xref: /openbmc/linux/drivers/scsi/sd.c (revision e59e80cfef60366ce4dda96e9322a0b5947158a6)
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, 0, 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, unsigned int dld)
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 	cmd->cmnd[11] = dld & 0x07;
1052 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1053 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1054 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1055 
1056 	return BLK_STS_OK;
1057 }
1058 
1059 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1060 				       sector_t lba, unsigned int nr_blocks,
1061 				       unsigned char flags, unsigned int dld)
1062 {
1063 	cmd->cmd_len  = 16;
1064 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1065 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1066 	cmd->cmnd[14] = (dld & 0x03) << 6;
1067 	cmd->cmnd[15] = 0;
1068 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1069 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1070 
1071 	return BLK_STS_OK;
1072 }
1073 
1074 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1075 				       sector_t lba, unsigned int nr_blocks,
1076 				       unsigned char flags)
1077 {
1078 	cmd->cmd_len = 10;
1079 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1080 	cmd->cmnd[1] = flags;
1081 	cmd->cmnd[6] = 0;
1082 	cmd->cmnd[9] = 0;
1083 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1084 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1085 
1086 	return BLK_STS_OK;
1087 }
1088 
1089 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1090 				      sector_t lba, unsigned int nr_blocks,
1091 				      unsigned char flags)
1092 {
1093 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1094 	if (WARN_ON_ONCE(nr_blocks == 0))
1095 		return BLK_STS_IOERR;
1096 
1097 	if (unlikely(flags & 0x8)) {
1098 		/*
1099 		 * This happens only if this drive failed 10byte rw
1100 		 * command with ILLEGAL_REQUEST during operation and
1101 		 * thus turned off use_10_for_rw.
1102 		 */
1103 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1104 		return BLK_STS_IOERR;
1105 	}
1106 
1107 	cmd->cmd_len = 6;
1108 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1109 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1110 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1111 	cmd->cmnd[3] = lba & 0xff;
1112 	cmd->cmnd[4] = nr_blocks;
1113 	cmd->cmnd[5] = 0;
1114 
1115 	return BLK_STS_OK;
1116 }
1117 
1118 /*
1119  * Check if a command has a duration limit set. If it does, and the target
1120  * device supports CDL and the feature is enabled, return the limit
1121  * descriptor index to use. Return 0 (no limit) otherwise.
1122  */
1123 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1124 {
1125 	struct scsi_device *sdp = sdkp->device;
1126 	int hint;
1127 
1128 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1129 		return 0;
1130 
1131 	/*
1132 	 * Use "no limit" if the request ioprio does not specify a duration
1133 	 * limit hint.
1134 	 */
1135 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1136 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1137 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1138 		return 0;
1139 
1140 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1141 }
1142 
1143 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1144 {
1145 	struct request *rq = scsi_cmd_to_rq(cmd);
1146 	struct scsi_device *sdp = cmd->device;
1147 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1148 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1149 	sector_t threshold;
1150 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1151 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1152 	bool write = rq_data_dir(rq) == WRITE;
1153 	unsigned char protect, fua;
1154 	unsigned int dld;
1155 	blk_status_t ret;
1156 	unsigned int dif;
1157 	bool dix;
1158 
1159 	ret = scsi_alloc_sgtables(cmd);
1160 	if (ret != BLK_STS_OK)
1161 		return ret;
1162 
1163 	ret = BLK_STS_IOERR;
1164 	if (!scsi_device_online(sdp) || sdp->changed) {
1165 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1166 		goto fail;
1167 	}
1168 
1169 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1170 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1171 		goto fail;
1172 	}
1173 
1174 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1175 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1176 		goto fail;
1177 	}
1178 
1179 	/*
1180 	 * Some SD card readers can't handle accesses which touch the
1181 	 * last one or two logical blocks. Split accesses as needed.
1182 	 */
1183 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1184 
1185 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1186 		if (lba < threshold) {
1187 			/* Access up to the threshold but not beyond */
1188 			nr_blocks = threshold - lba;
1189 		} else {
1190 			/* Access only a single logical block */
1191 			nr_blocks = 1;
1192 		}
1193 	}
1194 
1195 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1196 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1197 		if (ret)
1198 			goto fail;
1199 	}
1200 
1201 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1202 	dix = scsi_prot_sg_count(cmd);
1203 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1204 	dld = sd_cdl_dld(sdkp, cmd);
1205 
1206 	if (dif || dix)
1207 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1208 	else
1209 		protect = 0;
1210 
1211 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1212 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1213 					 protect | fua, dld);
1214 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1215 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1216 					 protect | fua, dld);
1217 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1218 		   sdp->use_10_for_rw || protect) {
1219 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1220 					 protect | fua);
1221 	} else {
1222 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1223 					protect | fua);
1224 	}
1225 
1226 	if (unlikely(ret != BLK_STS_OK))
1227 		goto fail;
1228 
1229 	/*
1230 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1231 	 * host adapter, it's safe to assume that we can at least transfer
1232 	 * this many bytes between each connect / disconnect.
1233 	 */
1234 	cmd->transfersize = sdp->sector_size;
1235 	cmd->underflow = nr_blocks << 9;
1236 	cmd->allowed = sdkp->max_retries;
1237 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1238 
1239 	SCSI_LOG_HLQUEUE(1,
1240 			 scmd_printk(KERN_INFO, cmd,
1241 				     "%s: block=%llu, count=%d\n", __func__,
1242 				     (unsigned long long)blk_rq_pos(rq),
1243 				     blk_rq_sectors(rq)));
1244 	SCSI_LOG_HLQUEUE(2,
1245 			 scmd_printk(KERN_INFO, cmd,
1246 				     "%s %d/%u 512 byte blocks.\n",
1247 				     write ? "writing" : "reading", nr_blocks,
1248 				     blk_rq_sectors(rq)));
1249 
1250 	/*
1251 	 * This indicates that the command is ready from our end to be queued.
1252 	 */
1253 	return BLK_STS_OK;
1254 fail:
1255 	scsi_free_sgtables(cmd);
1256 	return ret;
1257 }
1258 
1259 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1260 {
1261 	struct request *rq = scsi_cmd_to_rq(cmd);
1262 
1263 	switch (req_op(rq)) {
1264 	case REQ_OP_DISCARD:
1265 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1266 		case SD_LBP_UNMAP:
1267 			return sd_setup_unmap_cmnd(cmd);
1268 		case SD_LBP_WS16:
1269 			return sd_setup_write_same16_cmnd(cmd, true);
1270 		case SD_LBP_WS10:
1271 			return sd_setup_write_same10_cmnd(cmd, true);
1272 		case SD_LBP_ZERO:
1273 			return sd_setup_write_same10_cmnd(cmd, false);
1274 		default:
1275 			return BLK_STS_TARGET;
1276 		}
1277 	case REQ_OP_WRITE_ZEROES:
1278 		return sd_setup_write_zeroes_cmnd(cmd);
1279 	case REQ_OP_FLUSH:
1280 		return sd_setup_flush_cmnd(cmd);
1281 	case REQ_OP_READ:
1282 	case REQ_OP_WRITE:
1283 	case REQ_OP_ZONE_APPEND:
1284 		return sd_setup_read_write_cmnd(cmd);
1285 	case REQ_OP_ZONE_RESET:
1286 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1287 						   false);
1288 	case REQ_OP_ZONE_RESET_ALL:
1289 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1290 						   true);
1291 	case REQ_OP_ZONE_OPEN:
1292 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1293 	case REQ_OP_ZONE_CLOSE:
1294 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1295 	case REQ_OP_ZONE_FINISH:
1296 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1297 	default:
1298 		WARN_ON_ONCE(1);
1299 		return BLK_STS_NOTSUPP;
1300 	}
1301 }
1302 
1303 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1304 {
1305 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1306 
1307 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1308 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1309 }
1310 
1311 static bool sd_need_revalidate(struct block_device *bdev,
1312 		struct scsi_disk *sdkp)
1313 {
1314 	if (sdkp->device->removable || sdkp->write_prot) {
1315 		if (bdev_check_media_change(bdev))
1316 			return true;
1317 	}
1318 
1319 	/*
1320 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1321 	 * nothing to do with partitions, BLKRRPART is used to force a full
1322 	 * revalidate after things like a format for historical reasons.
1323 	 */
1324 	return test_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1325 }
1326 
1327 /**
1328  *	sd_open - open a scsi disk device
1329  *	@bdev: Block device of the scsi disk to open
1330  *	@mode: FMODE_* mask
1331  *
1332  *	Returns 0 if successful. Returns a negated errno value in case
1333  *	of error.
1334  *
1335  *	Note: This can be called from a user context (e.g. fsck(1) )
1336  *	or from within the kernel (e.g. as a result of a mount(1) ).
1337  *	In the latter case @inode and @filp carry an abridged amount
1338  *	of information as noted above.
1339  *
1340  *	Locking: called with bdev->bd_disk->open_mutex held.
1341  **/
1342 static int sd_open(struct block_device *bdev, fmode_t mode)
1343 {
1344 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1345 	struct scsi_device *sdev = sdkp->device;
1346 	int retval;
1347 
1348 	if (scsi_device_get(sdev))
1349 		return -ENXIO;
1350 
1351 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1352 
1353 	/*
1354 	 * If the device is in error recovery, wait until it is done.
1355 	 * If the device is offline, then disallow any access to it.
1356 	 */
1357 	retval = -ENXIO;
1358 	if (!scsi_block_when_processing_errors(sdev))
1359 		goto error_out;
1360 
1361 	if (sd_need_revalidate(bdev, sdkp))
1362 		sd_revalidate_disk(bdev->bd_disk);
1363 
1364 	/*
1365 	 * If the drive is empty, just let the open fail.
1366 	 */
1367 	retval = -ENOMEDIUM;
1368 	if (sdev->removable && !sdkp->media_present && !(mode & FMODE_NDELAY))
1369 		goto error_out;
1370 
1371 	/*
1372 	 * If the device has the write protect tab set, have the open fail
1373 	 * if the user expects to be able to write to the thing.
1374 	 */
1375 	retval = -EROFS;
1376 	if (sdkp->write_prot && (mode & FMODE_WRITE))
1377 		goto error_out;
1378 
1379 	/*
1380 	 * It is possible that the disk changing stuff resulted in
1381 	 * the device being taken offline.  If this is the case,
1382 	 * report this to the user, and don't pretend that the
1383 	 * open actually succeeded.
1384 	 */
1385 	retval = -ENXIO;
1386 	if (!scsi_device_online(sdev))
1387 		goto error_out;
1388 
1389 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1390 		if (scsi_block_when_processing_errors(sdev))
1391 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1392 	}
1393 
1394 	return 0;
1395 
1396 error_out:
1397 	scsi_device_put(sdev);
1398 	return retval;
1399 }
1400 
1401 /**
1402  *	sd_release - invoked when the (last) close(2) is called on this
1403  *	scsi disk.
1404  *	@disk: disk to release
1405  *	@mode: FMODE_* mask
1406  *
1407  *	Returns 0.
1408  *
1409  *	Note: may block (uninterruptible) if error recovery is underway
1410  *	on this disk.
1411  *
1412  *	Locking: called with bdev->bd_disk->open_mutex held.
