xref: /openbmc/linux/drivers/scsi/sd.c (revision c29b9772)
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 mempool_t *sd_page_pool;
125 static struct lock_class_key sd_bio_compl_lkclass;
126 
127 static const char *sd_cache_types[] = {
128 	"write through", "none", "write back",
129 	"write back, no read (daft)"
130 };
131 
132 static void sd_set_flush_flag(struct scsi_disk *sdkp)
133 {
134 	bool wc = false, fua = false;
135 
136 	if (sdkp->WCE) {
137 		wc = true;
138 		if (sdkp->DPOFUA)
139 			fua = true;
140 	}
141 
142 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
143 }
144 
145 static ssize_t
146 cache_type_store(struct device *dev, struct device_attribute *attr,
147 		 const char *buf, size_t count)
148 {
149 	int ct, rcd, wce, sp;
150 	struct scsi_disk *sdkp = to_scsi_disk(dev);
151 	struct scsi_device *sdp = sdkp->device;
152 	char buffer[64];
153 	char *buffer_data;
154 	struct scsi_mode_data data;
155 	struct scsi_sense_hdr sshdr;
156 	static const char temp[] = "temporary ";
157 	int len;
158 
159 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
160 		/* no cache control on RBC devices; theoretically they
161 		 * can do it, but there's probably so many exceptions
162 		 * it's not worth the risk */
163 		return -EINVAL;
164 
165 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
166 		buf += sizeof(temp) - 1;
167 		sdkp->cache_override = 1;
168 	} else {
169 		sdkp->cache_override = 0;
170 	}
171 
172 	ct = sysfs_match_string(sd_cache_types, buf);
173 	if (ct < 0)
174 		return -EINVAL;
175 
176 	rcd = ct & 0x01 ? 1 : 0;
177 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
178 
179 	if (sdkp->cache_override) {
180 		sdkp->WCE = wce;
181 		sdkp->RCD = rcd;
182 		sd_set_flush_flag(sdkp);
183 		return count;
184 	}
185 
186 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
187 			    sdkp->max_retries, &data, NULL))
188 		return -EINVAL;
189 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
190 		  data.block_descriptor_length);
191 	buffer_data = buffer + data.header_length +
192 		data.block_descriptor_length;
193 	buffer_data[2] &= ~0x05;
194 	buffer_data[2] |= wce << 2 | rcd;
195 	sp = buffer_data[0] & 0x80 ? 1 : 0;
196 	buffer_data[0] &= ~0x80;
197 
198 	/*
199 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
200 	 * received mode parameter buffer before doing MODE SELECT.
201 	 */
202 	data.device_specific = 0;
203 
204 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
205 			     sdkp->max_retries, &data, &sshdr)) {
206 		if (scsi_sense_valid(&sshdr))
207 			sd_print_sense_hdr(sdkp, &sshdr);
208 		return -EINVAL;
209 	}
210 	sd_revalidate_disk(sdkp->disk);
211 	return count;
212 }
213 
214 static ssize_t
215 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
216 		       char *buf)
217 {
218 	struct scsi_disk *sdkp = to_scsi_disk(dev);
219 	struct scsi_device *sdp = sdkp->device;
220 
221 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
222 }
223 
224 static ssize_t
225 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
226 			const char *buf, size_t count)
227 {
228 	struct scsi_disk *sdkp = to_scsi_disk(dev);
229 	struct scsi_device *sdp = sdkp->device;
230 	bool v;
231 
232 	if (!capable(CAP_SYS_ADMIN))
233 		return -EACCES;
234 
235 	if (kstrtobool(buf, &v))
236 		return -EINVAL;
237 
238 	sdp->manage_start_stop = v;
239 
240 	return count;
241 }
242 static DEVICE_ATTR_RW(manage_start_stop);
243 
244 static ssize_t
245 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
246 {
247 	struct scsi_disk *sdkp = to_scsi_disk(dev);
248 
249 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
250 }
251 
252 static ssize_t
253 allow_restart_store(struct device *dev, struct device_attribute *attr,
254 		    const char *buf, size_t count)
255 {
256 	bool v;
257 	struct scsi_disk *sdkp = to_scsi_disk(dev);
258 	struct scsi_device *sdp = sdkp->device;
259 
260 	if (!capable(CAP_SYS_ADMIN))
261 		return -EACCES;
262 
263 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
264 		return -EINVAL;
265 
266 	if (kstrtobool(buf, &v))
267 		return -EINVAL;
268 
269 	sdp->allow_restart = v;
270 
271 	return count;
272 }
273 static DEVICE_ATTR_RW(allow_restart);
274 
275 static ssize_t
276 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
277 {
278 	struct scsi_disk *sdkp = to_scsi_disk(dev);
279 	int ct = sdkp->RCD + 2*sdkp->WCE;
280 
281 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
282 }
283 static DEVICE_ATTR_RW(cache_type);
284 
285 static ssize_t
286 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
287 {
288 	struct scsi_disk *sdkp = to_scsi_disk(dev);
289 
290 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
291 }
292 static DEVICE_ATTR_RO(FUA);
293 
294 static ssize_t
295 protection_type_show(struct device *dev, struct device_attribute *attr,
296 		     char *buf)
297 {
298 	struct scsi_disk *sdkp = to_scsi_disk(dev);
299 
300 	return sprintf(buf, "%u\n", sdkp->protection_type);
301 }
302 
303 static ssize_t
304 protection_type_store(struct device *dev, struct device_attribute *attr,
305 		      const char *buf, size_t count)
306 {
307 	struct scsi_disk *sdkp = to_scsi_disk(dev);
308 	unsigned int val;
309 	int err;
310 
311 	if (!capable(CAP_SYS_ADMIN))
312 		return -EACCES;
313 
314 	err = kstrtouint(buf, 10, &val);
315 
316 	if (err)
317 		return err;
318 
319 	if (val <= T10_PI_TYPE3_PROTECTION)
320 		sdkp->protection_type = val;
321 
322 	return count;
323 }
324 static DEVICE_ATTR_RW(protection_type);
325 
326 static ssize_t
327 protection_mode_show(struct device *dev, struct device_attribute *attr,
328 		     char *buf)
329 {
330 	struct scsi_disk *sdkp = to_scsi_disk(dev);
331 	struct scsi_device *sdp = sdkp->device;
332 	unsigned int dif, dix;
333 
334 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
335 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
336 
337 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
338 		dif = 0;
339 		dix = 1;
340 	}
341 
342 	if (!dif && !dix)
343 		return sprintf(buf, "none\n");
344 
345 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
346 }
347 static DEVICE_ATTR_RO(protection_mode);
348 
349 static ssize_t
350 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
351 {
352 	struct scsi_disk *sdkp = to_scsi_disk(dev);
353 
354 	return sprintf(buf, "%u\n", sdkp->ATO);
355 }
356 static DEVICE_ATTR_RO(app_tag_own);
357 
358 static ssize_t
359 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
360 		       char *buf)
361 {
362 	struct scsi_disk *sdkp = to_scsi_disk(dev);
363 
364 	return sprintf(buf, "%u\n", sdkp->lbpme);
365 }
366 static DEVICE_ATTR_RO(thin_provisioning);
367 
368 /* sysfs_match_string() requires dense arrays */
369 static const char *lbp_mode[] = {
370 	[SD_LBP_FULL]		= "full",
371 	[SD_LBP_UNMAP]		= "unmap",
372 	[SD_LBP_WS16]		= "writesame_16",
373 	[SD_LBP_WS10]		= "writesame_10",
374 	[SD_LBP_ZERO]		= "writesame_zero",
375 	[SD_LBP_DISABLE]	= "disabled",
376 };
377 
378 static ssize_t
379 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
380 		       char *buf)
381 {
382 	struct scsi_disk *sdkp = to_scsi_disk(dev);
383 
384 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
385 }
386 
387 static ssize_t
388 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
389 			const char *buf, size_t count)
390 {
391 	struct scsi_disk *sdkp = to_scsi_disk(dev);
392 	struct scsi_device *sdp = sdkp->device;
393 	int mode;
394 
395 	if (!capable(CAP_SYS_ADMIN))
396 		return -EACCES;
397 
398 	if (sd_is_zoned(sdkp)) {
399 		sd_config_discard(sdkp, SD_LBP_DISABLE);
400 		return count;
401 	}
402 
403 	if (sdp->type != TYPE_DISK)
404 		return -EINVAL;
405 
406 	mode = sysfs_match_string(lbp_mode, buf);
407 	if (mode < 0)
408 		return -EINVAL;
409 
410 	sd_config_discard(sdkp, mode);
411 
412 	return count;
413 }
414 static DEVICE_ATTR_RW(provisioning_mode);
415 
416 /* sysfs_match_string() requires dense arrays */
417 static const char *zeroing_mode[] = {
418 	[SD_ZERO_WRITE]		= "write",
419 	[SD_ZERO_WS]		= "writesame",
420 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
421 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
422 };
423 
424 static ssize_t
425 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
426 		  char *buf)
427 {
428 	struct scsi_disk *sdkp = to_scsi_disk(dev);
429 
430 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
431 }
432 
433 static ssize_t
434 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
435 		   const char *buf, size_t count)
436 {
437 	struct scsi_disk *sdkp = to_scsi_disk(dev);
438 	int mode;
439 
440 	if (!capable(CAP_SYS_ADMIN))
441 		return -EACCES;
442 
443 	mode = sysfs_match_string(zeroing_mode, buf);
444 	if (mode < 0)
445 		return -EINVAL;
446 
447 	sdkp->zeroing_mode = mode;
448 
449 	return count;
450 }
451 static DEVICE_ATTR_RW(zeroing_mode);
452 
453 static ssize_t
454 max_medium_access_timeouts_show(struct device *dev,
455 				struct device_attribute *attr, char *buf)
456 {
457 	struct scsi_disk *sdkp = to_scsi_disk(dev);
458 
459 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
460 }
461 
462 static ssize_t
463 max_medium_access_timeouts_store(struct device *dev,
464 				 struct device_attribute *attr, const char *buf,
465 				 size_t count)
466 {
467 	struct scsi_disk *sdkp = to_scsi_disk(dev);
468 	int err;
469 
470 	if (!capable(CAP_SYS_ADMIN))
471 		return -EACCES;
472 
473 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
474 
475 	return err ? err : count;
476 }
477 static DEVICE_ATTR_RW(max_medium_access_timeouts);
478 
479 static ssize_t
480 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
481 			   char *buf)
482 {
483 	struct scsi_disk *sdkp = to_scsi_disk(dev);
484 
485 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
486 }
487 
488 static ssize_t
489 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
490 			    const char *buf, size_t count)
491 {
492 	struct scsi_disk *sdkp = to_scsi_disk(dev);
493 	struct scsi_device *sdp = sdkp->device;
494 	unsigned long max;
495 	int err;
496 
497 	if (!capable(CAP_SYS_ADMIN))
498 		return -EACCES;
499 
500 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
501 		return -EINVAL;
502 
503 	err = kstrtoul(buf, 10, &max);
504 
505 	if (err)
506 		return err;
507 
508 	if (max == 0)
509 		sdp->no_write_same = 1;
510 	else if (max <= SD_MAX_WS16_BLOCKS) {
511 		sdp->no_write_same = 0;
512 		sdkp->max_ws_blocks = max;
513 	}
514 
515 	sd_config_write_same(sdkp);
516 
517 	return count;
518 }
519 static DEVICE_ATTR_RW(max_write_same_blocks);
520 
521 static ssize_t
522 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
523 {
524 	struct scsi_disk *sdkp = to_scsi_disk(dev);
525 
526 	if (sdkp->device->type == TYPE_ZBC)
527 		return sprintf(buf, "host-managed\n");
528 	if (sdkp->zoned == 1)
529 		return sprintf(buf, "host-aware\n");
530 	if (sdkp->zoned == 2)
531 		return sprintf(buf, "drive-managed\n");
532 	return sprintf(buf, "none\n");
533 }
534 static DEVICE_ATTR_RO(zoned_cap);
535 
536 static ssize_t
537 max_retries_store(struct device *dev, struct device_attribute *attr,
538 		  const char *buf, size_t count)
539 {
540 	struct scsi_disk *sdkp = to_scsi_disk(dev);
541 	struct scsi_device *sdev = sdkp->device;
542 	int retries, err;
543 
544 	err = kstrtoint(buf, 10, &retries);
545 	if (err)
546 		return err;
547 
548 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
549 		sdkp->max_retries = retries;
550 		return count;
551 	}
552 
553 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
554 		    SD_MAX_RETRIES);
555 	return -EINVAL;
556 }
557 
558 static ssize_t
559 max_retries_show(struct device *dev, struct device_attribute *attr,
560 		 char *buf)
561 {
562 	struct scsi_disk *sdkp = to_scsi_disk(dev);
563 
564 	return sprintf(buf, "%d\n", sdkp->max_retries);
565 }
566 
567 static DEVICE_ATTR_RW(max_retries);
568 
569 static struct attribute *sd_disk_attrs[] = {
570 	&dev_attr_cache_type.attr,
571 	&dev_attr_FUA.attr,
572 	&dev_attr_allow_restart.attr,
573 	&dev_attr_manage_start_stop.attr,
574 	&dev_attr_protection_type.attr,
575 	&dev_attr_protection_mode.attr,
576 	&dev_attr_app_tag_own.attr,
577 	&dev_attr_thin_provisioning.attr,
578 	&dev_attr_provisioning_mode.attr,
579 	&dev_attr_zeroing_mode.attr,
580 	&dev_attr_max_write_same_blocks.attr,
581 	&dev_attr_max_medium_access_timeouts.attr,
582 	&dev_attr_zoned_cap.attr,
583 	&dev_attr_max_retries.attr,
584 	NULL,
585 };
586 ATTRIBUTE_GROUPS(sd_disk);
587 
588 static struct class sd_disk_class = {
589 	.name		= "scsi_disk",
590 	.owner		= THIS_MODULE,
591 	.dev_release	= scsi_disk_release,
592 	.dev_groups	= sd_disk_groups,
593 };
594 
595 static const struct dev_pm_ops sd_pm_ops = {
596 	.suspend		= sd_suspend_system,
597 	.resume			= sd_resume_system,
598 	.poweroff		= sd_suspend_system,
599 	.restore		= sd_resume_system,
600 	.runtime_suspend	= sd_suspend_runtime,
601 	.runtime_resume		= sd_resume_runtime,
602 };
603 
604 static struct scsi_driver sd_template = {
605 	.gendrv = {
606 		.name		= "sd",
607 		.owner		= THIS_MODULE,
608 		.probe		= sd_probe,
609 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
610 		.remove		= sd_remove,
611 		.shutdown	= sd_shutdown,
612 		.pm		= &sd_pm_ops,
613 	},
614 	.rescan			= sd_rescan,
615 	.init_command		= sd_init_command,
616 	.uninit_command		= sd_uninit_command,
617 	.done			= sd_done,
618 	.eh_action		= sd_eh_action,
619 	.eh_reset		= sd_eh_reset,
620 };
621 
622 /*
623  * Don't request a new module, as that could deadlock in multipath
624  * environment.
