xref: /openbmc/linux/drivers/scsi/sd.c (revision 99398d2070ab03d13f90b758ad397e19a65fffb0)
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 }
214 static DEVICE_ATTR_RO(manage_start_stop);
215 
216 static ssize_t
217 manage_system_start_stop_show(struct device *dev,
218 			      struct device_attribute *attr, char *buf)
219 {
220 	struct scsi_disk *sdkp = to_scsi_disk(dev);
221 	struct scsi_device *sdp = sdkp->device;
222 
223 	return sysfs_emit(buf, "%u\n", sdp->manage_system_start_stop);
224 }
225 
226 static ssize_t
227 manage_system_start_stop_store(struct device *dev,
228 			       struct device_attribute *attr,
229 			       const char *buf, size_t count)
230 {
231 	struct scsi_disk *sdkp = to_scsi_disk(dev);
232 	struct scsi_device *sdp = sdkp->device;
233 	bool v;
234 
235 	if (!capable(CAP_SYS_ADMIN))
236 		return -EACCES;
237 
238 	if (kstrtobool(buf, &v))
239 		return -EINVAL;
240 
241 	sdp->manage_system_start_stop = v;
242 
243 	return count;
244 }
245 static DEVICE_ATTR_RW(manage_system_start_stop);
246 
247 static ssize_t
248 manage_runtime_start_stop_show(struct device *dev,
249 			       struct device_attribute *attr, char *buf)
250 {
251 	struct scsi_disk *sdkp = to_scsi_disk(dev);
252 	struct scsi_device *sdp = sdkp->device;
253 
254 	return sysfs_emit(buf, "%u\n", sdp->manage_runtime_start_stop);
255 }
256 
257 static ssize_t
258 manage_runtime_start_stop_store(struct device *dev,
259 				struct device_attribute *attr,
260 				const char *buf, size_t count)
261 {
262 	struct scsi_disk *sdkp = to_scsi_disk(dev);
263 	struct scsi_device *sdp = sdkp->device;
264 	bool v;
265 
266 	if (!capable(CAP_SYS_ADMIN))
267 		return -EACCES;
268 
269 	if (kstrtobool(buf, &v))
270 		return -EINVAL;
271 
272 	sdp->manage_runtime_start_stop = v;
273 
274 	return count;
275 }
276 static DEVICE_ATTR_RW(manage_runtime_start_stop);
277 
278 static ssize_t
279 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
280 {
281 	struct scsi_disk *sdkp = to_scsi_disk(dev);
282 
283 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
284 }
285 
286 static ssize_t
287 allow_restart_store(struct device *dev, struct device_attribute *attr,
288 		    const char *buf, size_t count)
289 {
290 	bool v;
291 	struct scsi_disk *sdkp = to_scsi_disk(dev);
292 	struct scsi_device *sdp = sdkp->device;
293 
294 	if (!capable(CAP_SYS_ADMIN))
295 		return -EACCES;
296 
297 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
298 		return -EINVAL;
299 
300 	if (kstrtobool(buf, &v))
301 		return -EINVAL;
302 
303 	sdp->allow_restart = v;
304 
305 	return count;
306 }
307 static DEVICE_ATTR_RW(allow_restart);
308 
309 static ssize_t
310 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
311 {
312 	struct scsi_disk *sdkp = to_scsi_disk(dev);
313 	int ct = sdkp->RCD + 2*sdkp->WCE;
314 
315 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
316 }
317 static DEVICE_ATTR_RW(cache_type);
318 
319 static ssize_t
320 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
321 {
322 	struct scsi_disk *sdkp = to_scsi_disk(dev);
323 
324 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
325 }
326 static DEVICE_ATTR_RO(FUA);
327 
328 static ssize_t
329 protection_type_show(struct device *dev, struct device_attribute *attr,
330 		     char *buf)
331 {
332 	struct scsi_disk *sdkp = to_scsi_disk(dev);
333 
334 	return sprintf(buf, "%u\n", sdkp->protection_type);
335 }
336 
337 static ssize_t
338 protection_type_store(struct device *dev, struct device_attribute *attr,
339 		      const char *buf, size_t count)
340 {
341 	struct scsi_disk *sdkp = to_scsi_disk(dev);
342 	unsigned int val;
343 	int err;
344 
345 	if (!capable(CAP_SYS_ADMIN))
346 		return -EACCES;
347 
348 	err = kstrtouint(buf, 10, &val);
349 
350 	if (err)
351 		return err;
352 
353 	if (val <= T10_PI_TYPE3_PROTECTION)
354 		sdkp->protection_type = val;
355 
356 	return count;
357 }
358 static DEVICE_ATTR_RW(protection_type);
359 
360 static ssize_t
361 protection_mode_show(struct device *dev, struct device_attribute *attr,
362 		     char *buf)
363 {
364 	struct scsi_disk *sdkp = to_scsi_disk(dev);
365 	struct scsi_device *sdp = sdkp->device;
366 	unsigned int dif, dix;
367 
368 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
369 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
370 
371 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
372 		dif = 0;
373 		dix = 1;
374 	}
375 
376 	if (!dif && !dix)
377 		return sprintf(buf, "none\n");
378 
379 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
380 }
381 static DEVICE_ATTR_RO(protection_mode);
382 
383 static ssize_t
384 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
385 {
386 	struct scsi_disk *sdkp = to_scsi_disk(dev);
387 
388 	return sprintf(buf, "%u\n", sdkp->ATO);
389 }
390 static DEVICE_ATTR_RO(app_tag_own);
391 
392 static ssize_t
393 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
394 		       char *buf)
395 {
396 	struct scsi_disk *sdkp = to_scsi_disk(dev);
397 
398 	return sprintf(buf, "%u\n", sdkp->lbpme);
399 }
400 static DEVICE_ATTR_RO(thin_provisioning);
401 
402 /* sysfs_match_string() requires dense arrays */
403 static const char *lbp_mode[] = {
404 	[SD_LBP_FULL]		= "full",
405 	[SD_LBP_UNMAP]		= "unmap",
406 	[SD_LBP_WS16]		= "writesame_16",
407 	[SD_LBP_WS10]		= "writesame_10",
408 	[SD_LBP_ZERO]		= "writesame_zero",
409 	[SD_LBP_DISABLE]	= "disabled",
410 };
411 
412 static ssize_t
413 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
414 		       char *buf)
415 {
416 	struct scsi_disk *sdkp = to_scsi_disk(dev);
417 
418 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
419 }
420 
421 static ssize_t
422 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
423 			const char *buf, size_t count)
424 {
425 	struct scsi_disk *sdkp = to_scsi_disk(dev);
426 	struct scsi_device *sdp = sdkp->device;
427 	int mode;
428 
429 	if (!capable(CAP_SYS_ADMIN))
430 		return -EACCES;
431 
432 	if (sd_is_zoned(sdkp)) {
433 		sd_config_discard(sdkp, SD_LBP_DISABLE);
434 		return count;
435 	}
436 
437 	if (sdp->type != TYPE_DISK)
438 		return -EINVAL;
439 
440 	mode = sysfs_match_string(lbp_mode, buf);
441 	if (mode < 0)
442 		return -EINVAL;
443 
444 	sd_config_discard(sdkp, mode);
445 
446 	return count;
447 }
448 static DEVICE_ATTR_RW(provisioning_mode);
449 
450 /* sysfs_match_string() requires dense arrays */
451 static const char *zeroing_mode[] = {
452 	[SD_ZERO_WRITE]		= "write",
453 	[SD_ZERO_WS]		= "writesame",
454 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
455 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
456 };
457 
458 static ssize_t
459 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
460 		  char *buf)
461 {
462 	struct scsi_disk *sdkp = to_scsi_disk(dev);
463 
464 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
465 }
466 
467 static ssize_t
468 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
469 		   const char *buf, size_t count)
470 {
471 	struct scsi_disk *sdkp = to_scsi_disk(dev);
472 	int mode;
473 
474 	if (!capable(CAP_SYS_ADMIN))
475 		return -EACCES;
476 
477 	mode = sysfs_match_string(zeroing_mode, buf);
478 	if (mode < 0)
479 		return -EINVAL;
480 
481 	sdkp->zeroing_mode = mode;
482 
483 	return count;
484 }
485 static DEVICE_ATTR_RW(zeroing_mode);
486 
487 static ssize_t
488 max_medium_access_timeouts_show(struct device *dev,
489 				struct device_attribute *attr, char *buf)
490 {
491 	struct scsi_disk *sdkp = to_scsi_disk(dev);
492 
493 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
494 }
495 
496 static ssize_t
497 max_medium_access_timeouts_store(struct device *dev,
498 				 struct device_attribute *attr, const char *buf,
499 				 size_t count)
500 {
501 	struct scsi_disk *sdkp = to_scsi_disk(dev);
502 	int err;
503 
504 	if (!capable(CAP_SYS_ADMIN))
505 		return -EACCES;
506 
507 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
508 
509 	return err ? err : count;
510 }
511 static DEVICE_ATTR_RW(max_medium_access_timeouts);
512 
513 static ssize_t
514 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
515 			   char *buf)
516 {
517 	struct scsi_disk *sdkp = to_scsi_disk(dev);
518 
519 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
520 }
521 
522 static ssize_t
523 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
524 			    const char *buf, size_t count)
525 {
526 	struct scsi_disk *sdkp = to_scsi_disk(dev);
527 	struct scsi_device *sdp = sdkp->device;
528 	unsigned long max;
529 	int err;
530 
531 	if (!capable(CAP_SYS_ADMIN))
532 		return -EACCES;
533 
534 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
535 		return -EINVAL;
536 
537 	err = kstrtoul(buf, 10, &max);
538 
539 	if (err)
540 		return err;
541 
542 	if (max == 0)
543 		sdp->no_write_same = 1;
544 	else if (max <= SD_MAX_WS16_BLOCKS) {
545 		sdp->no_write_same = 0;
546 		sdkp->max_ws_blocks = max;
547 	}
548 
549 	sd_config_write_same(sdkp);
550 
551 	return count;
552 }
553 static DEVICE_ATTR_RW(max_write_same_blocks);
554 
555 static ssize_t
556 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
557 {
558 	struct scsi_disk *sdkp = to_scsi_disk(dev);
559 
560 	if (sdkp->device->type == TYPE_ZBC)
561 		return sprintf(buf, "host-managed\n");
562 	if (sdkp->zoned == 1)
563 		return sprintf(buf, "host-aware\n");
564 	if (sdkp->zoned == 2)
565 		return sprintf(buf, "drive-managed\n");
566 	return sprintf(buf, "none\n");
567 }
568 static DEVICE_ATTR_RO(zoned_cap);
569 
570 static ssize_t
571 max_retries_store(struct device *dev, struct device_attribute *attr,
572 		  const char *buf, size_t count)
573 {
574 	struct scsi_disk *sdkp = to_scsi_disk(dev);
575 	struct scsi_device *sdev = sdkp->device;
576 	int retries, err;
577 
578 	err = kstrtoint(buf, 10, &retries);
579 	if (err)
580 		return err;
581 
582 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
583 		sdkp->max_retries = retries;
584 		return count;
585 	}
586 
587 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
588 		    SD_MAX_RETRIES);
589 	return -EINVAL;
590 }
591 
592 static ssize_t
593 max_retries_show(struct device *dev, struct device_attribute *attr,
594 		 char *buf)
595 {
596 	struct scsi_disk *sdkp = to_scsi_disk(dev);
597 
598 	return sprintf(buf, "%d\n", sdkp->max_retries);
599 }
600 
601 static DEVICE_ATTR_RW(max_retries);
602 
603 static struct attribute *sd_disk_attrs[] = {
604 	&dev_attr_cache_type.attr,
605 	&dev_attr_FUA.attr,
606 	&dev_attr_allow_restart.attr,
607 	&dev_attr_manage_start_stop.attr,
608 	&dev_attr_manage_system_start_stop.attr,
609 	&dev_attr_manage_runtime_start_stop.attr,
610 	&dev_attr_protection_type.attr,
611 	&dev_attr_protection_mode.attr,
612 	&dev_attr_app_tag_own.attr,
613 	&dev_attr_thin_provisioning.attr,
614 	&dev_attr_provisioning_mode.attr,
615 	&dev_attr_zeroing_mode.attr,
616 	&dev_attr_max_write_same_blocks.attr,
617 	&dev_attr_max_medium_access_timeouts.attr,
618 	&dev_attr_zoned_cap.attr,
619 	&dev_attr_max_retries.attr,
620 	NULL,
621 };
622 ATTRIBUTE_GROUPS(sd_disk);
623 
624 static struct class sd_disk_class = {
625 	.name		= "scsi_disk",
626 	.dev_release	= scsi_disk_release,
627 	.dev_groups	= sd_disk_groups,
628 };
629 
630 /*
631  * Don't request a new module, as that could deadlock in multipath
632  * environment.
633  */
634 static void sd_default_probe(dev_t devt)
635 {
636 }
637 
638 /*
639  * Device no to disk mapping:
640  *
641  *       major         disc2     disc  p1
642  *   |............|.............|....|....| <- dev_t
643  *    31        20 19          8 7  4 3  0
644  *
645  * Inside a major, we have 16k disks, however mapped non-
646  * contiguously. The first 16 disks are for major0, the next
647  * ones with major1, ... Disk 256 is for major0 again, disk 272
648  * for major1, ...
649  * As we stay compatible with our numbering scheme, we can reuse
650  * the well-know SCSI majors 8, 65--71, 136--143.
651  */
652 static int sd_major(int major_idx)
653 {
654 	switch (major_idx) {
655 	case 0:
656 		return SCSI_DISK0_MAJOR;
657 	case 1 ... 7:
658 		return SCSI_DISK1_MAJOR + major_idx - 1;
659 	case 8 ... 15:
660 		return SCSI_DISK8_MAJOR + major_idx - 8;
661 	default:
662 		BUG();
663 		return 0;	/* shut up gcc */
664 	}
665 }
666 
667 #ifdef CONFIG_BLK_SED_OPAL
668 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
669 		size_t len, bool send)
670 {
671 	struct scsi_disk *sdkp = data;
672 	struct scsi_device *sdev = sdkp->device;
673 	u8 cdb[12] = { 0, };
674 	const struct scsi_exec_args exec_args = {
675 		.req_flags = BLK_MQ_REQ_PM,
676 	};
677 	int ret;
678 
679 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
680 	cdb[1] = secp;
681 	put_unaligned_be16(spsp, &cdb[2]);
682 	put_unaligned_be32(len, &cdb[6]);
683 
684 	ret = scsi_execute_cmd(sdev, cdb, send ? REQ_OP_DRV_OUT : REQ_OP_DRV_IN,
685 			       buffer, len, SD_TIMEOUT, sdkp->max_retries,
686 			       &exec_args);
687 	return ret <= 0 ? ret : -EIO;
688 }
689 #endif /* CONFIG_BLK_SED_OPAL */
690 
691 /*
692  * Look up the DIX operation based on whether the command is read or
693  * write and whether dix and dif are enabled.
694  */
695 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
696 {
697 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
698 	static const unsigned int ops[] = {	/* wrt dix dif */
699 		SCSI_PROT_NORMAL,		/*  0	0   0  */
700 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
701 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
702 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
703 		SCSI_PROT_NORMAL,		/*  1	0   0  */
704 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
705 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
706 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
707 	};
708 
709 	return ops[write << 2 | dix << 1 | dif];
710 }
711 
712 /*
713  * Returns a mask of the protection flags that are valid for a given DIX
714  * operation.
