xref: /openbmc/linux/drivers/scsi/sd.c (revision 9322af3e6aeae04c7eda3e6a0c977e97a13cf6ed)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *      sd.c Copyright (C) 1992 Drew Eckhardt
4  *           Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5  *
6  *      Linux scsi disk driver
7  *              Initial versions: Drew Eckhardt
8  *              Subsequent revisions: Eric Youngdale
9  *	Modification history:
10  *       - Drew Eckhardt <drew@colorado.edu> original
11  *       - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12  *         outstanding request, and other enhancements.
13  *         Support loadable low-level scsi drivers.
14  *       - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15  *         eight major numbers.
16  *       - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17  *	 - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18  *	   sd_init and cleanups.
19  *	 - Alex Davis <letmein@erols.com> Fix problem where partition info
20  *	   not being read in sd_open. Fix problem where removable media
21  *	   could be ejected after sd_open.
22  *	 - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23  *	 - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24  *	   <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25  *	   Support 32k/1M disks.
26  *
27  *	Logging policy (needs CONFIG_SCSI_LOGGING defined):
28  *	 - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29  *	 - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30  *	 - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31  *	 - entering other commands: SCSI_LOG_HLQUEUE level 3
32  *	Note: when the logging level is set by the user, it must be greater
33  *	than the level indicated above to trigger output.
34  */
35 
36 #include <linux/module.h>
37 #include <linux/fs.h>
38 #include <linux/kernel.h>
39 #include <linux/mm.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
56 #include <linux/pr.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
60 
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 
71 #include "sd.h"
72 #include "scsi_priv.h"
73 #include "scsi_logging.h"
74 
75 MODULE_AUTHOR("Eric Youngdale");
76 MODULE_DESCRIPTION("SCSI disk (sd) driver");
77 MODULE_LICENSE("GPL");
78 
79 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
95 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
99 
100 #define SD_MINORS	16
101 
102 static void sd_config_discard(struct scsi_disk *, unsigned int);
103 static void sd_config_write_same(struct scsi_disk *);
104 static int  sd_revalidate_disk(struct gendisk *);
105 static void sd_unlock_native_capacity(struct gendisk *disk);
106 static int  sd_probe(struct device *);
107 static int  sd_remove(struct device *);
108 static void sd_shutdown(struct device *);
109 static int sd_suspend_system(struct device *);
110 static int sd_suspend_runtime(struct device *);
111 static int sd_resume_system(struct device *);
112 static int sd_resume_runtime(struct device *);
113 static void sd_rescan(struct device *);
114 static blk_status_t sd_init_command(struct scsi_cmnd *SCpnt);
115 static void sd_uninit_command(struct scsi_cmnd *SCpnt);
116 static int sd_done(struct scsi_cmnd *);
117 static void sd_eh_reset(struct scsi_cmnd *);
118 static int sd_eh_action(struct scsi_cmnd *, int);
119 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
120 static void scsi_disk_release(struct device *cdev);
121 
122 static DEFINE_IDA(sd_index_ida);
123 
124 static struct kmem_cache *sd_cdb_cache;
125 static mempool_t *sd_page_pool;
126 static struct lock_class_key sd_bio_compl_lkclass;
127 
128 static const char *sd_cache_types[] = {
129 	"write through", "none", "write back",
130 	"write back, no read (daft)"
131 };
132 
133 static void sd_set_flush_flag(struct scsi_disk *sdkp)
134 {
135 	bool wc = false, fua = false;
136 
137 	if (sdkp->WCE) {
138 		wc = true;
139 		if (sdkp->DPOFUA)
140 			fua = true;
141 	}
142 
143 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
144 }
145 
146 static ssize_t
147 cache_type_store(struct device *dev, struct device_attribute *attr,
148 		 const char *buf, size_t count)
149 {
150 	int ct, rcd, wce, sp;
151 	struct scsi_disk *sdkp = to_scsi_disk(dev);
152 	struct scsi_device *sdp = sdkp->device;
153 	char buffer[64];
154 	char *buffer_data;
155 	struct scsi_mode_data data;
156 	struct scsi_sense_hdr sshdr;
157 	static const char temp[] = "temporary ";
158 	int len;
159 
160 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
161 		/* no cache control on RBC devices; theoretically they
162 		 * can do it, but there's probably so many exceptions
163 		 * it's not worth the risk */
164 		return -EINVAL;
165 
166 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
167 		buf += sizeof(temp) - 1;
168 		sdkp->cache_override = 1;
169 	} else {
170 		sdkp->cache_override = 0;
171 	}
172 
173 	ct = sysfs_match_string(sd_cache_types, buf);
174 	if (ct < 0)
175 		return -EINVAL;
176 
177 	rcd = ct & 0x01 ? 1 : 0;
178 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
179 
180 	if (sdkp->cache_override) {
181 		sdkp->WCE = wce;
182 		sdkp->RCD = rcd;
183 		sd_set_flush_flag(sdkp);
184 		return count;
185 	}
186 
187 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
188 			    sdkp->max_retries, &data, NULL))
189 		return -EINVAL;
190 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
191 		  data.block_descriptor_length);
192 	buffer_data = buffer + data.header_length +
193 		data.block_descriptor_length;
194 	buffer_data[2] &= ~0x05;
195 	buffer_data[2] |= wce << 2 | rcd;
196 	sp = buffer_data[0] & 0x80 ? 1 : 0;
197 	buffer_data[0] &= ~0x80;
198 
199 	/*
200 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
201 	 * received mode parameter buffer before doing MODE SELECT.
202 	 */
203 	data.device_specific = 0;
204 
205 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
206 			     sdkp->max_retries, &data, &sshdr)) {
207 		if (scsi_sense_valid(&sshdr))
208 			sd_print_sense_hdr(sdkp, &sshdr);
209 		return -EINVAL;
210 	}
211 	sd_revalidate_disk(sdkp->disk);
212 	return count;
213 }
214 
215 static ssize_t
216 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
217 		       char *buf)
218 {
219 	struct scsi_disk *sdkp = to_scsi_disk(dev);
220 	struct scsi_device *sdp = sdkp->device;
221 
222 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
223 }
224 
225 static ssize_t
226 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
227 			const char *buf, size_t count)
228 {
229 	struct scsi_disk *sdkp = to_scsi_disk(dev);
230 	struct scsi_device *sdp = sdkp->device;
231 	bool v;
232 
233 	if (!capable(CAP_SYS_ADMIN))
234 		return -EACCES;
235 
236 	if (kstrtobool(buf, &v))
237 		return -EINVAL;
238 
239 	sdp->manage_start_stop = v;
240 
241 	return count;
242 }
243 static DEVICE_ATTR_RW(manage_start_stop);
244 
245 static ssize_t
246 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
247 {
248 	struct scsi_disk *sdkp = to_scsi_disk(dev);
249 
250 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
251 }
252 
253 static ssize_t
254 allow_restart_store(struct device *dev, struct device_attribute *attr,
255 		    const char *buf, size_t count)
256 {
257 	bool v;
258 	struct scsi_disk *sdkp = to_scsi_disk(dev);
259 	struct scsi_device *sdp = sdkp->device;
260 
261 	if (!capable(CAP_SYS_ADMIN))
262 		return -EACCES;
263 
264 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
265 		return -EINVAL;
266 
267 	if (kstrtobool(buf, &v))
268 		return -EINVAL;
269 
270 	sdp->allow_restart = v;
271 
272 	return count;
273 }
274 static DEVICE_ATTR_RW(allow_restart);
275 
276 static ssize_t
277 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
278 {
279 	struct scsi_disk *sdkp = to_scsi_disk(dev);
280 	int ct = sdkp->RCD + 2*sdkp->WCE;
281 
282 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
283 }
284 static DEVICE_ATTR_RW(cache_type);
285 
286 static ssize_t
287 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
288 {
289 	struct scsi_disk *sdkp = to_scsi_disk(dev);
290 
291 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
292 }
293 static DEVICE_ATTR_RO(FUA);
294 
295 static ssize_t
296 protection_type_show(struct device *dev, struct device_attribute *attr,
297 		     char *buf)
298 {
299 	struct scsi_disk *sdkp = to_scsi_disk(dev);
300 
301 	return sprintf(buf, "%u\n", sdkp->protection_type);
302 }
303 
304 static ssize_t
305 protection_type_store(struct device *dev, struct device_attribute *attr,
306 		      const char *buf, size_t count)
307 {
308 	struct scsi_disk *sdkp = to_scsi_disk(dev);
309 	unsigned int val;
310 	int err;
311 
312 	if (!capable(CAP_SYS_ADMIN))
313 		return -EACCES;
314 
315 	err = kstrtouint(buf, 10, &val);
316 
317 	if (err)
318 		return err;
319 
320 	if (val <= T10_PI_TYPE3_PROTECTION)
321 		sdkp->protection_type = val;
322 
323 	return count;
324 }
325 static DEVICE_ATTR_RW(protection_type);
326 
327 static ssize_t
328 protection_mode_show(struct device *dev, struct device_attribute *attr,
329 		     char *buf)
330 {
331 	struct scsi_disk *sdkp = to_scsi_disk(dev);
332 	struct scsi_device *sdp = sdkp->device;
333 	unsigned int dif, dix;
334 
335 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
336 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
337 
338 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
339 		dif = 0;
340 		dix = 1;
341 	}
342 
343 	if (!dif && !dix)
344 		return sprintf(buf, "none\n");
345 
346 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
347 }
348 static DEVICE_ATTR_RO(protection_mode);
349 
350 static ssize_t
351 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
352 {
353 	struct scsi_disk *sdkp = to_scsi_disk(dev);
354 
355 	return sprintf(buf, "%u\n", sdkp->ATO);
356 }
357 static DEVICE_ATTR_RO(app_tag_own);
358 
359 static ssize_t
360 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
361 		       char *buf)
362 {
363 	struct scsi_disk *sdkp = to_scsi_disk(dev);
364 
365 	return sprintf(buf, "%u\n", sdkp->lbpme);
366 }
367 static DEVICE_ATTR_RO(thin_provisioning);
368 
369 /* sysfs_match_string() requires dense arrays */
370 static const char *lbp_mode[] = {
371 	[SD_LBP_FULL]		= "full",
372 	[SD_LBP_UNMAP]		= "unmap",
373 	[SD_LBP_WS16]		= "writesame_16",
374 	[SD_LBP_WS10]		= "writesame_10",
375 	[SD_LBP_ZERO]		= "writesame_zero",
376 	[SD_LBP_DISABLE]	= "disabled",
377 };
378 
379 static ssize_t
380 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
381 		       char *buf)
382 {
383 	struct scsi_disk *sdkp = to_scsi_disk(dev);
384 
385 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
386 }
387 
388 static ssize_t
389 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
390 			const char *buf, size_t count)
391 {
392 	struct scsi_disk *sdkp = to_scsi_disk(dev);
393 	struct scsi_device *sdp = sdkp->device;
394 	int mode;
395 
396 	if (!capable(CAP_SYS_ADMIN))
397 		return -EACCES;
398 
399 	if (sd_is_zoned(sdkp)) {
400 		sd_config_discard(sdkp, SD_LBP_DISABLE);
401 		return count;
402 	}
403 
404 	if (sdp->type != TYPE_DISK)
405 		return -EINVAL;
406 
407 	mode = sysfs_match_string(lbp_mode, buf);
408 	if (mode < 0)
409 		return -EINVAL;
410 
411 	sd_config_discard(sdkp, mode);
412 
413 	return count;
414 }
415 static DEVICE_ATTR_RW(provisioning_mode);
416 
417 /* sysfs_match_string() requires dense arrays */
418 static const char *zeroing_mode[] = {
419 	[SD_ZERO_WRITE]		= "write",
420 	[SD_ZERO_WS]		= "writesame",
421 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
422 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
423 };
424 
425 static ssize_t
426 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
427 		  char *buf)
428 {
429 	struct scsi_disk *sdkp = to_scsi_disk(dev);
430 
431 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
432 }
433 
434 static ssize_t
435 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
436 		   const char *buf, size_t count)
437 {
438 	struct scsi_disk *sdkp = to_scsi_disk(dev);
439 	int mode;
440 
441 	if (!capable(CAP_SYS_ADMIN))
442 		return -EACCES;
443 
444 	mode = sysfs_match_string(zeroing_mode, buf);
445 	if (mode < 0)
446 		return -EINVAL;
447 
448 	sdkp->zeroing_mode = mode;
449 
450 	return count;
451 }
452 static DEVICE_ATTR_RW(zeroing_mode);
453 
454 static ssize_t
455 max_medium_access_timeouts_show(struct device *dev,
456 				struct device_attribute *attr, char *buf)
457 {
458 	struct scsi_disk *sdkp = to_scsi_disk(dev);
459 
460 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
461 }
462 
463 static ssize_t
464 max_medium_access_timeouts_store(struct device *dev,
465 				 struct device_attribute *attr, const char *buf,
466 				 size_t count)
467 {
468 	struct scsi_disk *sdkp = to_scsi_disk(dev);
469 	int err;
470 
471 	if (!capable(CAP_SYS_ADMIN))
472 		return -EACCES;
473 
474 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
475 
476 	return err ? err : count;
477 }
478 static DEVICE_ATTR_RW(max_medium_access_timeouts);
479 
480 static ssize_t
481 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
482 			   char *buf)
483 {
484 	struct scsi_disk *sdkp = to_scsi_disk(dev);
485 
486 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
487 }
488 
489 static ssize_t
490 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
491 			    const char *buf, size_t count)
492 {
493 	struct scsi_disk *sdkp = to_scsi_disk(dev);
494 	struct scsi_device *sdp = sdkp->device;
495 	unsigned long max;
496 	int err;
497 
498 	if (!capable(CAP_SYS_ADMIN))
499 		return -EACCES;
500 
501 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
502 		return -EINVAL;
503 
504 	err = kstrtoul(buf, 10, &max);
505 
506 	if (err)
507 		return err;
508 
509 	if (max == 0)
510 		sdp->no_write_same = 1;
511 	else if (max <= SD_MAX_WS16_BLOCKS) {
512 		sdp->no_write_same = 0;
513 		sdkp->max_ws_blocks = max;
514 	}
515 
516 	sd_config_write_same(sdkp);
517 
518 	return count;
519 }
520 static DEVICE_ATTR_RW(max_write_same_blocks);
521 
522 static ssize_t
523 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
524 {
525 	struct scsi_disk *sdkp = to_scsi_disk(dev);
526 
527 	if (sdkp->device->type == TYPE_ZBC)
528 		return sprintf(buf, "host-managed\n");
529 	if (sdkp->zoned == 1)
530 		return sprintf(buf, "host-aware\n");
531 	if (sdkp->zoned == 2)
532 		return sprintf(buf, "drive-managed\n");
533 	return sprintf(buf, "none\n");
534 }
535 static DEVICE_ATTR_RO(zoned_cap);
536 
537 static ssize_t
538 max_retries_store(struct device *dev, struct device_attribute *attr,
539 		  const char *buf, size_t count)
540 {
541 	struct scsi_disk *sdkp = to_scsi_disk(dev);
542 	struct scsi_device *sdev = sdkp->device;
543 	int retries, err;
544 
545 	err = kstrtoint(buf, 10, &retries);
546 	if (err)
547 		return err;
548 
549 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
550 		sdkp->max_retries = retries;
551 		return count;
552 	}
553 
554 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
555 		    SD_MAX_RETRIES);
556 	return -EINVAL;
557 }
558 
559 static ssize_t
560 max_retries_show(struct device *dev, struct device_attribute *attr,
561 		 char *buf)
562 {
563 	struct scsi_disk *sdkp = to_scsi_disk(dev);
564 
565 	return sprintf(buf, "%d\n", sdkp->max_retries);
566 }
567 
568 static DEVICE_ATTR_RW(max_retries);
569 
570 static struct attribute *sd_disk_attrs[] = {
571 	&dev_attr_cache_type.attr,
572 	&dev_attr_FUA.attr,
573 	&dev_attr_allow_restart.attr,
574 	&dev_attr_manage_start_stop.attr,
575 	&dev_attr_protection_type.attr,
576 	&dev_attr_protection_mode.attr,
577 	&dev_attr_app_tag_own.attr,
578 	&dev_attr_thin_provisioning.attr,
579 	&dev_attr_provisioning_mode.attr,
580 	&dev_attr_zeroing_mode.attr,
581 	&dev_attr_max_write_same_blocks.attr,
582 	&dev_attr_max_medium_access_timeouts.attr,
583 	&dev_attr_zoned_cap.attr,
584 	&dev_attr_max_retries.attr,
585 	NULL,
586 };
587 ATTRIBUTE_GROUPS(sd_disk);
588 
589 static struct class sd_disk_class = {
590 	.name		= "scsi_disk",
591 	.owner		= THIS_MODULE,
592 	.dev_release	= scsi_disk_release,
593 	.dev_groups	= sd_disk_groups,
594 };
595 
596 static const struct dev_pm_ops sd_pm_ops = {
597 	.suspend		= sd_suspend_system,
598 	.resume			= sd_resume_system,
599 	.poweroff		= sd_suspend_system,
600 	.restore		= sd_resume_system,
601 	.runtime_suspend	= sd_suspend_runtime,
602 	.runtime_resume		= sd_resume_runtime,
603 };
604 
605 static struct scsi_driver sd_template = {
606 	.gendrv = {
607 		.name		= "sd",
608 		.owner		= THIS_MODULE,
609 		.probe		= sd_probe,
610 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
611 		.remove		= sd_remove,
612 		.shutdown	= sd_shutdown,
613 		.pm		= &sd_pm_ops,
614 	},
615 	.rescan			= sd_rescan,
616 	.init_command		= sd_init_command,
617 	.uninit_command		= sd_uninit_command,
618 	.done			= sd_done,
619 	.eh_action		= sd_eh_action,
620 	.eh_reset		= sd_eh_reset,
621 };
622 
623 /*
624  * Don't request a new module, as that could deadlock in multipath
625  * environment.