1413  **/
1414 static void sd_release(struct gendisk *disk, fmode_t mode)
1415 {
1416 	struct scsi_disk *sdkp = scsi_disk(disk);
1417 	struct scsi_device *sdev = sdkp->device;
1418 
1419 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1420 
1421 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1422 		if (scsi_block_when_processing_errors(sdev))
1423 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1424 	}
1425 
1426 	scsi_device_put(sdev);
1427 }
1428 
1429 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1430 {
1431 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1432 	struct scsi_device *sdp = sdkp->device;
1433 	struct Scsi_Host *host = sdp->host;
1434 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1435 	int diskinfo[4];
1436 
1437 	/* default to most commonly used values */
1438 	diskinfo[0] = 0x40;	/* 1 << 6 */
1439 	diskinfo[1] = 0x20;	/* 1 << 5 */
1440 	diskinfo[2] = capacity >> 11;
1441 
1442 	/* override with calculated, extended default, or driver values */
1443 	if (host->hostt->bios_param)
1444 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1445 	else
1446 		scsicam_bios_param(bdev, capacity, diskinfo);
1447 
1448 	geo->heads = diskinfo[0];
1449 	geo->sectors = diskinfo[1];
1450 	geo->cylinders = diskinfo[2];
1451 	return 0;
1452 }
1453 
1454 /**
1455  *	sd_ioctl - process an ioctl
1456  *	@bdev: target block device
1457  *	@mode: FMODE_* mask
1458  *	@cmd: ioctl command number
1459  *	@arg: this is third argument given to ioctl(2) system call.
1460  *	Often contains a pointer.
1461  *
1462  *	Returns 0 if successful (some ioctls return positive numbers on
1463  *	success as well). Returns a negated errno value in case of error.
1464  *
1465  *	Note: most ioctls are forward onto the block subsystem or further
1466  *	down in the scsi subsystem.
1467  **/
1468 static int sd_ioctl(struct block_device *bdev, fmode_t mode,
1469 		    unsigned int cmd, unsigned long arg)
1470 {
1471 	struct gendisk *disk = bdev->bd_disk;
1472 	struct scsi_disk *sdkp = scsi_disk(disk);
1473 	struct scsi_device *sdp = sdkp->device;
1474 	void __user *p = (void __user *)arg;
1475 	int error;
1476 
1477 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1478 				    "cmd=0x%x\n", disk->disk_name, cmd));
1479 
1480 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1481 		return -ENOIOCTLCMD;
1482 
1483 	/*
1484 	 * If we are in the middle of error recovery, don't let anyone
1485 	 * else try and use this device.  Also, if error recovery fails, it
1486 	 * may try and take the device offline, in which case all further
1487 	 * access to the device is prohibited.
1488 	 */
1489 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1490 			(mode & FMODE_NDELAY) != 0);
1491 	if (error)
1492 		return error;
1493 
1494 	if (is_sed_ioctl(cmd))
1495 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1496 	return scsi_ioctl(sdp, mode, cmd, p);
1497 }
1498 
1499 static void set_media_not_present(struct scsi_disk *sdkp)
1500 {
1501 	if (sdkp->media_present)
1502 		sdkp->device->changed = 1;
1503 
1504 	if (sdkp->device->removable) {
1505 		sdkp->media_present = 0;
1506 		sdkp->capacity = 0;
1507 	}
1508 }
1509 
1510 static int media_not_present(struct scsi_disk *sdkp,
1511 			     struct scsi_sense_hdr *sshdr)
1512 {
1513 	if (!scsi_sense_valid(sshdr))
1514 		return 0;
1515 
1516 	/* not invoked for commands that could return deferred errors */
1517 	switch (sshdr->sense_key) {
1518 	case UNIT_ATTENTION:
1519 	case NOT_READY:
1520 		/* medium not present */
1521 		if (sshdr->asc == 0x3A) {
1522 			set_media_not_present(sdkp);
1523 			return 1;
1524 		}
1525 	}
1526 	return 0;
1527 }
1528 
1529 /**
1530  *	sd_check_events - check media events
1531  *	@disk: kernel device descriptor
1532  *	@clearing: disk events currently being cleared
1533  *
1534  *	Returns mask of DISK_EVENT_*.
1535  *
1536  *	Note: this function is invoked from the block subsystem.
1537  **/
1538 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1539 {
1540 	struct scsi_disk *sdkp = disk->private_data;
1541 	struct scsi_device *sdp;
1542 	int retval;
1543 	bool disk_changed;
1544 
1545 	if (!sdkp)
1546 		return 0;
1547 
1548 	sdp = sdkp->device;
1549 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1550 
1551 	/*
1552 	 * If the device is offline, don't send any commands - just pretend as
1553 	 * if the command failed.  If the device ever comes back online, we
1554 	 * can deal with it then.  It is only because of unrecoverable errors
1555 	 * that we would ever take a device offline in the first place.
1556 	 */
1557 	if (!scsi_device_online(sdp)) {
1558 		set_media_not_present(sdkp);
1559 		goto out;
1560 	}
1561 
1562 	/*
1563 	 * Using TEST_UNIT_READY enables differentiation between drive with
1564 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1565 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1566 	 *
1567 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1568 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1569 	 * sd_revalidate() is called.
1570 	 */
1571 	if (scsi_block_when_processing_errors(sdp)) {
1572 		struct scsi_sense_hdr sshdr = { 0, };
1573 
1574 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1575 					      &sshdr);
1576 
1577 		/* failed to execute TUR, assume media not present */
1578 		if (retval < 0 || host_byte(retval)) {
1579 			set_media_not_present(sdkp);
1580 			goto out;
1581 		}
1582 
1583 		if (media_not_present(sdkp, &sshdr))
1584 			goto out;
1585 	}
1586 
1587 	/*
1588 	 * For removable scsi disk we have to recognise the presence
1589 	 * of a disk in the drive.
1590 	 */
1591 	if (!sdkp->media_present)
1592 		sdp->changed = 1;
1593 	sdkp->media_present = 1;
1594 out:
1595 	/*
1596 	 * sdp->changed is set under the following conditions:
1597 	 *
1598 	 *	Medium present state has changed in either direction.
1599 	 *	Device has indicated UNIT_ATTENTION.
1600 	 */
1601 	disk_changed = sdp->changed;
1602 	sdp->changed = 0;
1603 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1604 }
1605 
1606 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1607 {
1608 	int retries, res;
1609 	struct scsi_device *sdp = sdkp->device;
1610 	const int timeout = sdp->request_queue->rq_timeout
1611 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1612 	struct scsi_sense_hdr my_sshdr;
1613 	const struct scsi_exec_args exec_args = {
1614 		.req_flags = BLK_MQ_REQ_PM,
1615 		/* caller might not be interested in sense, but we need it */
1616 		.sshdr = sshdr ? : &my_sshdr,
1617 	};
1618 
1619 	if (!scsi_device_online(sdp))
1620 		return -ENODEV;
1621 
1622 	sshdr = exec_args.sshdr;
1623 
1624 	for (retries = 3; retries > 0; --retries) {
1625 		unsigned char cmd[16] = { 0 };
1626 
1627 		if (sdp->use_16_for_sync)
1628 			cmd[0] = SYNCHRONIZE_CACHE_16;
1629 		else
1630 			cmd[0] = SYNCHRONIZE_CACHE;
1631 		/*
1632 		 * Leave the rest of the command zero to indicate
1633 		 * flush everything.
1634 		 */
1635 		res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1636 				       timeout, sdkp->max_retries, &exec_args);
1637 		if (res == 0)
1638 			break;
1639 	}
1640 
1641 	if (res) {
1642 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1643 
1644 		if (res < 0)
1645 			return res;
1646 
1647 		if (scsi_status_is_check_condition(res) &&
1648 		    scsi_sense_valid(sshdr)) {
1649 			sd_print_sense_hdr(sdkp, sshdr);
1650 
1651 			/* we need to evaluate the error return  */
1652 			if (sshdr->asc == 0x3a ||	/* medium not present */
1653 			    sshdr->asc == 0x20 ||	/* invalid command */
1654 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1655 				/* this is no error here */
1656 				return 0;
1657 		}
1658 
1659 		switch (host_byte(res)) {
1660 		/* ignore errors due to racing a disconnection */
1661 		case DID_BAD_TARGET:
1662 		case DID_NO_CONNECT:
1663 			return 0;
1664 		/* signal the upper layer it might try again */
1665 		case DID_BUS_BUSY:
1666 		case DID_IMM_RETRY:
1667 		case DID_REQUEUE:
1668 		case DID_SOFT_ERROR:
1669 			return -EBUSY;
1670 		default:
1671 			return -EIO;
1672 		}
1673 	}
1674 	return 0;
1675 }
1676 
1677 static void sd_rescan(struct device *dev)
1678 {
1679 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1680 
1681 	sd_revalidate_disk(sdkp->disk);
1682 }
1683 
1684 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1685 		enum blk_unique_id type)
1686 {
1687 	struct scsi_device *sdev = scsi_disk(disk)->device;
1688 	const struct scsi_vpd *vpd;
1689 	const unsigned char *d;
1690 	int ret = -ENXIO, len;
1691 
1692 	rcu_read_lock();
1693 	vpd = rcu_dereference(sdev->vpd_pg83);
1694 	if (!vpd)
1695 		goto out_unlock;
1696 
1697 	ret = -EINVAL;
1698 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1699 		/* we only care about designators with LU association */
1700 		if (((d[1] >> 4) & 0x3) != 0x00)
1701 			continue;
1702 		if ((d[1] & 0xf) != type)
1703 			continue;
1704 
1705 		/*
1706 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1707 		 * keep looking as one with more entropy might still show up.