715  */
716 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
717 {
718 	static const unsigned int flag_mask[] = {
719 		[SCSI_PROT_NORMAL]		= 0,
720 
721 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
722 						  SCSI_PROT_GUARD_CHECK |
723 						  SCSI_PROT_REF_CHECK |
724 						  SCSI_PROT_REF_INCREMENT,
725 
726 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
727 						  SCSI_PROT_IP_CHECKSUM,
728 
729 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
730 						  SCSI_PROT_GUARD_CHECK |
731 						  SCSI_PROT_REF_CHECK |
732 						  SCSI_PROT_REF_INCREMENT |
733 						  SCSI_PROT_IP_CHECKSUM,
734 
735 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
736 						  SCSI_PROT_REF_INCREMENT,
737 
738 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
739 						  SCSI_PROT_REF_CHECK |
740 						  SCSI_PROT_REF_INCREMENT |
741 						  SCSI_PROT_IP_CHECKSUM,
742 
743 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
744 						  SCSI_PROT_GUARD_CHECK |
745 						  SCSI_PROT_REF_CHECK |
746 						  SCSI_PROT_REF_INCREMENT |
747 						  SCSI_PROT_IP_CHECKSUM,
748 	};
749 
750 	return flag_mask[prot_op];
751 }
752 
753 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
754 					   unsigned int dix, unsigned int dif)
755 {
756 	struct request *rq = scsi_cmd_to_rq(scmd);
757 	struct bio *bio = rq->bio;
758 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
759 	unsigned int protect = 0;
760 
761 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
762 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
763 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
764 
765 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
766 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
767 	}
768 
769 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
770 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
771 
772 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
773 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
774 	}
775 
776 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
777 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
778 
779 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
780 			protect = 3 << 5;	/* Disable target PI checking */
781 		else
782 			protect = 1 << 5;	/* Enable target PI checking */
783 	}
784 
785 	scsi_set_prot_op(scmd, prot_op);
786 	scsi_set_prot_type(scmd, dif);
787 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
788 
789 	return protect;
790 }
791 
792 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
793 {
794 	struct request_queue *q = sdkp->disk->queue;
795 	unsigned int logical_block_size = sdkp->device->sector_size;
796 	unsigned int max_blocks = 0;
797 
798 	q->limits.discard_alignment =
799 		sdkp->unmap_alignment * logical_block_size;
800 	q->limits.discard_granularity =
801 		max(sdkp->physical_block_size,
802 		    sdkp->unmap_granularity * logical_block_size);
803 	sdkp->provisioning_mode = mode;
804 
805 	switch (mode) {
806 
807 	case SD_LBP_FULL:
808 	case SD_LBP_DISABLE:
809 		blk_queue_max_discard_sectors(q, 0);
810 		return;
811 
812 	case SD_LBP_UNMAP:
813 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
814 					  (u32)SD_MAX_WS16_BLOCKS);
815 		break;
816 
817 	case SD_LBP_WS16:
818 		if (sdkp->device->unmap_limit_for_ws)
819 			max_blocks = sdkp->max_unmap_blocks;
820 		else
821 			max_blocks = sdkp->max_ws_blocks;
822 
823 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
824 		break;
825 
826 	case SD_LBP_WS10:
827 		if (sdkp->device->unmap_limit_for_ws)
828 			max_blocks = sdkp->max_unmap_blocks;
829 		else
830 			max_blocks = sdkp->max_ws_blocks;
831 
832 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
833 		break;
834 
835 	case SD_LBP_ZERO:
836 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
837 					  (u32)SD_MAX_WS10_BLOCKS);
838 		break;
839 	}
840 
841 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
842 }
843 
844 static void *sd_set_special_bvec(struct request *rq, unsigned int data_len)
845 {
846 	struct page *page;
847 
848 	page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
849 	if (!page)
850 		return NULL;
851 	clear_highpage(page);
852 	bvec_set_page(&rq->special_vec, page, data_len, 0);
853 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
854 	return bvec_virt(&rq->special_vec);
855 }
856 
857 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
858 {
859 	struct scsi_device *sdp = cmd->device;
860 	struct request *rq = scsi_cmd_to_rq(cmd);
861 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
862 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
863 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
864 	unsigned int data_len = 24;
865 	char *buf;
866 
867 	buf = sd_set_special_bvec(rq, data_len);
868 	if (!buf)
869 		return BLK_STS_RESOURCE;
870 
871 	cmd->cmd_len = 10;
872 	cmd->cmnd[0] = UNMAP;
873 	cmd->cmnd[8] = 24;
874 
875 	put_unaligned_be16(6 + 16, &buf[0]);
876 	put_unaligned_be16(16, &buf[2]);
877 	put_unaligned_be64(lba, &buf[8]);
878 	put_unaligned_be32(nr_blocks, &buf[16]);
879 
880 	cmd->allowed = sdkp->max_retries;
881 	cmd->transfersize = data_len;
882 	rq->timeout = SD_TIMEOUT;
883 
884 	return scsi_alloc_sgtables(cmd);
885 }
886 
887 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
888 		bool unmap)
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 	u32 data_len = sdp->sector_size;
896 
897 	if (!sd_set_special_bvec(rq, data_len))
898 		return BLK_STS_RESOURCE;
899 
900 	cmd->cmd_len = 16;
901 	cmd->cmnd[0] = WRITE_SAME_16;
902 	if (unmap)
903 		cmd->cmnd[1] = 0x8; /* UNMAP */
904 	put_unaligned_be64(lba, &cmd->cmnd[2]);
905 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
906 
907 	cmd->allowed = sdkp->max_retries;
908 	cmd->transfersize = data_len;
909 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
910 
911 	return scsi_alloc_sgtables(cmd);
912 }
913 
914 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
915 		bool unmap)
916 {
917 	struct scsi_device *sdp = cmd->device;
918 	struct request *rq = scsi_cmd_to_rq(cmd);
919 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
920 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
921 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
922 	u32 data_len = sdp->sector_size;
923 
924 	if (!sd_set_special_bvec(rq, data_len))
925 		return BLK_STS_RESOURCE;
926 
927 	cmd->cmd_len = 10;
928 	cmd->cmnd[0] = WRITE_SAME;
929 	if (unmap)
930 		cmd->cmnd[1] = 0x8; /* UNMAP */
931 	put_unaligned_be32(lba, &cmd->cmnd[2]);
932 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
933 
934 	cmd->allowed = sdkp->max_retries;
935 	cmd->transfersize = data_len;
936 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
937 
938 	return scsi_alloc_sgtables(cmd);
939 }
940 
941 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
942 {
943 	struct request *rq = scsi_cmd_to_rq(cmd);
944 	struct scsi_device *sdp = cmd->device;
945 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
946 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
947 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
948 
949 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
950 		switch (sdkp->zeroing_mode) {
951 		case SD_ZERO_WS16_UNMAP:
952 			return sd_setup_write_same16_cmnd(cmd, true);
953 		case SD_ZERO_WS10_UNMAP:
954 			return sd_setup_write_same10_cmnd(cmd, true);
955 		}
956 	}
957 
958 	if (sdp->no_write_same) {
959 		rq->rq_flags |= RQF_QUIET;
960 		return BLK_STS_TARGET;
961 	}
962 
963 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
964 		return sd_setup_write_same16_cmnd(cmd, false);
965 
966 	return sd_setup_write_same10_cmnd(cmd, false);
967 }
968 
969 static void sd_config_write_same(struct scsi_disk *sdkp)
970 {
971 	struct request_queue *q = sdkp->disk->queue;
972 	unsigned int logical_block_size = sdkp->device->sector_size;
973 
974 	if (sdkp->device->no_write_same) {
975 		sdkp->max_ws_blocks = 0;
976 		goto out;
977 	}
978 
979 	/* Some devices can not handle block counts above 0xffff despite
980 	 * supporting WRITE SAME(16). Consequently we default to 64k
981 	 * blocks per I/O unless the device explicitly advertises a
982 	 * bigger limit.
983 	 */
984 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
985 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
986 						   (u32)SD_MAX_WS16_BLOCKS);
987 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
988 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
989 						   (u32)SD_MAX_WS10_BLOCKS);
990 	else {
991 		sdkp->device->no_write_same = 1;
992 		sdkp->max_ws_blocks = 0;
993 	}
994 
995 	if (sdkp->lbprz && sdkp->lbpws)
996 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
997 	else if (sdkp->lbprz && sdkp->lbpws10)
998 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
999 	else if (sdkp->max_ws_blocks)
1000 		sdkp->zeroing_mode = SD_ZERO_WS;
1001 	else
1002 		sdkp->zeroing_mode = SD_ZERO_WRITE;
1003 
1004 	if (sdkp->max_ws_blocks &&
1005 	    sdkp->physical_block_size > logical_block_size) {
1006 		/*
1007 		 * Reporting a maximum number of blocks that is not aligned
1008 		 * on the device physical size would cause a large write same
1009 		 * request to be split into physically unaligned chunks by
1010 		 * __blkdev_issue_write_zeroes() even if the caller of this
1011 		 * functions took care to align the large request. So make sure
1012 		 * the maximum reported is aligned to the device physical block
1013 		 * size. This is only an optional optimization for regular
1014 		 * disks, but this is mandatory to avoid failure of large write
1015 		 * same requests directed at sequential write required zones of
1016 		 * host-managed ZBC disks.
1017 		 */
1018 		sdkp->max_ws_blocks =
1019 			round_down(sdkp->max_ws_blocks,
1020 				   bytes_to_logical(sdkp->device,
1021 						    sdkp->physical_block_size));
1022 	}
1023 
1024 out:
1025 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1026 					 (logical_block_size >> 9));
1027 }
1028 
1029 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1030 {
1031 	struct request *rq = scsi_cmd_to_rq(cmd);
1032 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1033 
1034 	/* flush requests don't perform I/O, zero the S/G table */
1035 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1036 
1037 	if (cmd->device->use_16_for_sync) {
1038 		cmd->cmnd[0] = SYNCHRONIZE_CACHE_16;
1039 		cmd->cmd_len = 16;
1040 	} else {
1041 		cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1042 		cmd->cmd_len = 10;
1043 	}
1044 	cmd->transfersize = 0;
1045 	cmd->allowed = sdkp->max_retries;
1046 
1047 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1048 	return BLK_STS_OK;
1049 }
1050 
1051 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1052 				       sector_t lba, unsigned int nr_blocks,
1053 				       unsigned char flags, unsigned int dld)
1054 {
1055 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1056 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1057 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1058 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1059 	cmd->cmnd[10] = flags;
1060 	cmd->cmnd[11] = dld & 0x07;
1061 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1062 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1063 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1064 
1065 	return BLK_STS_OK;
1066 }
1067 
1068 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1069 				       sector_t lba, unsigned int nr_blocks,
1070 				       unsigned char flags, unsigned int dld)
1071 {
1072 	cmd->cmd_len  = 16;
1073 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1074 	cmd->cmnd[1]  = flags | ((dld >> 2) & 0x01);
1075 	cmd->cmnd[14] = (dld & 0x03) << 6;
1076 	cmd->cmnd[15] = 0;
1077 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1078 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1079 
1080 	return BLK_STS_OK;
1081 }
1082 
1083 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1084 				       sector_t lba, unsigned int nr_blocks,
1085 				       unsigned char flags)
1086 {
1087 	cmd->cmd_len = 10;
1088 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1089 	cmd->cmnd[1] = flags;
1090 	cmd->cmnd[6] = 0;
1091 	cmd->cmnd[9] = 0;
1092 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1093 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1094 
1095 	return BLK_STS_OK;
1096 }
1097 
1098 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1099 				      sector_t lba, unsigned int nr_blocks,
1100 				      unsigned char flags)
1101 {
1102 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1103 	if (WARN_ON_ONCE(nr_blocks == 0))
1104 		return BLK_STS_IOERR;
1105 
1106 	if (unlikely(flags & 0x8)) {
1107 		/*
1108 		 * This happens only if this drive failed 10byte rw
1109 		 * command with ILLEGAL_REQUEST during operation and
1110 		 * thus turned off use_10_for_rw.
1111 		 */
1112 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1113 		return BLK_STS_IOERR;
1114 	}
1115 
1116 	cmd->cmd_len = 6;
1117 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1118 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1119 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1120 	cmd->cmnd[3] = lba & 0xff;
1121 	cmd->cmnd[4] = nr_blocks;
1122 	cmd->cmnd[5] = 0;
1123 
1124 	return BLK_STS_OK;
1125 }
1126 
1127 /*
1128  * Check if a command has a duration limit set. If it does, and the target
1129  * device supports CDL and the feature is enabled, return the limit
1130  * descriptor index to use. Return 0 (no limit) otherwise.
1131  */
1132 static int sd_cdl_dld(struct scsi_disk *sdkp, struct scsi_cmnd *scmd)
1133 {
1134 	struct scsi_device *sdp = sdkp->device;
1135 	int hint;
1136 
1137 	if (!sdp->cdl_supported || !sdp->cdl_enable)
1138 		return 0;
1139 
1140 	/*
1141 	 * Use "no limit" if the request ioprio does not specify a duration
1142 	 * limit hint.
1143 	 */
1144 	hint = IOPRIO_PRIO_HINT(req_get_ioprio(scsi_cmd_to_rq(scmd)));
1145 	if (hint < IOPRIO_HINT_DEV_DURATION_LIMIT_1 ||
1146 	    hint > IOPRIO_HINT_DEV_DURATION_LIMIT_7)
1147 		return 0;
1148 
1149 	return (hint - IOPRIO_HINT_DEV_DURATION_LIMIT_1) + 1;
1150 }
1151 
1152 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1153 {
1154 	struct request *rq = scsi_cmd_to_rq(cmd);
1155 	struct scsi_device *sdp = cmd->device;
1156 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1157 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1158 	sector_t threshold;
1159 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1160 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1161 	bool write = rq_data_dir(rq) == WRITE;
1162 	unsigned char protect, fua;
1163 	unsigned int dld;
1164 	blk_status_t ret;
1165 	unsigned int dif;
1166 	bool dix;
1167 
1168 	ret = scsi_alloc_sgtables(cmd);
1169 	if (ret != BLK_STS_OK)
1170 		return ret;
1171 
1172 	ret = BLK_STS_IOERR;
1173 	if (!scsi_device_online(sdp) || sdp->changed) {
1174 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1175 		goto fail;
1176 	}
1177 
1178 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1179 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1180 		goto fail;
1181 	}
1182 
1183 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1184 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1185 		goto fail;
1186 	}
1187 
1188 	/*
1189 	 * Some SD card readers can't handle accesses which touch the
1190 	 * last one or two logical blocks. Split accesses as needed.
1191 	 */
1192 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1193 
1194 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1195 		if (lba < threshold) {
1196 			/* Access up to the threshold but not beyond */
1197 			nr_blocks = threshold - lba;
1198 		} else {
1199 			/* Access only a single logical block */
1200 			nr_blocks = 1;
1201 		}
1202 	}
1203 
1204 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1205 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1206 		if (ret)
1207 			goto fail;
1208 	}
1209 
1210 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1211 	dix = scsi_prot_sg_count(cmd);
1212 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1213 	dld = sd_cdl_dld(sdkp, cmd);
1214 
1215 	if (dif || dix)
1216 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1217 	else
1218 		protect = 0;
1219 
1220 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1221 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1222 					 protect | fua, dld);
1223 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1224 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1225 					 protect | fua, dld);
1226 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1227 		   sdp->use_10_for_rw || protect) {
1228 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1229 					 protect | fua);
1230 	} else {
1231 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1232 					protect | fua);
1233 	}
1234 
1235 	if (unlikely(ret != BLK_STS_OK))
1236 		goto fail;
1237 
1238 	/*
1239 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1240 	 * host adapter, it's safe to assume that we can at least transfer
1241 	 * this many bytes between each connect / disconnect.