626  */
627 static void sd_default_probe(dev_t devt)
628 {
629 }
630 
631 /*
632  * Device no to disk mapping:
633  *
634  *       major         disc2     disc  p1
635  *   |............|.............|....|....| <- dev_t
636  *    31        20 19          8 7  4 3  0
637  *
638  * Inside a major, we have 16k disks, however mapped non-
639  * contiguously. The first 16 disks are for major0, the next
640  * ones with major1, ... Disk 256 is for major0 again, disk 272
641  * for major1, ...
642  * As we stay compatible with our numbering scheme, we can reuse
643  * the well-know SCSI majors 8, 65--71, 136--143.
644  */
645 static int sd_major(int major_idx)
646 {
647 	switch (major_idx) {
648 	case 0:
649 		return SCSI_DISK0_MAJOR;
650 	case 1 ... 7:
651 		return SCSI_DISK1_MAJOR + major_idx - 1;
652 	case 8 ... 15:
653 		return SCSI_DISK8_MAJOR + major_idx - 8;
654 	default:
655 		BUG();
656 		return 0;	/* shut up gcc */
657 	}
658 }
659 
660 #ifdef CONFIG_BLK_SED_OPAL
661 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
662 		size_t len, bool send)
663 {
664 	struct scsi_disk *sdkp = data;
665 	struct scsi_device *sdev = sdkp->device;
666 	u8 cdb[12] = { 0, };
667 	int ret;
668 
669 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
670 	cdb[1] = secp;
671 	put_unaligned_be16(spsp, &cdb[2]);
672 	put_unaligned_be32(len, &cdb[6]);
673 
674 	ret = scsi_execute(sdev, cdb, send ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
675 		buffer, len, NULL, NULL, SD_TIMEOUT, sdkp->max_retries, 0,
676 		RQF_PM, NULL);
677 	return ret <= 0 ? ret : -EIO;
678 }
679 #endif /* CONFIG_BLK_SED_OPAL */
680 
681 /*
682  * Look up the DIX operation based on whether the command is read or
683  * write and whether dix and dif are enabled.
684  */
685 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
686 {
687 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
688 	static const unsigned int ops[] = {	/* wrt dix dif */
689 		SCSI_PROT_NORMAL,		/*  0	0   0  */
690 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
691 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
692 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
693 		SCSI_PROT_NORMAL,		/*  1	0   0  */
694 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
695 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
696 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
697 	};
698 
699 	return ops[write << 2 | dix << 1 | dif];
700 }
701 
702 /*
703  * Returns a mask of the protection flags that are valid for a given DIX
704  * operation.
705  */
706 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
707 {
708 	static const unsigned int flag_mask[] = {
709 		[SCSI_PROT_NORMAL]		= 0,
710 
711 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
712 						  SCSI_PROT_GUARD_CHECK |
713 						  SCSI_PROT_REF_CHECK |
714 						  SCSI_PROT_REF_INCREMENT,
715 
716 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
717 						  SCSI_PROT_IP_CHECKSUM,
718 
719 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
720 						  SCSI_PROT_GUARD_CHECK |
721 						  SCSI_PROT_REF_CHECK |
722 						  SCSI_PROT_REF_INCREMENT |
723 						  SCSI_PROT_IP_CHECKSUM,
724 
725 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
726 						  SCSI_PROT_REF_INCREMENT,
727 
728 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
729 						  SCSI_PROT_REF_CHECK |
730 						  SCSI_PROT_REF_INCREMENT |
731 						  SCSI_PROT_IP_CHECKSUM,
732 
733 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
734 						  SCSI_PROT_GUARD_CHECK |
735 						  SCSI_PROT_REF_CHECK |
736 						  SCSI_PROT_REF_INCREMENT |
737 						  SCSI_PROT_IP_CHECKSUM,
738 	};
739 
740 	return flag_mask[prot_op];
741 }
742 
743 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
744 					   unsigned int dix, unsigned int dif)
745 {
746 	struct request *rq = scsi_cmd_to_rq(scmd);
747 	struct bio *bio = rq->bio;
748 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
749 	unsigned int protect = 0;
750 
751 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
752 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
753 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
754 
755 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
756 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
757 	}
758 
759 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
760 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
761 
762 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
763 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
764 	}
765 
766 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
767 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
768 
769 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
770 			protect = 3 << 5;	/* Disable target PI checking */
771 		else
772 			protect = 1 << 5;	/* Enable target PI checking */
773 	}
774 
775 	scsi_set_prot_op(scmd, prot_op);
776 	scsi_set_prot_type(scmd, dif);
777 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
778 
779 	return protect;
780 }
781 
782 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
783 {
784 	struct request_queue *q = sdkp->disk->queue;
785 	unsigned int logical_block_size = sdkp->device->sector_size;
786 	unsigned int max_blocks = 0;
787 
788 	q->limits.discard_alignment =
789 		sdkp->unmap_alignment * logical_block_size;
790 	q->limits.discard_granularity =
791 		max(sdkp->physical_block_size,
792 		    sdkp->unmap_granularity * logical_block_size);
793 	sdkp->provisioning_mode = mode;
794 
795 	switch (mode) {
796 
797 	case SD_LBP_FULL:
798 	case SD_LBP_DISABLE:
799 		blk_queue_max_discard_sectors(q, 0);
800 		return;
801 
802 	case SD_LBP_UNMAP:
803 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
804 					  (u32)SD_MAX_WS16_BLOCKS);
805 		break;
806 
807 	case SD_LBP_WS16:
808 		if (sdkp->device->unmap_limit_for_ws)
809 			max_blocks = sdkp->max_unmap_blocks;
810 		else
811 			max_blocks = sdkp->max_ws_blocks;
812 
813 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
814 		break;
815 
816 	case SD_LBP_WS10:
817 		if (sdkp->device->unmap_limit_for_ws)
818 			max_blocks = sdkp->max_unmap_blocks;
819 		else
820 			max_blocks = sdkp->max_ws_blocks;
821 
822 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
823 		break;
824 
825 	case SD_LBP_ZERO:
826 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
827 					  (u32)SD_MAX_WS10_BLOCKS);
828 		break;
829 	}
830 
831 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
832 }
833 
834 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
835 {
836 	struct scsi_device *sdp = cmd->device;
837 	struct request *rq = scsi_cmd_to_rq(cmd);
838 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
839 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
840 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
841 	unsigned int data_len = 24;
842 	char *buf;
843 
844 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
845 	if (!rq->special_vec.bv_page)
846 		return BLK_STS_RESOURCE;
847 	clear_highpage(rq->special_vec.bv_page);
848 	rq->special_vec.bv_offset = 0;
849 	rq->special_vec.bv_len = data_len;
850 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
851 
852 	cmd->cmd_len = 10;
853 	cmd->cmnd[0] = UNMAP;
854 	cmd->cmnd[8] = 24;
855 
856 	buf = bvec_virt(&rq->special_vec);
857 	put_unaligned_be16(6 + 16, &buf[0]);
858 	put_unaligned_be16(16, &buf[2]);
859 	put_unaligned_be64(lba, &buf[8]);
860 	put_unaligned_be32(nr_blocks, &buf[16]);
861 
862 	cmd->allowed = sdkp->max_retries;
863 	cmd->transfersize = data_len;
864 	rq->timeout = SD_TIMEOUT;
865 
866 	return scsi_alloc_sgtables(cmd);
867 }
868 
869 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
870 		bool unmap)
871 {
872 	struct scsi_device *sdp = cmd->device;
873 	struct request *rq = scsi_cmd_to_rq(cmd);
874 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
875 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
876 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
877 	u32 data_len = sdp->sector_size;
878 
879 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
880 	if (!rq->special_vec.bv_page)
881 		return BLK_STS_RESOURCE;
882 	clear_highpage(rq->special_vec.bv_page);
883 	rq->special_vec.bv_offset = 0;
884 	rq->special_vec.bv_len = data_len;
885 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
886 
887 	cmd->cmd_len = 16;
888 	cmd->cmnd[0] = WRITE_SAME_16;
889 	if (unmap)
890 		cmd->cmnd[1] = 0x8; /* UNMAP */
891 	put_unaligned_be64(lba, &cmd->cmnd[2]);
892 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
893 
894 	cmd->allowed = sdkp->max_retries;
895 	cmd->transfersize = data_len;
896 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
897 
898 	return scsi_alloc_sgtables(cmd);
899 }
900 
901 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
902 		bool unmap)
903 {
904 	struct scsi_device *sdp = cmd->device;
905 	struct request *rq = scsi_cmd_to_rq(cmd);
906 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
907 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
908 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
909 	u32 data_len = sdp->sector_size;
910 
911 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
912 	if (!rq->special_vec.bv_page)
913 		return BLK_STS_RESOURCE;
914 	clear_highpage(rq->special_vec.bv_page);
915 	rq->special_vec.bv_offset = 0;
916 	rq->special_vec.bv_len = data_len;
917 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
918 
919 	cmd->cmd_len = 10;
920 	cmd->cmnd[0] = WRITE_SAME;
921 	if (unmap)
922 		cmd->cmnd[1] = 0x8; /* UNMAP */
923 	put_unaligned_be32(lba, &cmd->cmnd[2]);
924 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
925 
926 	cmd->allowed = sdkp->max_retries;
927 	cmd->transfersize = data_len;
928 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
929 
930 	return scsi_alloc_sgtables(cmd);
931 }
932 
933 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
934 {
935 	struct request *rq = scsi_cmd_to_rq(cmd);
936 	struct scsi_device *sdp = cmd->device;
937 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
938 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
939 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
940 
941 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
942 		switch (sdkp->zeroing_mode) {
943 		case SD_ZERO_WS16_UNMAP:
944 			return sd_setup_write_same16_cmnd(cmd, true);
945 		case SD_ZERO_WS10_UNMAP:
946 			return sd_setup_write_same10_cmnd(cmd, true);
947 		}
948 	}
949 
950 	if (sdp->no_write_same) {
951 		rq->rq_flags |= RQF_QUIET;
952 		return BLK_STS_TARGET;
953 	}
954 
955 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
956 		return sd_setup_write_same16_cmnd(cmd, false);
957 
958 	return sd_setup_write_same10_cmnd(cmd, false);
959 }
960 
961 static void sd_config_write_same(struct scsi_disk *sdkp)
962 {
963 	struct request_queue *q = sdkp->disk->queue;
964 	unsigned int logical_block_size = sdkp->device->sector_size;
965 
966 	if (sdkp->device->no_write_same) {
967 		sdkp->max_ws_blocks = 0;
968 		goto out;
969 	}
970 
971 	/* Some devices can not handle block counts above 0xffff despite
972 	 * supporting WRITE SAME(16). Consequently we default to 64k
973 	 * blocks per I/O unless the device explicitly advertises a
974 	 * bigger limit.
975 	 */
976 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
977 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
978 						   (u32)SD_MAX_WS16_BLOCKS);
979 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
980 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
981 						   (u32)SD_MAX_WS10_BLOCKS);
982 	else {
983 		sdkp->device->no_write_same = 1;
984 		sdkp->max_ws_blocks = 0;
985 	}
986 
987 	if (sdkp->lbprz && sdkp->lbpws)
988 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
989 	else if (sdkp->lbprz && sdkp->lbpws10)
990 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
991 	else if (sdkp->max_ws_blocks)
992 		sdkp->zeroing_mode = SD_ZERO_WS;
993 	else
994 		sdkp->zeroing_mode = SD_ZERO_WRITE;
995 
996 	if (sdkp->max_ws_blocks &&
997 	    sdkp->physical_block_size > logical_block_size) {
998 		/*
999 		 * Reporting a maximum number of blocks that is not aligned
1000 		 * on the device physical size would cause a large write same
1001 		 * request to be split into physically unaligned chunks by
1002 		 * __blkdev_issue_write_zeroes() even if the caller of this
1003 		 * functions took care to align the large request. So make sure
1004 		 * the maximum reported is aligned to the device physical block
1005 		 * size. This is only an optional optimization for regular
1006 		 * disks, but this is mandatory to avoid failure of large write
1007 		 * same requests directed at sequential write required zones of
1008 		 * host-managed ZBC disks.
