xref: /openbmc/linux/drivers/scsi/sd.c (revision bbaf1ff0)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *      sd.c Copyright (C) 1992 Drew Eckhardt
4  *           Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5  *
6  *      Linux scsi disk driver
7  *              Initial versions: Drew Eckhardt
8  *              Subsequent revisions: Eric Youngdale
9  *	Modification history:
10  *       - Drew Eckhardt <drew@colorado.edu> original
11  *       - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12  *         outstanding request, and other enhancements.
13  *         Support loadable low-level scsi drivers.
14  *       - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15  *         eight major numbers.
16  *       - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17  *	 - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18  *	   sd_init and cleanups.
19  *	 - Alex Davis <letmein@erols.com> Fix problem where partition info
20  *	   not being read in sd_open. Fix problem where removable media
21  *	   could be ejected after sd_open.
22  *	 - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23  *	 - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24  *	   <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25  *	   Support 32k/1M disks.
26  *
27  *	Logging policy (needs CONFIG_SCSI_LOGGING defined):
28  *	 - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29  *	 - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30  *	 - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31  *	 - entering other commands: SCSI_LOG_HLQUEUE level 3
32  *	Note: when the logging level is set by the user, it must be greater
33  *	than the level indicated above to trigger output.
34  */
35 
36 #include <linux/module.h>
37 #include <linux/fs.h>
38 #include <linux/kernel.h>
39 #include <linux/mm.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
56 #include <linux/pr.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
60 
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 
71 #include "sd.h"
72 #include "scsi_priv.h"
73 #include "scsi_logging.h"
74 
75 MODULE_AUTHOR("Eric Youngdale");
76 MODULE_DESCRIPTION("SCSI disk (sd) driver");
77 MODULE_LICENSE("GPL");
78 
79 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
95 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
99 
100 #define SD_MINORS	16
101 
102 static void sd_config_discard(struct scsi_disk *, unsigned int);
103 static void sd_config_write_same(struct scsi_disk *);
104 static int  sd_revalidate_disk(struct gendisk *);
105 static void sd_unlock_native_capacity(struct gendisk *disk);
106 static int  sd_probe(struct device *);
107 static int  sd_remove(struct device *);
108 static void sd_shutdown(struct device *);
109 static int sd_suspend_system(struct device *);
110 static int sd_suspend_runtime(struct device *);
111 static int sd_resume_system(struct device *);
112 static int sd_resume_runtime(struct device *);
113 static void sd_rescan(struct device *);
114 static blk_status_t sd_init_command(struct scsi_cmnd *SCpnt);
115 static void sd_uninit_command(struct scsi_cmnd *SCpnt);
116 static int sd_done(struct scsi_cmnd *);
117 static void sd_eh_reset(struct scsi_cmnd *);
118 static int sd_eh_action(struct scsi_cmnd *, int);
119 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
120 static void scsi_disk_release(struct device *cdev);
121 
122 static DEFINE_IDA(sd_index_ida);
123 
124 static mempool_t *sd_page_pool;
125 static struct lock_class_key sd_bio_compl_lkclass;
126 
127 static const char *sd_cache_types[] = {
128 	"write through", "none", "write back",
129 	"write back, no read (daft)"
130 };
131 
132 static void sd_set_flush_flag(struct scsi_disk *sdkp)
133 {
134 	bool wc = false, fua = false;
135 
136 	if (sdkp->WCE) {
137 		wc = true;
138 		if (sdkp->DPOFUA)
139 			fua = true;
140 	}
141 
142 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
143 }
144 
145 static ssize_t
146 cache_type_store(struct device *dev, struct device_attribute *attr,
147 		 const char *buf, size_t count)
148 {
149 	int ct, rcd, wce, sp;
150 	struct scsi_disk *sdkp = to_scsi_disk(dev);
151 	struct scsi_device *sdp = sdkp->device;
152 	char buffer[64];
153 	char *buffer_data;
154 	struct scsi_mode_data data;
155 	struct scsi_sense_hdr sshdr;
156 	static const char temp[] = "temporary ";
157 	int len;
158 
159 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
160 		/* no cache control on RBC devices; theoretically they
161 		 * can do it, but there's probably so many exceptions
162 		 * it's not worth the risk */
163 		return -EINVAL;
164 
165 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
166 		buf += sizeof(temp) - 1;
167 		sdkp->cache_override = 1;
168 	} else {
169 		sdkp->cache_override = 0;
170 	}
171 
172 	ct = sysfs_match_string(sd_cache_types, buf);
173 	if (ct < 0)
174 		return -EINVAL;
175 
176 	rcd = ct & 0x01 ? 1 : 0;
177 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
178 
179 	if (sdkp->cache_override) {
180 		sdkp->WCE = wce;
181 		sdkp->RCD = rcd;
182 		sd_set_flush_flag(sdkp);
183 		return count;
184 	}
185 
186 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
187 			    sdkp->max_retries, &data, NULL))
188 		return -EINVAL;
189 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
190 		  data.block_descriptor_length);
191 	buffer_data = buffer + data.header_length +
192 		data.block_descriptor_length;
193 	buffer_data[2] &= ~0x05;
194 	buffer_data[2] |= wce << 2 | rcd;
195 	sp = buffer_data[0] & 0x80 ? 1 : 0;
196 	buffer_data[0] &= ~0x80;
197 
198 	/*
199 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
200 	 * received mode parameter buffer before doing MODE SELECT.
201 	 */
202 	data.device_specific = 0;
203 
204 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
205 			     sdkp->max_retries, &data, &sshdr)) {
206 		if (scsi_sense_valid(&sshdr))
207 			sd_print_sense_hdr(sdkp, &sshdr);
208 		return -EINVAL;
209 	}
210 	sd_revalidate_disk(sdkp->disk);
211 	return count;
212 }
213 
214 static ssize_t
215 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
216 		       char *buf)
217 {
218 	struct scsi_disk *sdkp = to_scsi_disk(dev);
219 	struct scsi_device *sdp = sdkp->device;
220 
221 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
222 }
223 
224 static ssize_t
225 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
226 			const char *buf, size_t count)
227 {
228 	struct scsi_disk *sdkp = to_scsi_disk(dev);
229 	struct scsi_device *sdp = sdkp->device;
230 	bool v;
231 
232 	if (!capable(CAP_SYS_ADMIN))
233 		return -EACCES;
234 
235 	if (kstrtobool(buf, &v))
236 		return -EINVAL;
237 
238 	sdp->manage_start_stop = v;
239 
240 	return count;
241 }
242 static DEVICE_ATTR_RW(manage_start_stop);
243 
244 static ssize_t
245 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
246 {
247 	struct scsi_disk *sdkp = to_scsi_disk(dev);
248 
249 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
250 }
251 
252 static ssize_t
253 allow_restart_store(struct device *dev, struct device_attribute *attr,
254 		    const char *buf, size_t count)
255 {
256 	bool v;
257 	struct scsi_disk *sdkp = to_scsi_disk(dev);
258 	struct scsi_device *sdp = sdkp->device;
259 
260 	if (!capable(CAP_SYS_ADMIN))
261 		return -EACCES;
262 
263 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
264 		return -EINVAL;
265 
266 	if (kstrtobool(buf, &v))
267 		return -EINVAL;
268 
269 	sdp->allow_restart = v;
270 
271 	return count;
272 }
273 static DEVICE_ATTR_RW(allow_restart);
274 
275 static ssize_t
276 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
277 {
278 	struct scsi_disk *sdkp = to_scsi_disk(dev);
279 	int ct = sdkp->RCD + 2*sdkp->WCE;
280 
281 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
282 }
283 static DEVICE_ATTR_RW(cache_type);
284 
285 static ssize_t
286 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
287 {
288 	struct scsi_disk *sdkp = to_scsi_disk(dev);
289 
290 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
291 }
292 static DEVICE_ATTR_RO(FUA);
293 
294 static ssize_t
295 protection_type_show(struct device *dev, struct device_attribute *attr,
296 		     char *buf)
297 {
298 	struct scsi_disk *sdkp = to_scsi_disk(dev);
299 
300 	return sprintf(buf, "%u\n", sdkp->protection_type);
301 }
302 
303 static ssize_t
304 protection_type_store(struct device *dev, struct device_attribute *attr,
305 		      const char *buf, size_t count)
306 {
307 	struct scsi_disk *sdkp = to_scsi_disk(dev);
308 	unsigned int val;
309 	int err;
310 
311 	if (!capable(CAP_SYS_ADMIN))
312 		return -EACCES;
313 
314 	err = kstrtouint(buf, 10, &val);
315 
316 	if (err)
317 		return err;
318 
319 	if (val <= T10_PI_TYPE3_PROTECTION)
320 		sdkp->protection_type = val;
321 
322 	return count;
323 }
324 static DEVICE_ATTR_RW(protection_type);
325 
326 static ssize_t
327 protection_mode_show(struct device *dev, struct device_attribute *attr,
328 		     char *buf)
329 {
330 	struct scsi_disk *sdkp = to_scsi_disk(dev);
331 	struct scsi_device *sdp = sdkp->device;
332 	unsigned int dif, dix;
333 
334 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
335 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
336 
337 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
338 		dif = 0;
339 		dix = 1;
340 	}
341 
342 	if (!dif && !dix)
343 		return sprintf(buf, "none\n");
344 
345 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
346 }
347 static DEVICE_ATTR_RO(protection_mode);
348 
349 static ssize_t
350 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
351 {
352 	struct scsi_disk *sdkp = to_scsi_disk(dev);
353 
354 	return sprintf(buf, "%u\n", sdkp->ATO);
355 }
356 static DEVICE_ATTR_RO(app_tag_own);
357 
358 static ssize_t
359 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
360 		       char *buf)
361 {
362 	struct scsi_disk *sdkp = to_scsi_disk(dev);
363 
364 	return sprintf(buf, "%u\n", sdkp->lbpme);
365 }
366 static DEVICE_ATTR_RO(thin_provisioning);
367 
368 /* sysfs_match_string() requires dense arrays */
369 static const char *lbp_mode[] = {
370 	[SD_LBP_FULL]		= "full",
371 	[SD_LBP_UNMAP]		= "unmap",
372 	[SD_LBP_WS16]		= "writesame_16",
373 	[SD_LBP_WS10]		= "writesame_10",
374 	[SD_LBP_ZERO]		= "writesame_zero",
375 	[SD_LBP_DISABLE]	= "disabled",
376 };
377 
378 static ssize_t
379 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
380 		       char *buf)
381 {
382 	struct scsi_disk *sdkp = to_scsi_disk(dev);
383 
384 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
385 }
386 
387 static ssize_t
388 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
389 			const char *buf, size_t count)
390 {
391 	struct scsi_disk *sdkp = to_scsi_disk(dev);
392 	struct scsi_device *sdp = sdkp->device;
393 	int mode;
394 
395 	if (!capable(CAP_SYS_ADMIN))
396 		return -EACCES;
397 
398 	if (sd_is_zoned(sdkp)) {
399 		sd_config_discard(sdkp, SD_LBP_DISABLE);
400 		return count;
401 	}
402 
403 	if (sdp->type != TYPE_DISK)
404 		return -EINVAL;
405 
406 	mode = sysfs_match_string(lbp_mode, buf);
407 	if (mode < 0)
408 		return -EINVAL;
409 
410 	sd_config_discard(sdkp, mode);
411 
412 	return count;
413 }
414 static DEVICE_ATTR_RW(provisioning_mode);
415 
416 /* sysfs_match_string() requires dense arrays */
417 static const char *zeroing_mode[] = {
418 	[SD_ZERO_WRITE]		= "write",
419 	[SD_ZERO_WS]		= "writesame",
420 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
421 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
422 };
423 
424 static ssize_t
425 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
426 		  char *buf)
427 {
428 	struct scsi_disk *sdkp = to_scsi_disk(dev);
429 
430 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
431 }
432 
433 static ssize_t
434 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
435 		   const char *buf, size_t count)
436 {
437 	struct scsi_disk *sdkp = to_scsi_disk(dev);
438 	int mode;
439 
440 	if (!capable(CAP_SYS_ADMIN))
441 		return -EACCES;
442 
443 	mode = sysfs_match_string(zeroing_mode, buf);
444 	if (mode < 0)
445 		return -EINVAL;
446 
447 	sdkp->zeroing_mode = mode;
448 
449 	return count;
450 }
451 static DEVICE_ATTR_RW(zeroing_mode);
452 
453 static ssize_t
454 max_medium_access_timeouts_show(struct device *dev,
455 				struct device_attribute *attr, char *buf)
456 {
457 	struct scsi_disk *sdkp = to_scsi_disk(dev);
458 
459 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
460 }
461 
462 static ssize_t
463 max_medium_access_timeouts_store(struct device *dev,
464 				 struct device_attribute *attr, const char *buf,
465 				 size_t count)
466 {
467 	struct scsi_disk *sdkp = to_scsi_disk(dev);
468 	int err;
469 
470 	if (!capable(CAP_SYS_ADMIN))
471 		return -EACCES;
472 
473 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
474 
475 	return err ? err : count;
476 }
477 static DEVICE_ATTR_RW(max_medium_access_timeouts);
478 
479 static ssize_t
480 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
481 			   char *buf)
482 {
483 	struct scsi_disk *sdkp = to_scsi_disk(dev);
484 
485 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
486 }
487 
488 static ssize_t
489 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
490 			    const char *buf, size_t count)
491 {
492 	struct scsi_disk *sdkp = to_scsi_disk(dev);
493 	struct scsi_device *sdp = sdkp->device;
494 	unsigned long max;
495 	int err;
496 
497 	if (!capable(CAP_SYS_ADMIN))
498 		return -EACCES;
499 
500 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
501 		return -EINVAL;
502 
503 	err = kstrtoul(buf, 10, &max);
504 
505 	if (err)
506 		return err;
507 
508 	if (max == 0)
509 		sdp->no_write_same = 1;
510 	else if (max <= SD_MAX_WS16_BLOCKS) {
511 		sdp->no_write_same = 0;
512 		sdkp->max_ws_blocks = max;
513 	}
514 
515 	sd_config_write_same(sdkp);
516 
517 	return count;
518 }
519 static DEVICE_ATTR_RW(max_write_same_blocks);
520 
521 static ssize_t
522 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
523 {
524 	struct scsi_disk *sdkp = to_scsi_disk(dev);
525 
526 	if (sdkp->device->type == TYPE_ZBC)
527 		return sprintf(buf, "host-managed\n");
528 	if (sdkp->zoned == 1)
529 		return sprintf(buf, "host-aware\n");
530 	if (sdkp->zoned == 2)
531 		return sprintf(buf, "drive-managed\n");
532 	return sprintf(buf, "none\n");
533 }
534 static DEVICE_ATTR_RO(zoned_cap);
535 
536 static ssize_t
537 max_retries_store(struct device *dev, struct device_attribute *attr,
538 		  const char *buf, size_t count)
539 {
540 	struct scsi_disk *sdkp = to_scsi_disk(dev);
541 	struct scsi_device *sdev = sdkp->device;
542 	int retries, err;
543 
544 	err = kstrtoint(buf, 10, &retries);
545 	if (err)
546 		return err;
547 
548 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
549 		sdkp->max_retries = retries;
550 		return count;
551 	}
552 
553 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
554 		    SD_MAX_RETRIES);
555 	return -EINVAL;
556 }
557 
558 static ssize_t
559 max_retries_show(struct device *dev, struct device_attribute *attr,
560 		 char *buf)
561 {
562 	struct scsi_disk *sdkp = to_scsi_disk(dev);
563 
564 	return sprintf(buf, "%d\n", sdkp->max_retries);
565 }
566 
567 static DEVICE_ATTR_RW(max_retries);
568 
569 static struct attribute *sd_disk_attrs[] = {
570 	&dev_attr_cache_type.attr,
571 	&dev_attr_FUA.attr,
572 	&dev_attr_allow_restart.attr,
573 	&dev_attr_manage_start_stop.attr,
574 	&dev_attr_protection_type.attr,
575 	&dev_attr_protection_mode.attr,
576 	&dev_attr_app_tag_own.attr,
577 	&dev_attr_thin_provisioning.attr,
578 	&dev_attr_provisioning_mode.attr,
579 	&dev_attr_zeroing_mode.attr,
580 	&dev_attr_max_write_same_blocks.attr,
581 	&dev_attr_max_medium_access_timeouts.attr,
582 	&dev_attr_zoned_cap.attr,
583 	&dev_attr_max_retries.attr,
584 	NULL,
585 };
586 ATTRIBUTE_GROUPS(sd_disk);
587 
588 static struct class sd_disk_class = {
589 	.name		= "scsi_disk",
590 	.dev_release	= scsi_disk_release,
591 	.dev_groups	= sd_disk_groups,
592 };
593 
594 static const struct dev_pm_ops sd_pm_ops = {
595 	.suspend		= sd_suspend_system,
596 	.resume			= sd_resume_system,
597 	.poweroff		= sd_suspend_system,
598 	.restore		= sd_resume_system,
599 	.runtime_suspend	= sd_suspend_runtime,
600 	.runtime_resume		= sd_resume_runtime,
601 };
602 
603 static struct scsi_driver sd_template = {
604 	.gendrv = {
605 		.name		= "sd",
606 		.owner		= THIS_MODULE,
607 		.probe		= sd_probe,
608 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
609 		.remove		= sd_remove,
610 		.shutdown	= sd_shutdown,
611 		.pm		= &sd_pm_ops,
612 	},
613 	.rescan			= sd_rescan,
614 	.init_command		= sd_init_command,
615 	.uninit_command		= sd_uninit_command,
616 	.done			= sd_done,
617 	.eh_action		= sd_eh_action,
618 	.eh_reset		= sd_eh_reset,
619 };
620 
621 /*
622  * Don't request a new module, as that could deadlock in multipath
623  * environment.