1242 	 */
1243 	cmd->transfersize = sdp->sector_size;
1244 	cmd->underflow = nr_blocks << 9;
1245 	cmd->allowed = sdkp->max_retries;
1246 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1247 
1248 	SCSI_LOG_HLQUEUE(1,
1249 			 scmd_printk(KERN_INFO, cmd,
1250 				     "%s: block=%llu, count=%d\n", __func__,
1251 				     (unsigned long long)blk_rq_pos(rq),
1252 				     blk_rq_sectors(rq)));
1253 	SCSI_LOG_HLQUEUE(2,
1254 			 scmd_printk(KERN_INFO, cmd,
1255 				     "%s %d/%u 512 byte blocks.\n",
1256 				     write ? "writing" : "reading", nr_blocks,
1257 				     blk_rq_sectors(rq)));
1258 
1259 	/*
1260 	 * This indicates that the command is ready from our end to be queued.
1261 	 */
1262 	return BLK_STS_OK;
1263 fail:
1264 	scsi_free_sgtables(cmd);
1265 	return ret;
1266 }
1267 
1268 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1269 {
1270 	struct request *rq = scsi_cmd_to_rq(cmd);
1271 
1272 	switch (req_op(rq)) {
1273 	case REQ_OP_DISCARD:
1274 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1275 		case SD_LBP_UNMAP:
1276 			return sd_setup_unmap_cmnd(cmd);
1277 		case SD_LBP_WS16:
1278 			return sd_setup_write_same16_cmnd(cmd, true);
1279 		case SD_LBP_WS10:
1280 			return sd_setup_write_same10_cmnd(cmd, true);
1281 		case SD_LBP_ZERO:
1282 			return sd_setup_write_same10_cmnd(cmd, false);
1283 		default:
1284 			return BLK_STS_TARGET;
1285 		}
1286 	case REQ_OP_WRITE_ZEROES:
1287 		return sd_setup_write_zeroes_cmnd(cmd);
1288 	case REQ_OP_FLUSH:
1289 		return sd_setup_flush_cmnd(cmd);
1290 	case REQ_OP_READ:
1291 	case REQ_OP_WRITE:
1292 	case REQ_OP_ZONE_APPEND:
1293 		return sd_setup_read_write_cmnd(cmd);
1294 	case REQ_OP_ZONE_RESET:
1295 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1296 						   false);
1297 	case REQ_OP_ZONE_RESET_ALL:
1298 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1299 						   true);
1300 	case REQ_OP_ZONE_OPEN:
1301 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1302 	case REQ_OP_ZONE_CLOSE:
1303 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1304 	case REQ_OP_ZONE_FINISH:
1305 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1306 	default:
1307 		WARN_ON_ONCE(1);
1308 		return BLK_STS_NOTSUPP;
1309 	}
1310 }
1311 
1312 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1313 {
1314 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1315 
1316 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1317 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1318 }
1319 
1320 static bool sd_need_revalidate(struct gendisk *disk, struct scsi_disk *sdkp)
1321 {
1322 	if (sdkp->device->removable || sdkp->write_prot) {
1323 		if (disk_check_media_change(disk))
1324 			return true;
1325 	}
1326 
1327 	/*
1328 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1329 	 * nothing to do with partitions, BLKRRPART is used to force a full
1330 	 * revalidate after things like a format for historical reasons.
1331 	 */
1332 	return test_bit(GD_NEED_PART_SCAN, &disk->state);
1333 }
1334 
1335 /**
1336  *	sd_open - open a scsi disk device
1337  *	@disk: disk to open
1338  *	@mode: open mode
1339  *
1340  *	Returns 0 if successful. Returns a negated errno value in case
1341  *	of error.
1342  *
1343  *	Note: This can be called from a user context (e.g. fsck(1) )
1344  *	or from within the kernel (e.g. as a result of a mount(1) ).
1345  *	In the latter case @inode and @filp carry an abridged amount
1346  *	of information as noted above.
1347  *
1348  *	Locking: called with disk->open_mutex held.
1349  **/
1350 static int sd_open(struct gendisk *disk, blk_mode_t mode)
1351 {
1352 	struct scsi_disk *sdkp = scsi_disk(disk);
1353 	struct scsi_device *sdev = sdkp->device;
1354 	int retval;
1355 
1356 	if (scsi_device_get(sdev))
1357 		return -ENXIO;
1358 
1359 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1360 
1361 	/*
1362 	 * If the device is in error recovery, wait until it is done.
1363 	 * If the device is offline, then disallow any access to it.
1364 	 */
1365 	retval = -ENXIO;
1366 	if (!scsi_block_when_processing_errors(sdev))
1367 		goto error_out;
1368 
1369 	if (sd_need_revalidate(disk, sdkp))
1370 		sd_revalidate_disk(disk);
1371 
1372 	/*
1373 	 * If the drive is empty, just let the open fail.
1374 	 */
1375 	retval = -ENOMEDIUM;
1376 	if (sdev->removable && !sdkp->media_present &&
1377 	    !(mode & BLK_OPEN_NDELAY))
1378 		goto error_out;
1379 
1380 	/*
1381 	 * If the device has the write protect tab set, have the open fail
1382 	 * if the user expects to be able to write to the thing.
1383 	 */
1384 	retval = -EROFS;
1385 	if (sdkp->write_prot && (mode & BLK_OPEN_WRITE))
1386 		goto error_out;
1387 
1388 	/*
1389 	 * It is possible that the disk changing stuff resulted in
1390 	 * the device being taken offline.  If this is the case,
1391 	 * report this to the user, and don't pretend that the
1392 	 * open actually succeeded.
1393 	 */
1394 	retval = -ENXIO;
1395 	if (!scsi_device_online(sdev))
1396 		goto error_out;
1397 
1398 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1399 		if (scsi_block_when_processing_errors(sdev))
1400 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1401 	}
1402 
1403 	return 0;
1404 
1405 error_out:
1406 	scsi_device_put(sdev);
1407 	return retval;
1408 }
1409 
1410 /**
1411  *	sd_release - invoked when the (last) close(2) is called on this
1412  *	scsi disk.
1413  *	@disk: disk to release
1414  *
1415  *	Returns 0.
1416  *
1417  *	Note: may block (uninterruptible) if error recovery is underway
1418  *	on this disk.
1419  *
1420  *	Locking: called with disk->open_mutex held.
1421  **/
1422 static void sd_release(struct gendisk *disk)
1423 {
1424 	struct scsi_disk *sdkp = scsi_disk(disk);
1425 	struct scsi_device *sdev = sdkp->device;
1426 
1427 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1428 
1429 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1430 		if (scsi_block_when_processing_errors(sdev))
1431 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1432 	}
1433 
1434 	scsi_device_put(sdev);
1435 }
1436 
1437 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1438 {
1439 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1440 	struct scsi_device *sdp = sdkp->device;
1441 	struct Scsi_Host *host = sdp->host;
1442 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1443 	int diskinfo[4];
1444 
1445 	/* default to most commonly used values */
1446 	diskinfo[0] = 0x40;	/* 1 << 6 */
1447 	diskinfo[1] = 0x20;	/* 1 << 5 */
1448 	diskinfo[2] = capacity >> 11;
1449 
1450 	/* override with calculated, extended default, or driver values */
1451 	if (host->hostt->bios_param)
1452 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1453 	else
1454 		scsicam_bios_param(bdev, capacity, diskinfo);
1455 
1456 	geo->heads = diskinfo[0];
1457 	geo->sectors = diskinfo[1];
1458 	geo->cylinders = diskinfo[2];
1459 	return 0;
1460 }
1461 
1462 /**
1463  *	sd_ioctl - process an ioctl
1464  *	@bdev: target block device
1465  *	@mode: open mode
1466  *	@cmd: ioctl command number
1467  *	@arg: this is third argument given to ioctl(2) system call.
1468  *	Often contains a pointer.
1469  *
1470  *	Returns 0 if successful (some ioctls return positive numbers on
1471  *	success as well). Returns a negated errno value in case of error.
1472  *
1473  *	Note: most ioctls are forward onto the block subsystem or further
1474  *	down in the scsi subsystem.
1475  **/
1476 static int sd_ioctl(struct block_device *bdev, blk_mode_t mode,
1477 		    unsigned int cmd, unsigned long arg)
1478 {
1479 	struct gendisk *disk = bdev->bd_disk;
1480 	struct scsi_disk *sdkp = scsi_disk(disk);
1481 	struct scsi_device *sdp = sdkp->device;
1482 	void __user *p = (void __user *)arg;
1483 	int error;
1484 
1485 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1486 				    "cmd=0x%x\n", disk->disk_name, cmd));
1487 
1488 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1489 		return -ENOIOCTLCMD;
1490 
1491 	/*
1492 	 * If we are in the middle of error recovery, don't let anyone
1493 	 * else try and use this device.  Also, if error recovery fails, it
1494 	 * may try and take the device offline, in which case all further
1495 	 * access to the device is prohibited.
1496 	 */
1497 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1498 			(mode & BLK_OPEN_NDELAY));
1499 	if (error)
1500 		return error;
1501 
1502 	if (is_sed_ioctl(cmd))
1503 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1504 	return scsi_ioctl(sdp, mode & BLK_OPEN_WRITE, cmd, p);
1505 }
1506 
1507 static void set_media_not_present(struct scsi_disk *sdkp)
1508 {
1509 	if (sdkp->media_present)
1510 		sdkp->device->changed = 1;
1511 
1512 	if (sdkp->device->removable) {
1513 		sdkp->media_present = 0;
1514 		sdkp->capacity = 0;
1515 	}
1516 }
1517 
1518 static int media_not_present(struct scsi_disk *sdkp,
1519 			     struct scsi_sense_hdr *sshdr)
1520 {
1521 	if (!scsi_sense_valid(sshdr))
1522 		return 0;
1523 
1524 	/* not invoked for commands that could return deferred errors */
1525 	switch (sshdr->sense_key) {
1526 	case UNIT_ATTENTION:
1527 	case NOT_READY:
1528 		/* medium not present */
1529 		if (sshdr->asc == 0x3A) {
1530 			set_media_not_present(sdkp);
1531 			return 1;
1532 		}
1533 	}
1534 	return 0;
1535 }
1536 
1537 /**
1538  *	sd_check_events - check media events
1539  *	@disk: kernel device descriptor
1540  *	@clearing: disk events currently being cleared
1541  *
1542  *	Returns mask of DISK_EVENT_*.
1543  *
1544  *	Note: this function is invoked from the block subsystem.
1545  **/
1546 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1547 {
1548 	struct scsi_disk *sdkp = disk->private_data;
1549 	struct scsi_device *sdp;
1550 	int retval;
1551 	bool disk_changed;
1552 
1553 	if (!sdkp)
1554 		return 0;
1555 
1556 	sdp = sdkp->device;
1557 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1558 
1559 	/*
1560 	 * If the device is offline, don't send any commands - just pretend as
1561 	 * if the command failed.  If the device ever comes back online, we
1562 	 * can deal with it then.  It is only because of unrecoverable errors
1563 	 * that we would ever take a device offline in the first place.
1564 	 */
1565 	if (!scsi_device_online(sdp)) {
1566 		set_media_not_present(sdkp);
1567 		goto out;
1568 	}
1569 
1570 	/*
1571 	 * Using TEST_UNIT_READY enables differentiation between drive with
1572 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1573 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1574 	 *
1575 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1576 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1577 	 * sd_revalidate() is called.
1578 	 */
1579 	if (scsi_block_when_processing_errors(sdp)) {
1580 		struct scsi_sense_hdr sshdr = { 0, };
1581 
1582 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1583 					      &sshdr);
1584 
1585 		/* failed to execute TUR, assume media not present */
1586 		if (retval < 0 || host_byte(retval)) {
1587 			set_media_not_present(sdkp);
1588 			goto out;
1589 		}
1590 
1591 		if (media_not_present(sdkp, &sshdr))
1592 			goto out;
1593 	}
1594 
1595 	/*
1596 	 * For removable scsi disk we have to recognise the presence
1597 	 * of a disk in the drive.
1598 	 */
1599 	if (!sdkp->media_present)
1600 		sdp->changed = 1;
1601 	sdkp->media_present = 1;
1602 out:
1603 	/*
1604 	 * sdp->changed is set under the following conditions:
1605 	 *
1606 	 *	Medium present state has changed in either direction.
1607 	 *	Device has indicated UNIT_ATTENTION.
1608 	 */
1609 	disk_changed = sdp->changed;
1610 	sdp->changed = 0;
1611 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1612 }
1613 
1614 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1615 {
1616 	int retries, res;
1617 	struct scsi_device *sdp = sdkp->device;
1618 	const int timeout = sdp->request_queue->rq_timeout
1619 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1620 	struct scsi_sense_hdr my_sshdr;
1621 	const struct scsi_exec_args exec_args = {
1622 		.req_flags = BLK_MQ_REQ_PM,
1623 		/* caller might not be interested in sense, but we need it */
1624 		.sshdr = sshdr ? : &my_sshdr,
1625 	};
1626 
1627 	if (!scsi_device_online(sdp))
1628 		return -ENODEV;
1629 
1630 	sshdr = exec_args.sshdr;
1631 
1632 	for (retries = 3; retries > 0; --retries) {
1633 		unsigned char cmd[16] = { 0 };
1634 
1635 		if (sdp->use_16_for_sync)
1636 			cmd[0] = SYNCHRONIZE_CACHE_16;
1637 		else
1638 			cmd[0] = SYNCHRONIZE_CACHE;
1639 		/*
1640 		 * Leave the rest of the command zero to indicate
1641 		 * flush everything.
1642 		 */
1643 		res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
1644 				       timeout, sdkp->max_retries, &exec_args);
1645 		if (res == 0)
1646 			break;
1647 	}
1648 
1649 	if (res) {
1650 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1651 
1652 		if (res < 0)
1653 			return res;
1654 
1655 		if (scsi_status_is_check_condition(res) &&
1656 		    scsi_sense_valid(sshdr)) {
1657 			sd_print_sense_hdr(sdkp, sshdr);
1658 
1659 			/* we need to evaluate the error return  */
1660 			if (sshdr->asc == 0x3a ||	/* medium not present */
1661 			    sshdr->asc == 0x20 ||	/* invalid command */
1662 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1663 				/* this is no error here */
1664 				return 0;
1665 		}
1666 
1667 		switch (host_byte(res)) {
1668 		/* ignore errors due to racing a disconnection */
1669 		case DID_BAD_TARGET:
1670 		case DID_NO_CONNECT:
1671 			return 0;
1672 		/* signal the upper layer it might try again */
1673 		case DID_BUS_BUSY:
1674 		case DID_IMM_RETRY:
1675 		case DID_REQUEUE:
1676 		case DID_SOFT_ERROR:
1677 			return -EBUSY;
1678 		default:
1679 			return -EIO;
1680 		}
1681 	}
1682 	return 0;
1683 }
1684 
1685 static void sd_rescan(struct device *dev)
1686 {
1687 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1688 
1689 	sd_revalidate_disk(sdkp->disk);
1690 }
1691 
1692 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1693 		enum blk_unique_id type)
1694 {
1695 	struct scsi_device *sdev = scsi_disk(disk)->device;
1696 	const struct scsi_vpd *vpd;
1697 	const unsigned char *d;
1698 	int ret = -ENXIO, len;
1699 
1700 	rcu_read_lock();
1701 	vpd = rcu_dereference(sdev->vpd_pg83);
1702 	if (!vpd)
1703 		goto out_unlock;
1704 
1705 	ret = -EINVAL;
1706 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1707 		/* we only care about designators with LU association */
1708 		if (((d[1] >> 4) & 0x3) != 0x00)
1709 			continue;
1710 		if ((d[1] & 0xf) != type)
1711 			continue;
1712 
1713 		/*
1714 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1715 		 * keep looking as one with more entropy might still show up.