1009 		 */
1010 		sdkp->max_ws_blocks =
1011 			round_down(sdkp->max_ws_blocks,
1012 				   bytes_to_logical(sdkp->device,
1013 						    sdkp->physical_block_size));
1014 	}
1015 
1016 out:
1017 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1018 					 (logical_block_size >> 9));
1019 }
1020 
1021 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1022 {
1023 	struct request *rq = scsi_cmd_to_rq(cmd);
1024 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1025 
1026 	/* flush requests don't perform I/O, zero the S/G table */
1027 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1028 
1029 	if (cmd->device->use_16_for_sync) {
1030 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1031 		cmd->cmd_len = 16;
1032 	} else {
1033 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1034 		cmd->cmd_len = 10;
1035 	}
1036 	cmd->transfersize = 0;
1037 	cmd->allowed = sdkp->max_retries;
1038 
1039 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1040 	return BLK_STS_OK;
1041 }
1042 
1043 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1044 				       sector_t lba, unsigned int nr_blocks,
1045 				       unsigned char flags)
1046 {
1047 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1048 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1049 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1050 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1051 	cmd->cmnd[10] = flags;
1052 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1053 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1054 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1055 
1056 	return BLK_STS_OK;
1057 }
1058 
1059 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1060 				       sector_t lba, unsigned int nr_blocks,
1061 				       unsigned char flags)
1062 {
1063 	cmd->cmd_len  = 16;
1064 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1065 	cmd->cmnd[1]  = flags;
1066 	cmd->cmnd[14] = 0;
1067 	cmd->cmnd[15] = 0;
1068 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1069 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1070 
1071 	return BLK_STS_OK;
1072 }
1073 
1074 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1075 				       sector_t lba, unsigned int nr_blocks,
1076 				       unsigned char flags)
1077 {
1078 	cmd->cmd_len = 10;
1079 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1080 	cmd->cmnd[1] = flags;
1081 	cmd->cmnd[6] = 0;
1082 	cmd->cmnd[9] = 0;
1083 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1084 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1085 
1086 	return BLK_STS_OK;
1087 }
1088 
1089 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1090 				      sector_t lba, unsigned int nr_blocks,
1091 				      unsigned char flags)
1092 {
1093 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1094 	if (WARN_ON_ONCE(nr_blocks == 0))
1095 		return BLK_STS_IOERR;
1096 
1097 	if (unlikely(flags & 0x8)) {
1098 		/*
1099 		 * This happens only if this drive failed 10byte rw
1100 		 * command with ILLEGAL_REQUEST during operation and
1101 		 * thus turned off use_10_for_rw.
1102 		 */
1103 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1104 		return BLK_STS_IOERR;
1105 	}
1106 
1107 	cmd->cmd_len = 6;
1108 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1109 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1110 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1111 	cmd->cmnd[3] = lba & 0xff;
1112 	cmd->cmnd[4] = nr_blocks;
1113 	cmd->cmnd[5] = 0;
1114 
1115 	return BLK_STS_OK;
1116 }
1117 
1118 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1119 {
1120 	struct request *rq = scsi_cmd_to_rq(cmd);
1121 	struct scsi_device *sdp = cmd->device;
1122 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1123 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1124 	sector_t threshold;
1125 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1126 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1127 	bool write = rq_data_dir(rq) == WRITE;
1128 	unsigned char protect, fua;
1129 	blk_status_t ret;
1130 	unsigned int dif;
1131 	bool dix;
1132 
1133 	ret = scsi_alloc_sgtables(cmd);
1134 	if (ret != BLK_STS_OK)
1135 		return ret;
1136 
1137 	ret = BLK_STS_IOERR;
1138 	if (!scsi_device_online(sdp) || sdp->changed) {
1139 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1140 		goto fail;
1141 	}
1142 
1143 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1144 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1145 		goto fail;
1146 	}
1147 
1148 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1149 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1150 		goto fail;
1151 	}
1152 
1153 	/*
1154 	 * Some SD card readers can't handle accesses which touch the
1155 	 * last one or two logical blocks. Split accesses as needed.
1156 	 */
1157 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1158 
1159 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1160 		if (lba < threshold) {
1161 			/* Access up to the threshold but not beyond */
1162 			nr_blocks = threshold - lba;
1163 		} else {
1164 			/* Access only a single logical block */
1165 			nr_blocks = 1;
1166 		}
1167 	}
1168 
1169 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1170 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1171 		if (ret)
1172 			goto fail;
1173 	}
1174 
1175 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1176 	dix = scsi_prot_sg_count(cmd);
1177 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1178 
1179 	if (dif || dix)
1180 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1181 	else
1182 		protect = 0;
1183 
1184 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1185 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1186 					 protect | fua);
1187 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1188 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1189 					 protect | fua);
1190 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1191 		   sdp->use_10_for_rw || protect) {
1192 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1193 					 protect | fua);
1194 	} else {
1195 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1196 					protect | fua);
1197 	}
1198 
1199 	if (unlikely(ret != BLK_STS_OK))
1200 		goto fail;
1201 
1202 	/*
1203 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1204 	 * host adapter, it's safe to assume that we can at least transfer
1205 	 * this many bytes between each connect / disconnect.
1206 	 */
1207 	cmd->transfersize = sdp->sector_size;
1208 	cmd->underflow = nr_blocks << 9;
1209 	cmd->allowed = sdkp->max_retries;
1210 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1211 
1212 	SCSI_LOG_HLQUEUE(1,
1213 			 scmd_printk(KERN_INFO, cmd,
1214 				     "%s: block=%llu, count=%d\n", __func__,
1215 				     (unsigned long long)blk_rq_pos(rq),
1216 				     blk_rq_sectors(rq)));
1217 	SCSI_LOG_HLQUEUE(2,
1218 			 scmd_printk(KERN_INFO, cmd,
1219 				     "%s %d/%u 512 byte blocks.\n",
1220 				     write ? "writing" : "reading", nr_blocks,
1221 				     blk_rq_sectors(rq)));
1222 
1223 	/*
1224 	 * This indicates that the command is ready from our end to be queued.
1225 	 */
1226 	return BLK_STS_OK;
1227 fail:
1228 	scsi_free_sgtables(cmd);
1229 	return ret;
1230 }
1231 
1232 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1233 {
1234 	struct request *rq = scsi_cmd_to_rq(cmd);
1235 
1236 	switch (req_op(rq)) {
1237 	case REQ_OP_DISCARD:
1238 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1239 		case SD_LBP_UNMAP:
1240 			return sd_setup_unmap_cmnd(cmd);
1241 		case SD_LBP_WS16:
1242 			return sd_setup_write_same16_cmnd(cmd, true);
1243 		case SD_LBP_WS10:
1244 			return sd_setup_write_same10_cmnd(cmd, true);
1245 		case SD_LBP_ZERO:
1246 			return sd_setup_write_same10_cmnd(cmd, false);
1247 		default:
1248 			return BLK_STS_TARGET;
1249 		}
1250 	case REQ_OP_WRITE_ZEROES:
1251 		return sd_setup_write_zeroes_cmnd(cmd);
1252 	case REQ_OP_FLUSH:
1253 		return sd_setup_flush_cmnd(cmd);
1254 	case REQ_OP_READ:
1255 	case REQ_OP_WRITE:
1256 	case REQ_OP_ZONE_APPEND:
1257 		return sd_setup_read_write_cmnd(cmd);
1258 	case REQ_OP_ZONE_RESET:
1259 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1260 						   false);
1261 	case REQ_OP_ZONE_RESET_ALL:
1262 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1263 						   true);
1264 	case REQ_OP_ZONE_OPEN:
1265 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1266 	case REQ_OP_ZONE_CLOSE:
1267 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1268 	case REQ_OP_ZONE_FINISH:
1269 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1270 	default:
1271 		WARN_ON_ONCE(1);
1272 		return BLK_STS_NOTSUPP;
1273 	}
1274 }
1275 
1276 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1277 {
1278 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1279 
1280 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1281 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1282 }
1283 
1284 static bool sd_need_revalidate(struct block_device *bdev,
1285 		struct scsi_disk *sdkp)
1286 {
1287 	if (sdkp->device->removable || sdkp->write_prot) {
1288 		if (bdev_check_media_change(bdev))
1289 			return true;
1290 	}
1291 
1292 	/*
1293 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1294 	 * nothing to do with partitions, BLKRRPART is used to force a full
1295 	 * revalidate after things like a format for historical reasons.
1296 	 */
1297 	return test_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1298 }
1299 
1300 /**
1301  *	sd_open - open a scsi disk device
1302  *	@bdev: Block device of the scsi disk to open
1303  *	@mode: FMODE_* mask
1304  *
1305  *	Returns 0 if successful. Returns a negated errno value in case
1306  *	of error.
1307  *
1308  *	Note: This can be called from a user context (e.g. fsck(1) )
1309  *	or from within the kernel (e.g. as a result of a mount(1) ).
1310  *	In the latter case @inode and @filp carry an abridged amount
1311  *	of information as noted above.
1312  *
1313  *	Locking: called with bdev->bd_disk->open_mutex held.
1314  **/
1315 static int sd_open(struct block_device *bdev, fmode_t mode)
1316 {
1317 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1318 	struct scsi_device *sdev = sdkp->device;
1319 	int retval;
1320 
1321 	if (scsi_device_get(sdev))
1322 		return -ENXIO;
1323 
1324 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1325 
1326 	/*
1327 	 * If the device is in error recovery, wait until it is done.
1328 	 * If the device is offline, then disallow any access to it.
1329 	 */
1330 	retval = -ENXIO;
1331 	if (!scsi_block_when_processing_errors(sdev))
1332 		goto error_out;
1333 
1334 	if (sd_need_revalidate(bdev, sdkp))
1335 		sd_revalidate_disk(bdev->bd_disk);
1336 
1337 	/*
1338 	 * If the drive is empty, just let the open fail.
1339 	 */
1340 	retval = -ENOMEDIUM;
1341 	if (sdev->removable && !sdkp->media_present && !(mode & FMODE_NDELAY))
1342 		goto error_out;
1343 
1344 	/*
1345 	 * If the device has the write protect tab set, have the open fail
1346 	 * if the user expects to be able to write to the thing.
1347 	 */
1348 	retval = -EROFS;
1349 	if (sdkp->write_prot && (mode & FMODE_WRITE))
1350 		goto error_out;
1351 
1352 	/*
1353 	 * It is possible that the disk changing stuff resulted in
1354 	 * the device being taken offline.  If this is the case,
1355 	 * report this to the user, and don't pretend that the
1356 	 * open actually succeeded.
1357 	 */
1358 	retval = -ENXIO;
1359 	if (!scsi_device_online(sdev))
1360 		goto error_out;
1361 
1362 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1363 		if (scsi_block_when_processing_errors(sdev))
1364 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1365 	}
1366 
1367 	return 0;
1368 
1369 error_out:
1370 	scsi_device_put(sdev);
1371 	return retval;
1372 }
1373 
1374 /**
1375  *	sd_release - invoked when the (last) close(2) is called on this
1376  *	scsi disk.
1377  *	@disk: disk to release
1378  *	@mode: FMODE_* mask
1379  *
1380  *	Returns 0.
1381  *
1382  *	Note: may block (uninterruptible) if error recovery is underway
1383  *	on this disk.
1384  *
1385  *	Locking: called with bdev->bd_disk->open_mutex held.
1386  **/
1387 static void sd_release(struct gendisk *disk, fmode_t mode)
1388 {
1389 	struct scsi_disk *sdkp = scsi_disk(disk);
1390 	struct scsi_device *sdev = sdkp->device;
1391 
1392 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1393 
1394 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1395 		if (scsi_block_when_processing_errors(sdev))
1396 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1397 	}
1398 
1399 	scsi_device_put(sdev);
1400 }
1401 
1402 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1403 {
1404 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1405 	struct scsi_device *sdp = sdkp->device;
1406 	struct Scsi_Host *host = sdp->host;
1407 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1408 	int diskinfo[4];
1409 
1410 	/* default to most commonly used values */
1411 	diskinfo[0] = 0x40;	/* 1 << 6 */
1412 	diskinfo[1] = 0x20;	/* 1 << 5 */
1413 	diskinfo[2] = capacity >> 11;
1414 
1415 	/* override with calculated, extended default, or driver values */
1416 	if (host->hostt->bios_param)
1417 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1418 	else
1419 		scsicam_bios_param(bdev, capacity, diskinfo);
1420 
1421 	geo->heads = diskinfo[0];
1422 	geo->sectors = diskinfo[1];
1423 	geo->cylinders = diskinfo[2];
1424 	return 0;
1425 }
1426 
1427 /**
1428  *	sd_ioctl - process an ioctl
1429  *	@bdev: target block device
1430  *	@mode: FMODE_* mask
1431  *	@cmd: ioctl command number
1432  *	@arg: this is third argument given to ioctl(2) system call.
1433  *	Often contains a pointer.
1434  *
1435  *	Returns 0 if successful (some ioctls return positive numbers on
1436  *	success as well). Returns a negated errno value in case of error.
1437  *
1438  *	Note: most ioctls are forward onto the block subsystem or further
1439  *	down in the scsi subsystem.
1440  **/
1441 static int sd_ioctl(struct block_device *bdev, fmode_t mode,
1442 		    unsigned int cmd, unsigned long arg)
1443 {
1444 	struct gendisk *disk = bdev->bd_disk;
1445 	struct scsi_disk *sdkp = scsi_disk(disk);
1446 	struct scsi_device *sdp = sdkp->device;
1447 	void __user *p = (void __user *)arg;
1448 	int error;
1449 
1450 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1451 				    "cmd=0x%x\n", disk->disk_name, cmd));
1452 
1453 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1454 		return -ENOIOCTLCMD;
1455 
1456 	/*
1457 	 * If we are in the middle of error recovery, don't let anyone
1458 	 * else try and use this device.  Also, if error recovery fails, it
1459 	 * may try and take the device offline, in which case all further
1460 	 * access to the device is prohibited.
1461 	 */
1462 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1463 			(mode & FMODE_NDELAY) != 0);
1464 	if (error)
1465 		return error;
1466 
1467 	if (is_sed_ioctl(cmd))
1468 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1469 	return scsi_ioctl(sdp, mode, cmd, p);
1470 }
1471 
1472 static void set_media_not_present(struct scsi_disk *sdkp)
1473 {
1474 	if (sdkp->media_present)
1475 		sdkp->device->changed = 1;
1476 
1477 	if (sdkp->device->removable) {
1478 		sdkp->media_present = 0;
1479 		sdkp->capacity = 0;
1480 	}
1481 }
1482 
1483 static int media_not_present(struct scsi_disk *sdkp,
1484 			     struct scsi_sense_hdr *sshdr)
1485 {
1486 	if (!scsi_sense_valid(sshdr))
1487 		return 0;
1488 
1489 	/* not invoked for commands that could return deferred errors */
1490 	switch (sshdr->sense_key) {
1491 	case UNIT_ATTENTION:
1492 	case NOT_READY:
1493 		/* medium not present */
1494 		if (sshdr->asc == 0x3A) {
1495 			set_media_not_present(sdkp);
1496 			return 1;
1497 		}
1498 	}
1499 	return 0;
1500 }
1501 
1502 /**
1503  *	sd_check_events - check media events
1504  *	@disk: kernel device descriptor
1505  *	@clearing: disk events currently being cleared
1506  *
1507  *	Returns mask of DISK_EVENT_*.