625  */
626 static void sd_default_probe(dev_t devt)
627 {
628 }
629 
630 /*
631  * Device no to disk mapping:
632  *
633  *       major         disc2     disc  p1
634  *   |............|.............|....|....| <- dev_t
635  *    31        20 19          8 7  4 3  0
636  *
637  * Inside a major, we have 16k disks, however mapped non-
638  * contiguously. The first 16 disks are for major0, the next
639  * ones with major1, ... Disk 256 is for major0 again, disk 272
640  * for major1, ...
641  * As we stay compatible with our numbering scheme, we can reuse
642  * the well-know SCSI majors 8, 65--71, 136--143.
643  */
644 static int sd_major(int major_idx)
645 {
646 	switch (major_idx) {
647 	case 0:
648 		return SCSI_DISK0_MAJOR;
649 	case 1 ... 7:
650 		return SCSI_DISK1_MAJOR + major_idx - 1;
651 	case 8 ... 15:
652 		return SCSI_DISK8_MAJOR + major_idx - 8;
653 	default:
654 		BUG();
655 		return 0;	/* shut up gcc */
656 	}
657 }
658 
659 #ifdef CONFIG_BLK_SED_OPAL
660 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
661 		size_t len, bool send)
662 {
663 	struct scsi_disk *sdkp = data;
664 	struct scsi_device *sdev = sdkp->device;
665 	u8 cdb[12] = { 0, };
666 	const struct scsi_exec_args exec_args = {
667 		.req_flags = BLK_MQ_REQ_PM,
668 	};
669 	int ret;
670 
671 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
672 	cdb[1] = secp;
673 	put_unaligned_be16(spsp, &cdb[2]);
674 	put_unaligned_be32(len, &cdb[6]);
675 
676 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
677 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
678 			       &exec_args);
679 	return ret <= 0 ? ret : -EIO;
680 }
681 #endif /* CONFIG_BLK_SED_OPAL */
682 
683 /*
684  * Look up the DIX operation based on whether the command is read or
685  * write and whether dix and dif are enabled.
686  */
687 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
688 {
689 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
690 	static const unsigned int ops[] = {	/* wrt dix dif */
691 		SCSI_PROT_NORMAL,		/*  0	0   0  */
692 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
693 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
694 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
695 		SCSI_PROT_NORMAL,		/*  1	0   0  */
696 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
697 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
698 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
699 	};
700 
701 	return ops[write << 2 | dix << 1 | dif];
702 }
703 
704 /*
705  * Returns a mask of the protection flags that are valid for a given DIX
706  * operation.
707  */
708 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
709 {
710 	static const unsigned int flag_mask[] = {
711 		[SCSI_PROT_NORMAL]		= 0,
712 
713 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
714 						  SCSI_PROT_GUARD_CHECK |
715 						  SCSI_PROT_REF_CHECK |
716 						  SCSI_PROT_REF_INCREMENT,
717 
718 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
719 						  SCSI_PROT_IP_CHECKSUM,
720 
721 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
722 						  SCSI_PROT_GUARD_CHECK |
723 						  SCSI_PROT_REF_CHECK |
724 						  SCSI_PROT_REF_INCREMENT |
725 						  SCSI_PROT_IP_CHECKSUM,
726 
727 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
728 						  SCSI_PROT_REF_INCREMENT,
729 
730 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
731 						  SCSI_PROT_REF_CHECK |
732 						  SCSI_PROT_REF_INCREMENT |
733 						  SCSI_PROT_IP_CHECKSUM,
734 
735 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
736 						  SCSI_PROT_GUARD_CHECK |
737 						  SCSI_PROT_REF_CHECK |
738 						  SCSI_PROT_REF_INCREMENT |
739 						  SCSI_PROT_IP_CHECKSUM,
740 	};
741 
742 	return flag_mask[prot_op];
743 }
744 
745 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
746 					   unsigned int dix, unsigned int dif)
747 {
748 	struct request *rq = scsi_cmd_to_rq(scmd);
749 	struct bio *bio = rq->bio;
750 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
751 	unsigned int protect = 0;
752 
753 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
754 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
755 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
756 
757 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
758 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
759 	}
760 
761 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
762 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
763 
764 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
765 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
766 	}
767 
768 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
769 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
770 
771 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
772 			protect = 3 << 5;	/* Disable target PI checking */
773 		else
774 			protect = 1 << 5;	/* Enable target PI checking */
775 	}
776 
777 	scsi_set_prot_op(scmd, prot_op);
778 	scsi_set_prot_type(scmd, dif);
779 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
780 
781 	return protect;
782 }
783 
784 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
785 {
786 	struct request_queue *q = sdkp->disk->queue;
787 	unsigned int logical_block_size = sdkp->device->sector_size;
788 	unsigned int max_blocks = 0;
789 
790 	q->limits.discard_alignment =
791 		sdkp->unmap_alignment * logical_block_size;
792 	q->limits.discard_granularity =
793 		max(sdkp->physical_block_size,
794 		    sdkp->unmap_granularity * logical_block_size);
795 	sdkp->provisioning_mode = mode;
796 
797 	switch (mode) {
798 
799 	case SD_LBP_FULL:
800 	case SD_LBP_DISABLE:
801 		blk_queue_max_discard_sectors(q, 0);
802 		return;
803 
804 	case SD_LBP_UNMAP:
805 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
806 					  (u32)SD_MAX_WS16_BLOCKS);
807 		break;
808 
809 	case SD_LBP_WS16:
810 		if (sdkp->device->unmap_limit_for_ws)
811 			max_blocks = sdkp->max_unmap_blocks;
812 		else
813 			max_blocks = sdkp->max_ws_blocks;
814 
815 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
816 		break;
817 
818 	case SD_LBP_WS10:
819 		if (sdkp->device->unmap_limit_for_ws)
820 			max_blocks = sdkp->max_unmap_blocks;
821 		else
822 			max_blocks = sdkp->max_ws_blocks;
823 
824 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
825 		break;
826 
827 	case SD_LBP_ZERO:
828 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
829 					  (u32)SD_MAX_WS10_BLOCKS);
830 		break;
831 	}
832 
833 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
834 }
835 
836 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
837 {
838 	struct page *page;
839 
840 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
841 	if (!page)
842 		return NULL;
843 	clear_highpage(page);
844 	bvec_set_page(&rq->special_vec, page, data_len, 0);
845 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
846 	return bvec_virt(&rq->special_vec);
847 }
848 
849 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
850 {
851 	struct scsi_device *sdp = cmd->device;
852 	struct request *rq = scsi_cmd_to_rq(cmd);
853 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
854 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
855 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
856 	unsigned int data_len = 24;
857 	char *buf;
858 
859 	buf = sd_set_special_bvec(rq, data_len);
860 	if (!buf)
861 		return BLK_STS_RESOURCE;
862 
863 	cmd->cmd_len = 10;
864 	cmd->cmnd[0] = UNMAP;
865 	cmd->cmnd[8] = 24;
866 
867 	put_unaligned_be16(6 + 16, &buf[0]);
868 	put_unaligned_be16(16, &buf[2]);
869 	put_unaligned_be64(lba, &buf[8]);
870 	put_unaligned_be32(nr_blocks, &buf[16]);
871 
872 	cmd->allowed = sdkp->max_retries;
873 	cmd->transfersize = data_len;
874 	rq->timeout = SD_TIMEOUT;
875 
876 	return scsi_alloc_sgtables(cmd);
877 }
878 
879 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
880 		bool unmap)
881 {
882 	struct scsi_device *sdp = cmd->device;
883 	struct request *rq = scsi_cmd_to_rq(cmd);
884 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
885 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
886 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
887 	u32 data_len = sdp->sector_size;
888 
889 	if (!sd_set_special_bvec(rq, data_len))
890 		return BLK_STS_RESOURCE;
891 
892 	cmd->cmd_len = 16;
893 	cmd->cmnd[0] = WRITE_SAME_16;
894 	if (unmap)
895 		cmd->cmnd[1] = 0x8; /* UNMAP */
896 	put_unaligned_be64(lba, &cmd->cmnd[2]);
897 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
898 
899 	cmd->allowed = sdkp->max_retries;
900 	cmd->transfersize = data_len;
901 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
902 
903 	return scsi_alloc_sgtables(cmd);
904 }
905 
906 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
907 		bool unmap)
908 {
909 	struct scsi_device *sdp = cmd->device;
910 	struct request *rq = scsi_cmd_to_rq(cmd);
911 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
912 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
913 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
914 	u32 data_len = sdp->sector_size;
915 
916 	if (!sd_set_special_bvec(rq, data_len))
917 		return BLK_STS_RESOURCE;
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 	const struct scsi_exec_args exec_args = {
1587 		.req_flags = BLK_MQ_REQ_PM,
1588 		/* caller might not be interested in sense, but we need it */
1589 		.sshdr = sshdr ? : &my_sshdr,
1590 	};
1591 
1592 	if (!scsi_device_online(sdp))
1593 		return -ENODEV;
1594 
1595 	sshdr = exec_args.sshdr;
1596 
1597 	for (retries = 3; retries > 0; --retries) {
1598 		unsigned char cmd[16] = { 0 };
1599 
1600 		if (sdp->use_16_for_sync)
1601 			cmd[0] = SYNCHRONIZE_CACHE_16;
1602 		else
1603 			cmd[0] = SYNCHRONIZE_CACHE;
1604 		/*
1605 		 * Leave the rest of the command zero to indicate
1606 		 * flush everything.
