;
3154 	}
3155 
3156 	kfree(buffer);
3157 }
3158 
3159 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3160 {
3161 	struct scsi_device *sdp = sdkp->device;
3162 	unsigned int min_xfer_bytes =
3163 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3164 
3165 	if (sdkp->min_xfer_blocks == 0)
3166 		return false;
3167 
3168 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3169 		sd_first_printk(KERN_WARNING, sdkp,
3170 				"Preferred minimum I/O size %u bytes not a " \
3171 				"multiple of physical block size (%u bytes)\n",
3172 				min_xfer_bytes, sdkp->physical_block_size);
3173 		sdkp->min_xfer_blocks = 0;
3174 		return false;
3175 	}
3176 
3177 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3178 			min_xfer_bytes);
3179 	return true;
3180 }
3181 
3182 /*
3183  * Determine the device's preferred I/O size for reads and writes
3184  * unless the reported value is unreasonably small, large, not a
3185  * multiple of the physical block size, or simply garbage.
3186  */
3187 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3188 				      unsigned int dev_max)
3189 {
3190 	struct scsi_device *sdp = sdkp->device;
3191 	unsigned int opt_xfer_bytes =
3192 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3193 	unsigned int min_xfer_bytes =
3194 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3195 
3196 	if (sdkp->opt_xfer_blocks == 0)
3197 		return false;
3198 
3199 	if (sdkp->opt_xfer_blocks > dev_max) {
3200 		sd_first_printk(KERN_WARNING, sdkp,
3201 				"Optimal transfer size %u logical blocks " \
3202 				"> dev_max (%u logical blocks)\n",
3203 				sdkp->opt_xfer_blocks, dev_max);
3204 		return false;
3205 	}
3206 
3207 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3208 		sd_first_printk(KERN_WARNING, sdkp,
3209 				"Optimal transfer size %u logical blocks " \
3210 				"> sd driver limit (%u logical blocks)\n",
3211 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3212 		return false;
3213 	}
3214 
3215 	if (opt_xfer_bytes < PAGE_SIZE) {
3216 		sd_first_printk(KERN_WARNING, sdkp,
3217 				"Optimal transfer size %u bytes < " \
3218 				"PAGE_SIZE (%u bytes)\n",
3219 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3220 		return false;
3221 	}
3222 
3223 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3224 		sd_first_printk(KERN_WARNING, sdkp,
3225 				"Optimal transfer size %u bytes not a " \
3226 				"multiple of preferred minimum block " \
3227 				"size (%u bytes)\n",
3228 				opt_xfer_bytes, min_xfer_bytes);
3229 		return false;
3230 	}
3231 
3232 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3233 		sd_first_printk(KERN_WARNING, sdkp,
3234 				"Optimal transfer size %u bytes not a " \
3235 				"multiple of physical block size (%u bytes)\n",
3236 				opt_xfer_bytes, sdkp->physical_block_size);
3237 		return false;
3238 	}
3239 
3240 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3241 			opt_xfer_bytes);
3242 	return true;
3243 }
3244 
3245 /**
3246  *	sd_revalidate_disk - called the first time a new disk is seen,
3247  *	performs disk spin up, read_capacity, etc.
3248  *	@disk: struct gendisk we care about
3249  **/
3250 static int sd_revalidate_disk(struct gendisk *disk)
3251 {
3252 	struct scsi_disk *sdkp = scsi_disk(disk);
3253 	struct scsi_device *sdp = sdkp->device;
3254 	struct request_queue *q = sdkp->disk->queue;
3255 	sector_t old_capacity = sdkp->capacity;
3256 	unsigned char *buffer;
3257 	unsigned int dev_max, rw_max;
3258 
3259 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3260 				      "sd_revalidate_disk\n"));
3261 
3262 	/*
3263 	 * If the device is offline, don't try and read capacity or any
3264 	 * of the other niceties.