1508  *
1509  *	Note: this function is invoked from the block subsystem.
1510  **/
1511 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1512 {
1513 	struct scsi_disk *sdkp = disk->private_data;
1514 	struct scsi_device *sdp;
1515 	int retval;
1516 	bool disk_changed;
1517 
1518 	if (!sdkp)
1519 		return 0;
1520 
1521 	sdp = sdkp->device;
1522 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1523 
1524 	/*
1525 	 * If the device is offline, don't send any commands - just pretend as
1526 	 * if the command failed.  If the device ever comes back online, we
1527 	 * can deal with it then.  It is only because of unrecoverable errors
1528 	 * that we would ever take a device offline in the first place.
1529 	 */
1530 	if (!scsi_device_online(sdp)) {
1531 		set_media_not_present(sdkp);
1532 		goto out;
1533 	}
1534 
1535 	/*
1536 	 * Using TEST_UNIT_READY enables differentiation between drive with
1537 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1538 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1539 	 *
1540 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1541 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1542 	 * sd_revalidate() is called.
1543 	 */
1544 	if (scsi_block_when_processing_errors(sdp)) {
1545 		struct scsi_sense_hdr sshdr = { 0, };
1546 
1547 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1548 					      &sshdr);
1549 
1550 		/* failed to execute TUR, assume media not present */
1551 		if (retval < 0 || host_byte(retval)) {
1552 			set_media_not_present(sdkp);
1553 			goto out;
1554 		}
1555 
1556 		if (media_not_present(sdkp, &sshdr))
1557 			goto out;
1558 	}
1559 
1560 	/*
1561 	 * For removable scsi disk we have to recognise the presence
1562 	 * of a disk in the drive.
1563 	 */
1564 	if (!sdkp->media_present)
1565 		sdp->changed = 1;
1566 	sdkp->media_present = 1;
1567 out:
1568 	/*
1569 	 * sdp->changed is set under the following conditions:
1570 	 *
1571 	 *	Medium present state has changed in either direction.
1572 	 *	Device has indicated UNIT_ATTENTION.
1573 	 */
1574 	disk_changed = sdp->changed;
1575 	sdp->changed = 0;
1576 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1577 }
1578 
1579 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1580 {
1581 	int retries, res;
1582 	struct scsi_device *sdp = sdkp->device;
1583 	const int timeout = sdp->request_queue->rq_timeout
1584 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1585 	struct scsi_sense_hdr my_sshdr;
1586 
1587 	if (!scsi_device_online(sdp))
1588 		return -ENODEV;
1589 
1590 	/* caller might not be interested in sense, but we need it */
1591 	if (!sshdr)
1592 		sshdr = &my_sshdr;
1593 
1594 	for (retries = 3; retries > 0; --retries) {
1595 		unsigned char cmd[16] = { 0 };
1596 
1597 		if (sdp->use_16_for_sync)
1598 			cmd[0] = SYNCHRONIZE_CACHE_16;
1599 		else
1600 			cmd[0] = SYNCHRONIZE_CACHE;
1601 		/*
1602 		 * Leave the rest of the command zero to indicate
1603 		 * flush everything.
1604 		 */
1605 		res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, sshdr,
1606 				timeout, sdkp->max_retries, 0, RQF_PM, NULL);
1607 		if (res == 0)
1608 			break;
1609 	}
1610 
1611 	if (res) {
1612 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1613 
1614 		if (res < 0)
1615 			return res;
1616 
1617 		if (scsi_status_is_check_condition(res) &&
1618 		    scsi_sense_valid(sshdr)) {
1619 			sd_print_sense_hdr(sdkp, sshdr);
1620 
1621 			/* we need to evaluate the error return  */
1622 			if (sshdr->asc == 0x3a ||	/* medium not present */
1623 			    sshdr->asc == 0x20 ||	/* invalid command */
1624 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1625 				/* this is no error here */
1626 				return 0;
1627 		}
1628 
1629 		switch (host_byte(res)) {
1630 		/* ignore errors due to racing a disconnection */
1631 		case DID_BAD_TARGET:
1632 		case DID_NO_CONNECT:
1633 			return 0;
1634 		/* signal the upper layer it might try again */
1635 		case DID_BUS_BUSY:
1636 		case DID_IMM_RETRY:
1637 		case DID_REQUEUE:
1638 		case DID_SOFT_ERROR:
1639 			return -EBUSY;
1640 		default:
1641 			return -EIO;
1642 		}
1643 	}
1644 	return 0;
1645 }
1646 
1647 static void sd_rescan(struct device *dev)
1648 {
1649 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1650 
1651 	sd_revalidate_disk(sdkp->disk);
1652 }
1653 
1654 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1655 		enum blk_unique_id type)
1656 {
1657 	struct scsi_device *sdev = scsi_disk(disk)->device;
1658 	const struct scsi_vpd *vpd;
1659 	const unsigned char *d;
1660 	int ret = -ENXIO, len;
1661 
1662 	rcu_read_lock();
1663 	vpd = rcu_dereference(sdev->vpd_pg83);
1664 	if (!vpd)
1665 		goto out_unlock;
1666 
1667 	ret = -EINVAL;
1668 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1669 		/* we only care about designators with LU association */
1670 		if (((d[1] >> 4) & 0x3) != 0x00)
1671 			continue;
1672 		if ((d[1] & 0xf) != type)
1673 			continue;
1674 
1675 		/*
1676 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1677 		 * keep looking as one with more entropy might still show up.
1678 		 */
1679 		len = d[3];
1680 		if (len != 8 && len != 12 && len != 16)
1681 			continue;
1682 		ret = len;
1683 		memcpy(id, d + 4, len);
1684 		if (len == 16)
1685 			break;
1686 	}
1687 out_unlock:
1688 	rcu_read_unlock();
1689 	return ret;
1690 }
1691 
1692 static char sd_pr_type(enum pr_type type)
1693 {
1694 	switch (type) {
1695 	case PR_WRITE_EXCLUSIVE:
1696 		return 0x01;
1697 	case PR_EXCLUSIVE_ACCESS:
1698 		return 0x03;
1699 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1700 		return 0x05;
1701 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1702 		return 0x06;
1703 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1704 		return 0x07;
1705 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1706 		return 0x08;
1707 	default:
1708 		return 0;
1709 	}
1710 };
1711 
1712 static int sd_pr_command(struct block_device *bdev, u8 sa,
1713 		u64 key, u64 sa_key, u8 type, u8 flags)
1714 {
1715 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1716 	struct scsi_device *sdev = sdkp->device;
1717 	struct scsi_sense_hdr sshdr;
1718 	int result;
1719 	u8 cmd[16] = { 0, };
1720 	u8 data[24] = { 0, };
1721 
1722 	cmd[0] = PERSISTENT_RESERVE_OUT;
1723 	cmd[1] = sa;
1724 	cmd[2] = type;
1725 	put_unaligned_be32(sizeof(data), &cmd[5]);
1726 
1727 	put_unaligned_be64(key, &data[0]);
1728 	put_unaligned_be64(sa_key, &data[8]);
1729 	data[20] = flags;
1730 
1731 	result = scsi_execute_req(sdev, cmd, DMA_TO_DEVICE, &data, sizeof(data),
1732 			&sshdr, SD_TIMEOUT, sdkp->max_retries, NULL);
1733 
1734 	if (scsi_status_is_check_condition(result) &&
1735 	    scsi_sense_valid(&sshdr)) {
1736 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1737 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1738 	}
1739 
1740 	return result;
1741 }
1742 
1743 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1744 		u32 flags)
1745 {
1746 	if (flags & ~PR_FL_IGNORE_KEY)
1747 		return -EOPNOTSUPP;
1748 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1749 			old_key, new_key, 0,
1750 			(1 << 0) /* APTPL */);
1751 }
1752 
1753 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1754 		u32 flags)
1755 {
1756 	if (flags)
1757 		return -EOPNOTSUPP;
1758 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1759 }
1760 
1761 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1762 {
1763 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1764 }
1765 
1766 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1767 		enum pr_type type, bool abort)
1768 {
1769 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1770 			     sd_pr_type(type), 0);
1771 }
1772 
1773 static int sd_pr_clear(struct block_device *bdev, u64 key)
1774 {
1775 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1776 }
1777 
1778 static const struct pr_ops sd_pr_ops = {
1779 	.pr_register	= sd_pr_register,
1780 	.pr_reserve	= sd_pr_reserve,
1781 	.pr_release	= sd_pr_release,
1782 	.pr_preempt	= sd_pr_preempt,
1783 	.pr_clear	= sd_pr_clear,
1784 };
1785 
1786 static void scsi_disk_free_disk(struct gendisk *disk)
1787 {
1788 	struct scsi_disk *sdkp = scsi_disk(disk);
1789 
1790 	put_device(&sdkp->disk_dev);
1791 }
1792 
1793 static const struct block_device_operations sd_fops = {
1794 	.owner			= THIS_MODULE,
1795 	.open			= sd_open,
1796 	.release		= sd_release,
1797 	.ioctl			= sd_ioctl,
1798 	.getgeo			= sd_getgeo,
1799 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1800 	.check_events		= sd_check_events,
1801 	.unlock_native_capacity	= sd_unlock_native_capacity,
1802 	.report_zones		= sd_zbc_report_zones,
1803 	.get_unique_id		= sd_get_unique_id,
1804 	.free_disk		= scsi_disk_free_disk,
1805 	.pr_ops			= &sd_pr_ops,
1806 };
1807 
1808 /**
1809  *	sd_eh_reset - reset error handling callback
1810  *	@scmd:		sd-issued command that has failed
1811  *
1812  *	This function is called by the SCSI midlayer before starting
1813  *	SCSI EH. When counting medium access failures we have to be
1814  *	careful to register it only only once per device and SCSI EH run;
1815  *	there might be several timed out commands which will cause the
1816  *	'max_medium_access_timeouts' counter to trigger after the first
1817  *	SCSI EH run already and set the device to offline.
1818  *	So this function resets the internal counter before starting SCSI EH.
1819  **/
1820 static void sd_eh_reset(struct scsi_cmnd *scmd)
1821 {
1822 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1823 
1824 	/* New SCSI EH run, reset gate variable */
1825 	sdkp->ignore_medium_access_errors = false;
1826 }
1827 
1828 /**
1829  *	sd_eh_action - error handling callback
1830  *	@scmd:		sd-issued command that has failed
1831  *	@eh_disp:	The recovery disposition suggested by the midlayer
1832  *
1833  *	This function is called by the SCSI midlayer upon completion of an
1834  *	error test command (currently TEST UNIT READY). The result of sending
1835  *	the eh command is passed in eh_disp.  We're looking for devices that
1836  *	fail medium access commands but are OK with non access commands like
1837  *	test unit ready (so wrongly see the device as having a successful
1838  *	recovery)
1839  **/
1840 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1841 {
1842 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1843 	struct scsi_device *sdev = scmd->device;
1844 
1845 	if (!scsi_device_online(sdev) ||
1846 	    !scsi_medium_access_command(scmd) ||
1847 	    host_byte(scmd->result) != DID_TIME_OUT ||
1848 	    eh_disp != SUCCESS)
1849 		return eh_disp;
1850 
1851 	/*
1852 	 * The device has timed out executing a medium access command.
1853 	 * However, the TEST UNIT READY command sent during error
1854 	 * handling completed successfully. Either the device is in the
1855 	 * process of recovering or has it suffered an internal failure
1856 	 * that prevents access to the storage medium.
1857 	 */
1858 	if (!sdkp->ignore_medium_access_errors) {
1859 		sdkp->medium_access_timed_out++;
1860 		sdkp->ignore_medium_access_errors = true;
1861 	}
1862 
1863 	/*
1864 	 * If the device keeps failing read/write commands but TEST UNIT
1865 	 * READY always completes successfully we assume that medium
1866 	 * access is no longer possible and take the device offline.
1867 	 */
1868 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1869 		scmd_printk(KERN_ERR, scmd,
1870 			    "Medium access timeout failure. Offlining disk!\n");
1871 		mutex_lock(&sdev->state_mutex);
1872 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1873 		mutex_unlock(&sdev->state_mutex);
1874 
1875 		return SUCCESS;
1876 	}
1877 
1878 	return eh_disp;
1879 }
1880 
1881 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1882 {
1883 	struct request *req = scsi_cmd_to_rq(scmd);
1884 	struct scsi_device *sdev = scmd->device;
1885 	unsigned int transferred, good_bytes;
1886 	u64 start_lba, end_lba, bad_lba;
1887 
1888 	/*
1889 	 * Some commands have a payload smaller than the device logical
1890 	 * block size (e.g. INQUIRY on a 4K disk).
1891 	 */
1892 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1893 		return 0;
1894 
1895 	/* Check if we have a 'bad_lba' information */
1896 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1897 				     SCSI_SENSE_BUFFERSIZE,
1898 				     &bad_lba))
1899 		return 0;
1900 
1901 	/*
1902 	 * If the bad lba was reported incorrectly, we have no idea where
1903 	 * the error is.
1904 	 */
1905 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1906 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1907 	if (bad_lba < start_lba || bad_lba >= end_lba)
1908 		return 0;
1909 
1910 	/*
1911 	 * resid is optional but mostly filled in.  When it's unused,
1912 	 * its value is zero, so we assume the whole buffer transferred
1913 	 */
1914 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1915 
1916 	/* This computation should always be done in terms of the
1917 	 * resolution of the device's medium.
1918 	 */
1919 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1920 
1921 	return min(good_bytes, transferred);
1922 }
1923 
1924 /**
1925  *	sd_done - bottom half handler: called when the lower level
1926  *	driver has completed (successfully or otherwise) a scsi command.
1927  *	@SCpnt: mid-level's per command structure.
1928  *
1929  *	Note: potentially run from within an ISR. Must not block.