1708 		 */
1709 		len = d[3];
1710 		if (len != 8 && len != 12 && len != 16)
1711 			continue;
1712 		ret = len;
1713 		memcpy(id, d + 4, len);
1714 		if (len == 16)
1715 			break;
1716 	}
1717 out_unlock:
1718 	rcu_read_unlock();
1719 	return ret;
1720 }
1721 
1722 static char sd_pr_type(enum pr_type type)
1723 {
1724 	switch (type) {
1725 	case PR_WRITE_EXCLUSIVE:
1726 		return 0x01;
1727 	case PR_EXCLUSIVE_ACCESS:
1728 		return 0x03;
1729 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1730 		return 0x05;
1731 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1732 		return 0x06;
1733 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1734 		return 0x07;
1735 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1736 		return 0x08;
1737 	default:
1738 		return 0;
1739 	}
1740 };
1741 
1742 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1743 {
1744 	switch (host_byte(result)) {
1745 	case DID_TRANSPORT_MARGINAL:
1746 	case DID_TRANSPORT_DISRUPTED:
1747 	case DID_BUS_BUSY:
1748 		return PR_STS_RETRY_PATH_FAILURE;
1749 	case DID_NO_CONNECT:
1750 		return PR_STS_PATH_FAILED;
1751 	case DID_TRANSPORT_FAILFAST:
1752 		return PR_STS_PATH_FAST_FAILED;
1753 	}
1754 
1755 	switch (status_byte(result)) {
1756 	case SAM_STAT_RESERVATION_CONFLICT:
1757 		return PR_STS_RESERVATION_CONFLICT;
1758 	case SAM_STAT_CHECK_CONDITION:
1759 		if (!scsi_sense_valid(sshdr))
1760 			return PR_STS_IOERR;
1761 
1762 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1763 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1764 			return -EINVAL;
1765 
1766 		fallthrough;
1767 	default:
1768 		return PR_STS_IOERR;
1769 	}
1770 }
1771 
1772 static int sd_pr_command(struct block_device *bdev, u8 sa,
1773 		u64 key, u64 sa_key, u8 type, u8 flags)
1774 {
1775 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1776 	struct scsi_device *sdev = sdkp->device;
1777 	struct scsi_sense_hdr sshdr;
1778 	const struct scsi_exec_args exec_args = {
1779 		.sshdr = &sshdr,
1780 	};
1781 	int result;
1782 	u8 cmd[16] = { 0, };
1783 	u8 data[24] = { 0, };
1784 
1785 	cmd[0] = PERSISTENT_RESERVE_OUT;
1786 	cmd[1] = sa;
1787 	cmd[2] = type;
1788 	put_unaligned_be32(sizeof(data), &cmd[5]);
1789 
1790 	put_unaligned_be64(key, &data[0]);
1791 	put_unaligned_be64(sa_key, &data[8]);
1792 	data[20] = flags;
1793 
1794 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1795 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1796 				  &exec_args);
1797 
1798 	if (scsi_status_is_check_condition(result) &&
1799 	    scsi_sense_valid(&sshdr)) {
1800 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1801 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1802 	}
1803 
1804 	if (result <= 0)
1805 		return result;
1806 
1807 	return sd_scsi_to_pr_err(&sshdr, result);
1808 }
1809 
1810 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1811 		u32 flags)
1812 {
1813 	if (flags & ~PR_FL_IGNORE_KEY)
1814 		return -EOPNOTSUPP;
1815 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1816 			old_key, new_key, 0,
1817 			(1 << 0) /* APTPL */);
1818 }
1819 
1820 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1821 		u32 flags)
1822 {
1823 	if (flags)
1824 		return -EOPNOTSUPP;
1825 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1826 }
1827 
1828 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1829 {
1830 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1831 }
1832 
1833 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1834 		enum pr_type type, bool abort)
1835 {
1836 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1837 			     sd_pr_type(type), 0);
1838 }
1839 
1840 static int sd_pr_clear(struct block_device *bdev, u64 key)
1841 {
1842 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1843 }
1844 
1845 static const struct pr_ops sd_pr_ops = {
1846 	.pr_register	= sd_pr_register,
1847 	.pr_reserve	= sd_pr_reserve,
1848 	.pr_release	= sd_pr_release,
1849 	.pr_preempt	= sd_pr_preempt,
1850 	.pr_clear	= sd_pr_clear,
1851 };
1852 
1853 static void scsi_disk_free_disk(struct gendisk *disk)
1854 {
1855 	struct scsi_disk *sdkp = scsi_disk(disk);
1856 
1857 	put_device(&sdkp->disk_dev);
1858 }
1859 
1860 static const struct block_device_operations sd_fops = {
1861 	.owner			= THIS_MODULE,
1862 	.open			= sd_open,
1863 	.release		= sd_release,
1864 	.ioctl			= sd_ioctl,
1865 	.getgeo			= sd_getgeo,
1866 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1867 	.check_events		= sd_check_events,
1868 	.unlock_native_capacity	= sd_unlock_native_capacity,
1869 	.report_zones		= sd_zbc_report_zones,
1870 	.get_unique_id		= sd_get_unique_id,
1871 	.free_disk		= scsi_disk_free_disk,
1872 	.pr_ops			= &sd_pr_ops,
1873 };
1874 
1875 /**
1876  *	sd_eh_reset - reset error handling callback
1877  *	@scmd:		sd-issued command that has failed
1878  *
1879  *	This function is called by the SCSI midlayer before starting
1880  *	SCSI EH. When counting medium access failures we have to be
1881  *	careful to register it only only once per device and SCSI EH run;
1882  *	there might be several timed out commands which will cause the
1883  *	'max_medium_access_timeouts' counter to trigger after the first
1884  *	SCSI EH run already and set the device to offline.
1885  *	So this function resets the internal counter before starting SCSI EH.
1886  **/
1887 static void sd_eh_reset(struct scsi_cmnd *scmd)
1888 {
1889 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1890 
1891 	/* New SCSI EH run, reset gate variable */
1892 	sdkp->ignore_medium_access_errors = false;
1893 }
1894 
1895 /**
1896  *	sd_eh_action - error handling callback
1897  *	@scmd:		sd-issued command that has failed
1898  *	@eh_disp:	The recovery disposition suggested by the midlayer
1899  *
1900  *	This function is called by the SCSI midlayer upon completion of an
1901  *	error test command (currently TEST UNIT READY). The result of sending
1902  *	the eh command is passed in eh_disp.  We're looking for devices that
1903  *	fail medium access commands but are OK with non access commands like
1904  *	test unit ready (so wrongly see the device as having a successful
1905  *	recovery)
1906  **/
1907 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1908 {
1909 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1910 	struct scsi_device *sdev = scmd->device;
1911 
1912 	if (!scsi_device_online(sdev) ||
1913 	    !scsi_medium_access_command(scmd) ||
1914 	    host_byte(scmd->result) != DID_TIME_OUT ||
1915 	    eh_disp != SUCCESS)
1916 		return eh_disp;
1917 
1918 	/*
1919 	 * The device has timed out executing a medium access command.
1920 	 * However, the TEST UNIT READY command sent during error
1921 	 * handling completed successfully. Either the device is in the
1922 	 * process of recovering or has it suffered an internal failure
1923 	 * that prevents access to the storage medium.
1924 	 */
1925 	if (!sdkp->ignore_medium_access_errors) {
1926 		sdkp->medium_access_timed_out++;
1927 		sdkp->ignore_medium_access_errors = true;
1928 	}
1929 
1930 	/*
1931 	 * If the device keeps failing read/write commands but TEST UNIT
1932 	 * READY always completes successfully we assume that medium
1933 	 * access is no longer possible and take the device offline.
1934 	 */
1935 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1936 		scmd_printk(KERN_ERR, scmd,
1937 			    "Medium access timeout failure. Offlining disk!\n");
1938 		mutex_lock(&sdev->state_mutex);
1939 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1940 		mutex_unlock(&sdev->state_mutex);
1941 
1942 		return SUCCESS;
1943 	}
1944 
1945 	return eh_disp;
1946 }
1947 
1948 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1949 {
1950 	struct request *req = scsi_cmd_to_rq(scmd);
1951 	struct scsi_device *sdev = scmd->device;
1952 	unsigned int transferred, good_bytes;
1953 	u64 start_lba, end_lba, bad_lba;
1954 
1955 	/*
1956 	 * Some commands have a payload smaller than the device logical
1957 	 * block size (e.g. INQUIRY on a 4K disk).
1958 	 */
1959 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1960 		return 0;
1961 
1962 	/* Check if we have a 'bad_lba' information */
1963 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1964 				     SCSI_SENSE_BUFFERSIZE,
1965 				     &bad_lba))
1966 		return 0;
1967 
1968 	/*
1969 	 * If the bad lba was reported incorrectly, we have no idea where
1970 	 * the error is.
1971 	 */
1972 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1973 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1974 	if (bad_lba < start_lba || bad_lba >= end_lba)
1975 		return 0;
1976 
1977 	/*
1978 	 * resid is optional but mostly filled in.  When it's unused,
1979 	 * its value is zero, so we assume the whole buffer transferred
1980 	 */
1981 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1982 
1983 	/* This computation should always be done in terms of the
1984 	 * resolution of the device's medium.
1985 	 */
1986 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1987 
1988 	return min(good_bytes, transferred);
1989 }
1990 
1991 /**
1992  *	sd_done - bottom half handler: called when the lower level
1993  *	driver has completed (successfully or otherwise) a scsi command.
1994  *	@SCpnt: mid-level's per command structure.
1995  *
1996  *	Note: potentially run from within an ISR. Must not block.
1997  **/
1998 static int sd_done(struct scsi_cmnd *SCpnt)
1999 {
2000 	int result = SCpnt->result;
2001 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2002 	unsigned int sector_size = SCpnt->device->sector_size;
2003 	unsigned int resid;
2004 	struct scsi_sense_hdr sshdr;
2005 	struct request *req = scsi_cmd_to_rq(SCpnt);
2006 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2007 	int sense_valid = 0;
2008 	int sense_deferred = 0;
2009 
2010 	switch (req_op(req)) {
2011 	case REQ_OP_DISCARD:
2012 	case REQ_OP_WRITE_ZEROES:
2013 	case REQ_OP_ZONE_RESET:
2014 	case REQ_OP_ZONE_RESET_ALL:
2015 	case REQ_OP_ZONE_OPEN:
2016 	case REQ_OP_ZONE_CLOSE:
2017 	case REQ_OP_ZONE_FINISH:
2018 		if (!result) {
2019 			good_bytes = blk_rq_bytes(req);
2020 			scsi_set_resid(SCpnt, 0);
2021 		} else {
2022 			good_bytes = 0;
2023 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2024 		}
2025 		break;
2026 	default:
2027 		/*
2028 		 * In case of bogus fw or device, we could end up having
2029 		 * an unaligned partial completion. Check this here and force
2030 		 * alignment.
2031 		 */
2032 		resid = scsi_get_resid(SCpnt);
2033 		if (resid & (sector_size - 1)) {
2034 			sd_printk(KERN_INFO, sdkp,
2035 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2036 				resid, sector_size);
2037 			scsi_print_command(SCpnt);
2038 			resid = min(scsi_bufflen(SCpnt),
2039 				    round_up(resid, sector_size));
2040 			scsi_set_resid(SCpnt, resid);
2041 		}
2042 	}
2043 
2044 	if (result) {
2045 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2046 		if (sense_valid)
2047 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2048 	}
2049 	sdkp->medium_access_timed_out = 0;
2050 
2051 	if (!scsi_status_is_check_condition(result) &&
2052 	    (!sense_valid || sense_deferred))
2053 		goto out;
2054 
2055 	switch (sshdr.sense_key) {
2056 	case HARDWARE_ERROR:
2057 	case MEDIUM_ERROR:
2058 		good_bytes = sd_completed_bytes(SCpnt);
2059 		break;
2060 	case RECOVERED_ERROR:
2061 		good_bytes = scsi_bufflen(SCpnt);
2062 		break;
2063 	case NO_SENSE:
2064 		/* This indicates a false check condition, so ignore it.  An
2065 		 * unknown amount of data was transferred so treat it as an
2066 		 * error.