624  */
625 static void sd_default_probe(dev_t devt)
626 {
627 }
628 
629 /*
630  * Device no to disk mapping:
631  *
632  *       major         disc2     disc  p1
633  *   |............|.............|....|....| <- dev_t
634  *    31        20 19          8 7  4 3  0
635  *
636  * Inside a major, we have 16k disks, however mapped non-
637  * contiguously. The first 16 disks are for major0, the next
638  * ones with major1, ... Disk 256 is for major0 again, disk 272
639  * for major1, ...
640  * As we stay compatible with our numbering scheme, we can reuse
641  * the well-know SCSI majors 8, 65--71, 136--143.
642  */
643 static int sd_major(int major_idx)
644 {
645 	switch (major_idx) {
646 	case 0:
647 		return SCSI_DISK0_MAJOR;
648 	case 1 ... 7:
649 		return SCSI_DISK1_MAJOR + major_idx - 1;
650 	case 8 ... 15:
651 		return SCSI_DISK8_MAJOR + major_idx - 8;
652 	default:
653 		BUG();
654 		return 0;	/* shut up gcc */
655 	}
656 }
657 
658 #ifdef CONFIG_BLK_SED_OPAL
659 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
660 		size_t len, bool send)
661 {
662 	struct scsi_disk *sdkp = data;
663 	struct scsi_device *sdev = sdkp->device;
664 	u8 cdb[12] = { 0, };
665 	const struct scsi_exec_args exec_args = {
666 		.req_flags = BLK_MQ_REQ_PM,
667 	};
668 	int ret;
669 
670 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
671 	cdb[1] = secp;
672 	put_unaligned_be16(spsp, &cdb[2]);
673 	put_unaligned_be32(len, &cdb[6]);
674 
675 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
676 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
677 			       &exec_args);
678 	return ret <= 0 ? ret : -EIO;
679 }
680 #endif /* CONFIG_BLK_SED_OPAL */
681 
682 /*
683  * Look up the DIX operation based on whether the command is read or
684  * write and whether dix and dif are enabled.
685  */
686 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
687 {
688 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
689 	static const unsigned int ops[] = {	/* wrt dix dif */
690 		SCSI_PROT_NORMAL,		/*  0	0   0  */
691 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
692 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
693 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
694 		SCSI_PROT_NORMAL,		/*  1	0   0  */
695 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
696 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
697 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
698 	};
699 
700 	return ops[write << 2 | dix << 1 | dif];
701 }
702 
703 /*
704  * Returns a mask of the protection flags that are valid for a given DIX
705  * operation.
706  */
707 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
708 {
709 	static const unsigned int flag_mask[] = {
710 		[SCSI_PROT_NORMAL]		= 0,
711 
712 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
713 						  SCSI_PROT_GUARD_CHECK |
714 						  SCSI_PROT_REF_CHECK |
715 						  SCSI_PROT_REF_INCREMENT,
716 
717 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
718 						  SCSI_PROT_IP_CHECKSUM,
719 
720 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
721 						  SCSI_PROT_GUARD_CHECK |
722 						  SCSI_PROT_REF_CHECK |
723 						  SCSI_PROT_REF_INCREMENT |
724 						  SCSI_PROT_IP_CHECKSUM,
725 
726 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
727 						  SCSI_PROT_REF_INCREMENT,
728 
729 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
730 						  SCSI_PROT_REF_CHECK |
731 						  SCSI_PROT_REF_INCREMENT |
732 						  SCSI_PROT_IP_CHECKSUM,
733 
734 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
735 						  SCSI_PROT_GUARD_CHECK |
736 						  SCSI_PROT_REF_CHECK |
737 						  SCSI_PROT_REF_INCREMENT |
738 						  SCSI_PROT_IP_CHECKSUM,
739 	};
740 
741 	return flag_mask[prot_op];
742 }
743 
744 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
745 					   unsigned int dix, unsigned int dif)
746 {
747 	struct request *rq = scsi_cmd_to_rq(scmd);
748 	struct bio *bio = rq->bio;
749 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
750 	unsigned int protect = 0;
751 
752 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
753 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
754 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
755 
756 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
757 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
758 	}
759 
760 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
761 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
762 
763 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
764 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
765 	}
766 
767 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
768 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
769 
770 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
771 			protect = 3 << 5;	/* Disable target PI checking */
772 		else
773 			protect = 1 << 5;	/* Enable target PI checking */
774 	}
775 
776 	scsi_set_prot_op(scmd, prot_op);
777 	scsi_set_prot_type(scmd, dif);
778 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
779 
780 	return protect;
781 }
782 
783 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
784 {
785 	struct request_queue *q = sdkp->disk->queue;
786 	unsigned int logical_block_size = sdkp->device->sector_size;
787 	unsigned int max_blocks = 0;
788 
789 	q->limits.discard_alignment =
790 		sdkp->unmap_alignment * logical_block_size;
791 	q->limits.discard_granularity =
792 		max(sdkp->physical_block_size,
793 		    sdkp->unmap_granularity * logical_block_size);
794 	sdkp->provisioning_mode = mode;
795 
796 	switch (mode) {
797 
798 	case SD_LBP_FULL:
799 	case SD_LBP_DISABLE:
800 		blk_queue_max_discard_sectors(q, 0);
801 		return;
802 
803 	case SD_LBP_UNMAP:
804 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
805 					  (u32)SD_MAX_WS16_BLOCKS);
806 		break;
807 
808 	case SD_LBP_WS16:
809 		if (sdkp->device->unmap_limit_for_ws)
810 			max_blocks = sdkp->max_unmap_blocks;
811 		else
812 			max_blocks = sdkp->max_ws_blocks;
813 
814 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
815 		break;
816 
817 	case SD_LBP_WS10:
818 		if (sdkp->device->unmap_limit_for_ws)
819 			max_blocks = sdkp->max_unmap_blocks;
820 		else
821 			max_blocks = sdkp->max_ws_blocks;
822 
823 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
824 		break;
825 
826 	case SD_LBP_ZERO:
827 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
828 					  (u32)SD_MAX_WS10_BLOCKS);
829 		break;
830 	}
831 
832 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
833 }
834 
835 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
836 {
837 	struct page *page;
838 
839 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
840 	if (!page)
841 		return NULL;
842 	clear_highpage(page);
843 	bvec_set_page(&rq->special_vec, page, data_len, 0);
844 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
845 	return bvec_virt(&rq->special_vec);
846 }
847 
848 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
849 {
850 	struct scsi_device *sdp = cmd->device;
851 	struct request *rq = scsi_cmd_to_rq(cmd);
852 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
853 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
854 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
855 	unsigned int data_len = 24;
856 	char *buf;
857 
858 	buf = sd_set_special_bvec(rq, data_len);
859 	if (!buf)
860 		return BLK_STS_RESOURCE;
861 
862 	cmd->cmd_len = 10;
863 	cmd->cmnd[0] = UNMAP;
864 	cmd->cmnd[8] = 24;
865 
866 	put_unaligned_be16(6 + 16, &buf[0]);
867 	put_unaligned_be16(16, &buf[2]);
868 	put_unaligned_be64(lba, &buf[8]);
869 	put_unaligned_be32(nr_blocks, &buf[16]);
870 
871 	cmd->allowed = sdkp->max_retries;
872 	cmd->transfersize = data_len;
873 	rq->timeout = SD_TIMEOUT;
874 
875 	return scsi_alloc_sgtables(cmd);
876 }
877 
878 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
879 		bool unmap)
880 {
881 	struct scsi_device *sdp = cmd->device;
882 	struct request *rq = scsi_cmd_to_rq(cmd);
883 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
884 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
885 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
886 	u32 data_len = sdp->sector_size;
887 
888 	if (!sd_set_special_bvec(rq, data_len))
889 		return BLK_STS_RESOURCE;
890 
891 	cmd->cmd_len = 16;
892 	cmd->cmnd[0] = WRITE_SAME_16;
893 	if (unmap)
894 		cmd->cmnd[1] = 0x8; /* UNMAP */
895 	put_unaligned_be64(lba, &cmd->cmnd[2]);
896 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
897 
898 	cmd->allowed = sdkp->max_retries;
899 	cmd->transfersize = data_len;
900 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
901 
902 	return scsi_alloc_sgtables(cmd);
903 }
904 
905 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
906 		bool unmap)
907 {
908 	struct scsi_device *sdp = cmd->device;
909 	struct request *rq = scsi_cmd_to_rq(cmd);
910 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
911 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
912 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
913 	u32 data_len = sdp->sector_size;
914 
915 	if (!sd_set_special_bvec(rq, data_len))
916 		return BLK_STS_RESOURCE;
917 
918 	cmd->cmd_len = 10;
919 	cmd->cmnd[0] = WRITE_SAME;
920 	if (unmap)
921 		cmd->cmnd[1] = 0x8; /* UNMAP */
922 	put_unaligned_be32(lba, &cmd->cmnd[2]);
923 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
924 
925 	cmd->allowed = sdkp->max_retries;
926 	cmd->transfersize = data_len;
927 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
928 
929 	return scsi_alloc_sgtables(cmd);
930 }
931 
932 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
933 {
934 	struct request *rq = scsi_cmd_to_rq(cmd);
935 	struct scsi_device *sdp = cmd->device;
936 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
937 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
938 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
939 
940 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
941 		switch (sdkp->zeroing_mode) {
942 		case SD_ZERO_WS16_UNMAP:
943 			return sd_setup_write_same16_cmnd(cmd, true);
944 		case SD_ZERO_WS10_UNMAP:
945 			return sd_setup_write_same10_cmnd(cmd, true);
946 		}
947 	}
948 
949 	if (sdp->no_write_same) {
950 		rq->rq_flags |= RQF_QUIET;
951 		return BLK_STS_TARGET;
952 	}
953 
954 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
955 		return sd_setup_write_same16_cmnd(cmd, false);
956 
957 	return sd_setup_write_same10_cmnd(cmd, false);
958 }
959 
960 static void sd_config_write_same(struct scsi_disk *sdkp)
961 {
962 	struct request_queue *q = sdkp->disk->queue;
963 	unsigned int logical_block_size = sdkp->device->sector_size;
964 
965 	if (sdkp->device->no_write_same) {
966 		sdkp->max_ws_blocks = 0;
967 		goto out;
968 	}
969 
970 	/* Some devices can not handle block counts above 0xffff despite
971 	 * supporting WRITE SAME(16). Consequently we default to 64k
972 	 * blocks per I/O unless the device explicitly advertises a
973 	 * bigger limit.
974 	 */
975 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
976 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
977 						   (u32)SD_MAX_WS16_BLOCKS);
978 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
979 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
980 						   (u32)SD_MAX_WS10_BLOCKS);
981 	else {
982 		sdkp->device->no_write_same = 1;
983 		sdkp->max_ws_blocks = 0;
984 	}
985 
986 	if (sdkp->lbprz && sdkp->lbpws)
987 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
988 	else if (sdkp->lbprz && sdkp->lbpws10)
989 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
990 	else if (sdkp->max_ws_blocks)
991 		sdkp->zeroing_mode = SD_ZERO_WS;
992 	else
993 		sdkp->zeroing_mode = SD_ZERO_WRITE;
994 
995 	if (sdkp->max_ws_blocks &&
996 	    sdkp->physical_block_size > logical_block_size) {
997 		/*
998 		 * Reporting a maximum number of blocks that is not aligned
999 		 * on the device physical size would cause a large write same
1000 		 * request to be split into physically unaligned chunks by
1001 		 * __blkdev_issue_write_zeroes() even if the caller of this
1002 		 * functions took care to align the large request. So make sure
1003 		 * the maximum reported is aligned to the device physical block
1004 		 * size. This is only an optional optimization for regular
1005 		 * disks, but this is mandatory to avoid failure of large write
1006 		 * same requests directed at sequential write required zones of
1007 		 * host-managed ZBC disks.