1930  **/
1931 static int sd_done(struct scsi_cmnd *SCpnt)
1932 {
1933 	int result = SCpnt->result;
1934 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1935 	unsigned int sector_size = SCpnt->device->sector_size;
1936 	unsigned int resid;
1937 	struct scsi_sense_hdr sshdr;
1938 	struct request *req = scsi_cmd_to_rq(SCpnt);
1939 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1940 	int sense_valid = 0;
1941 	int sense_deferred = 0;
1942 
1943 	switch (req_op(req)) {
1944 	case REQ_OP_DISCARD:
1945 	case REQ_OP_WRITE_ZEROES:
1946 	case REQ_OP_ZONE_RESET:
1947 	case REQ_OP_ZONE_RESET_ALL:
1948 	case REQ_OP_ZONE_OPEN:
1949 	case REQ_OP_ZONE_CLOSE:
1950 	case REQ_OP_ZONE_FINISH:
1951 		if (!result) {
1952 			good_bytes = blk_rq_bytes(req);
1953 			scsi_set_resid(SCpnt, 0);
1954 		} else {
1955 			good_bytes = 0;
1956 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1957 		}
1958 		break;
1959 	default:
1960 		/*
1961 		 * In case of bogus fw or device, we could end up having
1962 		 * an unaligned partial completion. Check this here and force
1963 		 * alignment.
1964 		 */
1965 		resid = scsi_get_resid(SCpnt);
1966 		if (resid & (sector_size - 1)) {
1967 			sd_printk(KERN_INFO, sdkp,
1968 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
1969 				resid, sector_size);
1970 			scsi_print_command(SCpnt);
1971 			resid = min(scsi_bufflen(SCpnt),
1972 				    round_up(resid, sector_size));
1973 			scsi_set_resid(SCpnt, resid);
1974 		}
1975 	}
1976 
1977 	if (result) {
1978 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
1979 		if (sense_valid)
1980 			sense_deferred = scsi_sense_is_deferred(&sshdr);
1981 	}
1982 	sdkp->medium_access_timed_out = 0;
1983 
1984 	if (!scsi_status_is_check_condition(result) &&
1985 	    (!sense_valid || sense_deferred))
1986 		goto out;
1987 
1988 	switch (sshdr.sense_key) {
1989 	case HARDWARE_ERROR:
1990 	case MEDIUM_ERROR:
1991 		good_bytes = sd_completed_bytes(SCpnt);
1992 		break;
1993 	case RECOVERED_ERROR:
1994 		good_bytes = scsi_bufflen(SCpnt);
1995 		break;
1996 	case NO_SENSE:
1997 		/* This indicates a false check condition, so ignore it.  An
1998 		 * unknown amount of data was transferred so treat it as an
1999 		 * error.
2000 		 */
2001 		SCpnt->result = 0;
2002 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2003 		break;
2004 	case ABORTED_COMMAND:
2005 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2006 			good_bytes = sd_completed_bytes(SCpnt);
2007 		break;
2008 	case ILLEGAL_REQUEST:
2009 		switch (sshdr.asc) {
2010 		case 0x10:	/* DIX: Host detected corruption */
2011 			good_bytes = sd_completed_bytes(SCpnt);
2012 			break;
2013 		case 0x20:	/* INVALID COMMAND OPCODE */
2014 		case 0x24:	/* INVALID FIELD IN CDB */
2015 			switch (SCpnt->cmnd[0]) {
2016 			case UNMAP:
2017 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2018 				break;
2019 			case WRITE_SAME_16:
2020 			case WRITE_SAME:
2021 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2022 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2023 				} else {
2024 					sdkp->device->no_write_same = 1;
2025 					sd_config_write_same(sdkp);
2026 					req->rq_flags |= RQF_QUIET;
2027 				}
2028 				break;
2029 			}
2030 		}
2031 		break;
2032 	default:
2033 		break;
2034 	}
2035 
2036  out:
2037 	if (sd_is_zoned(sdkp))
2038 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2039 
2040 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2041 					   "sd_done: completed %d of %d bytes\n",
2042 					   good_bytes, scsi_bufflen(SCpnt)));
2043 
2044 	return good_bytes;
2045 }
2046 
2047 /*
2048  * spinup disk - called only in sd_revalidate_disk()
2049  */
2050 static void
2051 sd_spinup_disk(struct scsi_disk *sdkp)
2052 {
2053 	unsigned char cmd[10];
2054 	unsigned long spintime_expire = 0;
2055 	int retries, spintime;
2056 	unsigned int the_result;
2057 	struct scsi_sense_hdr sshdr;
2058 	int sense_valid = 0;
2059 
2060 	spintime = 0;
2061 
2062 	/* Spin up drives, as required.  Only do this at boot time */
2063 	/* Spinup needs to be done for module loads too. */
2064 	do {
2065 		retries = 0;
2066 
2067 		do {
2068 			bool media_was_present = sdkp->media_present;
2069 
2070 			cmd[0] = TEST_UNIT_READY;
2071 			memset((void *) &cmd[1], 0, 9);
2072 
2073 			the_result = scsi_execute_req(sdkp->device, cmd,
2074 						      DMA_NONE, NULL, 0,
2075 						      &sshdr, SD_TIMEOUT,
2076 						      sdkp->max_retries, NULL);
2077 
2078 			/*
2079 			 * If the drive has indicated to us that it
2080 			 * doesn't have any media in it, don't bother
2081 			 * with any more polling.
2082 			 */
2083 			if (media_not_present(sdkp, &sshdr)) {
2084 				if (media_was_present)
2085 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2086 				return;
2087 			}
2088 
2089 			if (the_result)
2090 				sense_valid = scsi_sense_valid(&sshdr);
2091 			retries++;
2092 		} while (retries < 3 &&
2093 			 (!scsi_status_is_good(the_result) ||
2094 			  (scsi_status_is_check_condition(the_result) &&
2095 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2096 
2097 		if (!scsi_status_is_check_condition(the_result)) {
2098 			/* no sense, TUR either succeeded or failed
2099 			 * with a status error */
2100 			if(!spintime && !scsi_status_is_good(the_result)) {
2101 				sd_print_result(sdkp, "Test Unit Ready failed",
2102 						the_result);
2103 			}
2104 			break;
2105 		}
2106 
2107 		/*
2108 		 * The device does not want the automatic start to be issued.
2109 		 */
2110 		if (sdkp->device->no_start_on_add)
2111 			break;
2112 
2113 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2114 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2115 				break;	/* manual intervention required */
2116 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2117 				break;	/* standby */
2118 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2119 				break;	/* unavailable */
2120 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2121 				break;	/* sanitize in progress */
2122 			/*
2123 			 * Issue command to spin up drive when not ready
2124 			 */
2125 			if (!spintime) {
2126 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2127 				cmd[0] = START_STOP;
2128 				cmd[1] = 1;	/* Return immediately */
2129 				memset((void *) &cmd[2], 0, 8);
2130 				cmd[4] = 1;	/* Start spin cycle */
2131 				if (sdkp->device->start_stop_pwr_cond)
2132 					cmd[4] |= 1 << 4;
2133 				scsi_execute_req(sdkp->device, cmd, DMA_NONE,
2134 						 NULL, 0, &sshdr,
2135 						 SD_TIMEOUT, sdkp->max_retries,
2136 						 NULL);
2137 				spintime_expire = jiffies + 100 * HZ;
2138 				spintime = 1;
2139 			}
2140 			/* Wait 1 second for next try */
2141 			msleep(1000);
2142 			printk(KERN_CONT ".");
2143 
2144 		/*
2145 		 * Wait for USB flash devices with slow firmware.
2146 		 * Yes, this sense key/ASC combination shouldn't
2147 		 * occur here.  It's characteristic of these devices.
2148 		 */
2149 		} else if (sense_valid &&
2150 				sshdr.sense_key == UNIT_ATTENTION &&
2151 				sshdr.asc == 0x28) {
2152 			if (!spintime) {
2153 				spintime_expire = jiffies + 5 * HZ;
2154 				spintime = 1;
2155 			}
2156 			/* Wait 1 second for next try */
2157 			msleep(1000);
2158 		} else {
2159 			/* we don't understand the sense code, so it's
2160 			 * probably pointless to loop */
2161 			if(!spintime) {
2162 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2163 				sd_print_sense_hdr(sdkp, &sshdr);
2164 			}
2165 			break;
2166 		}
2167 
2168 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2169 
2170 	if (spintime) {
2171 		if (scsi_status_is_good(the_result))
2172 			printk(KERN_CONT "ready\n");
2173 		else
2174 			printk(KERN_CONT "not responding...\n");
2175 	}
2176 }
2177 
2178 /*
2179  * Determine whether disk supports Data Integrity Field.
2180  */
2181 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2182 {
2183 	struct scsi_device *sdp = sdkp->device;
2184 	u8 type;
2185 
2186 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2187 		sdkp->protection_type = 0;
2188 		return 0;
2189 	}
2190 
2191 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2192 
2193 	if (type > T10_PI_TYPE3_PROTECTION) {
2194 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2195 			  " protection type %u. Disabling disk!\n",
2196 			  type);
2197 		sdkp->protection_type = 0;
2198 		return -ENODEV;
2199 	}
2200 
2201 	sdkp->protection_type = type;
2202 
2203 	return 0;
2204 }
2205 
2206 static void sd_config_protection(struct scsi_disk *sdkp)
2207 {
2208 	struct scsi_device *sdp = sdkp->device;
2209 
2210 	if (!sdkp->first_scan)
2211 		return;
2212 
2213 	sd_dif_config_host(sdkp);
2214 
2215 	if (!sdkp->protection_type)
2216 		return;
2217 
2218 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2219 		sd_printk(KERN_NOTICE, sdkp,
2220 			  "Disabling DIF Type %u protection\n",
2221 			  sdkp->protection_type);
2222 		sdkp->protection_type = 0;
2223 	}
2224 
2225 	sd_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2226 		  sdkp->protection_type);
2227 }
2228 
2229 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2230 			struct scsi_sense_hdr *sshdr, int sense_valid,
2231 			int the_result)
2232 {
2233 	if (sense_valid)
2234 		sd_print_sense_hdr(sdkp, sshdr);
2235 	else
2236 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2237 
2238 	/*
2239 	 * Set dirty bit for removable devices if not ready -
2240 	 * sometimes drives will not report this properly.
2241 	 */
2242 	if (sdp->removable &&
2243 	    sense_valid && sshdr->sense_key == NOT_READY)
2244 		set_media_not_present(sdkp);
2245 
2246 	/*
2247 	 * We used to set media_present to 0 here to indicate no media
2248 	 * in the drive, but some drives fail read capacity even with
2249 	 * media present, so we can't do that.
2250 	 */
2251 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2252 }
2253 
2254 #define RC16_LEN 32
2255 #if RC16_LEN > SD_BUF_SIZE
2256 #error RC16_LEN must not be more than SD_BUF_SIZE
2257 #endif
2258 
2259 #define READ_CAPACITY_RETRIES_ON_RESET	10
2260 
2261 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2262 						unsigned char *buffer)
2263 {
2264 	unsigned char cmd[16];
2265 	struct scsi_sense_hdr sshdr;
2266 	int sense_valid = 0;
2267 	int the_result;
2268 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2269 	unsigned int alignment;
2270 	unsigned long long lba;
2271 	unsigned sector_size;
2272 
2273 	if (sdp->no_read_capacity_16)
2274 		return -EINVAL;
2275 
2276 	do {
2277 		memset(cmd, 0, 16);
2278 		cmd[0] = SERVICE_ACTION_IN_16;
2279 		cmd[1] = SAI_READ_CAPACITY_16;
2280 		cmd[13] = RC16_LEN;
2281 		memset(buffer, 0, RC16_LEN);
2282 
2283 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2284 					buffer, RC16_LEN, &sshdr,
2285 					SD_TIMEOUT, sdkp->max_retries, NULL);
2286 
2287 		if (media_not_present(sdkp, &sshdr))
2288 			return -ENODEV;
2289 
2290 		if (the_result > 0) {
2291 			sense_valid = scsi_sense_valid(&sshdr);
2292 			if (sense_valid &&
2293 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2294 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2295 			    sshdr.ascq == 0x00)
2296 				/* Invalid Command Operation Code or
2297 				 * Invalid Field in CDB, just retry
2298 				 * silently with RC10 */
2299 				return -EINVAL;
2300 			if (sense_valid &&
2301 			    sshdr.sense_key == UNIT_ATTENTION &&
2302 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2303 				/* Device reset might occur several times,
2304 				 * give it one more chance */
2305 				if (--reset_retries > 0)
2306 					continue;
2307 		}
2308 		retries--;
2309 
2310 	} while (the_result && retries);
2311 
2312 	if (the_result) {
2313 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2314 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2315 		return -EINVAL;
2316 	}
2317 
2318 	sector_size = get_unaligned_be32(&buffer[8]);
2319 	lba = get_unaligned_be64(&buffer[0]);
2320 
2321 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2322 		sdkp->capacity = 0;
2323 		return -ENODEV;
2324 	}
2325 
2326 	/* Logical blocks per physical block exponent */
2327 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2328 
2329 	/* RC basis */
2330 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2331 
2332 	/* Lowest aligned logical block */
2333 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2334 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2335 	if (alignment && sdkp->first_scan)
2336 		sd_printk(KERN_NOTICE, sdkp,
2337 			  "physical block alignment offset: %u\n", alignment);
2338 
2339 	if (buffer[14] & 0x80) { /* LBPME */
2340 		sdkp->lbpme = 1;
2341 
2342 		if (buffer[14] & 0x40) /* LBPRZ */
2343 			sdkp->lbprz = 1;
2344 
2345 		sd_config_discard(sdkp, SD_LBP_WS16);
2346 	}
2347 
2348 	sdkp->capacity = lba + 1;
2349 	return sector_size;
2350 }
2351 
2352 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2353 						unsigned char *buffer)
2354 {
2355 	unsigned char cmd[16];
2356 	struct scsi_sense_hdr sshdr;
2357 	int sense_valid = 0;
2358 	int the_result;
2359 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2360 	sector_t lba;
2361 	unsigned sector_size;
2362 
2363 	do {
2364 		cmd[0] = READ_CAPACITY;
2365 		memset(&cmd[1], 0, 9);
2366 		memset(buffer, 0, 8);
2367 
2368 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2369 					buffer, 8, &sshdr,
2370 					SD_TIMEOUT, sdkp->max_retries, NULL);
2371 
2372 		if (media_not_present(sdkp, &sshdr))
2373 			return -ENODEV;
2374 
2375 		if (the_result > 0) {
2376 			sense_valid = scsi_sense_valid(&sshdr);
2377 			if (sense_valid &&
2378 			    sshdr.sense_key == UNIT_ATTENTION &&
2379 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2380 				/* Device reset might occur several times,
2381 				 * give it one more chance */
2382 				if (--reset_retries > 0)
2383 					continue;
2384 		}
2385 		retries--;
2386 
2387 	} while (the_result && retries);
2388 
2389 	if (the_result) {
2390 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2391 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2392 		return -EINVAL;
2393 	}
2394 
2395 	sector_size = get_unaligned_be32(&buffer[4]);
2396 	lba = get_unaligned_be32(&buffer[0]);
2397 
2398 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2399 		/* Some buggy (usb cardreader) devices return an lba of
2400 		   0xffffffff when the want to report a size of 0 (with
2401 		   which they really mean no media is present) */
2402 		sdkp->capacity = 0;
2403 		sdkp->physical_block_size = sector_size;
2404 		return sector_size;
2405 	}
2406 
2407 	sdkp->capacity = lba + 1;
2408 	sdkp->physical_block_size = sector_size;
2409 	return sector_size;
2410 }
2411 
2412 static int sd_try_rc16_first(struct scsi_device *sdp)
2413 {
2414 	if (sdp->host->max_cmd_len < 16)
2415 		return 0;
2416 	if (sdp->try_rc_10_first)
2417 		return 0;
2418 	if (sdp->scsi_level > SCSI_SPC_2)
2419 		return 1;
2420 	if (scsi_device_protection(sdp))
2421 		return 1;
2422 	return 0;
2423 }
2424 
2425 /*
2426  * read disk capacity
2427  */
2428 static void
2429 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2430 {
2431 	int sector_size;
2432 	struct scsi_device *sdp = sdkp->device;
2433 
2434 	if (sd_try_rc16_first(sdp)) {
2435 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2436 		if (sector_size == -EOVERFLOW)
2437 			goto got_data;
2438 		if (sector_size == -ENODEV)
2439 			return;
2440 		if (sector_size < 0)
2441 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2442 		if (sector_size < 0)
2443 			return;
2444 	} else {
2445 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2446 		if (sector_size == -EOVERFLOW)
2447 			goto got_data;
2448 		if (sector_size < 0)
2449 			return;
2450 		if ((sizeof(sdkp->capacity) > 4) &&
2451 		    (sdkp->capacity > 0xffffffffULL)) {
2452 			int old_sector_size = sector_size;
2453 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2454 					"Trying to use READ CAPACITY(16).\n");
2455 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2456 			if (sector_size < 0) {
2457 				sd_printk(KERN_NOTICE, sdkp,
2458 					"Using 0xffffffff as device size\n");
2459 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2460 				sector_size = old_sector_size;
2461 				goto got_data;
2462 			}
2463 			/* Remember that READ CAPACITY(16) succeeded */
2464 			sdp->try_rc_10_first = 0;
2465 		}
2466 	}
2467 
2468 	/* Some devices are known to return the total number of blocks,
2469 	 * not the highest block number.  Some devices have versions
2470 	 * which do this and others which do not.  Some devices we might
2471 	 * suspect of doing this but we don't know for certain.