1607 		 */
1608 		res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1609 				       timeout, sdkp->max_retries, &exec_args);
1610 		if (res == 0)
1611 			break;
1612 	}
1613 
1614 	if (res) {
1615 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1616 
1617 		if (res < 0)
1618 			return res;
1619 
1620 		if (scsi_status_is_check_condition(res) &&
1621 		    scsi_sense_valid(sshdr)) {
1622 			sd_print_sense_hdr(sdkp, sshdr);
1623 
1624 			/* we need to evaluate the error return  */
1625 			if (sshdr->asc == 0x3a ||	/* medium not present */
1626 			    sshdr->asc == 0x20 ||	/* invalid command */
1627 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1628 				/* this is no error here */
1629 				return 0;
1630 		}
1631 
1632 		switch (host_byte(res)) {
1633 		/* ignore errors due to racing a disconnection */
1634 		case DID_BAD_TARGET:
1635 		case DID_NO_CONNECT:
1636 			return 0;
1637 		/* signal the upper layer it might try again */
1638 		case DID_BUS_BUSY:
1639 		case DID_IMM_RETRY:
1640 		case DID_REQUEUE:
1641 		case DID_SOFT_ERROR:
1642 			return -EBUSY;
1643 		default:
1644 			return -EIO;
1645 		}
1646 	}
1647 	return 0;
1648 }
1649 
1650 static void sd_rescan(struct device *dev)
1651 {
1652 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1653 
1654 	sd_revalidate_disk(sdkp->disk);
1655 }
1656 
1657 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1658 		enum blk_unique_id type)
1659 {
1660 	struct scsi_device *sdev = scsi_disk(disk)->device;
1661 	const struct scsi_vpd *vpd;
1662 	const unsigned char *d;
1663 	int ret = -ENXIO, len;
1664 
1665 	rcu_read_lock();
1666 	vpd = rcu_dereference(sdev->vpd_pg83);
1667 	if (!vpd)
1668 		goto out_unlock;
1669 
1670 	ret = -EINVAL;
1671 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1672 		/* we only care about designators with LU association */
1673 		if (((d[1] >> 4) & 0x3) != 0x00)
1674 			continue;
1675 		if ((d[1] & 0xf) != type)
1676 			continue;
1677 
1678 		/*
1679 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1680 		 * keep looking as one with more entropy might still show up.
1681 		 */
1682 		len = d[3];
1683 		if (len != 8 && len != 12 && len != 16)
1684 			continue;
1685 		ret = len;
1686 		memcpy(id, d + 4, len);
1687 		if (len == 16)
1688 			break;
1689 	}
1690 out_unlock:
1691 	rcu_read_unlock();
1692 	return ret;
1693 }
1694 
1695 static char sd_pr_type(enum pr_type type)
1696 {
1697 	switch (type) {
1698 	case PR_WRITE_EXCLUSIVE:
1699 		return 0x01;
1700 	case PR_EXCLUSIVE_ACCESS:
1701 		return 0x03;
1702 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1703 		return 0x05;
1704 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1705 		return 0x06;
1706 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1707 		return 0x07;
1708 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1709 		return 0x08;
1710 	default:
1711 		return 0;
1712 	}
1713 };
1714 
1715 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1716 {
1717 	switch (host_byte(result)) {
1718 	case DID_TRANSPORT_MARGINAL:
1719 	case DID_TRANSPORT_DISRUPTED:
1720 	case DID_BUS_BUSY:
1721 		return PR_STS_RETRY_PATH_FAILURE;
1722 	case DID_NO_CONNECT:
1723 		return PR_STS_PATH_FAILED;
1724 	case DID_TRANSPORT_FAILFAST:
1725 		return PR_STS_PATH_FAST_FAILED;
1726 	}
1727 
1728 	switch (status_byte(result)) {
1729 	case SAM_STAT_RESERVATION_CONFLICT:
1730 		return PR_STS_RESERVATION_CONFLICT;
1731 	case SAM_STAT_CHECK_CONDITION:
1732 		if (!scsi_sense_valid(sshdr))
1733 			return PR_STS_IOERR;
1734 
1735 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1736 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1737 			return -EINVAL;
1738 
1739 		fallthrough;
1740 	default:
1741 		return PR_STS_IOERR;
1742 	}
1743 }
1744 
1745 static int sd_pr_command(struct block_device *bdev, u8 sa,
1746 		u64 key, u64 sa_key, u8 type, u8 flags)
1747 {
1748 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1749 	struct scsi_device *sdev = sdkp->device;
1750 	struct scsi_sense_hdr sshdr;
1751 	const struct scsi_exec_args exec_args = {
1752 		.sshdr = &sshdr,
1753 	};
1754 	int result;
1755 	u8 cmd[16] = { 0, };
1756 	u8 data[24] = { 0, };
1757 
1758 	cmd[0] = PERSISTENT_RESERVE_OUT;
1759 	cmd[1] = sa;
1760 	cmd[2] = type;
1761 	put_unaligned_be32(sizeof(data), &cmd[5]);
1762 
1763 	put_unaligned_be64(key, &data[0]);
1764 	put_unaligned_be64(sa_key, &data[8]);
1765 	data[20] = flags;
1766 
1767 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1768 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1769 				  &exec_args);
1770 
1771 	if (scsi_status_is_check_condition(result) &&
1772 	    scsi_sense_valid(&sshdr)) {
1773 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1774 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1775 	}
1776 
1777 	if (result <= 0)
1778 		return result;
1779 
1780 	return sd_scsi_to_pr_err(&sshdr, result);
1781 }
1782 
1783 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1784 		u32 flags)
1785 {
1786 	if (flags & ~PR_FL_IGNORE_KEY)
1787 		return -EOPNOTSUPP;
1788 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1789 			old_key, new_key, 0,
1790 			(1 << 0) /* APTPL */);
1791 }
1792 
1793 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1794 		u32 flags)
1795 {
1796 	if (flags)
1797 		return -EOPNOTSUPP;
1798 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1799 }
1800 
1801 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1802 {
1803 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1804 }
1805 
1806 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1807 		enum pr_type type, bool abort)
1808 {
1809 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1810 			     sd_pr_type(type), 0);
1811 }
1812 
1813 static int sd_pr_clear(struct block_device *bdev, u64 key)
1814 {
1815 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1816 }
1817 
1818 static const struct pr_ops sd_pr_ops = {
1819 	.pr_register	= sd_pr_register,
1820 	.pr_reserve	= sd_pr_reserve,
1821 	.pr_release	= sd_pr_release,
1822 	.pr_preempt	= sd_pr_preempt,
1823 	.pr_clear	= sd_pr_clear,
1824 };
1825 
1826 static void scsi_disk_free_disk(struct gendisk *disk)
1827 {
1828 	struct scsi_disk *sdkp = scsi_disk(disk);
1829 
1830 	put_device(&sdkp->disk_dev);
1831 }
1832 
1833 static const struct block_device_operations sd_fops = {
1834 	.owner			= THIS_MODULE,
1835 	.open			= sd_open,
1836 	.release		= sd_release,
1837 	.ioctl			= sd_ioctl,
1838 	.getgeo			= sd_getgeo,
1839 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1840 	.check_events		= sd_check_events,
1841 	.unlock_native_capacity	= sd_unlock_native_capacity,
1842 	.report_zones		= sd_zbc_report_zones,
1843 	.get_unique_id		= sd_get_unique_id,
1844 	.free_disk		= scsi_disk_free_disk,
1845 	.pr_ops			= &sd_pr_ops,
1846 };
1847 
1848 /**
1849  *	sd_eh_reset - reset error handling callback
1850  *	@scmd:		sd-issued command that has failed
1851  *
1852  *	This function is called by the SCSI midlayer before starting
1853  *	SCSI EH. When counting medium access failures we have to be
1854  *	careful to register it only only once per device and SCSI EH run;
1855  *	there might be several timed out commands which will cause the
1856  *	'max_medium_access_timeouts' counter to trigger after the first
1857  *	SCSI EH run already and set the device to offline.
1858  *	So this function resets the internal counter before starting SCSI EH.
1859  **/
1860 static void sd_eh_reset(struct scsi_cmnd *scmd)
1861 {
1862 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1863 
1864 	/* New SCSI EH run, reset gate variable */
1865 	sdkp->ignore_medium_access_errors = false;
1866 }
1867 
1868 /**
1869  *	sd_eh_action - error handling callback
1870  *	@scmd:		sd-issued command that has failed
1871  *	@eh_disp:	The recovery disposition suggested by the midlayer
1872  *
1873  *	This function is called by the SCSI midlayer upon completion of an
1874  *	error test command (currently TEST UNIT READY). The result of sending
1875  *	the eh command is passed in eh_disp.  We're looking for devices that
1876  *	fail medium access commands but are OK with non access commands like
1877  *	test unit ready (so wrongly see the device as having a successful
1878  *	recovery)
1879  **/
1880 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1881 {
1882 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1883 	struct scsi_device *sdev = scmd->device;
1884 
1885 	if (!scsi_device_online(sdev) ||
1886 	    !scsi_medium_access_command(scmd) ||
1887 	    host_byte(scmd->result) != DID_TIME_OUT ||
1888 	    eh_disp != SUCCESS)
1889 		return eh_disp;
1890 
1891 	/*
1892 	 * The device has timed out executing a medium access command.
1893 	 * However, the TEST UNIT READY command sent during error
1894 	 * handling completed successfully. Either the device is in the
1895 	 * process of recovering or has it suffered an internal failure
1896 	 * that prevents access to the storage medium.
1897 	 */
1898 	if (!sdkp->ignore_medium_access_errors) {
1899 		sdkp->medium_access_timed_out++;
1900 		sdkp->ignore_medium_access_errors = true;
1901 	}
1902 
1903 	/*
1904 	 * If the device keeps failing read/write commands but TEST UNIT
1905 	 * READY always completes successfully we assume that medium
1906 	 * access is no longer possible and take the device offline.
1907 	 */
1908 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1909 		scmd_printk(KERN_ERR, scmd,
1910 			    "Medium access timeout failure. Offlining disk!\n");
1911 		mutex_lock(&sdev->state_mutex);
1912 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1913 		mutex_unlock(&sdev->state_mutex);
1914 
1915 		return SUCCESS;
1916 	}
1917 
1918 	return eh_disp;
1919 }
1920 
1921 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1922 {
1923 	struct request *req = scsi_cmd_to_rq(scmd);
1924 	struct scsi_device *sdev = scmd->device;
1925 	unsigned int transferred, good_bytes;
1926 	u64 start_lba, end_lba, bad_lba;
1927 
1928 	/*
1929 	 * Some commands have a payload smaller than the device logical
1930 	 * block size (e.g. INQUIRY on a 4K disk).
1931 	 */
1932 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1933 		return 0;
1934 
1935 	/* Check if we have a 'bad_lba' information */
1936 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1937 				     SCSI_SENSE_BUFFERSIZE,
1938 				     &bad_lba))
1939 		return 0;
1940 
1941 	/*
1942 	 * If the bad lba was reported incorrectly, we have no idea where
1943 	 * the error is.
1944 	 */
1945 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1946 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1947 	if (bad_lba < start_lba || bad_lba >= end_lba)
1948 		return 0;
1949 
1950 	/*
1951 	 * resid is optional but mostly filled in.  When it's unused,
1952 	 * its value is zero, so we assume the whole buffer transferred
1953 	 */
1954 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1955 
1956 	/* This computation should always be done in terms of the
1957 	 * resolution of the device's medium.
1958 	 */
1959 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1960 
1961 	return min(good_bytes, transferred);
1962 }
1963 
1964 /**
1965  *	sd_done - bottom half handler: called when the lower level
1966  *	driver has completed (successfully or otherwise) a scsi command.
1967  *	@SCpnt: mid-level's per command structure.
