xref: /openbmc/linux/drivers/scsi/sd.c (revision d171c85d)
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 #include <scsi/scsi_common.h>
71 
72 #include "sd.h"
73 #include "scsi_priv.h"
74 #include "scsi_logging.h"
75 
76 MODULE_AUTHOR("Eric Youngdale");
77 MODULE_DESCRIPTION("SCSI disk (sd) driver");
78 MODULE_LICENSE("GPL");
79 
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
95 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
99 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
100 
101 #define SD_MINORS	16
102 
103 static void sd_config_discard(struct scsi_disk *, unsigned int);
104 static void sd_config_write_same(struct scsi_disk *);
105 static int  sd_revalidate_disk(struct gendisk *);
106 static void sd_unlock_native_capacity(struct gendisk *disk);
107 static void sd_shutdown(struct device *);
108 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
109 static void scsi_disk_release(struct device *cdev);
110 
111 static DEFINE_IDA(sd_index_ida);
112 
113 static mempool_t *sd_page_pool;
114 static struct lock_class_key sd_bio_compl_lkclass;
115 
116 static const char *sd_cache_types[] = {
117 	"write through", "none", "write back",
118 	"write back, no read (daft)"
119 };
120 
121 static void sd_set_flush_flag(struct scsi_disk *sdkp)
122 {
123 	bool wc = false, fua = false;
124 
125 	if (sdkp->WCE) {
126 		wc = true;
127 		if (sdkp->DPOFUA)
128 			fua = true;
129 	}
130 
131 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
132 }
133 
134 static ssize_t
135 cache_type_store(struct device *dev, struct device_attribute *attr,
136 		 const char *buf, size_t count)
137 {
138 	int ct, rcd, wce, sp;
139 	struct scsi_disk *sdkp = to_scsi_disk(dev);
140 	struct scsi_device *sdp = sdkp->device;
141 	char buffer[64];
142 	char *buffer_data;
143 	struct scsi_mode_data data;
144 	struct scsi_sense_hdr sshdr;
145 	static const char temp[] = "temporary ";
146 	int len;
147 
148 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
149 		/* no cache control on RBC devices; theoretically they
150 		 * can do it, but there's probably so many exceptions
151 		 * it's not worth the risk */
152 		return -EINVAL;
153 
154 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
155 		buf += sizeof(temp) - 1;
156 		sdkp->cache_override = 1;
157 	} else {
158 		sdkp->cache_override = 0;
159 	}
160 
161 	ct = sysfs_match_string(sd_cache_types, buf);
162 	if (ct < 0)
163 		return -EINVAL;
164 
165 	rcd = ct & 0x01 ? 1 : 0;
166 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
167 
168 	if (sdkp->cache_override) {
169 		sdkp->WCE = wce;
170 		sdkp->RCD = rcd;
171 		sd_set_flush_flag(sdkp);
172 		return count;
173 	}
174 
175 	if (scsi_mode_sense(sdp, 0x08, 8, 0, buffer, sizeof(buffer), SD_TIMEOUT,
176 			    sdkp->max_retries, &data, NULL))
177 		return -EINVAL;
178 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
179 		  data.block_descriptor_length);
180 	buffer_data = buffer + data.header_length +
181 		data.block_descriptor_length;
182 	buffer_data[2] &= ~0x05;
183 	buffer_data[2] |= wce << 2 | rcd;
184 	sp = buffer_data[0] & 0x80 ? 1 : 0;
185 	buffer_data[0] &= ~0x80;
186 
187 	/*
188 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
189 	 * received mode parameter buffer before doing MODE SELECT.
190 	 */
191 	data.device_specific = 0;
192 
193 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
194 			     sdkp->max_retries, &data, &sshdr)) {
195 		if (scsi_sense_valid(&sshdr))
196 			sd_print_sense_hdr(sdkp, &sshdr);
197 		return -EINVAL;
198 	}
199 	sd_revalidate_disk(sdkp->disk);
200 	return count;
201 }
202 
203 static ssize_t
204 manage_start_stop_show(struct device *dev,
205 		       struct device_attribute *attr, char *buf)
206 {
207 	struct scsi_disk *sdkp = to_scsi_disk(dev);
208 	struct scsi_device *sdp = sdkp->device;
209 
210 	return sysfs_emit(buf, "%u\n",
211 			  sdp->manage_system_start_stop &&
212 			  sdp->manage_runtime_start_stop &&
213 			  sdp->manage_shutdown);
214 }
215 static DEVICE_ATTR_RO(manage_start_stop);
216 
217 static ssize_t
218 manage_system_start_stop_show(struct device *dev,
219 			      struct device_attribute *attr, char *buf)
220 {
221 	struct scsi_disk *sdkp = to_scsi_disk(dev);
222 	struct scsi_device *sdp = sdkp->device;
223 
224 	return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
225 }
226 
227 static ssize_t
228 manage_system_start_stop_store(struct device *dev,
229 			       struct device_attribute *attr,
230 			       const char *buf, size_t count)
231 {
232 	struct scsi_disk *sdkp = to_scsi_disk(dev);
233 	struct scsi_device *sdp = sdkp->device;
234 	bool v;
235 
236 	if (!capable(CAP_SYS_ADMIN))
237 		return -EACCES;
238 
239 	if (kstrtobool(buf, &v))
240 		return -EINVAL;
241 
242 	sdp->manage_system_start_stop = v;
243 
244 	return count;
245 }
246 static DEVICE_ATTR_RW(manage_system_start_stop);
247 
248 static ssize_t
249 manage_runtime_start_stop_show(struct device *dev,
250 			       struct device_attribute *attr, char *buf)
251 {
252 	struct scsi_disk *sdkp = to_scsi_disk(dev);
253 	struct scsi_device *sdp = sdkp->device;
254 
255 	return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
256 }
257 
258 static ssize_t
259 manage_runtime_start_stop_store(struct device *dev,
260 				struct device_attribute *attr,
261 				const char *buf, size_t count)
262 {
263 	struct scsi_disk *sdkp = to_scsi_disk(dev);
264 	struct scsi_device *sdp = sdkp->device;
265 	bool v;
266 
267 	if (!capable(CAP_SYS_ADMIN))
268 		return -EACCES;
269 
270 	if (kstrtobool(buf, &v))
271 		return -EINVAL;
272 
273 	sdp->manage_runtime_start_stop = v;
274 
275 	return count;
276 }
277 static DEVICE_ATTR_RW(manage_runtime_start_stop);
278 
279 static ssize_t manage_shutdown_show(struct device *dev,
280 				    struct device_attribute *attr, char *buf)
281 {
282 	struct scsi_disk *sdkp = to_scsi_disk(dev);
283 	struct scsi_device *sdp = sdkp->device;
284 
285 	return sysfs_emit(buf, "%u\n", sdp->manage_shutdown);
286 }
287 
288 static ssize_t manage_shutdown_store(struct device *dev,
289 				     struct device_attribute *attr,
290 				     const char *buf, size_t count)
291 {
292 	struct scsi_disk *sdkp = to_scsi_disk(dev);
293 	struct scsi_device *sdp = sdkp->device;
294 	bool v;
295 
296 	if (!capable(CAP_SYS_ADMIN))
297 		return -EACCES;
298 
299 	if (kstrtobool(buf, &v))
300 		return -EINVAL;
301 
302 	sdp->manage_shutdown = v;
303 
304 	return count;
305 }
306 static DEVICE_ATTR_RW(manage_shutdown);
307 
308 static ssize_t
309 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
310 {
311 	struct scsi_disk *sdkp = to_scsi_disk(dev);
312 
313 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
314 }
315 
316 static ssize_t
317 allow_restart_store(struct device *dev, struct device_attribute *attr,
318 		    const char *buf, size_t count)
319 {
320 	bool v;
321 	struct scsi_disk *sdkp = to_scsi_disk(dev);
322 	struct scsi_device *sdp = sdkp->device;
323 
324 	if (!capable(CAP_SYS_ADMIN))
325 		return -EACCES;
326 
327 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
328 		return -EINVAL;
329 
330 	if (kstrtobool(buf, &v))
331 		return -EINVAL;
332 
333 	sdp->allow_restart = v;
334 
335 	return count;
336 }
337 static DEVICE_ATTR_RW(allow_restart);
338 
339 static ssize_t
340 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
341 {
342 	struct scsi_disk *sdkp = to_scsi_disk(dev);
343 	int ct = sdkp->RCD + 2*sdkp->WCE;
344 
345 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
346 }
347 static DEVICE_ATTR_RW(cache_type);
348 
349 static ssize_t
350 FUA_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->DPOFUA);
355 }
356 static DEVICE_ATTR_RO(FUA);
357 
358 static ssize_t
359 protection_type_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->protection_type);
365 }
366 
367 static ssize_t
368 protection_type_store(struct device *dev, struct device_attribute *attr,
369 		      const char *buf, size_t count)
370 {
371 	struct scsi_disk *sdkp = to_scsi_disk(dev);
372 	unsigned int val;
373 	int err;
374 
375 	if (!capable(CAP_SYS_ADMIN))
376 		return -EACCES;
377 
378 	err = kstrtouint(buf, 10, &val);
379 
380 	if (err)
381 		return err;
382 
383 	if (val <= T10_PI_TYPE3_PROTECTION)
384 		sdkp->protection_type = val;
385 
386 	return count;
387 }
388 static DEVICE_ATTR_RW(protection_type);
389 
390 static ssize_t
391 protection_mode_show(struct device *dev, struct device_attribute *attr,
392 		     char *buf)
393 {
394 	struct scsi_disk *sdkp = to_scsi_disk(dev);
395 	struct scsi_device *sdp = sdkp->device;
396 	unsigned int dif, dix;
397 
398 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
399 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
400 
401 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
402 		dif = 0;
403 		dix = 1;
404 	}
405 
406 	if (!dif && !dix)
407 		return sprintf(buf, "none\n");
408 
409 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
410 }
411 static DEVICE_ATTR_RO(protection_mode);
412 
413 static ssize_t
414 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
415 {
416 	struct scsi_disk *sdkp = to_scsi_disk(dev);
417 
418 	return sprintf(buf, "%u\n", sdkp->ATO);
419 }
420 static DEVICE_ATTR_RO(app_tag_own);
421 
422 static ssize_t
423 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
424 		       char *buf)
425 {
426 	struct scsi_disk *sdkp = to_scsi_disk(dev);
427 
428 	return sprintf(buf, "%u\n", sdkp->lbpme);
429 }
430 static DEVICE_ATTR_RO(thin_provisioning);
431 
432 /* sysfs_match_string() requires dense arrays */
433 static const char *lbp_mode[] = {
434 	[SD_LBP_FULL]		= "full",
435 	[SD_LBP_UNMAP]		= "unmap",
436 	[SD_LBP_WS16]		= "writesame_16",
437 	[SD_LBP_WS10]		= "writesame_10",
438 	[SD_LBP_ZERO]		= "writesame_zero",
439 	[SD_LBP_DISABLE]	= "disabled",
440 };
441 
442 static ssize_t
443 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
444 		       char *buf)
445 {
446 	struct scsi_disk *sdkp = to_scsi_disk(dev);
447 
448 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
449 }
450 
451 static ssize_t
452 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
453 			const char *buf, size_t count)
454 {
455 	struct scsi_disk *sdkp = to_scsi_disk(dev);
456 	struct scsi_device *sdp = sdkp->device;
457 	int mode;
458 
459 	if (!capable(CAP_SYS_ADMIN))
460 		return -EACCES;
461 
462 	if (sd_is_zoned(sdkp)) {
463 		sd_config_discard(sdkp, SD_LBP_DISABLE);
464 		return count;
465 	}
466 
467 	if (sdp->type != TYPE_DISK)
468 		return -EINVAL;
469 
470 	mode = sysfs_match_string(lbp_mode, buf);
471 	if (mode < 0)
472 		return -EINVAL;
473 
474 	sd_config_discard(sdkp, mode);
475 
476 	return count;
477 }
478 static DEVICE_ATTR_RW(provisioning_mode);
479 
480 /* sysfs_match_string() requires dense arrays */
481 static const char *zeroing_mode[] = {
482 	[SD_ZERO_WRITE]		= "write",
483 	[SD_ZERO_WS]		= "writesame",
484 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
485 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
486 };
487 
488 static ssize_t
489 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
490 		  char *buf)
491 {
492 	struct scsi_disk *sdkp = to_scsi_disk(dev);
493 
494 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
495 }
496 
497 static ssize_t
498 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
499 		   const char *buf, size_t count)
500 {
501 	struct scsi_disk *sdkp = to_scsi_disk(dev);
502 	int mode;
503 
504 	if (!capable(CAP_SYS_ADMIN))
505 		return -EACCES;
506 
507 	mode = sysfs_match_string(zeroing_mode, buf);
508 	if (mode < 0)
509 		return -EINVAL;
510 
511 	sdkp->zeroing_mode = mode;
512 
513 	return count;
514 }
515 static DEVICE_ATTR_RW(zeroing_mode);
516 
517 static ssize_t
518 max_medium_access_timeouts_show(struct device *dev,
519 				struct device_attribute *attr, char *buf)
520 {
521 	struct scsi_disk *sdkp = to_scsi_disk(dev);
522 
523 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
524 }
525 
526 static ssize_t
527 max_medium_access_timeouts_store(struct device *dev,
528 				 struct device_attribute *attr, const char *buf,
529 				 size_t count)
530 {
531 	struct scsi_disk *sdkp = to_scsi_disk(dev);
532 	int err;
533 
534 	if (!capable(CAP_SYS_ADMIN))
535 		return -EACCES;
536 
537 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
538 
539 	return err ? err : count;
540 }
541 static DEVICE_ATTR_RW(max_medium_access_timeouts);
542 
543 static ssize_t
544 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
545 			   char *buf)
546 {
547 	struct scsi_disk *sdkp = to_scsi_disk(dev);
548 
549 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
550 }
551 
552 static ssize_t
553 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
554 			    const char *buf, size_t count)
555 {
556 	struct scsi_disk *sdkp = to_scsi_disk(dev);
557 	struct scsi_device *sdp = sdkp->device;
558 	unsigned long max;
559 	int err;
560 
561 	if (!capable(CAP_SYS_ADMIN))
562 		return -EACCES;
563 
564 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
565 		return -EINVAL;
566 
567 	err = kstrtoul(buf, 10, &max);
568 
569 	if (err)
570 		return err;
571 
572 	if (max == 0)
573 		sdp->no_write_same = 1;
574 	else if (max <= SD_MAX_WS16_BLOCKS) {
575 		sdp->no_write_same = 0;
576 		sdkp->max_ws_blocks = max;
577 	}
578 
579 	sd_config_write_same(sdkp);
580 
581 	return count;
582 }
583 static DEVICE_ATTR_RW(max_write_same_blocks);
584 
585 static ssize_t
586 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
587 {
588 	struct scsi_disk *sdkp = to_scsi_disk(dev);
589 
590 	if (sdkp->device->type == TYPE_ZBC)
591 		return sprintf(buf, "host-managed\n");
592 	if (sdkp->zoned == 1)
593 		return sprintf(buf, "host-aware\n");
594 	if (sdkp->zoned == 2)
595 		return sprintf(buf, "drive-managed\n");
596 	return sprintf(buf, "none\n");
597 }
598 static DEVICE_ATTR_RO(zoned_cap);
599 
600 static ssize_t
601 max_retries_store(struct device *dev, struct device_attribute *attr,
602 		  const char *buf, size_t count)
603 {
604 	struct scsi_disk *sdkp = to_scsi_disk(dev);
605 	struct scsi_device *sdev = sdkp->device;
606 	int retries, err;
607 
608 	err = kstrtoint(buf, 10, &retries);
609 	if (err)
610 		return err;
611 
612 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
613 		sdkp->max_retries = retries;
614 		return count;
615 	}
616 
617 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
618 		    SD_MAX_RETRIES);
619 	return -EINVAL;
620 }
621 
622 static ssize_t
623 max_retries_show(struct device *dev, struct device_attribute *attr,
624 		 char *buf)
625 {
626 	struct scsi_disk *sdkp = to_scsi_disk(dev);
627 
628 	return sprintf(buf, "%d\n", sdkp->max_retries);
629 }
630 
631 static DEVICE_ATTR_RW(max_retries);
632 
633 static struct attribute *sd_disk_attrs[] = {
634 	&dev_attr_cache_type.attr,
635 	&dev_attr_FUA.attr,
636 	&dev_attr_allow_restart.attr,
637 	&dev_attr_manage_start_stop.attr,
638 	&dev_attr_manage_system_start_stop.attr,
639 	&dev_attr_manage_runtime_start_stop.attr,
640 	&dev_attr_manage_shutdown.attr,
641 	&dev_attr_protection_type.attr,
642 	&dev_attr_protection_mode.attr,
643 	&dev_attr_app_tag_own.attr,
644 	&dev_attr_thin_provisioning.attr,
645 	&dev_attr_provisioning_mode.attr,
646 	&dev_attr_zeroing_mode.attr,
647 	&dev_attr_max_write_same_blocks.attr,
648 	&dev_attr_max_medium_access_timeouts.attr,
649 	&dev_attr_zoned_cap.attr,
650 	&dev_attr_max_retries.attr,
651 	NULL,
652 };
653 ATTRIBUTE_GROUPS(sd_disk);
654 
655 static struct class sd_disk_class = {
656 	.name		= "scsi_disk",
657 	.dev_release	= scsi_disk_release,
658 	.dev_groups	= sd_disk_groups,
659 };
660 
661 /*
662  * Don't request a new module, as that could deadlock in multipath
663  * environment.