1716 		 */
1717 		len = d[3];
1718 		if (len != 8 && len != 12 && len != 16)
1719 			continue;
1720 		ret = len;
1721 		memcpy(id, d + 4, len);
1722 		if (len == 16)
1723 			break;
1724 	}
1725 out_unlock:
1726 	rcu_read_unlock();
1727 	return ret;
1728 }
1729 
1730 static int sd_scsi_to_pr_err(struct scsi_sense_hdr *sshdr, int result)
1731 {
1732 	switch (host_byte(result)) {
1733 	case DID_TRANSPORT_MARGINAL:
1734 	case DID_TRANSPORT_DISRUPTED:
1735 	case DID_BUS_BUSY:
1736 		return PR_STS_RETRY_PATH_FAILURE;
1737 	case DID_NO_CONNECT:
1738 		return PR_STS_PATH_FAILED;
1739 	case DID_TRANSPORT_FAILFAST:
1740 		return PR_STS_PATH_FAST_FAILED;
1741 	}
1742 
1743 	switch (status_byte(result)) {
1744 	case SAM_STAT_RESERVATION_CONFLICT:
1745 		return PR_STS_RESERVATION_CONFLICT;
1746 	case SAM_STAT_CHECK_CONDITION:
1747 		if (!scsi_sense_valid(sshdr))
1748 			return PR_STS_IOERR;
1749 
1750 		if (sshdr->sense_key == ILLEGAL_REQUEST &&
1751 		    (sshdr->asc == 0x26 || sshdr->asc == 0x24))
1752 			return -EINVAL;
1753 
1754 		fallthrough;
1755 	default:
1756 		return PR_STS_IOERR;
1757 	}
1758 }
1759 
1760 static int sd_pr_in_command(struct block_device *bdev, u8 sa,
1761 			    unsigned char *data, int data_len)
1762 {
1763 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1764 	struct scsi_device *sdev = sdkp->device;
1765 	struct scsi_sense_hdr sshdr;
1766 	u8 cmd[10] = { PERSISTENT_RESERVE_IN, sa };
1767 	const struct scsi_exec_args exec_args = {
1768 		.sshdr = &sshdr,
1769 	};
1770 	int result;
1771 
1772 	put_unaligned_be16(data_len, &cmd[7]);
1773 
1774 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_IN, data, data_len,
1775 				  SD_TIMEOUT, sdkp->max_retries, &exec_args);
1776 	if (scsi_status_is_check_condition(result) &&
1777 	    scsi_sense_valid(&sshdr)) {
1778 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1779 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1780 	}
1781 
1782 	if (result <= 0)
1783 		return result;
1784 
1785 	return sd_scsi_to_pr_err(&sshdr, result);
1786 }
1787 
1788 static int sd_pr_read_keys(struct block_device *bdev, struct pr_keys *keys_info)
1789 {
1790 	int result, i, data_offset, num_copy_keys;
1791 	u32 num_keys = keys_info->num_keys;
1792 	int data_len = num_keys * 8 + 8;
1793 	u8 *data;
1794 
1795 	data = kzalloc(data_len, GFP_KERNEL);
1796 	if (!data)
1797 		return -ENOMEM;
1798 
1799 	result = sd_pr_in_command(bdev, READ_KEYS, data, data_len);
1800 	if (result)
1801 		goto free_data;
1802 
1803 	keys_info->generation = get_unaligned_be32(&data[0]);
1804 	keys_info->num_keys = get_unaligned_be32(&data[4]) / 8;
1805 
1806 	data_offset = 8;
1807 	num_copy_keys = min(num_keys, keys_info->num_keys);
1808 
1809 	for (i = 0; i < num_copy_keys; i++) {
1810 		keys_info->keys[i] = get_unaligned_be64(&data[data_offset]);
1811 		data_offset += 8;
1812 	}
1813 
1814 free_data:
1815 	kfree(data);
1816 	return result;
1817 }
1818 
1819 static int sd_pr_read_reservation(struct block_device *bdev,
1820 				  struct pr_held_reservation *rsv)
1821 {
1822 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1823 	struct scsi_device *sdev = sdkp->device;
1824 	u8 data[24] = { };
1825 	int result, len;
1826 
1827 	result = sd_pr_in_command(bdev, READ_RESERVATION, data, sizeof(data));
1828 	if (result)
1829 		return result;
1830 
1831 	len = get_unaligned_be32(&data[4]);
1832 	if (!len)
1833 		return 0;
1834 
1835 	/* Make sure we have at least the key and type */
1836 	if (len < 14) {
1837 		sdev_printk(KERN_INFO, sdev,
1838 			    "READ RESERVATION failed due to short return buffer of %d bytes\n",
1839 			    len);
1840 		return -EINVAL;
1841 	}
1842 
1843 	rsv->generation = get_unaligned_be32(&data[0]);
1844 	rsv->key = get_unaligned_be64(&data[8]);
1845 	rsv->type = scsi_pr_type_to_block(data[21] & 0x0f);
1846 	return 0;
1847 }
1848 
1849 static int sd_pr_out_command(struct block_device *bdev, u8 sa, u64 key,
1850 			     u64 sa_key, enum scsi_pr_type type, u8 flags)
1851 {
1852 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1853 	struct scsi_device *sdev = sdkp->device;
1854 	struct scsi_sense_hdr sshdr;
1855 	const struct scsi_exec_args exec_args = {
1856 		.sshdr = &sshdr,
1857 	};
1858 	int result;
1859 	u8 cmd[16] = { 0, };
1860 	u8 data[24] = { 0, };
1861 
1862 	cmd[0] = PERSISTENT_RESERVE_OUT;
1863 	cmd[1] = sa;
1864 	cmd[2] = type;
1865 	put_unaligned_be32(sizeof(data), &cmd[5]);
1866 
1867 	put_unaligned_be64(key, &data[0]);
1868 	put_unaligned_be64(sa_key, &data[8]);
1869 	data[20] = flags;
1870 
1871 	result = scsi_execute_cmd(sdev, cmd, REQ_OP_DRV_OUT, &data,
1872 				  sizeof(data), SD_TIMEOUT, sdkp->max_retries,
1873 				  &exec_args);
1874 
1875 	if (scsi_status_is_check_condition(result) &&
1876 	    scsi_sense_valid(&sshdr)) {
1877 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1878 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1879 	}
1880 
1881 	if (result <= 0)
1882 		return result;
1883 
1884 	return sd_scsi_to_pr_err(&sshdr, result);
1885 }
1886 
1887 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1888 		u32 flags)
1889 {
1890 	if (flags & ~PR_FL_IGNORE_KEY)
1891 		return -EOPNOTSUPP;
1892 	return sd_pr_out_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1893 			old_key, new_key, 0,
1894 			(1 << 0) /* APTPL */);
1895 }
1896 
1897 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1898 		u32 flags)
1899 {
1900 	if (flags)
1901 		return -EOPNOTSUPP;
1902 	return sd_pr_out_command(bdev, 0x01, key, 0,
1903 				 block_pr_type_to_scsi(type), 0);
1904 }
1905 
1906 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1907 {
1908 	return sd_pr_out_command(bdev, 0x02, key, 0,
1909 				 block_pr_type_to_scsi(type), 0);
1910 }
1911 
1912 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1913 		enum pr_type type, bool abort)
1914 {
1915 	return sd_pr_out_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1916 				 block_pr_type_to_scsi(type), 0);
1917 }
1918 
1919 static int sd_pr_clear(struct block_device *bdev, u64 key)
1920 {
1921 	return sd_pr_out_command(bdev, 0x03, key, 0, 0, 0);
1922 }
1923 
1924 static const struct pr_ops sd_pr_ops = {
1925 	.pr_register	= sd_pr_register,
1926 	.pr_reserve	= sd_pr_reserve,
1927 	.pr_release	= sd_pr_release,
1928 	.pr_preempt	= sd_pr_preempt,
1929 	.pr_clear	= sd_pr_clear,
1930 	.pr_read_keys	= sd_pr_read_keys,
1931 	.pr_read_reservation = sd_pr_read_reservation,
1932 };
1933 
1934 static void scsi_disk_free_disk(struct gendisk *disk)
1935 {
1936 	struct scsi_disk *sdkp = scsi_disk(disk);
1937 
1938 	put_device(&sdkp->disk_dev);
1939 }
1940 
1941 static const struct block_device_operations sd_fops = {
1942 	.owner			= THIS_MODULE,
1943 	.open			= sd_open,
1944 	.release		= sd_release,
1945 	.ioctl			= sd_ioctl,
1946 	.getgeo			= sd_getgeo,
1947 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1948 	.check_events		= sd_check_events,
1949 	.unlock_native_capacity	= sd_unlock_native_capacity,
1950 	.report_zones		= sd_zbc_report_zones,
1951 	.get_unique_id		= sd_get_unique_id,
1952 	.free_disk		= scsi_disk_free_disk,
1953 	.pr_ops			= &sd_pr_ops,
1954 };
1955 
1956 /**
1957  *	sd_eh_reset - reset error handling callback
1958  *	@scmd:		sd-issued command that has failed
1959  *
1960  *	This function is called by the SCSI midlayer before starting
1961  *	SCSI EH. When counting medium access failures we have to be
1962  *	careful to register it only only once per device and SCSI EH run;
1963  *	there might be several timed out commands which will cause the
1964  *	'max_medium_access_timeouts' counter to trigger after the first
1965  *	SCSI EH run already and set the device to offline.
1966  *	So this function resets the internal counter before starting SCSI EH.
1967  **/
1968 static void sd_eh_reset(struct scsi_cmnd *scmd)
1969 {
1970 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1971 
1972 	/* New SCSI EH run, reset gate variable */
1973 	sdkp->ignore_medium_access_errors = false;
1974 }
1975 
1976 /**
1977  *	sd_eh_action - error handling callback
1978  *	@scmd:		sd-issued command that has failed
1979  *	@eh_disp:	The recovery disposition suggested by the midlayer
1980  *
1981  *	This function is called by the SCSI midlayer upon completion of an
1982  *	error test command (currently TEST UNIT READY). The result of sending
1983  *	the eh command is passed in eh_disp.  We're looking for devices that
1984  *	fail medium access commands but are OK with non access commands like
1985  *	test unit ready (so wrongly see the device as having a successful
1986  *	recovery)
1987  **/
1988 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1989 {
1990 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1991 	struct scsi_device *sdev = scmd->device;
1992 
1993 	if (!scsi_device_online(sdev) ||
1994 	    !scsi_medium_access_command(scmd) ||
1995 	    host_byte(scmd->result) != DID_TIME_OUT ||
1996 	    eh_disp != SUCCESS)
1997 		return eh_disp;
1998 
1999 	/*
2000 	 * The device has timed out executing a medium access command.
2001 	 * However, the TEST UNIT READY command sent during error
2002 	 * handling completed successfully. Either the device is in the
2003 	 * process of recovering or has it suffered an internal failure
2004 	 * that prevents access to the storage medium.
2005 	 */
2006 	if (!sdkp->ignore_medium_access_errors) {
2007 		sdkp->medium_access_timed_out++;
2008 		sdkp->ignore_medium_access_errors = true;
2009 	}
2010 
2011 	/*
2012 	 * If the device keeps failing read/write commands but TEST UNIT
2013 	 * READY always completes successfully we assume that medium
2014 	 * access is no longer possible and take the device offline.
2015 	 */
2016 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
2017 		scmd_printk(KERN_ERR, scmd,
2018 			    "Medium access timeout failure. Offlining disk!\n");
2019 		mutex_lock(&sdev->state_mutex);
2020 		scsi_device_set_state(sdev, SDEV_OFFLINE);
2021 		mutex_unlock(&sdev->state_mutex);
2022 
2023 		return SUCCESS;
2024 	}
2025 
2026 	return eh_disp;
2027 }
2028 
2029 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
2030 {
2031 	struct request *req = scsi_cmd_to_rq(scmd);
2032 	struct scsi_device *sdev = scmd->device;
2033 	unsigned int transferred, good_bytes;
2034 	u64 start_lba, end_lba, bad_lba;
2035 
2036 	/*
2037 	 * Some commands have a payload smaller than the device logical
2038 	 * block size (e.g. INQUIRY on a 4K disk).
2039 	 */
2040 	if (scsi_bufflen(scmd) <= sdev->sector_size)
2041 		return 0;
2042 
2043 	/* Check if we have a 'bad_lba' information */
2044 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
2045 				     SCSI_SENSE_BUFFERSIZE,
2046 				     &bad_lba))
2047 		return 0;
2048 
2049 	/*
2050 	 * If the bad lba was reported incorrectly, we have no idea where
2051 	 * the error is.
2052 	 */
2053 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
2054 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
2055 	if (bad_lba < start_lba || bad_lba >= end_lba)
2056 		return 0;
2057 
2058 	/*
2059 	 * resid is optional but mostly filled in.  When it's unused,
2060 	 * its value is zero, so we assume the whole buffer transferred
2061 	 */
2062 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
2063 
2064 	/* This computation should always be done in terms of the
2065 	 * resolution of the device's medium.
2066 	 */
2067 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
2068 
2069 	return min(good_bytes, transferred);
2070 }
2071 
2072 /**
2073  *	sd_done - bottom half handler: called when the lower level
2074  *	driver has completed (successfully or otherwise) a scsi command.
2075  *	@SCpnt: mid-level's per command structure.
2076  *
2077  *	Note: potentially run from within an ISR. Must not block.
2078  **/
2079 static int sd_done(struct scsi_cmnd *SCpnt)
2080 {
2081 	int result = SCpnt->result;
2082 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
2083 	unsigned int sector_size = SCpnt->device->sector_size;
2084 	unsigned int resid;
2085 	struct scsi_sense_hdr sshdr;
2086 	struct request *req = scsi_cmd_to_rq(SCpnt);
2087 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
2088 	int sense_valid = 0;
2089 	int sense_deferred = 0;
2090 
2091 	switch (req_op(req)) {
2092 	case REQ_OP_DISCARD:
2093 	case REQ_OP_WRITE_ZEROES:
2094 	case REQ_OP_ZONE_RESET:
2095 	case REQ_OP_ZONE_RESET_ALL:
2096 	case REQ_OP_ZONE_OPEN:
2097 	case REQ_OP_ZONE_CLOSE:
2098 	case REQ_OP_ZONE_FINISH:
2099 		if (!result) {
2100 			good_bytes = blk_rq_bytes(req);
2101 			scsi_set_resid(SCpnt, 0);
2102 		} else {
2103 			good_bytes = 0;
2104 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
2105 		}
2106 		break;
2107 	default:
2108 		/*
2109 		 * In case of bogus fw or device, we could end up having
2110 		 * an unaligned partial completion. Check this here and force
2111 		 * alignment.