2067 		 */
2068 		SCpnt->result = 0;
2069 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2070 		break;
2071 	case ABORTED_COMMAND:
2072 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2073 			good_bytes = sd_completed_bytes(SCpnt);
2074 		break;
2075 	case ILLEGAL_REQUEST:
2076 		switch (sshdr.asc) {
2077 		case 0x10:	/* DIX: Host detected corruption */
2078 			good_bytes = sd_completed_bytes(SCpnt);
2079 			break;
2080 		case 0x20:	/* INVALID COMMAND OPCODE */
2081 		case 0x24:	/* INVALID FIELD IN CDB */
2082 			switch (SCpnt->cmnd[0]) {
2083 			case UNMAP:
2084 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2085 				break;
2086 			case WRITE_SAME_16:
2087 			case WRITE_SAME:
2088 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2089 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2090 				} else {
2091 					sdkp->device->no_write_same = 1;
2092 					sd_config_write_same(sdkp);
2093 					req->rq_flags |= RQF_QUIET;
2094 				}
2095 				break;
2096 			}
2097 		}
2098 		break;
2099 	default:
2100 		break;
2101 	}
2102 
2103  out:
2104 	if (sd_is_zoned(sdkp))
2105 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2106 
2107 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2108 					   "sd_done: completed %d of %d bytes\n",
2109 					   good_bytes, scsi_bufflen(SCpnt)));
2110 
2111 	return good_bytes;
2112 }
2113 
2114 /*
2115  * spinup disk - called only in sd_revalidate_disk()
2116  */
2117 static void
2118 sd_spinup_disk(struct scsi_disk *sdkp)
2119 {
2120 	unsigned char cmd[10];
2121 	unsigned long spintime_expire = 0;
2122 	int retries, spintime;
2123 	unsigned int the_result;
2124 	struct scsi_sense_hdr sshdr;
2125 	const struct scsi_exec_args exec_args = {
2126 		.sshdr = &sshdr,
2127 	};
2128 	int sense_valid = 0;
2129 
2130 	spintime = 0;
2131 
2132 	/* Spin up drives, as required.  Only do this at boot time */
2133 	/* Spinup needs to be done for module loads too. */
2134 	do {
2135 		retries = 0;
2136 
2137 		do {
2138 			bool media_was_present = sdkp->media_present;
2139 
2140 			cmd[0] = TEST_UNIT_READY;
2141 			memset((void *) &cmd[1], 0, 9);
2142 
2143 			the_result = scsi_execute_cmd(sdkp->device, cmd,
2144 						      REQ_OP_DRV_IN, NULL, 0,
2145 						      SD_TIMEOUT,
2146 						      sdkp->max_retries,
2147 						      &exec_args);
2148 
2149 			/*
2150 			 * If the drive has indicated to us that it
2151 			 * doesn't have any media in it, don't bother
2152 			 * with any more polling.
2153 			 */
2154 			if (media_not_present(sdkp, &sshdr)) {
2155 				if (media_was_present)
2156 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2157 				return;
2158 			}
2159 
2160 			if (the_result)
2161 				sense_valid = scsi_sense_valid(&sshdr);
2162 			retries++;
2163 		} while (retries < 3 &&
2164 			 (!scsi_status_is_good(the_result) ||
2165 			  (scsi_status_is_check_condition(the_result) &&
2166 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2167 
2168 		if (!scsi_status_is_check_condition(the_result)) {
2169 			/* no sense, TUR either succeeded or failed
2170 			 * with a status error */
2171 			if(!spintime && !scsi_status_is_good(the_result)) {
2172 				sd_print_result(sdkp, "Test Unit Ready failed",
2173 						the_result);
2174 			}
2175 			break;
2176 		}
2177 
2178 		/*
2179 		 * The device does not want the automatic start to be issued.
2180 		 */
2181 		if (sdkp->device->no_start_on_add)
2182 			break;
2183 
2184 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2185 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2186 				break;	/* manual intervention required */
2187 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2188 				break;	/* standby */
2189 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2190 				break;	/* unavailable */
2191 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2192 				break;	/* sanitize in progress */
2193 			/*
2194 			 * Issue command to spin up drive when not ready
2195 			 */
2196 			if (!spintime) {
2197 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2198 				cmd[0] = START_STOP;
2199 				cmd[1] = 1;	/* Return immediately */
2200 				memset((void *) &cmd[2], 0, 8);
2201 				cmd[4] = 1;	/* Start spin cycle */
2202 				if (sdkp->device->start_stop_pwr_cond)
2203 					cmd[4] |= 1 << 4;
2204 				scsi_execute_cmd(sdkp->device, cmd,
2205 						 REQ_OP_DRV_IN, NULL, 0,
2206 						 SD_TIMEOUT, sdkp->max_retries,
2207 						 &exec_args);
2208 				spintime_expire = jiffies + 100 * HZ;
2209 				spintime = 1;
2210 			}
2211 			/* Wait 1 second for next try */
2212 			msleep(1000);
2213 			printk(KERN_CONT ".");
2214 
2215 		/*
2216 		 * Wait for USB flash devices with slow firmware.
2217 		 * Yes, this sense key/ASC combination shouldn't
2218 		 * occur here.  It's characteristic of these devices.
2219 		 */
2220 		} else if (sense_valid &&
2221 				sshdr.sense_key == UNIT_ATTENTION &&
2222 				sshdr.asc == 0x28) {
2223 			if (!spintime) {
2224 				spintime_expire = jiffies + 5 * HZ;
2225 				spintime = 1;
2226 			}
2227 			/* Wait 1 second for next try */
2228 			msleep(1000);
2229 		} else {
2230 			/* we don't understand the sense code, so it's
2231 			 * probably pointless to loop */
2232 			if(!spintime) {
2233 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2234 				sd_print_sense_hdr(sdkp, &sshdr);
2235 			}
2236 			break;
2237 		}
2238 
2239 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2240 
2241 	if (spintime) {
2242 		if (scsi_status_is_good(the_result))
2243 			printk(KERN_CONT "ready\n");
2244 		else
2245 			printk(KERN_CONT "not responding...\n");
2246 	}
2247 }
2248 
2249 /*
2250  * Determine whether disk supports Data Integrity Field.
2251  */
2252 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2253 {
2254 	struct scsi_device *sdp = sdkp->device;
2255 	u8 type;
2256 
2257 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2258 		sdkp->protection_type = 0;
2259 		return 0;
2260 	}
2261 
2262 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2263 
2264 	if (type > T10_PI_TYPE3_PROTECTION) {
2265 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2266 			  " protection type %u. Disabling disk!\n",
2267 			  type);
2268 		sdkp->protection_type = 0;
2269 		return -ENODEV;
2270 	}
2271 
2272 	sdkp->protection_type = type;
2273 
2274 	return 0;
2275 }
2276 
2277 static void sd_config_protection(struct scsi_disk *sdkp)
2278 {
2279 	struct scsi_device *sdp = sdkp->device;
2280 
2281 	sd_dif_config_host(sdkp);
2282 
2283 	if (!sdkp->protection_type)
2284 		return;
2285 
2286 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2287 		sd_first_printk(KERN_NOTICE, sdkp,
2288 				"Disabling DIF Type %u protection\n",
2289 				sdkp->protection_type);
2290 		sdkp->protection_type = 0;
2291 	}
2292 
2293 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2294 			sdkp->protection_type);
2295 }
2296 
2297 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2298 			struct scsi_sense_hdr *sshdr, int sense_valid,
2299 			int the_result)
2300 {
2301 	if (sense_valid)
2302 		sd_print_sense_hdr(sdkp, sshdr);
2303 	else
2304 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2305 
2306 	/*
2307 	 * Set dirty bit for removable devices if not ready -
2308 	 * sometimes drives will not report this properly.
2309 	 */
2310 	if (sdp->removable &&
2311 	    sense_valid && sshdr->sense_key == NOT_READY)
2312 		set_media_not_present(sdkp);
2313 
2314 	/*
2315 	 * We used to set media_present to 0 here to indicate no media
2316 	 * in the drive, but some drives fail read capacity even with
2317 	 * media present, so we can't do that.
2318 	 */
2319 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2320 }
2321 
2322 #define RC16_LEN 32
2323 #if RC16_LEN > SD_BUF_SIZE
2324 #error RC16_LEN must not be more than SD_BUF_SIZE
2325 #endif
2326 
2327 #define READ_CAPACITY_RETRIES_ON_RESET	10
2328 
2329 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2330 						unsigned char *buffer)
2331 {
2332 	unsigned char cmd[16];
2333 	struct scsi_sense_hdr sshdr;
2334 	const struct scsi_exec_args exec_args = {
2335 		.sshdr = &sshdr,
2336 	};
2337 	int sense_valid = 0;
2338 	int the_result;
2339 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2340 	unsigned int alignment;
2341 	unsigned long long lba;
2342 	unsigned sector_size;
2343 
2344 	if (sdp->no_read_capacity_16)
2345 		return -EINVAL;
2346 
2347 	do {
2348 		memset(cmd, 0, 16);
2349 		cmd[0] = SERVICE_ACTION_IN_16;
2350 		cmd[1] = SAI_READ_CAPACITY_16;
2351 		cmd[13] = RC16_LEN;
2352 		memset(buffer, 0, RC16_LEN);
2353 
2354 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2355 					      buffer, RC16_LEN, SD_TIMEOUT,
2356 					      sdkp->max_retries, &exec_args);
2357 
2358 		if (media_not_present(sdkp, &sshdr))
2359 			return -ENODEV;
2360 
2361 		if (the_result > 0) {
2362 			sense_valid = scsi_sense_valid(&sshdr);
2363 			if (sense_valid &&
2364 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2365 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2366 			    sshdr.ascq == 0x00)
2367 				/* Invalid Command Operation Code or
2368 				 * Invalid Field in CDB, just retry
2369 				 * silently with RC10 */
2370 				return -EINVAL;
2371 			if (sense_valid &&
2372 			    sshdr.sense_key == UNIT_ATTENTION &&
2373 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2374 				/* Device reset might occur several times,
2375 				 * give it one more chance */
2376 				if (--reset_retries > 0)
2377 					continue;
2378 		}
2379 		retries--;
2380 
2381 	} while (the_result && retries);
2382 
2383 	if (the_result) {
2384 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2385 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2386 		return -EINVAL;
2387 	}
2388 
2389 	sector_size = get_unaligned_be32(&buffer[8]);
2390 	lba = get_unaligned_be64(&buffer[0]);
2391 
2392 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2393 		sdkp->capacity = 0;
2394 		return -ENODEV;
2395 	}
2396 
2397 	/* Logical blocks per physical block exponent */
2398 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2399 
2400 	/* RC basis */
2401 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2402 
2403 	/* Lowest aligned logical block */
2404 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2405 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2406 	if (alignment && sdkp->first_scan)
2407 		sd_printk(KERN_NOTICE, sdkp,
2408 			  "physical block alignment offset: %u\n", alignment);
2409 
2410 	if (buffer[14] & 0x80) { /* LBPME */
2411 		sdkp->lbpme = 1;
2412 
2413 		if (buffer[14] & 0x40) /* LBPRZ */
2414 			sdkp->lbprz = 1;
2415 
2416 		sd_config_discard(sdkp, SD_LBP_WS16);
2417 	}
2418 
2419 	sdkp->capacity = lba + 1;
2420 	return sector_size;
2421 }
2422 
2423 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2424 						unsigned char *buffer)
2425 {
2426 	unsigned char cmd[16];
2427 	struct scsi_sense_hdr sshdr;
2428 	const struct scsi_exec_args exec_args = {
2429 		.sshdr = &sshdr,
2430 	};
2431 	int sense_valid = 0;
2432 	int the_result;
2433 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2434 	sector_t lba;
2435 	unsigned sector_size;
2436 
2437 	do {
2438 		cmd[0] = READ_CAPACITY;
2439 		memset(&cmd[1], 0, 9);
2440 		memset(buffer, 0, 8);
2441 
2442 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2443 					      8, SD_TIMEOUT, sdkp->max_retries,
2444 					      &exec_args);
2445 
2446 		if (media_not_present(sdkp, &sshdr))
2447 			return -ENODEV;
2448 
2449 		if (the_result > 0) {
2450 			sense_valid = scsi_sense_valid(&sshdr);
2451 			if (sense_valid &&
2452 			    sshdr.sense_key == UNIT_ATTENTION &&
2453 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2454 				/* Device reset might occur several times,
2455 				 * give it one more chance */
2456 				if (--reset_retries > 0)
2457 					continue;
2458 		}
2459 		retries--;
2460 
2461 	} while (the_result && retries);
2462 
2463 	if (the_result) {
2464 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2465 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2466 		return -EINVAL;
2467 	}
2468 
2469 	sector_size = get_unaligned_be32(&buffer[4]);
2470 	lba = get_unaligned_be32(&buffer[0]);
2471 
2472 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2473 		/* Some buggy (usb cardreader) devices return an lba of
2474 		   0xffffffff when the want to report a size of 0 (with
2475 		   which they really mean no media is present) */
2476 		sdkp->capacity = 0;
2477 		sdkp->physical_block_size = sector_size;
2478 		return sector_size;
2479 	}
2480 
2481 	sdkp->capacity = lba + 1;
2482 	sdkp->physical_block_size = sector_size;
2483 	return sector_size;
2484 }
2485 
2486 static int sd_try_rc16_first(struct scsi_device *sdp)
2487 {
2488 	if (sdp->host->max_cmd_len < 16)
2489 		return 0;
2490 	if (sdp->try_rc_10_first)
2491 		return 0;
2492 	if (sdp->scsi_level > SCSI_SPC_2)
2493 		return 1;
2494 	if (scsi_device_protection(sdp))
2495 		return 1;
2496 	return 0;
2497 }
2498 
2499 /*
2500  * read disk capacity
2501  */
2502 static void
2503 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2504 {
2505 	int sector_size;
2506 	struct scsi_device *sdp = sdkp->device;
2507 
2508 	if (sd_try_rc16_first(sdp)) {
2509 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2510 		if (sector_size == -EOVERFLOW)
2511 			goto got_data;
2512 		if (sector_size == -ENODEV)
2513 			return;
2514 		if (sector_size < 0)
2515 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2516 		if (sector_size < 0)
2517 			return;
2518 	} else {
2519 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2520 		if (sector_size == -EOVERFLOW)
2521 			goto got_data;
2522 		if (sector_size < 0)
2523 			return;
2524 		if ((sizeof(sdkp->capacity) > 4) &&
2525 		    (sdkp->capacity > 0xffffffffULL)) {
2526 			int old_sector_size = sector_size;
2527 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2528 					"Trying to use READ CAPACITY(16).\n");
2529 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2530 			if (sector_size < 0) {
2531 				sd_printk(KERN_NOTICE, sdkp,
2532 					"Using 0xffffffff as device size\n");
2533 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2534 				sector_size = old_sector_size;
2535 				goto got_data;
2536 			}
2537 			/* Remember that READ CAPACITY(16) succeeded */
2538 			sdp->try_rc_10_first = 0;
2539 		}
2540 	}
2541 
2542 	/* Some devices are known to return the total number of blocks,
2543 	 * not the highest block number.  Some devices have versions
2544 	 * which do this and others which do not.  Some devices we might
2545 	 * suspect of doing this but we don't know for certain.