1968  *
1969  *	Note: potentially run from within an ISR. Must not block.
1970  **/
1971 static int sd_done(struct scsi_cmnd *SCpnt)
1972 {
1973 	int result = SCpnt->result;
1974 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1975 	unsigned int sector_size = SCpnt->device->sector_size;
1976 	unsigned int resid;
1977 	struct scsi_sense_hdr sshdr;
1978 	struct request *req = scsi_cmd_to_rq(SCpnt);
1979 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1980 	int sense_valid = 0;
1981 	int sense_deferred = 0;
1982 
1983 	switch (req_op(req)) {
1984 	case REQ_OP_DISCARD:
1985 	case REQ_OP_WRITE_ZEROES:
1986 	case REQ_OP_ZONE_RESET:
1987 	case REQ_OP_ZONE_RESET_ALL:
1988 	case REQ_OP_ZONE_OPEN:
1989 	case REQ_OP_ZONE_CLOSE:
1990 	case REQ_OP_ZONE_FINISH:
1991 		if (!result) {
1992 			good_bytes = blk_rq_bytes(req);
1993 			scsi_set_resid(SCpnt, 0);
1994 		} else {
1995 			good_bytes = 0;
1996 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1997 		}
1998 		break;
1999 	default:
2000 		/*
2001 		 * In case of bogus fw or device, we could end up having
2002 		 * an unaligned partial completion. Check this here and force
2003 		 * alignment.
2004 		 */
2005 		resid = scsi_get_resid(SCpnt);
2006 		if (resid & (sector_size - 1)) {
2007 			sd_printk(KERN_INFO, sdkp,
2008 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2009 				resid, sector_size);
2010 			scsi_print_command(SCpnt);
2011 			resid = min(scsi_bufflen(SCpnt),
2012 				    round_up(resid, sector_size));
2013 			scsi_set_resid(SCpnt, resid);
2014 		}
2015 	}
2016 
2017 	if (result) {
2018 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2019 		if (sense_valid)
2020 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2021 	}
2022 	sdkp->medium_access_timed_out = 0;
2023 
2024 	if (!scsi_status_is_check_condition(result) &&
2025 	    (!sense_valid || sense_deferred))
2026 		goto out;
2027 
2028 	switch (sshdr.sense_key) {
2029 	case HARDWARE_ERROR:
2030 	case MEDIUM_ERROR:
2031 		good_bytes = sd_completed_bytes(SCpnt);
2032 		break;
2033 	case RECOVERED_ERROR:
2034 		good_bytes = scsi_bufflen(SCpnt);
2035 		break;
2036 	case NO_SENSE:
2037 		/* This indicates a false check condition, so ignore it.  An
2038 		 * unknown amount of data was transferred so treat it as an
2039 		 * error.
2040 		 */
2041 		SCpnt->result = 0;
2042 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2043 		break;
2044 	case ABORTED_COMMAND:
2045 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2046 			good_bytes = sd_completed_bytes(SCpnt);
2047 		break;
2048 	case ILLEGAL_REQUEST:
2049 		switch (sshdr.asc) {
2050 		case 0x10:	/* DIX: Host detected corruption */
2051 			good_bytes = sd_completed_bytes(SCpnt);
2052 			break;
2053 		case 0x20:	/* INVALID COMMAND OPCODE */
2054 		case 0x24:	/* INVALID FIELD IN CDB */
2055 			switch (SCpnt->cmnd[0]) {
2056 			case UNMAP:
2057 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2058 				break;
2059 			case WRITE_SAME_16:
2060 			case WRITE_SAME:
2061 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2062 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2063 				} else {
2064 					sdkp->device->no_write_same = 1;
2065 					sd_config_write_same(sdkp);
2066 					req->rq_flags |= RQF_QUIET;
2067 				}
2068 				break;
2069 			}
2070 		}
2071 		break;
2072 	default:
2073 		break;
2074 	}
2075 
2076  out:
2077 	if (sd_is_zoned(sdkp))
2078 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2079 
2080 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2081 					   "sd_done: completed %d of %d bytes\n",
2082 					   good_bytes, scsi_bufflen(SCpnt)));
2083 
2084 	return good_bytes;
2085 }
2086 
2087 /*
2088  * spinup disk - called only in sd_revalidate_disk()
2089  */
2090 static void
2091 sd_spinup_disk(struct scsi_disk *sdkp)
2092 {
2093 	unsigned char cmd[10];
2094 	unsigned long spintime_expire = 0;
2095 	int retries, spintime;
2096 	unsigned int the_result;
2097 	struct scsi_sense_hdr sshdr;
2098 	const struct scsi_exec_args exec_args = {
2099 		.sshdr = &sshdr,
2100 	};
2101 	int sense_valid = 0;
2102 
2103 	spintime = 0;
2104 
2105 	/* Spin up drives, as required.  Only do this at boot time */
2106 	/* Spinup needs to be done for module loads too. */
2107 	do {
2108 		retries = 0;
2109 
2110 		do {
2111 			bool media_was_present = sdkp->media_present;
2112 
2113 			cmd[0] = TEST_UNIT_READY;
2114 			memset((void *) &cmd[1], 0, 9);
2115 
2116 			the_result = scsi_execute_cmd(sdkp->device, cmd,
2117 						      REQ_OP_DRV_IN, NULL, 0,
2118 						      SD_TIMEOUT,
2119 						      sdkp->max_retries,
2120 						      &exec_args);
2121 
2122 			/*
2123 			 * If the drive has indicated to us that it
2124 			 * doesn't have any media in it, don't bother
2125 			 * with any more polling.
2126 			 */
2127 			if (media_not_present(sdkp, &sshdr)) {
2128 				if (media_was_present)
2129 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2130 				return;
2131 			}
2132 
2133 			if (the_result)
2134 				sense_valid = scsi_sense_valid(&sshdr);
2135 			retries++;
2136 		} while (retries < 3 &&
2137 			 (!scsi_status_is_good(the_result) ||
2138 			  (scsi_status_is_check_condition(the_result) &&
2139 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2140 
2141 		if (!scsi_status_is_check_condition(the_result)) {
2142 			/* no sense, TUR either succeeded or failed
2143 			 * with a status error */
2144 			if(!spintime && !scsi_status_is_good(the_result)) {
2145 				sd_print_result(sdkp, "Test Unit Ready failed",
2146 						the_result);
2147 			}
2148 			break;
2149 		}
2150 
2151 		/*
2152 		 * The device does not want the automatic start to be issued.
2153 		 */
2154 		if (sdkp->device->no_start_on_add)
2155 			break;
2156 
2157 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2158 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2159 				break;	/* manual intervention required */
2160 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2161 				break;	/* standby */
2162 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2163 				break;	/* unavailable */
2164 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2165 				break;	/* sanitize in progress */
2166 			/*
2167 			 * Issue command to spin up drive when not ready
2168 			 */
2169 			if (!spintime) {
2170 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2171 				cmd[0] = START_STOP;
2172 				cmd[1] = 1;	/* Return immediately */
2173 				memset((void *) &cmd[2], 0, 8);
2174 				cmd[4] = 1;	/* Start spin cycle */
2175 				if (sdkp->device->start_stop_pwr_cond)
2176 					cmd[4] |= 1 << 4;
2177 				scsi_execute_cmd(sdkp->device, cmd,
2178 						 REQ_OP_DRV_IN, NULL, 0,
2179 						 SD_TIMEOUT, sdkp->max_retries,
2180 						 &exec_args);
2181 				spintime_expire = jiffies + 100 * HZ;
2182 				spintime = 1;
2183 			}
2184 			/* Wait 1 second for next try */
2185 			msleep(1000);
2186 			printk(KERN_CONT ".");
2187 
2188 		/*
2189 		 * Wait for USB flash devices with slow firmware.
2190 		 * Yes, this sense key/ASC combination shouldn't
2191 		 * occur here.  It's characteristic of these devices.
2192 		 */
2193 		} else if (sense_valid &&
2194 				sshdr.sense_key == UNIT_ATTENTION &&
2195 				sshdr.asc == 0x28) {
2196 			if (!spintime) {
2197 				spintime_expire = jiffies + 5 * HZ;
2198 				spintime = 1;
2199 			}
2200 			/* Wait 1 second for next try */
2201 			msleep(1000);
2202 		} else {
2203 			/* we don't understand the sense code, so it's
2204 			 * probably pointless to loop */
2205 			if(!spintime) {
2206 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2207 				sd_print_sense_hdr(sdkp, &sshdr);
2208 			}
2209 			break;
2210 		}
2211 
2212 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2213 
2214 	if (spintime) {
2215 		if (scsi_status_is_good(the_result))
2216 			printk(KERN_CONT "ready\n");
2217 		else
2218 			printk(KERN_CONT "not responding...\n");
2219 	}
2220 }
2221 
2222 /*
2223  * Determine whether disk supports Data Integrity Field.
2224  */
2225 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2226 {
2227 	struct scsi_device *sdp = sdkp->device;
2228 	u8 type;
2229 
2230 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2231 		sdkp->protection_type = 0;
2232 		return 0;
2233 	}
2234 
2235 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2236 
2237 	if (type > T10_PI_TYPE3_PROTECTION) {
2238 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2239 			  " protection type %u. Disabling disk!\n",
2240 			  type);
2241 		sdkp->protection_type = 0;
2242 		return -ENODEV;
2243 	}
2244 
2245 	sdkp->protection_type = type;
2246 
2247 	return 0;
2248 }
2249 
2250 static void sd_config_protection(struct scsi_disk *sdkp)
2251 {
2252 	struct scsi_device *sdp = sdkp->device;
2253 
2254 	sd_dif_config_host(sdkp);
2255 
2256 	if (!sdkp->protection_type)
2257 		return;
2258 
2259 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2260 		sd_first_printk(KERN_NOTICE, sdkp,
2261 				"Disabling DIF Type %u protection\n",
2262 				sdkp->protection_type);
2263 		sdkp->protection_type = 0;
2264 	}
2265 
2266 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2267 			sdkp->protection_type);
2268 }
2269 
2270 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2271 			struct scsi_sense_hdr *sshdr, int sense_valid,
2272 			int the_result)
2273 {
2274 	if (sense_valid)
2275 		sd_print_sense_hdr(sdkp, sshdr);
2276 	else
2277 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2278 
2279 	/*
2280 	 * Set dirty bit for removable devices if not ready -
2281 	 * sometimes drives will not report this properly.
2282 	 */
2283 	if (sdp->removable &&
2284 	    sense_valid && sshdr->sense_key == NOT_READY)
2285 		set_media_not_present(sdkp);
2286 
2287 	/*
2288 	 * We used to set media_present to 0 here to indicate no media
2289 	 * in the drive, but some drives fail read capacity even with
2290 	 * media present, so we can't do that.