664  */
665 static void sd_default_probe(dev_t devt)
666 {
667 }
668 
669 /*
670  * Device no to disk mapping:
671  *
672  *       major         disc2     disc  p1
673  *   |............|.............|....|....| <- dev_t
674  *    31        20 19          8 7  4 3  0
675  *
676  * Inside a major, we have 16k disks, however mapped non-
677  * contiguously. The first 16 disks are for major0, the next
678  * ones with major1, ... Disk 256 is for major0 again, disk 272
679  * for major1, ...
680  * As we stay compatible with our numbering scheme, we can reuse
681  * the well-know SCSI majors 8, 65--71, 136--143.
682  */
683 static int sd_major(int major_idx)
684 {
685 	switch (major_idx) {
686 	case 0:
687 		return SCSI_DISK0_MAJOR;
688 	case 1 ... 7:
689 		return SCSI_DISK1_MAJOR + major_idx - 1;
690 	case 8 ... 15:
691 		return SCSI_DISK8_MAJOR + major_idx - 8;
692 	default:
693 		BUG();
694 		return 0;	/* shut up gcc */
695 	}
696 }
697 
698 #ifdef CONFIG_BLK_SED_OPAL
699 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
700 		size_t len, bool send)
701 {
702 	struct scsi_disk *sdkp = data;
703 	struct scsi_device *sdev = sdkp->device;
704 	u8 cdb[12] = { 0, };
705 	const struct scsi_exec_args exec_args = {
706 		.req_flags = BLK_MQ_REQ_PM,
707 	};
708 	int ret;
709 
710 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
711 	cdb[1] = secp;
712 	put_unaligned_be16(spsp, &cdb[2]);
713 	put_unaligned_be32(len, &cdb[6]);
714 
715 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
716 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
717 			       &exec_args);
718 	return ret <= 0 ? ret : -EIO;
719 }
720 #endif /* CONFIG_BLK_SED_OPAL */
721 
722 /*
723  * Look up the DIX operation based on whether the command is read or
724  * write and whether dix and dif are enabled.
725  */
726 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
727 {
728 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
729 	static const unsigned int ops[] = {	/* wrt dix dif */
730 		SCSI_PROT_NORMAL,		/*  0	0   0  */
731 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
732 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
733 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
734 		SCSI_PROT_NORMAL,		/*  1	0   0  */
735 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
736 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
737 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
738 	};
739 
740 	return ops[write << 2 | dix << 1 | dif];
741 }
742 
743 /*
744  * Returns a mask of the protection flags that are valid for a given DIX
745  * operation.
746  */
747 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
748 {
749 	static const unsigned int flag_mask[] = {
750 		[SCSI_PROT_NORMAL]		= 0,
751 
752 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
753 						  SCSI_PROT_GUARD_CHECK |
754 						  SCSI_PROT_REF_CHECK |
755 						  SCSI_PROT_REF_INCREMENT,
756 
757 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
758 						  SCSI_PROT_IP_CHECKSUM,
759 
760 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
761 						  SCSI_PROT_GUARD_CHECK |
762 						  SCSI_PROT_REF_CHECK |
763 						  SCSI_PROT_REF_INCREMENT |
764 						  SCSI_PROT_IP_CHECKSUM,
765 
766 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
767 						  SCSI_PROT_REF_INCREMENT,
768 
769 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
770 						  SCSI_PROT_REF_CHECK |
771 						  SCSI_PROT_REF_INCREMENT |
772 						  SCSI_PROT_IP_CHECKSUM,
773 
774 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
775 						  SCSI_PROT_GUARD_CHECK |
776 						  SCSI_PROT_REF_CHECK |
777 						  SCSI_PROT_REF_INCREMENT |
778 						  SCSI_PROT_IP_CHECKSUM,
779 	};
780 
781 	return flag_mask[prot_op];
782 }
783 
784 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
785 					   unsigned int dix, unsigned int dif)
786 {
787 	struct request *rq = scsi_cmd_to_rq(scmd);
788 	struct bio *bio = rq->bio;
789 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
790 	unsigned int protect = 0;
791 
792 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
793 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
794 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
795 
796 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
797 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
798 	}
799 
800 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
801 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
802 
803 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
804 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
805 	}
806 
807 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
808 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
809 
810 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
811 			protect = 3 << 5;	/* Disable target PI checking */
812 		else
813 			protect = 1 << 5;	/* Enable target PI checking */
814 	}
815 
816 	scsi_set_prot_op(scmd, prot_op);
817 	scsi_set_prot_type(scmd, dif);
818 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
819 
820 	return protect;
821 }
822 
823 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
824 {
825 	struct request_queue *q = sdkp->disk->queue;
826 	unsigned int logical_block_size = sdkp->device->sector_size;
827 	unsigned int max_blocks = 0;
828 
829 	q->limits.discard_alignment =
830 		sdkp->unmap_alignment * logical_block_size;
831 	q->limits.discard_granularity =
832 		max(sdkp->physical_block_size,
833 		    sdkp->unmap_granularity * logical_block_size);
834 	sdkp->provisioning_mode = mode;
835 
836 	switch (mode) {
837 
838 	case SD_LBP_FULL:
839 	case SD_LBP_DISABLE:
840 		blk_queue_max_discard_sectors(q, 0);
841 		return;
842 
843 	case SD_LBP_UNMAP:
844 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
845 					  (u32)SD_MAX_WS16_BLOCKS);
846 		break;
847 
848 	case SD_LBP_WS16:
849 		if (sdkp->device->unmap_limit_for_ws)
850 			max_blocks = sdkp->max_unmap_blocks;
851 		else
852 			max_blocks = sdkp->max_ws_blocks;
853 
854 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
855 		break;
856 
857 	case SD_LBP_WS10:
858 		if (sdkp->device->unmap_limit_for_ws)
859 			max_blocks = sdkp->max_unmap_blocks;
860 		else
861 			max_blocks = sdkp->max_ws_blocks;
862 
863 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
864 		break;
865 
866 	case SD_LBP_ZERO:
867 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
868 					  (u32)SD_MAX_WS10_BLOCKS);
869 		break;
870 	}
871 
872 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
873 }
874 
875 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
876 {
877 	struct page *page;
878 
879 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
880 	if (!page)
881 		return NULL;
882 	clear_highpage(page);
883 	bvec_set_page(&rq->special_vec, page, data_len, 0);
884 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
885 	return bvec_virt(&rq->special_vec);
886 }
887 
888 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
889 {
890 	struct scsi_device *sdp = cmd->device;
891 	struct request *rq = scsi_cmd_to_rq(cmd);
892 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
893 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
894 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
895 	unsigned int data_len = 24;
896 	char *buf;
897 
898 	buf = sd_set_special_bvec(rq, data_len);
899 	if (!buf)
900 		return BLK_STS_RESOURCE;
901 
902 	cmd->cmd_len = 10;
903 	cmd->cmnd[0] = UNMAP;
904 	cmd->cmnd[8] = 24;
905 
906 	put_unaligned_be16(6 + 16, &buf[0]);
907 	put_unaligned_be16(16, &buf[2]);
908 	put_unaligned_be64(lba, &buf[8]);
909 	put_unaligned_be32(nr_blocks, &buf[16]);
910 
911 	cmd->allowed = sdkp->max_retries;
912 	cmd->transfersize = data_len;
913 	rq->timeout = SD_TIMEOUT;
914 
915 	return scsi_alloc_sgtables(cmd);
916 }
917 
918 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
919 		bool unmap)
920 {
921 	struct scsi_device *sdp = cmd->device;
922 	struct request *rq = scsi_cmd_to_rq(cmd);
923 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
924 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
925 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
926 	u32 data_len = sdp->sector_size;
927 
928 	if (!sd_set_special_bvec(rq, data_len))
929 		return BLK_STS_RESOURCE;
930 
931 	cmd->cmd_len = 16;
932 	cmd->cmnd[0] = WRITE_SAME_16;
933 	if (unmap)
934 		cmd->cmnd[1] = 0x8; /* UNMAP */
935 	put_unaligned_be64(lba, &cmd->cmnd[2]);
936 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
937 
938 	cmd->allowed = sdkp->max_retries;
939 	cmd->transfersize = data_len;
940 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
941 
942 	return scsi_alloc_sgtables(cmd);
943 }
944 
945 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
946 		bool unmap)
947 {
948 	struct scsi_device *sdp = cmd->device;
949 	struct request *rq = scsi_cmd_to_rq(cmd);
950 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
951 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
952 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
953 	u32 data_len = sdp->sector_size;
954 
955 	if (!sd_set_special_bvec(rq, data_len))
956 		return BLK_STS_RESOURCE;
957 
958 	cmd->cmd_len = 10;
959 	cmd->cmnd[0] = WRITE_SAME;
960 	if (unmap)
961 		cmd->cmnd[1] = 0x8; /* UNMAP */
962 	put_unaligned_be32(lba, &cmd->cmnd[2]);
963 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
964 
965 	cmd->allowed = sdkp->max_retries;
966 	cmd->transfersize = data_len;
967 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
968 
969 	return scsi_alloc_sgtables(cmd);
970 }
971 
972 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
973 {
974 	struct request *rq = scsi_cmd_to_rq(cmd);
975 	struct scsi_device *sdp = cmd->device;
976 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
977 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
978 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
979 
980 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
981 		switch (sdkp->zeroing_mode) {
982 		case SD_ZERO_WS16_UNMAP:
983 			return sd_setup_write_same16_cmnd(cmd, true);
984 		case SD_ZERO_WS10_UNMAP:
985 			return sd_setup_write_same10_cmnd(cmd, true);
986 		}
987 	}
988 
989 	if (sdp->no_write_same) {
990 		rq->rq_flags |= RQF_QUIET;
991 		return BLK_STS_TARGET;
992 	}
993 
994 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
995 		return sd_setup_write_same16_cmnd(cmd, false);
996 
997 	return sd_setup_write_same10_cmnd(cmd, false);
998 }
999 
1000 static void sd_config_write_same(struct scsi_disk *sdkp)
1001 {
1002 	struct request_queue *q = sdkp->disk->queue;
1003 	unsigned int logical_block_size = sdkp->device->sector_size;
1004 
1005 	if (sdkp->device->no_write_same) {
1006 		sdkp->max_ws_blocks = 0;
1007 		goto out;
1008 	}
1009 
1010 	/* Some devices can not handle block counts above 0xffff despite
1011 	 * supporting WRITE SAME(16). Consequently we default to 64k
1012 	 * blocks per I/O unless the device explicitly advertises a
1013 	 * bigger limit.
1014 	 */
1015 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
1016 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1017 						   (u32)SD_MAX_WS16_BLOCKS);
1018 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
1019 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
1020 						   (u32)SD_MAX_WS10_BLOCKS);
1021 	else {
1022 		sdkp->device->no_write_same = 1;
1023 		sdkp->max_ws_blocks = 0;
1024 	}
1025 
1026 	if (sdkp->lbprz && sdkp->lbpws)
1027 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
1028 	else if (sdkp->lbprz && sdkp->lbpws10)
1029 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
1030 	else if (sdkp->max_ws_blocks)
1031 		sdkp->zeroing_mode = SD_ZERO_WS;
1032 	else
1033 		sdkp->zeroing_mode = SD_ZERO_WRITE;
1034 
1035 	if (sdkp->max_ws_blocks &&
1036 	    sdkp->physical_block_size > logical_block_size) {
1037 		/*
1038 		 * Reporting a maximum number of blocks that is not aligned
1039 		 * on the device physical size would cause a large write same
1040 		 * request to be split into physically unaligned chunks by
1041 		 * __blkdev_issue_write_zeroes() even if the caller of this
1042 		 * functions took care to align the large request. So make sure
1043 		 * the maximum reported is aligned to the device physical block
1044 		 * size. This is only an optional optimization for regular
1045 		 * disks, but this is mandatory to avoid failure of large write
1046 		 * same requests directed at sequential write required zones of
1047 		 * host-managed ZBC disks.
1048 		 */
1049 		sdkp->max_ws_blocks =
1050 			round_down(sdkp->max_ws_blocks,
1051 				   bytes_to_logical(sdkp->device,
1052 						    sdkp->physical_block_size));
1053 	}
1054 
1055 out:
1056 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1057 					 (logical_block_size >> 9));
1058 }
1059 
1060 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1061 {
1062 	struct request *rq = scsi_cmd_to_rq(cmd);
1063 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1064 
1065 	/* flush requests don't perform I/O, zero the S/G table */
1066 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1067 
1068 	if (cmd->device->use_16_for_sync) {
1069 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1070 		cmd->cmd_len = 16;
1071 	} else {
1072 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1073 		cmd->cmd_len = 10;
1074 	}
1075 	cmd->transfersize = 0;
1076 	cmd->allowed = sdkp->max_retries;
1077 
1078 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1079 	return BLK_STS_OK;
1080 }
1081 
1082 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1083 				       sector_t lba, unsigned int nr_blocks,
1084 				       unsigned char flags, unsigned int dld)
1085 {
1086 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1087 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1088 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1089 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1090 	cmd->cmnd[10] = flags;
1091 	cmd->cmnd[11] = dld & 0x07;
1092 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1093 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1094 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1095 
1096 	return BLK_STS_OK;
1097 }
1098 
1099 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1100 				       sector_t lba, unsigned int nr_blocks,
1101 				       unsigned char flags, unsigned int dld)
1102 {
1103 	cmd->cmd_len  = 16;
1104 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1105 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1106 	cmd->cmnd[14] = (dld & 0x03) << 6;
1107 	cmd->cmnd[15] = 0;
1108 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1109 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1110 
1111 	return BLK_STS_OK;
1112 }
1113 
1114 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1115 				       sector_t lba, unsigned int nr_blocks,
1116 				       unsigned char flags)
1117 {
1118 	cmd->cmd_len = 10;
1119 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1120 	cmd->cmnd[1] = flags;
1121 	cmd->cmnd[6] = 0;
1122 	cmd->cmnd[9] = 0;
1123 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1124 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1125 
1126 	return BLK_STS_OK;
1127 }
1128 
1129 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1130 				      sector_t lba, unsigned int nr_blocks,
1131 				      unsigned char flags)
1132 {
1133 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1134 	if (WARN_ON_ONCE(nr_blocks == 0))
1135 		return BLK_STS_IOERR;
1136 
1137 	if (unlikely(flags & 0x8)) {
1138 		/*
1139 		 * This happens only if this drive failed 10byte rw
1140 		 * command with ILLEGAL_REQUEST during operation and
1141 		 * thus turned off use_10_for_rw.