1008 		 */
1009 		sdkp->max_ws_blocks =
1010 			round_down(sdkp->max_ws_blocks,
1011 				   bytes_to_logical(sdkp->device,
1012 						    sdkp->physical_block_size));
1013 	}
1014 
1015 out:
1016 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1017 					 (logical_block_size >> 9));
1018 }
1019 
1020 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1021 {
1022 	struct request *rq = scsi_cmd_to_rq(cmd);
1023 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1024 
1025 	/* flush requests don't perform I/O, zero the S/G table */
1026 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1027 
1028 	if (cmd->device->use_16_for_sync) {
1029 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1030 		cmd->cmd_len = 16;
1031 	} else {
1032 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1033 		cmd->cmd_len = 10;
1034 	}
1035 	cmd->transfersize = 0;
1036 	cmd->allowed = sdkp->max_retries;
1037 
1038 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1039 	return BLK_STS_OK;
1040 }
1041 
1042 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1043 				       sector_t lba, unsigned int nr_blocks,
1044 				       unsigned char flags)
1045 {
1046 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1047 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1048 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1049 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1050 	cmd->cmnd[10] = flags;
1051 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1052 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1053 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1054 
1055 	return BLK_STS_OK;
1056 }
1057 
1058 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1059 				       sector_t lba, unsigned int nr_blocks,
1060 				       unsigned char flags)
1061 {
1062 	cmd->cmd_len  = 16;
1063 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1064 	cmd->cmnd[1]  = flags;
1065 	cmd->cmnd[14] = 0;
1066 	cmd->cmnd[15] = 0;
1067 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1068 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1069 
1070 	return BLK_STS_OK;
1071 }
1072 
1073 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1074 				       sector_t lba, unsigned int nr_blocks,
1075 				       unsigned char flags)
1076 {
1077 	cmd->cmd_len = 10;
1078 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1079 	cmd->cmnd[1] = flags;
1080 	cmd->cmnd[6] = 0;
1081 	cmd->cmnd[9] = 0;
1082 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1083 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1084 
1085 	return BLK_STS_OK;
1086 }
1087 
1088 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1089 				      sector_t lba, unsigned int nr_blocks,
1090 				      unsigned char flags)
1091 {
1092 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1093 	if (WARN_ON_ONCE(nr_blocks == 0))
1094 		return BLK_STS_IOERR;
1095 
1096 	if (unlikely(flags & 0x8)) {
1097 		/*
1098 		 * This happens only if this drive failed 10byte rw
1099 		 * command with ILLEGAL_REQUEST during operation and
1100 		 * thus turned off use_10_for_rw.
1101 		 */
1102 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1103 		return BLK_STS_IOERR;
1104 	}
1105 
1106 	cmd->cmd_len = 6;
1107 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1108 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1109 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1110 	cmd->cmnd[3] = lba & 0xff;
1111 	cmd->cmnd[4] = nr_blocks;
1112 	cmd->cmnd[5] = 0;
1113 
1114 	return BLK_STS_OK;
1115 }
1116 
1117 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1118 {
1119 	struct request *rq = scsi_cmd_to_rq(cmd);
1120 	struct scsi_device *sdp = cmd->device;
1121 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1122 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1123 	sector_t threshold;
1124 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1125 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1126 	bool write = rq_data_dir(rq) == WRITE;
1127 	unsigned char protect, fua;
1128 	blk_status_t ret;
1129 	unsigned int dif;
1130 	bool dix;
1131 
1132 	ret = scsi_alloc_sgtables(cmd);
1133 	if (ret != BLK_STS_OK)
1134 		return ret;
1135 
1136 	ret = BLK_STS_IOERR;
1137 	if (!scsi_device_online(sdp) || sdp->changed) {
1138 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1139 		goto fail;
1140 	}
1141 
1142 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1143 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1144 		goto fail;
1145 	}
1146 
1147 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1148 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1149 		goto fail;
1150 	}
1151 
1152 	/*
1153 	 * Some SD card readers can't handle accesses which touch the
1154 	 * last one or two logical blocks. Split accesses as needed.
1155 	 */
1156 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1157 
1158 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1159 		if (lba < threshold) {
1160 			/* Access up to the threshold but not beyond */
1161 			nr_blocks = threshold - lba;
1162 		} else {
1163 			/* Access only a single logical block */
1164 			nr_blocks = 1;
1165 		}
1166 	}
1167 
1168 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1169 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1170 		if (ret)
1171 			goto fail;
1172 	}
1173 
1174 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1175 	dix = scsi_prot_sg_count(cmd);
1176 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1177 
1178 	if (dif || dix)
1179 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1180 	else
1181 		protect = 0;
1182 
1183 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1184 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1185 					 protect | fua);
1186 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1187 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1188 					 protect | fua);
1189 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1190 		   sdp->use_10_for_rw || protect) {
1191 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1192 					 protect | fua);
1193 	} else {
1194 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1195 					protect | fua);
1196 	}
1197 
1198 	if (unlikely(ret != BLK_STS_OK))
1199 		goto fail;
1200 
1201 	/*
1202 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1203 	 * host adapter, it's safe to assume that we can at least transfer
1204 	 * this many bytes between each connect / disconnect.
1205 	 */
1206 	cmd->transfersize = sdp->sector_size;
1207 	cmd->underflow = nr_blocks << 9;
1208 	cmd->allowed = sdkp->max_retries;
1209 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1210 
1211 	SCSI_LOG_HLQUEUE(1,
1212 			 scmd_printk(KERN_INFO, cmd,
1213 				     "%s: block=%llu, count=%d\n", __func__,
1214 				     (unsigned long long)blk_rq_pos(rq),
1215 				     blk_rq_sectors(rq)));
1216 	SCSI_LOG_HLQUEUE(2,
1217 			 scmd_printk(KERN_INFO, cmd,
1218 				     "%s %d/%u 512 byte blocks.\n",
1219 				     write ? "writing" : "reading", nr_blocks,
1220 				     blk_rq_sectors(rq)));
1221 
1222 	/*
1223 	 * This indicates that the command is ready from our end to be queued.
1224 	 */
1225 	return BLK_STS_OK;
1226 fail:
1227 	scsi_free_sgtables(cmd);
1228 	return ret;
1229 }
1230 
1231 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1232 {
1233 	struct request *rq = scsi_cmd_to_rq(cmd);
1234 
1235 	switch (req_op(rq)) {
1236 	case REQ_OP_DISCARD:
1237 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1238 		case SD_LBP_UNMAP:
1239 			return sd_setup_unmap_cmnd(cmd);
1240 		case SD_LBP_WS16:
1241 			return sd_setup_write_same16_cmnd(cmd, true);
1242 		case SD_LBP_WS10:
1243 			return sd_setup_write_same10_cmnd(cmd, true);
1244 		case SD_LBP_ZERO:
1245 			return sd_setup_write_same10_cmnd(cmd, false);
1246 		default:
1247 			return BLK_STS_TARGET;
1248 		}
1249 	case REQ_OP_WRITE_ZEROES:
1250 		return sd_setup_write_zeroes_cmnd(cmd);
1251 	case REQ_OP_FLUSH:
1252 		return sd_setup_flush_cmnd(cmd);
1253 	case REQ_OP_READ:
1254 	case REQ_OP_WRITE:
1255 	case REQ_OP_ZONE_APPEND:
1256 		return sd_setup_read_write_cmnd(cmd);
1257 	case REQ_OP_ZONE_RESET:
1258 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1259 						   false);
1260 	case REQ_OP_ZONE_RESET_ALL:
1261 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1262 						   true);
1263 	case REQ_OP_ZONE_OPEN:
1264 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1265 	case REQ_OP_ZONE_CLOSE:
1266 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1267 	case REQ_OP_ZONE_FINISH:
1268 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1269 	default:
1270 		WARN_ON_ONCE(1);
1271 		return BLK_STS_NOTSUPP;
1272 	}
1273 }
1274 
1275 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1276 {
1277 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1278 
1279 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1280 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1281 }
1282 
1283 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1284 {
1285 	if (sdkp->device->removable || sdkp->write_prot) {
1286 		if (disk_check_media_change(disk))
1287 			return true;
1288 	}
1289 
1290 	/*
1291 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1292 	 * nothing to do with partitions, BLKRRPART is used to force a full
1293 	 * revalidate after things like a format for historical reasons.
1294 	 */
1295 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1296 }
1297 
1298 /**
1299  *	sd_open - open a scsi disk device
1300  *	@disk: disk to open
1301  *	@mode: open mode
1302  *
1303  *	Returns 0 if successful. Returns a negated errno value in case
1304  *	of error.
1305  *
1306  *	Note: This can be called from a user context (e.g. fsck(1) )
1307  *	or from within the kernel (e.g. as a result of a mount(1) ).
1308  *	In the latter case @inode and @filp carry an abridged amount
1309  *	of information as noted above.
1310  *
1311  *	Locking: called with disk->open_mutex held.
1312  **/
1313 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1314 {
1315 	struct scsi_disk *sdkp = scsi_disk(disk);
1316 	struct scsi_device *sdev = sdkp->device;
1317 	int retval;
1318 
1319 	if (scsi_device_get(sdev))
1320 		return -ENXIO;
1321 
1322 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1323 
1324 	/*
1325 	 * If the device is in error recovery, wait until it is done.
1326 	 * If the device is offline, then disallow any access to it.
1327 	 */
1328 	retval = -ENXIO;
1329 	if (!scsi_block_when_processing_errors(sdev))
1330 		goto error_out;
1331 
1332 	if (sd_need_revalidate(disk, sdkp))
1333 		sd_revalidate_disk(disk);
1334 
1335 	/*
1336 	 * If the drive is empty, just let the open fail.
1337 	 */
1338 	retval = -ENOMEDIUM;
1339 	if (sdev->removable && !sdkp->media_present &&
1340 	    !(mode & BLK_OPEN_NDELAY))
1341 		goto error_out;
1342 
1343 	/*
1344 	 * If the device has the write protect tab set, have the open fail
1345 	 * if the user expects to be able to write to the thing.
1346 	 */
1347 	retval = -EROFS;
1348 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1349 		goto error_out;
1350 
1351 	/*
1352 	 * It is possible that the disk changing stuff resulted in
1353 	 * the device being taken offline.  If this is the case,
1354 	 * report this to the user, and don't pretend that the
1355 	 * open actually succeeded.
1356 	 */
1357 	retval = -ENXIO;
1358 	if (!scsi_device_online(sdev))
1359 		goto error_out;
1360 
1361 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1362 		if (scsi_block_when_processing_errors(sdev))
1363 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1364 	}
1365 
1366 	return 0;
1367 
1368 error_out:
1369 	scsi_device_put(sdev);
1370 	return retval;
1371 }
1372 
1373 /**
1374  *	sd_release - invoked when the (last) close(2) is called on this
1375  *	scsi disk.
1376  *	@disk: disk to release
1377  *
1378  *	Returns 0.
1379  *
1380  *	Note: may block (uninterruptible) if error recovery is underway
1381  *	on this disk.
1382  *
1383  *	Locking: called with disk->open_mutex held.
1384  **/
1385 static void sd_release(struct gendisk *disk)
1386 {
1387 	struct scsi_disk *sdkp = scsi_disk(disk);
1388 	struct scsi_device *sdev = sdkp->device;
1389 
1390 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1391 
1392 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1393 		if (scsi_block_when_processing_errors(sdev))
1394 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1395 	}
1396 
1397 	scsi_device_put(sdev);
1398 }
1399 
1400 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1401 {
1402 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1403 	struct scsi_device *sdp = sdkp->device;
1404 	struct Scsi_Host *host = sdp->host;
1405 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1406 	int diskinfo[4];
1407 
1408 	/* default to most commonly used values */
1409 	diskinfo[0] = 0x40;	/* 1 << 6 */
1410 	diskinfo[1] = 0x20;	/* 1 << 5 */
1411 	diskinfo[2] = capacity >> 11;
1412 
1413 	/* override with calculated, extended default, or driver values */
1414 	if (host->hostt->bios_param)
1415 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1416 	else
1417 		scsicam_bios_param(bdev, capacity, diskinfo);
1418 
1419 	geo->heads = diskinfo[0];
1420 	geo->sectors = diskinfo[1];
1421 	geo->cylinders = diskinfo[2];
1422 	return 0;
1423 }
1424 
1425 /**
1426  *	sd_ioctl - process an ioctl
1427  *	@bdev: target block device
1428  *	@mode: open mode
1429  *	@cmd: ioctl command number
1430  *	@arg: this is third argument given to ioctl(2) system call.
1431  *	Often contains a pointer.
1432  *
1433  *	Returns 0 if successful (some ioctls return positive numbers on
1434  *	success as well). Returns a negated errno value in case of error.
1435  *
1436  *	Note: most ioctls are forward onto the block subsystem or further
1437  *	down in the scsi subsystem.
1438  **/
1439 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1440 		    unsigned int cmd, unsigned long arg)
1441 {
1442 	struct gendisk *disk = bdev->bd_disk;
1443 	struct scsi_disk *sdkp = scsi_disk(disk);
1444 	struct scsi_device *sdp = sdkp->device;
1445 	void __user *p = (void __user *)arg;
1446 	int error;
1447 
1448 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1449 				    "cmd=0x%x\n", disk->disk_name, cmd));
1450 
1451 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1452 		return -ENOIOCTLCMD;
1453 
1454 	/*
1455 	 * If we are in the middle of error recovery, don't let anyone
1456 	 * else try and use this device.  Also, if error recovery fails, it
1457 	 * may try and take the device offline, in which case all further
1458 	 * access to the device is prohibited.
1459 	 */
1460 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1461 			(mode & BLK_OPEN_NDELAY));
1462 	if (error)
1463 		return error;
1464 
1465 	if (is_sed_ioctl(cmd))
1466 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1467 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1468 }
1469 
1470 static void set_media_not_present(struct scsi_disk *sdkp)
1471 {
1472 	if (sdkp->media_present)
1473 		sdkp->device->changed = 1;
1474 
1475 	if (sdkp->device->removable) {
1476 		sdkp->media_present = 0;
1477 		sdkp->capacity = 0;
1478 	}
1479 }
1480 
1481 static int media_not_present(struct scsi_disk *sdkp,
1482 			     struct scsi_sense_hdr *sshdr)
1483 {
1484 	if (!scsi_sense_valid(sshdr))
1485 		return 0;
1486 
1487 	/* not invoked for commands that could return deferred errors */
1488 	switch (sshdr->sense_key) {
1489 	case UNIT_ATTENTION:
1490 	case NOT_READY:
1491 		/* medium not present */
1492 		if (sshdr->asc == 0x3A) {
1493 			set_media_not_present(sdkp);
1494 			return 1;
1495 		}
1496 	}
1497 	return 0;
1498 }
1499 
1500 /**
1501  *	sd_check_events - check media events
1502  *	@disk: kernel device descriptor
1503  *	@clearing: disk events currently being cleared
1504  *
1505  *	Returns mask of DISK_EVENT_*.
1506  *
1507  *	Note: this function is invoked from the block subsystem.
1508  **/
1509 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1510 {
1511 	struct scsi_disk *sdkp = disk->private_data;
1512 	struct scsi_device *sdp;
1513 	int retval;
1514 	bool disk_changed;
1515 
1516 	if (!sdkp)
1517 		return 0;
1518 
1519 	sdp = sdkp->device;
1520 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1521 
1522 	/*
1523 	 * If the device is offline, don't send any commands - just pretend as
1524 	 * if the command failed.  If the device ever comes back online, we
1525 	 * can deal with it then.  It is only because of unrecoverable errors
1526 	 * that we would ever take a device offline in the first place.