2291 	 */
2292 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2293 }
2294 
2295 #define RC16_LEN 32
2296 #if RC16_LEN > SD_BUF_SIZE
2297 #error RC16_LEN must not be more than SD_BUF_SIZE
2298 #endif
2299 
2300 #define READ_CAPACITY_RETRIES_ON_RESET	10
2301 
2302 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2303 						unsigned char *buffer)
2304 {
2305 	unsigned char cmd[16];
2306 	struct scsi_sense_hdr sshdr;
2307 	const struct scsi_exec_args exec_args = {
2308 		.sshdr = &sshdr,
2309 	};
2310 	int sense_valid = 0;
2311 	int the_result;
2312 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2313 	unsigned int alignment;
2314 	unsigned long long lba;
2315 	unsigned sector_size;
2316 
2317 	if (sdp->no_read_capacity_16)
2318 		return -EINVAL;
2319 
2320 	do {
2321 		memset(cmd, 0, 16);
2322 		cmd[0] = SERVICE_ACTION_IN_16;
2323 		cmd[1] = SAI_READ_CAPACITY_16;
2324 		cmd[13] = RC16_LEN;
2325 		memset(buffer, 0, RC16_LEN);
2326 
2327 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2328 					      buffer, RC16_LEN, SD_TIMEOUT,
2329 					      sdkp->max_retries, &exec_args);
2330 
2331 		if (media_not_present(sdkp, &sshdr))
2332 			return -ENODEV;
2333 
2334 		if (the_result > 0) {
2335 			sense_valid = scsi_sense_valid(&sshdr);
2336 			if (sense_valid &&
2337 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2338 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2339 			    sshdr.ascq == 0x00)
2340 				/* Invalid Command Operation Code or
2341 				 * Invalid Field in CDB, just retry
2342 				 * silently with RC10 */
2343 				return -EINVAL;
2344 			if (sense_valid &&
2345 			    sshdr.sense_key == UNIT_ATTENTION &&
2346 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2347 				/* Device reset might occur several times,
2348 				 * give it one more chance */
2349 				if (--reset_retries > 0)
2350 					continue;
2351 		}
2352 		retries--;
2353 
2354 	} while (the_result && retries);
2355 
2356 	if (the_result) {
2357 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2358 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2359 		return -EINVAL;
2360 	}
2361 
2362 	sector_size = get_unaligned_be32(&buffer[8]);
2363 	lba = get_unaligned_be64(&buffer[0]);
2364 
2365 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2366 		sdkp->capacity = 0;
2367 		return -ENODEV;
2368 	}
2369 
2370 	/* Logical blocks per physical block exponent */
2371 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2372 
2373 	/* RC basis */
2374 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2375 
2376 	/* Lowest aligned logical block */
2377 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2378 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2379 	if (alignment && sdkp->first_scan)
2380 		sd_printk(KERN_NOTICE, sdkp,
2381 			  "physical block alignment offset: %u\n", alignment);
2382 
2383 	if (buffer[14] & 0x80) { /* LBPME */
2384 		sdkp->lbpme = 1;
2385 
2386 		if (buffer[14] & 0x40) /* LBPRZ */
2387 			sdkp->lbprz = 1;
2388 
2389 		sd_config_discard(sdkp, SD_LBP_WS16);
2390 	}
2391 
2392 	sdkp->capacity = lba + 1;
2393 	return sector_size;
2394 }
2395 
2396 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2397 						unsigned char *buffer)
2398 {
2399 	unsigned char cmd[16];
2400 	struct scsi_sense_hdr sshdr;
2401 	const struct scsi_exec_args exec_args = {
2402 		.sshdr = &sshdr,
2403 	};
2404 	int sense_valid = 0;
2405 	int the_result;
2406 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2407 	sector_t lba;
2408 	unsigned sector_size;
2409 
2410 	do {
2411 		cmd[0] = READ_CAPACITY;
2412 		memset(&cmd[1], 0, 9);
2413 		memset(buffer, 0, 8);
2414 
2415 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2416 					      8, SD_TIMEOUT, sdkp->max_retries,
2417 					      &exec_args);
2418 
2419 		if (media_not_present(sdkp, &sshdr))
2420 			return -ENODEV;
2421 
2422 		if (the_result > 0) {
2423 			sense_valid = scsi_sense_valid(&sshdr);
2424 			if (sense_valid &&
2425 			    sshdr.sense_key == UNIT_ATTENTION &&
2426 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2427 				/* Device reset might occur several times,
2428 				 * give it one more chance */
2429 				if (--reset_retries > 0)
2430 					continue;
2431 		}
2432 		retries--;
2433 
2434 	} while (the_result && retries);
2435 
2436 	if (the_result) {
2437 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2438 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2439 		return -EINVAL;
2440 	}
2441 
2442 	sector_size = get_unaligned_be32(&buffer[4]);
2443 	lba = get_unaligned_be32(&buffer[0]);
2444 
2445 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2446 		/* Some buggy (usb cardreader) devices return an lba of
2447 		   0xffffffff when the want to report a size of 0 (with
2448 		   which they really mean no media is present) */
2449 		sdkp->capacity = 0;
2450 		sdkp->physical_block_size = sector_size;
2451 		return sector_size;
2452 	}
2453 
2454 	sdkp->capacity = lba + 1;
2455 	sdkp->physical_block_size = sector_size;
2456 	return sector_size;
2457 }
2458 
2459 static int sd_try_rc16_first(struct scsi_device *sdp)
2460 {
2461 	if (sdp->host->max_cmd_len < 16)
2462 		return 0;
2463 	if (sdp->try_rc_10_first)
2464 		return 0;
2465 	if (sdp->scsi_level > SCSI_SPC_2)
2466 		return 1;
2467 	if (scsi_device_protection(sdp))
2468 		return 1;
2469 	return 0;
2470 }
2471 
2472 /*
2473  * read disk capacity
2474  */
2475 static void
2476 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2477 {
2478 	int sector_size;
2479 	struct scsi_device *sdp = sdkp->device;
2480 
2481 	if (sd_try_rc16_first(sdp)) {
2482 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2483 		if (sector_size == -EOVERFLOW)
2484 			goto got_data;
2485 		if (sector_size == -ENODEV)
2486 			return;
2487 		if (sector_size < 0)
2488 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2489 		if (sector_size < 0)
2490 			return;
2491 	} else {
2492 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2493 		if (sector_size == -EOVERFLOW)
2494 			goto got_data;
2495 		if (sector_size < 0)
2496 			return;
2497 		if ((sizeof(sdkp->capacity) > 4) &&
2498 		    (sdkp->capacity > 0xffffffffULL)) {
2499 			int old_sector_size = sector_size;
2500 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2501 					"Trying to use READ CAPACITY(16).\n");
2502 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2503 			if (sector_size < 0) {
2504 				sd_printk(KERN_NOTICE, sdkp,
2505 					"Using 0xffffffff as device size\n");
2506 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2507 				sector_size = old_sector_size;
2508 				goto got_data;
2509 			}
2510 			/* Remember that READ CAPACITY(16) succeeded */
2511 			sdp->try_rc_10_first = 0;
2512 		}
2513 	}
2514 
2515 	/* Some devices are known to return the total number of blocks,
2516 	 * not the highest block number.  Some devices have versions
2517 	 * which do this and others which do not.  Some devices we might
2518 	 * suspect of doing this but we don't know for certain.
2519 	 *
2520 	 * If we know the reported capacity is wrong, decrement it.  If
2521 	 * we can only guess, then assume the number of blocks is even
2522 	 * (usually true but not always) and err on the side of lowering
2523 	 * the capacity.
2524 	 */
2525 	if (sdp->fix_capacity ||
2526 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2527 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2528 				"from its reported value: %llu\n",
2529 				(unsigned long long) sdkp->capacity);
2530 		--sdkp->capacity;
2531 	}
2532 
2533 got_data:
2534 	if (sector_size == 0) {
2535 		sector_size = 512;
2536 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2537 			  "assuming 512.\n");
2538 	}
2539 
2540 	if (sector_size != 512 &&
2541 	    sector_size != 1024 &&
2542 	    sector_size != 2048 &&
2543 	    sector_size != 4096) {
2544 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2545 			  sector_size);
2546 		/*
2547 		 * The user might want to re-format the drive with
2548 		 * a supported sectorsize.  Once this happens, it
2549 		 * would be relatively trivial to set the thing up.
2550 		 * For this reason, we leave the thing in the table.
2551 		 */
2552 		sdkp->capacity = 0;
2553 		/*
2554 		 * set a bogus sector size so the normal read/write
2555 		 * logic in the block layer will eventually refuse any
2556 		 * request on this device without tripping over power
2557 		 * of two sector size assumptions
2558 		 */
2559 		sector_size = 512;
2560 	}
2561 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2562 	blk_queue_physical_block_size(sdp->request_queue,
2563 				      sdkp->physical_block_size);
2564 	sdkp->device->sector_size = sector_size;
2565 
2566 	if (sdkp->capacity > 0xffffffff)
2567 		sdp->use_16_for_rw = 1;
2568 
2569 }
2570 
2571 /*
2572  * Print disk capacity
2573  */
2574 static void
2575 sd_print_capacity(struct scsi_disk *sdkp,
2576 		  sector_t old_capacity)
2577 {
2578 	int sector_size = sdkp->device->sector_size;
2579 	char cap_str_2[10], cap_str_10[10];
2580 
2581 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2582 		return;
2583 
2584 	string_get_size(sdkp->capacity, sector_size,
2585 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2586 	string_get_size(sdkp->capacity, sector_size,
2587 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2588 
2589 	sd_printk(KERN_NOTICE, sdkp,
2590 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2591 		  (unsigned long long)sdkp->capacity,
2592 		  sector_size, cap_str_10, cap_str_2);
2593 
2594 	if (sdkp->physical_block_size != sector_size)
2595 		sd_printk(KERN_NOTICE, sdkp,
2596 			  "%u-byte physical blocks\n",
2597 			  sdkp->physical_block_size);
2598 }
2599 
2600 /* called with buffer of length 512 */
2601 static inline int
2602 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2603 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2604 		 struct scsi_sense_hdr *sshdr)
2605 {
2606 	/*
2607 	 * If we must use MODE SENSE(10), make sure that the buffer length
2608 	 * is at least 8 bytes so that the mode sense header fits.
2609 	 */
2610 	if (sdkp->device->use_10_for_ms && len < 8)
2611 		len = 8;
2612 
2613 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2614 			       SD_TIMEOUT, sdkp->max_retries, data,
2615 			       sshdr);
2616 }
2617 
2618 /*
2619  * read write protect setting, if possible - called only in sd_revalidate_disk()
2620  * called with buffer of length SD_BUF_SIZE
2621  */
2622 static void
2623 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2624 {
2625 	int res;
2626 	struct scsi_device *sdp = sdkp->device;
2627 	struct scsi_mode_data data;
2628 	int old_wp = sdkp->write_prot;
2629 
2630 	set_disk_ro(sdkp->disk, 0);
2631 	if (sdp->skip_ms_page_3f) {
2632 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2633 		return;
2634 	}
2635 
2636 	if (sdp->use_192_bytes_for_3f) {
2637 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2638 	} else {
2639 		/*
2640 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2641 		 * We have to start carefully: some devices hang if we ask
2642 		 * for more than is available.
2643 		 */
2644 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2645 
2646 		/*
2647 		 * Second attempt: ask for page 0 When only page 0 is
2648 		 * implemented, a request for page 3F may return Sense Key
2649 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2650 		 * CDB.
2651 		 */
2652 		if (res < 0)
2653 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2654 
2655 		/*
2656 		 * Third attempt: ask 255 bytes, as we did earlier.
2657 		 */
2658 		if (res < 0)
2659 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2660 					       &data, NULL);
2661 	}
2662 
2663 	if (res < 0) {
2664 		sd_first_printk(KERN_WARNING, sdkp,
2665 			  "Test WP failed, assume Write Enabled\n");
2666 	} else {
2667 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2668 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2669 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2670 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2671 				  sdkp->write_prot ? "on" : "off");
2672 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2673 		}
2674 	}
2675 }
2676 
2677 /*
2678  * sd_read_cache_type - called only from sd_revalidate_disk()
2679  * called with buffer of length SD_BUF_SIZE
2680  */
2681 static void
2682 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2683 {
2684 	int len = 0, res;
2685 	struct scsi_device *sdp = sdkp->device;
2686 
2687 	int dbd;
2688 	int modepage;
2689 	int first_len;
2690 	struct scsi_mode_data data;
2691 	struct scsi_sense_hdr sshdr;
2692 	int old_wce = sdkp->WCE;
2693 	int old_rcd = sdkp->RCD;
2694 	int old_dpofua = sdkp->DPOFUA;
2695 
2696 
2697 	if (sdkp->cache_override)
2698 		return;
2699 
2700 	first_len = 4;
2701 	if (sdp->skip_ms_page_8) {
2702 		if (sdp->type == TYPE_RBC)
2703 			goto defaults;
2704 		else {
2705 			if (sdp->skip_ms_page_3f)
2706 				goto defaults;
2707 			modepage = 0x3F;
2708 			if (sdp->use_192_bytes_for_3f)
2709 				first_len = 192;
2710 			dbd = 0;
2711 		}
2712 	} else if (sdp->type == TYPE_RBC) {
2713 		modepage = 6;
2714 		dbd = 8;
2715 	} else {
2716 		modepage = 8;
2717 		dbd = 0;
2718 	}
2719 
2720 	/* cautiously ask */
2721 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2722 			&data, &sshdr);
2723 
2724 	if (res < 0)
2725 		goto bad_sense;
2726 
2727 	if (!data.header_length) {
2728 		modepage = 6;
2729 		first_len = 0;
2730 		sd_first_printk(KERN_ERR, sdkp,
2731 				"Missing header in MODE_SENSE response\n");
2732 	}
2733 
2734 	/* that went OK, now ask for the proper length */
2735 	len = data.length;
2736 
2737 	/*
2738 	 * We're only interested in the first three bytes, actually.