1142 		 */
1143 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1144 		return BLK_STS_IOERR;
1145 	}
1146 
1147 	cmd->cmd_len = 6;
1148 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1149 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1150 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1151 	cmd->cmnd[3] = lba & 0xff;
1152 	cmd->cmnd[4] = nr_blocks;
1153 	cmd->cmnd[5] = 0;
1154 
1155 	return BLK_STS_OK;
1156 }
1157 
1158 /*
1159  * Check if a command has a duration limit set. If it does, and the target
1160  * device supports CDL and the feature is enabled, return the limit
1161  * descriptor index to use. Return 0 (no limit) otherwise.
1162  */
1163 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1164 {
1165 	struct scsi_device *sdp = sdkp->device;
1166 	int hint;
1167 
1168 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1169 		return 0;
1170 
1171 	/*
1172 	 * Use "no limit" if the request ioprio does not specify a duration
1173 	 * limit hint.
1174 	 */
1175 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1176 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1177 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1178 		return 0;
1179 
1180 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1181 }
1182 
1183 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1184 {
1185 	struct request *rq = scsi_cmd_to_rq(cmd);
1186 	struct scsi_device *sdp = cmd->device;
1187 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1188 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1189 	sector_t threshold;
1190 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1191 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1192 	bool write = rq_data_dir(rq) == WRITE;
1193 	unsigned char protect, fua;
1194 	unsigned int dld;
1195 	blk_status_t ret;
1196 	unsigned int dif;
1197 	bool dix;
1198 
1199 	ret = scsi_alloc_sgtables(cmd);
1200 	if (ret != BLK_STS_OK)
1201 		return ret;
1202 
1203 	ret = BLK_STS_IOERR;
1204 	if (!scsi_device_online(sdp) || sdp->changed) {
1205 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1206 		goto fail;
1207 	}
1208 
1209 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1210 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1211 		goto fail;
1212 	}
1213 
1214 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1215 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1216 		goto fail;
1217 	}
1218 
1219 	/*
1220 	 * Some SD card readers can't handle accesses which touch the
1221 	 * last one or two logical blocks. Split accesses as needed.
1222 	 */
1223 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1224 
1225 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1226 		if (lba < threshold) {
1227 			/* Access up to the threshold but not beyond */
1228 			nr_blocks = threshold - lba;
1229 		} else {
1230 			/* Access only a single logical block */
1231 			nr_blocks = 1;
1232 		}
1233 	}
1234 
1235 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1236 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1237 		if (ret)
1238 			goto fail;
1239 	}
1240 
1241 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1242 	dix = scsi_prot_sg_count(cmd);
1243 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1244 	dld = sd_cdl_dld(sdkp, cmd);
1245 
1246 	if (dif || dix)
1247 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1248 	else
1249 		protect = 0;
1250 
1251 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1252 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1253 					 protect | fua, dld);
1254 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1255 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1256 					 protect | fua, dld);
1257 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1258 		   sdp->use_10_for_rw || protect) {
1259 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1260 					 protect | fua);
1261 	} else {
1262 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1263 					protect | fua);
1264 	}
1265 
1266 	if (unlikely(ret != BLK_STS_OK))
1267 		goto fail;
1268 
1269 	/*
1270 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1271 	 * host adapter, it's safe to assume that we can at least transfer
1272 	 * this many bytes between each connect / disconnect.
1273 	 */
1274 	cmd->transfersize = sdp->sector_size;
1275 	cmd->underflow = nr_blocks << 9;
1276 	cmd->allowed = sdkp->max_retries;
1277 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1278 
1279 	SCSI_LOG_HLQUEUE(1,
1280 			 scmd_printk(KERN_INFO, cmd,
1281 				     "%s: block=%llu, count=%d\n", __func__,
1282 				     (unsigned long long)blk_rq_pos(rq),
1283 				     blk_rq_sectors(rq)));
1284 	SCSI_LOG_HLQUEUE(2,
1285 			 scmd_printk(KERN_INFO, cmd,
1286 				     "%s %d/%u 512 byte blocks.\n",
1287 				     write ? "writing" : "reading", nr_blocks,
1288 				     blk_rq_sectors(rq)));
1289 
1290 	/*
1291 	 * This indicates that the command is ready from our end to be queued.
1292 	 */
1293 	return BLK_STS_OK;
1294 fail:
1295 	scsi_free_sgtables(cmd);
1296 	return ret;
1297 }
1298 
1299 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1300 {
1301 	struct request *rq = scsi_cmd_to_rq(cmd);
1302 
1303 	switch (req_op(rq)) {
1304 	case REQ_OP_DISCARD:
1305 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1306 		case SD_LBP_UNMAP:
1307 			return sd_setup_unmap_cmnd(cmd);
1308 		case SD_LBP_WS16:
1309 			return sd_setup_write_same16_cmnd(cmd, true);
1310 		case SD_LBP_WS10:
1311 			return sd_setup_write_same10_cmnd(cmd, true);
1312 		case SD_LBP_ZERO:
1313 			return sd_setup_write_same10_cmnd(cmd, false);
1314 		default:
1315 			return BLK_STS_TARGET;
1316 		}
1317 	case REQ_OP_WRITE_ZEROES:
1318 		return sd_setup_write_zeroes_cmnd(cmd);
1319 	case REQ_OP_FLUSH:
1320 		return sd_setup_flush_cmnd(cmd);
1321 	case REQ_OP_READ:
1322 	case REQ_OP_WRITE:
1323 	case REQ_OP_ZONE_APPEND:
1324 		return sd_setup_read_write_cmnd(cmd);
1325 	case REQ_OP_ZONE_RESET:
1326 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1327 						   false);
1328 	case REQ_OP_ZONE_RESET_ALL:
1329 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1330 						   true);
1331 	case REQ_OP_ZONE_OPEN:
1332 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1333 	case REQ_OP_ZONE_CLOSE:
1334 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1335 	case REQ_OP_ZONE_FINISH:
1336 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1337 	default:
1338 		WARN_ON_ONCE(1);
1339 		return BLK_STS_NOTSUPP;
1340 	}
1341 }
1342 
1343 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1344 {
1345 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1346 
1347 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1348 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1349 }
1350 
1351 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1352 {
1353 	if (sdkp->device->removable || sdkp->write_prot) {
1354 		if (disk_check_media_change(disk))
1355 			return true;
1356 	}
1357 
1358 	/*
1359 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1360 	 * nothing to do with partitions, BLKRRPART is used to force a full
1361 	 * revalidate after things like a format for historical reasons.
1362 	 */
1363 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1364 }
1365 
1366 /**
1367  *	sd_open - open a scsi disk device
1368  *	@disk: disk to open
1369  *	@mode: open mode
1370  *
1371  *	Returns 0 if successful. Returns a negated errno value in case
1372  *	of error.
1373  *
1374  *	Note: This can be called from a user context (e.g. fsck(1) )
1375  *	or from within the kernel (e.g. as a result of a mount(1) ).
1376  *	In the latter case @inode and @filp carry an abridged amount
1377  *	of information as noted above.
1378  *
1379  *	Locking: called with disk->open_mutex held.
1380  **/
1381 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1382 {
1383 	struct scsi_disk *sdkp = scsi_disk(disk);
1384 	struct scsi_device *sdev = sdkp->device;
1385 	int retval;
1386 
1387 	if (scsi_device_get(sdev))
1388 		return -ENXIO;
1389 
1390 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1391 
1392 	/*
1393 	 * If the device is in error recovery, wait until it is done.
1394 	 * If the device is offline, then disallow any access to it.
1395 	 */
1396 	retval = -ENXIO;
1397 	if (!scsi_block_when_processing_errors(sdev))
1398 		goto error_out;
1399 
1400 	if (sd_need_revalidate(disk, sdkp))
1401 		sd_revalidate_disk(disk);
1402 
1403 	/*
1404 	 * If the drive is empty, just let the open fail.
1405 	 */
1406 	retval = -ENOMEDIUM;
1407 	if (sdev->removable && !sdkp->media_present &&
1408 	    !(mode & BLK_OPEN_NDELAY))
1409 		goto error_out;
1410 
1411 	/*
1412 	 * If the device has the write protect tab set, have the open fail
1413 	 * if the user expects to be able to write to the thing.
1414 	 */
1415 	retval = -EROFS;
1416 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1417 		goto error_out;
1418 
1419 	/*
1420 	 * It is possible that the disk changing stuff resulted in
1421 	 * the device being taken offline.  If this is the case,
1422 	 * report this to the user, and don't pretend that the
1423 	 * open actually succeeded.
1424 	 */
1425 	retval = -ENXIO;
1426 	if (!scsi_device_online(sdev))
1427 		goto error_out;
1428 
1429 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1430 		if (scsi_block_when_processing_errors(sdev))
1431 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1432 	}
1433 
1434 	return 0;
1435 
1436 error_out:
1437 	scsi_device_put(sdev);
1438 	return retval;
1439 }
1440 
1441 /**
1442  *	sd_release - invoked when the (last) close(2) is called on this
1443  *	scsi disk.
1444  *	@disk: disk to release
1445  *
1446  *	Returns 0.
1447  *
1448  *	Note: may block (uninterruptible) if error recovery is underway
1449  *	on this disk.
1450  *
1451  *	Locking: called with disk->open_mutex held.
1452  **/
1453 static void sd_release(struct gendisk *disk)
1454 {
1455 	struct scsi_disk *sdkp = scsi_disk(disk);
1456 	struct scsi_device *sdev = sdkp->device;
1457 
1458 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1459 
1460 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1461 		if (scsi_block_when_processing_errors(sdev))
1462 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1463 	}
1464 
1465 	scsi_device_put(sdev);
1466 }
1467 
1468 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1469 {
1470 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1471 	struct scsi_device *sdp = sdkp->device;
1472 	struct Scsi_Host *host = sdp->host;
1473 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1474 	int diskinfo[4];
1475 
1476 	/* default to most commonly used values */
1477 	diskinfo[0] = 0x40;	/* 1 << 6 */
1478 	diskinfo[1] = 0x20;	/* 1 << 5 */
1479 	diskinfo[2] = capacity >> 11;
1480 
1481 	/* override with calculated, extended default, or driver values */
1482 	if (host->hostt->bios_param)
1483 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1484 	else
1485 		scsicam_bios_param(bdev, capacity, diskinfo);
1486 
1487 	geo->heads = diskinfo[0];
1488 	geo->sectors = diskinfo[1];
1489 	geo->cylinders = diskinfo[2];
1490 	return 0;
1491 }
1492 
1493 /**
1494  *	sd_ioctl - process an ioctl
1495  *	@bdev: target block device
1496  *	@mode: open mode
1497  *	@cmd: ioctl command number
1498  *	@arg: this is third argument given to ioctl(2) system call.
1499  *	Often contains a pointer.
1500  *
1501  *	Returns 0 if successful (some ioctls return positive numbers on
1502  *	success as well). Returns a negated errno value in case of error.
1503  *
1504  *	Note: most ioctls are forward onto the block subsystem or further
1505  *	down in the scsi subsystem.
1506  **/
1507 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1508 		    unsigned int cmd, unsigned long arg)
1509 {
1510 	struct gendisk *disk = bdev->bd_disk;
1511 	struct scsi_disk *sdkp = scsi_disk(disk);
1512 	struct scsi_device *sdp = sdkp->device;
1513 	void __user *p = (void __user *)arg;
1514 	int error;
1515 
1516 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1517 				    "cmd=0x%x\n", disk->disk_name, cmd));
1518 
1519 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1520 		return -ENOIOCTLCMD;
1521 
1522 	/*
1523 	 * If we are in the middle of error recovery, don't let anyone
1524 	 * else try and use this device.  Also, if error recovery fails, it
1525 	 * may try and take the device offline, in which case all further
1526 	 * access to the device is prohibited.
1527 	 */
1528 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1529 			(mode & BLK_OPEN_NDELAY));
1530 	if (error)
1531 		return error;
1532 
1533 	if (is_sed_ioctl(cmd))
1534 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1535 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1536 }
1537 
1538 static void set_media_not_present(struct scsi_disk *sdkp)
1539 {
1540 	if (sdkp->media_present)
1541 		sdkp->device->changed = 1;
1542 
1543 	if (sdkp->device->removable) {
1544 		sdkp->media_present = 0;
1545 		sdkp->capacity = 0;
1546 	}
1547 }
1548 
1549 static int media_not_present(struct scsi_disk *sdkp,
1550 			     struct scsi_sense_hdr *sshdr)
1551 {
1552 	if (!scsi_sense_valid(sshdr))
1553 		return 0;
1554 
1555 	/* not invoked for commands that could return deferred errors */
1556 	switch (sshdr->sense_key) {
1557 	case UNIT_ATTENTION:
1558 	case NOT_READY:
1559 		/* medium not present */
1560 		if (sshdr->asc == 0x3A) {
1561 			set_media_not_present(sdkp);
1562 			return 1;
1563 		}
1564 	}
1565 	return 0;
1566 }
1567 
1568 /**
1569  *	sd_check_events - check media events
1570  *	@disk: kernel device descriptor
1571  *	@clearing: disk events currently being cleared
1572  *
1573  *	Returns mask of DISK_EVENT_*.
1574  *
1575  *	Note: this function is invoked from the block subsystem.
1576  **/
1577 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1578 {
1579 	struct scsi_disk *sdkp = disk->private_data;
1580 	struct scsi_device *sdp;
1581 	int retval;
1582 	bool disk_changed;
1583 
1584 	if (!sdkp)
1585 		return 0;
1586 
1587 	sdp = sdkp->device;
1588 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1589 
1590 	/*
1591 	 * If the device is offline, don't send any commands - just pretend as
1592 	 * if the command failed.  If the device ever comes back online, we
1593 	 * can deal with it then.  It is only because of unrecoverable errors
1594 	 * that we would ever take a device offline in the first place.
1595 	 */
1596 	if (!scsi_device_online(sdp)) {
1597 		set_media_not_present(sdkp);
1598 		goto out;
1599 	}
1600 
1601 	/*
1602 	 * Using TEST_UNIT_READY enables differentiation between drive with
1603 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1604 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1605 	 *
1606 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1607 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1608 	 * sd_revalidate() is called.
1609 	 */
1610 	if (scsi_block_when_processing_errors(sdp)) {
1611 		struct scsi_sense_hdr sshdr = { 0, };
1612 
1613 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1614 					      &sshdr);
1615 
1616 		/* failed to execute TUR, assume media not present */
1617 		if (retval < 0 || host_byte(retval)) {
1618 			set_media_not_present(sdkp);
1619 			goto out;
1620 		}
1621 
1622 		if (media_not_present(sdkp, &sshdr))
1623 			goto out;
1624 	}
1625 
1626 	/*
1627 	 * For removable scsi disk we have to recognise the presence
1628 	 * of a disk in the drive.
1629 	 */
1630 	if (!sdkp->media_present)
1631 		sdp->changed = 1;
1632 	sdkp->media_present = 1;
1633 out:
1634 	/*
1635 	 * sdp->changed is set under the following conditions:
1636 	 *
1637 	 *	Medium present state has changed in either direction.
1638 	 *	Device has indicated UNIT_ATTENTION.