1527 	 */
1528 	if (!scsi_device_online(sdp)) {
1529 		set_media_not_present(sdkp);
1530 		goto out;
1531 	}
1532 
1533 	/*
1534 	 * Using TEST_UNIT_READY enables differentiation between drive with
1535 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1536 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1537 	 *
1538 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1539 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1540 	 * sd_revalidate() is called.
1541 	 */
1542 	if (scsi_block_when_processing_errors(sdp)) {
1543 		struct scsi_sense_hdr sshdr = { 0, };
1544 
1545 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1546 					      &sshdr);
1547 
1548 		/* failed to execute TUR, assume media not present */
1549 		if (retval < 0 || host_byte(retval)) {
1550 			set_media_not_present(sdkp);
1551 			goto out;
1552 		}
1553 
1554 		if (media_not_present(sdkp, &sshdr))
1555 			goto out;
1556 	}
1557 
1558 	/*
1559 	 * For removable scsi disk we have to recognise the presence
1560 	 * of a disk in the drive.
1561 	 */
1562 	if (!sdkp->media_present)
1563 		sdp->changed = 1;
1564 	sdkp->media_present = 1;
1565 out:
1566 	/*
1567 	 * sdp->changed is set under the following conditions:
1568 	 *
1569 	 *	Medium present state has changed in either direction.
1570 	 *	Device has indicated UNIT_ATTENTION.
1571 	 */
1572 	disk_changed = sdp->changed;
1573 	sdp->changed = 0;
1574 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1575 }
1576 
1577 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1578 {
1579 	int retries, res;
1580 	struct scsi_device *sdp = sdkp->device;
1581 	const int timeout = sdp->request_queue->rq_timeout
1582 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1583 	struct scsi_sense_hdr my_sshdr;
1584 	const struct scsi_exec_args exec_args = {
1585 		.req_flags = BLK_MQ_REQ_PM,
1586 		/* caller might not be interested in sense, but we need it */
1587 		.sshdr = sshdr ? : &my_sshdr,
1588 	};
1589 
1590 	if (!scsi_device_online(sdp))
1591 		return -ENODEV;
1592 
1593 	sshdr = exec_args.sshdr;
1594 
1595 	for (retries = 3; retries > 0; --retries) {
1596 		unsigned char cmd[16] = { 0 };
1597 
1598 		if (sdp->use_16_for_sync)
1599 			cmd[0] = SYNCHRONIZE_CACHE_16;
1600 		else
1601 			cmd[0] = SYNCHRONIZE_CACHE;
1602 		/*
1603 		 * Leave the rest of the command zero to indicate
1604 		 * flush everything.
1605 		 */
1606 		res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1607 				       timeout, sdkp->max_retries, &exec_args);
1608 		if (res == 0)
1609 			break;
1610 	}
1611 
1612 	if (res) {
1613 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1614 
1615 		if (res < 0)
1616 			return res;
1617 
1618 		if (scsi_status_is_check_condition(res) &&
1619 		    scsi_sense_valid(sshdr)) {
1620 			sd_print_sense_hdr(sdkp, sshdr);
1621 
1622 			/* we need to evaluate the error return  */
1623 			if (sshdr->asc == 0x3a ||	/* medium not present */
1624 			    sshdr->asc == 0x20 ||	/* invalid command */
1625 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1626 				/* this is no error here */
1627 				return 0;
1628 		}
1629 
1630 		switch (host_byte(res)) {
1631 		/* ignore errors due to racing a disconnection */
1632 		case DID_BAD_TARGET:
1633 		case DID_NO_CONNECT:
1634 			return 0;
1635 		/* signal the upper layer it might try again */
1636 		case DID_BUS_BUSY:
1637 		case DID_IMM_RETRY:
1638 		case DID_REQUEUE:
1639 		case DID_SOFT_ERROR:
1640 			return -EBUSY;
1641 		default:
1642 			return -EIO;
1643 		}
1644 	}
1645 	return 0;
1646 }
1647 
1648 static void sd_rescan(struct device *dev)
1649 {
1650 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1651 
1652 	sd_revalidate_disk(sdkp->disk);
1653 }
1654 
1655 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1656 		enum blk_unique_id type)
1657 {
1658 	struct scsi_device *sdev = scsi_disk(disk)->device;
1659 	const struct scsi_vpd *vpd;
1660 	const unsigned char *d;
1661 	int ret = -ENXIO, len;
1662 
1663 	rcu_read_lock();
1664 	vpd = rcu_dereference(sdev->vpd_pg83);
1665 	if (!vpd)
1666 		goto out_unlock;
1667 
1668 	ret = -EINVAL;
1669 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1670 		/* we only care about designators with LU association */
1671 		if (((d[1] >> 4) & 0x3) != 0x00)
1672 			continue;
1673 		if ((d[1] & 0xf) != type)
1674 			continue;
1675 
1676 		/*
1677 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1678 		 * keep looking as one with more entropy might still show up.
1679 		 */
1680 		len = d[3];
1681 		if (len != 8 && len != 12 && len != 16)
1682 			continue;
1683 		ret = len;
1684 		memcpy(id, d + 4, len);
1685 		if (len == 16)
1686 			break;
1687 	}
1688 out_unlock:
1689 	rcu_read_unlock();
1690 	return ret;
1691 }
1692 
1693 static char sd_pr_type(enum pr_type type)
1694 {
1695 	switch (type) {
1696 	case PR_WRITE_EXCLUSIVE:
1697 		return 0x01;
1698 	case PR_EXCLUSIVE_ACCESS:
1699 		return 0x03;
1700 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1701 		return 0x05;
1702 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1703 		return 0x06;
1704 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1705 		return 0x07;
1706 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1707 		return 0x08;
1708 	default:
1709 		return 0;
1710 	}
1711 };
1712 
1713 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1714 {
1715 	switch (host_byte(result)) {
1716 	case DID_TRANSPORT_MARGINAL:
1717 	case DID_TRANSPORT_DISRUPTED:
1718 	case DID_BUS_BUSY:
1719 		return PR_STS_RETRY_PATH_FAILURE;
1720 	case DID_NO_CONNECT:
1721 		return PR_STS_PATH_FAILED;
1722 	case DID_TRANSPORT_FAILFAST:
1723 		return PR_STS_PATH_FAST_FAILED;
1724 	}
1725 
1726 	switch (status_byte(result)) {
1727 	case SAM_STAT_RESERVATION_CONFLICT:
1728 		return PR_STS_RESERVATION_CONFLICT;
1729 	case SAM_STAT_CHECK_CONDITION:
1730 		if (!scsi_sense_valid(sshdr))
1731 			return PR_STS_IOERR;
1732 
1733 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1734 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1735 			return -EINVAL;
1736 
1737 		fallthrough;
1738 	default:
1739 		return PR_STS_IOERR;
1740 	}
1741 }
1742 
1743 static int sd_pr_command(struct block_device *bdev, u8 sa,
1744 		u64 key, u64 sa_key, u8 type, u8 flags)
1745 {
1746 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1747 	struct scsi_device *sdev = sdkp->device;
1748 	struct scsi_sense_hdr sshdr;
1749 	const struct scsi_exec_args exec_args = {
1750 		.sshdr = &sshdr,
1751 	};
1752 	int result;
1753 	u8 cmd[16] = { 0, };
1754 	u8 data[24] = { 0, };
1755 
1756 	cmd[0] = PERSISTENT_RESERVE_OUT;
1757 	cmd[1] = sa;
1758 	cmd[2] = type;
1759 	put_unaligned_be32(sizeof(data), &cmd[5]);
1760 
1761 	put_unaligned_be64(key, &data[0]);
1762 	put_unaligned_be64(sa_key, &data[8]);
1763 	data[20] = flags;
1764 
1765 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1766 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1767 				  &exec_args);
1768 
1769 	if (scsi_status_is_check_condition(result) &&
1770 	    scsi_sense_valid(&sshdr)) {
1771 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1772 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1773 	}
1774 
1775 	if (result <= 0)
1776 		return result;
1777 
1778 	return sd_scsi_to_pr_err(&sshdr, result);
1779 }
1780 
1781 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1782 		u32 flags)
1783 {
1784 	if (flags & ~PR_FL_IGNORE_KEY)
1785 		return -EOPNOTSUPP;
1786 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1787 			old_key, new_key, 0,
1788 			(1 << 0) /* APTPL */);
1789 }
1790 
1791 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1792 		u32 flags)
1793 {
1794 	if (flags)
1795 		return -EOPNOTSUPP;
1796 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1797 }
1798 
1799 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1800 {
1801 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1802 }
1803 
1804 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1805 		enum pr_type type, bool abort)
1806 {
1807 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1808 			     sd_pr_type(type), 0);
1809 }
1810 
1811 static int sd_pr_clear(struct block_device *bdev, u64 key)
1812 {
1813 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1814 }
1815 
1816 static const struct pr_ops sd_pr_ops = {
1817 	.pr_register	= sd_pr_register,
1818 	.pr_reserve	= sd_pr_reserve,
1819 	.pr_release	= sd_pr_release,
1820 	.pr_preempt	= sd_pr_preempt,
1821 	.pr_clear	= sd_pr_clear,
1822 };
1823 
1824 static void scsi_disk_free_disk(struct gendisk *disk)
1825 {
1826 	struct scsi_disk *sdkp = scsi_disk(disk);
1827 
1828 	put_device(&sdkp->disk_dev);
1829 }
1830 
1831 static const struct block_device_operations sd_fops = {
1832 	.owner			= THIS_MODULE,
1833 	.open			= sd_open,
1834 	.release		= sd_release,
1835 	.ioctl			= sd_ioctl,
1836 	.getgeo			= sd_getgeo,
1837 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1838 	.check_events		= sd_check_events,
1839 	.unlock_native_capacity	= sd_unlock_native_capacity,
1840 	.report_zones		= sd_zbc_report_zones,
1841 	.get_unique_id		= sd_get_unique_id,
1842 	.free_disk		= scsi_disk_free_disk,
1843 	.pr_ops			= &sd_pr_ops,
1844 };
1845 
1846 /**
1847  *	sd_eh_reset - reset error handling callback
1848  *	@scmd:		sd-issued command that has failed
1849  *
1850  *	This function is called by the SCSI midlayer before starting
1851  *	SCSI EH. When counting medium access failures we have to be
1852  *	careful to register it only only once per device and SCSI EH run;
1853  *	there might be several timed out commands which will cause the
1854  *	'max_medium_access_timeouts' counter to trigger after the first
1855  *	SCSI EH run already and set the device to offline.
1856  *	So this function resets the internal counter before starting SCSI EH.
1857  **/
1858 static void sd_eh_reset(struct scsi_cmnd *scmd)
1859 {
1860 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1861 
1862 	/* New SCSI EH run, reset gate variable */
1863 	sdkp->ignore_medium_access_errors = false;
1864 }
1865 
1866 /**
1867  *	sd_eh_action - error handling callback
1868  *	@scmd:		sd-issued command that has failed
1869  *	@eh_disp:	The recovery disposition suggested by the midlayer
1870  *
1871  *	This function is called by the SCSI midlayer upon completion of an
1872  *	error test command (currently TEST UNIT READY). The result of sending
1873  *	the eh command is passed in eh_disp.  We're looking for devices that
1874  *	fail medium access commands but are OK with non access commands like
1875  *	test unit ready (so wrongly see the device as having a successful
1876  *	recovery)
1877  **/
1878 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1879 {
1880 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1881 	struct scsi_device *sdev = scmd->device;
1882 
1883 	if (!scsi_device_online(sdev) ||
1884 	    !scsi_medium_access_command(scmd) ||
1885 	    host_byte(scmd->result) != DID_TIME_OUT ||
1886 	    eh_disp != SUCCESS)
1887 		return eh_disp;
1888 
1889 	/*
1890 	 * The device has timed out executing a medium access command.
1891 	 * However, the TEST UNIT READY command sent during error
1892 	 * handling completed successfully. Either the device is in the
1893 	 * process of recovering or has it suffered an internal failure
1894 	 * that prevents access to the storage medium.
1895 	 */
1896 	if (!sdkp->ignore_medium_access_errors) {
1897 		sdkp->medium_access_timed_out++;
1898 		sdkp->ignore_medium_access_errors = true;
1899 	}
1900 
1901 	/*
1902 	 * If the device keeps failing read/write commands but TEST UNIT
1903 	 * READY always completes successfully we assume that medium
1904 	 * access is no longer possible and take the device offline.
1905 	 */
1906 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1907 		scmd_printk(KERN_ERR, scmd,
1908 			    "Medium access timeout failure. Offlining disk!\n");
1909 		mutex_lock(&sdev->state_mutex);
1910 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1911 		mutex_unlock(&sdev->state_mutex);
1912 
1913 		return SUCCESS;
1914 	}
1915 
1916 	return eh_disp;
1917 }
1918 
1919 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1920 {
1921 	struct request *req = scsi_cmd_to_rq(scmd);
1922 	struct scsi_device *sdev = scmd->device;
1923 	unsigned int transferred, good_bytes;
1924 	u64 start_lba, end_lba, bad_lba;
1925 
1926 	/*
1927 	 * Some commands have a payload smaller than the device logical
1928 	 * block size (e.g. INQUIRY on a 4K disk).
1929 	 */
1930 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1931 		return 0;
1932 
1933 	/* Check if we have a 'bad_lba' information */
1934 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1935 				     SCSI_SENSE_BUFFERSIZE,
1936 				     &bad_lba))
1937 		return 0;
1938 
1939 	/*
1940 	 * If the bad lba was reported incorrectly, we have no idea where
1941 	 * the error is.
1942 	 */
1943 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1944 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1945 	if (bad_lba < start_lba || bad_lba >= end_lba)
1946 		return 0;
1947 
1948 	/*
1949 	 * resid is optional but mostly filled in.  When it's unused,
1950 	 * its value is zero, so we assume the whole buffer transferred
1951 	 */
1952 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1953 
1954 	/* This computation should always be done in terms of the
1955 	 * resolution of the device's medium.
1956 	 */
1957 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1958 
1959 	return min(good_bytes, transferred);
1960 }
1961 
1962 /**
1963  *	sd_done - bottom half handler: called when the lower level
1964  *	driver has completed (successfully or otherwise) a scsi command.
1965  *	@SCpnt: mid-level's per command structure.
1966  *
1967  *	Note: potentially run from within an ISR. Must not block.