3265 	 */
3266 	if (!scsi_device_online(sdp))
3267 		goto out;
3268 
3269 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3270 	if (!buffer) {
3271 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3272 			  "allocation failure.\n");
3273 		goto out;
3274 	}
3275 
3276 	sd_spinup_disk(sdkp);
3277 
3278 	/*
3279 	 * Without media there is no reason to ask; moreover, some devices
3280 	 * react badly if we do.
3281 	 */
3282 	if (sdkp->media_present) {
3283 		sd_read_capacity(sdkp, buffer);
3284 
3285 		/*
3286 		 * set the default to rotational.  All non-rotational devices
3287 		 * support the block characteristics VPD page, which will
3288 		 * cause this to be updated correctly and any device which
3289 		 * doesn't support it should be treated as rotational.
3290 		 */
3291 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3292 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3293 
3294 		if (scsi_device_supports_vpd(sdp)) {
3295 			sd_read_block_provisioning(sdkp);
3296 			sd_read_block_limits(sdkp);
3297 			sd_read_block_characteristics(sdkp);
3298 			sd_zbc_read_zones(sdkp, buffer);
3299 			sd_read_cpr(sdkp);
3300 		}
3301 
3302 		sd_print_capacity(sdkp, old_capacity);
3303 
3304 		sd_read_write_protect_flag(sdkp, buffer);
3305 		sd_read_cache_type(sdkp, buffer);
3306 		sd_read_app_tag_own(sdkp, buffer);
3307 		sd_read_write_same(sdkp, buffer);
3308 		sd_read_security(sdkp, buffer);
3309 		sd_config_protection(sdkp);
3310 	}
3311 
3312 	/*
3313 	 * We now have all cache related info, determine how we deal
3314 	 * with flush requests.
3315 	 */
3316 	sd_set_flush_flag(sdkp);
3317 
3318 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3319 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3320 
3321 	/* Some devices report a maximum block count for READ/WRITE requests. */
3322 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3323 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3324 
3325 	if (sd_validate_min_xfer_size(sdkp))
3326 		blk_queue_io_min(sdkp->disk->queue,
3327 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3328 	else
3329 		blk_queue_io_min(sdkp->disk->queue, 0);
3330 
3331 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3332 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3333 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3334 	} else {
3335 		q->limits.io_opt = 0;
3336 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3337 				      (sector_t)BLK_DEF_MAX_SECTORS);
3338 	}
3339 
3340 	/*
3341 	 * Limit default to SCSI host optimal sector limit if set. There may be
3342 	 * an impact on performance for when the size of a request exceeds this
3343 	 * host limit.
3344 	 */
3345 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3346 
3347 	/* Do not exceed controller limit */
3348 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3349 
3350 	/*
3351 	 * Only update max_sectors if previously unset or if the current value
3352 	 * exceeds the capabilities of the hardware.
3353 	 */
3354 	if (sdkp->first_scan ||
3355 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3356 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3357 		q->limits.max_sectors = rw_max;
3358 
3359 	sdkp->first_scan = 0;
3360 
3361 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3362 	sd_config_write_same(sdkp);
3363 	kfree(buffer);
3364 
3365 	/*
3366 	 * For a zoned drive, revalidating the zones can be done only once
3367 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3368 	 * capacity to 0.
3369 	 */
3370 	if (sd_zbc_revalidate_zones(sdkp))
3371 		set_capacity_and_notify(disk, 0);
3372 
3373  out:
3374 	return 0;
3375 }
3376 
3377 /**
3378  *	sd_unlock_native_capacity - unlock native capacity
3379  *	@disk: struct gendisk to set capacity for
3380  *
3381  *	Block layer calls this function if it detects that partitions
3382  *	on @disk reach beyond the end of the device.  If the SCSI host
3383  *	implements ->unlock_native_capacity() method, it's invoked to
3384  *	give it a chance to adjust the device capacity.
3385  *
3386  *	CONTEXT:
3387  *	Defined by block layer.  Might sleep.