2112 		 */
2113 		resid = scsi_get_resid(SCpnt);
2114 		if (resid & (sector_size - 1)) {
2115 			sd_printk(KERN_INFO, sdkp,
2116 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
2117 				resid, sector_size);
2118 			scsi_print_command(SCpnt);
2119 			resid = min(scsi_bufflen(SCpnt),
2120 				    round_up(resid, sector_size));
2121 			scsi_set_resid(SCpnt, resid);
2122 		}
2123 	}
2124 
2125 	if (result) {
2126 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
2127 		if (sense_valid)
2128 			sense_deferred = scsi_sense_is_deferred(&sshdr);
2129 	}
2130 	sdkp->medium_access_timed_out = 0;
2131 
2132 	if (!scsi_status_is_check_condition(result) &&
2133 	    (!sense_valid || sense_deferred))
2134 		goto out;
2135 
2136 	switch (sshdr.sense_key) {
2137 	case HARDWARE_ERROR:
2138 	case MEDIUM_ERROR:
2139 		good_bytes = sd_completed_bytes(SCpnt);
2140 		break;
2141 	case RECOVERED_ERROR:
2142 		good_bytes = scsi_bufflen(SCpnt);
2143 		break;
2144 	case NO_SENSE:
2145 		/* This indicates a false check condition, so ignore it.  An
2146 		 * unknown amount of data was transferred so treat it as an
2147 		 * error.
2148 		 */
2149 		SCpnt->result = 0;
2150 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
2151 		break;
2152 	case ABORTED_COMMAND:
2153 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2154 			good_bytes = sd_completed_bytes(SCpnt);
2155 		break;
2156 	case ILLEGAL_REQUEST:
2157 		switch (sshdr.asc) {
2158 		case 0x10:	/* DIX: Host detected corruption */
2159 			good_bytes = sd_completed_bytes(SCpnt);
2160 			break;
2161 		case 0x20:	/* INVALID COMMAND OPCODE */
2162 		case 0x24:	/* INVALID FIELD IN CDB */
2163 			switch (SCpnt->cmnd[0]) {
2164 			case UNMAP:
2165 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2166 				break;
2167 			case WRITE_SAME_16:
2168 			case WRITE_SAME:
2169 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2170 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2171 				} else {
2172 					sdkp->device->no_write_same = 1;
2173 					sd_config_write_same(sdkp);
2174 					req->rq_flags |= RQF_QUIET;
2175 				}
2176 				break;
2177 			}
2178 		}
2179 		break;
2180 	default:
2181 		break;
2182 	}
2183 
2184  out:
2185 	if (sd_is_zoned(sdkp))
2186 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2187 
2188 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2189 					   "sd_done: completed %d of %d bytes\n",
2190 					   good_bytes, scsi_bufflen(SCpnt)));
2191 
2192 	return good_bytes;
2193 }
2194 
2195 /*
2196  * spinup disk - called only in sd_revalidate_disk()
2197  */
2198 static void
2199 sd_spinup_disk(struct scsi_disk *sdkp)
2200 {
2201 	unsigned char cmd[10];
2202 	unsigned long spintime_expire = 0;
2203 	int retries, spintime;
2204 	unsigned int the_result;
2205 	struct scsi_sense_hdr sshdr;
2206 	const struct scsi_exec_args exec_args = {
2207 		.sshdr = &sshdr,
2208 	};
2209 	int sense_valid = 0;
2210 
2211 	spintime = 0;
2212 
2213 	/* Spin up drives, as required.  Only do this at boot time */
2214 	/* Spinup needs to be done for module loads too. */
2215 	do {
2216 		retries = 0;
2217 
2218 		do {
2219 			bool media_was_present = sdkp->media_present;
2220 
2221 			cmd[0] = TEST_UNIT_READY;
2222 			memset((void *) &cmd[1], 0, 9);
2223 
2224 			the_result = scsi_execute_cmd(sdkp->device, cmd,
2225 						      REQ_OP_DRV_IN, NULL, 0,
2226 						      SD_TIMEOUT,
2227 						      sdkp->max_retries,
2228 						      &exec_args);
2229 
2230 			/*
2231 			 * If the drive has indicated to us that it
2232 			 * doesn't have any media in it, don't bother
2233 			 * with any more polling.
2234 			 */
2235 			if (media_not_present(sdkp, &sshdr)) {
2236 				if (media_was_present)
2237 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2238 				return;
2239 			}
2240 
2241 			if (the_result)
2242 				sense_valid = scsi_sense_valid(&sshdr);
2243 			retries++;
2244 		} while (retries < 3 &&
2245 			 (!scsi_status_is_good(the_result) ||
2246 			  (scsi_status_is_check_condition(the_result) &&
2247 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2248 
2249 		if (!scsi_status_is_check_condition(the_result)) {
2250 			/* no sense, TUR either succeeded or failed
2251 			 * with a status error */
2252 			if(!spintime && !scsi_status_is_good(the_result)) {
2253 				sd_print_result(sdkp, "Test Unit Ready failed",
2254 						the_result);
2255 			}
2256 			break;
2257 		}
2258 
2259 		/*
2260 		 * The device does not want the automatic start to be issued.
2261 		 */
2262 		if (sdkp->device->no_start_on_add)
2263 			break;
2264 
2265 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2266 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2267 				break;	/* manual intervention required */
2268 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2269 				break;	/* standby */
2270 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2271 				break;	/* unavailable */
2272 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2273 				break;	/* sanitize in progress */
2274 			/*
2275 			 * Issue command to spin up drive when not ready
2276 			 */
2277 			if (!spintime) {
2278 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2279 				cmd[0] = START_STOP;
2280 				cmd[1] = 1;	/* Return immediately */
2281 				memset((void *) &cmd[2], 0, 8);
2282 				cmd[4] = 1;	/* Start spin cycle */
2283 				if (sdkp->device->start_stop_pwr_cond)
2284 					cmd[4] |= 1 << 4;
2285 				scsi_execute_cmd(sdkp->device, cmd,
2286 						 REQ_OP_DRV_IN, NULL, 0,
2287 						 SD_TIMEOUT, sdkp->max_retries,
2288 						 &exec_args);
2289 				spintime_expire = jiffies + 100 * HZ;
2290 				spintime = 1;
2291 			}
2292 			/* Wait 1 second for next try */
2293 			msleep(1000);
2294 			printk(KERN_CONT ".");
2295 
2296 		/*
2297 		 * Wait for USB flash devices with slow firmware.
2298 		 * Yes, this sense key/ASC combination shouldn't
2299 		 * occur here.  It's characteristic of these devices.
2300 		 */
2301 		} else if (sense_valid &&
2302 				sshdr.sense_key == UNIT_ATTENTION &&
2303 				sshdr.asc == 0x28) {
2304 			if (!spintime) {
2305 				spintime_expire = jiffies + 5 * HZ;
2306 				spintime = 1;
2307 			}
2308 			/* Wait 1 second for next try */
2309 			msleep(1000);
2310 		} else {
2311 			/* we don't understand the sense code, so it's
2312 			 * probably pointless to loop */
2313 			if(!spintime) {
2314 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2315 				sd_print_sense_hdr(sdkp, &sshdr);
2316 			}
2317 			break;
2318 		}
2319 
2320 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2321 
2322 	if (spintime) {
2323 		if (scsi_status_is_good(the_result))
2324 			printk(KERN_CONT "ready\n");
2325 		else
2326 			printk(KERN_CONT "not responding...\n");
2327 	}
2328 }
2329 
2330 /*
2331  * Determine whether disk supports Data Integrity Field.
2332  */
2333 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2334 {
2335 	struct scsi_device *sdp = sdkp->device;
2336 	u8 type;
2337 
2338 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2339 		sdkp->protection_type = 0;
2340 		return 0;
2341 	}
2342 
2343 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2344 
2345 	if (type > T10_PI_TYPE3_PROTECTION) {
2346 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2347 			  " protection type %u. Disabling disk!\n",
2348 			  type);
2349 		sdkp->protection_type = 0;
2350 		return -ENODEV;
2351 	}
2352 
2353 	sdkp->protection_type = type;
2354 
2355 	return 0;
2356 }
2357 
2358 static void sd_config_protection(struct scsi_disk *sdkp)
2359 {
2360 	struct scsi_device *sdp = sdkp->device;
2361 
2362 	sd_dif_config_host(sdkp);
2363 
2364 	if (!sdkp->protection_type)
2365 		return;
2366 
2367 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2368 		sd_first_printk(KERN_NOTICE, sdkp,
2369 				"Disabling DIF Type %u protection\n",
2370 				sdkp->protection_type);
2371 		sdkp->protection_type = 0;
2372 	}
2373 
2374 	sd_first_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2375 			sdkp->protection_type);
2376 }
2377 
2378 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2379 			struct scsi_sense_hdr *sshdr, int sense_valid,
2380 			int the_result)
2381 {
2382 	if (sense_valid)
2383 		sd_print_sense_hdr(sdkp, sshdr);
2384 	else
2385 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2386 
2387 	/*
2388 	 * Set dirty bit for removable devices if not ready -
2389 	 * sometimes drives will not report this properly.
2390 	 */
2391 	if (sdp->removable &&
2392 	    sense_valid && sshdr->sense_key == NOT_READY)
2393 		set_media_not_present(sdkp);
2394 
2395 	/*
2396 	 * We used to set media_present to 0 here to indicate no media
2397 	 * in the drive, but some drives fail read capacity even with
2398 	 * media present, so we can't do that.
2399 	 */
2400 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2401 }
2402 
2403 #define RC16_LEN 32
2404 #if RC16_LEN > SD_BUF_SIZE
2405 #error RC16_LEN must not be more than SD_BUF_SIZE
2406 #endif
2407 
2408 #define READ_CAPACITY_RETRIES_ON_RESET	10
2409 
2410 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2411 						unsigned char *buffer)
2412 {
2413 	unsigned char cmd[16];
2414 	struct scsi_sense_hdr sshdr;
2415 	const struct scsi_exec_args exec_args = {
2416 		.sshdr = &sshdr,
2417 	};
2418 	int sense_valid = 0;
2419 	int the_result;
2420 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2421 	unsigned int alignment;
2422 	unsigned long long lba;
2423 	unsigned sector_size;
2424 
2425 	if (sdp->no_read_capacity_16)
2426 		return -EINVAL;
2427 
2428 	do {
2429 		memset(cmd, 0, 16);
2430 		cmd[0] = SERVICE_ACTION_IN_16;
2431 		cmd[1] = SAI_READ_CAPACITY_16;
2432 		cmd[13] = RC16_LEN;
2433 		memset(buffer, 0, RC16_LEN);
2434 
2435 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN,
2436 					      buffer, RC16_LEN, SD_TIMEOUT,
2437 					      sdkp->max_retries, &exec_args);
2438 
2439 		if (media_not_present(sdkp, &sshdr))
2440 			return -ENODEV;
2441 
2442 		if (the_result > 0) {
2443 			sense_valid = scsi_sense_valid(&sshdr);
2444 			if (sense_valid &&
2445 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2446 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2447 			    sshdr.ascq == 0x00)
2448 				/* Invalid Command Operation Code or
2449 				 * Invalid Field in CDB, just retry
2450 				 * silently with RC10 */
2451 				return -EINVAL;
2452 			if (sense_valid &&
2453 			    sshdr.sense_key == UNIT_ATTENTION &&
2454 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2455 				/* Device reset might occur several times,
2456 				 * give it one more chance */
2457 				if (--reset_retries > 0)
2458 					continue;
2459 		}
2460 		retries--;
2461 
2462 	} while (the_result && retries);
2463 
2464 	if (the_result) {
2465 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2466 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2467 		return -EINVAL;
2468 	}
2469 
2470 	sector_size = get_unaligned_be32(&buffer[8]);
2471 	lba = get_unaligned_be64(&buffer[0]);
2472 
2473 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2474 		sdkp->capacity = 0;
2475 		return -ENODEV;
2476 	}
2477 
2478 	/* Logical blocks per physical block exponent */
2479 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2480 
2481 	/* RC basis */
2482 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2483 
2484 	/* Lowest aligned logical block */
2485 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2486 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2487 	if (alignment && sdkp->first_scan)
2488 		sd_printk(KERN_NOTICE, sdkp,
2489 			  "physical block alignment offset: %u\n", alignment);
2490 
2491 	if (buffer[14] & 0x80) { /* LBPME */
2492 		sdkp->lbpme = 1;
2493 
2494 		if (buffer[14] & 0x40) /* LBPRZ */
2495 			sdkp->lbprz = 1;
2496 
2497 		sd_config_discard(sdkp, SD_LBP_WS16);
2498 	}
2499 
2500 	sdkp->capacity = lba + 1;
2501 	return sector_size;
2502 }
2503 
2504 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2505 						unsigned char *buffer)
2506 {
2507 	unsigned char cmd[16];
2508 	struct scsi_sense_hdr sshdr;
2509 	const struct scsi_exec_args exec_args = {
2510 		.sshdr = &sshdr,
2511 	};
2512 	int sense_valid = 0;
2513 	int the_result;
2514 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2515 	sector_t lba;
2516 	unsigned sector_size;
2517 
2518 	do {
2519 		cmd[0] = READ_CAPACITY;
2520 		memset(&cmd[1], 0, 9);
2521 		memset(buffer, 0, 8);
2522 
2523 		the_result = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, buffer,
2524 					      8, SD_TIMEOUT, sdkp->max_retries,
2525 					      &exec_args);
2526 
2527 		if (media_not_present(sdkp, &sshdr))
2528 			return -ENODEV;
2529 
2530 		if (the_result > 0) {
2531 			sense_valid = scsi_sense_valid(&sshdr);
2532 			if (sense_valid &&
2533 			    sshdr.sense_key == UNIT_ATTENTION &&
2534 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2535 				/* Device reset might occur several times,
2536 				 * give it one more chance */
2537 				if (--reset_retries > 0)
2538 					continue;
2539 		}
2540 		retries--;
2541 
2542 	} while (the_result && retries);
2543 
2544 	if (the_result) {
2545 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2546 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2547 		return -EINVAL;
2548 	}
2549 
2550 	sector_size = get_unaligned_be32(&buffer[4]);
2551 	lba = get_unaligned_be32(&buffer[0]);
2552 
2553 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2554 		/* Some buggy (usb cardreader) devices return an lba of
2555 		   0xffffffff when the want to report a size of 0 (with
2556 		   which they really mean no media is present) */
2557 		sdkp->capacity = 0;
2558 		sdkp->physical_block_size = sector_size;
2559 		return sector_size;
2560 	}
2561 
2562 	sdkp->capacity = lba + 1;
2563 	sdkp->physical_block_size = sector_size;
2564 	return sector_size;
2565 }
2566 
2567 static int sd_try_rc16_first(struct scsi_device *sdp)
2568 {
2569 	if (sdp->host->max_cmd_len < 16)
2570 		return 0;
2571 	if (sdp->try_rc_10_first)
2572 		return 0;
2573 	if (sdp->scsi_level > SCSI_SPC_2)
2574 		return 1;
2575 	if (scsi_device_protection(sdp))
2576 		return 1;
2577 	return 0;
2578 }
2579 
2580 /*
2581  * read disk capacity
2582  */
2583 static void
2584 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2585 {
2586 	int sector_size;
2587 	struct scsi_device *sdp = sdkp->device;
2588 
2589 	if (sd_try_rc16_first(sdp)) {
2590 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2591 		if (sector_size == -EOVERFLOW)
2592 			goto got_data;
2593 		if (sector_size == -ENODEV)
2594 			return;
2595 		if (sector_size < 0)
2596 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2597 		if (sector_size < 0)
2598 			return;
2599 	} else {
2600 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2601 		if (sector_size == -EOVERFLOW)
2602 			goto got_data;
2603 		if (sector_size < 0)
2604 			return;
2605 		if ((sizeof(sdkp->capacity) > 4) &&
2606 		    (sdkp->capacity > 0xffffffffULL)) {
2607 			int old_sector_size = sector_size;
2608 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2609 					"Trying to use READ CAPACITY(16).\n");
2610 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2611 			if (sector_size < 0) {
2612 				sd_printk(KERN_NOTICE, sdkp,
2613 					"Using 0xffffffff as device size\n");
2614 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2615 				sector_size = old_sector_size;
2616 				goto got_data;
2617 			}
2618 			/* Remember that READ CAPACITY(16) succeeded */
2619 			sdp->try_rc_10_first = 0;
2620 		}
2621 	}
2622 
2623 	/* Some devices are known to return the total number of blocks,
2624 	 * not the highest block number.  Some devices have versions
2625 	 * which do this and others which do not.  Some devices we might
2626 	 * suspect of doing this but we don't know for certain.