1968  **/
1969 static int sd_done(struct scsi_cmnd *SCpnt)
1970 {
1971 	int result = SCpnt->result;
1972 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1973 	unsigned int sector_size = SCpnt->device->sector_size;
1974 	unsigned int resid;
1975 	struct scsi_sense_hdr sshdr;
1976 	struct request *req = scsi_cmd_to_rq(SCpnt);
1977 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1978 	int sense_valid = 0;
1979 	int sense_deferred = 0;
1980 
1981 	switch (req_op(req)) {
1982 	case REQ_OP_DISCARD:
1983 	case REQ_OP_WRITE_ZEROES:
1984 	case REQ_OP_ZONE_RESET:
1985 	case REQ_OP_ZONE_RESET_ALL:
1986 	case REQ_OP_ZONE_OPEN:
1987 	case REQ_OP_ZONE_CLOSE:
1988 	case REQ_OP_ZONE_FINISH:
1989 		if (!result) {
1990 			good_bytes = blk_rq_bytes(req);
1991 			scsi_set_resid(SCpnt, 0);
1992 		} else {
1993 			good_bytes = 0;
1994 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1995 		}
1996 		break;
1997 	default:
1998 		/*
1999 		 * In case of bogus fw or device, we could end up having
2000 		 * an unaligned partial completion. Check this here and force
2001 		 * alignment.
2002 		 */
2003 		resid = scsi_get_resid(SCpnt);
2004 		if (resid & (sector_size - 1)) {
2005 			sd_printk(KERN_INFO, sdkp,
2006 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2007 				resid, sector_size);
2008 			scsi_print_command(SCpnt);
2009 			resid = min(scsi_bufflen(SCpnt),
2010 				    round_up(resid, sector_size));
2011 			scsi_set_resid(SCpnt, resid);
2012 		}
2013 	}
2014 
2015 	if (result) {
2016 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2017 		if (sense_valid)
2018 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2019 	}
2020 	sdkp->medium_access_timed_out = 0;
2021 
2022 	if (!scsi_status_is_check_condition(result) &&
2023 	    (!sense_valid || sense_deferred))
2024 		goto out;
2025 
2026 	switch (sshdr.sense_key) {
2027 	case HARDWARE_ERROR:
2028 	case MEDIUM_ERROR:
2029 		good_bytes = sd_completed_bytes(SCpnt);
2030 		break;
2031 	case RECOVERED_ERROR:
2032 		good_bytes = scsi_bufflen(SCpnt);
2033 		break;
2034 	case NO_SENSE:
2035 		/* This indicates a false check condition, so ignore it.  An
2036 		 * unknown amount of data was transferred so treat it as an
2037 		 * error.
2038 		 */
2039 		SCpnt->result = 0;
2040 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2041 		break;
2042 	case ABORTED_COMMAND:
2043 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2044 			good_bytes = sd_completed_bytes(SCpnt);
2045 		break;
2046 	case ILLEGAL_REQUEST:
2047 		switch (sshdr.asc) {
2048 		case 0x10:	/* DIX: Host detected corruption */
2049 			good_bytes = sd_completed_bytes(SCpnt);
2050 			break;
2051 		case 0x20:	/* INVALID COMMAND OPCODE */
2052 		case 0x24:	/* INVALID FIELD IN CDB */
2053 			switch (SCpnt->cmnd[0]) {
2054 			case UNMAP:
2055 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2056 				break;
2057 			case WRITE_SAME_16:
2058 			case WRITE_SAME:
2059 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2060 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2061 				} else {
2062 					sdkp->device->no_write_same = 1;
2063 					sd_config_write_same(sdkp);
2064 					req->rq_flags |= RQF_QUIET;
2065 				}
2066 				break;
2067 			}
2068 		}
2069 		break;
2070 	default:
2071 		break;
2072 	}
2073 
2074  out:
2075 	if (sd_is_zoned(sdkp))
2076 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2077 
2078 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2079 					   "sd_done: completed %d of %d bytes\n",
2080 					   good_bytes, scsi_bufflen(SCpnt)));
2081 
2082 	return good_bytes;
2083 }
2084 
2085 /*
2086  * spinup disk - called only in sd_revalidate_disk()
2087  */
2088 static void
2089 sd_spinup_disk(struct scsi_disk *sdkp)
2090 {
2091 	unsigned char cmd[10];
2092 	unsigned long spintime_expire = 0;
2093 	int retries, spintime;
2094 	unsigned int the_result;
2095 	struct scsi_sense_hdr sshdr;
2096 	const struct scsi_exec_args exec_args = {
2097 		.sshdr = &sshdr,
2098 	};
2099 	int sense_valid = 0;
2100 
2101 	spintime = 0;
2102 
2103 	/* Spin up drives, as required.  Only do this at boot time */
2104 	/* Spinup needs to be done for module loads too. */
2105 	do {
2106 		retries = 0;
2107 
2108 		do {
2109 			bool media_was_present = sdkp->media_present;
2110 
2111 			cmd[0] = TEST_UNIT_READY;
2112 			memset((void *) &cmd[1], 0, 9);
2113 
2114 			the_result = scsi_execute_cmd(sdkp->device, cmd,
2115 						      REQ_OP_DRV_IN, NULL, 0,
2116 						      SD_TIMEOUT,
2117 						      sdkp->max_retries,
2118 						      &exec_args);
2119 
2120 			/*
2121 			 * If the drive has indicated to us that it
2122 			 * doesn't have any media in it, don't bother
2123 			 * with any more polling.
2124 			 */
2125 			if (media_not_present(sdkp, &sshdr)) {
2126 				if (media_was_present)
2127 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2128 				return;
2129 			}
2130 
2131 			if (the_result)
2132 				sense_valid = scsi_sense_valid(&sshdr);
2133 			retries++;
2134 		} while (retries < 3 &&
2135 			 (!scsi_status_is_good(the_result) ||
2136 			  (scsi_status_is_check_condition(the_result) &&
2137 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2138 
2139 		if (!scsi_status_is_check_condition(the_result)) {
2140 			/* no sense, TUR either succeeded or failed
2141 			 * with a status error */
2142 			if(!spintime && !scsi_status_is_good(the_result)) {
2143 				sd_print_result(sdkp, "Test Unit Ready failed",
2144 						the_result);
2145 			}
2146 			break;
2147 		}
2148 
2149 		/*
2150 		 * The device does not want the automatic start to be issued.
2151 		 */
2152 		if (sdkp->device->no_start_on_add)
2153 			break;
2154 
2155 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2156 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2157 				break;	/* manual intervention required */
2158 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2159 				break;	/* standby */
2160 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2161 				break;	/* unavailable */
2162 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2163 				break;	/* sanitize in progress */
2164 			/*
2165 			 * Issue command to spin up drive when not ready
2166 			 */
2167 			if (!spintime) {
2168 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2169 				cmd[0] = START_STOP;
2170 				cmd[1] = 1;	/* Return immediately */
2171 				memset((void *) &cmd[2], 0, 8);
2172 				cmd[4] = 1;	/* Start spin cycle */
2173 				if (sdkp->device->start_stop_pwr_cond)
2174 					cmd[4] |= 1 << 4;
2175 				scsi_execute_cmd(sdkp->device, cmd,
2176 						 REQ_OP_DRV_IN, NULL, 0,
2177 						 SD_TIMEOUT, sdkp->max_retries,
2178 						 &exec_args);
2179 				spintime_expire = jiffies + 100 * HZ;
2180 				spintime = 1;
2181 			}
2182 			/* Wait 1 second for next try */
2183 			msleep(1000);
2184 			printk(KERN_CONT ".");
2185 
2186 		/*
2187 		 * Wait for USB flash devices with slow firmware.
2188 		 * Yes, this sense key/ASC combination shouldn't
2189 		 * occur here.  It's characteristic of these devices.
2190 		 */
2191 		} else if (sense_valid &&
2192 				sshdr.sense_key == UNIT_ATTENTION &&
2193 				sshdr.asc == 0x28) {
2194 			if (!spintime) {
2195 				spintime_expire = jiffies + 5 * HZ;
2196 				spintime = 1;
2197 			}
2198 			/* Wait 1 second for next try */
2199 			msleep(1000);
2200 		} else {
2201 			/* we don't understand the sense code, so it's
2202 			 * probably pointless to loop */
2203 			if(!spintime) {
2204 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2205 				sd_print_sense_hdr(sdkp, &sshdr);
2206 			}
2207 			break;
2208 		}
2209 
2210 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2211 
2212 	if (spintime) {
2213 		if (scsi_status_is_good(the_result))
2214 			printk(KERN_CONT "ready\n");
2215 		else
2216 			printk(KERN_CONT "not responding...\n");
2217 	}
2218 }
2219 
2220 /*
2221  * Determine whether disk supports Data Integrity Field.
2222  */
2223 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2224 {
2225 	struct scsi_device *sdp = sdkp->device;
2226 	u8 type;
2227 
2228 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2229 		sdkp->protection_type = 0;
2230 		return 0;
2231 	}
2232 
2233 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2234 
2235 	if (type > T10_PI_TYPE3_PROTECTION) {
2236 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2237 			  " protection type %u. Disabling disk!\n",
2238 			  type);
2239 		sdkp->protection_type = 0;
2240 		return -ENODEV;
2241 	}
2242 
2243 	sdkp->protection_type = type;
2244 
2245 	return 0;
2246 }
2247 
2248 static void sd_config_protection(struct scsi_disk *sdkp)
2249 {
2250 	struct scsi_device *sdp = sdkp->device;
2251 
2252 	sd_dif_config_host(sdkp);
2253 
2254 	if (!sdkp->protection_type)
2255 		return;
2256 
2257 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2258 		sd_first_printk(KERN_NOTICE, sdkp,
2259 				"Disabling DIF Type %u protection\n",
2260 				sdkp->protection_type);
2261 		sdkp->protection_type = 0;
2262 	}
2263 
2264 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2265 			sdkp->protection_type);
2266 }
2267 
2268 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2269 			struct scsi_sense_hdr *sshdr, int sense_valid,
2270 			int the_result)
2271 {
2272 	if (sense_valid)
2273 		sd_print_sense_hdr(sdkp, sshdr);
2274 	else
2275 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2276 
2277 	/*
2278 	 * Set dirty bit for removable devices if not ready -
2279 	 * sometimes drives will not report this properly.
2280 	 */
2281 	if (sdp->removable &&
2282 	    sense_valid && sshdr->sense_key == NOT_READY)
2283 		set_media_not_present(sdkp);
2284 
2285 	/*
2286 	 * We used to set media_present to 0 here to indicate no media
2287 	 * in the drive, but some drives fail read capacity even with
2288 	 * media present, so we can't do that.
2289 	 */
2290 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2291 }
2292 
2293 #define RC16_LEN 32
2294 #if RC16_LEN > SD_BUF_SIZE
2295 #error RC16_LEN must not be more than SD_BUF_SIZE
2296 #endif
2297 
2298 #define READ_CAPACITY_RETRIES_ON_RESET	10
2299 
2300 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2301 						unsigned char *buffer)
2302 {
2303 	unsigned char cmd[16];
2304 	struct scsi_sense_hdr sshdr;
2305 	const struct scsi_exec_args exec_args = {
2306 		.sshdr = &sshdr,
2307 	};
2308 	int sense_valid = 0;
2309 	int the_result;
2310 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2311 	unsigned int alignment;
2312 	unsigned long long lba;
2313 	unsigned sector_size;
2314 
2315 	if (sdp->no_read_capacity_16)
2316 		return -EINVAL;
2317 
2318 	do {
2319 		memset(cmd, 0, 16);
2320 		cmd[0] = SERVICE_ACTION_IN_16;
2321 		cmd[1] = SAI_READ_CAPACITY_16;
2322 		cmd[13] = RC16_LEN;
2323 		memset(buffer, 0, RC16_LEN);
2324 
2325 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2326 					      buffer, RC16_LEN, SD_TIMEOUT,
2327 					      sdkp->max_retries, &exec_args);
2328 
2329 		if (media_not_present(sdkp, &sshdr))
2330 			return -ENODEV;
2331 
2332 		if (the_result > 0) {
2333 			sense_valid = scsi_sense_valid(&sshdr);
2334 			if (sense_valid &&
2335 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2336 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2337 			    sshdr.ascq == 0x00)
2338 				/* Invalid Command Operation Code or
2339 				 * Invalid Field in CDB, just retry
2340 				 * silently with RC10 */
2341 				return -EINVAL;
2342 			if (sense_valid &&
2343 			    sshdr.sense_key == UNIT_ATTENTION &&
2344 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2345 				/* Device reset might occur several times,
2346 				 * give it one more chance */
2347 				if (--reset_retries > 0)
2348 					continue;
2349 		}
2350 		retries--;
2351 
2352 	} while (the_result && retries);
2353 
2354 	if (the_result) {
2355 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2356 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2357 		return -EINVAL;
2358 	}
2359 
2360 	sector_size = get_unaligned_be32(&buffer[8]);
2361 	lba = get_unaligned_be64(&buffer[0]);
2362 
2363 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2364 		sdkp->capacity = 0;
2365 		return -ENODEV;
2366 	}
2367 
2368 	/* Logical blocks per physical block exponent */
2369 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2370 
2371 	/* RC basis */
2372 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2373 
2374 	/* Lowest aligned logical block */
2375 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2376 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2377 	if (alignment && sdkp->first_scan)
2378 		sd_printk(KERN_NOTICE, sdkp,
2379 			  "physical block alignment offset: %u\n", alignment);
2380 
2381 	if (buffer[14] & 0x80) { /* LBPME */
2382 		sdkp->lbpme = 1;
2383 
2384 		if (buffer[14] & 0x40) /* LBPRZ */
2385 			sdkp->lbprz = 1;
2386 
2387 		sd_config_discard(sdkp, SD_LBP_WS16);
2388 	}
2389 
2390 	sdkp->capacity = lba + 1;
2391 	return sector_size;
2392 }
2393 
2394 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2395 						unsigned char *buffer)
2396 {
2397 	unsigned char cmd[16];
2398 	struct scsi_sense_hdr sshdr;
2399 	const struct scsi_exec_args exec_args = {
2400 		.sshdr = &sshdr,
2401 	};
2402 	int sense_valid = 0;
2403 	int the_result;
2404 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2405 	sector_t lba;
2406 	unsigned sector_size;
2407 
2408 	do {
2409 		cmd[0] = READ_CAPACITY;
2410 		memset(&cmd[1], 0, 9);
2411 		memset(buffer, 0, 8);
2412 
2413 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2414 					      8, SD_TIMEOUT, sdkp->max_retries,
2415 					      &exec_args);
2416 
2417 		if (media_not_present(sdkp, &sshdr))
2418 			return -ENODEV;
2419 
2420 		if (the_result > 0) {
2421 			sense_valid = scsi_sense_valid(&sshdr);
2422 			if (sense_valid &&
2423 			    sshdr.sense_key == UNIT_ATTENTION &&
2424 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2425 				/* Device reset might occur several times,
2426 				 * give it one more chance */
2427 				if (--reset_retries > 0)
2428 					continue;
2429 		}
2430 		retries--;
2431 
2432 	} while (the_result && retries);
2433 
2434 	if (the_result) {
2435 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2436 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2437 		return -EINVAL;
2438 	}
2439 
2440 	sector_size = get_unaligned_be32(&buffer[4]);
2441 	lba = get_unaligned_be32(&buffer[0]);
2442 
2443 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2444 		/* Some buggy (usb cardreader) devices return an lba of
2445 		   0xffffffff when the want to report a size of 0 (with
2446 		   which they really mean no media is present) */
2447 		sdkp->capacity = 0;
2448 		sdkp->physical_block_size = sector_size;
2449 		return sector_size;
2450 	}
2451 
2452 	sdkp->capacity = lba + 1;
2453 	sdkp->physical_block_size = sector_size;
2454 	return sector_size;
2455 }
2456 
2457 static int sd_try_rc16_first(struct scsi_device *sdp)
2458 {
2459 	if (sdp->host->max_cmd_len < 16)
2460 		return 0;
2461 	if (sdp->try_rc_10_first)
2462 		return 0;
2463 	if (sdp->scsi_level > SCSI_SPC_2)
2464 		return 1;
2465 	if (scsi_device_protection(sdp))
2466 		return 1;
2467 	return 0;
2468 }
2469 
2470 /*
2471  * read disk capacity
2472  */
2473 static void
2474 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2475 {
2476 	int sector_size;
2477 	struct scsi_device *sdp = sdkp->device;
2478 
2479 	if (sd_try_rc16_first(sdp)) {
2480 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2481 		if (sector_size == -EOVERFLOW)
2482 			goto got_data;
2483 		if (sector_size == -ENODEV)
2484 			return;
2485 		if (sector_size < 0)
2486 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2487 		if (sector_size < 0)
2488 			return;
2489 	} else {
2490 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2491 		if (sector_size == -EOVERFLOW)
2492 			goto got_data;
2493 		if (sector_size < 0)
2494 			return;
2495 		if ((sizeof(sdkp->capacity) > 4) &&
2496 		    (sdkp->capacity > 0xffffffffULL)) {
2497 			int old_sector_size = sector_size;
2498 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2499 					"Trying to use READ CAPACITY(16).\n");
2500 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2501 			if (sector_size < 0) {
2502 				sd_printk(KERN_NOTICE, sdkp,
2503 					"Using 0xffffffff as device size\n");
2504 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2505 				sector_size = old_sector_size;
2506 				goto got_data;
2507 			}
2508 			/* Remember that READ CAPACITY(16) succeeded */
2509 			sdp->try_rc_10_first = 0;
2510 		}
2511 	}
2512 
2513 	/* Some devices are known to return the total number of blocks,
2514 	 * not the highest block number.  Some devices have versions
2515 	 * which do this and others which do not.  Some devices we might
2516 	 * suspect of doing this but we don't know for certain.