3388  */
3389 static void sd_unlock_native_capacity(struct gendisk *disk)
3390 {
3391 	struct scsi_device *sdev = scsi_disk(disk)->device;
3392 
3393 	if (sdev->host->hostt->unlock_native_capacity)
3394 		sdev->host->hostt->unlock_native_capacity(sdev);
3395 }
3396 
3397 /**
3398  *	sd_format_disk_name - format disk name
3399  *	@prefix: name prefix - ie. "sd" for SCSI disks
3400  *	@index: index of the disk to format name for
3401  *	@buf: output buffer
3402  *	@buflen: length of the output buffer
3403  *
3404  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3405  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3406  *	which is followed by sdaaa.
3407  *
3408  *	This is basically 26 base counting with one extra 'nil' entry
3409  *	at the beginning from the second digit on and can be
3410  *	determined using similar method as 26 base conversion with the
3411  *	index shifted -1 after each digit is computed.
3412  *
3413  *	CONTEXT:
3414  *	Don't care.
3415  *
3416  *	RETURNS:
3417  *	0 on success, -errno on failure.
3418  */
3419 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3420 {
3421 	const int base = 'z' - 'a' + 1;
3422 	char *begin = buf + strlen(prefix);
3423 	char *end = buf + buflen;
3424 	char *p;
3425 	int unit;
3426 
3427 	p = end - 1;
3428 	*p = '\0';
3429 	unit = base;
3430 	do {
3431 		if (p == begin)
3432 			return -EINVAL;
3433 		*--p = 'a' + (index % unit);
3434 		index = (index / unit) - 1;
3435 	} while (index >= 0);
3436 
3437 	memmove(begin, p, end - p);
3438 	memcpy(buf, prefix, strlen(prefix));
3439 
3440 	return 0;
3441 }
3442 
3443 /**
3444  *	sd_probe - called during driver initialization and whenever a
3445  *	new scsi device is attached to the system. It is called once
3446  *	for each scsi device (not just disks) present.
3447  *	@dev: pointer to device object
3448  *
3449  *	Returns 0 if successful (or not interested in this scsi device
3450  *	(e.g. scanner)); 1 when there is an error.
3451  *
3452  *	Note: this function is invoked from the scsi mid-level.
3453  *	This function sets up the mapping between a given
3454  *	<host,channel,id,lun> (found in sdp) and new device name
3455  *	(e.g. /dev/sda). More precisely it is the block device major
3456  *	and minor number that is chosen here.
3457  *
3458  *	Assume sd_probe is not re-entrant (for time being)
3459  *	Also think about sd_probe() and sd_remove() running coincidentally.
3460  **/
3461 static int sd_probe(struct device *dev)
3462 {
3463 	struct scsi_device *sdp = to_scsi_device(dev);
3464 	struct scsi_disk *sdkp;
3465 	struct gendisk *gd;
3466 	int index;
3467 	int error;
3468 
3469 	scsi_autopm_get_device(sdp);
3470 	error = -ENODEV;
3471 	if (sdp->type != TYPE_DISK &&
3472 	    sdp->type != TYPE_ZBC &&
3473 	    sdp->type != TYPE_MOD &&
3474 	    sdp->type != TYPE_RBC)
3475 		goto out;
3476 
3477 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3478 		sdev_printk(KERN_WARNING, sdp,
3479 			    "Unsupported ZBC host-managed device.\n");
3480 		goto out;
3481 	}
3482 
3483 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3484 					"sd_probe\n"));
3485 
3486 	error = -ENOMEM;
3487 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3488 	if (!sdkp)
3489 		goto out;
3490 
3491 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3492 					 &sd_bio_compl_lkclass);
3493 	if (!gd)
3494 		goto out_free;
3495 