1639 	 */
1640 	disk_changed = sdp->changed;
1641 	sdp->changed = 0;
1642 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1643 }
1644 
1645 static int sd_sync_cache(struct scsi_disk *sdkp)
1646 {
1647 	int retries, res;
1648 	struct scsi_device *sdp = sdkp->device;
1649 	const int timeout = sdp->request_queue->rq_timeout
1650 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1651 	struct scsi_sense_hdr sshdr;
1652 	const struct scsi_exec_args exec_args = {
1653 		.req_flags = BLK_MQ_REQ_PM,
1654 		.sshdr = &sshdr,
1655 	};
1656 
1657 	if (!scsi_device_online(sdp))
1658 		return -ENODEV;
1659 
1660 	for (retries = 3; retries > 0; --retries) {
1661 		unsigned char cmd[16] = { 0 };
1662 
1663 		if (sdp->use_16_for_sync)
1664 			cmd[0] = SYNCHRONIZE_CACHE_16;
1665 		else
1666 			cmd[0] = SYNCHRONIZE_CACHE;
1667 		/*
1668 		 * Leave the rest of the command zero to indicate
1669 		 * flush everything.
1670 		 */
1671 		res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1672 				       timeout, sdkp->max_retries, &exec_args);
1673 		if (res == 0)
1674 			break;
1675 	}
1676 
1677 	if (res) {
1678 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1679 
1680 		if (res < 0)
1681 			return res;
1682 
1683 		if (scsi_status_is_check_condition(res) &&
1684 		    scsi_sense_valid(&sshdr)) {
1685 			sd_print_sense_hdr(sdkp, &sshdr);
1686 
1687 			/* we need to evaluate the error return  */
1688 			if (sshdr.asc == 0x3a ||	/* medium not present */
1689 			    sshdr.asc == 0x20 ||	/* invalid command */
1690 			    (sshdr.asc == 0x74 && sshdr.ascq == 0x71))	/* drive is password locked */
1691 				/* this is no error here */
1692 				return 0;
1693 			/*
1694 			 * This drive doesn't support sync and there's not much
1695 			 * we can do because this is called during shutdown
1696 			 * or suspend so just return success so those operations
1697 			 * can proceed.
1698 			 */
1699 			if (sshdr.sense_key == ILLEGAL_REQUEST)
1700 				return 0;
1701 		}
1702 
1703 		switch (host_byte(res)) {
1704 		/* ignore errors due to racing a disconnection */
1705 		case DID_BAD_TARGET:
1706 		case DID_NO_CONNECT:
1707 			return 0;
1708 		/* signal the upper layer it might try again */
1709 		case DID_BUS_BUSY:
1710 		case DID_IMM_RETRY:
1711 		case DID_REQUEUE:
1712 		case DID_SOFT_ERROR:
1713 			return -EBUSY;
1714 		default:
1715 			return -EIO;
1716 		}
1717 	}
1718 	return 0;
1719 }
1720 
1721 static void sd_rescan(struct device *dev)
1722 {
1723 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1724 
1725 	sd_revalidate_disk(sdkp->disk);
1726 }
1727 
1728 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1729 		enum blk_unique_id type)
1730 {
1731 	struct scsi_device *sdev = scsi_disk(disk)->device;
1732 	const struct scsi_vpd *vpd;
1733 	const unsigned char *d;
1734 	int ret = -ENXIO, len;
1735 
1736 	rcu_read_lock();
1737 	vpd = rcu_dereference(sdev->vpd_pg83);
1738 	if (!vpd)
1739 		goto out_unlock;
1740 
1741 	ret = -EINVAL;
1742 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1743 		/* we only care about designators with LU association */
1744 		if (((d[1] >> 4) & 0x3) != 0x00)
1745 			continue;
1746 		if ((d[1] & 0xf) != type)
1747 			continue;
1748 
1749 		/*
1750 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1751 		 * keep looking as one with more entropy might still show up.
1752 		 */
1753 		len = d[3];
1754 		if (len != 8 && len != 12 && len != 16)
1755 			continue;
1756 		ret = len;
1757 		memcpy(id, d + 4, len);
1758 		if (len == 16)
1759 			break;
1760 	}
1761 out_unlock:
1762 	rcu_read_unlock();
1763 	return ret;
1764 }
1765 
1766 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1767 {
1768 	switch (host_byte(result)) {
1769 	case DID_TRANSPORT_MARGINAL:
1770 	case DID_TRANSPORT_DISRUPTED:
1771 	case DID_BUS_BUSY:
1772 		return PR_STS_RETRY_PATH_FAILURE;
1773 	case DID_NO_CONNECT:
1774 		return PR_STS_PATH_FAILED;
1775 	case DID_TRANSPORT_FAILFAST:
1776 		return PR_STS_PATH_FAST_FAILED;
1777 	}
1778 
1779 	switch (status_byte(result)) {
1780 	case SAM_STAT_RESERVATION_CONFLICT:
1781 		return PR_STS_RESERVATION_CONFLICT;
1782 	case SAM_STAT_CHECK_CONDITION:
1783 		if (!scsi_sense_valid(sshdr))
1784 			return PR_STS_IOERR;
1785 
1786 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1787 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1788 			return -EINVAL;
1789 
1790 		fallthrough;
1791 	default:
1792 		return PR_STS_IOERR;
1793 	}
1794 }
1795 
1796 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1797 			    unsigned char *data, int data_len)
1798 {
1799 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1800 	struct scsi_device *sdev = sdkp->device;
1801 	struct scsi_sense_hdr sshdr;
1802 	u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1803 	const struct scsi_exec_args exec_args = {
1804 		.sshdr = &sshdr,
1805 	};
1806 	int result;
1807 
1808 	put_unaligned_be16(data_len, &cmd[7]);
1809 
1810 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1811 				  SD_TIMEOUT, sdkp->max_retries, &exec_args);
1812 	if (scsi_status_is_check_condition(result) &&
1813 	    scsi_sense_valid(&sshdr)) {
1814 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1815 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1816 	}
1817 
1818 	if (result <= 0)
1819 		return result;
1820 
1821 	return sd_scsi_to_pr_err(&sshdr, result);
1822 }
1823 
1824 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1825 {
1826 	int result, i, data_offset, num_copy_keys;
1827 	u32 num_keys = keys_info->num_keys;
1828 	int data_len = num_keys * 8 + 8;
1829 	u8 *data;
1830 
1831 	data = kzalloc(data_len, GFP_KERNEL);
1832 	if (!data)
1833 		return -ENOMEM;
1834 
1835 	result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1836 	if (result)
1837 		goto free_data;
1838 
1839 	keys_info->generation = get_unaligned_be32(&data[0]);
1840 	keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1841 
1842 	data_offset = 8;
1843 	num_copy_keys = min(num_keys, keys_info->num_keys);
1844 
1845 	for (i = 0; i < num_copy_keys; i++) {
1846 		keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1847 		data_offset += 8;
1848 	}
1849 
1850 free_data:
1851 	kfree(data);
1852 	return result;
1853 }
1854 
1855 static int sd_pr_read_reservation(struct block_device *bdev,
1856 				  struct pr_held_reservation *rsv)
1857 {
1858 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1859 	struct scsi_device *sdev = sdkp->device;
1860 	u8 data[24] = { };
1861 	int result, len;
1862 
1863 	result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1864 	if (result)
1865 		return result;
1866 
1867 	len = get_unaligned_be32(&data[4]);
1868 	if (!len)
1869 		return 0;
1870 
1871 	/* Make sure we have at least the key and type */
1872 	if (len < 14) {
1873 		sdev_printk(KERN_INFO, sdev,
1874 			    "READ RESERVATION failed due to short return buffer of %d bytes\n",
1875 			    len);
1876 		return -EINVAL;
1877 	}
1878 
1879 	rsv->generation = get_unaligned_be32(&data[0]);
1880 	rsv->key = get_unaligned_be64(&data[8]);
1881 	rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1882 	return 0;
1883 }
1884 
1885 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1886 			     u64 sa_key, enum scsi_pr_type type, u8 flags)
1887 {
1888 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1889 	struct scsi_device *sdev = sdkp->device;
1890 	struct scsi_sense_hdr sshdr;
1891 	const struct scsi_exec_args exec_args = {
1892 		.sshdr = &sshdr,
1893 	};
1894 	int result;
1895 	u8 cmd[16] = { 0, };
1896 	u8 data[24] = { 0, };
1897 
1898 	cmd[0] = PERSISTENT_RESERVE_OUT;
1899 	cmd[1] = sa;
1900 	cmd[2] = type;
1901 	put_unaligned_be32(sizeof(data), &cmd[5]);
1902 
1903 	put_unaligned_be64(key, &data[0]);
1904 	put_unaligned_be64(sa_key, &data[8]);
1905 	data[20] = flags;
1906 
1907 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1908 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1909 				  &exec_args);
1910 
1911 	if (scsi_status_is_check_condition(result) &&
1912 	    scsi_sense_valid(&sshdr)) {
1913 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1914 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1915 	}
1916 
1917 	if (result <= 0)
1918 		return result;
1919 
1920 	return sd_scsi_to_pr_err(&sshdr, result);
1921 }
1922 
1923 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1924 		u32 flags)
1925 {
1926 	if (flags & ~PR_FL_IGNORE_KEY)
1927 		return -EOPNOTSUPP;
1928 	return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1929 			old_key, new_key, 0,
1930 			(1 << 0) /* APTPL */);
1931 }
1932 
1933 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1934 		u32 flags)
1935 {
1936 	if (flags)
1937 		return -EOPNOTSUPP;
1938 	return sd_pr_out_command(bdev, 0x01, key, 0,
1939 				 block_pr_type_to_scsi(type), 0);
1940 }
1941 
1942 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1943 {
1944 	return sd_pr_out_command(bdev, 0x02, key, 0,
1945 				 block_pr_type_to_scsi(type), 0);
1946 }
1947 
1948 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1949 		enum pr_type type, bool abort)
1950 {
1951 	return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1952 				 block_pr_type_to_scsi(type), 0);
1953 }
1954 
1955 static int sd_pr_clear(struct block_device *bdev, u64 key)
1956 {
1957 	return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
1958 }
1959 
1960 static const struct pr_ops sd_pr_ops = {
1961 	.pr_register	= sd_pr_register,
1962 	.pr_reserve	= sd_pr_reserve,
1963 	.pr_release	= sd_pr_release,
1964 	.pr_preempt	= sd_pr_preempt,
1965 	.pr_clear	= sd_pr_clear,
1966 	.pr_read_keys	= sd_pr_read_keys,
1967 	.pr_read_reservation = sd_pr_read_reservation,
1968 };
1969 
1970 static void scsi_disk_free_disk(struct gendisk *disk)
1971 {
1972 	struct scsi_disk *sdkp = scsi_disk(disk);
1973 
1974 	put_device(&sdkp->disk_dev);
1975 }
1976 
1977 static const struct block_device_operations sd_fops = {
1978 	.owner			= THIS_MODULE,
1979 	.open			= sd_open,
1980 	.release		= sd_release,
1981 	.ioctl			= sd_ioctl,
1982 	.getgeo			= sd_getgeo,
1983 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1984 	.check_events		= sd_check_events,
1985 	.unlock_native_capacity	= sd_unlock_native_capacity,
1986 	.report_zones		= sd_zbc_report_zones,
1987 	.get_unique_id		= sd_get_unique_id,
1988 	.free_disk		= scsi_disk_free_disk,
1989 	.pr_ops			= &sd_pr_ops,
1990 };
1991 
1992 /**
1993  *	sd_eh_reset - reset error handling callback
1994  *	@scmd:		sd-issued command that has failed
1995  *
1996  *	This function is called by the SCSI midlayer before starting
1997  *	SCSI EH. When counting medium access failures we have to be
1998  *	careful to register it only only once per device and SCSI EH run;
1999  *	there might be several timed out commands which will cause the
2000  *	'max_medium_access_timeouts' counter to trigger after the first
2001  *	SCSI EH run already and set the device to offline.
2002  *	So this function resets the internal counter before starting SCSI EH.
2003  **/
2004 static void sd_eh_reset(struct scsi_cmnd *scmd)
2005 {
2006 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2007 
2008 	/* New SCSI EH run, reset gate variable */
2009 	sdkp->ignore_medium_access_errors = false;
2010 }
2011 
2012 /**
2013  *	sd_eh_action - error handling callback
2014  *	@scmd:		sd-issued command that has failed
2015  *	@eh_disp:	The recovery disposition suggested by the midlayer
2016  *
2017  *	This function is called by the SCSI midlayer upon completion of an
2018  *	error test command (currently TEST UNIT READY). The result of sending
2019  *	the eh command is passed in eh_disp.  We're looking for devices that
2020  *	fail medium access commands but are OK with non access commands like
2021  *	test unit ready (so wrongly see the device as having a successful
2022  *	recovery)
2023  **/
2024 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
2025 {
2026 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
2027 	struct scsi_device *sdev = scmd->device;
2028 
2029 	if (!scsi_device_online(sdev) ||
2030 	    !scsi_medium_access_command(scmd) ||
2031 	    host_byte(scmd->result) != DID_TIME_OUT ||
2032 	    eh_disp != SUCCESS)
2033 		return eh_disp;
2034 
2035 	/*
2036 	 * The device has timed out executing a medium access command.
2037 	 * However, the TEST UNIT READY command sent during error
2038 	 * handling completed successfully. Either the device is in the
2039 	 * process of recovering or has it suffered an internal failure
2040 	 * that prevents access to the storage medium.
2041 	 */
2042 	if (!sdkp->ignore_medium_access_errors) {
2043 		sdkp->medium_access_timed_out++;
2044 		sdkp->ignore_medium_access_errors = true;
2045 	}
2046 
2047 	/*
2048 	 * If the device keeps failing read/write commands but TEST UNIT
2049 	 * READY always completes successfully we assume that medium
2050 	 * access is no longer possible and take the device offline.
2051 	 */
2052 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2053 		scmd_printk(KERN_ERR, scmd,
2054 			    "Medium access timeout failure. Offlining disk!\n");
2055 		mutex_lock(&sdev->state_mutex);
2056 		scsi_device_set_state(sdev, SDEV_OFFLINE);
2057 		mutex_unlock(&sdev->state_mutex);
2058 
2059 		return SUCCESS;
2060 	}
2061 
2062 	return eh_disp;
2063 }
2064 
2065 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2066 {
2067 	struct request *req = scsi_cmd_to_rq(scmd);
2068 	struct scsi_device *sdev = scmd->device;
2069 	unsigned int transferred, good_bytes;
2070 	u64 start_lba, end_lba, bad_lba;
2071 
2072 	/*
2073 	 * Some commands have a payload smaller than the device logical
2074 	 * block size (e.g. INQUIRY on a 4K disk).
2075 	 */
2076 	if (scsi_bufflen(scmd) <= sdev->sector_size)
2077 		return 0;
2078 
2079 	/* Check if we have a 'bad_lba' information */
2080 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2081 				     SCSI_SENSE_BUFFERSIZE,
2082 				     &bad_lba))
2083 		return 0;
2084 
2085 	/*
2086 	 * If the bad lba was reported incorrectly, we have no idea where
2087 	 * the error is.
2088 	 */
2089 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2090 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2091 	if (bad_lba < start_lba || bad_lba >= end_lba)
2092 		return 0;
2093 
2094 	/*
2095 	 * resid is optional but mostly filled in.  When it's unused,
2096 	 * its value is zero, so we assume the whole buffer transferred
2097 	 */
2098 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2099 
2100 	/* This computation should always be done in terms of the
2101 	 * resolution of the device's medium.
2102 	 */
2103 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2104 
2105 	return min(good_bytes, transferred);
2106 }
2107 
2108 /**
2109  *	sd_done - bottom half handler: called when the lower level
2110  *	driver has completed (successfully or otherwise) a scsi command.
2111  *	@SCpnt: mid-level's per command structure.
2112  *
2113  *	Note: potentially run from within an ISR. Must not block.