2472 	 *
2473 	 * If we know the reported capacity is wrong, decrement it.  If
2474 	 * we can only guess, then assume the number of blocks is even
2475 	 * (usually true but not always) and err on the side of lowering
2476 	 * the capacity.
2477 	 */
2478 	if (sdp->fix_capacity ||
2479 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2480 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2481 				"from its reported value: %llu\n",
2482 				(unsigned long long) sdkp->capacity);
2483 		--sdkp->capacity;
2484 	}
2485 
2486 got_data:
2487 	if (sector_size == 0) {
2488 		sector_size = 512;
2489 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2490 			  "assuming 512.\n");
2491 	}
2492 
2493 	if (sector_size != 512 &&
2494 	    sector_size != 1024 &&
2495 	    sector_size != 2048 &&
2496 	    sector_size != 4096) {
2497 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2498 			  sector_size);
2499 		/*
2500 		 * The user might want to re-format the drive with
2501 		 * a supported sectorsize.  Once this happens, it
2502 		 * would be relatively trivial to set the thing up.
2503 		 * For this reason, we leave the thing in the table.
2504 		 */
2505 		sdkp->capacity = 0;
2506 		/*
2507 		 * set a bogus sector size so the normal read/write
2508 		 * logic in the block layer will eventually refuse any
2509 		 * request on this device without tripping over power
2510 		 * of two sector size assumptions
2511 		 */
2512 		sector_size = 512;
2513 	}
2514 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2515 	blk_queue_physical_block_size(sdp->request_queue,
2516 				      sdkp->physical_block_size);
2517 	sdkp->device->sector_size = sector_size;
2518 
2519 	if (sdkp->capacity > 0xffffffff)
2520 		sdp->use_16_for_rw = 1;
2521 
2522 }
2523 
2524 /*
2525  * Print disk capacity
2526  */
2527 static void
2528 sd_print_capacity(struct scsi_disk *sdkp,
2529 		  sector_t old_capacity)
2530 {
2531 	int sector_size = sdkp->device->sector_size;
2532 	char cap_str_2[10], cap_str_10[10];
2533 
2534 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2535 		return;
2536 
2537 	string_get_size(sdkp->capacity, sector_size,
2538 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2539 	string_get_size(sdkp->capacity, sector_size,
2540 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2541 
2542 	sd_printk(KERN_NOTICE, sdkp,
2543 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2544 		  (unsigned long long)sdkp->capacity,
2545 		  sector_size, cap_str_10, cap_str_2);
2546 
2547 	if (sdkp->physical_block_size != sector_size)
2548 		sd_printk(KERN_NOTICE, sdkp,
2549 			  "%u-byte physical blocks\n",
2550 			  sdkp->physical_block_size);
2551 }
2552 
2553 /* called with buffer of length 512 */
2554 static inline int
2555 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2556 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2557 		 struct scsi_sense_hdr *sshdr)
2558 {
2559 	/*
2560 	 * If we must use MODE SENSE(10), make sure that the buffer length
2561 	 * is at least 8 bytes so that the mode sense header fits.
2562 	 */
2563 	if (sdkp->device->use_10_for_ms && len < 8)
2564 		len = 8;
2565 
2566 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2567 			       SD_TIMEOUT, sdkp->max_retries, data,
2568 			       sshdr);
2569 }
2570 
2571 /*
2572  * read write protect setting, if possible - called only in sd_revalidate_disk()
2573  * called with buffer of length SD_BUF_SIZE
2574  */
2575 static void
2576 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2577 {
2578 	int res;
2579 	struct scsi_device *sdp = sdkp->device;
2580 	struct scsi_mode_data data;
2581 	int old_wp = sdkp->write_prot;
2582 
2583 	set_disk_ro(sdkp->disk, 0);
2584 	if (sdp->skip_ms_page_3f) {
2585 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2586 		return;
2587 	}
2588 
2589 	if (sdp->use_192_bytes_for_3f) {
2590 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2591 	} else {
2592 		/*
2593 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2594 		 * We have to start carefully: some devices hang if we ask
2595 		 * for more than is available.
2596 		 */
2597 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2598 
2599 		/*
2600 		 * Second attempt: ask for page 0 When only page 0 is
2601 		 * implemented, a request for page 3F may return Sense Key
2602 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2603 		 * CDB.
2604 		 */
2605 		if (res < 0)
2606 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2607 
2608 		/*
2609 		 * Third attempt: ask 255 bytes, as we did earlier.
2610 		 */
2611 		if (res < 0)
2612 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2613 					       &data, NULL);
2614 	}
2615 
2616 	if (res < 0) {
2617 		sd_first_printk(KERN_WARNING, sdkp,
2618 			  "Test WP failed, assume Write Enabled\n");
2619 	} else {
2620 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2621 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2622 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2623 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2624 				  sdkp->write_prot ? "on" : "off");
2625 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2626 		}
2627 	}
2628 }
2629 
2630 /*
2631  * sd_read_cache_type - called only from sd_revalidate_disk()
2632  * called with buffer of length SD_BUF_SIZE
2633  */
2634 static void
2635 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2636 {
2637 	int len = 0, res;
2638 	struct scsi_device *sdp = sdkp->device;
2639 
2640 	int dbd;
2641 	int modepage;
2642 	int first_len;
2643 	struct scsi_mode_data data;
2644 	struct scsi_sense_hdr sshdr;
2645 	int old_wce = sdkp->WCE;
2646 	int old_rcd = sdkp->RCD;
2647 	int old_dpofua = sdkp->DPOFUA;
2648 
2649 
2650 	if (sdkp->cache_override)
2651 		return;
2652 
2653 	first_len = 4;
2654 	if (sdp->skip_ms_page_8) {
2655 		if (sdp->type == TYPE_RBC)
2656 			goto defaults;
2657 		else {
2658 			if (sdp->skip_ms_page_3f)
2659 				goto defaults;
2660 			modepage = 0x3F;
2661 			if (sdp->use_192_bytes_for_3f)
2662 				first_len = 192;
2663 			dbd = 0;
2664 		}
2665 	} else if (sdp->type == TYPE_RBC) {
2666 		modepage = 6;
2667 		dbd = 8;
2668 	} else {
2669 		modepage = 8;
2670 		dbd = 0;
2671 	}
2672 
2673 	/* cautiously ask */
2674 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2675 			&data, &sshdr);
2676 
2677 	if (res < 0)
2678 		goto bad_sense;
2679 
2680 	if (!data.header_length) {
2681 		modepage = 6;
2682 		first_len = 0;
2683 		sd_first_printk(KERN_ERR, sdkp,
2684 				"Missing header in MODE_SENSE response\n");
2685 	}
2686 
2687 	/* that went OK, now ask for the proper length */
2688 	len = data.length;
2689 
2690 	/*
2691 	 * We're only interested in the first three bytes, actually.
2692 	 * But the data cache page is defined for the first 20.
2693 	 */
2694 	if (len < 3)
2695 		goto bad_sense;
2696 	else if (len > SD_BUF_SIZE) {
2697 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2698 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2699 		len = SD_BUF_SIZE;
2700 	}
2701 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2702 		len = 192;
2703 
2704 	/* Get the data */
2705 	if (len > first_len)
2706 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2707 				&data, &sshdr);
2708 
2709 	if (!res) {
2710 		int offset = data.header_length + data.block_descriptor_length;
2711 
2712 		while (offset < len) {
2713 			u8 page_code = buffer[offset] & 0x3F;
2714 			u8 spf       = buffer[offset] & 0x40;
2715 
2716 			if (page_code == 8 || page_code == 6) {
2717 				/* We're interested only in the first 3 bytes.
2718 				 */
2719 				if (len - offset <= 2) {
2720 					sd_first_printk(KERN_ERR, sdkp,
2721 						"Incomplete mode parameter "
2722 							"data\n");
2723 					goto defaults;
2724 				} else {
2725 					modepage = page_code;
2726 					goto Page_found;
2727 				}
2728 			} else {
2729 				/* Go to the next page */
2730 				if (spf && len - offset > 3)
2731 					offset += 4 + (buffer[offset+2] << 8) +
2732 						buffer[offset+3];
2733 				else if (!spf && len - offset > 1)
2734 					offset += 2 + buffer[offset+1];
2735 				else {
2736 					sd_first_printk(KERN_ERR, sdkp,
2737 							"Incomplete mode "
2738 							"parameter data\n");
2739 					goto defaults;
2740 				}
2741 			}
2742 		}
2743 
2744 		sd_first_printk(KERN_WARNING, sdkp,
2745 				"No Caching mode page found\n");
2746 		goto defaults;
2747 
2748 	Page_found:
2749 		if (modepage == 8) {
2750 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2751 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2752 		} else {
2753 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2754 			sdkp->RCD = 0;
2755 		}
2756 
2757 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2758 		if (sdp->broken_fua) {
2759 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2760 			sdkp->DPOFUA = 0;
2761 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2762 			   !sdkp->device->use_16_for_rw) {
2763 			sd_first_printk(KERN_NOTICE, sdkp,
2764 				  "Uses READ/WRITE(6), disabling FUA\n");
2765 			sdkp->DPOFUA = 0;
2766 		}
2767 
2768 		/* No cache flush allowed for write protected devices */
2769 		if (sdkp->WCE && sdkp->write_prot)
2770 			sdkp->WCE = 0;
2771 
2772 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2773 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2774 			sd_printk(KERN_NOTICE, sdkp,
2775 				  "Write cache: %s, read cache: %s, %s\n",
2776 				  sdkp->WCE ? "enabled" : "disabled",
2777 				  sdkp->RCD ? "disabled" : "enabled",
2778 				  sdkp->DPOFUA ? "supports DPO and FUA"
2779 				  : "doesn't support DPO or FUA");
2780 
2781 		return;
2782 	}
2783 
2784 bad_sense:
2785 	if (scsi_sense_valid(&sshdr) &&
2786 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2787 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2788 		/* Invalid field in CDB */
2789 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2790 	else
2791 		sd_first_printk(KERN_ERR, sdkp,
2792 				"Asking for cache data failed\n");
2793 
2794 defaults:
2795 	if (sdp->wce_default_on) {
2796 		sd_first_printk(KERN_NOTICE, sdkp,
2797 				"Assuming drive cache: write back\n");
2798 		sdkp->WCE = 1;
2799 	} else {
2800 		sd_first_printk(KERN_WARNING, sdkp,
2801 				"Assuming drive cache: write through\n");
2802 		sdkp->WCE = 0;
2803 	}
2804 	sdkp->RCD = 0;
2805 	sdkp->DPOFUA = 0;
2806 }
2807 
2808 /*
2809  * The ATO bit indicates whether the DIF application tag is available
2810  * for use by the operating system.