3496 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3497 	if (index < 0) {
3498 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3499 		goto out_put;
3500 	}
3501 
3502 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3503 	if (error) {
3504 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3505 		goto out_free_index;
3506 	}
3507 
3508 	sdkp->device = sdp;
3509 	sdkp->disk = gd;
3510 	sdkp->index = index;
3511 	sdkp->max_retries = SD_MAX_RETRIES;
3512 	atomic_set(&sdkp->openers, 0);
3513 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3514 
3515 	if (!sdp->request_queue->rq_timeout) {
3516 		if (sdp->type != TYPE_MOD)
3517 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3518 		else
3519 			blk_queue_rq_timeout(sdp->request_queue,
3520 					     SD_MOD_TIMEOUT);
3521 	}
3522 
3523 	device_initialize(&sdkp->disk_dev);
3524 	sdkp->disk_dev.parent = get_device(dev);
3525 	sdkp->disk_dev.class = &sd_disk_class;
3526 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3527 
3528 	error = device_add(&sdkp->disk_dev);
3529 	if (error) {
3530 		put_device(&sdkp->disk_dev);
3531 		goto out;
3532 	}
3533 
3534 	dev_set_drvdata(dev, sdkp);
3535 
3536 	gd->major = sd_major((index & 0xf0) >> 4);
3537 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3538 	gd->minors = SD_MINORS;
3539 
3540 	gd->fops = &sd_fops;
3541 	gd->private_data = sdkp;
3542 
3543 	/* defaults, until the device tells us otherwise */
3544 	sdp->sector_size = 512;
3545 	sdkp->capacity = 0;
3546 	sdkp->media_present = 1;
3547 	sdkp->write_prot = 0;
3548 	sdkp->cache_override = 0;
3549 	sdkp->WCE = 0;
3550 	sdkp->RCD = 0;
3551 	sdkp->ATO = 0;
3552 	sdkp->first_scan = 1;
3553 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3554 
3555 	sd_revalidate_disk(gd);
3556 
3557 	if (sdp->removable) {
3558 		gd->flags |= GENHD_FL_REMOVABLE;
3559 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3560 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3561 	}
3562 
3563 	blk_pm_runtime_init(sdp->request_queue, dev);
3564 	if (sdp->rpm_autosuspend) {
3565 		pm_runtime_set_autosuspend_delay(dev,
3566 			sdp->host->hostt->rpm_autosuspend_delay);
3567 	}
3568 
3569 	error = device_add_disk(dev, gd, NULL);
3570 	if (error) {
3571 		put_device(&sdkp->disk_dev);
3572 		put_disk(gd);
3573 		goto out;
3574 	}
3575 
3576 	if (sdkp->security) {
3577 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3578 		if (sdkp->opal_dev)
3579 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3580 	}
3581 
3582 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3583 		  sdp->removable ? "removable " : "");
3584 	scsi_autopm_put_device(sdp);
3585 
3586 	return 0;
3587 
3588  out_free_index:
3589 	ida_free(&sd_index_ida, index);
3590  out_put:
3591 	put_disk(gd);
3592  out_free:
3593 	kfree(sdkp);
3594  out:
3595 	scsi_autopm_put_device(sdp);
3596 	return error;
3597 }
3598 
3599 /**
3600  *	sd_remove - called whenever a scsi disk (previously recognized by
3601  *	sd_probe) is detached from the system. It is called (potentially
3602  *	multiple times) during sd module unload.
3603  *	@dev: pointer to device object
3604  *
3605  *	Note: this function is invoked from the scsi mid-level.
3606  *	This function potentially frees up a device name (e.g. /dev/sdc)
3607  *	that could be re-used by a subsequent sd_probe().
3608  *	This function is not called when the built-in sd driver is "exit-ed".