2627 	 *
2628 	 * If we know the reported capacity is wrong, decrement it.  If
2629 	 * we can only guess, then assume the number of blocks is even
2630 	 * (usually true but not always) and err on the side of lowering
2631 	 * the capacity.
2632 	 */
2633 	if (sdp->fix_capacity ||
2634 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2635 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2636 				"from its reported value: %llu\n",
2637 				(unsigned long long) sdkp->capacity);
2638 		--sdkp->capacity;
2639 	}
2640 
2641 got_data:
2642 	if (sector_size == 0) {
2643 		sector_size = 512;
2644 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2645 			  "assuming 512.\n");
2646 	}
2647 
2648 	if (sector_size != 512 &&
2649 	    sector_size != 1024 &&
2650 	    sector_size != 2048 &&
2651 	    sector_size != 4096) {
2652 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2653 			  sector_size);
2654 		/*
2655 		 * The user might want to re-format the drive with
2656 		 * a supported sectorsize.  Once this happens, it
2657 		 * would be relatively trivial to set the thing up.
2658 		 * For this reason, we leave the thing in the table.
2659 		 */
2660 		sdkp->capacity = 0;
2661 		/*
2662 		 * set a bogus sector size so the normal read/write
2663 		 * logic in the block layer will eventually refuse any
2664 		 * request on this device without tripping over power
2665 		 * of two sector size assumptions
2666 		 */
2667 		sector_size = 512;
2668 	}
2669 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2670 	blk_queue_physical_block_size(sdp->request_queue,
2671 				      sdkp->physical_block_size);
2672 	sdkp->device->sector_size = sector_size;
2673 
2674 	if (sdkp->capacity > 0xffffffff)
2675 		sdp->use_16_for_rw = 1;
2676 
2677 }
2678 
2679 /*
2680  * Print disk capacity
2681  */
2682 static void
2683 sd_print_capacity(struct scsi_disk *sdkp,
2684 		  sector_t old_capacity)
2685 {
2686 	int sector_size = sdkp->device->sector_size;
2687 	char cap_str_2[10], cap_str_10[10];
2688 
2689 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2690 		return;
2691 
2692 	string_get_size(sdkp->capacity, sector_size,
2693 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2694 	string_get_size(sdkp->capacity, sector_size,
2695 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2696 
2697 	sd_printk(KERN_NOTICE, sdkp,
2698 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2699 		  (unsigned long long)sdkp->capacity,
2700 		  sector_size, cap_str_10, cap_str_2);
2701 
2702 	if (sdkp->physical_block_size != sector_size)
2703 		sd_printk(KERN_NOTICE, sdkp,
2704 			  "%u-byte physical blocks\n",
2705 			  sdkp->physical_block_size);
2706 }
2707 
2708 /* called with buffer of length 512 */
2709 static inline int
2710 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2711 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2712 		 struct scsi_sense_hdr *sshdr)
2713 {
2714 	/*
2715 	 * If we must use MODE SENSE(10), make sure that the buffer length
2716 	 * is at least 8 bytes so that the mode sense header fits.
2717 	 */
2718 	if (sdkp->device->use_10_for_ms && len < 8)
2719 		len = 8;
2720 
2721 	return scsi_mode_sense(sdkp->device, dbd, modepage, 0, buffer, len,
2722 			       SD_TIMEOUT, sdkp->max_retries, data, sshdr);
2723 }
2724 
2725 /*
2726  * read write protect setting, if possible - called only in sd_revalidate_disk()
2727  * called with buffer of length SD_BUF_SIZE
2728  */
2729 static void
2730 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2731 {
2732 	int res;
2733 	struct scsi_device *sdp = sdkp->device;
2734 	struct scsi_mode_data data;
2735 	int old_wp = sdkp->write_prot;
2736 
2737 	set_disk_ro(sdkp->disk, 0);
2738 	if (sdp->skip_ms_page_3f) {
2739 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2740 		return;
2741 	}
2742 
2743 	if (sdp->use_192_bytes_for_3f) {
2744 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2745 	} else {
2746 		/*
2747 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2748 		 * We have to start carefully: some devices hang if we ask
2749 		 * for more than is available.
2750 		 */
2751 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2752 
2753 		/*
2754 		 * Second attempt: ask for page 0 When only page 0 is
2755 		 * implemented, a request for page 3F may return Sense Key
2756 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2757 		 * CDB.
2758 		 */
2759 		if (res < 0)
2760 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2761 
2762 		/*
2763 		 * Third attempt: ask 255 bytes, as we did earlier.
2764 		 */
2765 		if (res < 0)
2766 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2767 					       &data, NULL);
2768 	}
2769 
2770 	if (res < 0) {
2771 		sd_first_printk(KERN_WARNING, sdkp,
2772 			  "Test WP failed, assume Write Enabled\n");
2773 	} else {
2774 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2775 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2776 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2777 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2778 				  sdkp->write_prot ? "on" : "off");
2779 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2780 		}
2781 	}
2782 }
2783 
2784 /*
2785  * sd_read_cache_type - called only from sd_revalidate_disk()
2786  * called with buffer of length SD_BUF_SIZE
2787  */
2788 static void
2789 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2790 {
2791 	int len = 0, res;
2792 	struct scsi_device *sdp = sdkp->device;
2793 
2794 	int dbd;
2795 	int modepage;
2796 	int first_len;
2797 	struct scsi_mode_data data;
2798 	struct scsi_sense_hdr sshdr;
2799 	int old_wce = sdkp->WCE;
2800 	int old_rcd = sdkp->RCD;
2801 	int old_dpofua = sdkp->DPOFUA;
2802 
2803 
2804 	if (sdkp->cache_override)
2805 		return;
2806 
2807 	first_len = 4;
2808 	if (sdp->skip_ms_page_8) {
2809 		if (sdp->type == TYPE_RBC)
2810 			goto defaults;
2811 		else {
2812 			if (sdp->skip_ms_page_3f)
2813 				goto defaults;
2814 			modepage = 0x3F;
2815 			if (sdp->use_192_bytes_for_3f)
2816 				first_len = 192;
2817 			dbd = 0;
2818 		}
2819 	} else if (sdp->type == TYPE_RBC) {
2820 		modepage = 6;
2821 		dbd = 8;
2822 	} else {
2823 		modepage = 8;
2824 		dbd = 0;
2825 	}
2826 
2827 	/* cautiously ask */
2828 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2829 			&data, &sshdr);
2830 
2831 	if (res < 0)
2832 		goto bad_sense;
2833 
2834 	if (!data.header_length) {
2835 		modepage = 6;
2836 		first_len = 0;
2837 		sd_first_printk(KERN_ERR, sdkp,
2838 				"Missing header in MODE_SENSE response\n");
2839 	}
2840 
2841 	/* that went OK, now ask for the proper length */
2842 	len = data.length;
2843 
2844 	/*
2845 	 * We're only interested in the first three bytes, actually.
2846 	 * But the data cache page is defined for the first 20.
2847 	 */
2848 	if (len < 3)
2849 		goto bad_sense;
2850 	else if (len > SD_BUF_SIZE) {
2851 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2852 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2853 		len = SD_BUF_SIZE;
2854 	}
2855 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2856 		len = 192;
2857 
2858 	/* Get the data */
2859 	if (len > first_len)
2860 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2861 				&data, &sshdr);
2862 
2863 	if (!res) {
2864 		int offset = data.header_length + data.block_descriptor_length;
2865 
2866 		while (offset < len) {
2867 			u8 page_code = buffer[offset] & 0x3F;
2868 			u8 spf       = buffer[offset] & 0x40;
2869 
2870 			if (page_code == 8 || page_code == 6) {
2871 				/* We're interested only in the first 3 bytes.
2872 				 */
2873 				if (len - offset <= 2) {
2874 					sd_first_printk(KERN_ERR, sdkp,
2875 						"Incomplete mode parameter "
2876 							"data\n");
2877 					goto defaults;
2878 				} else {
2879 					modepage = page_code;
2880 					goto Page_found;
2881 				}
2882 			} else {
2883 				/* Go to the next page */
2884 				if (spf && len - offset > 3)
2885 					offset += 4 + (buffer[offset+2] << 8) +
2886 						buffer[offset+3];
2887 				else if (!spf && len - offset > 1)
2888 					offset += 2 + buffer[offset+1];
2889 				else {
2890 					sd_first_printk(KERN_ERR, sdkp,
2891 							"Incomplete mode "
2892 							"parameter data\n");
2893 					goto defaults;
2894 				}
2895 			}
2896 		}
2897 
2898 		sd_first_printk(KERN_WARNING, sdkp,
2899 				"No Caching mode page found\n");
2900 		goto defaults;
2901 
2902 	Page_found:
2903 		if (modepage == 8) {
2904 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2905 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2906 		} else {
2907 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2908 			sdkp->RCD = 0;
2909 		}
2910 
2911 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2912 		if (sdp->broken_fua) {
2913 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2914 			sdkp->DPOFUA = 0;
2915 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2916 			   !sdkp->device->use_16_for_rw) {
2917 			sd_first_printk(KERN_NOTICE, sdkp,
2918 				  "Uses READ/WRITE(6), disabling FUA\n");
2919 			sdkp->DPOFUA = 0;
2920 		}
2921 
2922 		/* No cache flush allowed for write protected devices */
2923 		if (sdkp->WCE && sdkp->write_prot)
2924 			sdkp->WCE = 0;
2925 
2926 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2927 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2928 			sd_printk(KERN_NOTICE, sdkp,
2929 				  "Write cache: %s, read cache: %s, %s\n",
2930 				  sdkp->WCE ? "enabled" : "disabled",
2931 				  sdkp->RCD ? "disabled" : "enabled",
2932 				  sdkp->DPOFUA ? "supports DPO and FUA"
2933 				  : "doesn't support DPO or FUA");
2934 
2935 		return;
2936 	}
2937 
2938 bad_sense:
2939 	if (scsi_sense_valid(&sshdr) &&
2940 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2941 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2942 		/* Invalid field in CDB */
2943 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2944 	else
2945 		sd_first_printk(KERN_ERR, sdkp,
2946 				"Asking for cache data failed\n");
2947 
2948 defaults:
2949 	if (sdp->wce_default_on) {
2950 		sd_first_printk(KERN_NOTICE, sdkp,
2951 				"Assuming drive cache: write back\n");
2952 		sdkp->WCE = 1;
2953 	} else {
2954 		sd_first_printk(KERN_WARNING, sdkp,
2955 				"Assuming drive cache: write through\n");
2956 		sdkp->WCE = 0;
2957 	}
2958 	sdkp->RCD = 0;
2959 	sdkp->DPOFUA = 0;
2960 }
2961 
2962 /*
2963  * The ATO bit indicates whether the DIF application tag is available
2964  * for use by the operating system.
2965  */
2966 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2967 {
2968 	int res, offset;
2969 	struct scsi_device *sdp = sdkp->device;
2970 	struct scsi_mode_data data;
2971 	struct scsi_sense_hdr sshdr;
2972 
2973 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2974 		return;
2975 
2976 	if (sdkp->protection_type == 0)
2977 		return;
2978 
2979 	res = scsi_mode_sense(sdp, 1, 0x0a, 0, buffer, 36, SD_TIMEOUT,
2980 			      sdkp->max_retries, &data, &sshdr);
2981 
2982 	if (res < 0 || !data.header_length ||
2983 	    data.length < 6) {
2984 		sd_first_printk(KERN_WARNING, sdkp,
2985 			  "getting Control mode page failed, assume no ATO\n");
2986 
2987 		if (scsi_sense_valid(&sshdr))
2988 			sd_print_sense_hdr(sdkp, &sshdr);
2989 
2990 		return;
2991 	}
2992 
2993 	offset = data.header_length + data.block_descriptor_length;
2994 
2995 	if ((buffer[offset] & 0x3f) != 0x0a) {
2996 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2997 		return;
2998 	}
2999 
3000 	if ((buffer[offset + 5] & 0x80) == 0)
3001 		return;
3002 
3003 	sdkp->ATO = 1;
3004 
3005 	return;
3006 }
3007 
3008 /**
3009  * sd_read_block_limits - Query disk device for preferred I/O sizes.
3010  * @sdkp: disk to query
3011  */
3012 static void sd_read_block_limits(struct scsi_disk *sdkp)
3013 {
3014 	struct scsi_vpd *vpd;
3015 
3016 	rcu_read_lock();
3017 
3018 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
3019 	if (!vpd || vpd->len < 16)
3020 		goto out;
3021 
3022 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
3023 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
3024 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
3025 
3026 	if (vpd->len >= 64) {
3027 		unsigned int lba_count, desc_count;
3028 
3029 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
3030 
3031 		if (!sdkp->lbpme)
3032 			goto out;
3033 
3034 		lba_count = get_unaligned_be32(&vpd->data[20]);
3035 		desc_count = get_unaligned_be32(&vpd->data[24]);
3036 
3037 		if (lba_count && desc_count)
3038 			sdkp->max_unmap_blocks = lba_count;
3039 
3040 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
3041 
3042 		if (vpd->data[32] & 0x80)
3043 			sdkp->unmap_alignment =
3044 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
3045 
3046 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
3047 
3048 			if (sdkp->max_unmap_blocks)
3049 				sd_config_discard(sdkp, SD_LBP_UNMAP);
3050 			else
3051 				sd_config_discard(sdkp, SD_LBP_WS16);
3052 
3053 		} else {	/* LBP VPD page tells us what to use */
3054 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
3055 				sd_config_discard(sdkp, SD_LBP_UNMAP);
3056 			else if (sdkp->lbpws)
3057 				sd_config_discard(sdkp, SD_LBP_WS16);
3058 			else if (sdkp->lbpws10)
3059 				sd_config_discard(sdkp, SD_LBP_WS10);
3060 			else
3061 				sd_config_discard(sdkp, SD_LBP_DISABLE);
3062 		}
3063 	}
3064 
3065  out:
3066 	rcu_read_unlock();
3067 }
3068 
3069 /**
3070  * sd_read_block_characteristics - Query block dev. characteristics
3071  * @sdkp: disk to query
3072  */
3073 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
3074 {
3075 	struct request_queue *q = sdkp->disk->queue;
3076 	struct scsi_vpd *vpd;
3077 	u16 rot;
3078 	u8 zoned;
3079 
3080 	rcu_read_lock();
3081 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
3082 
3083 	if (!vpd || vpd->len < 8) {
3084 		rcu_read_unlock();
3085 	        return;
3086 	}
3087 
3088 	rot = get_unaligned_be16(&vpd->data[4]);
3089 	zoned = (vpd->data[8] >> 4) & 3;
3090 	rcu_read_unlock();
3091 
3092 	if (rot == 1) {
3093 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
3094 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
3095 	}
3096 
3097 	if (sdkp->device->type == TYPE_ZBC) {
3098 		/*
3099 		 * Host-managed: Per ZBC and ZAC specifications, writes in
3100 		 * sequential write required zones of host-managed devices must
3101 		 * be aligned to the device physical block size.