2811  */
2812 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2813 {
2814 	int res, offset;
2815 	struct scsi_device *sdp = sdkp->device;
2816 	struct scsi_mode_data data;
2817 	struct scsi_sense_hdr sshdr;
2818 
2819 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2820 		return;
2821 
2822 	if (sdkp->protection_type == 0)
2823 		return;
2824 
2825 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2826 			      sdkp->max_retries, &data, &sshdr);
2827 
2828 	if (res < 0 || !data.header_length ||
2829 	    data.length < 6) {
2830 		sd_first_printk(KERN_WARNING, sdkp,
2831 			  "getting Control mode page failed, assume no ATO\n");
2832 
2833 		if (scsi_sense_valid(&sshdr))
2834 			sd_print_sense_hdr(sdkp, &sshdr);
2835 
2836 		return;
2837 	}
2838 
2839 	offset = data.header_length + data.block_descriptor_length;
2840 
2841 	if ((buffer[offset] & 0x3f) != 0x0a) {
2842 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2843 		return;
2844 	}
2845 
2846 	if ((buffer[offset + 5] & 0x80) == 0)
2847 		return;
2848 
2849 	sdkp->ATO = 1;
2850 
2851 	return;
2852 }
2853 
2854 /**
2855  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2856  * @sdkp: disk to query
2857  */
2858 static void sd_read_block_limits(struct scsi_disk *sdkp)
2859 {
2860 	struct scsi_vpd *vpd;
2861 
2862 	rcu_read_lock();
2863 
2864 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2865 	if (!vpd || vpd->len < 16)
2866 		goto out;
2867 
2868 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2869 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2870 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2871 
2872 	if (vpd->len >= 64) {
2873 		unsigned int lba_count, desc_count;
2874 
2875 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2876 
2877 		if (!sdkp->lbpme)
2878 			goto out;
2879 
2880 		lba_count = get_unaligned_be32(&vpd->data[20]);
2881 		desc_count = get_unaligned_be32(&vpd->data[24]);
2882 
2883 		if (lba_count && desc_count)
2884 			sdkp->max_unmap_blocks = lba_count;
2885 
2886 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2887 
2888 		if (vpd->data[32] & 0x80)
2889 			sdkp->unmap_alignment =
2890 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2891 
2892 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2893 
2894 			if (sdkp->max_unmap_blocks)
2895 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2896 			else
2897 				sd_config_discard(sdkp, SD_LBP_WS16);
2898 
2899 		} else {	/* LBP VPD page tells us what to use */
2900 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2901 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2902 			else if (sdkp->lbpws)
2903 				sd_config_discard(sdkp, SD_LBP_WS16);
2904 			else if (sdkp->lbpws10)
2905 				sd_config_discard(sdkp, SD_LBP_WS10);
2906 			else
2907 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2908 		}
2909 	}
2910 
2911  out:
2912 	rcu_read_unlock();
2913 }
2914 
2915 /**
2916  * sd_read_block_characteristics - Query block dev. characteristics
2917  * @sdkp: disk to query
2918  */
2919 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2920 {
2921 	struct request_queue *q = sdkp->disk->queue;
2922 	struct scsi_vpd *vpd;
2923 	u16 rot;
2924 	u8 zoned;
2925 
2926 	rcu_read_lock();
2927 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
2928 
2929 	if (!vpd || vpd->len < 8) {
2930 		rcu_read_unlock();
2931 	        return;
2932 	}
2933 
2934 	rot = get_unaligned_be16(&vpd->data[4]);
2935 	zoned = (vpd->data[8] >> 4) & 3;
2936 	rcu_read_unlock();
2937 
2938 	if (rot == 1) {
2939 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2940 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2941 	}
2942 
2943 	if (sdkp->device->type == TYPE_ZBC) {
2944 		/* Host-managed */
2945 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
2946 	} else {
2947 		sdkp->zoned = zoned;
2948 		if (sdkp->zoned == 1) {
2949 			/* Host-aware */
2950 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
2951 		} else {
2952 			/* Regular disk or drive managed disk */
2953 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
2954 		}
2955 	}
2956 
2957 	if (!sdkp->first_scan)
2958 		return;
2959 
2960 	if (blk_queue_is_zoned(q)) {
2961 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
2962 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
2963 	} else {
2964 		if (sdkp->zoned == 1)
2965 			sd_printk(KERN_NOTICE, sdkp,
2966 				  "Host-aware SMR disk used as regular disk\n");
2967 		else if (sdkp->zoned == 2)
2968 			sd_printk(KERN_NOTICE, sdkp,
2969 				  "Drive-managed SMR disk\n");
2970 	}
2971 }
2972 
2973 /**
2974  * sd_read_block_provisioning - Query provisioning VPD page
2975  * @sdkp: disk to query
2976  */
2977 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
2978 {
2979 	struct scsi_vpd *vpd;
2980 
2981 	if (sdkp->lbpme == 0)
2982 		return;
2983 
2984 	rcu_read_lock();
2985 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
2986 
2987 	if (!vpd || vpd->len < 8) {
2988 		rcu_read_unlock();
2989 		return;
2990 	}
2991 
2992 	sdkp->lbpvpd	= 1;
2993 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
2994 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
2995 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
2996 	rcu_read_unlock();
2997 }
2998 
2999 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3000 {
3001 	struct scsi_device *sdev = sdkp->device;
3002 
3003 	if (sdev->host->no_write_same) {
3004 		sdev->no_write_same = 1;
3005 
3006 		return;
3007 	}
3008 
3009 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
3010 		struct scsi_vpd *vpd;
3011 
3012 		sdev->no_report_opcodes = 1;
3013 
3014 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3015 		 * CODES is unsupported and the device has an ATA
3016 		 * Information VPD page (SAT).
3017 		 */
3018 		rcu_read_lock();
3019 		vpd = rcu_dereference(sdev->vpd_pg89);
3020 		if (vpd)
3021 			sdev->no_write_same = 1;
3022 		rcu_read_unlock();
3023 	}
3024 
3025 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3026 		sdkp->ws16 = 1;
3027 
3028 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3029 		sdkp->ws10 = 1;
3030 }
3031 
3032 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3033 {
3034 	struct scsi_device *sdev = sdkp->device;
3035 
3036 	if (!sdev->security_supported)
3037 		return;
3038 
3039 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3040 			SECURITY_PROTOCOL_IN) == 1 &&
3041 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3042 			SECURITY_PROTOCOL_OUT) == 1)
3043 		sdkp->security = 1;
3044 }
3045 
3046 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3047 {
3048 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3049 }
3050 
3051 /**
3052  * sd_read_cpr - Query concurrent positioning ranges
3053  * @sdkp:	disk to query
3054  */
3055 static void sd_read_cpr(struct scsi_disk *sdkp)
3056 {
3057 	struct blk_independent_access_ranges *iars = NULL;
3058 	unsigned char *buffer = NULL;
3059 	unsigned int nr_cpr = 0;
3060 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3061 	u8 *desc;
3062 
3063 	/*
3064 	 * We need to have the capacity set first for the block layer to be
3065 	 * able to check the ranges.
3066 	 */
3067 	if (sdkp->first_scan)
3068 		return;
3069 
3070 	if (!sdkp->capacity)
3071 		goto out;
3072 
3073 	/*
3074 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3075 	 * leading to a maximum page size of 64 + 256*32 bytes.
3076 	 */
3077 	buf_len = 64 + 256*32;
3078 	buffer = kmalloc(buf_len, GFP_KERNEL);
3079 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3080 		goto out;
3081 
3082 	/* We must have at least a 64B header and one 32B range descriptor */
3083 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3084 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3085 		sd_printk(KERN_ERR, sdkp,
3086 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3087 		goto out;
3088 	}
3089 
3090 	nr_cpr = (vpd_len - 64) / 32;
3091 	if (nr_cpr == 1) {
3092 		nr_cpr = 0;
3093 		goto out;
3094 	}
3095 
3096 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3097 	if (!iars) {
3098 		nr_cpr = 0;
3099 		goto out;
3100 	}
3101 
3102 	desc = &buffer[64];
3103 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3104 		if (desc[0] != i) {
3105 			sd_printk(KERN_ERR, sdkp,
3106 				"Invalid Concurrent Positioning Range number\n");
3107 			nr_cpr = 0;
3108 			break;
3109 		}
3110 
3111 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3112 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3113 	}
3114 
3115 out:
3116 	disk_set_independent_access_ranges(sdkp->disk, iars);
3117 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3118 		sd_printk(KERN_NOTICE, sdkp,
3119 			  "%u concurrent positioning ranges\n", nr_cpr);
3120 		sdkp->nr_actuators = nr_cpr;
3121 	}
3122 
3123 	kfree(buffer);
3124 }
3125 
3126 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3127 {
3128 	struct scsi_device *sdp = sdkp->device;
3129 	unsigned int min_xfer_bytes =
3130 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3131 
3132 	if (sdkp->min_xfer_blocks == 0)
3133 		return false;
3134 
3135 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3136 		sd_first_printk(KERN_WARNING, sdkp,
3137 				"Preferred minimum I/O size %u bytes not a " \
3138 				"multiple of physical block size (%u bytes)\n",
3139 				min_xfer_bytes, sdkp->physical_block_size);
3140 		sdkp->min_xfer_blocks = 0;
3141 		return false;
3142 	}
3143 
3144 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3145 			min_xfer_bytes);
3146 	return true;
3147 }
3148 
3149 /*
3150  * Determine the device's preferred I/O size for reads and writes
3151  * unless the reported value is unreasonably small, large, not a
3152  * multiple of the physical block size, or simply garbage.
3153  */
3154 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3155 				      unsigned int dev_max)
3156 {
3157 	struct scsi_device *sdp = sdkp->device;
3158 	unsigned int opt_xfer_bytes =
3159 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3160 	unsigned int min_xfer_bytes =
3161 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3162 
3163 	if (sdkp->opt_xfer_blocks == 0)
3164 		return false;
3165 
3166 	if (sdkp->opt_xfer_blocks > dev_max) {
3167 		sd_first_printk(KERN_WARNING, sdkp,
3168 				"Optimal transfer size %u logical blocks " \
3169 				"> dev_max (%u logical blocks)\n",
3170 				sdkp->opt_xfer_blocks, dev_max);
3171 		return false;
3172 	}
3173 
3174 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3175 		sd_first_printk(KERN_WARNING, sdkp,
3176 				"Optimal transfer size %u logical blocks " \
3177 				"> sd driver limit (%u logical blocks)\n",
3178 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3179 		return false;
3180 	}
3181 
3182 	if (opt_xfer_bytes < PAGE_SIZE) {
3183 		sd_first_printk(KERN_WARNING, sdkp,
3184 				"Optimal transfer size %u bytes < " \
3185 				"PAGE_SIZE (%u bytes)\n",
3186 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3187 		return false;
3188 	}
3189 
3190 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3191 		sd_first_printk(KERN_WARNING, sdkp,
3192 				"Optimal transfer size %u bytes not a " \
3193 				"multiple of preferred minimum block " \
3194 				"size (%u bytes)\n",
3195 				opt_xfer_bytes, min_xfer_bytes);
3196 		return false;
3197 	}
3198 
3199 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3200 		sd_first_printk(KERN_WARNING, sdkp,
3201 				"Optimal transfer size %u bytes not a " \
3202 				"multiple of physical block size (%u bytes)\n",
3203 				opt_xfer_bytes, sdkp->physical_block_size);
3204 		return false;
3205 	}
3206 
3207 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3208 			opt_xfer_bytes);
3209 	return true;
3210 }
3211 
3212 /**
3213  *	sd_revalidate_disk - called the first time a new disk is seen,
3214  *	performs disk spin up, read_capacity, etc.
3215  *	@disk: struct gendisk we care about
3216  **/
3217 static int sd_revalidate_disk(struct gendisk *disk)
3218 {
3219 	struct scsi_disk *sdkp = scsi_disk(disk);
3220 	struct scsi_device *sdp = sdkp->device;
3221 	struct request_queue *q = sdkp->disk->queue;
3222 	sector_t old_capacity = sdkp->capacity;
3223 	unsigned char *buffer;
3224 	unsigned int dev_max, rw_max;
3225 
3226 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3227 				      "sd_revalidate_disk\n"));
3228 
3229 	/*
3230 	 * If the device is offline, don't try and read capacity or any
3231 	 * of the other niceties.
3232 	 */
3233 	if (!scsi_device_online(sdp))
3234 		goto out;
3235 
3236 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3237 	if (!buffer) {
3238 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3239 			  "allocation failure.\n");
3240 		goto out;
3241 	}
3242 
3243 	sd_spinup_disk(sdkp);
3244 
3245 	/*
3246 	 * Without media there is no reason to ask; moreover, some devices
3247 	 * react badly if we do.
3248 	 */
3249 	if (sdkp->media_present) {
3250 		sd_read_capacity(sdkp, buffer);
3251 
3252 		/*
3253 		 * set the default to rotational.  All non-rotational devices
3254 		 * support the block characteristics VPD page, which will
3255 		 * cause this to be updated correctly and any device which
3256 		 * doesn't support it should be treated as rotational.
3257 		 */
3258 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3259 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3260 
3261 		if (scsi_device_supports_vpd(sdp)) {
3262 			sd_read_block_provisioning(sdkp);
3263 			sd_read_block_limits(sdkp);
3264 			sd_read_block_characteristics(sdkp);
3265 			sd_zbc_read_zones(sdkp, buffer);
3266 			sd_read_cpr(sdkp);
3267 		}
3268 
3269 		sd_print_capacity(sdkp, old_capacity);
3270 
3271 		sd_read_write_protect_flag(sdkp, buffer);
3272 		sd_read_cache_type(sdkp, buffer);
3273 		sd_read_app_tag_own(sdkp, buffer);
3274 		sd_read_write_same(sdkp, buffer);
3275 		sd_read_security(sdkp, buffer);
3276 		sd_config_protection(sdkp);
3277 	}
3278 
3279 	/*
3280 	 * We now have all cache related info, determine how we deal
3281 	 * with flush requests.
3282 	 */
3283 	sd_set_flush_flag(sdkp);
3284 
3285 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3286 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3287 
3288 	/* Some devices report a maximum block count for READ/WRITE requests. */
3289 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3290 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3291 
3292 	if (sd_validate_min_xfer_size(sdkp))
3293 		blk_queue_io_min(sdkp->disk->queue,
3294 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3295 	else
3296 		blk_queue_io_min(sdkp->disk->queue, 0);
3297 
3298 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3299 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3300 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3301 	} else {
3302 		q->limits.io_opt = 0;
3303 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3304 				      (sector_t)BLK_DEF_MAX_SECTORS);
3305 	}
3306 
3307 	/*
3308 	 * Limit default to SCSI host optimal sector limit if set. There may be
3309 	 * an impact on performance for when the size of a request exceeds this
3310 	 * host limit.
3311 	 */
3312 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3313 
3314 	/* Do not exceed controller limit */
3315 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3316 
3317 	/*
3318 	 * Only update max_sectors if previously unset or if the current value
3319 	 * exceeds the capabilities of the hardware.
3320 	 */
3321 	if (sdkp->first_scan ||
3322 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3323 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3324 		q->limits.max_sectors = rw_max;
3325 
3326 	sdkp->first_scan = 0;
3327 
3328 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3329 	sd_config_write_same(sdkp);
3330 	kfree(buffer);
3331 
3332 	/*
3333 	 * For a zoned drive, revalidating the zones can be done only once
3334 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3335 	 * capacity to 0.
3336 	 */
3337 	if (sd_zbc_revalidate_zones(sdkp))
3338 		set_capacity_and_notify(disk, 0);
3339 
3340  out:
3341 	return 0;
3342 }
3343 
3344 /**
3345  *	sd_unlock_native_capacity - unlock native capacity
3346  *	@disk: struct gendisk to set capacity for
3347  *
3348  *	Block layer calls this function if it detects that partitions
3349  *	on @disk reach beyond the end of the device.  If the SCSI host
3350  *	implements ->unlock_native_capacity() method, it's invoked to
3351  *	give it a chance to adjust the device capacity.
3352  *
3353  *	CONTEXT:
3354  *	Defined by block layer.  Might sleep.
3355  */
3356 static void sd_unlock_native_capacity(struct gendisk *disk)
3357 {
3358 	struct scsi_device *sdev = scsi_disk(disk)->device;
3359 
3360 	if (sdev->host->hostt->unlock_native_capacity)
3361 		sdev->host->hostt->unlock_native_capacity(sdev);
3362 }
3363 
3364 /**
3365  *	sd_format_disk_name - format disk name
3366  *	@prefix: name prefix - ie. "sd" for SCSI disks
3367  *	@index: index of the disk to format name for
3368  *	@buf: output buffer
3369  *	@buflen: length of the output buffer
3370  *
3371  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3372  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3373  *	which is followed by sdaaa.
3374  *
3375  *	This is basically 26 base counting with one extra 'nil' entry
3376  *	at the beginning from the second digit on and can be
3377  *	determined using similar method as 26 base conversion with the
3378  *	index shifted -1 after each digit is computed.
3379  *
3380  *	CONTEXT:
3381  *	Don't care.