2114  **/
2115 static int sd_done(struct scsi_cmnd *SCpnt)
2116 {
2117 	int result = SCpnt->result;
2118 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2119 	unsigned int sector_size = SCpnt->device->sector_size;
2120 	unsigned int resid;
2121 	struct scsi_sense_hdr sshdr;
2122 	struct request *req = scsi_cmd_to_rq(SCpnt);
2123 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2124 	int sense_valid = 0;
2125 	int sense_deferred = 0;
2126 
2127 	switch (req_op(req)) {
2128 	case REQ_OP_DISCARD:
2129 	case REQ_OP_WRITE_ZEROES:
2130 	case REQ_OP_ZONE_RESET:
2131 	case REQ_OP_ZONE_RESET_ALL:
2132 	case REQ_OP_ZONE_OPEN:
2133 	case REQ_OP_ZONE_CLOSE:
2134 	case REQ_OP_ZONE_FINISH:
2135 		if (!result) {
2136 			good_bytes = blk_rq_bytes(req);
2137 			scsi_set_resid(SCpnt, 0);
2138 		} else {
2139 			good_bytes = 0;
2140 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2141 		}
2142 		break;
2143 	default:
2144 		/*
2145 		 * In case of bogus fw or device, we could end up having
2146 		 * an unaligned partial completion. Check this here and force
2147 		 * alignment.
2148 		 */
2149 		resid = scsi_get_resid(SCpnt);
2150 		if (resid & (sector_size - 1)) {
2151 			sd_printk(KERN_INFO, sdkp,
2152 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2153 				resid, sector_size);
2154 			scsi_print_command(SCpnt);
2155 			resid = min(scsi_bufflen(SCpnt),
2156 				    round_up(resid, sector_size));
2157 			scsi_set_resid(SCpnt, resid);
2158 		}
2159 	}
2160 
2161 	if (result) {
2162 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2163 		if (sense_valid)
2164 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2165 	}
2166 	sdkp->medium_access_timed_out = 0;
2167 
2168 	if (!scsi_status_is_check_condition(result) &&
2169 	    (!sense_valid || sense_deferred))
2170 		goto out;
2171 
2172 	switch (sshdr.sense_key) {
2173 	case HARDWARE_ERROR:
2174 	case MEDIUM_ERROR:
2175 		good_bytes = sd_completed_bytes(SCpnt);
2176 		break;
2177 	case RECOVERED_ERROR:
2178 		good_bytes = scsi_bufflen(SCpnt);
2179 		break;
2180 	case NO_SENSE:
2181 		/* This indicates a false check condition, so ignore it.  An
2182 		 * unknown amount of data was transferred so treat it as an
2183 		 * error.
2184 		 */
2185 		SCpnt->result = 0;
2186 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2187 		break;
2188 	case ABORTED_COMMAND:
2189 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2190 			good_bytes = sd_completed_bytes(SCpnt);
2191 		break;
2192 	case ILLEGAL_REQUEST:
2193 		switch (sshdr.asc) {
2194 		case 0x10:	/* DIX: Host detected corruption */
2195 			good_bytes = sd_completed_bytes(SCpnt);
2196 			break;
2197 		case 0x20:	/* INVALID COMMAND OPCODE */
2198 		case 0x24:	/* INVALID FIELD IN CDB */
2199 			switch (SCpnt->cmnd[0]) {
2200 			case UNMAP:
2201 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2202 				break;
2203 			case WRITE_SAME_16:
2204 			case WRITE_SAME:
2205 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2206 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2207 				} else {
2208 					sdkp->device->no_write_same = 1;
2209 					sd_config_write_same(sdkp);
2210 					req->rq_flags |= RQF_QUIET;
2211 				}
2212 				break;
2213 			}
2214 		}
2215 		break;
2216 	default:
2217 		break;
2218 	}
2219 
2220  out:
2221 	if (sd_is_zoned(sdkp))
2222 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2223 
2224 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2225 					   "sd_done: completed %d of %d bytes\n",
2226 					   good_bytes, scsi_bufflen(SCpnt)));
2227 
2228 	return good_bytes;
2229 }
2230 
2231 /*
2232  * spinup disk - called only in sd_revalidate_disk()
2233  */
2234 static void
2235 sd_spinup_disk(struct scsi_disk *sdkp)
2236 {
2237 	unsigned char cmd[10];
2238 	unsigned long spintime_expire = 0;
2239 	int retries, spintime;
2240 	unsigned int the_result;
2241 	struct scsi_sense_hdr sshdr;
2242 	const struct scsi_exec_args exec_args = {
2243 		.sshdr = &sshdr,
2244 	};
2245 	int sense_valid = 0;
2246 
2247 	spintime = 0;
2248 
2249 	/* Spin up drives, as required.  Only do this at boot time */
2250 	/* Spinup needs to be done for module loads too. */
2251 	do {
2252 		retries = 0;
2253 
2254 		do {
2255 			bool media_was_present = sdkp->media_present;
2256 
2257 			cmd[0] = TEST_UNIT_READY;
2258 			memset((void *) &cmd[1], 0, 9);
2259 
2260 			the_result = scsi_execute_cmd(sdkp->device, cmd,
2261 						      REQ_OP_DRV_IN, NULL, 0,
2262 						      SD_TIMEOUT,
2263 						      sdkp->max_retries,
2264 						      &exec_args);
2265 
2266 			/*
2267 			 * If the drive has indicated to us that it
2268 			 * doesn't have any media in it, don't bother
2269 			 * with any more polling.
2270 			 */
2271 			if (media_not_present(sdkp, &sshdr)) {
2272 				if (media_was_present)
2273 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2274 				return;
2275 			}
2276 
2277 			if (the_result)
2278 				sense_valid = scsi_sense_valid(&sshdr);
2279 			retries++;
2280 		} while (retries < 3 &&
2281 			 (!scsi_status_is_good(the_result) ||
2282 			  (scsi_status_is_check_condition(the_result) &&
2283 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2284 
2285 		if (!scsi_status_is_check_condition(the_result)) {
2286 			/* no sense, TUR either succeeded or failed
2287 			 * with a status error */
2288 			if(!spintime && !scsi_status_is_good(the_result)) {
2289 				sd_print_result(sdkp, "Test Unit Ready failed",
2290 						the_result);
2291 			}
2292 			break;
2293 		}
2294 
2295 		/*
2296 		 * The device does not want the automatic start to be issued.
2297 		 */
2298 		if (sdkp->device->no_start_on_add)
2299 			break;
2300 
2301 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2302 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2303 				break;	/* manual intervention required */
2304 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2305 				break;	/* standby */
2306 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2307 				break;	/* unavailable */
2308 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2309 				break;	/* sanitize in progress */
2310 			/*
2311 			 * Issue command to spin up drive when not ready
2312 			 */
2313 			if (!spintime) {
2314 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2315 				cmd[0] = START_STOP;
2316 				cmd[1] = 1;	/* Return immediately */
2317 				memset((void *) &cmd[2], 0, 8);
2318 				cmd[4] = 1;	/* Start spin cycle */
2319 				if (sdkp->device->start_stop_pwr_cond)
2320 					cmd[4] |= 1 << 4;
2321 				scsi_execute_cmd(sdkp->device, cmd,
2322 						 REQ_OP_DRV_IN, NULL, 0,
2323 						 SD_TIMEOUT, sdkp->max_retries,
2324 						 &exec_args);
2325 				spintime_expire = jiffies + 100 * HZ;
2326 				spintime = 1;
2327 			}
2328 			/* Wait 1 second for next try */
2329 			msleep(1000);
2330 			printk(KERN_CONT ".");
2331 
2332 		/*
2333 		 * Wait for USB flash devices with slow firmware.
2334 		 * Yes, this sense key/ASC combination shouldn't
2335 		 * occur here.  It's characteristic of these devices.
2336 		 */
2337 		} else if (sense_valid &&
2338 				sshdr.sense_key == UNIT_ATTENTION &&
2339 				sshdr.asc == 0x28) {
2340 			if (!spintime) {
2341 				spintime_expire = jiffies + 5 * HZ;
2342 				spintime = 1;
2343 			}
2344 			/* Wait 1 second for next try */
2345 			msleep(1000);
2346 		} else {
2347 			/* we don't understand the sense code, so it's
2348 			 * probably pointless to loop */
2349 			if(!spintime) {
2350 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2351 				sd_print_sense_hdr(sdkp, &sshdr);
2352 			}
2353 			break;
2354 		}
2355 
2356 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2357 
2358 	if (spintime) {
2359 		if (scsi_status_is_good(the_result))
2360 			printk(KERN_CONT "ready\n");
2361 		else
2362 			printk(KERN_CONT "not responding...\n");
2363 	}
2364 }
2365 
2366 /*
2367  * Determine whether disk supports Data Integrity Field.
2368  */
2369 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2370 {
2371 	struct scsi_device *sdp = sdkp->device;
2372 	u8 type;
2373 
2374 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2375 		sdkp->protection_type = 0;
2376 		return 0;
2377 	}
2378 
2379 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2380 
2381 	if (type > T10_PI_TYPE3_PROTECTION) {
2382 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2383 			  " protection type %u. Disabling disk!\n",
2384 			  type);
2385 		sdkp->protection_type = 0;
2386 		return -ENODEV;
2387 	}
2388 
2389 	sdkp->protection_type = type;
2390 
2391 	return 0;
2392 }
2393 
2394 static void sd_config_protection(struct scsi_disk *sdkp)
2395 {
2396 	struct scsi_device *sdp = sdkp->device;
2397 
2398 	sd_dif_config_host(sdkp);
2399 
2400 	if (!sdkp->protection_type)
2401 		return;
2402 
2403 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2404 		sd_first_printk(KERN_NOTICE, sdkp,
2405 				"Disabling DIF Type %u protection\n",
2406 				sdkp->protection_type);
2407 		sdkp->protection_type = 0;
2408 	}
2409 
2410 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2411 			sdkp->protection_type);
2412 }
2413 
2414 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2415 			struct scsi_sense_hdr *sshdr, int sense_valid,
2416 			int the_result)
2417 {
2418 	if (sense_valid)
2419 		sd_print_sense_hdr(sdkp, sshdr);
2420 	else
2421 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2422 
2423 	/*
2424 	 * Set dirty bit for removable devices if not ready -
2425 	 * sometimes drives will not report this properly.
2426 	 */
2427 	if (sdp->removable &&
2428 	    sense_valid && sshdr->sense_key == NOT_READY)
2429 		set_media_not_present(sdkp);
2430 
2431 	/*
2432 	 * We used to set media_present to 0 here to indicate no media
2433 	 * in the drive, but some drives fail read capacity even with
2434 	 * media present, so we can't do that.
2435 	 */
2436 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2437 }
2438 
2439 #define RC16_LEN 32
2440 #if RC16_LEN > SD_BUF_SIZE
2441 #error RC16_LEN must not be more than SD_BUF_SIZE
2442 #endif
2443 
2444 #define READ_CAPACITY_RETRIES_ON_RESET	10
2445 
2446 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2447 						unsigned char *buffer)
2448 {
2449 	unsigned char cmd[16];
2450 	struct scsi_sense_hdr sshdr;
2451 	const struct scsi_exec_args exec_args = {
2452 		.sshdr = &sshdr,
2453 	};
2454 	int sense_valid = 0;
2455 	int the_result;
2456 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2457 	unsigned int alignment;
2458 	unsigned long long lba;
2459 	unsigned sector_size;
2460 
2461 	if (sdp->no_read_capacity_16)
2462 		return -EINVAL;
2463 
2464 	do {
2465 		memset(cmd, 0, 16);
2466 		cmd[0] = SERVICE_ACTION_IN_16;
2467 		cmd[1] = SAI_READ_CAPACITY_16;
2468 		cmd[13] = RC16_LEN;
2469 		memset(buffer, 0, RC16_LEN);
2470 
2471 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2472 					      buffer, RC16_LEN, SD_TIMEOUT,
2473 					      sdkp->max_retries, &exec_args);
2474 
2475 		if (media_not_present(sdkp, &sshdr))
2476 			return -ENODEV;
2477 
2478 		if (the_result > 0) {
2479 			sense_valid = scsi_sense_valid(&sshdr);
2480 			if (sense_valid &&
2481 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2482 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2483 			    sshdr.ascq == 0x00)
2484 				/* Invalid Command Operation Code or
2485 				 * Invalid Field in CDB, just retry
2486 				 * silently with RC10 */
2487 				return -EINVAL;
2488 			if (sense_valid &&
2489 			    sshdr.sense_key == UNIT_ATTENTION &&
2490 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2491 				/* Device reset might occur several times,
2492 				 * give it one more chance */
2493 				if (--reset_retries > 0)
2494 					continue;
2495 		}
2496 		retries--;
2497 
2498 	} while (the_result && retries);
2499 
2500 	if (the_result) {
2501 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2502 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2503 		return -EINVAL;
2504 	}
2505 
2506 	sector_size = get_unaligned_be32(&buffer[8]);
2507 	lba = get_unaligned_be64(&buffer[0]);
2508 
2509 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2510 		sdkp->capacity = 0;
2511 		return -ENODEV;
2512 	}
2513 
2514 	/* Logical blocks per physical block exponent */
2515 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2516 
2517 	/* RC basis */
2518 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2519 
2520 	/* Lowest aligned logical block */
2521 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2522 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2523 	if (alignment && sdkp->first_scan)
2524 		sd_printk(KERN_NOTICE, sdkp,
2525 			  "physical block alignment offset: %u\n", alignment);
2526 
2527 	if (buffer[14] & 0x80) { /* LBPME */
2528 		sdkp->lbpme = 1;
2529 
2530 		if (buffer[14] & 0x40) /* LBPRZ */
2531 			sdkp->lbprz = 1;
2532 
2533 		sd_config_discard(sdkp, SD_LBP_WS16);
2534 	}
2535 
2536 	sdkp->capacity = lba + 1;
2537 	return sector_size;
2538 }
2539 
2540 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2541 						unsigned char *buffer)
2542 {
2543 	unsigned char cmd[16];
2544 	struct scsi_sense_hdr sshdr;
2545 	const struct scsi_exec_args exec_args = {
2546 		.sshdr = &sshdr,
2547 	};
2548 	int sense_valid = 0;
2549 	int the_result;
2550 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2551 	sector_t lba;
2552 	unsigned sector_size;
2553 
2554 	do {
2555 		cmd[0] = READ_CAPACITY;
2556 		memset(&cmd[1], 0, 9);
2557 		memset(buffer, 0, 8);
2558 
2559 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2560 					      8, SD_TIMEOUT, sdkp->max_retries,
2561 					      &exec_args);
2562 
2563 		if (media_not_present(sdkp, &sshdr))
2564 			return -ENODEV;
2565 
2566 		if (the_result > 0) {
2567 			sense_valid = scsi_sense_valid(&sshdr);
2568 			if (sense_valid &&
2569 			    sshdr.sense_key == UNIT_ATTENTION &&
2570 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2571 				/* Device reset might occur several times,
2572 				 * give it one more chance */
2573 				if (--reset_retries > 0)
2574 					continue;
2575 		}
2576 		retries--;
2577 
2578 	} while (the_result && retries);
2579 
2580 	if (the_result) {
2581 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2582 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2583 		return -EINVAL;
2584 	}
2585 
2586 	sector_size = get_unaligned_be32(&buffer[4]);
2587 	lba = get_unaligned_be32(&buffer[0]);
2588 
2589 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2590 		/* Some buggy (usb cardreader) devices return an lba of
2591 		   0xffffffff when the want to report a size of 0 (with
2592 		   which they really mean no media is present) */
2593 		sdkp->capacity = 0;
2594 		sdkp->physical_block_size = sector_size;
2595 		return sector_size;
2596 	}
2597 
2598 	sdkp->capacity = lba + 1;
2599 	sdkp->physical_block_size = sector_size;
2600 	return sector_size;
2601 }
2602 
2603 static int sd_try_rc16_first(struct scsi_device *sdp)
2604 {
2605 	if (sdp->host->max_cmd_len < 16)
2606 		return 0;
2607 	if (sdp->try_rc_10_first)
2608 		return 0;
2609 	if (sdp->scsi_level > SCSI_SPC_2)
2610 		return 1;
2611 	if (scsi_device_protection(sdp))
2612 		return 1;
2613 	return 0;
2614 }
2615 
2616 /*
2617  * read disk capacity
2618  */
2619 static void
2620 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2621 {
2622 	int sector_size;
2623 	struct scsi_device *sdp = sdkp->device;
2624 
2625 	if (sd_try_rc16_first(sdp)) {
2626 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2627 		if (sector_size == -EOVERFLOW)
2628 			goto got_data;
2629 		if (sector_size == -ENODEV)
2630 			return;
2631 		if (sector_size < 0)
2632 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2633 		if (sector_size < 0)
2634 			return;
2635 	} else {
2636 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2637 		if (sector_size == -EOVERFLOW)
2638 			goto got_data;
2639 		if (sector_size < 0)
2640 			return;
2641 		if ((sizeof(sdkp->capacity) > 4) &&
2642 		    (sdkp->capacity > 0xffffffffULL)) {
2643 			int old_sector_size = sector_size;
2644 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2645 					"Trying to use READ CAPACITY(16).\n");
2646 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2647 			if (sector_size < 0) {
2648 				sd_printk(KERN_NOTICE, sdkp,
2649 					"Using 0xffffffff as device size\n");
2650 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2651 				sector_size = old_sector_size;
2652 				goto got_data;
2653 			}
2654 			/* Remember that READ CAPACITY(16) succeeded */
2655 			sdp->try_rc_10_first = 0;
2656 		}
2657 	}
2658 
2659 	/* Some devices are known to return the total number of blocks,
2660 	 * not the highest block number.  Some devices have versions
2661 	 * which do this and others which do not.  Some devices we might
2662 	 * suspect of doing this but we don't know for certain.
