xref: /openbmc/linux/drivers/scsi/sd.c (revision 4094981d)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *      sd.c Copyright (C) 1992 Drew Eckhardt
4  *           Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5  *
6  *      Linux scsi disk driver
7  *              Initial versions: Drew Eckhardt
8  *              Subsequent revisions: Eric Youngdale
9  *	Modification history:
10  *       - Drew Eckhardt <drew@colorado.edu> original
11  *       - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12  *         outstanding request, and other enhancements.
13  *         Support loadable low-level scsi drivers.
14  *       - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15  *         eight major numbers.
16  *       - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17  *	 - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18  *	   sd_init and cleanups.
19  *	 - Alex Davis <letmein@erols.com> Fix problem where partition info
20  *	   not being read in sd_open. Fix problem where removable media
21  *	   could be ejected after sd_open.
22  *	 - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23  *	 - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24  *	   <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25  *	   Support 32k/1M disks.
26  *
27  *	Logging policy (needs CONFIG_SCSI_LOGGING defined):
28  *	 - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29  *	 - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30  *	 - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31  *	 - entering other commands: SCSI_LOG_HLQUEUE level 3
32  *	Note: when the logging level is set by the user, it must be greater
33  *	than the level indicated above to trigger output.
34  */
35 
36 #include <linux/module.h>
37 #include <linux/fs.h>
38 #include <linux/kernel.h>
39 #include <linux/mm.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
56 #include <linux/pr.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
60 
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 
71 #include "sd.h"
72 #include "scsi_priv.h"
73 #include "scsi_logging.h"
74 
75 MODULE_AUTHOR("Eric Youngdale");
76 MODULE_DESCRIPTION("SCSI disk (sd) driver");
77 MODULE_LICENSE("GPL");
78 
79 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
95 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
99 
100 #define SD_MINORS	16
101 
102 static void sd_config_discard(struct scsi_disk *, unsigned int);
103 static void sd_config_write_same(struct scsi_disk *);
104 static int  sd_revalidate_disk(struct gendisk *);
105 static void sd_unlock_native_capacity(struct gendisk *disk);
106 static int  sd_probe(struct device *);
107 static int  sd_remove(struct device *);
108 static void sd_shutdown(struct device *);
109 static int sd_suspend_system(struct device *);
110 static int sd_suspend_runtime(struct device *);
111 static int sd_resume_system(struct device *);
112 static int sd_resume_runtime(struct device *);
113 static void sd_rescan(struct device *);
114 static blk_status_t sd_init_command(struct scsi_cmnd *SCpnt);
115 static void sd_uninit_command(struct scsi_cmnd *SCpnt);
116 static int sd_done(struct scsi_cmnd *);
117 static void sd_eh_reset(struct scsi_cmnd *);
118 static int sd_eh_action(struct scsi_cmnd *, int);
119 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
120 static void scsi_disk_release(struct device *cdev);
121 
122 static DEFINE_IDA(sd_index_ida);
123 
124 static struct kmem_cache *sd_cdb_cache;
125 static mempool_t *sd_page_pool;
126 static struct lock_class_key sd_bio_compl_lkclass;
127 
128 static const char *sd_cache_types[] = {
129 	"write through", "none", "write back",
130 	"write back, no read (daft)"
131 };
132 
133 static void sd_set_flush_flag(struct scsi_disk *sdkp)
134 {
135 	bool wc = false, fua = false;
136 
137 	if (sdkp->WCE) {
138 		wc = true;
139 		if (sdkp->DPOFUA)
140 			fua = true;
141 	}
142 
143 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
144 }
145 
146 static ssize_t
147 cache_type_store(struct device *dev, struct device_attribute *attr,
148 		 const char *buf, size_t count)
149 {
150 	int ct, rcd, wce, sp;
151 	struct scsi_disk *sdkp = to_scsi_disk(dev);
152 	struct scsi_device *sdp = sdkp->device;
153 	char buffer[64];
154 	char *buffer_data;
155 	struct scsi_mode_data data;
156 	struct scsi_sense_hdr sshdr;
157 	static const char temp[] = "temporary ";
158 	int len;
159 
160 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
161 		/* no cache control on RBC devices; theoretically they
162 		 * can do it, but there's probably so many exceptions
163 		 * it's not worth the risk */
164 		return -EINVAL;
165 
166 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
167 		buf += sizeof(temp) - 1;
168 		sdkp->cache_override = 1;
169 	} else {
170 		sdkp->cache_override = 0;
171 	}
172 
173 	ct = sysfs_match_string(sd_cache_types, buf);
174 	if (ct < 0)
175 		return -EINVAL;
176 
177 	rcd = ct & 0x01 ? 1 : 0;
178 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
179 
180 	if (sdkp->cache_override) {
181 		sdkp->WCE = wce;
182 		sdkp->RCD = rcd;
183 		sd_set_flush_flag(sdkp);
184 		return count;
185 	}
186 
187 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
188 			    sdkp->max_retries, &data, NULL))
189 		return -EINVAL;
190 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
191 		  data.block_descriptor_length);
192 	buffer_data = buffer + data.header_length +
193 		data.block_descriptor_length;
194 	buffer_data[2] &= ~0x05;
195 	buffer_data[2] |= wce << 2 | rcd;
196 	sp = buffer_data[0] & 0x80 ? 1 : 0;
197 	buffer_data[0] &= ~0x80;
198 
199 	/*
200 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
201 	 * received mode parameter buffer before doing MODE SELECT.