3102 		 */
3103 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
3104 		blk_queue_zone_write_granularity(q, sdkp->physical_block_size);
3105 	} else {
3106 		sdkp->zoned = zoned;
3107 		if (sdkp->zoned == 1) {
3108 			/* Host-aware */
3109 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
3110 		} else {
3111 			/* Regular disk or drive managed disk */
3112 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
3113 		}
3114 	}
3115 
3116 	if (!sdkp->first_scan)
3117 		return;
3118 
3119 	if (blk_queue_is_zoned(q)) {
3120 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
3121 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
3122 	} else {
3123 		if (sdkp->zoned == 1)
3124 			sd_printk(KERN_NOTICE, sdkp,
3125 				  "Host-aware SMR disk used as regular disk\n");
3126 		else if (sdkp->zoned == 2)
3127 			sd_printk(KERN_NOTICE, sdkp,
3128 				  "Drive-managed SMR disk\n");
3129 	}
3130 }
3131 
3132 /**
3133  * sd_read_block_provisioning - Query provisioning VPD page
3134  * @sdkp: disk to query
3135  */
3136 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
3137 {
3138 	struct scsi_vpd *vpd;
3139 
3140 	if (sdkp->lbpme == 0)
3141 		return;
3142 
3143 	rcu_read_lock();
3144 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
3145 
3146 	if (!vpd || vpd->len < 8) {
3147 		rcu_read_unlock();
3148 		return;
3149 	}
3150 
3151 	sdkp->lbpvpd	= 1;
3152 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
3153 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
3154 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
3155 	rcu_read_unlock();
3156 }
3157 
3158 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
3159 {
3160 	struct scsi_device *sdev = sdkp->device;
3161 
3162 	if (sdev->host->no_write_same) {
3163 		sdev->no_write_same = 1;
3164 
3165 		return;
3166 	}
3167 
3168 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY, 0) < 0) {
3169 		struct scsi_vpd *vpd;
3170 
3171 		sdev->no_report_opcodes = 1;
3172 
3173 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3174 		 * CODES is unsupported and the device has an ATA
3175 		 * Information VPD page (SAT).
3176 		 */
3177 		rcu_read_lock();
3178 		vpd = rcu_dereference(sdev->vpd_pg89);
3179 		if (vpd)
3180 			sdev->no_write_same = 1;
3181 		rcu_read_unlock();
3182 	}
3183 
3184 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16, 0) == 1)
3185 		sdkp->ws16 = 1;
3186 
3187 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME, 0) == 1)
3188 		sdkp->ws10 = 1;
3189 }
3190 
3191 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3192 {
3193 	struct scsi_device *sdev = sdkp->device;
3194 
3195 	if (!sdev->security_supported)
3196 		return;
3197 
3198 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3199 			SECURITY_PROTOCOL_IN, 0) == 1 &&
3200 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3201 			SECURITY_PROTOCOL_OUT, 0) == 1)
3202 		sdkp->security = 1;
3203 }
3204 
3205 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3206 {
3207 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3208 }
3209 
3210 /**
3211  * sd_read_cpr - Query concurrent positioning ranges
3212  * @sdkp:	disk to query
3213  */
3214 static void sd_read_cpr(struct scsi_disk *sdkp)
3215 {
3216 	struct blk_independent_access_ranges *iars = NULL;
3217 	unsigned char *buffer = NULL;
3218 	unsigned int nr_cpr = 0;
3219 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3220 	u8 *desc;
3221 
3222 	/*
3223 	 * We need to have the capacity set first for the block layer to be
3224 	 * able to check the ranges.
3225 	 */
3226 	if (sdkp->first_scan)
3227 		return;
3228 
3229 	if (!sdkp->capacity)
3230 		goto out;
3231 
3232 	/*
3233 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3234 	 * leading to a maximum page size of 64 + 256*32 bytes.
3235 	 */
3236 	buf_len = 64 + 256*32;
3237 	buffer = kmalloc(buf_len, GFP_KERNEL);
3238 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3239 		goto out;
3240 
3241 	/* We must have at least a 64B header and one 32B range descriptor */
3242 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3243 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3244 		sd_printk(KERN_ERR, sdkp,
3245 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3246 		goto out;
3247 	}
3248 
3249 	nr_cpr = (vpd_len - 64) / 32;
3250 	if (nr_cpr == 1) {
3251 		nr_cpr = 0;
3252 		goto out;
3253 	}
3254 
3255 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3256 	if (!iars) {
3257 		nr_cpr = 0;
3258 		goto out;
3259 	}
3260 
3261 	desc = &buffer[64];
3262 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3263 		if (desc[0] != i) {
3264 			sd_printk(KERN_ERR, sdkp,
3265 				"Invalid Concurrent Positioning Range number\n");
3266 			nr_cpr = 0;
3267 			break;
3268 		}
3269 
3270 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3271 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3272 	}
3273 
3274 out:
3275 	disk_set_independent_access_ranges(sdkp->disk, iars);
3276 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3277 		sd_printk(KERN_NOTICE, sdkp,
3278 			  "%u concurrent positioning ranges\n", nr_cpr);
3279 		sdkp->nr_actuators = nr_cpr;
3280 	}
3281 
3282 	kfree(buffer);
3283 }
3284 
3285 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3286 {
3287 	struct scsi_device *sdp = sdkp->device;
3288 	unsigned int min_xfer_bytes =
3289 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3290 
3291 	if (sdkp->min_xfer_blocks == 0)
3292 		return false;
3293 
3294 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3295 		sd_first_printk(KERN_WARNING, sdkp,
3296 				"Preferred minimum I/O size %u bytes not a " \
3297 				"multiple of physical block size (%u bytes)\n",
3298 				min_xfer_bytes, sdkp->physical_block_size);
3299 		sdkp->min_xfer_blocks = 0;
3300 		return false;
3301 	}
3302 
3303 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3304 			min_xfer_bytes);
3305 	return true;
3306 }
3307 
3308 /*
3309  * Determine the device's preferred I/O size for reads and writes
3310  * unless the reported value is unreasonably small, large, not a
3311  * multiple of the physical block size, or simply garbage.
3312  */
3313 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3314 				      unsigned int dev_max)
3315 {
3316 	struct scsi_device *sdp = sdkp->device;
3317 	unsigned int opt_xfer_bytes =
3318 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3319 	unsigned int min_xfer_bytes =
3320 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3321 
3322 	if (sdkp->opt_xfer_blocks == 0)
3323 		return false;
3324 
3325 	if (sdkp->opt_xfer_blocks > dev_max) {
3326 		sd_first_printk(KERN_WARNING, sdkp,
3327 				"Optimal transfer size %u logical blocks " \
3328 				"> dev_max (%u logical blocks)\n",
3329 				sdkp->opt_xfer_blocks, dev_max);
3330 		return false;
3331 	}
3332 
3333 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3334 		sd_first_printk(KERN_WARNING, sdkp,
3335 				"Optimal transfer size %u logical blocks " \
3336 				"> sd driver limit (%u logical blocks)\n",
3337 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3338 		return false;
3339 	}
3340 
3341 	if (opt_xfer_bytes < PAGE_SIZE) {
3342 		sd_first_printk(KERN_WARNING, sdkp,
3343 				"Optimal transfer size %u bytes < " \
3344 				"PAGE_SIZE (%u bytes)\n",
3345 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3346 		return false;
3347 	}
3348 
3349 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3350 		sd_first_printk(KERN_WARNING, sdkp,
3351 				"Optimal transfer size %u bytes not a " \
3352 				"multiple of preferred minimum block " \
3353 				"size (%u bytes)\n",
3354 				opt_xfer_bytes, min_xfer_bytes);
3355 		return false;
3356 	}
3357 
3358 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3359 		sd_first_printk(KERN_WARNING, sdkp,
3360 				"Optimal transfer size %u bytes not a " \
3361 				"multiple of physical block size (%u bytes)\n",
3362 				opt_xfer_bytes, sdkp->physical_block_size);
3363 		return false;
3364 	}
3365 
3366 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3367 			opt_xfer_bytes);
3368 	return true;
3369 }
3370 
3371 /**
3372  *	sd_revalidate_disk - called the first time a new disk is seen,
3373  *	performs disk spin up, read_capacity, etc.
3374  *	@disk: struct gendisk we care about
3375  **/
3376 static int sd_revalidate_disk(struct gendisk *disk)
3377 {
3378 	struct scsi_disk *sdkp = scsi_disk(disk);
3379 	struct scsi_device *sdp = sdkp->device;
3380 	struct request_queue *q = sdkp->disk->queue;
3381 	sector_t old_capacity = sdkp->capacity;
3382 	unsigned char *buffer;
3383 	unsigned int dev_max, rw_max;
3384 
3385 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3386 				      "sd_revalidate_disk\n"));
3387 
3388 	/*
3389 	 * If the device is offline, don't try and read capacity or any
3390 	 * of the other niceties.
3391 	 */
3392 	if (!scsi_device_online(sdp))
3393 		goto out;
3394 
3395 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3396 	if (!buffer) {
3397 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3398 			  "allocation failure.\n");
3399 		goto out;
3400 	}
3401 
3402 	sd_spinup_disk(sdkp);
3403 
3404 	/*
3405 	 * Without media there is no reason to ask; moreover, some devices
3406 	 * react badly if we do.
3407 	 */
3408 	if (sdkp->media_present) {
3409 		sd_read_capacity(sdkp, buffer);
3410 
3411 		/*
3412 		 * set the default to rotational.  All non-rotational devices
3413 		 * support the block characteristics VPD page, which will
3414 		 * cause this to be updated correctly and any device which
3415 		 * doesn't support it should be treated as rotational.
3416 		 */
3417 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3418 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3419 
3420 		if (scsi_device_supports_vpd(sdp)) {
3421 			sd_read_block_provisioning(sdkp);
3422 			sd_read_block_limits(sdkp);
3423 			sd_read_block_characteristics(sdkp);
3424 			sd_zbc_read_zones(sdkp, buffer);
3425 			sd_read_cpr(sdkp);
3426 		}
3427 
3428 		sd_print_capacity(sdkp, old_capacity);
3429 
3430 		sd_read_write_protect_flag(sdkp, buffer);
3431 		sd_read_cache_type(sdkp, buffer);
3432 		sd_read_app_tag_own(sdkp, buffer);
3433 		sd_read_write_same(sdkp, buffer);
3434 		sd_read_security(sdkp, buffer);
3435 		sd_config_protection(sdkp);
3436 	}
3437 
3438 	/*
3439 	 * We now have all cache related info, determine how we deal
3440 	 * with flush requests.
3441 	 */
3442 	sd_set_flush_flag(sdkp);
3443 
3444 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3445 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3446 
3447 	/* Some devices report a maximum block count for READ/WRITE requests. */
3448 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3449 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3450 
3451 	if (sd_validate_min_xfer_size(sdkp))
3452 		blk_queue_io_min(sdkp->disk->queue,
3453 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3454 	else
3455 		blk_queue_io_min(sdkp->disk->queue, 0);
3456 
3457 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3458 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3459 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3460 	} else {
3461 		q->limits.io_opt = 0;
3462 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3463 				      (sector_t)BLK_DEF_MAX_SECTORS);
3464 	}
3465 
3466 	/*
3467 	 * Limit default to SCSI host optimal sector limit if set. There may be
3468 	 * an impact on performance for when the size of a request exceeds this
3469 	 * host limit.
3470 	 */
3471 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3472 
3473 	/* Do not exceed controller limit */
3474 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3475 
3476 	/*
3477 	 * Only update max_sectors if previously unset or if the current value
3478 	 * exceeds the capabilities of the hardware.
3479 	 */
3480 	if (sdkp->first_scan ||
3481 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3482 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3483 		q->limits.max_sectors = rw_max;
3484 
3485 	sdkp->first_scan = 0;
3486 
3487 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3488 	sd_config_write_same(sdkp);
3489 	kfree(buffer);
3490 
3491 	/*
3492 	 * For a zoned drive, revalidating the zones can be done only once
3493 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3494 	 * capacity to 0.
3495 	 */
3496 	if (sd_zbc_revalidate_zones(sdkp))
3497 		set_capacity_and_notify(disk, 0);
3498 
3499  out:
3500 	return 0;
3501 }
3502 
3503 /**
3504  *	sd_unlock_native_capacity - unlock native capacity
3505  *	@disk: struct gendisk to set capacity for
3506  *
3507  *	Block layer calls this function if it detects that partitions
3508  *	on @disk reach beyond the end of the device.  If the SCSI host
3509  *	implements ->unlock_native_capacity() method, it's invoked to
3510  *	give it a chance to adjust the device capacity.
3511  *
3512  *	CONTEXT:
3513  *	Defined by block layer.  Might sleep.
3514  */
3515 static void sd_unlock_native_capacity(struct gendisk *disk)
3516 {
3517 	struct scsi_device *sdev = scsi_disk(disk)->device;
3518 
3519 	if (sdev->host->hostt->unlock_native_capacity)
3520 		sdev->host->hostt->unlock_native_capacity(sdev);
3521 }
3522 
3523 /**
3524  *	sd_format_disk_name - format disk name
3525  *	@prefix: name prefix - ie. "sd" for SCSI disks
3526  *	@index: index of the disk to format name for
3527  *	@buf: output buffer
3528  *	@buflen: length of the output buffer
3529  *
3530  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3531  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3532  *	which is followed by sdaaa.
3533  *
3534  *	This is basically 26 base counting with one extra 'nil' entry
3535  *	at the beginning from the second digit on and can be
3536  *	determined using similar method as 26 base conversion with the
3537  *	index shifted -1 after each digit is computed.
3538  *
3539  *	CONTEXT:
3540  *	Don't care.
3541  *
3542  *	RETURNS:
3543  *	0 on success, -errno on failure.
3544  */
3545 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3546 {
3547 	const int base = 'z' - 'a' + 1;
3548 	char *begin = buf + strlen(prefix);
3549 	char *end = buf + buflen;
3550 	char *p;
3551 	int unit;
3552 
3553 	p = end - 1;
3554 	*p = '\0';
3555 	unit = base;
3556 	do {
3557 		if (p == begin)
3558 			return -EINVAL;
3559 		*--p = 'a' + (index % unit);
3560 		index = (index / unit) - 1;
3561 	} while (index >= 0);
3562 
3563 	memmove(begin, p, end - p);
3564 	memcpy(buf, prefix, strlen(prefix));
3565 
3566 	return 0;
3567 }
3568 
3569 /**
3570  *	sd_probe - called during driver initialization and whenever a
3571  *	new scsi device is attached to the system. It is called once
3572  *	for each scsi device (not just disks) present.
3573  *	@dev: pointer to device object
3574  *
3575  *	Returns 0 if successful (or not interested in this scsi device
3576  *	(e.g. scanner)); 1 when there is an error.