2517 	 *
2518 	 * If we know the reported capacity is wrong, decrement it.  If
2519 	 * we can only guess, then assume the number of blocks is even
2520 	 * (usually true but not always) and err on the side of lowering
2521 	 * the capacity.
2522 	 */
2523 	if (sdp->fix_capacity ||
2524 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2525 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2526 				"from its reported value: %llu\n",
2527 				(unsigned long long) sdkp->capacity);
2528 		--sdkp->capacity;
2529 	}
2530 
2531 got_data:
2532 	if (sector_size == 0) {
2533 		sector_size = 512;
2534 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2535 			  "assuming 512.\n");
2536 	}
2537 
2538 	if (sector_size != 512 &&
2539 	    sector_size != 1024 &&
2540 	    sector_size != 2048 &&
2541 	    sector_size != 4096) {
2542 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2543 			  sector_size);
2544 		/*
2545 		 * The user might want to re-format the drive with
2546 		 * a supported sectorsize.  Once this happens, it
2547 		 * would be relatively trivial to set the thing up.
2548 		 * For this reason, we leave the thing in the table.
2549 		 */
2550 		sdkp->capacity = 0;
2551 		/*
2552 		 * set a bogus sector size so the normal read/write
2553 		 * logic in the block layer will eventually refuse any
2554 		 * request on this device without tripping over power
2555 		 * of two sector size assumptions
2556 		 */
2557 		sector_size = 512;
2558 	}
2559 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2560 	blk_queue_physical_block_size(sdp->request_queue,
2561 				      sdkp->physical_block_size);
2562 	sdkp->device->sector_size = sector_size;
2563 
2564 	if (sdkp->capacity > 0xffffffff)
2565 		sdp->use_16_for_rw = 1;
2566 
2567 }
2568 
2569 /*
2570  * Print disk capacity
2571  */
2572 static void
2573 sd_print_capacity(struct scsi_disk *sdkp,
2574 		  sector_t old_capacity)
2575 {
2576 	int sector_size = sdkp->device->sector_size;
2577 	char cap_str_2[10], cap_str_10[10];
2578 
2579 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2580 		return;
2581 
2582 	string_get_size(sdkp->capacity, sector_size,
2583 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2584 	string_get_size(sdkp->capacity, sector_size,
2585 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2586 
2587 	sd_printk(KERN_NOTICE, sdkp,
2588 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2589 		  (unsigned long long)sdkp->capacity,
2590 		  sector_size, cap_str_10, cap_str_2);
2591 
2592 	if (sdkp->physical_block_size != sector_size)
2593 		sd_printk(KERN_NOTICE, sdkp,
2594 			  "%u-byte physical blocks\n",
2595 			  sdkp->physical_block_size);
2596 }
2597 
2598 /* called with buffer of length 512 */
2599 static inline int
2600 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2601 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2602 		 struct scsi_sense_hdr *sshdr)
2603 {
2604 	/*
2605 	 * If we must use MODE SENSE(10), make sure that the buffer length
2606 	 * is at least 8 bytes so that the mode sense header fits.
2607 	 */
2608 	if (sdkp->device->use_10_for_ms && len < 8)
2609 		len = 8;
2610 
2611 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2612 			       SD_TIMEOUT, sdkp->max_retries, data,
2613 			       sshdr);
2614 }
2615 
2616 /*
2617  * read write protect setting, if possible - called only in sd_revalidate_disk()
2618  * called with buffer of length SD_BUF_SIZE
2619  */
2620 static void
2621 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2622 {
2623 	int res;
2624 	struct scsi_device *sdp = sdkp->device;
2625 	struct scsi_mode_data data;
2626 	int old_wp = sdkp->write_prot;
2627 
2628 	set_disk_ro(sdkp->disk, 0);
2629 	if (sdp->skip_ms_page_3f) {
2630 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2631 		return;
2632 	}
2633 
2634 	if (sdp->use_192_bytes_for_3f) {
2635 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2636 	} else {
2637 		/*
2638 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2639 		 * We have to start carefully: some devices hang if we ask
2640 		 * for more than is available.
2641 		 */
2642 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2643 
2644 		/*
2645 		 * Second attempt: ask for page 0 When only page 0 is
2646 		 * implemented, a request for page 3F may return Sense Key
2647 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2648 		 * CDB.
2649 		 */
2650 		if (res < 0)
2651 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2652 
2653 		/*
2654 		 * Third attempt: ask 255 bytes, as we did earlier.
2655 		 */
2656 		if (res < 0)
2657 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2658 					       &data, NULL);
2659 	}
2660 
2661 	if (res < 0) {
2662 		sd_first_printk(KERN_WARNING, sdkp,
2663 			  "Test WP failed, assume Write Enabled\n");
2664 	} else {
2665 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2666 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2667 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2668 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2669 				  sdkp->write_prot ? "on" : "off");
2670 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2671 		}
2672 	}
2673 }
2674 
2675 /*
2676  * sd_read_cache_type - called only from sd_revalidate_disk()
2677  * called with buffer of length SD_BUF_SIZE
2678  */
2679 static void
2680 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2681 {
2682 	int len = 0, res;
2683 	struct scsi_device *sdp = sdkp->device;
2684 
2685 	int dbd;
2686 	int modepage;
2687 	int first_len;
2688 	struct scsi_mode_data data;
2689 	struct scsi_sense_hdr sshdr;
2690 	int old_wce = sdkp->WCE;
2691 	int old_rcd = sdkp->RCD;
2692 	int old_dpofua = sdkp->DPOFUA;
2693 
2694 
2695 	if (sdkp->cache_override)
2696 		return;
2697 
2698 	first_len = 4;
2699 	if (sdp->skip_ms_page_8) {
2700 		if (sdp->type == TYPE_RBC)
2701 			goto defaults;
2702 		else {
2703 			if (sdp->skip_ms_page_3f)
2704 				goto defaults;
2705 			modepage = 0x3F;
2706 			if (sdp->use_192_bytes_for_3f)
2707 				first_len = 192;
2708 			dbd = 0;
2709 		}
2710 	} else if (sdp->type == TYPE_RBC) {
2711 		modepage = 6;
2712 		dbd = 8;
2713 	} else {
2714 		modepage = 8;
2715 		dbd = 0;
2716 	}
2717 
2718 	/* cautiously ask */
2719 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2720 			&data, &sshdr);
2721 
2722 	if (res < 0)
2723 		goto bad_sense;
2724 
2725 	if (!data.header_length) {
2726 		modepage = 6;
2727 		first_len = 0;
2728 		sd_first_printk(KERN_ERR, sdkp,
2729 				"Missing header in MODE_SENSE response\n");
2730 	}
2731 
2732 	/* that went OK, now ask for the proper length */
2733 	len = data.length;
2734 
2735 	/*
2736 	 * We're only interested in the first three bytes, actually.
2737 	 * But the data cache page is defined for the first 20.
2738 	 */
2739 	if (len < 3)
2740 		goto bad_sense;
2741 	else if (len > SD_BUF_SIZE) {
2742 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2743 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2744 		len = SD_BUF_SIZE;
2745 	}
2746 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2747 		len = 192;
2748 
2749 	/* Get the data */
2750 	if (len > first_len)
2751 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2752 				&data, &sshdr);
2753 
2754 	if (!res) {
2755 		int offset = data.header_length + data.block_descriptor_length;
2756 
2757 		while (offset < len) {
2758 			u8 page_code = buffer[offset] & 0x3F;
2759 			u8 spf       = buffer[offset] & 0x40;
2760 
2761 			if (page_code == 8 || page_code == 6) {
2762 				/* We're interested only in the first 3 bytes.
2763 				 */
2764 				if (len - offset <= 2) {
2765 					sd_first_printk(KERN_ERR, sdkp,
2766 						"Incomplete mode parameter "
2767 							"data\n");
2768 					goto defaults;
2769 				} else {
2770 					modepage = page_code;
2771 					goto Page_found;
2772 				}
2773 			} else {
2774 				/* Go to the next page */
2775 				if (spf && len - offset > 3)
2776 					offset += 4 + (buffer[offset+2] << 8) +
2777 						buffer[offset+3];
2778 				else if (!spf && len - offset > 1)
2779 					offset += 2 + buffer[offset+1];
2780 				else {
2781 					sd_first_printk(KERN_ERR, sdkp,
2782 							"Incomplete mode "
2783 							"parameter data\n");
2784 					goto defaults;
2785 				}
2786 			}
2787 		}
2788 
2789 		sd_first_printk(KERN_WARNING, sdkp,
2790 				"No Caching mode page found\n");
2791 		goto defaults;
2792 
2793 	Page_found:
2794 		if (modepage == 8) {
2795 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2796 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2797 		} else {
2798 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2799 			sdkp->RCD = 0;
2800 		}
2801 
2802 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2803 		if (sdp->broken_fua) {
2804 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2805 			sdkp->DPOFUA = 0;
2806 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2807 			   !sdkp->device->use_16_for_rw) {
2808 			sd_first_printk(KERN_NOTICE, sdkp,
2809 				  "Uses READ/WRITE(6), disabling FUA\n");
2810 			sdkp->DPOFUA = 0;
2811 		}
2812 
2813 		/* No cache flush allowed for write protected devices */
2814 		if (sdkp->WCE && sdkp->write_prot)
2815 			sdkp->WCE = 0;
2816 
2817 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2818 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2819 			sd_printk(KERN_NOTICE, sdkp,
2820 				  "Write cache: %s, read cache: %s, %s\n",
2821 				  sdkp->WCE ? "enabled" : "disabled",
2822 				  sdkp->RCD ? "disabled" : "enabled",
2823 				  sdkp->DPOFUA ? "supports DPO and FUA"
2824 				  : "doesn't support DPO or FUA");
2825 
2826 		return;
2827 	}
2828 
2829 bad_sense:
2830 	if (scsi_sense_valid(&sshdr) &&
2831 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2832 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2833 		/* Invalid field in CDB */
2834 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2835 	else
2836 		sd_first_printk(KERN_ERR, sdkp,
2837 				"Asking for cache data failed\n");
2838 
2839 defaults:
2840 	if (sdp->wce_default_on) {
2841 		sd_first_printk(KERN_NOTICE, sdkp,
2842 				"Assuming drive cache: write back\n");
2843 		sdkp->WCE = 1;
2844 	} else {
2845 		sd_first_printk(KERN_WARNING, sdkp,
2846 				"Assuming drive cache: write through\n");
2847 		sdkp->WCE = 0;
2848 	}
2849 	sdkp->RCD = 0;
2850 	sdkp->DPOFUA = 0;
2851 }
2852 
2853 /*
2854  * The ATO bit indicates whether the DIF application tag is available
2855  * for use by the operating system.