2546 	 *
2547 	 * If we know the reported capacity is wrong, decrement it.  If
2548 	 * we can only guess, then assume the number of blocks is even
2549 	 * (usually true but not always) and err on the side of lowering
2550 	 * the capacity.
2551 	 */
2552 	if (sdp->fix_capacity ||
2553 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2554 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2555 				"from its reported value: %llu\n",
2556 				(unsigned long long) sdkp->capacity);
2557 		--sdkp->capacity;
2558 	}
2559 
2560 got_data:
2561 	if (sector_size == 0) {
2562 		sector_size = 512;
2563 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2564 			  "assuming 512.\n");
2565 	}
2566 
2567 	if (sector_size != 512 &&
2568 	    sector_size != 1024 &&
2569 	    sector_size != 2048 &&
2570 	    sector_size != 4096) {
2571 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2572 			  sector_size);
2573 		/*
2574 		 * The user might want to re-format the drive with
2575 		 * a supported sectorsize.  Once this happens, it
2576 		 * would be relatively trivial to set the thing up.
2577 		 * For this reason, we leave the thing in the table.
2578 		 */
2579 		sdkp->capacity = 0;
2580 		/*
2581 		 * set a bogus sector size so the normal read/write
2582 		 * logic in the block layer will eventually refuse any
2583 		 * request on this device without tripping over power
2584 		 * of two sector size assumptions
2585 		 */
2586 		sector_size = 512;
2587 	}
2588 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2589 	blk_queue_physical_block_size(sdp->request_queue,
2590 				      sdkp->physical_block_size);
2591 	sdkp->device->sector_size = sector_size;
2592 
2593 	if (sdkp->capacity > 0xffffffff)
2594 		sdp->use_16_for_rw = 1;
2595 
2596 }
2597 
2598 /*
2599  * Print disk capacity
2600  */
2601 static void
2602 sd_print_capacity(struct scsi_disk *sdkp,
2603 		  sector_t old_capacity)
2604 {
2605 	int sector_size = sdkp->device->sector_size;
2606 	char cap_str_2[10], cap_str_10[10];
2607 
2608 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2609 		return;
2610 
2611 	string_get_size(sdkp->capacity, sector_size,
2612 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2613 	string_get_size(sdkp->capacity, sector_size,
2614 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2615 
2616 	sd_printk(KERN_NOTICE, sdkp,
2617 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2618 		  (unsigned long long)sdkp->capacity,
2619 		  sector_size, cap_str_10, cap_str_2);
2620 
2621 	if (sdkp->physical_block_size != sector_size)
2622 		sd_printk(KERN_NOTICE, sdkp,
2623 			  "%u-byte physical blocks\n",
2624 			  sdkp->physical_block_size);
2625 }
2626 
2627 /* called with buffer of length 512 */
2628 static inline int
2629 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2630 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2631 		 struct scsi_sense_hdr *sshdr)
2632 {
2633 	/*
2634 	 * If we must use MODE SENSE(10), make sure that the buffer length
2635 	 * is at least 8 bytes so that the mode sense header fits.
2636 	 */
2637 	if (sdkp->device->use_10_for_ms && len < 8)
2638 		len = 8;
2639 
2640 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2641 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2642 }
2643 
2644 /*
2645  * read write protect setting, if possible - called only in sd_revalidate_disk()
2646  * called with buffer of length SD_BUF_SIZE
2647  */
2648 static void
2649 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2650 {
2651 	int res;
2652 	struct scsi_device *sdp = sdkp->device;
2653 	struct scsi_mode_data data;
2654 	int old_wp = sdkp->write_prot;
2655 
2656 	set_disk_ro(sdkp->disk, 0);
2657 	if (sdp->skip_ms_page_3f) {
2658 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2659 		return;
2660 	}
2661 
2662 	if (sdp->use_192_bytes_for_3f) {
2663 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2664 	} else {
2665 		/*
2666 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2667 		 * We have to start carefully: some devices hang if we ask
2668 		 * for more than is available.
2669 		 */
2670 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2671 
2672 		/*
2673 		 * Second attempt: ask for page 0 When only page 0 is
2674 		 * implemented, a request for page 3F may return Sense Key
2675 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2676 		 * CDB.
2677 		 */
2678 		if (res < 0)
2679 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2680 
2681 		/*
2682 		 * Third attempt: ask 255 bytes, as we did earlier.
2683 		 */
2684 		if (res < 0)
2685 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2686 					       &data, NULL);
2687 	}
2688 
2689 	if (res < 0) {
2690 		sd_first_printk(KERN_WARNING, sdkp,
2691 			  "Test WP failed, assume Write Enabled\n");
2692 	} else {
2693 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2694 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2695 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2696 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2697 				  sdkp->write_prot ? "on" : "off");
2698 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2699 		}
2700 	}
2701 }
2702 
2703 /*
2704  * sd_read_cache_type - called only from sd_revalidate_disk()
2705  * called with buffer of length SD_BUF_SIZE
2706  */
2707 static void
2708 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2709 {
2710 	int len = 0, res;
2711 	struct scsi_device *sdp = sdkp->device;
2712 
2713 	int dbd;
2714 	int modepage;
2715 	int first_len;
2716 	struct scsi_mode_data data;
2717 	struct scsi_sense_hdr sshdr;
2718 	int old_wce = sdkp->WCE;
2719 	int old_rcd = sdkp->RCD;
2720 	int old_dpofua = sdkp->DPOFUA;
2721 
2722 
2723 	if (sdkp->cache_override)
2724 		return;
2725 
2726 	first_len = 4;
2727 	if (sdp->skip_ms_page_8) {
2728 		if (sdp->type == TYPE_RBC)
2729 			goto defaults;
2730 		else {
2731 			if (sdp->skip_ms_page_3f)
2732 				goto defaults;
2733 			modepage = 0x3F;
2734 			if (sdp->use_192_bytes_for_3f)
2735 				first_len = 192;
2736 			dbd = 0;
2737 		}
2738 	} else if (sdp->type == TYPE_RBC) {
2739 		modepage = 6;
2740 		dbd = 8;
2741 	} else {
2742 		modepage = 8;
2743 		dbd = 0;
2744 	}
2745 
2746 	/* cautiously ask */
2747 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2748 			&data, &sshdr);
2749 
2750 	if (res < 0)
2751 		goto bad_sense;
2752 
2753 	if (!data.header_length) {
2754 		modepage = 6;
2755 		first_len = 0;
2756 		sd_first_printk(KERN_ERR, sdkp,
2757 				"Missing header in MODE_SENSE response\n");
2758 	}
2759 
2760 	/* that went OK, now ask for the proper length */
2761 	len = data.length;
2762 
2763 	/*
2764 	 * We're only interested in the first three bytes, actually.
2765 	 * But the data cache page is defined for the first 20.
2766 	 */
2767 	if (len < 3)
2768 		goto bad_sense;
2769 	else if (len > SD_BUF_SIZE) {
2770 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2771 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2772 		len = SD_BUF_SIZE;
2773 	}
2774 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2775 		len = 192;
2776 
2777 	/* Get the data */
2778 	if (len > first_len)
2779 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2780 				&data, &sshdr);
2781 
2782 	if (!res) {
2783 		int offset = data.header_length + data.block_descriptor_length;
2784 
2785 		while (offset < len) {
2786 			u8 page_code = buffer[offset] & 0x3F;
2787 			u8 spf       = buffer[offset] & 0x40;
2788 
2789 			if (page_code == 8 || page_code == 6) {
2790 				/* We're interested only in the first 3 bytes.