3609  **/
3610 static int sd_remove(struct device *dev)
3611 {
3612 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3613 
3614 	scsi_autopm_get_device(sdkp->device);
3615 
3616 	device_del(&sdkp->disk_dev);
3617 	del_gendisk(sdkp->disk);
3618 	sd_shutdown(dev);
3619 
3620 	put_disk(sdkp->disk);
3621 	return 0;
3622 }
3623 
3624 static void scsi_disk_release(struct device *dev)
3625 {
3626 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3627 
3628 	ida_free(&sd_index_ida, sdkp->index);
3629 	sd_zbc_free_zone_info(sdkp);
3630 	put_device(&sdkp->device->sdev_gendev);
3631 	free_opal_dev(sdkp->opal_dev);
3632 
3633 	kfree(sdkp);
3634 }
3635 
3636 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3637 {
3638 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3639 	struct scsi_sense_hdr sshdr;
3640 	struct scsi_device *sdp = sdkp->device;
3641 	int res;
3642 
3643 	if (start)
3644 		cmd[4] |= 1;	/* START */
3645 
3646 	if (sdp->start_stop_pwr_cond)
3647 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3648 
3649 	if (!scsi_device_online(sdp))
3650 		return -ENODEV;
3651 
3652 	res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, &sshdr,
3653 			SD_TIMEOUT, sdkp->max_retries, 0, RQF_PM, NULL);
3654 	if (res) {
3655 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3656 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3657 			sd_print_sense_hdr(sdkp, &sshdr);
3658 			/* 0x3a is medium not present */
3659 			if (sshdr.asc == 0x3a)
3660 				res = 0;
3661 		}
3662 	}
3663 
3664 	/* SCSI error codes must not go to the generic layer */
3665 	if (res)
3666 		return -EIO;
3667 
3668 	return 0;
3669 }
3670 
3671 /*
3672  * Send a SYNCHRONIZE CACHE instruction down to the device through
3673  * the normal SCSI command structure.  Wait for the command to
3674  * complete.
3675  */
3676 static void sd_shutdown(struct device *dev)
3677 {
3678 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3679 
3680 	if (!sdkp)
3681 		return;         /* this can happen */
3682 
3683 	if (pm_runtime_suspended(dev))
3684 		return;
3685 
3686 	if (sdkp->WCE && sdkp->media_present) {
3687 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3688 		sd_sync_cache(sdkp, NULL);
3689 	}
3690 
3691 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3692 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3693 		sd_start_stop_device(sdkp, 0);
3694 	}
3695 }
3696 
3697 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3698 {
3699 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3700 	struct scsi_sense_hdr sshdr;
3701 	int ret = 0;
3702 
3703 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3704 		return 0;
3705 
3706 	if (sdkp->WCE && sdkp->media_present) {
3707 		if (!sdkp->device->silence_suspend)
3708 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3709 		ret = sd_sync_cache(sdkp, &sshdr);
3710 
3711 		if (ret) {
3712 			/* ignore OFFLINE device */
3713 			if (ret == -ENODEV)
3714 				return 0;
3715 
3716 			if (!scsi_sense_valid(&sshdr) ||
3717 			    sshdr.sense_key != ILLEGAL_REQUEST)
3718 				return ret;
3719 
3720 			/*
3721 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3722 			 * doesn't support sync. There's not much to do and
3723 			 * suspend shouldn't fail.
3724 			 */
3725 			ret = 0;
3726 		}
3727 	}
3728 
3729 	if (sdkp->device->manage_start_stop) {
3730 		if (!sdkp->device->silence_suspend)
3731 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3732 		/* an error is not worth aborting a system sleep */
3733 		ret = sd_start_stop_device(sdkp, 0);
3734 		if (ignore_stop_errors)
3735 			ret = 0;
3736 	}
3737 
3738 	return ret;
3739 }
3740 
3741 static int sd_suspend_system(struct device *dev)
3742 {
3743 	if (pm_runtime_suspended(dev))
3744 		return 0;
3745 
3746 	return sd_suspend_common(dev, true);
3747 }
3748 
3749 static int sd_suspend_runtime(struct device *dev)
3750 {
3751 	return sd_suspend_common(dev, false);
3752 }
3753 
3754 static int sd_resume(struct device *dev)
3755 {
3756 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3757 	int ret;
3758 
3759 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3760 		return 0;
3761 
3762 	if (!sdkp->device->manage_start_stop)
3763 		return 0;
3764 
3765 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3766 	ret = sd_start_stop_device(sdkp, 1);
3767 	if (!ret)
3768 		opal_unlock_from_suspend(sdkp->opal_dev);
3769 	return ret;
3770 }
3771 
3772 static int sd_resume_system(struct device *dev)
3773 {
3774 	if (pm_runtime_suspended(dev))
3775 		return 0;
3776 
3777 	return sd_resume(dev);
3778 }
3779 
3780 static int sd_resume_runtime(struct device *dev)
3781 {
3782 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3783 	struct scsi_device *sdp;
3784 
3785 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3786 		return 0;
3787 
3788 	sdp = sdkp->device;
3789 
3790 	if (sdp->ignore_media_change) {
3791 		/* clear the device's sense data */
3792 		static const u8 cmd[10] = { REQUEST_SENSE };
3793 
3794 		if (scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL,
3795 				 NULL, sdp->request_queue->rq_timeout, 1, 0,
3796 				 RQF_PM, NULL))
3797 			sd_printk(KERN_NOTICE, sdkp,
3798 				  "Failed to clear sense data\n");
3799 	}
3800 
3801 	return sd_resume(dev);
3802 }
3803 
3804 /**
3805  *	init_sd - entry point for this driver (both when built in or when
3806  *	a module).