3382  *
3383  *	RETURNS:
3384  *	0 on success, -errno on failure.
3385  */
3386 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3387 {
3388 	const int base = 'z' - 'a' + 1;
3389 	char *begin = buf + strlen(prefix);
3390 	char *end = buf + buflen;
3391 	char *p;
3392 	int unit;
3393 
3394 	p = end - 1;
3395 	*p = '\0';
3396 	unit = base;
3397 	do {
3398 		if (p == begin)
3399 			return -EINVAL;
3400 		*--p = 'a' + (index % unit);
3401 		index = (index / unit) - 1;
3402 	} while (index >= 0);
3403 
3404 	memmove(begin, p, end - p);
3405 	memcpy(buf, prefix, strlen(prefix));
3406 
3407 	return 0;
3408 }
3409 
3410 /**
3411  *	sd_probe - called during driver initialization and whenever a
3412  *	new scsi device is attached to the system. It is called once
3413  *	for each scsi device (not just disks) present.
3414  *	@dev: pointer to device object
3415  *
3416  *	Returns 0 if successful (or not interested in this scsi device
3417  *	(e.g. scanner)); 1 when there is an error.
3418  *
3419  *	Note: this function is invoked from the scsi mid-level.
3420  *	This function sets up the mapping between a given
3421  *	<host,channel,id,lun> (found in sdp) and new device name
3422  *	(e.g. /dev/sda). More precisely it is the block device major
3423  *	and minor number that is chosen here.
3424  *
3425  *	Assume sd_probe is not re-entrant (for time being)
3426  *	Also think about sd_probe() and sd_remove() running coincidentally.
3427  **/
3428 static int sd_probe(struct device *dev)
3429 {
3430 	struct scsi_device *sdp = to_scsi_device(dev);
3431 	struct scsi_disk *sdkp;
3432 	struct gendisk *gd;
3433 	int index;
3434 	int error;
3435 
3436 	scsi_autopm_get_device(sdp);
3437 	error = -ENODEV;
3438 	if (sdp->type != TYPE_DISK &&
3439 	    sdp->type != TYPE_ZBC &&
3440 	    sdp->type != TYPE_MOD &&
3441 	    sdp->type != TYPE_RBC)
3442 		goto out;
3443 
3444 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3445 		sdev_printk(KERN_WARNING, sdp,
3446 			    "Unsupported ZBC host-managed device.\n");
3447 		goto out;
3448 	}
3449 
3450 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3451 					"sd_probe\n"));
3452 
3453 	error = -ENOMEM;
3454 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3455 	if (!sdkp)
3456 		goto out;
3457 
3458 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3459 					 &sd_bio_compl_lkclass);
3460 	if (!gd)
3461 		goto out_free;
3462 
3463 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3464 	if (index < 0) {
3465 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3466 		goto out_put;
3467 	}
3468 
3469 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3470 	if (error) {
3471 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3472 		goto out_free_index;
3473 	}
3474 
3475 	sdkp->device = sdp;
3476 	sdkp->disk = gd;
3477 	sdkp->index = index;
3478 	sdkp->max_retries = SD_MAX_RETRIES;
3479 	atomic_set(&sdkp->openers, 0);
3480 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3481 
3482 	if (!sdp->request_queue->rq_timeout) {
3483 		if (sdp->type != TYPE_MOD)
3484 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3485 		else
3486 			blk_queue_rq_timeout(sdp->request_queue,
3487 					     SD_MOD_TIMEOUT);
3488 	}
3489 
3490 	device_initialize(&sdkp->disk_dev);
3491 	sdkp->disk_dev.parent = get_device(dev);
3492 	sdkp->disk_dev.class = &sd_disk_class;
3493 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3494 
3495 	error = device_add(&sdkp->disk_dev);
3496 	if (error) {
3497 		put_device(&sdkp->disk_dev);
3498 		goto out;
3499 	}
3500 
3501 	dev_set_drvdata(dev, sdkp);
3502 
3503 	gd->major = sd_major((index & 0xf0) >> 4);
3504 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3505 	gd->minors = SD_MINORS;
3506 
3507 	gd->fops = &sd_fops;
3508 	gd->private_data = sdkp;
3509 
3510 	/* defaults, until the device tells us otherwise */
3511 	sdp->sector_size = 512;
3512 	sdkp->capacity = 0;
3513 	sdkp->media_present = 1;
3514 	sdkp->write_prot = 0;
3515 	sdkp->cache_override = 0;
3516 	sdkp->WCE = 0;
3517 	sdkp->RCD = 0;
3518 	sdkp->ATO = 0;
3519 	sdkp->first_scan = 1;
3520 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3521 
3522 	sd_revalidate_disk(gd);
3523 
3524 	if (sdp->removable) {
3525 		gd->flags |= GENHD_FL_REMOVABLE;
3526 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3527 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3528 	}
3529 
3530 	blk_pm_runtime_init(sdp->request_queue, dev);
3531 	if (sdp->rpm_autosuspend) {
3532 		pm_runtime_set_autosuspend_delay(dev,
3533 			sdp->host->hostt->rpm_autosuspend_delay);
3534 	}
3535 
3536 	error = device_add_disk(dev, gd, NULL);
3537 	if (error) {
3538 		put_device(&sdkp->disk_dev);
3539 		put_disk(gd);
3540 		goto out;
3541 	}
3542 
3543 	if (sdkp->security) {
3544 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3545 		if (sdkp->opal_dev)
3546 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3547 	}
3548 
3549 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3550 		  sdp->removable ? "removable " : "");
3551 	scsi_autopm_put_device(sdp);
3552 
3553 	return 0;
3554 
3555  out_free_index:
3556 	ida_free(&sd_index_ida, index);
3557  out_put:
3558 	put_disk(gd);
3559  out_free:
3560 	kfree(sdkp);
3561  out:
3562 	scsi_autopm_put_device(sdp);
3563 	return error;
3564 }
3565 
3566 /**
3567  *	sd_remove - called whenever a scsi disk (previously recognized by
3568  *	sd_probe) is detached from the system. It is called (potentially
3569  *	multiple times) during sd module unload.
3570  *	@dev: pointer to device object
3571  *
3572  *	Note: this function is invoked from the scsi mid-level.
3573  *	This function potentially frees up a device name (e.g. /dev/sdc)
3574  *	that could be re-used by a subsequent sd_probe().
3575  *	This function is not called when the built-in sd driver is "exit-ed".
3576  **/
3577 static int sd_remove(struct device *dev)
3578 {
3579 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3580 
3581 	scsi_autopm_get_device(sdkp->device);
3582 
3583 	device_del(&sdkp->disk_dev);
3584 	del_gendisk(sdkp->disk);
3585 	sd_shutdown(dev);
3586 
3587 	put_disk(sdkp->disk);
3588 	return 0;
3589 }
3590 
3591 static void scsi_disk_release(struct device *dev)
3592 {
3593 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3594 
3595 	ida_free(&sd_index_ida, sdkp->index);
3596 	sd_zbc_free_zone_info(sdkp);
3597 	put_device(&sdkp->device->sdev_gendev);
3598 	free_opal_dev(sdkp->opal_dev);
3599 
3600 	kfree(sdkp);
3601 }
3602 
3603 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3604 {
3605 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3606 	struct scsi_sense_hdr sshdr;
3607 	struct scsi_device *sdp = sdkp->device;
3608 	int res;
3609 
3610 	if (start)
3611 		cmd[4] |= 1;	/* START */
3612 
3613 	if (sdp->start_stop_pwr_cond)
3614 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3615 
3616 	if (!scsi_device_online(sdp))
3617 		return -ENODEV;
3618 
3619 	res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, &sshdr,
3620 			SD_TIMEOUT, sdkp->max_retries, 0, RQF_PM, NULL);
3621 	if (res) {
3622 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3623 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3624 			sd_print_sense_hdr(sdkp, &sshdr);
3625 			/* 0x3a is medium not present */
3626 			if (sshdr.asc == 0x3a)
3627 				res = 0;
3628 		}
3629 	}
3630 
3631 	/* SCSI error codes must not go to the generic layer */
3632 	if (res)
3633 		return -EIO;
3634 
3635 	return 0;
3636 }
3637 
3638 /*
3639  * Send a SYNCHRONIZE CACHE instruction down to the device through
3640  * the normal SCSI command structure.  Wait for the command to
3641  * complete.
3642  */
3643 static void sd_shutdown(struct device *dev)
3644 {
3645 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3646 
3647 	if (!sdkp)
3648 		return;         /* this can happen */
3649 
3650 	if (pm_runtime_suspended(dev))
3651 		return;
3652 
3653 	if (sdkp->WCE && sdkp->media_present) {
3654 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3655 		sd_sync_cache(sdkp, NULL);
3656 	}
3657 
3658 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3659 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3660 		sd_start_stop_device(sdkp, 0);
3661 	}
3662 }
3663 
3664 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3665 {
3666 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3667 	struct scsi_sense_hdr sshdr;
3668 	int ret = 0;
3669 
3670 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3671 		return 0;
3672 
3673 	if (sdkp->WCE && sdkp->media_present) {
3674 		if (!sdkp->device->silence_suspend)
3675 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3676 		ret = sd_sync_cache(sdkp, &sshdr);
3677 
3678 		if (ret) {
3679 			/* ignore OFFLINE device */
3680 			if (ret == -ENODEV)
3681 				return 0;
3682 
3683 			if (!scsi_sense_valid(&sshdr) ||
3684 			    sshdr.sense_key != ILLEGAL_REQUEST)
3685 				return ret;
3686 
3687 			/*
3688 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3689 			 * doesn't support sync. There's not much to do and
3690 			 * suspend shouldn't fail.
3691 			 */
3692 			ret = 0;
3693 		}
3694 	}
3695 
3696 	if (sdkp->device->manage_start_stop) {
3697 		if (!sdkp->device->silence_suspend)
3698 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3699 		/* an error is not worth aborting a system sleep */
3700 		ret = sd_start_stop_device(sdkp, 0);
3701 		if (ignore_stop_errors)
3702 			ret = 0;
3703 	}
3704 
3705 	return ret;
3706 }
3707 
3708 static int sd_suspend_system(struct device *dev)
3709 {
3710 	if (pm_runtime_suspended(dev))
3711 		return 0;
3712 
3713 	return sd_suspend_common(dev, true);
3714 }
3715 
3716 static int sd_suspend_runtime(struct device *dev)
3717 {
3718 	return sd_suspend_common(dev, false);
3719 }
3720 
3721 static int sd_resume(struct device *dev)
3722 {
3723 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3724 	int ret;
3725 
3726 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3727 		return 0;
3728 
3729 	if (!sdkp->device->manage_start_stop)
3730 		return 0;
3731 
3732 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3733 	ret = sd_start_stop_device(sdkp, 1);
3734 	if (!ret)
3735 		opal_unlock_from_suspend(sdkp->opal_dev);
3736 	return ret;
3737 }
3738 
3739 static int sd_resume_system(struct device *dev)
3740 {
3741 	if (pm_runtime_suspended(dev))
3742 		return 0;
3743 
3744 	return sd_resume(dev);
3745 }
3746 
3747 static int sd_resume_runtime(struct device *dev)
3748 {
3749 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3750 	struct scsi_device *sdp;
3751 
3752 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3753 		return 0;
3754 
3755 	sdp = sdkp->device;
3756 
3757 	if (sdp->ignore_media_change) {
3758 		/* clear the device's sense data */
3759 		static const u8 cmd[10] = { REQUEST_SENSE };
3760 
3761 		if (scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL,
3762 				 NULL, sdp->request_queue->rq_timeout, 1, 0,
3763 				 RQF_PM, NULL))
3764 			sd_printk(KERN_NOTICE, sdkp,
3765 				  "Failed to clear sense data\n");
3766 	}
3767 
3768 	return sd_resume(dev);
3769 }
3770 
3771 /**
3772  *	init_sd - entry point for this driver (both when built in or when
3773  *	a module).
3774  *
3775  *	Note: this function registers this driver with the scsi mid-level.
3776  **/
3777 static int __init init_sd(void)
3778 {
3779 	int majors = 0, i, err;
3780 
3781 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3782 
3783 	for (i = 0; i < SD_MAJORS; i++) {
3784 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3785 			continue;
3786 		majors++;
3787 	}
3788 
3789 	if (!majors)
3790 		return -ENODEV;
3791 
3792 	err = class_register(&sd_disk_class);
3793 	if (err)
3794 		goto err_out;
3795 
3796 	sd_cdb_cache = kmem_cache_create("sd_ext_cdb", SD_EXT_CDB_SIZE,
3797 					 0, 0, NULL);
3798 	if (!sd_cdb_cache) {
3799 		printk(KERN_ERR "sd: can't init extended cdb cache\n");
3800 		err = -ENOMEM;
3801 		goto err_out_class;
3802 	}
3803 
3804 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3805 	if (!sd_page_pool) {
3806 		printk(KERN_ERR "sd: can't init discard page pool\n");
3807 		err = -ENOMEM;
3808 		goto err_out_cache;
3809 	}
3810 
3811 	err = scsi_register_driver(&sd_template.gendrv);
3812 	if (err)
3813 		goto err_out_driver;
3814 
3815 	return 0;
3816 
3817 err_out_driver:
3818 	mempool_destroy(sd_page_pool);
3819 
3820 err_out_cache:
3821 	kmem_cache_destroy(sd_cdb_cache);
3822 
3823 err_out_class:
3824 	class_unregister(&sd_disk_class);
3825 err_out:
3826 	for (i = 0; i < SD_MAJORS; i++)
3827 		unregister_blkdev(sd_major(i), "sd");
3828 	return err;
3829 }
3830 
3831 /**
3832  *	exit_sd - exit point for this driver (when it is a module).
3833  *
3834  *	Note: this function unregisters this driver from the scsi mid-level.
3835  **/
3836 static void __exit exit_sd(void)
3837 {
3838 	int i;
3839 
3840 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3841 
3842 	scsi_unregister_driver(&sd_template.gendrv);
3843 	mempool_destroy(sd_page_pool);
3844 	kmem_cache_destroy(sd_cdb_cache);
3845 
3846 	class_unregister(&sd_disk_class);
3847 
3848 	for (i = 0; i < SD_MAJORS; i++)
3849 		unregister_blkdev(sd_major(i), "sd");
3850 }
3851 
3852 module_init(init_sd);
3853 module_exit(exit_sd);
3854 
3855 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3856 {
3857 	scsi_print_sense_hdr(sdkp->device,
3858 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3859 }
3860 
3861 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3862 {
3863 	const char *hb_string = scsi_hostbyte_string(result);
3864 
3865 	if (hb_string)
3866 		sd_printk(KERN_INFO, sdkp,
3867 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3868 			  hb_string ? hb_string : "invalid",
3869 			  "DRIVER_OK");
3870 	else
3871 		sd_printk(KERN_INFO, sdkp,
3872 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3873 			  msg, host_byte(result), "DRIVER_OK");
3874 }
3875