2791 				 */
2792 				if (len - offset <= 2) {
2793 					sd_first_printk(KERN_ERR, sdkp,
2794 						"Incomplete mode parameter "
2795 							"data\n");
2796 					goto defaults;
2797 				} else {
2798 					modepage = page_code;
2799 					goto Page_found;
2800 				}
2801 			} else {
2802 				/* Go to the next page */
2803 				if (spf && len - offset > 3)
2804 					offset += 4 + (buffer[offset+2] << 8) +
2805 						buffer[offset+3];
2806 				else if (!spf && len - offset > 1)
2807 					offset += 2 + buffer[offset+1];
2808 				else {
2809 					sd_first_printk(KERN_ERR, sdkp,
2810 							"Incomplete mode "
2811 							"parameter data\n");
2812 					goto defaults;
2813 				}
2814 			}
2815 		}
2816 
2817 		sd_first_printk(KERN_WARNING, sdkp,
2818 				"No Caching mode page found\n");
2819 		goto defaults;
2820 
2821 	Page_found:
2822 		if (modepage == 8) {
2823 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2824 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2825 		} else {
2826 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2827 			sdkp->RCD = 0;
2828 		}
2829 
2830 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2831 		if (sdp->broken_fua) {
2832 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2833 			sdkp->DPOFUA = 0;
2834 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2835 			   !sdkp->device->use_16_for_rw) {
2836 			sd_first_printk(KERN_NOTICE, sdkp,
2837 				  "Uses READ/WRITE(6), disabling FUA\n");
2838 			sdkp->DPOFUA = 0;
2839 		}
2840 
2841 		/* No cache flush allowed for write protected devices */
2842 		if (sdkp->WCE && sdkp->write_prot)
2843 			sdkp->WCE = 0;
2844 
2845 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2846 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2847 			sd_printk(KERN_NOTICE, sdkp,
2848 				  "Write cache: %s, read cache: %s, %s\n",
2849 				  sdkp->WCE ? "enabled" : "disabled",
2850 				  sdkp->RCD ? "disabled" : "enabled",
2851 				  sdkp->DPOFUA ? "supports DPO and FUA"
2852 				  : "doesn't support DPO or FUA");
2853 
2854 		return;
2855 	}
2856 
2857 bad_sense:
2858 	if (scsi_sense_valid(&sshdr) &&
2859 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2860 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2861 		/* Invalid field in CDB */
2862 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2863 	else
2864 		sd_first_printk(KERN_ERR, sdkp,
2865 				"Asking for cache data failed\n");
2866 
2867 defaults:
2868 	if (sdp->wce_default_on) {
2869 		sd_first_printk(KERN_NOTICE, sdkp,
2870 				"Assuming drive cache: write back\n");
2871 		sdkp->WCE = 1;
2872 	} else {
2873 		sd_first_printk(KERN_WARNING, sdkp,
2874 				"Assuming drive cache: write through\n");
2875 		sdkp->WCE = 0;
2876 	}
2877 	sdkp->RCD = 0;
2878 	sdkp->DPOFUA = 0;
2879 }
2880 
2881 /*
2882  * The ATO bit indicates whether the DIF application tag is available
2883  * for use by the operating system.
2884  */
2885 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2886 {
2887 	int res, offset;
2888 	struct scsi_device *sdp = sdkp->device;
2889 	struct scsi_mode_data data;
2890 	struct scsi_sense_hdr sshdr;
2891 
2892 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2893 		return;
2894 
2895 	if (sdkp->protection_type == 0)
2896 		return;
2897 
2898 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
2899 			      sdkp->max_retries, &data, &sshdr);
2900 
2901 	if (res < 0 || !data.header_length ||
2902 	    data.length < 6) {
2903 		sd_first_printk(KERN_WARNING, sdkp,
2904 			  "getting Control mode page failed, assume no ATO\n");
2905 
2906 		if (scsi_sense_valid(&sshdr))
2907 			sd_print_sense_hdr(sdkp, &sshdr);
2908 
2909 		return;
2910 	}
2911 
2912 	offset = data.header_length + data.block_descriptor_length;
2913 
2914 	if ((buffer[offset] & 0x3f) != 0x0a) {
2915 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2916 		return;
2917 	}
2918 
2919 	if ((buffer[offset + 5] & 0x80) == 0)
2920 		return;
2921 
2922 	sdkp->ATO = 1;
2923 
2924 	return;
2925 }
2926 
2927 /**
2928  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2929  * @sdkp: disk to query
2930  */
2931 static void sd_read_block_limits(struct scsi_disk *sdkp)
2932 {
2933 	struct scsi_vpd *vpd;
2934 
2935 	rcu_read_lock();
2936 
2937 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2938 	if (!vpd || vpd->len < 16)
2939 		goto out;
2940 
2941 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2942 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2943 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2944 
2945 	if (vpd->len >= 64) {
2946 		unsigned int lba_count, desc_count;
2947 
2948 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2949 
2950 		if (!sdkp->lbpme)
2951 			goto out;
2952 
2953 		lba_count = get_unaligned_be32(&vpd->data[20]);
2954 		desc_count = get_unaligned_be32(&vpd->data[24]);
2955 
2956 		if (lba_count && desc_count)
2957 			sdkp->max_unmap_blocks = lba_count;
2958 
2959 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2960 
2961 		if (vpd->data[32] & 0x80)
2962 			sdkp->unmap_alignment =
2963 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2964 
2965 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2966 
2967 			if (sdkp->max_unmap_blocks)
2968 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2969 			else
2970 				sd_config_discard(sdkp, SD_LBP_WS16);
2971 
2972 		} else {	/* LBP VPD page tells us what to use */
2973 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2974 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2975 			else if (sdkp->lbpws)
2976 				sd_config_discard(sdkp, SD_LBP_WS16);
2977 			else if (sdkp->lbpws10)
2978 				sd_config_discard(sdkp, SD_LBP_WS10);
2979 			else
2980 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2981 		}
2982 	}
2983 
2984  out:
2985 	rcu_read_unlock();
2986 }
2987 
2988 /**
2989  * sd_read_block_characteristics - Query block dev. characteristics
2990  * @sdkp: disk to query
2991  */
2992 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2993 {
2994 	struct request_queue *q = sdkp->disk->queue;
2995 	struct scsi_vpd *vpd;
2996 	u16 rot;
2997 	u8 zoned;
2998 
2999 	rcu_read_lock();
3000 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3001 
3002 	if (!vpd || vpd->len < 8) {
3003 		rcu_read_unlock();
3004 	        return;
3005 	}
3006 
3007 	rot = get_unaligned_be16(&vpd->data[4]);
3008 	zoned = (vpd->data[8] >> 4) & 3;
3009 	rcu_read_unlock();
3010 
3011 	if (rot == 1) {
3012 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3013 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3014 	}
3015 
3016 	if (sdkp->device->type == TYPE_ZBC) {
3017 		/*
3018 		 * Host-managed: Per ZBC and ZAC specifications, writes in
3019 		 * sequential write required zones of host-managed devices must
3020 		 * be aligned to the device physical block size.
3021 		 */
3022 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
3023 		blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3024 	} else {
3025 		sdkp->zoned = zoned;
3026 		if (sdkp->zoned == 1) {
3027 			/* Host-aware */
3028 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
3029 		} else {
3030 			/* Regular disk or drive managed disk */
3031 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3032 		}
3033 	}
3034 
3035 	if (!sdkp->first_scan)
3036 		return;
3037 
3038 	if (blk_queue_is_zoned(q)) {
3039 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3040 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3041 	} else {
3042 		if (sdkp->zoned == 1)
3043 			sd_printk(KERN_NOTICE, sdkp,
3044 				  "Host-aware SMR disk used as regular disk\n");
3045 		else if (sdkp->zoned == 2)
3046 			sd_printk(KERN_NOTICE, sdkp,
3047 				  "Drive-managed SMR disk\n");
3048 	}
3049 }
3050 
3051 /**
3052  * sd_read_block_provisioning - Query provisioning VPD page
3053  * @sdkp: disk to query
3054  */
3055 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3056 {
3057 	struct scsi_vpd *vpd;
3058 
3059 	if (sdkp->lbpme == 0)
3060 		return;
3061 
3062 	rcu_read_lock();
3063 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3064 
3065 	if (!vpd || vpd->len < 8) {
3066 		rcu_read_unlock();
3067 		return;
3068 	}
3069 
3070 	sdkp->lbpvpd	= 1;
3071 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3072 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3073 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3074 	rcu_read_unlock();
3075 }
3076 
3077 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3078 {
3079 	struct scsi_device *sdev = sdkp->device;
3080 
3081 	if (sdev->host->no_write_same) {
3082 		sdev->no_write_same = 1;
3083 
3084 		return;
3085 	}
3086 
3087 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3088 		struct scsi_vpd *vpd;
3089 
3090 		sdev->no_report_opcodes = 1;
3091 
3092 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3093 		 * CODES is unsupported and the device has an ATA
3094 		 * Information VPD page (SAT).
3095 		 */
3096 		rcu_read_lock();
3097 		vpd = rcu_dereference(sdev->vpd_pg89);
3098 		if (vpd)
3099 			sdev->no_write_same = 1;
3100 		rcu_read_unlock();
3101 	}
3102 
3103 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3104 		sdkp->ws16 = 1;
3105 
3106 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3107 		sdkp->ws10 = 1;
3108 }
3109 
3110 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3111 {
3112 	struct scsi_device *sdev = sdkp->device;
3113 
3114 	if (!sdev->security_supported)
3115 		return;
3116 
3117 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3118 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3119 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3120 			SECURITY_PROTOCOL_OUT, 0) == 1)
3121 		sdkp->security = 1;
3122 }
3123 
3124 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3125 {
3126 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3127 }
3128 
3129 /**
3130  * sd_read_cpr - Query concurrent positioning ranges
3131  * @sdkp:	disk to query
3132  */
3133 static void sd_read_cpr(struct scsi_disk *sdkp)
3134 {
3135 	struct blk_independent_access_ranges *iars = NULL;
3136 	unsigned char *buffer = NULL;
3137 	unsigned int nr_cpr = 0;
3138 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3139 	u8 *desc;
3140 
3141 	/*
3142 	 * We need to have the capacity set first for the block layer to be
3143 	 * able to check the ranges.
3144 	 */
3145 	if (sdkp->first_scan)
3146 		return;
3147 
3148 	if (!sdkp->capacity)
3149 		goto out;
3150 
3151 	/*
3152 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3153 	 * leading to a maximum page size of 64 + 256*32 bytes.
3154 	 */
3155 	buf_len = 64 + 256*32;
3156 	buffer = kmalloc(buf_len, GFP_KERNEL);
3157 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3158 		goto out;
3159 
3160 	/* We must have at least a 64B header and one 32B range descriptor */
3161 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3162 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3163 		sd_printk(KERN_ERR, sdkp,
3164 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3165 		goto out;
3166 	}
3167 
3168 	nr_cpr = (vpd_len - 64) / 32;
3169 	if (nr_cpr == 1) {
3170 		nr_cpr = 0;
3171 		goto out;
3172 	}
3173 
3174 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3175 	if (!iars) {
3176 		nr_cpr = 0;
3177 		goto out;
3178 	}
3179 
3180 	desc = &buffer[64];
3181 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3182 		if (desc[0] != i) {
3183 			sd_printk(KERN_ERR, sdkp,
3184 				"Invalid Concurrent Positioning Range number\n");
3185 			nr_cpr = 0;
3186 			break;
3187 		}
3188 
3189 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3190 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3191 	}
3192 
3193 out:
3194 	disk_set_independent_access_ranges(sdkp->disk, iars);
3195 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3196 		sd_printk(KERN_NOTICE, sdkp,
3197 			  "%u concurrent positioning ranges\n", nr_cpr);
3198 		sdkp->nr_actuators = nr_cpr;
3199 	}
3200 
3201 	kfree(buffer);
3202 }
3203 
3204 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3205 {
3206 	struct scsi_device *sdp = sdkp->device;
3207 	unsigned int min_xfer_bytes =
3208 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3209 
3210 	if (sdkp->min_xfer_blocks == 0)
3211 		return false;
3212 
3213 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3214 		sd_first_printk(KERN_WARNING, sdkp,
3215 				"Preferred minimum I/O size %u bytes not a " \
3216 				"multiple of physical block size (%u bytes)\n",
3217 				min_xfer_bytes, sdkp->physical_block_size);
3218 		sdkp->min_xfer_blocks = 0;
3219 		return false;
3220 	}
3221 
3222 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3223 			min_xfer_bytes);
3224 	return true;
3225 }
3226 
3227 /*
3228  * Determine the device's preferred I/O size for reads and writes
3229  * unless the reported value is unreasonably small, large, not a
3230  * multiple of the physical block size, or simply garbage.