2856  */
2857 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2858 {
2859 	int res, offset;
2860 	struct scsi_device *sdp = sdkp->device;
2861 	struct scsi_mode_data data;
2862 	struct scsi_sense_hdr sshdr;
2863 
2864 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2865 		return;
2866 
2867 	if (sdkp->protection_type == 0)
2868 		return;
2869 
2870 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2871 			      sdkp->max_retries, &data, &sshdr);
2872 
2873 	if (res < 0 || !data.header_length ||
2874 	    data.length < 6) {
2875 		sd_first_printk(KERN_WARNING, sdkp,
2876 			  "getting Control mode page failed, assume no ATO\n");
2877 
2878 		if (scsi_sense_valid(&sshdr))
2879 			sd_print_sense_hdr(sdkp, &sshdr);
2880 
2881 		return;
2882 	}
2883 
2884 	offset = data.header_length + data.block_descriptor_length;
2885 
2886 	if ((buffer[offset] & 0x3f) != 0x0a) {
2887 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2888 		return;
2889 	}
2890 
2891 	if ((buffer[offset + 5] & 0x80) == 0)
2892 		return;
2893 
2894 	sdkp->ATO = 1;
2895 
2896 	return;
2897 }
2898 
2899 /**
2900  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2901  * @sdkp: disk to query
2902  */
2903 static void sd_read_block_limits(struct scsi_disk *sdkp)
2904 {
2905 	struct scsi_vpd *vpd;
2906 
2907 	rcu_read_lock();
2908 
2909 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2910 	if (!vpd || vpd->len < 16)
2911 		goto out;
2912 
2913 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2914 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2915 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2916 
2917 	if (vpd->len >= 64) {
2918 		unsigned int lba_count, desc_count;
2919 
2920 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2921 
2922 		if (!sdkp->lbpme)
2923 			goto out;
2924 
2925 		lba_count = get_unaligned_be32(&vpd->data[20]);
2926 		desc_count = get_unaligned_be32(&vpd->data[24]);
2927 
2928 		if (lba_count && desc_count)
2929 			sdkp->max_unmap_blocks = lba_count;
2930 
2931 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2932 
2933 		if (vpd->data[32] & 0x80)
2934 			sdkp->unmap_alignment =
2935 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2936 
2937 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2938 
2939 			if (sdkp->max_unmap_blocks)
2940 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2941 			else
2942 				sd_config_discard(sdkp, SD_LBP_WS16);
2943 
2944 		} else {	/* LBP VPD page tells us what to use */
2945 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2946 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2947 			else if (sdkp->lbpws)
2948 				sd_config_discard(sdkp, SD_LBP_WS16);
2949 			else if (sdkp->lbpws10)
2950 				sd_config_discard(sdkp, SD_LBP_WS10);
2951 			else
2952 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2953 		}
2954 	}
2955 
2956  out:
2957 	rcu_read_unlock();
2958 }
2959 
2960 /**
2961  * sd_read_block_characteristics - Query block dev. characteristics
2962  * @sdkp: disk to query
2963  */
2964 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2965 {
2966 	struct request_queue *q = sdkp->disk->queue;
2967 	struct scsi_vpd *vpd;
2968 	u16 rot;
2969 	u8 zoned;
2970 
2971 	rcu_read_lock();
2972 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
2973 
2974 	if (!vpd || vpd->len < 8) {
2975 		rcu_read_unlock();
2976 	        return;
2977 	}
2978 
2979 	rot = get_unaligned_be16(&vpd->data[4]);
2980 	zoned = (vpd->data[8] >> 4) & 3;
2981 	rcu_read_unlock();
2982 
2983 	if (rot == 1) {
2984 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2985 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2986 	}
2987 
2988 	if (sdkp->device->type == TYPE_ZBC) {
2989 		/*
2990 		 * Host-managed: Per ZBC and ZAC specifications, writes in
2991 		 * sequential write required zones of host-managed devices must
2992 		 * be aligned to the device physical block size.
2993 		 */
2994 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
2995 		blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
2996 	} else {
2997 		sdkp->zoned = zoned;
2998 		if (sdkp->zoned == 1) {
2999 			/* Host-aware */
3000 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
3001 		} else {
3002 			/* Regular disk or drive managed disk */
3003 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3004 		}
3005 	}
3006 
3007 	if (!sdkp->first_scan)
3008 		return;
3009 
3010 	if (blk_queue_is_zoned(q)) {
3011 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3012 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3013 	} else {
3014 		if (sdkp->zoned == 1)
3015 			sd_printk(KERN_NOTICE, sdkp,
3016 				  "Host-aware SMR disk used as regular disk\n");
3017 		else if (sdkp->zoned == 2)
3018 			sd_printk(KERN_NOTICE, sdkp,
3019 				  "Drive-managed SMR disk\n");
3020 	}
3021 }
3022 
3023 /**
3024  * sd_read_block_provisioning - Query provisioning VPD page
3025  * @sdkp: disk to query
3026  */
3027 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3028 {
3029 	struct scsi_vpd *vpd;
3030 
3031 	if (sdkp->lbpme == 0)
3032 		return;
3033 
3034 	rcu_read_lock();
3035 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3036 
3037 	if (!vpd || vpd->len < 8) {
3038 		rcu_read_unlock();
3039 		return;
3040 	}
3041 
3042 	sdkp->lbpvpd	= 1;
3043 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3044 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3045 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3046 	rcu_read_unlock();
3047 }
3048 
3049 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3050 {
3051 	struct scsi_device *sdev = sdkp->device;
3052 
3053 	if (sdev->host->no_write_same) {
3054 		sdev->no_write_same = 1;
3055 
3056 		return;
3057 	}
3058 
3059 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
3060 		struct scsi_vpd *vpd;
3061 
3062 		sdev->no_report_opcodes = 1;
3063 
3064 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3065 		 * CODES is unsupported and the device has an ATA
3066 		 * Information VPD page (SAT).
3067 		 */
3068 		rcu_read_lock();
3069 		vpd = rcu_dereference(sdev->vpd_pg89);
3070 		if (vpd)
3071 			sdev->no_write_same = 1;
3072 		rcu_read_unlock();
3073 	}
3074 
3075 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3076 		sdkp->ws16 = 1;
3077 
3078 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3079 		sdkp->ws10 = 1;
3080 }
3081 
3082 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3083 {
3084 	struct scsi_device *sdev = sdkp->device;
3085 
3086 	if (!sdev->security_supported)
3087 		return;
3088 
3089 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3090 			SECURITY_PROTOCOL_IN) == 1 &&
3091 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3092 			SECURITY_PROTOCOL_OUT) == 1)
3093 		sdkp->security = 1;
3094 }
3095 
3096 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3097 {
3098 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3099 }
3100 
3101 /**
3102  * sd_read_cpr - Query concurrent positioning ranges
3103  * @sdkp:	disk to query
3104  */
3105 static void sd_read_cpr(struct scsi_disk *sdkp)
3106 {
3107 	struct blk_independent_access_ranges *iars = NULL;
3108 	unsigned char *buffer = NULL;
3109 	unsigned int nr_cpr = 0;
3110 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3111 	u8 *desc;
3112 
3113 	/*
3114 	 * We need to have the capacity set first for the block layer to be
3115 	 * able to check the ranges.
3116 	 */
3117 	if (sdkp->first_scan)
3118 		return;
3119 
3120 	if (!sdkp->capacity)
3121 		goto out;
3122 
3123 	/*
3124 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3125 	 * leading to a maximum page size of 64 + 256*32 bytes.
3126 	 */
3127 	buf_len = 64 + 256*32;
3128 	buffer = kmalloc(buf_len, GFP_KERNEL);
3129 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3130 		goto out;
3131 
3132 	/* We must have at least a 64B header and one 32B range descriptor */
3133 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3134 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3135 		sd_printk(KERN_ERR, sdkp,
3136 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3137 		goto out;
3138 	}
3139 
3140 	nr_cpr = (vpd_len - 64) / 32;
3141 	if (nr_cpr == 1) {
3142 		nr_cpr = 0;
3143 		goto out;
3144 	}
3145 
3146 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3147 	if (!iars) {
3148 		nr_cpr = 0;
3149 		goto out;
3150 	}
3151 
3152 	desc = &buffer[64];
3153 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3154 		if (desc[0] != i) {
3155 			sd_printk(KERN_ERR, sdkp,
3156 				"Invalid Concurrent Positioning Range number\n");
3157 			nr_cpr = 0;
3158 			break;
3159 		}
3160 
3161 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3162 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3163 	}
3164 
3165 out:
3166 	disk_set_independent_access_ranges(sdkp->disk, iars);
3167 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3168 		sd_printk(KERN_NOTICE, sdkp,
3169 			  "%u concurrent positioning ranges\n", nr_cpr);
3170 		sdkp->nr_actuators = nr_cpr;
3171 	}
3172 
3173 	kfree(buffer);
3174 }
3175 
3176 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3177 {
3178 	struct scsi_device *sdp = sdkp->device;
3179 	unsigned int min_xfer_bytes =
3180 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3181 
3182 	if (sdkp->min_xfer_blocks == 0)
3183 		return false;
3184 
3185 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3186 		sd_first_printk(KERN_WARNING, sdkp,
3187 				"Preferred minimum I/O size %u bytes not a " \
3188 				"multiple of physical block size (%u bytes)\n",
3189 				min_xfer_bytes, sdkp->physical_block_size);
3190 		sdkp->min_xfer_blocks = 0;
3191 		return false;
3192 	}
3193 
3194 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3195 			min_xfer_bytes);
3196 	return true;
3197 }
3198 
3199 /*
3200  * Determine the device's preferred I/O size for reads and writes
3201  * unless the reported value is unreasonably small, large, not a
3202  * multiple of the physical block size, or simply garbage.
3203  */
3204 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3205 				      unsigned int dev_max)
3206 {
3207 	struct scsi_device *sdp = sdkp->device;
3208 	unsigned int opt_xfer_bytes =
3209 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3210 	unsigned int min_xfer_bytes =
3211 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3212 
3213 	if (sdkp->opt_xfer_blocks == 0)
3214 		return false;
3215 
3216 	if (sdkp->opt_xfer_blocks > dev_max) {
3217 		sd_first_printk(KERN_WARNING, sdkp,
3218 				"Optimal transfer size %u logical blocks " \
3219 				"> dev_max (%u logical blocks)\n",
3220 				sdkp->opt_xfer_blocks, dev_max);
3221 		return false;
3222 	}
3223 
3224 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3225 		sd_first_printk(KERN_WARNING, sdkp,
3226 				"Optimal transfer size %u logical blocks " \
3227 				"> sd driver limit (%u logical blocks)\n",
3228 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3229 		return false;
3230 	}
3231 
3232 	if (opt_xfer_bytes < PAGE_SIZE) {
3233 		sd_first_printk(KERN_WARNING, sdkp,
3234 				"Optimal transfer size %u bytes < " \
3235 				"PAGE_SIZE (%u bytes)\n",
3236 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3237 		return false;
3238 	}
3239 
3240 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3241 		sd_first_printk(KERN_WARNING, sdkp,
3242 				"Optimal transfer size %u bytes not a " \
3243 				"multiple of preferred minimum block " \
3244 				"size (%u bytes)\n",
3245 				opt_xfer_bytes, min_xfer_bytes);
3246 		return false;
3247 	}
3248 
3249 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3250 		sd_first_printk(KERN_WARNING, sdkp,
3251 				"Optimal transfer size %u bytes not a " \
3252 				"multiple of physical block size (%u bytes)\n",
3253 				opt_xfer_bytes, sdkp->physical_block_size);
3254 		return false;
3255 	}
3256 
3257 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3258 			opt_xfer_bytes);
3259 	return true;
3260 }
3261 
3262 /**
3263  *	sd_revalidate_disk - called the first time a new disk is seen,
3264  *	performs disk spin up, read_capacity, etc.
3265  *	@disk: struct gendisk we care about
3266  **/
3267 static int sd_revalidate_disk(struct gendisk *disk)
3268 {
3269 	struct scsi_disk *sdkp = scsi_disk(disk);
3270 	struct scsi_device *sdp = sdkp->device;
3271 	struct request_queue *q = sdkp->disk->queue;
3272 	sector_t old_capacity = sdkp->capacity;
3273 	unsigned char *buffer;
3274 	unsigned int dev_max, rw_max;
3275 
3276 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3277 				      "sd_revalidate_disk\n"));
3278 
3279 	/*
3280 	 * If the device is offline, don't try and read capacity or any
3281 	 * of the other niceties.
3282 	 */
3283 	if (!scsi_device_online(sdp))
3284 		goto out;
3285 
3286 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3287 	if (!buffer) {
3288 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3289 			  "allocation failure.\n");
3290 		goto out;
3291 	}
3292 
3293 	sd_spinup_disk(sdkp);
3294 
3295 	/*
3296 	 * Without media there is no reason to ask; moreover, some devices
3297 	 * react badly if we do.
3298 	 */
3299 	if (sdkp->media_present) {
3300 		sd_read_capacity(sdkp, buffer);
3301 
3302 		/*
3303 		 * set the default to rotational.  All non-rotational devices
3304 		 * support the block characteristics VPD page, which will
3305 		 * cause this to be updated correctly and any device which
3306 		 * doesn't support it should be treated as rotational.
3307 		 */
3308 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3309 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3310 
3311 		if (scsi_device_supports_vpd(sdp)) {
3312 			sd_read_block_provisioning(sdkp);
3313 			sd_read_block_limits(sdkp);
3314 			sd_read_block_characteristics(sdkp);
3315 			sd_zbc_read_zones(sdkp, buffer);
3316 			sd_read_cpr(sdkp);
3317 		}
3318 
3319 		sd_print_capacity(sdkp, old_capacity);
3320 
3321 		sd_read_write_protect_flag(sdkp, buffer);
3322 		sd_read_cache_type(sdkp, buffer);
3323 		sd_read_app_tag_own(sdkp, buffer);
3324 		sd_read_write_same(sdkp, buffer);
3325 		sd_read_security(sdkp, buffer);
3326 		sd_config_protection(sdkp);
3327 	}
3328 
3329 	/*
3330 	 * We now have all cache related info, determine how we deal
3331 	 * with flush requests.
3332 	 */
3333 	sd_set_flush_flag(sdkp);
3334 
3335 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3336 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3337 
3338 	/* Some devices report a maximum block count for READ/WRITE requests. */
3339 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3340 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3341 
3342 	if (sd_validate_min_xfer_size(sdkp))
3343 		blk_queue_io_min(sdkp->disk->queue,
3344 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3345 	else
3346 		blk_queue_io_min(sdkp->disk->queue, 0);
3347 
3348 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3349 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3350 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3351 	} else {
3352 		q->limits.io_opt = 0;
3353 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3354 				      (sector_t)BLK_DEF_MAX_SECTORS);
3355 	}
3356 
3357 	/*
3358 	 * Limit default to SCSI host optimal sector limit if set. There may be
3359 	 * an impact on performance for when the size of a request exceeds this
3360 	 * host limit.
3361 	 */
3362 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3363 
3364 	/* Do not exceed controller limit */
3365 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3366 
3367 	/*
3368 	 * Only update max_sectors if previously unset or if the current value
3369 	 * exceeds the capabilities of the hardware.
3370 	 */
3371 	if (sdkp->first_scan ||
3372 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3373 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3374 		q->limits.max_sectors = rw_max;
3375 
3376 	sdkp->first_scan = 0;
3377 
3378 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3379 	sd_config_write_same(sdkp);
3380 	kfree(buffer);
3381 
3382 	/*
3383 	 * For a zoned drive, revalidating the zones can be done only once
3384 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3385 	 * capacity to 0.
3386 	 */
3387 	if (sd_zbc_revalidate_zones(sdkp))
3388 		set_capacity_and_notify(disk, 0);
3389 
3390  out:
3391 	return 0;
3392 }
3393 
3394 /**
3395  *	sd_unlock_native_capacity - unlock native capacity
3396  *	@disk: struct gendisk to set capacity for
3397  *
3398  *	Block layer calls this function if it detects that partitions
3399  *	on @disk reach beyond the end of the device.  If the SCSI host
3400  *	implements ->unlock_native_capacity() method, it's invoked to
3401  *	give it a chance to adjust the device capacity.
3402  *
3403  *	CONTEXT:
3404  *	Defined by block layer.  Might sleep.
3405  */
3406 static void sd_unlock_native_capacity(struct gendisk *disk)
3407 {
3408 	struct scsi_device *sdev = scsi_disk(disk)->device;
3409 
3410 	if (sdev->host->hostt->unlock_native_capacity)
3411 		sdev->host->hostt->unlock_native_capacity(sdev);
3412 }
3413 
3414 /**
3415  *	sd_format_disk_name - format disk name
3416  *	@prefix: name prefix - ie. "sd" for SCSI disks
3417  *	@index: index of the disk to format name for
3418  *	@buf: output buffer
3419  *	@buflen: length of the output buffer
3420  *
3421  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3422  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3423  *	which is followed by sdaaa.