2663 	 *
2664 	 * If we know the reported capacity is wrong, decrement it.  If
2665 	 * we can only guess, then assume the number of blocks is even
2666 	 * (usually true but not always) and err on the side of lowering
2667 	 * the capacity.
2668 	 */
2669 	if (sdp->fix_capacity ||
2670 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2671 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2672 				"from its reported value: %llu\n",
2673 				(unsigned long long) sdkp->capacity);
2674 		--sdkp->capacity;
2675 	}
2676 
2677 got_data:
2678 	if (sector_size == 0) {
2679 		sector_size = 512;
2680 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2681 			  "assuming 512.\n");
2682 	}
2683 
2684 	if (sector_size != 512 &&
2685 	    sector_size != 1024 &&
2686 	    sector_size != 2048 &&
2687 	    sector_size != 4096) {
2688 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2689 			  sector_size);
2690 		/*
2691 		 * The user might want to re-format the drive with
2692 		 * a supported sectorsize.  Once this happens, it
2693 		 * would be relatively trivial to set the thing up.
2694 		 * For this reason, we leave the thing in the table.
2695 		 */
2696 		sdkp->capacity = 0;
2697 		/*
2698 		 * set a bogus sector size so the normal read/write
2699 		 * logic in the block layer will eventually refuse any
2700 		 * request on this device without tripping over power
2701 		 * of two sector size assumptions
2702 		 */
2703 		sector_size = 512;
2704 	}
2705 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2706 	blk_queue_physical_block_size(sdp->request_queue,
2707 				      sdkp->physical_block_size);
2708 	sdkp->device->sector_size = sector_size;
2709 
2710 	if (sdkp->capacity > 0xffffffff)
2711 		sdp->use_16_for_rw = 1;
2712 
2713 }
2714 
2715 /*
2716  * Print disk capacity
2717  */
2718 static void
2719 sd_print_capacity(struct scsi_disk *sdkp,
2720 		  sector_t old_capacity)
2721 {
2722 	int sector_size = sdkp->device->sector_size;
2723 	char cap_str_2[10], cap_str_10[10];
2724 
2725 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2726 		return;
2727 
2728 	string_get_size(sdkp->capacity, sector_size,
2729 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2730 	string_get_size(sdkp->capacity, sector_size,
2731 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2732 
2733 	sd_printk(KERN_NOTICE, sdkp,
2734 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2735 		  (unsigned long long)sdkp->capacity,
2736 		  sector_size, cap_str_10, cap_str_2);
2737 
2738 	if (sdkp->physical_block_size != sector_size)
2739 		sd_printk(KERN_NOTICE, sdkp,
2740 			  "%u-byte physical blocks\n",
2741 			  sdkp->physical_block_size);
2742 }
2743 
2744 /* called with buffer of length 512 */
2745 static inline int
2746 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2747 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2748 		 struct scsi_sense_hdr *sshdr)
2749 {
2750 	/*
2751 	 * If we must use MODE SENSE(10), make sure that the buffer length
2752 	 * is at least 8 bytes so that the mode sense header fits.
2753 	 */
2754 	if (sdkp->device->use_10_for_ms && len < 8)
2755 		len = 8;
2756 
2757 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2758 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2759 }
2760 
2761 /*
2762  * read write protect setting, if possible - called only in sd_revalidate_disk()
2763  * called with buffer of length SD_BUF_SIZE
2764  */
2765 static void
2766 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2767 {
2768 	int res;
2769 	struct scsi_device *sdp = sdkp->device;
2770 	struct scsi_mode_data data;
2771 	int old_wp = sdkp->write_prot;
2772 
2773 	set_disk_ro(sdkp->disk, 0);
2774 	if (sdp->skip_ms_page_3f) {
2775 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2776 		return;
2777 	}
2778 
2779 	if (sdp->use_192_bytes_for_3f) {
2780 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2781 	} else {
2782 		/*
2783 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2784 		 * We have to start carefully: some devices hang if we ask
2785 		 * for more than is available.
2786 		 */
2787 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2788 
2789 		/*
2790 		 * Second attempt: ask for page 0 When only page 0 is
2791 		 * implemented, a request for page 3F may return Sense Key
2792 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2793 		 * CDB.
2794 		 */
2795 		if (res < 0)
2796 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2797 
2798 		/*
2799 		 * Third attempt: ask 255 bytes, as we did earlier.
2800 		 */
2801 		if (res < 0)
2802 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2803 					       &data, NULL);
2804 	}
2805 
2806 	if (res < 0) {
2807 		sd_first_printk(KERN_WARNING, sdkp,
2808 			  "Test WP failed, assume Write Enabled\n");
2809 	} else {
2810 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2811 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2812 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2813 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2814 				  sdkp->write_prot ? "on" : "off");
2815 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2816 		}
2817 	}
2818 }
2819 
2820 /*
2821  * sd_read_cache_type - called only from sd_revalidate_disk()
2822  * called with buffer of length SD_BUF_SIZE
2823  */
2824 static void
2825 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2826 {
2827 	int len = 0, res;
2828 	struct scsi_device *sdp = sdkp->device;
2829 
2830 	int dbd;
2831 	int modepage;
2832 	int first_len;
2833 	struct scsi_mode_data data;
2834 	struct scsi_sense_hdr sshdr;
2835 	int old_wce = sdkp->WCE;
2836 	int old_rcd = sdkp->RCD;
2837 	int old_dpofua = sdkp->DPOFUA;
2838 
2839 
2840 	if (sdkp->cache_override)
2841 		return;
2842 
2843 	first_len = 4;
2844 	if (sdp->skip_ms_page_8) {
2845 		if (sdp->type == TYPE_RBC)
2846 			goto defaults;
2847 		else {
2848 			if (sdp->skip_ms_page_3f)
2849 				goto defaults;
2850 			modepage = 0x3F;
2851 			if (sdp->use_192_bytes_for_3f)
2852 				first_len = 192;
2853 			dbd = 0;
2854 		}
2855 	} else if (sdp->type == TYPE_RBC) {
2856 		modepage = 6;
2857 		dbd = 8;
2858 	} else {
2859 		modepage = 8;
2860 		dbd = 0;
2861 	}
2862 
2863 	/* cautiously ask */
2864 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2865 			&data, &sshdr);
2866 
2867 	if (res < 0)
2868 		goto bad_sense;
2869 
2870 	if (!data.header_length) {
2871 		modepage = 6;
2872 		first_len = 0;
2873 		sd_first_printk(KERN_ERR, sdkp,
2874 				"Missing header in MODE_SENSE response\n");
2875 	}
2876 
2877 	/* that went OK, now ask for the proper length */
2878 	len = data.length;
2879 
2880 	/*
2881 	 * We're only interested in the first three bytes, actually.
2882 	 * But the data cache page is defined for the first 20.
2883 	 */
2884 	if (len < 3)
2885 		goto bad_sense;
2886 	else if (len > SD_BUF_SIZE) {
2887 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2888 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2889 		len = SD_BUF_SIZE;
2890 	}
2891 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2892 		len = 192;
2893 
2894 	/* Get the data */
2895 	if (len > first_len)
2896 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2897 				&data, &sshdr);
2898 
2899 	if (!res) {
2900 		int offset = data.header_length + data.block_descriptor_length;
2901 
2902 		while (offset < len) {
2903 			u8 page_code = buffer[offset] & 0x3F;
2904 			u8 spf       = buffer[offset] & 0x40;
2905 
2906 			if (page_code == 8 || page_code == 6) {
2907 				/* We're interested only in the first 3 bytes.
2908 				 */
2909 				if (len - offset <= 2) {
2910 					sd_first_printk(KERN_ERR, sdkp,
2911 						"Incomplete mode parameter "
2912 							"data\n");
2913 					goto defaults;
2914 				} else {
2915 					modepage = page_code;
2916 					goto Page_found;
2917 				}
2918 			} else {
2919 				/* Go to the next page */
2920 				if (spf && len - offset > 3)
2921 					offset += 4 + (buffer[offset+2] << 8) +
2922 						buffer[offset+3];
2923 				else if (!spf && len - offset > 1)
2924 					offset += 2 + buffer[offset+1];
2925 				else {
2926 					sd_first_printk(KERN_ERR, sdkp,
2927 							"Incomplete mode "
2928 							"parameter data\n");
2929 					goto defaults;
2930 				}
2931 			}
2932 		}
2933 
2934 		sd_first_printk(KERN_WARNING, sdkp,
2935 				"No Caching mode page found\n");
2936 		goto defaults;
2937 
2938 	Page_found:
2939 		if (modepage == 8) {
2940 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2941 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2942 		} else {
2943 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2944 			sdkp->RCD = 0;
2945 		}
2946 
2947 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2948 		if (sdp->broken_fua) {
2949 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2950 			sdkp->DPOFUA = 0;
2951 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2952 			   !sdkp->device->use_16_for_rw) {
2953 			sd_first_printk(KERN_NOTICE, sdkp,
2954 				  "Uses READ/WRITE(6), disabling FUA\n");
2955 			sdkp->DPOFUA = 0;
2956 		}
2957 
2958 		/* No cache flush allowed for write protected devices */
2959 		if (sdkp->WCE && sdkp->write_prot)
2960 			sdkp->WCE = 0;
2961 
2962 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2963 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2964 			sd_printk(KERN_NOTICE, sdkp,
2965 				  "Write cache: %s, read cache: %s, %s\n",
2966 				  sdkp->WCE ? "enabled" : "disabled",
2967 				  sdkp->RCD ? "disabled" : "enabled",
2968 				  sdkp->DPOFUA ? "supports DPO and FUA"
2969 				  : "doesn't support DPO or FUA");
2970 
2971 		return;
2972 	}
2973 
2974 bad_sense:
2975 	if (scsi_sense_valid(&sshdr) &&
2976 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2977 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2978 		/* Invalid field in CDB */
2979 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2980 	else
2981 		sd_first_printk(KERN_ERR, sdkp,
2982 				"Asking for cache data failed\n");
2983 
2984 defaults:
2985 	if (sdp->wce_default_on) {
2986 		sd_first_printk(KERN_NOTICE, sdkp,
2987 				"Assuming drive cache: write back\n");
2988 		sdkp->WCE = 1;
2989 	} else {
2990 		sd_first_printk(KERN_WARNING, sdkp,
2991 				"Assuming drive cache: write through\n");
2992 		sdkp->WCE = 0;
2993 	}
2994 	sdkp->RCD = 0;
2995 	sdkp->DPOFUA = 0;
2996 }
2997 
2998 /*
2999  * The ATO bit indicates whether the DIF application tag is available
3000  * for use by the operating system.
3001  */
3002 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
3003 {
3004 	int res, offset;
3005 	struct scsi_device *sdp = sdkp->device;
3006 	struct scsi_mode_data data;
3007 	struct scsi_sense_hdr sshdr;
3008 
3009 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
3010 		return;
3011 
3012 	if (sdkp->protection_type == 0)
3013 		return;
3014 
3015 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
3016 			      sdkp->max_retries, &data, &sshdr);
3017 
3018 	if (res < 0 || !data.header_length ||
3019 	    data.length < 6) {
3020 		sd_first_printk(KERN_WARNING, sdkp,
3021 			  "getting Control mode page failed, assume no ATO\n");
3022 
3023 		if (scsi_sense_valid(&sshdr))
3024 			sd_print_sense_hdr(sdkp, &sshdr);
3025 
3026 		return;
3027 	}
3028 
3029 	offset = data.header_length + data.block_descriptor_length;
3030 
3031 	if ((buffer[offset] & 0x3f) != 0x0a) {
3032 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
3033 		return;
3034 	}
3035 
3036 	if ((buffer[offset + 5] & 0x80) == 0)
3037 		return;
3038 
3039 	sdkp->ATO = 1;
3040 
3041 	return;
3042 }
3043 
3044 /**
3045  * sd_read_block_limits - Query disk device for preferred I/O sizes.
3046  * @sdkp: disk to query
3047  */
3048 static void sd_read_block_limits(struct scsi_disk *sdkp)
3049 {
3050 	struct scsi_vpd *vpd;
3051 
3052 	rcu_read_lock();
3053 
3054 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3055 	if (!vpd || vpd->len < 16)
3056 		goto out;
3057 
3058 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3059 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3060 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3061 
3062 	if (vpd->len >= 64) {
3063 		unsigned int lba_count, desc_count;
3064 
3065 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3066 
3067 		if (!sdkp->lbpme)
3068 			goto out;
3069 
3070 		lba_count = get_unaligned_be32(&vpd->data[20]);
3071 		desc_count = get_unaligned_be32(&vpd->data[24]);
3072 
3073 		if (lba_count && desc_count)
3074 			sdkp->max_unmap_blocks = lba_count;
3075 
3076 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3077 
3078 		if (vpd->data[32] & 0x80)
3079 			sdkp->unmap_alignment =
3080 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3081 
3082 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
3083 
3084 			if (sdkp->max_unmap_blocks)
3085 				sd_config_discard(sdkp, SD_LBP_UNMAP);
3086 			else
3087 				sd_config_discard(sdkp, SD_LBP_WS16);
3088 
3089 		} else {	/* LBP VPD page tells us what to use */
3090 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
3091 				sd_config_discard(sdkp, SD_LBP_UNMAP);
3092 			else if (sdkp->lbpws)
3093 				sd_config_discard(sdkp, SD_LBP_WS16);
3094 			else if (sdkp->lbpws10)
3095 				sd_config_discard(sdkp, SD_LBP_WS10);
3096 			else
3097 				sd_config_discard(sdkp, SD_LBP_DISABLE);
3098 		}
3099 	}
3100 
3101  out:
3102 	rcu_read_unlock();
3103 }
3104 
3105 /**
3106  * sd_read_block_characteristics - Query block dev. characteristics
3107  * @sdkp: disk to query
3108  */
3109 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
3110 {
3111 	struct request_queue *q = sdkp->disk->queue;
3112 	struct scsi_vpd *vpd;
3113 	u16 rot;
3114 	u8 zoned;
3115 
3116 	rcu_read_lock();
3117 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3118 
3119 	if (!vpd || vpd->len < 8) {
3120 		rcu_read_unlock();
3121 	        return;
3122 	}
3123 
3124 	rot = get_unaligned_be16(&vpd->data[4]);
3125 	zoned = (vpd->data[8] >> 4) & 3;
3126 	rcu_read_unlock();
3127 
3128 	if (rot == 1) {
3129 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3130 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3131 	}
3132 
3133 	if (sdkp->device->type == TYPE_ZBC) {
3134 		/*
3135 		 * Host-managed: Per ZBC and ZAC specifications, writes in
3136 		 * sequential write required zones of host-managed devices must
3137 		 * be aligned to the device physical block size.