3231  */
3232 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3233 				      unsigned int dev_max)
3234 {
3235 	struct scsi_device *sdp = sdkp->device;
3236 	unsigned int opt_xfer_bytes =
3237 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3238 	unsigned int min_xfer_bytes =
3239 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3240 
3241 	if (sdkp->opt_xfer_blocks == 0)
3242 		return false;
3243 
3244 	if (sdkp->opt_xfer_blocks > dev_max) {
3245 		sd_first_printk(KERN_WARNING, sdkp,
3246 				"Optimal transfer size %u logical blocks " \
3247 				"> dev_max (%u logical blocks)\n",
3248 				sdkp->opt_xfer_blocks, dev_max);
3249 		return false;
3250 	}
3251 
3252 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3253 		sd_first_printk(KERN_WARNING, sdkp,
3254 				"Optimal transfer size %u logical blocks " \
3255 				"> sd driver limit (%u logical blocks)\n",
3256 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3257 		return false;
3258 	}
3259 
3260 	if (opt_xfer_bytes < PAGE_SIZE) {
3261 		sd_first_printk(KERN_WARNING, sdkp,
3262 				"Optimal transfer size %u bytes < " \
3263 				"PAGE_SIZE (%u bytes)\n",
3264 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3265 		return false;
3266 	}
3267 
3268 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3269 		sd_first_printk(KERN_WARNING, sdkp,
3270 				"Optimal transfer size %u bytes not a " \
3271 				"multiple of preferred minimum block " \
3272 				"size (%u bytes)\n",
3273 				opt_xfer_bytes, min_xfer_bytes);
3274 		return false;
3275 	}
3276 
3277 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3278 		sd_first_printk(KERN_WARNING, sdkp,
3279 				"Optimal transfer size %u bytes not a " \
3280 				"multiple of physical block size (%u bytes)\n",
3281 				opt_xfer_bytes, sdkp->physical_block_size);
3282 		return false;
3283 	}
3284 
3285 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3286 			opt_xfer_bytes);
3287 	return true;
3288 }
3289 
3290 /**
3291  *	sd_revalidate_disk - called the first time a new disk is seen,
3292  *	performs disk spin up, read_capacity, etc.
3293  *	@disk: struct gendisk we care about
3294  **/
3295 static int sd_revalidate_disk(struct gendisk *disk)
3296 {
3297 	struct scsi_disk *sdkp = scsi_disk(disk);
3298 	struct scsi_device *sdp = sdkp->device;
3299 	struct request_queue *q = sdkp->disk->queue;
3300 	sector_t old_capacity = sdkp->capacity;
3301 	unsigned char *buffer;
3302 	unsigned int dev_max, rw_max;
3303 
3304 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3305 				      "sd_revalidate_disk\n"));
3306 
3307 	/*
3308 	 * If the device is offline, don't try and read capacity or any
3309 	 * of the other niceties.
3310 	 */
3311 	if (!scsi_device_online(sdp))
3312 		goto out;
3313 
3314 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3315 	if (!buffer) {
3316 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3317 			  "allocation failure.\n");
3318 		goto out;
3319 	}
3320 
3321 	sd_spinup_disk(sdkp);
3322 
3323 	/*
3324 	 * Without media there is no reason to ask; moreover, some devices
3325 	 * react badly if we do.
3326 	 */
3327 	if (sdkp->media_present) {
3328 		sd_read_capacity(sdkp, buffer);
3329 
3330 		/*
3331 		 * set the default to rotational.  All non-rotational devices
3332 		 * support the block characteristics VPD page, which will
3333 		 * cause this to be updated correctly and any device which
3334 		 * doesn't support it should be treated as rotational.
3335 		 */
3336 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3337 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3338 
3339 		if (scsi_device_supports_vpd(sdp)) {
3340 			sd_read_block_provisioning(sdkp);
3341 			sd_read_block_limits(sdkp);
3342 			sd_read_block_characteristics(sdkp);
3343 			sd_zbc_read_zones(sdkp, buffer);
3344 			sd_read_cpr(sdkp);
3345 		}
3346 
3347 		sd_print_capacity(sdkp, old_capacity);
3348 
3349 		sd_read_write_protect_flag(sdkp, buffer);
3350 		sd_read_cache_type(sdkp, buffer);
3351 		sd_read_app_tag_own(sdkp, buffer);
3352 		sd_read_write_same(sdkp, buffer);
3353 		sd_read_security(sdkp, buffer);
3354 		sd_config_protection(sdkp);
3355 	}
3356 
3357 	/*
3358 	 * We now have all cache related info, determine how we deal
3359 	 * with flush requests.
3360 	 */
3361 	sd_set_flush_flag(sdkp);
3362 
3363 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3364 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3365 
3366 	/* Some devices report a maximum block count for READ/WRITE requests. */
3367 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3368 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3369 
3370 	if (sd_validate_min_xfer_size(sdkp))
3371 		blk_queue_io_min(sdkp->disk->queue,
3372 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3373 	else
3374 		blk_queue_io_min(sdkp->disk->queue, 0);
3375 
3376 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3377 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3378 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3379 	} else {
3380 		q->limits.io_opt = 0;
3381 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3382 				      (sector_t)BLK_DEF_MAX_SECTORS);
3383 	}
3384 
3385 	/*
3386 	 * Limit default to SCSI host optimal sector limit if set. There may be
3387 	 * an impact on performance for when the size of a request exceeds this
3388 	 * host limit.
3389 	 */
3390 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3391 
3392 	/* Do not exceed controller limit */
3393 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3394 
3395 	/*
3396 	 * Only update max_sectors if previously unset or if the current value
3397 	 * exceeds the capabilities of the hardware.
3398 	 */
3399 	if (sdkp->first_scan ||
3400 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3401 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3402 		q->limits.max_sectors = rw_max;
3403 
3404 	sdkp->first_scan = 0;
3405 
3406 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3407 	sd_config_write_same(sdkp);
3408 	kfree(buffer);
3409 
3410 	/*
3411 	 * For a zoned drive, revalidating the zones can be done only once
3412 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3413 	 * capacity to 0.
3414 	 */
3415 	if (sd_zbc_revalidate_zones(sdkp))
3416 		set_capacity_and_notify(disk, 0);
3417 
3418  out:
3419 	return 0;
3420 }
3421 
3422 /**
3423  *	sd_unlock_native_capacity - unlock native capacity
3424  *	@disk: struct gendisk to set capacity for
3425  *
3426  *	Block layer calls this function if it detects that partitions
3427  *	on @disk reach beyond the end of the device.  If the SCSI host
3428  *	implements ->unlock_native_capacity() method, it's invoked to
3429  *	give it a chance to adjust the device capacity.
3430  *
3431  *	CONTEXT:
3432  *	Defined by block layer.  Might sleep.
3433  */
3434 static void sd_unlock_native_capacity(struct gendisk *disk)
3435 {
3436 	struct scsi_device *sdev = scsi_disk(disk)->device;
3437 
3438 	if (sdev->host->hostt->unlock_native_capacity)
3439 		sdev->host->hostt->unlock_native_capacity(sdev);
3440 }
3441 
3442 /**
3443  *	sd_format_disk_name - format disk name
3444  *	@prefix: name prefix - ie. "sd" for SCSI disks
3445  *	@index: index of the disk to format name for
3446  *	@buf: output buffer
3447  *	@buflen: length of the output buffer
3448  *
3449  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3450  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3451  *	which is followed by sdaaa.
3452  *
3453  *	This is basically 26 base counting with one extra 'nil' entry
3454  *	at the beginning from the second digit on and can be
3455  *	determined using similar method as 26 base conversion with the
3456  *	index shifted -1 after each digit is computed.
3457  *
3458  *	CONTEXT:
3459  *	Don't care.
3460  *
3461  *	RETURNS:
3462  *	0 on success, -errno on failure.
3463  */
3464 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3465 {
3466 	const int base = 'z' - 'a' + 1;
3467 	char *begin = buf + strlen(prefix);
3468 	char *end = buf + buflen;
3469 	char *p;
3470 	int unit;
3471 
3472 	p = end - 1;
3473 	*p = '\0';
3474 	unit = base;
3475 	do {
3476 		if (p == begin)
3477 			return -EINVAL;
3478 		*--p = 'a' + (index % unit);
3479 		index = (index / unit) - 1;
3480 	} while (index >= 0);
3481 
3482 	memmove(begin, p, end - p);
3483 	memcpy(buf, prefix, strlen(prefix));
3484 
3485 	return 0;
3486 }
3487 
3488 /**
3489  *	sd_probe - called during driver initialization and whenever a
3490  *	new scsi device is attached to the system. It is called once
3491  *	for each scsi device (not just disks) present.
3492  *	@dev: pointer to device object
3493  *
3494  *	Returns 0 if successful (or not interested in this scsi device
3495  *	(e.g. scanner)); 1 when there is an error.
3496  *
3497  *	Note: this function is invoked from the scsi mid-level.
3498  *	This function sets up the mapping between a given
3499  *	<host,channel,id,lun> (found in sdp) and new device name
3500  *	(e.g. /dev/sda). More precisely it is the block device major
3501  *	and minor number that is chosen here.
3502  *
3503  *	Assume sd_probe is not re-entrant (for time being)
3504  *	Also think about sd_probe() and sd_remove() running coincidentally.