3424  *
3425  *	This is basically 26 base counting with one extra 'nil' entry
3426  *	at the beginning from the second digit on and can be
3427  *	determined using similar method as 26 base conversion with the
3428  *	index shifted -1 after each digit is computed.
3429  *
3430  *	CONTEXT:
3431  *	Don't care.
3432  *
3433  *	RETURNS:
3434  *	0 on success, -errno on failure.
3435  */
3436 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3437 {
3438 	const int base = 'z' - 'a' + 1;
3439 	char *begin = buf + strlen(prefix);
3440 	char *end = buf + buflen;
3441 	char *p;
3442 	int unit;
3443 
3444 	p = end - 1;
3445 	*p = '\0';
3446 	unit = base;
3447 	do {
3448 		if (p == begin)
3449 			return -EINVAL;
3450 		*--p = 'a' + (index % unit);
3451 		index = (index / unit) - 1;
3452 	} while (index >= 0);
3453 
3454 	memmove(begin, p, end - p);
3455 	memcpy(buf, prefix, strlen(prefix));
3456 
3457 	return 0;
3458 }
3459 
3460 /**
3461  *	sd_probe - called during driver initialization and whenever a
3462  *	new scsi device is attached to the system. It is called once
3463  *	for each scsi device (not just disks) present.
3464  *	@dev: pointer to device object
3465  *
3466  *	Returns 0 if successful (or not interested in this scsi device
3467  *	(e.g. scanner)); 1 when there is an error.
3468  *
3469  *	Note: this function is invoked from the scsi mid-level.
3470  *	This function sets up the mapping between a given
3471  *	<host,channel,id,lun> (found in sdp) and new device name
3472  *	(e.g. /dev/sda). More precisely it is the block device major
3473  *	and minor number that is chosen here.
3474  *
3475  *	Assume sd_probe is not re-entrant (for time being)
3476  *	Also think about sd_probe() and sd_remove() running coincidentally.
3477  **/
3478 static int sd_probe(struct device *dev)
3479 {
3480 	struct scsi_device *sdp = to_scsi_device(dev);
3481 	struct scsi_disk *sdkp;
3482 	struct gendisk *gd;
3483 	int index;
3484 	int error;
3485 
3486 	scsi_autopm_get_device(sdp);
3487 	error = -ENODEV;
3488 	if (sdp->type != TYPE_DISK &&
3489 	    sdp->type != TYPE_ZBC &&
3490 	    sdp->type != TYPE_MOD &&
3491 	    sdp->type != TYPE_RBC)
3492 		goto out;
3493 
3494 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3495 		sdev_printk(KERN_WARNING, sdp,
3496 			    "Unsupported ZBC host-managed device.\n");
3497 		goto out;
3498 	}
3499 
3500 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3501 					"sd_probe\n"));
3502 
3503 	error = -ENOMEM;
3504 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3505 	if (!sdkp)
3506 		goto out;
3507 
3508 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3509 					 &sd_bio_compl_lkclass);
3510 	if (!gd)
3511 		goto out_free;
3512 
3513 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3514 	if (index < 0) {
3515 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3516 		goto out_put;
3517 	}
3518 
3519 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3520 	if (error) {
3521 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3522 		goto out_free_index;
3523 	}
3524 
3525 	sdkp->device = sdp;
3526 	sdkp->disk = gd;
3527 	sdkp->index = index;
3528 	sdkp->max_retries = SD_MAX_RETRIES;
3529 	atomic_set(&sdkp->openers, 0);
3530 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3531 
3532 	if (!sdp->request_queue->rq_timeout) {
3533 		if (sdp->type != TYPE_MOD)
3534 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3535 		else
3536 			blk_queue_rq_timeout(sdp->request_queue,
3537 					     SD_MOD_TIMEOUT);
3538 	}
3539 
3540 	device_initialize(&sdkp->disk_dev);
3541 	sdkp->disk_dev.parent = get_device(dev);
3542 	sdkp->disk_dev.class = &sd_disk_class;
3543 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3544 
3545 	error = device_add(&sdkp->disk_dev);
3546 	if (error) {
3547 		put_device(&sdkp->disk_dev);
3548 		goto out;
3549 	}
3550 
3551 	dev_set_drvdata(dev, sdkp);
3552 
3553 	gd->major = sd_major((index & 0xf0) >> 4);
3554 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3555 	gd->minors = SD_MINORS;
3556 
3557 	gd->fops = &sd_fops;
3558 	gd->private_data = sdkp;
3559 
3560 	/* defaults, until the device tells us otherwise */
3561 	sdp->sector_size = 512;
3562 	sdkp->capacity = 0;
3563 	sdkp->media_present = 1;
3564 	sdkp->write_prot = 0;
3565 	sdkp->cache_override = 0;
3566 	sdkp->WCE = 0;
3567 	sdkp->RCD = 0;
3568 	sdkp->ATO = 0;
3569 	sdkp->first_scan = 1;
3570 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3571 
3572 	sd_revalidate_disk(gd);
3573 
3574 	if (sdp->removable) {
3575 		gd->flags |= GENHD_FL_REMOVABLE;
3576 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3577 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3578 	}
3579 
3580 	blk_pm_runtime_init(sdp->request_queue, dev);
3581 	if (sdp->rpm_autosuspend) {
3582 		pm_runtime_set_autosuspend_delay(dev,
3583 			sdp->host->hostt->rpm_autosuspend_delay);
3584 	}
3585 
3586 	error = device_add_disk(dev, gd, NULL);
3587 	if (error) {
3588 		put_device(&sdkp->disk_dev);
3589 		put_disk(gd);
3590 		goto out;
3591 	}
3592 
3593 	if (sdkp->security) {
3594 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3595 		if (sdkp->opal_dev)
3596 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3597 	}
3598 
3599 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3600 		  sdp->removable ? "removable " : "");
3601 	scsi_autopm_put_device(sdp);
3602 
3603 	return 0;
3604 
3605  out_free_index:
3606 	ida_free(&sd_index_ida, index);
3607  out_put:
3608 	put_disk(gd);
3609  out_free:
3610 	kfree(sdkp);
3611  out:
3612 	scsi_autopm_put_device(sdp);
3613 	return error;
3614 }
3615 
3616 /**
3617  *	sd_remove - called whenever a scsi disk (previously recognized by
3618  *	sd_probe) is detached from the system. It is called (potentially
3619  *	multiple times) during sd module unload.
3620  *	@dev: pointer to device object
3621  *
3622  *	Note: this function is invoked from the scsi mid-level.
3623  *	This function potentially frees up a device name (e.g. /dev/sdc)
3624  *	that could be re-used by a subsequent sd_probe().
3625  *	This function is not called when the built-in sd driver is "exit-ed".
3626  **/
3627 static int sd_remove(struct device *dev)
3628 {
3629 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3630 
3631 	scsi_autopm_get_device(sdkp->device);
3632 
3633 	device_del(&sdkp->disk_dev);
3634 	del_gendisk(sdkp->disk);
3635 	sd_shutdown(dev);
3636 
3637 	put_disk(sdkp->disk);
3638 	return 0;
3639 }
3640 
3641 static void scsi_disk_release(struct device *dev)
3642 {
3643 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3644 
3645 	ida_free(&sd_index_ida, sdkp->index);
3646 	sd_zbc_free_zone_info(sdkp);
3647 	put_device(&sdkp->device->sdev_gendev);
3648 	free_opal_dev(sdkp->opal_dev);
3649 
3650 	kfree(sdkp);
3651 }
3652 
3653 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3654 {
3655 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3656 	struct scsi_sense_hdr sshdr;
3657 	const struct scsi_exec_args exec_args = {
3658 		.sshdr = &sshdr,
3659 		.req_flags = BLK_MQ_REQ_PM,
3660 	};
3661 	struct scsi_device *sdp = sdkp->device;
3662 	int res;
3663 
3664 	if (start)
3665 		cmd[4] |= 1;	/* START */
3666 
3667 	if (sdp->start_stop_pwr_cond)
3668 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3669 
3670 	if (!scsi_device_online(sdp))
3671 		return -ENODEV;
3672 
3673 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3674 			       sdkp->max_retries, &exec_args);
3675 	if (res) {
3676 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3677 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3678 			sd_print_sense_hdr(sdkp, &sshdr);
3679 			/* 0x3a is medium not present */
3680 			if (sshdr.asc == 0x3a)
3681 				res = 0;
3682 		}
3683 	}
3684 
3685 	/* SCSI error codes must not go to the generic layer */
3686 	if (res)
3687 		return -EIO;
3688 
3689 	return 0;
3690 }
3691 
3692 /*
3693  * Send a SYNCHRONIZE CACHE instruction down to the device through
3694  * the normal SCSI command structure.  Wait for the command to
3695  * complete.
3696  */
3697 static void sd_shutdown(struct device *dev)
3698 {
3699 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3700 
3701 	if (!sdkp)
3702 		return;         /* this can happen */
3703 
3704 	if (pm_runtime_suspended(dev))
3705 		return;
3706 
3707 	if (sdkp->WCE && sdkp->media_present) {
3708 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3709 		sd_sync_cache(sdkp, NULL);
3710 	}
3711 
3712 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3713 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3714 		sd_start_stop_device(sdkp, 0);
3715 	}
3716 }
3717 
3718 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3719 {
3720 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3721 	struct scsi_sense_hdr sshdr;
3722 	int ret = 0;
3723 
3724 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3725 		return 0;
3726 
3727 	if (sdkp->WCE && sdkp->media_present) {
3728 		if (!sdkp->device->silence_suspend)
3729 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3730 		ret = sd_sync_cache(sdkp, &sshdr);
3731 
3732 		if (ret) {
3733 			/* ignore OFFLINE device */
3734 			if (ret == -ENODEV)
3735 				return 0;
3736 
3737 			if (!scsi_sense_valid(&sshdr) ||
3738 			    sshdr.sense_key != ILLEGAL_REQUEST)
3739 				return ret;
3740 
3741 			/*
3742 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3743 			 * doesn't support sync. There's not much to do and
3744 			 * suspend shouldn't fail.
3745 			 */
3746 			ret = 0;
3747 		}
3748 	}
3749 
3750 	if (sdkp->device->manage_start_stop) {
3751 		if (!sdkp->device->silence_suspend)
3752 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3753 		/* an error is not worth aborting a system sleep */
3754 		ret = sd_start_stop_device(sdkp, 0);
3755 		if (ignore_stop_errors)
3756 			ret = 0;
3757 	}
3758 
3759 	return ret;
3760 }
3761 
3762 static int sd_suspend_system(struct device *dev)
3763 {
3764 	if (pm_runtime_suspended(dev))
3765 		return 0;
3766 
3767 	return sd_suspend_common(dev, true);
3768 }
3769 
3770 static int sd_suspend_runtime(struct device *dev)
3771 {
3772 	return sd_suspend_common(dev, false);
3773 }
3774 
3775 static int sd_resume(struct device *dev)
3776 {
3777 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3778 	int ret;
3779 
3780 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3781 		return 0;
3782 
3783 	if (!sdkp->device->manage_start_stop)
3784 		return 0;
3785 
3786 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3787 	ret = sd_start_stop_device(sdkp, 1);
3788 	if (!ret)
3789 		opal_unlock_from_suspend(sdkp->opal_dev);
3790 	return ret;
3791 }
3792 
3793 static int sd_resume_system(struct device *dev)
3794 {
3795 	if (pm_runtime_suspended(dev))
3796 		return 0;
3797 
3798 	return sd_resume(dev);
3799 }
3800 
3801 static int sd_resume_runtime(struct device *dev)
3802 {
3803 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3804 	struct scsi_device *sdp;
3805 
3806 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3807 		return 0;
3808 
3809 	sdp = sdkp->device;
3810 
3811 	if (sdp->ignore_media_change) {
3812 		/* clear the device's sense data */
3813 		static const u8 cmd[10] = { REQUEST_SENSE };
3814 		const struct scsi_exec_args exec_args = {
3815 			.req_flags = BLK_MQ_REQ_PM,
3816 		};
3817 
3818 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
3819 				     sdp->request_queue->rq_timeout, 1,
3820 				     &exec_args))
3821 			sd_printk(KERN_NOTICE, sdkp,
3822 				  "Failed to clear sense data\n");
3823 	}
3824 
3825 	return sd_resume(dev);
3826 }
3827 
3828 /**
3829  *	init_sd - entry point for this driver (both when built in or when
3830  *	a module).
3831  *
3832  *	Note: this function registers this driver with the scsi mid-level.
3833  **/
3834 static int __init init_sd(void)
3835 {
3836 	int majors = 0, i, err;
3837 
3838 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3839 
3840 	for (i = 0; i < SD_MAJORS; i++) {
3841 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3842 			continue;
3843 		majors++;
3844 	}
3845 
3846 	if (!majors)
3847 		return -ENODEV;
3848 
3849 	err = class_register(&sd_disk_class);
3850 	if (err)
3851 		goto err_out;
3852 
3853 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3854 	if (!sd_page_pool) {
3855 		printk(KERN_ERR "sd: can't init discard page pool\n");
3856 		err = -ENOMEM;
3857 		goto err_out_class;
3858 	}
3859 
3860 	err = scsi_register_driver(&sd_template.gendrv);
3861 	if (err)
3862 		goto err_out_driver;
3863 
3864 	return 0;
3865 
3866 err_out_driver:
3867 	mempool_destroy(sd_page_pool);
3868 err_out_class:
3869 	class_unregister(&sd_disk_class);
3870 err_out:
3871 	for (i = 0; i < SD_MAJORS; i++)
3872 		unregister_blkdev(sd_major(i), "sd");
3873 	return err;
3874 }
3875 
3876 /**
3877  *	exit_sd - exit point for this driver (when it is a module).
3878  *
3879  *	Note: this function unregisters this driver from the scsi mid-level.
3880  **/
3881 static void __exit exit_sd(void)
3882 {
3883 	int i;
3884 
3885 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3886 
3887 	scsi_unregister_driver(&sd_template.gendrv);
3888 	mempool_destroy(sd_page_pool);
3889 
3890 	class_unregister(&sd_disk_class);
3891 
3892 	for (i = 0; i < SD_MAJORS; i++)
3893 		unregister_blkdev(sd_major(i), "sd");
3894 }
3895 
3896 module_init(init_sd);
3897 module_exit(exit_sd);
3898 
3899 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3900 {
3901 	scsi_print_sense_hdr(sdkp->device,
3902 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3903 }
3904 
3905 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3906 {
3907 	const char *hb_string = scsi_hostbyte_string(result);
3908 
3909 	if (hb_string)
3910 		sd_printk(KERN_INFO, sdkp,
3911 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3912 			  hb_string ? hb_string : "invalid",
3913 			  "DRIVER_OK");
3914 	else
3915 		sd_printk(KERN_INFO, sdkp,
3916 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3917 			  msg, host_byte(result), "DRIVER_OK");
3918 }
3919