3577  *
3578  *	Note: this function is invoked from the scsi mid-level.
3579  *	This function sets up the mapping between a given
3580  *	<host,channel,id,lun> (found in sdp) and new device name
3581  *	(e.g. /dev/sda). More precisely it is the block device major
3582  *	and minor number that is chosen here.
3583  *
3584  *	Assume sd_probe is not re-entrant (for time being)
3585  *	Also think about sd_probe() and sd_remove() running coincidentally.
3586  **/
3587 static int sd_probe(struct device *dev)
3588 {
3589 	struct scsi_device *sdp = to_scsi_device(dev);
3590 	struct scsi_disk *sdkp;
3591 	struct gendisk *gd;
3592 	int index;
3593 	int error;
3594 
3595 	scsi_autopm_get_device(sdp);
3596 	error = -ENODEV;
3597 	if (sdp->type != TYPE_DISK &&
3598 	    sdp->type != TYPE_ZBC &&
3599 	    sdp->type != TYPE_MOD &&
3600 	    sdp->type != TYPE_RBC)
3601 		goto out;
3602 
3603 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3604 		sdev_printk(KERN_WARNING, sdp,
3605 			    "Unsupported ZBC host-managed device.\n");
3606 		goto out;
3607 	}
3608 
3609 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3610 					"sd_probe\n"));
3611 
3612 	error = -ENOMEM;
3613 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3614 	if (!sdkp)
3615 		goto out;
3616 
3617 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3618 					 &sd_bio_compl_lkclass);
3619 	if (!gd)
3620 		goto out_free;
3621 
3622 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3623 	if (index < 0) {
3624 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3625 		goto out_put;
3626 	}
3627 
3628 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3629 	if (error) {
3630 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3631 		goto out_free_index;
3632 	}
3633 
3634 	sdkp->device = sdp;
3635 	sdkp->disk = gd;
3636 	sdkp->index = index;
3637 	sdkp->max_retries = SD_MAX_RETRIES;
3638 	atomic_set(&sdkp->openers, 0);
3639 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3640 
3641 	if (!sdp->request_queue->rq_timeout) {
3642 		if (sdp->type != TYPE_MOD)
3643 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3644 		else
3645 			blk_queue_rq_timeout(sdp->request_queue,
3646 					     SD_MOD_TIMEOUT);
3647 	}
3648 
3649 	device_initialize(&sdkp->disk_dev);
3650 	sdkp->disk_dev.parent = get_device(dev);
3651 	sdkp->disk_dev.class = &sd_disk_class;
3652 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3653 
3654 	error = device_add(&sdkp->disk_dev);
3655 	if (error) {
3656 		put_device(&sdkp->disk_dev);
3657 		goto out;
3658 	}
3659 
3660 	dev_set_drvdata(dev, sdkp);
3661 
3662 	gd->major = sd_major((index & 0xf0) >> 4);
3663 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3664 	gd->minors = SD_MINORS;
3665 
3666 	gd->fops = &sd_fops;
3667 	gd->private_data = sdkp;
3668 
3669 	/* defaults, until the device tells us otherwise */
3670 	sdp->sector_size = 512;
3671 	sdkp->capacity = 0;
3672 	sdkp->media_present = 1;
3673 	sdkp->write_prot = 0;
3674 	sdkp->cache_override = 0;
3675 	sdkp->WCE = 0;
3676 	sdkp->RCD = 0;
3677 	sdkp->ATO = 0;
3678 	sdkp->first_scan = 1;
3679 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3680 
3681 	sd_revalidate_disk(gd);
3682 
3683 	if (sdp->removable) {
3684 		gd->flags |= GENHD_FL_REMOVABLE;
3685 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3686 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3687 	}
3688 
3689 	blk_pm_runtime_init(sdp->request_queue, dev);
3690 	if (sdp->rpm_autosuspend) {
3691 		pm_runtime_set_autosuspend_delay(dev,
3692 			sdp->host->hostt->rpm_autosuspend_delay);
3693 	}
3694 
3695 	error = device_add_disk(dev, gd, NULL);
3696 	if (error) {
3697 		put_device(&sdkp->disk_dev);
3698 		put_disk(gd);
3699 		goto out;
3700 	}
3701 
3702 	if (sdkp->security) {
3703 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3704 		if (sdkp->opal_dev)
3705 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3706 	}
3707 
3708 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3709 		  sdp->removable ? "removable " : "");
3710 	scsi_autopm_put_device(sdp);
3711 
3712 	return 0;
3713 
3714  out_free_index:
3715 	ida_free(&sd_index_ida, index);
3716  out_put:
3717 	put_disk(gd);
3718  out_free:
3719 	kfree(sdkp);
3720  out:
3721 	scsi_autopm_put_device(sdp);
3722 	return error;
3723 }
3724 
3725 /**
3726  *	sd_remove - called whenever a scsi disk (previously recognized by
3727  *	sd_probe) is detached from the system. It is called (potentially
3728  *	multiple times) during sd module unload.
3729  *	@dev: pointer to device object
3730  *
3731  *	Note: this function is invoked from the scsi mid-level.
3732  *	This function potentially frees up a device name (e.g. /dev/sdc)
3733  *	that could be re-used by a subsequent sd_probe().
3734  *	This function is not called when the built-in sd driver is "exit-ed".
3735  **/
3736 static int sd_remove(struct device *dev)
3737 {
3738 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3739 
3740 	scsi_autopm_get_device(sdkp->device);
3741 
3742 	device_del(&sdkp->disk_dev);
3743 	del_gendisk(sdkp->disk);
3744 	if (!sdkp->suspended)
3745 		sd_shutdown(dev);
3746 
3747 	put_disk(sdkp->disk);
3748 	return 0;
3749 }
3750 
3751 static void scsi_disk_release(struct device *dev)
3752 {
3753 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3754 
3755 	ida_free(&sd_index_ida, sdkp->index);
3756 	sd_zbc_free_zone_info(sdkp);
3757 	put_device(&sdkp->device->sdev_gendev);
3758 	free_opal_dev(sdkp->opal_dev);
3759 
3760 	kfree(sdkp);
3761 }
3762 
3763 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3764 {
3765 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3766 	struct scsi_sense_hdr sshdr;
3767 	const struct scsi_exec_args exec_args = {
3768 		.sshdr = &sshdr,
3769 		.req_flags = BLK_MQ_REQ_PM,
3770 	};
3771 	struct scsi_device *sdp = sdkp->device;
3772 	int res;
3773 
3774 	if (start)
3775 		cmd[4] |= 1;	/* START */
3776 
3777 	if (sdp->start_stop_pwr_cond)
3778 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3779 
3780 	if (!scsi_device_online(sdp))
3781 		return -ENODEV;
3782 
3783 	res = scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0, SD_TIMEOUT,
3784 			       sdkp->max_retries, &exec_args);
3785 	if (res) {
3786 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3787 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3788 			sd_print_sense_hdr(sdkp, &sshdr);
3789 			/* 0x3a is medium not present */
3790 			if (sshdr.asc == 0x3a)
3791 				res = 0;
3792 		}
3793 	}
3794 
3795 	/* SCSI error codes must not go to the generic layer */
3796 	if (res)
3797 		return -EIO;
3798 
3799 	return 0;
3800 }
3801 
3802 /*
3803  * Send a SYNCHRONIZE CACHE instruction down to the device through
3804  * the normal SCSI command structure.  Wait for the command to
3805  * complete.
3806  */
3807 static void sd_shutdown(struct device *dev)
3808 {
3809 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3810 
3811 	if (!sdkp)
3812 		return;         /* this can happen */
3813 
3814 	if (pm_runtime_suspended(dev))
3815 		return;
3816 
3817 	if (sdkp->WCE && sdkp->media_present) {
3818 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3819 		sd_sync_cache(sdkp, NULL);
3820 	}
3821 
3822 	if (system_state != SYSTEM_RESTART &&
3823 	    sdkp->device->manage_system_start_stop) {
3824 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3825 		sd_start_stop_device(sdkp, 0);
3826 	}
3827 }
3828 
3829 static inline bool sd_do_start_stop(struct scsi_device *sdev, bool runtime)
3830 {
3831 	return (sdev->manage_system_start_stop && !runtime) ||
3832 		(sdev->manage_runtime_start_stop && runtime);
3833 }
3834 
3835 static int sd_suspend_common(struct device *dev, bool runtime)
3836 {
3837 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3838 	struct scsi_sense_hdr sshdr;
3839 	int ret = 0;
3840 
3841 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3842 		return 0;
3843 
3844 	if (sdkp->WCE && sdkp->media_present) {
3845 		if (!sdkp->device->silence_suspend)
3846 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3847 		ret = sd_sync_cache(sdkp, &sshdr);
3848 
3849 		if (ret) {
3850 			/* ignore OFFLINE device */
3851 			if (ret == -ENODEV)
3852 				return 0;
3853 
3854 			if (!scsi_sense_valid(&sshdr) ||
3855 			    sshdr.sense_key != ILLEGAL_REQUEST)
3856 				return ret;
3857 
3858 			/*
3859 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3860 			 * doesn't support sync. There's not much to do and
3861 			 * suspend shouldn't fail.
3862 			 */
3863 			ret = 0;
3864 		}
3865 	}
3866 
3867 	if (sd_do_start_stop(sdkp->device, runtime)) {
3868 		if (!sdkp->device->silence_suspend)
3869 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3870 		/* an error is not worth aborting a system sleep */
3871 		ret = sd_start_stop_device(sdkp, 0);
3872 		if (!runtime)
3873 			ret = 0;
3874 	}
3875 
3876 	if (!ret)
3877 		sdkp->suspended = true;
3878 
3879 	return ret;
3880 }
3881 
3882 static int sd_suspend_system(struct device *dev)
3883 {
3884 	if (pm_runtime_suspended(dev))
3885 		return 0;
3886 
3887 	return sd_suspend_common(dev, false);
3888 }
3889 
3890 static int sd_suspend_runtime(struct device *dev)
3891 {
3892 	return sd_suspend_common(dev, true);
3893 }
3894 
3895 static int sd_resume(struct device *dev, bool runtime)
3896 {
3897 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3898 	int ret = 0;
3899 
3900 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3901 		return 0;
3902 
3903 	if (!sd_do_start_stop(sdkp->device, runtime)) {
3904 		sdkp->suspended = false;
3905 		return 0;
3906 	}
3907 
3908 	if (!sdkp->device->no_start_on_resume) {
3909 		sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3910 		ret = sd_start_stop_device(sdkp, 1);
3911 	}
3912 
3913 	if (!ret) {
3914 		opal_unlock_from_suspend(sdkp->opal_dev);
3915 		sdkp->suspended = false;
3916 	}
3917 
3918 	return ret;
3919 }
3920 
3921 static int sd_resume_system(struct device *dev)
3922 {
3923 	if (pm_runtime_suspended(dev))
3924 		return 0;
3925 
3926 	return sd_resume(dev, false);
3927 }
3928 
3929 static int sd_resume_runtime(struct device *dev)
3930 {
3931 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3932 	struct scsi_device *sdp;
3933 
3934 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3935 		return 0;
3936 
3937 	sdp = sdkp->device;
3938 
3939 	if (sdp->ignore_media_change) {
3940 		/* clear the device's sense data */
3941 		static const u8 cmd[10] = { REQUEST_SENSE };
3942 		const struct scsi_exec_args exec_args = {
3943 			.req_flags = BLK_MQ_REQ_PM,
3944 		};
3945 
3946 		if (scsi_execute_cmd(sdp, cmd, REQ_OP_DRV_IN, NULL, 0,
3947 				     sdp->request_queue->rq_timeout, 1,
3948 				     &exec_args))
3949 			sd_printk(KERN_NOTICE, sdkp,
3950 				  "Failed to clear sense data\n");
3951 	}
3952 
3953 	return sd_resume(dev, true);
3954 }
3955 
3956 static const struct dev_pm_ops sd_pm_ops = {
3957 	.suspend		= sd_suspend_system,
3958 	.resume			= sd_resume_system,
3959 	.poweroff		= sd_suspend_system,
3960 	.restore		= sd_resume_system,
3961 	.runtime_suspend	= sd_suspend_runtime,
3962 	.runtime_resume		= sd_resume_runtime,
3963 };
3964 
3965 static struct scsi_driver sd_template = {
3966 	.gendrv = {
3967 		.name		= "sd",
3968 		.owner		= THIS_MODULE,
3969 		.probe		= sd_probe,
3970 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
3971 		.remove		= sd_remove,
3972 		.shutdown	= sd_shutdown,
3973 		.pm		= &sd_pm_ops,
3974 	},
3975 	.rescan			= sd_rescan,
3976 	.init_command		= sd_init_command,
3977 	.uninit_command		= sd_uninit_command,
3978 	.done			= sd_done,
3979 	.eh_action		= sd_eh_action,
3980 	.eh_reset		= sd_eh_reset,
3981 };
3982 
3983 /**
3984  *	init_sd - entry point for this driver (both when built in or when
3985  *	a module).
3986  *
3987  *	Note: this function registers this driver with the scsi mid-level.
3988  **/
3989 static int __init init_sd(void)
3990 {
3991 	int majors = 0, i, err;
3992 
3993 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3994 
3995 	for (i = 0; i < SD_MAJORS; i++) {
3996 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3997 			continue;
3998 		majors++;
3999 	}
4000 
4001 	if (!majors)
4002 		return -ENODEV;
4003 
4004 	err = class_register(&sd_disk_class);
4005 	if (err)
4006 		goto err_out;
4007 
4008 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
4009 	if (!sd_page_pool) {
4010 		printk(KERN_ERR "sd: can't init discard page pool\n");
4011 		err = -ENOMEM;
4012 		goto err_out_class;
4013 	}
4014 
4015 	err = scsi_register_driver(&sd_template.gendrv);
4016 	if (err)
4017 		goto err_out_driver;
4018 
4019 	return 0;
4020 
4021 err_out_driver:
4022 	mempool_destroy(sd_page_pool);
4023 err_out_class:
4024 	class_unregister(&sd_disk_class);
4025 err_out:
4026 	for (i = 0; i < SD_MAJORS; i++)
4027 		unregister_blkdev(sd_major(i), "sd");
4028 	return err;
4029 }
4030 
4031 /**
4032  *	exit_sd - exit point for this driver (when it is a module).
4033  *
4034  *	Note: this function unregisters this driver from the scsi mid-level.
4035  **/
4036 static void __exit exit_sd(void)
4037 {
4038 	int i;
4039 
4040 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
4041 
4042 	scsi_unregister_driver(&sd_template.gendrv);
4043 	mempool_destroy(sd_page_pool);
4044 
4045 	class_unregister(&sd_disk_class);
4046 
4047 	for (i = 0; i < SD_MAJORS; i++)
4048 		unregister_blkdev(sd_major(i), "sd");
4049 }
4050 
4051 module_init(init_sd);
4052 module_exit(exit_sd);
4053 
4054 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
4055 {
4056 	scsi_print_sense_hdr(sdkp->device,
4057 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
4058 }
4059 
4060 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
4061 {
4062 	const char *hb_string = scsi_hostbyte_string(result);
4063 
4064 	if (hb_string)
4065 		sd_printk(KERN_INFO, sdkp,
4066 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
4067 			  hb_string ? hb_string : "invalid",
4068 			  "DRIVER_OK");
4069 	else
4070 		sd_printk(KERN_INFO, sdkp,
4071 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
4072 			  msg, host_byte(result), "DRIVER_OK");
4073 }
4074