2739 	 * But the data cache page is defined for the first 20.
2740 	 */
2741 	if (len < 3)
2742 		goto bad_sense;
2743 	else if (len > SD_BUF_SIZE) {
2744 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2745 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2746 		len = SD_BUF_SIZE;
2747 	}
2748 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2749 		len = 192;
2750 
2751 	/* Get the data */
2752 	if (len > first_len)
2753 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2754 				&data, &sshdr);
2755 
2756 	if (!res) {
2757 		int offset = data.header_length + data.block_descriptor_length;
2758 
2759 		while (offset < len) {
2760 			u8 page_code = buffer[offset] & 0x3F;
2761 			u8 spf       = buffer[offset] & 0x40;
2762 
2763 			if (page_code == 8 || page_code == 6) {
2764 				/* We're interested only in the first 3 bytes.
2765 				 */
2766 				if (len - offset <= 2) {
2767 					sd_first_printk(KERN_ERR, sdkp,
2768 						"Incomplete mode parameter "
2769 							"data\n");
2770 					goto defaults;
2771 				} else {
2772 					modepage = page_code;
2773 					goto Page_found;
2774 				}
2775 			} else {
2776 				/* Go to the next page */
2777 				if (spf && len - offset > 3)
2778 					offset += 4 + (buffer[offset+2] << 8) +
2779 						buffer[offset+3];
2780 				else if (!spf && len - offset > 1)
2781 					offset += 2 + buffer[offset+1];
2782 				else {
2783 					sd_first_printk(KERN_ERR, sdkp,
2784 							"Incomplete mode "
2785 							"parameter data\n");
2786 					goto defaults;
2787 				}
2788 			}
2789 		}
2790 
2791 		sd_first_printk(KERN_WARNING, sdkp,
2792 				"No Caching mode page found\n");
2793 		goto defaults;
2794 
2795 	Page_found:
2796 		if (modepage == 8) {
2797 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2798 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2799 		} else {
2800 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2801 			sdkp->RCD = 0;
2802 		}
2803 
2804 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2805 		if (sdp->broken_fua) {
2806 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2807 			sdkp->DPOFUA = 0;
2808 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2809 			   !sdkp->device->use_16_for_rw) {
2810 			sd_first_printk(KERN_NOTICE, sdkp,
2811 				  "Uses READ/WRITE(6), disabling FUA\n");
2812 			sdkp->DPOFUA = 0;
2813 		}
2814 
2815 		/* No cache flush allowed for write protected devices */
2816 		if (sdkp->WCE && sdkp->write_prot)
2817 			sdkp->WCE = 0;
2818 
2819 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2820 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2821 			sd_printk(KERN_NOTICE, sdkp,
2822 				  "Write cache: %s, read cache: %s, %s\n",
2823 				  sdkp->WCE ? "enabled" : "disabled",
2824 				  sdkp->RCD ? "disabled" : "enabled",
2825 				  sdkp->DPOFUA ? "supports DPO and FUA"
2826 				  : "doesn't support DPO or FUA");
2827 
2828 		return;
2829 	}
2830 
2831 bad_sense:
2832 	if (scsi_sense_valid(&sshdr) &&
2833 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2834 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2835 		/* Invalid field in CDB */
2836 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2837 	else
2838 		sd_first_printk(KERN_ERR, sdkp,
2839 				"Asking for cache data failed\n");
2840 
2841 defaults:
2842 	if (sdp->wce_default_on) {
2843 		sd_first_printk(KERN_NOTICE, sdkp,
2844 				"Assuming drive cache: write back\n");
2845 		sdkp->WCE = 1;
2846 	} else {
2847 		sd_first_printk(KERN_WARNING, sdkp,
2848 				"Assuming drive cache: write through\n");
2849 		sdkp->WCE = 0;
2850 	}
2851 	sdkp->RCD = 0;
2852 	sdkp->DPOFUA = 0;
2853 }
2854 
2855 /*
2856  * The ATO bit indicates whether the DIF application tag is available
2857  * for use by the operating system.
2858  */
2859 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2860 {
2861 	int res, offset;
2862 	struct scsi_device *sdp = sdkp->device;
2863 	struct scsi_mode_data data;
2864 	struct scsi_sense_hdr sshdr;
2865 
2866 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2867 		return;
2868 
2869 	if (sdkp->protection_type == 0)
2870 		return;
2871 
2872 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2873 			      sdkp->max_retries, &data, &sshdr);
2874 
2875 	if (res < 0 || !data.header_length ||
2876 	    data.length < 6) {
2877 		sd_first_printk(KERN_WARNING, sdkp,
2878 			  "getting Control mode page failed, assume no ATO\n");
2879 
2880 		if (scsi_sense_valid(&sshdr))
2881 			sd_print_sense_hdr(sdkp, &sshdr);
2882 
2883 		return;
2884 	}
2885 
2886 	offset = data.header_length + data.block_descriptor_length;
2887 
2888 	if ((buffer[offset] & 0x3f) != 0x0a) {
2889 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2890 		return;
2891 	}
2892 
2893 	if ((buffer[offset + 5] & 0x80) == 0)
2894 		return;
2895 
2896 	sdkp->ATO = 1;
2897 
2898 	return;
2899 }
2900 
2901 /**
2902  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2903  * @sdkp: disk to query
2904  */
2905 static void sd_read_block_limits(struct scsi_disk *sdkp)
2906 {
2907 	struct scsi_vpd *vpd;
2908 
2909 	rcu_read_lock();
2910 
2911 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2912 	if (!vpd || vpd->len < 16)
2913 		goto out;
2914 
2915 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2916 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2917 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2918 
2919 	if (vpd->len >= 64) {
2920 		unsigned int lba_count, desc_count;
2921 
2922 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2923 
2924 		if (!sdkp->lbpme)
2925 			goto out;
2926 
2927 		lba_count = get_unaligned_be32(&vpd->data[20]);
2928 		desc_count = get_unaligned_be32(&vpd->data[24]);
2929 
2930 		if (lba_count && desc_count)
2931 			sdkp->max_unmap_blocks = lba_count;
2932 
2933 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2934 
2935 		if (vpd->data[32] & 0x80)
2936 			sdkp->unmap_alignment =
2937 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2938 
2939 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2940 
2941 			if (sdkp->max_unmap_blocks)
2942 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2943 			else
2944 				sd_config_discard(sdkp, SD_LBP_WS16);
2945 
2946 		} else {	/* LBP VPD page tells us what to use */
2947 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2948 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2949 			else if (sdkp->lbpws)
2950 				sd_config_discard(sdkp, SD_LBP_WS16);
2951 			else if (sdkp->lbpws10)
2952 				sd_config_discard(sdkp, SD_LBP_WS10);
2953 			else
2954 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2955 		}
2956 	}
2957 
2958  out:
2959 	rcu_read_unlock();
2960 }
2961 
2962 /**
2963  * sd_read_block_characteristics - Query block dev. characteristics
2964  * @sdkp: disk to query
2965  */
2966 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2967 {
2968 	struct request_queue *q = sdkp->disk->queue;
2969 	struct scsi_vpd *vpd;
2970 	u16 rot;
2971 	u8 zoned;
2972 
2973 	rcu_read_lock();
2974 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
2975 
2976 	if (!vpd || vpd->len < 8) {
2977 		rcu_read_unlock();
2978 	        return;
2979 	}
2980 
2981 	rot = get_unaligned_be16(&vpd->data[4]);
2982 	zoned = (vpd->data[8] >> 4) & 3;
2983 	rcu_read_unlock();
2984 
2985 	if (rot == 1) {
2986 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2987 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2988 	}
2989 
2990 	if (sdkp->device->type == TYPE_ZBC) {
2991 		/* Host-managed */
2992 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
2993 	} else {
2994 		sdkp->zoned = zoned;
2995 		if (sdkp->zoned == 1) {
2996 			/* Host-aware */
2997 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
2998 		} else {
2999 			/* Regular disk or drive managed disk */
3000 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3001 		}
3002 	}
3003 
3004 	if (!sdkp->first_scan)
3005 		return;
3006 
3007 	if (blk_queue_is_zoned(q)) {
3008 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3009 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3010 	} else {
3011 		if (sdkp->zoned == 1)
3012 			sd_printk(KERN_NOTICE, sdkp,
3013 				  "Host-aware SMR disk used as regular disk\n");
3014 		else if (sdkp->zoned == 2)
3015 			sd_printk(KERN_NOTICE, sdkp,
3016 				  "Drive-managed SMR disk\n");
3017 	}
3018 }
3019 
3020 /**
3021  * sd_read_block_provisioning - Query provisioning VPD page
3022  * @sdkp: disk to query
3023  */
3024 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3025 {
3026 	struct scsi_vpd *vpd;
3027 
3028 	if (sdkp->lbpme == 0)
3029 		return;
3030 
3031 	rcu_read_lock();
3032 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3033 
3034 	if (!vpd || vpd->len < 8) {
3035 		rcu_read_unlock();
3036 		return;
3037 	}
3038 
3039 	sdkp->lbpvpd	= 1;
3040 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3041 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3042 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3043 	rcu_read_unlock();
3044 }
3045 
3046 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3047 {
3048 	struct scsi_device *sdev = sdkp->device;
3049 
3050 	if (sdev->host->no_write_same) {
3051 		sdev->no_write_same = 1;
3052 
3053 		return;
3054 	}
3055 
3056 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
3057 		struct scsi_vpd *vpd;
3058 
3059 		sdev->no_report_opcodes = 1;
3060 
3061 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3062 		 * CODES is unsupported and the device has an ATA
3063 		 * Information VPD page (SAT).
3064 		 */
3065 		rcu_read_lock();
3066 		vpd = rcu_dereference(sdev->vpd_pg89);
3067 		if (vpd)
3068 			sdev->no_write_same = 1;
3069 		rcu_read_unlock();
3070 	}
3071 
3072 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3073 		sdkp->ws16 = 1;
3074 
3075 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3076 		sdkp->ws10 = 1;
3077 }
3078 
3079 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3080 {
3081 	struct scsi_device *sdev = sdkp->device;
3082 
3083 	if (!sdev->security_supported)
3084 		return;
3085 
3086 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3087 			SECURITY_PROTOCOL_IN) == 1 &&
3088 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3089 			SECURITY_PROTOCOL_OUT) == 1)
3090 		sdkp->security = 1;
3091 }
3092 
3093 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3094 {
3095 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3096 }
3097 
3098 /**
3099  * sd_read_cpr - Query concurrent positioning ranges
3100  * @sdkp:	disk to query
3101  */
3102 static void sd_read_cpr(struct scsi_disk *sdkp)
3103 {
3104 	struct blk_independent_access_ranges *iars = NULL;
3105 	unsigned char *buffer = NULL;
3106 	unsigned int nr_cpr = 0;
3107 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3108 	u8 *desc;
3109 
3110 	/*
3111 	 * We need to have the capacity set first for the block layer to be
3112 	 * able to check the ranges.
3113 	 */
3114 	if (sdkp->first_scan)
3115 		return;
3116 
3117 	if (!sdkp->capacity)
3118 		goto out;
3119 
3120 	/*
3121 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3122 	 * leading to a maximum page size of 64 + 256*32 bytes.