3138 		 */
3139 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
3140 		blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3141 	} else {
3142 		sdkp->zoned = zoned;
3143 		if (sdkp->zoned == 1) {
3144 			/* Host-aware */
3145 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
3146 		} else {
3147 			/* Regular disk or drive managed disk */
3148 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3149 		}
3150 	}
3151 
3152 	if (!sdkp->first_scan)
3153 		return;
3154 
3155 	if (blk_queue_is_zoned(q)) {
3156 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3157 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3158 	} else {
3159 		if (sdkp->zoned == 1)
3160 			sd_printk(KERN_NOTICE, sdkp,
3161 				  "Host-aware SMR disk used as regular disk\n");
3162 		else if (sdkp->zoned == 2)
3163 			sd_printk(KERN_NOTICE, sdkp,
3164 				  "Drive-managed SMR disk\n");
3165 	}
3166 }
3167 
3168 /**
3169  * sd_read_block_provisioning - Query provisioning VPD page
3170  * @sdkp: disk to query
3171  */
3172 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3173 {
3174 	struct scsi_vpd *vpd;
3175 
3176 	if (sdkp->lbpme == 0)
3177 		return;
3178 
3179 	rcu_read_lock();
3180 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3181 
3182 	if (!vpd || vpd->len < 8) {
3183 		rcu_read_unlock();
3184 		return;
3185 	}
3186 
3187 	sdkp->lbpvpd	= 1;
3188 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3189 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3190 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3191 	rcu_read_unlock();
3192 }
3193 
3194 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3195 {
3196 	struct scsi_device *sdev = sdkp->device;
3197 
3198 	if (sdev->host->no_write_same) {
3199 		sdev->no_write_same = 1;
3200 
3201 		return;
3202 	}
3203 
3204 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3205 		struct scsi_vpd *vpd;
3206 
3207 		sdev->no_report_opcodes = 1;
3208 
3209 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3210 		 * CODES is unsupported and the device has an ATA
3211 		 * Information VPD page (SAT).
3212 		 */
3213 		rcu_read_lock();
3214 		vpd = rcu_dereference(sdev->vpd_pg89);
3215 		if (vpd)
3216 			sdev->no_write_same = 1;
3217 		rcu_read_unlock();
3218 	}
3219 
3220 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3221 		sdkp->ws16 = 1;
3222 
3223 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3224 		sdkp->ws10 = 1;
3225 }
3226 
3227 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3228 {
3229 	struct scsi_device *sdev = sdkp->device;
3230 
3231 	if (!sdev->security_supported)
3232 		return;
3233 
3234 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3235 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3236 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3237 			SECURITY_PROTOCOL_OUT, 0) == 1)
3238 		sdkp->security = 1;
3239 }
3240 
3241 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3242 {
3243 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3244 }
3245 
3246 /**
3247  * sd_read_cpr - Query concurrent positioning ranges
3248  * @sdkp:	disk to query
3249  */
3250 static void sd_read_cpr(struct scsi_disk *sdkp)
3251 {
3252 	struct blk_independent_access_ranges *iars = NULL;
3253 	unsigned char *buffer = NULL;
3254 	unsigned int nr_cpr = 0;
3255 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3256 	u8 *desc;
3257 
3258 	/*
3259 	 * We need to have the capacity set first for the block layer to be
3260 	 * able to check the ranges.
3261 	 */
3262 	if (sdkp->first_scan)
3263 		return;
3264 
3265 	if (!sdkp->capacity)
3266 		goto out;
3267 
3268 	/*
3269 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3270 	 * leading to a maximum page size of 64 + 256*32 bytes.
3271 	 */
3272 	buf_len = 64 + 256*32;
3273 	buffer = kmalloc(buf_len, GFP_KERNEL);
3274 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3275 		goto out;
3276 
3277 	/* We must have at least a 64B header and one 32B range descriptor */
3278 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3279 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3280 		sd_printk(KERN_ERR, sdkp,
3281 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3282 		goto out;
3283 	}
3284 
3285 	nr_cpr = (vpd_len - 64) / 32;
3286 	if (nr_cpr == 1) {
3287 		nr_cpr = 0;
3288 		goto out;
3289 	}
3290 
3291 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3292 	if (!iars) {
3293 		nr_cpr = 0;
3294 		goto out;
3295 	}
3296 
3297 	desc = &buffer[64];
3298 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3299 		if (desc[0] != i) {
3300 			sd_printk(KERN_ERR, sdkp,
3301 				"Invalid Concurrent Positioning Range number\n");
3302 			nr_cpr = 0;
3303 			break;
3304 		}
3305 
3306 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3307 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3308 	}
3309 
3310 out:
3311 	disk_set_independent_access_ranges(sdkp->disk, iars);
3312 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3313 		sd_printk(KERN_NOTICE, sdkp,
3314 			  "%u concurrent positioning ranges\n", nr_cpr);
3315 		sdkp->nr_actuators = nr_cpr;
3316 	}
3317 
3318 	kfree(buffer);
3319 }
3320 
3321 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3322 {
3323 	struct scsi_device *sdp = sdkp->device;
3324 	unsigned int min_xfer_bytes =
3325 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3326 
3327 	if (sdkp->min_xfer_blocks == 0)
3328 		return false;
3329 
3330 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3331 		sd_first_printk(KERN_WARNING, sdkp,
3332 				"Preferred minimum I/O size %u bytes not a " \
3333 				"multiple of physical block size (%u bytes)\n",
3334 				min_xfer_bytes, sdkp->physical_block_size);
3335 		sdkp->min_xfer_blocks = 0;
3336 		return false;
3337 	}
3338 
3339 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3340 			min_xfer_bytes);
3341 	return true;
3342 }
3343 
3344 /*
3345  * Determine the device's preferred I/O size for reads and writes
3346  * unless the reported value is unreasonably small, large, not a
3347  * multiple of the physical block size, or simply garbage.
3348  */
3349 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3350 				      unsigned int dev_max)
3351 {
3352 	struct scsi_device *sdp = sdkp->device;
3353 	unsigned int opt_xfer_bytes =
3354 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3355 	unsigned int min_xfer_bytes =
3356 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3357 
3358 	if (sdkp->opt_xfer_blocks == 0)
3359 		return false;
3360 
3361 	if (sdkp->opt_xfer_blocks > dev_max) {
3362 		sd_first_printk(KERN_WARNING, sdkp,
3363 				"Optimal transfer size %u logical blocks " \
3364 				"> dev_max (%u logical blocks)\n",
3365 				sdkp->opt_xfer_blocks, dev_max);
3366 		return false;
3367 	}
3368 
3369 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3370 		sd_first_printk(KERN_WARNING, sdkp,
3371 				"Optimal transfer size %u logical blocks " \
3372 				"> sd driver limit (%u logical blocks)\n",
3373 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3374 		return false;
3375 	}
3376 
3377 	if (opt_xfer_bytes < PAGE_SIZE) {
3378 		sd_first_printk(KERN_WARNING, sdkp,
3379 				"Optimal transfer size %u bytes < " \
3380 				"PAGE_SIZE (%u bytes)\n",
3381 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3382 		return false;
3383 	}
3384 
3385 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3386 		sd_first_printk(KERN_WARNING, sdkp,
3387 				"Optimal transfer size %u bytes not a " \
3388 				"multiple of preferred minimum block " \
3389 				"size (%u bytes)\n",
3390 				opt_xfer_bytes, min_xfer_bytes);
3391 		return false;
3392 	}
3393 
3394 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3395 		sd_first_printk(KERN_WARNING, sdkp,
3396 				"Optimal transfer size %u bytes not a " \
3397 				"multiple of physical block size (%u bytes)\n",
3398 				opt_xfer_bytes, sdkp->physical_block_size);
3399 		return false;
3400 	}
3401 
3402 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3403 			opt_xfer_bytes);
3404 	return true;
3405 }
3406 
3407 static void sd_read_block_zero(struct scsi_disk *sdkp)
3408 {
3409 	unsigned int buf_len = sdkp->device->sector_size;
3410 	char *buffer, cmd[10] = { };
3411 
3412 	buffer = kmalloc(buf_len, GFP_KERNEL);
3413 	if (!buffer)
3414 		return;
3415 
3416 	cmd[0] = READ_10;
3417 	put_unaligned_be32(0, &cmd[2]); /* Logical block address 0 */
3418 	put_unaligned_be16(1, &cmd[7]);	/* Transfer 1 logical block */
3419 
3420 	scsi_execute_cmd(sdkp->device, cmd, REQ_OP_DRV_IN, buffer, buf_len,
3421 			 SD_TIMEOUT, sdkp->max_retries, NULL);
3422 	kfree(buffer);
3423 }
3424 
3425 /**
3426  *	sd_revalidate_disk - called the first time a new disk is seen,
3427  *	performs disk spin up, read_capacity, etc.
3428  *	@disk: struct gendisk we care about
3429  **/
3430 static int sd_revalidate_disk(struct gendisk *disk)
3431 {
3432 	struct scsi_disk *sdkp = scsi_disk(disk);
3433 	struct scsi_device *sdp = sdkp->device;
3434 	struct request_queue *q = sdkp->disk->queue;
3435 	sector_t old_capacity = sdkp->capacity;
3436 	unsigned char *buffer;
3437 	unsigned int dev_max, rw_max;
3438 
3439 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3440 				      "sd_revalidate_disk\n"));
3441 
3442 	/*
3443 	 * If the device is offline, don't try and read capacity or any
3444 	 * of the other niceties.
3445 	 */
3446 	if (!scsi_device_online(sdp))
3447 		goto out;
3448 
3449 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3450 	if (!buffer) {
3451 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3452 			  "allocation failure.\n");
3453 		goto out;
3454 	}
3455 
3456 	sd_spinup_disk(sdkp);
3457 
3458 	/*
3459 	 * Without media there is no reason to ask; moreover, some devices
3460 	 * react badly if we do.
3461 	 */
3462 	if (sdkp->media_present) {
3463 		sd_read_capacity(sdkp, buffer);
3464 		/*
3465 		 * Some USB/UAS devices return generic values for mode pages
3466 		 * until the media has been accessed. Trigger a READ operation
3467 		 * to force the device to populate mode pages.
3468 		 */
3469 		if (sdp->read_before_ms)
3470 			sd_read_block_zero(sdkp);
3471 		/*
3472 		 * set the default to rotational.  All non-rotational devices
3473 		 * support the block characteristics VPD page, which will
3474 		 * cause this to be updated correctly and any device which
3475 		 * doesn't support it should be treated as rotational.
3476 		 */
3477 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3478 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3479 
3480 		if (scsi_device_supports_vpd(sdp)) {
3481 			sd_read_block_provisioning(sdkp);
3482 			sd_read_block_limits(sdkp);
3483 			sd_read_block_characteristics(sdkp);
3484 			sd_zbc_read_zones(sdkp, buffer);
3485 			sd_read_cpr(sdkp);
3486 		}
3487 
3488 		sd_print_capacity(sdkp, old_capacity);
3489 
3490 		sd_read_write_protect_flag(sdkp, buffer);
3491 		sd_read_cache_type(sdkp, buffer);
3492 		sd_read_app_tag_own(sdkp, buffer);
3493 		sd_read_write_same(sdkp, buffer);
3494 		sd_read_security(sdkp, buffer);
3495 		sd_config_protection(sdkp);
3496 	}
3497 
3498 	/*
3499 	 * We now have all cache related info, determine how we deal
3500 	 * with flush requests.
3501 	 */
3502 	sd_set_flush_flag(sdkp);
3503 
3504 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3505 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3506 
3507 	/* Some devices report a maximum block count for READ/WRITE requests. */
3508 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3509 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3510 
3511 	if (sd_validate_min_xfer_size(sdkp))
3512 		blk_queue_io_min(sdkp->disk->queue,
3513 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3514 	else
3515 		blk_queue_io_min(sdkp->disk->queue, 0);
3516 
3517 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3518 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3519 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3520 	} else {
3521 		q->limits.io_opt = 0;
3522 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3523 				      (sector_t)BLK_DEF_MAX_SECTORS);
3524 	}
3525 
3526 	/*
3527 	 * Limit default to SCSI host optimal sector limit if set. There may be
3528 	 * an impact on performance for when the size of a request exceeds this
3529 	 * host limit.
3530 	 */
3531 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3532 
3533 	/* Do not exceed controller limit */
3534 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3535 
3536 	/*
3537 	 * Only update max_sectors if previously unset or if the current value
3538 	 * exceeds the capabilities of the hardware.
3539 	 */
3540 	if (sdkp->first_scan ||
3541 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3542 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3543 		q->limits.max_sectors = rw_max;
3544 
3545 	sdkp->first_scan = 0;
3546 
3547 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3548 	sd_config_write_same(sdkp);
3549 	kfree(buffer);
3550 
3551 	/*
3552 	 * For a zoned drive, revalidating the zones can be done only once
3553 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3554 	 * capacity to 0.
3555 	 */
3556 	if (sd_zbc_revalidate_zones(sdkp))
3557 		set_capacity_and_notify(disk, 0);
3558 
3559  out:
3560 	return 0;
3561 }
3562 
3563 /**
3564  *	sd_unlock_native_capacity - unlock native capacity
3565  *	@disk: struct gendisk to set capacity for
3566  *
3567  *	Block layer calls this function if it detects that partitions
3568  *	on @disk reach beyond the end of the device.  If the SCSI host
3569  *	implements ->unlock_native_capacity() method, it's invoked to
3570  *	give it a chance to adjust the device capacity.
3571  *
3572  *	CONTEXT:
3573  *	Defined by block layer.  Might sleep.
3574  */
3575 static void sd_unlock_native_capacity(struct gendisk *disk)
3576 {
3577 	struct scsi_device *sdev = scsi_disk(disk)->device;
3578 
3579 	if (sdev->host->hostt->unlock_native_capacity)
3580 		sdev->host->hostt->unlock_native_capacity(sdev);
3581 }
3582 
3583 /**
3584  *	sd_format_disk_name - format disk name
3585  *	@prefix: name prefix - ie. "sd" for SCSI disks
3586  *	@index: index of the disk to format name for
3587  *	@buf: output buffer
3588  *	@buflen: length of the output buffer
3589  *
3590  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3591  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3592  *	which is followed by sdaaa.
3593  *
3594  *	This is basically 26 base counting with one extra 'nil' entry
3595  *	at the beginning from the second digit on and can be
3596  *	determined using similar method as 26 base conversion with the
3597  *	index shifted -1 after each digit is computed.
3598  *
3599  *	CONTEXT:
3600  *	Don't care.
3601  *
3602  *	RETURNS:
3603  *	0 on success, -errno on failure.
3604  */
3605 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3606 {
3607 	const int base = 'z' - 'a' + 1;
3608 	char *begin = buf + strlen(prefix);
3609 	char *end = buf + buflen;
3610 	char *p;
3611 	int unit;
3612 
3613 	p = end - 1;
3614 	*p = '\0';
3615 	unit = base;
3616 	do {
3617 		if (p == begin)
3618 			return -EINVAL;
3619 		*--p = 'a' + (index % unit);
3620 		index = (index / unit) - 1;
3621 	} while (index >= 0);
3622 
3623 	memmove(begin, p, end - p);
3624 	memcpy(buf, prefix, strlen(prefix));
3625 
3626 	return 0;
3627 }
3628 
3629 /**
3630  *	sd_probe - called during driver initialization and whenever a
3631  *	new scsi device is attached to the system. It is called once
3632  *	for each scsi device (not just disks) present.