3505  **/
3506 static int sd_probe(struct device *dev)
3507 {
3508 	struct scsi_device *sdp = to_scsi_device(dev);
3509 	struct scsi_disk *sdkp;
3510 	struct gendisk *gd;
3511 	int index;
3512 	int error;
3513 
3514 	scsi_autopm_get_device(sdp);
3515 	error = -ENODEV;
3516 	if (sdp->type != TYPE_DISK &&
3517 	    sdp->type != TYPE_ZBC &&
3518 	    sdp->type != TYPE_MOD &&
3519 	    sdp->type != TYPE_RBC)
3520 		goto out;
3521 
3522 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3523 		sdev_printk(KERN_WARNING, sdp,
3524 			    "Unsupported ZBC host-managed device.\n");
3525 		goto out;
3526 	}
3527 
3528 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3529 					"sd_probe\n"));
3530 
3531 	error = -ENOMEM;
3532 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3533 	if (!sdkp)
3534 		goto out;
3535 
3536 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3537 					 &sd_bio_compl_lkclass);
3538 	if (!gd)
3539 		goto out_free;
3540 
3541 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3542 	if (index < 0) {
3543 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3544 		goto out_put;
3545 	}
3546 
3547 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3548 	if (error) {
3549 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3550 		goto out_free_index;
3551 	}
3552 
3553 	sdkp->device = sdp;
3554 	sdkp->disk = gd;
3555 	sdkp->index = index;
3556 	sdkp->max_retries = SD_MAX_RETRIES;
3557 	atomic_set(&sdkp->openers, 0);
3558 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3559 
3560 	if (!sdp->request_queue->rq_timeout) {
3561 		if (sdp->type != TYPE_MOD)
3562 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3563 		else
3564 			blk_queue_rq_timeout(sdp->request_queue,
3565 					     SD_MOD_TIMEOUT);
3566 	}
3567 
3568 	device_initialize(&sdkp->disk_dev);
3569 	sdkp->disk_dev.parent = get_device(dev);
3570 	sdkp->disk_dev.class = &sd_disk_class;
3571 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3572 
3573 	error = device_add(&sdkp->disk_dev);
3574 	if (error) {
3575 		put_device(&sdkp->disk_dev);
3576 		goto out;
3577 	}
3578 
3579 	dev_set_drvdata(dev, sdkp);
3580 
3581 	gd->major = sd_major((index & 0xf0) >> 4);
3582 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3583 	gd->minors = SD_MINORS;
3584 
3585 	gd->fops = &sd_fops;
3586 	gd->private_data = sdkp;
3587 
3588 	/* defaults, until the device tells us otherwise */
3589 	sdp->sector_size = 512;
3590 	sdkp->capacity = 0;
3591 	sdkp->media_present = 1;
3592 	sdkp->write_prot = 0;
3593 	sdkp->cache_override = 0;
3594 	sdkp->WCE = 0;
3595 	sdkp->RCD = 0;
3596 	sdkp->ATO = 0;
3597 	sdkp->first_scan = 1;
3598 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3599 
3600 	sd_revalidate_disk(gd);
3601 
3602 	if (sdp->removable) {
3603 		gd->flags |= GENHD_FL_REMOVABLE;
3604 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3605 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3606 	}
3607 
3608 	blk_pm_runtime_init(sdp->request_queue, dev);
3609 	if (sdp->rpm_autosuspend) {
3610 		pm_runtime_set_autosuspend_delay(dev,
3611 			sdp->host->hostt->rpm_autosuspend_delay);
3612 	}
3613 
3614 	error = device_add_disk(dev, gd, NULL);
3615 	if (error) {
3616 		put_device(&sdkp->disk_dev);
3617 		put_disk(gd);
3618 		goto out;
3619 	}
3620 
3621 	if (sdkp->security) {
3622 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3623 		if (sdkp->opal_dev)
3624 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3625 	}
3626 
3627 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3628 		  sdp->removable ? "removable " : "");
3629 	scsi_autopm_put_device(sdp);
3630 
3631 	return 0;
3632 
3633  out_free_index:
3634 	ida_free(&sd_index_ida, index);
3635  out_put:
3636 	put_disk(gd);
3637  out_free:
3638 	kfree(sdkp);
3639  out:
3640 	scsi_autopm_put_device(sdp);
3641 	return error;
3642 }
3643 
3644 /**
3645  *	sd_remove - called whenever a scsi disk (previously recognized by
3646  *	sd_probe) is detached from the system. It is called (potentially
3647  *	multiple times) during sd module unload.
3648  *	@dev: pointer to device object
3649  *
3650  *	Note: this function is invoked from the scsi mid-level.
3651  *	This function potentially frees up a device name (e.g. /dev/sdc)
3652  *	that could be re-used by a subsequent sd_probe().
3653  *	This function is not called when the built-in sd driver is "exit-ed".
3654  **/
3655 static int sd_remove(struct device *dev)
3656 {
3657 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3658 
3659 	scsi_autopm_get_device(sdkp->device);
3660 
3661 	device_del(&sdkp->disk_dev);
3662 	del_gendisk(sdkp->disk);
3663 	sd_shutdown(dev);
3664 
3665 	put_disk(sdkp->disk);
3666 	return 0;
3667 }
3668 
3669 static void scsi_disk_release(struct device *dev)
3670 {
3671 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3672 
3673 	ida_free(&sd_index_ida, sdkp->index);
3674 	sd_zbc_free_zone_info(sdkp);
3675 	put_device(&sdkp->device->sdev_gendev);
3676 	free_opal_dev(sdkp->opal_dev);
3677 
3678 	kfree(sdkp);
3679 }
3680 
3681 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3682 {
3683 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3684 	struct scsi_sense_hdr sshdr;
3685 	const struct scsi_exec_args exec_args = {
3686 		.sshdr = &sshdr,
3687 		.req_flags = BLK_MQ_REQ_PM,
3688 	};
3689 	struct scsi_device *sdp = sdkp->device;
3690 	int res;
3691 
3692 	if (start)
3693 		cmd[4] |= 1;	/* START */
3694 
3695 	if (sdp->start_stop_pwr_cond)
3696 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3697 
3698 	if (!scsi_device_online(sdp))
3699 		return -ENODEV;
3700 
3701 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3702 			       sdkp->max_retries, &exec_args);
3703 	if (res) {
3704 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3705 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3706 			sd_print_sense_hdr(sdkp, &sshdr);
3707 			/* 0x3a is medium not present */
3708 			if (sshdr.asc == 0x3a)
3709 				res = 0;
3710 		}
3711 	}
3712 
3713 	/* SCSI error codes must not go to the generic layer */
3714 	if (res)
3715 		return -EIO;
3716 
3717 	return 0;
3718 }
3719 
3720 /*
3721  * Send a SYNCHRONIZE CACHE instruction down to the device through
3722  * the normal SCSI command structure.  Wait for the command to
3723  * complete.
3724  */
3725 static void sd_shutdown(struct device *dev)
3726 {
3727 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3728 
3729 	if (!sdkp)
3730 		return;         /* this can happen */
3731 
3732 	if (pm_runtime_suspended(dev))
3733 		return;
3734 
3735 	if (sdkp->WCE && sdkp->media_present) {
3736 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3737 		sd_sync_cache(sdkp, NULL);
3738 	}
3739 
3740 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3741 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3742 		sd_start_stop_device(sdkp, 0);
3743 	}
3744 }
3745 
3746 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3747 {
3748 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3749 	struct scsi_sense_hdr sshdr;
3750 	int ret = 0;
3751 
3752 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3753 		return 0;
3754 
3755 	if (sdkp->WCE && sdkp->media_present) {
3756 		if (!sdkp->device->silence_suspend)
3757 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3758 		ret = sd_sync_cache(sdkp, &sshdr);
3759 
3760 		if (ret) {
3761 			/* ignore OFFLINE device */
3762 			if (ret == -ENODEV)
3763 				return 0;
3764 
3765 			if (!scsi_sense_valid(&sshdr) ||
3766 			    sshdr.sense_key != ILLEGAL_REQUEST)
3767 				return ret;
3768 
3769 			/*
3770 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3771 			 * doesn't support sync. There's not much to do and
3772 			 * suspend shouldn't fail.
3773 			 */
3774 			ret = 0;
3775 		}
3776 	}
3777 
3778 	if (sdkp->device->manage_start_stop) {
3779 		if (!sdkp->device->silence_suspend)
3780 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3781 		/* an error is not worth aborting a system sleep */
3782 		ret = sd_start_stop_device(sdkp, 0);
3783 		if (ignore_stop_errors)
3784 			ret = 0;
3785 	}
3786 
3787 	return ret;
3788 }
3789 
3790 static int sd_suspend_system(struct device *dev)
3791 {
3792 	if (pm_runtime_suspended(dev))
3793 		return 0;
3794 
3795 	return sd_suspend_common(dev, true);
3796 }
3797 
3798 static int sd_suspend_runtime(struct device *dev)
3799 {
3800 	return sd_suspend_common(dev, false);
3801 }
3802 
3803 static int sd_resume(struct device *dev)
3804 {
3805 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3806 	int ret;
3807 
3808 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3809 		return 0;
3810 
3811 	if (!sdkp->device->manage_start_stop)
3812 		return 0;
3813 
3814 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3815 	ret = sd_start_stop_device(sdkp, 1);
3816 	if (!ret)
3817 		opal_unlock_from_suspend(sdkp->opal_dev);
3818 	return ret;
3819 }
3820 
3821 static int sd_resume_system(struct device *dev)
3822 {
3823 	if (pm_runtime_suspended(dev))
3824 		return 0;
3825 
3826 	return sd_resume(dev);
3827 }
3828 
3829 static int sd_resume_runtime(struct device *dev)
3830 {
3831 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3832 	struct scsi_device *sdp;
3833 
3834 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3835 		return 0;
3836 
3837 	sdp = sdkp->device;
3838 
3839 	if (sdp->ignore_media_change) {
3840 		/* clear the device's sense data */
3841 		static const u8 cmd[10] = { REQUEST_SENSE };
3842 		const struct scsi_exec_args exec_args = {
3843 			.req_flags = BLK_MQ_REQ_PM,
3844 		};
3845 
3846 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
3847 				     sdp->request_queue->rq_timeout, 1,
3848 				     &exec_args))
3849 			sd_printk(KERN_NOTICE, sdkp,
3850 				  "Failed to clear sense data\n");
3851 	}
3852 
3853 	return sd_resume(dev);
3854 }
3855 
3856 /**
3857  *	init_sd - entry point for this driver (both when built in or when
3858  *	a module).
3859  *
3860  *	Note: this function registers this driver with the scsi mid-level.
3861  **/
3862 static int __init init_sd(void)
3863 {
3864 	int majors = 0, i, err;
3865 
3866 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3867 
3868 	for (i = 0; i < SD_MAJORS; i++) {
3869 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3870 			continue;
3871 		majors++;
3872 	}
3873 
3874 	if (!majors)
3875 		return -ENODEV;
3876 
3877 	err = class_register(&sd_disk_class);
3878 	if (err)
3879 		goto err_out;
3880 
3881 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3882 	if (!sd_page_pool) {
3883 		printk(KERN_ERR "sd: can't init discard page pool\n");
3884 		err = -ENOMEM;
3885 		goto err_out_class;
3886 	}
3887 
3888 	err = scsi_register_driver(&sd_template.gendrv);
3889 	if (err)
3890 		goto err_out_driver;
3891 
3892 	return 0;
3893 
3894 err_out_driver:
3895 	mempool_destroy(sd_page_pool);
3896 err_out_class:
3897 	class_unregister(&sd_disk_class);
3898 err_out:
3899 	for (i = 0; i < SD_MAJORS; i++)
3900 		unregister_blkdev(sd_major(i), "sd");
3901 	return err;
3902 }
3903 
3904 /**
3905  *	exit_sd - exit point for this driver (when it is a module).
3906  *
3907  *	Note: this function unregisters this driver from the scsi mid-level.
3908  **/
3909 static void __exit exit_sd(void)
3910 {
3911 	int i;
3912 
3913 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3914 
3915 	scsi_unregister_driver(&sd_template.gendrv);
3916 	mempool_destroy(sd_page_pool);
3917 
3918 	class_unregister(&sd_disk_class);
3919 
3920 	for (i = 0; i < SD_MAJORS; i++)
3921 		unregister_blkdev(sd_major(i), "sd");
3922 }
3923 
3924 module_init(init_sd);
3925 module_exit(exit_sd);
3926 
3927 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3928 {
3929 	scsi_print_sense_hdr(sdkp->device,
3930 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3931 }
3932 
3933 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3934 {
3935 	const char *hb_string = scsi_hostbyte_string(result);
3936 
3937 	if (hb_string)
3938 		sd_printk(KERN_INFO, sdkp,
3939 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3940 			  hb_string ? hb_string : "invalid",
3941 			  "DRIVER_OK");
3942 	else
3943 		sd_printk(KERN_INFO, sdkp,
3944 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3945 			  msg, host_byte(result), "DRIVER_OK");
3946 }
3947