3123 	 */
3124 	buf_len = 64 + 256*32;
3125 	buffer = kmalloc(buf_len, GFP_KERNEL);
3126 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3127 		goto out;
3128 
3129 	/* We must have at least a 64B header and one 32B range descriptor */
3130 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3131 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3132 		sd_printk(KERN_ERR, sdkp,
3133 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3134 		goto out;
3135 	}
3136 
3137 	nr_cpr = (vpd_len - 64) / 32;
3138 	if (nr_cpr == 1) {
3139 		nr_cpr = 0;
3140 		goto out;
3141 	}
3142 
3143 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3144 	if (!iars) {
3145 		nr_cpr = 0;
3146 		goto out;
3147 	}
3148 
3149 	desc = &buffer[64];
3150 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3151 		if (desc[0] != i) {
3152 			sd_printk(KERN_ERR, sdkp,
3153 				"Invalid Concurrent Positioning Range number\n");
3154 			nr_cpr = 0;
3155 			break;
3156 		}
3157 
3158 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3159 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3160 	}
3161 
3162 out:
3163 	disk_set_independent_access_ranges(sdkp->disk, iars);
3164 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3165 		sd_printk(KERN_NOTICE, sdkp,
3166 			  "%u concurrent positioning ranges\n", nr_cpr);
3167 		sdkp->nr_actuators = nr_cpr;
3168 	}
3169 
3170 	kfree(buffer);
3171 }
3172 
3173 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3174 {
3175 	struct scsi_device *sdp = sdkp->device;
3176 	unsigned int min_xfer_bytes =
3177 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3178 
3179 	if (sdkp->min_xfer_blocks == 0)
3180 		return false;
3181 
3182 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3183 		sd_first_printk(KERN_WARNING, sdkp,
3184 				"Preferred minimum I/O size %u bytes not a " \
3185 				"multiple of physical block size (%u bytes)\n",
3186 				min_xfer_bytes, sdkp->physical_block_size);
3187 		sdkp->min_xfer_blocks = 0;
3188 		return false;
3189 	}
3190 
3191 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3192 			min_xfer_bytes);
3193 	return true;
3194 }
3195 
3196 /*
3197  * Determine the device's preferred I/O size for reads and writes
3198  * unless the reported value is unreasonably small, large, not a
3199  * multiple of the physical block size, or simply garbage.
3200  */
3201 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3202 				      unsigned int dev_max)
3203 {
3204 	struct scsi_device *sdp = sdkp->device;
3205 	unsigned int opt_xfer_bytes =
3206 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3207 	unsigned int min_xfer_bytes =
3208 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3209 
3210 	if (sdkp->opt_xfer_blocks == 0)
3211 		return false;
3212 
3213 	if (sdkp->opt_xfer_blocks > dev_max) {
3214 		sd_first_printk(KERN_WARNING, sdkp,
3215 				"Optimal transfer size %u logical blocks " \
3216 				"> dev_max (%u logical blocks)\n",
3217 				sdkp->opt_xfer_blocks, dev_max);
3218 		return false;
3219 	}
3220 
3221 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3222 		sd_first_printk(KERN_WARNING, sdkp,
3223 				"Optimal transfer size %u logical blocks " \
3224 				"> sd driver limit (%u logical blocks)\n",
3225 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3226 		return false;
3227 	}
3228 
3229 	if (opt_xfer_bytes < PAGE_SIZE) {
3230 		sd_first_printk(KERN_WARNING, sdkp,
3231 				"Optimal transfer size %u bytes < " \
3232 				"PAGE_SIZE (%u bytes)\n",
3233 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3234 		return false;
3235 	}
3236 
3237 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3238 		sd_first_printk(KERN_WARNING, sdkp,
3239 				"Optimal transfer size %u bytes not a " \
3240 				"multiple of preferred minimum block " \
3241 				"size (%u bytes)\n",
3242 				opt_xfer_bytes, min_xfer_bytes);
3243 		return false;
3244 	}
3245 
3246 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3247 		sd_first_printk(KERN_WARNING, sdkp,
3248 				"Optimal transfer size %u bytes not a " \
3249 				"multiple of physical block size (%u bytes)\n",
3250 				opt_xfer_bytes, sdkp->physical_block_size);
3251 		return false;
3252 	}
3253 
3254 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3255 			opt_xfer_bytes);
3256 	return true;
3257 }
3258 
3259 /**
3260  *	sd_revalidate_disk - called the first time a new disk is seen,
3261  *	performs disk spin up, read_capacity, etc.
3262  *	@disk: struct gendisk we care about
3263  **/
3264 static int sd_revalidate_disk(struct gendisk *disk)
3265 {
3266 	struct scsi_disk *sdkp = scsi_disk(disk);
3267 	struct scsi_device *sdp = sdkp->device;
3268 	struct request_queue *q = sdkp->disk->queue;
3269 	sector_t old_capacity = sdkp->capacity;
3270 	unsigned char *buffer;
3271 	unsigned int dev_max, rw_max;
3272 
3273 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3274 				      "sd_revalidate_disk\n"));
3275 
3276 	/*
3277 	 * If the device is offline, don't try and read capacity or any
3278 	 * of the other niceties.
3279 	 */
3280 	if (!scsi_device_online(sdp))
3281 		goto out;
3282 
3283 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3284 	if (!buffer) {
3285 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3286 			  "allocation failure.\n");
3287 		goto out;
3288 	}
3289 
3290 	sd_spinup_disk(sdkp);
3291 
3292 	/*
3293 	 * Without media there is no reason to ask; moreover, some devices
3294 	 * react badly if we do.
3295 	 */
3296 	if (sdkp->media_present) {
3297 		sd_read_capacity(sdkp, buffer);
3298 
3299 		/*
3300 		 * set the default to rotational.  All non-rotational devices
3301 		 * support the block characteristics VPD page, which will
3302 		 * cause this to be updated correctly and any device which
3303 		 * doesn't support it should be treated as rotational.
3304 		 */
3305 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3306 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3307 
3308 		if (scsi_device_supports_vpd(sdp)) {
3309 			sd_read_block_provisioning(sdkp);
3310 			sd_read_block_limits(sdkp);
3311 			sd_read_block_characteristics(sdkp);
3312 			sd_zbc_read_zones(sdkp, buffer);
3313 			sd_read_cpr(sdkp);
3314 		}
3315 
3316 		sd_print_capacity(sdkp, old_capacity);
3317 
3318 		sd_read_write_protect_flag(sdkp, buffer);
3319 		sd_read_cache_type(sdkp, buffer);
3320 		sd_read_app_tag_own(sdkp, buffer);
3321 		sd_read_write_same(sdkp, buffer);
3322 		sd_read_security(sdkp, buffer);
3323 		sd_config_protection(sdkp);
3324 	}
3325 
3326 	/*
3327 	 * We now have all cache related info, determine how we deal
3328 	 * with flush requests.
3329 	 */
3330 	sd_set_flush_flag(sdkp);
3331 
3332 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3333 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3334 
3335 	/* Some devices report a maximum block count for READ/WRITE requests. */
3336 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3337 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3338 
3339 	if (sd_validate_min_xfer_size(sdkp))
3340 		blk_queue_io_min(sdkp->disk->queue,
3341 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3342 	else
3343 		blk_queue_io_min(sdkp->disk->queue, 0);
3344 
3345 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3346 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3347 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3348 	} else {
3349 		q->limits.io_opt = 0;
3350 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3351 				      (sector_t)BLK_DEF_MAX_SECTORS);
3352 	}
3353 
3354 	/*
3355 	 * Limit default to SCSI host optimal sector limit if set. There may be
3356 	 * an impact on performance for when the size of a request exceeds this
3357 	 * host limit.
3358 	 */
3359 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3360 
3361 	/* Do not exceed controller limit */
3362 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3363 
3364 	/*
3365 	 * Only update max_sectors if previously unset or if the current value
3366 	 * exceeds the capabilities of the hardware.
3367 	 */
3368 	if (sdkp->first_scan ||
3369 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3370 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3371 		q->limits.max_sectors = rw_max;
3372 
3373 	sdkp->first_scan = 0;
3374 
3375 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3376 	sd_config_write_same(sdkp);
3377 	kfree(buffer);
3378 
3379 	/*
3380 	 * For a zoned drive, revalidating the zones can be done only once
3381 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3382 	 * capacity to 0.
3383 	 */
3384 	if (sd_zbc_revalidate_zones(sdkp))
3385 		set_capacity_and_notify(disk, 0);
3386 
3387  out:
3388 	return 0;
3389 }
3390 
3391 /**
3392  *	sd_unlock_native_capacity - unlock native capacity
3393  *	@disk: struct gendisk to set capacity for
3394  *
3395  *	Block layer calls this function if it detects that partitions
3396  *	on @disk reach beyond the end of the device.  If the SCSI host
3397  *	implements ->unlock_native_capacity() method, it's invoked to
3398  *	give it a chance to adjust the device capacity.
3399  *
3400  *	CONTEXT:
3401  *	Defined by block layer.  Might sleep.
3402  */
3403 static void sd_unlock_native_capacity(struct gendisk *disk)
3404 {
3405 	struct scsi_device *sdev = scsi_disk(disk)->device;
3406 
3407 	if (sdev->host->hostt->unlock_native_capacity)
3408 		sdev->host->hostt->unlock_native_capacity(sdev);
3409 }
3410 
3411 /**
3412  *	sd_format_disk_name - format disk name
3413  *	@prefix: name prefix - ie. "sd" for SCSI disks
3414  *	@index: index of the disk to format name for
3415  *	@buf: output buffer
3416  *	@buflen: length of the output buffer
3417  *
3418  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3419  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3420  *	which is followed by sdaaa.
3421  *
3422  *	This is basically 26 base counting with one extra 'nil' entry
3423  *	at the beginning from the second digit on and can be
3424  *	determined using similar method as 26 base conversion with the
3425  *	index shifted -1 after each digit is computed.
3426  *
3427  *	CONTEXT:
3428  *	Don't care.
3429  *
3430  *	RETURNS:
3431  *	0 on success, -errno on failure.
3432  */
3433 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3434 {
3435 	const int base = 'z' - 'a' + 1;
3436 	char *begin = buf + strlen(prefix);
3437 	char *end = buf + buflen;
3438 	char *p;
3439 	int unit;
3440 
3441 	p = end - 1;
3442 	*p = '\0';
3443 	unit = base;
3444 	do {
3445 		if (p == begin)
3446 			return -EINVAL;
3447 		*--p = 'a' + (index % unit);
3448 		index = (index / unit) - 1;
3449 	} while (index >= 0);
3450 
3451 	memmove(begin, p, end - p);
3452 	memcpy(buf, prefix, strlen(prefix));
3453 
3454 	return 0;
3455 }
3456 
3457 /**
3458  *	sd_probe - called during driver initialization and whenever a
3459  *	new scsi device is attached to the system. It is called once
3460  *	for each scsi device (not just disks) present.
3461  *	@dev: pointer to device object
3462  *
3463  *	Returns 0 if successful (or not interested in this scsi device
3464  *	(e.g. scanner)); 1 when there is an error.
3465  *
3466  *	Note: this function is invoked from the scsi mid-level.
3467  *	This function sets up the mapping between a given
3468  *	<host,channel,id,lun> (found in sdp) and new device name
3469  *	(e.g. /dev/sda). More precisely it is the block device major
3470  *	and minor number that is chosen here.
3471  *
3472  *	Assume sd_probe is not re-entrant (for time being)
3473  *	Also think about sd_probe() and sd_remove() running coincidentally.