3633  *	@dev: pointer to device object
3634  *
3635  *	Returns 0 if successful (or not interested in this scsi device
3636  *	(e.g. scanner)); 1 when there is an error.
3637  *
3638  *	Note: this function is invoked from the scsi mid-level.
3639  *	This function sets up the mapping between a given
3640  *	<host,channel,id,lun> (found in sdp) and new device name
3641  *	(e.g. /dev/sda). More precisely it is the block device major
3642  *	and minor number that is chosen here.
3643  *
3644  *	Assume sd_probe is not re-entrant (for time being)
3645  *	Also think about sd_probe() and sd_remove() running coincidentally.
3646  **/
3647 static int sd_probe(struct device *dev)
3648 {
3649 	struct scsi_device *sdp = to_scsi_device(dev);
3650 	struct scsi_disk *sdkp;
3651 	struct gendisk *gd;
3652 	int index;
3653 	int error;
3654 
3655 	scsi_autopm_get_device(sdp);
3656 	error = -ENODEV;
3657 	if (sdp->type != TYPE_DISK &&
3658 	    sdp->type != TYPE_ZBC &&
3659 	    sdp->type != TYPE_MOD &&
3660 	    sdp->type != TYPE_RBC)
3661 		goto out;
3662 
3663 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3664 		sdev_printk(KERN_WARNING, sdp,
3665 			    "Unsupported ZBC host-managed device.\n");
3666 		goto out;
3667 	}
3668 
3669 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3670 					"sd_probe\n"));
3671 
3672 	error = -ENOMEM;
3673 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3674 	if (!sdkp)
3675 		goto out;
3676 
3677 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3678 					 &sd_bio_compl_lkclass);
3679 	if (!gd)
3680 		goto out_free;
3681 
3682 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3683 	if (index < 0) {
3684 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3685 		goto out_put;
3686 	}
3687 
3688 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3689 	if (error) {
3690 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3691 		goto out_free_index;
3692 	}
3693 
3694 	sdkp->device = sdp;
3695 	sdkp->disk = gd;
3696 	sdkp->index = index;
3697 	sdkp->max_retries = SD_MAX_RETRIES;
3698 	atomic_set(&sdkp->openers, 0);
3699 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3700 
3701 	if (!sdp->request_queue->rq_timeout) {
3702 		if (sdp->type != TYPE_MOD)
3703 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3704 		else
3705 			blk_queue_rq_timeout(sdp->request_queue,
3706 					     SD_MOD_TIMEOUT);
3707 	}
3708 
3709 	device_initialize(&sdkp->disk_dev);
3710 	sdkp->disk_dev.parent = get_device(dev);
3711 	sdkp->disk_dev.class = &sd_disk_class;
3712 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3713 
3714 	error = device_add(&sdkp->disk_dev);
3715 	if (error) {
3716 		put_device(&sdkp->disk_dev);
3717 		goto out;
3718 	}
3719 
3720 	dev_set_drvdata(dev, sdkp);
3721 
3722 	gd->major = sd_major((index & 0xf0) >> 4);
3723 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3724 	gd->minors = SD_MINORS;
3725 
3726 	gd->fops = &sd_fops;
3727 	gd->private_data = sdkp;
3728 
3729 	/* defaults, until the device tells us otherwise */
3730 	sdp->sector_size = 512;
3731 	sdkp->capacity = 0;
3732 	sdkp->media_present = 1;
3733 	sdkp->write_prot = 0;
3734 	sdkp->cache_override = 0;
3735 	sdkp->WCE = 0;
3736 	sdkp->RCD = 0;
3737 	sdkp->ATO = 0;
3738 	sdkp->first_scan = 1;
3739 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3740 
3741 	sd_revalidate_disk(gd);
3742 
3743 	if (sdp->removable) {
3744 		gd->flags |= GENHD_FL_REMOVABLE;
3745 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3746 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3747 	}
3748 
3749 	blk_pm_runtime_init(sdp->request_queue, dev);
3750 	if (sdp->rpm_autosuspend) {
3751 		pm_runtime_set_autosuspend_delay(dev,
3752 			sdp->host->hostt->rpm_autosuspend_delay);
3753 	}
3754 
3755 	error = device_add_disk(dev, gd, NULL);
3756 	if (error) {
3757 		device_unregister(&sdkp->disk_dev);
3758 		put_disk(gd);
3759 		goto out;
3760 	}
3761 
3762 	if (sdkp->security) {
3763 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3764 		if (sdkp->opal_dev)
3765 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3766 	}
3767 
3768 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3769 		  sdp->removable ? "removable " : "");
3770 	scsi_autopm_put_device(sdp);
3771 
3772 	return 0;
3773 
3774  out_free_index:
3775 	ida_free(&sd_index_ida, index);
3776  out_put:
3777 	put_disk(gd);
3778  out_free:
3779 	kfree(sdkp);
3780  out:
3781 	scsi_autopm_put_device(sdp);
3782 	return error;
3783 }
3784 
3785 /**
3786  *	sd_remove - called whenever a scsi disk (previously recognized by
3787  *	sd_probe) is detached from the system. It is called (potentially
3788  *	multiple times) during sd module unload.
3789  *	@dev: pointer to device object
3790  *
3791  *	Note: this function is invoked from the scsi mid-level.
3792  *	This function potentially frees up a device name (e.g. /dev/sdc)
3793  *	that could be re-used by a subsequent sd_probe().
3794  *	This function is not called when the built-in sd driver is "exit-ed".
3795  **/
3796 static int sd_remove(struct device *dev)
3797 {
3798 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3799 
3800 	scsi_autopm_get_device(sdkp->device);
3801 
3802 	device_del(&sdkp->disk_dev);
3803 	del_gendisk(sdkp->disk);
3804 	if (!sdkp->suspended)
3805 		sd_shutdown(dev);
3806 
3807 	put_disk(sdkp->disk);
3808 	return 0;
3809 }
3810 
3811 static void scsi_disk_release(struct device *dev)
3812 {
3813 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3814 
3815 	ida_free(&sd_index_ida, sdkp->index);
3816 	sd_zbc_free_zone_info(sdkp);
3817 	put_device(&sdkp->device->sdev_gendev);
3818 	free_opal_dev(sdkp->opal_dev);
3819 
3820 	kfree(sdkp);
3821 }
3822 
3823 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3824 {
3825 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3826 	struct scsi_sense_hdr sshdr;
3827 	const struct scsi_exec_args exec_args = {
3828 		.sshdr = &sshdr,
3829 		.req_flags = BLK_MQ_REQ_PM,
3830 	};
3831 	struct scsi_device *sdp = sdkp->device;
3832 	int res;
3833 
3834 	if (start)
3835 		cmd[4] |= 1;	/* START */
3836 
3837 	if (sdp->start_stop_pwr_cond)
3838 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3839 
3840 	if (!scsi_device_online(sdp))
3841 		return -ENODEV;
3842 
3843 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3844 			       sdkp->max_retries, &exec_args);
3845 	if (res) {
3846 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3847 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3848 			sd_print_sense_hdr(sdkp, &sshdr);
3849 			/* 0x3a is medium not present */
3850 			if (sshdr.asc == 0x3a)
3851 				res = 0;
3852 		}
3853 	}
3854 
3855 	/* SCSI error codes must not go to the generic layer */
3856 	if (res)
3857 		return -EIO;
3858 
3859 	return 0;
3860 }
3861 
3862 /*
3863  * Send a SYNCHRONIZE CACHE instruction down to the device through
3864  * the normal SCSI command structure.  Wait for the command to
3865  * complete.
3866  */
3867 static void sd_shutdown(struct device *dev)
3868 {
3869 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3870 
3871 	if (!sdkp)
3872 		return;         /* this can happen */
3873 
3874 	if (pm_runtime_suspended(dev))
3875 		return;
3876 
3877 	if (sdkp->WCE && sdkp->media_present) {
3878 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3879 		sd_sync_cache(sdkp);
3880 	}
3881 
3882 	if ((system_state != SYSTEM_RESTART &&
3883 	     sdkp->device->manage_system_start_stop) ||
3884 	    (system_state == SYSTEM_POWER_OFF &&
3885 	     sdkp->device->manage_shutdown)) {
3886 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3887 		sd_start_stop_device(sdkp, 0);
3888 	}
3889 }
3890 
3891 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
3892 {
3893 	return (sdev->manage_system_start_stop && !runtime) ||
3894 		(sdev->manage_runtime_start_stop && runtime);
3895 }
3896 
3897 static int sd_suspend_common(struct device *dev, bool runtime)
3898 {
3899 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3900 	int ret = 0;
3901 
3902 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3903 		return 0;
3904 
3905 	if (sdkp->WCE && sdkp->media_present) {
3906 		if (!sdkp->device->silence_suspend)
3907 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3908 		ret = sd_sync_cache(sdkp);
3909 		/* ignore OFFLINE device */
3910 		if (ret == -ENODEV)
3911 			return 0;
3912 
3913 		if (ret)
3914 			return ret;
3915 	}
3916 
3917 	if (sd_do_start_stop(sdkp->device, runtime)) {
3918 		if (!sdkp->device->silence_suspend)
3919 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3920 		/* an error is not worth aborting a system sleep */
3921 		ret = sd_start_stop_device(sdkp, 0);
3922 		if (!runtime)
3923 			ret = 0;
3924 	}
3925 
3926 	if (!ret)
3927 		sdkp->suspended = true;
3928 
3929 	return ret;
3930 }
3931 
3932 static int sd_suspend_system(struct device *dev)
3933 {
3934 	if (pm_runtime_suspended(dev))
3935 		return 0;
3936 
3937 	return sd_suspend_common(dev, false);
3938 }
3939 
3940 static int sd_suspend_runtime(struct device *dev)
3941 {
3942 	return sd_suspend_common(dev, true);
3943 }
3944 
3945 static int sd_resume(struct device *dev)
3946 {
3947 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3948 
3949 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3950 
3951 	if (opal_unlock_from_suspend(sdkp->opal_dev)) {
3952 		sd_printk(KERN_NOTICE, sdkp, "OPAL unlock failed\n");
3953 		return -EIO;
3954 	}
3955 
3956 	return 0;
3957 }
3958 
3959 static int sd_resume_common(struct device *dev, bool runtime)
3960 {
3961 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3962 	int ret = 0;
3963 
3964 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3965 		return 0;
3966 
3967 	if (!sd_do_start_stop(sdkp->device, runtime)) {
3968 		sdkp->suspended = false;
3969 		return 0;
3970 	}
3971 
3972 	if (!sdkp->device->no_start_on_resume) {
3973 		sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3974 		ret = sd_start_stop_device(sdkp, 1);
3975 	}
3976 
3977 	if (!ret) {
3978 		sd_resume(dev);
3979 		sdkp->suspended = false;
3980 	}
3981 
3982 	return ret;
3983 }
3984 
3985 static int sd_resume_system(struct device *dev)
3986 {
3987 	if (pm_runtime_suspended(dev)) {
3988 		struct scsi_disk *sdkp = dev_get_drvdata(dev);
3989 		struct scsi_device *sdp = sdkp ? sdkp->device : NULL;
3990 
3991 		if (sdp && sdp->force_runtime_start_on_system_start)
3992 			pm_request_resume(dev);
3993 
3994 		return 0;
3995 	}
3996 
3997 	return sd_resume_common(dev, false);
3998 }
3999 
4000 static int sd_resume_runtime(struct device *dev)
4001 {
4002 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
4003 	struct scsi_device *sdp;
4004 
4005 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
4006 		return 0;
4007 
4008 	sdp = sdkp->device;
4009 
4010 	if (sdp->ignore_media_change) {
4011 		/* clear the device's sense data */
4012 		static const u8 cmd[10] = { REQUEST_SENSE };
4013 		const struct scsi_exec_args exec_args = {
4014 			.req_flags = BLK_MQ_REQ_PM,
4015 		};
4016 
4017 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
4018 				     sdp->request_queue->rq_timeout, 1,
4019 				     &exec_args))
4020 			sd_printk(KERN_NOTICE, sdkp,
4021 				  "Failed to clear sense data\n");
4022 	}
4023 
4024 	return sd_resume_common(dev, true);
4025 }
4026 
4027 static const struct dev_pm_ops sd_pm_ops = {
4028 	.suspend		= sd_suspend_system,
4029 	.resume			= sd_resume_system,
4030 	.poweroff		= sd_suspend_system,
4031 	.restore		= sd_resume_system,
4032 	.runtime_suspend	= sd_suspend_runtime,
4033 	.runtime_resume		= sd_resume_runtime,
4034 };
4035 
4036 static struct scsi_driver sd_template = {
4037 	.gendrv = {
4038 		.name		= "sd",
4039 		.owner		= THIS_MODULE,
4040 		.probe		= sd_probe,
4041 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
4042 		.remove		= sd_remove,
4043 		.shutdown	= sd_shutdown,
4044 		.pm		= &sd_pm_ops,
4045 	},
4046 	.rescan			= sd_rescan,
4047 	.resume			= sd_resume,
4048 	.init_command		= sd_init_command,
4049 	.uninit_command		= sd_uninit_command,
4050 	.done			= sd_done,
4051 	.eh_action		= sd_eh_action,
4052 	.eh_reset		= sd_eh_reset,
4053 };
4054 
4055 /**
4056  *	init_sd - entry point for this driver (both when built in or when
4057  *	a module).
4058  *
4059  *	Note: this function registers this driver with the scsi mid-level.
4060  **/
4061 static int __init init_sd(void)
4062 {
4063 	int majors = 0, i, err;
4064 
4065 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
4066 
4067 	for (i = 0; i < SD_MAJORS; i++) {
4068 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
4069 			continue;
4070 		majors++;
4071 	}
4072 
4073 	if (!majors)
4074 		return -ENODEV;
4075 
4076 	err = class_register(&sd_disk_class);
4077 	if (err)
4078 		goto err_out;
4079 
4080 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4081 	if (!sd_page_pool) {
4082 		printk(KERN_ERR "sd: can't init discard page pool\n");
4083 		err = -ENOMEM;
4084 		goto err_out_class;
4085 	}
4086 
4087 	err = scsi_register_driver(&sd_template.gendrv);
4088 	if (err)
4089 		goto err_out_driver;
4090 
4091 	return 0;
4092 
4093 err_out_driver:
4094 	mempool_destroy(sd_page_pool);
4095 err_out_class:
4096 	class_unregister(&sd_disk_class);
4097 err_out:
4098 	for (i = 0; i < SD_MAJORS; i++)
4099 		unregister_blkdev(sd_major(i), "sd");
4100 	return err;
4101 }
4102 
4103 /**
4104  *	exit_sd - exit point for this driver (when it is a module).
4105  *
4106  *	Note: this function unregisters this driver from the scsi mid-level.
4107  **/
4108 static void __exit exit_sd(void)
4109 {
4110 	int i;
4111 
4112 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4113 
4114 	scsi_unregister_driver(&sd_template.gendrv);
4115 	mempool_destroy(sd_page_pool);
4116 
4117 	class_unregister(&sd_disk_class);
4118 
4119 	for (i = 0; i < SD_MAJORS; i++)
4120 		unregister_blkdev(sd_major(i), "sd");
4121 }
4122 
4123 module_init(init_sd);
4124 module_exit(exit_sd);
4125 
4126 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4127 {
4128 	scsi_print_sense_hdr(sdkp->device,
4129 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4130 }
4131 
4132 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4133 {
4134 	const char *hb_string = scsi_hostbyte_string(result);
4135 
4136 	if (hb_string)
4137 		sd_printk(KERN_INFO, sdkp,
4138 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4139 			  hb_string ? hb_string : "invalid",
4140 			  "DRIVER_OK");
4141 	else
4142 		sd_printk(KERN_INFO, sdkp,
4143 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4144 			  msg, host_byte(result), "DRIVER_OK");
4145 }
4146