3807  *
3808  *	Note: this function registers this driver with the scsi mid-level.
3809  **/
3810 static int __init init_sd(void)
3811 {
3812 	int majors = 0, i, err;
3813 
3814 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3815 
3816 	for (i = 0; i < SD_MAJORS; i++) {
3817 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3818 			continue;
3819 		majors++;
3820 	}
3821 
3822 	if (!majors)
3823 		return -ENODEV;
3824 
3825 	err = class_register(&sd_disk_class);
3826 	if (err)
3827 		goto err_out;
3828 
3829 	sd_cdb_cache = kmem_cache_create("sd_ext_cdb", SD_EXT_CDB_SIZE,
3830 					 0, 0, NULL);
3831 	if (!sd_cdb_cache) {
3832 		printk(KERN_ERR "sd: can't init extended cdb cache\n");
3833 		err = -ENOMEM;
3834 		goto err_out_class;
3835 	}
3836 
3837 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3838 	if (!sd_page_pool) {
3839 		printk(KERN_ERR "sd: can't init discard page pool\n");
3840 		err = -ENOMEM;
3841 		goto err_out_cache;
3842 	}
3843 
3844 	err = scsi_register_driver(&sd_template.gendrv);
3845 	if (err)
3846 		goto err_out_driver;
3847 
3848 	return 0;
3849 
3850 err_out_driver:
3851 	mempool_destroy(sd_page_pool);
3852 
3853 err_out_cache:
3854 	kmem_cache_destroy(sd_cdb_cache);
3855 
3856 err_out_class:
3857 	class_unregister(&sd_disk_class);
3858 err_out:
3859 	for (i = 0; i < SD_MAJORS; i++)
3860 		unregister_blkdev(sd_major(i), "sd");
3861 	return err;
3862 }
3863 
3864 /**
3865  *	exit_sd - exit point for this driver (when it is a module).
3866  *
3867  *	Note: this function unregisters this driver from the scsi mid-level.
3868  **/
3869 static void __exit exit_sd(void)
3870 {
3871 	int i;
3872 
3873 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3874 
3875 	scsi_unregister_driver(&sd_template.gendrv);
3876 	mempool_destroy(sd_page_pool);
3877 	kmem_cache_destroy(sd_cdb_cache);
3878 
3879 	class_unregister(&sd_disk_class);
3880 
3881 	for (i = 0; i < SD_MAJORS; i++)
3882 		unregister_blkdev(sd_major(i), "sd");
3883 }
3884 
3885 module_init(init_sd);
3886 module_exit(exit_sd);
3887 
3888 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3889 {
3890 	scsi_print_sense_hdr(sdkp->device,
3891 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3892 }
3893 
3894 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3895 {
3896 	const char *hb_string = scsi_hostbyte_string(result);
3897 
3898 	if (hb_string)
3899 		sd_printk(KERN_INFO, sdkp,
3900 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3901 			  hb_string ? hb_string : "invalid",
3902 			  "DRIVER_OK");
3903 	else
3904 		sd_printk(KERN_INFO, sdkp,
3905 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3906 			  msg, host_byte(result), "DRIVER_OK");
3907 }
3908