3474  **/
3475 static int sd_probe(struct device *dev)
3476 {
3477 	struct scsi_device *sdp = to_scsi_device(dev);
3478 	struct scsi_disk *sdkp;
3479 	struct gendisk *gd;
3480 	int index;
3481 	int error;
3482 
3483 	scsi_autopm_get_device(sdp);
3484 	error = -ENODEV;
3485 	if (sdp->type != TYPE_DISK &&
3486 	    sdp->type != TYPE_ZBC &&
3487 	    sdp->type != TYPE_MOD &&
3488 	    sdp->type != TYPE_RBC)
3489 		goto out;
3490 
3491 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3492 		sdev_printk(KERN_WARNING, sdp,
3493 			    "Unsupported ZBC host-managed device.\n");
3494 		goto out;
3495 	}
3496 
3497 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3498 					"sd_probe\n"));
3499 
3500 	error = -ENOMEM;
3501 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3502 	if (!sdkp)
3503 		goto out;
3504 
3505 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3506 					 &sd_bio_compl_lkclass);
3507 	if (!gd)
3508 		goto out_free;
3509 
3510 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3511 	if (index < 0) {
3512 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3513 		goto out_put;
3514 	}
3515 
3516 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3517 	if (error) {
3518 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3519 		goto out_free_index;
3520 	}
3521 
3522 	sdkp->device = sdp;
3523 	sdkp->disk = gd;
3524 	sdkp->index = index;
3525 	sdkp->max_retries = SD_MAX_RETRIES;
3526 	atomic_set(&sdkp->openers, 0);
3527 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3528 
3529 	if (!sdp->request_queue->rq_timeout) {
3530 		if (sdp->type != TYPE_MOD)
3531 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3532 		else
3533 			blk_queue_rq_timeout(sdp->request_queue,
3534 					     SD_MOD_TIMEOUT);
3535 	}
3536 
3537 	device_initialize(&sdkp->disk_dev);
3538 	sdkp->disk_dev.parent = get_device(dev);
3539 	sdkp->disk_dev.class = &sd_disk_class;
3540 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3541 
3542 	error = device_add(&sdkp->disk_dev);
3543 	if (error) {
3544 		put_device(&sdkp->disk_dev);
3545 		goto out;
3546 	}
3547 
3548 	dev_set_drvdata(dev, sdkp);
3549 
3550 	gd->major = sd_major((index & 0xf0) >> 4);
3551 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3552 	gd->minors = SD_MINORS;
3553 
3554 	gd->fops = &sd_fops;
3555 	gd->private_data = sdkp;
3556 
3557 	/* defaults, until the device tells us otherwise */
3558 	sdp->sector_size = 512;
3559 	sdkp->capacity = 0;
3560 	sdkp->media_present = 1;
3561 	sdkp->write_prot = 0;
3562 	sdkp->cache_override = 0;
3563 	sdkp->WCE = 0;
3564 	sdkp->RCD = 0;
3565 	sdkp->ATO = 0;
3566 	sdkp->first_scan = 1;
3567 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3568 
3569 	sd_revalidate_disk(gd);
3570 
3571 	if (sdp->removable) {
3572 		gd->flags |= GENHD_FL_REMOVABLE;
3573 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3574 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3575 	}
3576 
3577 	blk_pm_runtime_init(sdp->request_queue, dev);
3578 	if (sdp->rpm_autosuspend) {
3579 		pm_runtime_set_autosuspend_delay(dev,
3580 			sdp->host->hostt->rpm_autosuspend_delay);
3581 	}
3582 
3583 	error = device_add_disk(dev, gd, NULL);
3584 	if (error) {
3585 		put_device(&sdkp->disk_dev);
3586 		put_disk(gd);
3587 		goto out;
3588 	}
3589 
3590 	if (sdkp->security) {
3591 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3592 		if (sdkp->opal_dev)
3593 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3594 	}
3595 
3596 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3597 		  sdp->removable ? "removable " : "");
3598 	scsi_autopm_put_device(sdp);
3599 
3600 	return 0;
3601 
3602  out_free_index:
3603 	ida_free(&sd_index_ida, index);
3604  out_put:
3605 	put_disk(gd);
3606  out_free:
3607 	kfree(sdkp);
3608  out:
3609 	scsi_autopm_put_device(sdp);
3610 	return error;
3611 }
3612 
3613 /**
3614  *	sd_remove - called whenever a scsi disk (previously recognized by
3615  *	sd_probe) is detached from the system. It is called (potentially
3616  *	multiple times) during sd module unload.
3617  *	@dev: pointer to device object
3618  *
3619  *	Note: this function is invoked from the scsi mid-level.
3620  *	This function potentially frees up a device name (e.g. /dev/sdc)
3621  *	that could be re-used by a subsequent sd_probe().
3622  *	This function is not called when the built-in sd driver is "exit-ed".
3623  **/
3624 static int sd_remove(struct device *dev)
3625 {
3626 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3627 
3628 	scsi_autopm_get_device(sdkp->device);
3629 
3630 	device_del(&sdkp->disk_dev);
3631 	del_gendisk(sdkp->disk);
3632 	sd_shutdown(dev);
3633 
3634 	put_disk(sdkp->disk);
3635 	return 0;
3636 }
3637 
3638 static void scsi_disk_release(struct device *dev)
3639 {
3640 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3641 
3642 	ida_free(&sd_index_ida, sdkp->index);
3643 	sd_zbc_free_zone_info(sdkp);
3644 	put_device(&sdkp->device->sdev_gendev);
3645 	free_opal_dev(sdkp->opal_dev);
3646 
3647 	kfree(sdkp);
3648 }
3649 
3650 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3651 {
3652 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3653 	struct scsi_sense_hdr sshdr;
3654 	const struct scsi_exec_args exec_args = {
3655 		.sshdr = &sshdr,
3656 		.req_flags = BLK_MQ_REQ_PM,
3657 	};
3658 	struct scsi_device *sdp = sdkp->device;
3659 	int res;
3660 
3661 	if (start)
3662 		cmd[4] |= 1;	/* START */
3663 
3664 	if (sdp->start_stop_pwr_cond)
3665 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3666 
3667 	if (!scsi_device_online(sdp))
3668 		return -ENODEV;
3669 
3670 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3671 			       sdkp->max_retries, &exec_args);
3672 	if (res) {
3673 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3674 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3675 			sd_print_sense_hdr(sdkp, &sshdr);
3676 			/* 0x3a is medium not present */
3677 			if (sshdr.asc == 0x3a)
3678 				res = 0;
3679 		}
3680 	}
3681 
3682 	/* SCSI error codes must not go to the generic layer */
3683 	if (res)
3684 		return -EIO;
3685 
3686 	return 0;
3687 }
3688 
3689 /*
3690  * Send a SYNCHRONIZE CACHE instruction down to the device through
3691  * the normal SCSI command structure.  Wait for the command to
3692  * complete.
3693  */
3694 static void sd_shutdown(struct device *dev)
3695 {
3696 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3697 
3698 	if (!sdkp)
3699 		return;         /* this can happen */
3700 
3701 	if (pm_runtime_suspended(dev))
3702 		return;
3703 
3704 	if (sdkp->WCE && sdkp->media_present) {
3705 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3706 		sd_sync_cache(sdkp, NULL);
3707 	}
3708 
3709 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3710 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3711 		sd_start_stop_device(sdkp, 0);
3712 	}
3713 }
3714 
3715 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3716 {
3717 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3718 	struct scsi_sense_hdr sshdr;
3719 	int ret = 0;
3720 
3721 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3722 		return 0;
3723 
3724 	if (sdkp->WCE && sdkp->media_present) {
3725 		if (!sdkp->device->silence_suspend)
3726 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3727 		ret = sd_sync_cache(sdkp, &sshdr);
3728 
3729 		if (ret) {
3730 			/* ignore OFFLINE device */
3731 			if (ret == -ENODEV)
3732 				return 0;
3733 
3734 			if (!scsi_sense_valid(&sshdr) ||
3735 			    sshdr.sense_key != ILLEGAL_REQUEST)
3736 				return ret;
3737 
3738 			/*
3739 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3740 			 * doesn't support sync. There's not much to do and
3741 			 * suspend shouldn't fail.
3742 			 */
3743 			ret = 0;
3744 		}
3745 	}
3746 
3747 	if (sdkp->device->manage_start_stop) {
3748 		if (!sdkp->device->silence_suspend)
3749 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3750 		/* an error is not worth aborting a system sleep */
3751 		ret = sd_start_stop_device(sdkp, 0);
3752 		if (ignore_stop_errors)
3753 			ret = 0;
3754 	}
3755 
3756 	return ret;
3757 }
3758 
3759 static int sd_suspend_system(struct device *dev)
3760 {
3761 	if (pm_runtime_suspended(dev))
3762 		return 0;
3763 
3764 	return sd_suspend_common(dev, true);
3765 }
3766 
3767 static int sd_suspend_runtime(struct device *dev)
3768 {
3769 	return sd_suspend_common(dev, false);
3770 }
3771 
3772 static int sd_resume(struct device *dev)
3773 {
3774 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3775 	int ret;
3776 
3777 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3778 		return 0;
3779 
3780 	if (!sdkp->device->manage_start_stop)
3781 		return 0;
3782 
3783 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3784 	ret = sd_start_stop_device(sdkp, 1);
3785 	if (!ret)
3786 		opal_unlock_from_suspend(sdkp->opal_dev);
3787 	return ret;
3788 }
3789 
3790 static int sd_resume_system(struct device *dev)
3791 {
3792 	if (pm_runtime_suspended(dev))
3793 		return 0;
3794 
3795 	return sd_resume(dev);
3796 }
3797 
3798 static int sd_resume_runtime(struct device *dev)
3799 {
3800 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3801 	struct scsi_device *sdp;
3802 
3803 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3804 		return 0;
3805 
3806 	sdp = sdkp->device;
3807 
3808 	if (sdp->ignore_media_change) {
3809 		/* clear the device's sense data */
3810 		static const u8 cmd[10] = { REQUEST_SENSE };
3811 		const struct scsi_exec_args exec_args = {
3812 			.req_flags = BLK_MQ_REQ_PM,
3813 		};
3814 
3815 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
3816 				     sdp->request_queue->rq_timeout, 1,
3817 				     &exec_args))
3818 			sd_printk(KERN_NOTICE, sdkp,
3819 				  "Failed to clear sense data\n");
3820 	}
3821 
3822 	return sd_resume(dev);
3823 }
3824 
3825 /**
3826  *	init_sd - entry point for this driver (both when built in or when
3827  *	a module).
3828  *
3829  *	Note: this function registers this driver with the scsi mid-level.
3830  **/
3831 static int __init init_sd(void)
3832 {
3833 	int majors = 0, i, err;
3834 
3835 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3836 
3837 	for (i = 0; i < SD_MAJORS; i++) {
3838 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3839 			continue;
3840 		majors++;
3841 	}
3842 
3843 	if (!majors)
3844 		return -ENODEV;
3845 
3846 	err = class_register(&sd_disk_class);
3847 	if (err)
3848 		goto err_out;
3849 
3850 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3851 	if (!sd_page_pool) {
3852 		printk(KERN_ERR "sd: can't init discard page pool\n");
3853 		err = -ENOMEM;
3854 		goto err_out_class;
3855 	}
3856 
3857 	err = scsi_register_driver(&sd_template.gendrv);
3858 	if (err)
3859 		goto err_out_driver;
3860 
3861 	return 0;
3862 
3863 err_out_driver:
3864 	mempool_destroy(sd_page_pool);
3865 err_out_class:
3866 	class_unregister(&sd_disk_class);
3867 err_out:
3868 	for (i = 0; i < SD_MAJORS; i++)
3869 		unregister_blkdev(sd_major(i), "sd");
3870 	return err;
3871 }
3872 
3873 /**
3874  *	exit_sd - exit point for this driver (when it is a module).
3875  *
3876  *	Note: this function unregisters this driver from the scsi mid-level.
3877  **/
3878 static void __exit exit_sd(void)
3879 {
3880 	int i;
3881 
3882 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3883 
3884 	scsi_unregister_driver(&sd_template.gendrv);
3885 	mempool_destroy(sd_page_pool);
3886 
3887 	class_unregister(&sd_disk_class);
3888 
3889 	for (i = 0; i < SD_MAJORS; i++)
3890 		unregister_blkdev(sd_major(i), "sd");
3891 }
3892 
3893 module_init(init_sd);
3894 module_exit(exit_sd);
3895 
3896 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3897 {
3898 	scsi_print_sense_hdr(sdkp->device,
3899 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3900 }
3901 
3902 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3903 {
3904 	const char *hb_string = scsi_hostbyte_string(result);
3905 
3906 	if (hb_string)
3907 		sd_printk(KERN_INFO, sdkp,
3908 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3909 			  hb_string ? hb_string : "invalid",
3910 			  "DRIVER_OK");
3911 	else
3912 		sd_printk(KERN_INFO, sdkp,
3913 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3914 			  msg, host_byte(result), "DRIVER_OK");
3915 }
3916