202 	 */
203 	data.device_specific = 0;
204 
205 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
206 			     sdkp->max_retries, &data, &sshdr)) {
207 		if (scsi_sense_valid(&sshdr))
208 			sd_print_sense_hdr(sdkp, &sshdr);
209 		return -EINVAL;
210 	}
211 	sd_revalidate_disk(sdkp->disk);
212 	return count;
213 }
214 
215 static ssize_t
216 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
217 		       char *buf)
218 {
219 	struct scsi_disk *sdkp = to_scsi_disk(dev);
220 	struct scsi_device *sdp = sdkp->device;
221 
222 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
223 }
224 
225 static ssize_t
226 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
227 			const char *buf, size_t count)
228 {
229 	struct scsi_disk *sdkp = to_scsi_disk(dev);
230 	struct scsi_device *sdp = sdkp->device;
231 	bool v;
232 
233 	if (!capable(CAP_SYS_ADMIN))
234 		return -EACCES;
235 
236 	if (kstrtobool(buf, &v))
237 		return -EINVAL;
238 
239 	sdp->manage_start_stop = v;
240 
241 	return count;
242 }
243 static DEVICE_ATTR_RW(manage_start_stop);
244 
245 static ssize_t
246 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
247 {
248 	struct scsi_disk *sdkp = to_scsi_disk(dev);
249 
250 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
251 }
252 
253 static ssize_t
254 allow_restart_store(struct device *dev, struct device_attribute *attr,
255 		    const char *buf, size_t count)
256 {
257 	bool v;
258 	struct scsi_disk *sdkp = to_scsi_disk(dev);
259 	struct scsi_device *sdp = sdkp->device;
260 
261 	if (!capable(CAP_SYS_ADMIN))
262 		return -EACCES;
263 
264 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
265 		return -EINVAL;
266 
267 	if (kstrtobool(buf, &v))
268 		return -EINVAL;
269 
270 	sdp->allow_restart = v;
271 
272 	return count;
273 }
274 static DEVICE_ATTR_RW(allow_restart);
275 
276 static ssize_t
277 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
278 {
279 	struct scsi_disk *sdkp = to_scsi_disk(dev);
280 	int ct = sdkp->RCD + 2*sdkp->WCE;
281 
282 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
283 }
284 static DEVICE_ATTR_RW(cache_type);
285 
286 static ssize_t
287 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
288 {
289 	struct scsi_disk *sdkp = to_scsi_disk(dev);
290 
291 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
292 }
293 static DEVICE_ATTR_RO(FUA);
294 
295 static ssize_t
296 protection_type_show(struct device *dev, struct device_attribute *attr,
297 		     char *buf)
298 {
299 	struct scsi_disk *sdkp = to_scsi_disk(dev);
300 
301 	return sprintf(buf, "%u\n", sdkp->protection_type);
302 }
303 
304 static ssize_t
305 protection_type_store(struct device *dev, struct device_attribute *attr,
306 		      const char *buf, size_t count)
307 {
308 	struct scsi_disk *sdkp = to_scsi_disk(dev);
309 	unsigned int val;
310 	int err;
311 
312 	if (!capable(CAP_SYS_ADMIN))
313 		return -EACCES;
314 
315 	err = kstrtouint(buf, 10, &val);
316 
317 	if (err)
318 		return err;
319 
320 	if (val <= T10_PI_TYPE3_PROTECTION)
321 		sdkp->protection_type = val;
322 
323 	return count;
324 }
325 static DEVICE_ATTR_RW(protection_type);
326 
327 static ssize_t
328 protection_mode_show(struct device *dev, struct device_attribute *attr,
329 		     char *buf)
330 {
331 	struct scsi_disk *sdkp = to_scsi_disk(dev);
332 	struct scsi_device *sdp = sdkp->device;
333 	unsigned int dif, dix;
334 
335 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
336 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
337 
338 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
339 		dif = 0;
340 		dix = 1;
341 	}
342 
343 	if (!dif && !dix)
344 		return sprintf(buf, "none\n");
345 
346 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
347 }
348 static DEVICE_ATTR_RO(protection_mode);
349 
350 static ssize_t
351 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
352 {
353 	struct scsi_disk *sdkp = to_scsi_disk(dev);
354 
355 	return sprintf(buf, "%u\n", sdkp->ATO);
356 }
357 static DEVICE_ATTR_RO(app_tag_own);
358 
359 static ssize_t
360 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
361 		       char *buf)
362 {
363 	struct scsi_disk *sdkp = to_scsi_disk(dev);
364 
365 	return sprintf(buf, "%u\n", sdkp->lbpme);
366 }
367 static DEVICE_ATTR_RO(thin_provisioning);
368 
369 /* sysfs_match_string() requires dense arrays */
370 static const char *lbp_mode[] = {
371 	[SD_LBP_FULL]		= "full",
372 	[SD_LBP_UNMAP]		= "unmap",
373 	[SD_LBP_WS16]		= "writesame_16",
374 	[SD_LBP_WS10]		= "writesame_10",
375 	[SD_LBP_ZERO]		= "writesame_zero",
376 	[SD_LBP_DISABLE]	= "disabled",
377 };
378 
379 static ssize_t
380 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
381 		       char *buf)
382 {
383 	struct scsi_disk *sdkp = to_scsi_disk(dev);
384 
385 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
386 }
387 
388 static ssize_t
389 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
390 			const char *buf, size_t count)
391 {
392 	struct scsi_disk *sdkp = to_scsi_disk(dev);
393 	struct scsi_device *sdp = sdkp->device;
394 	int mode;
395 
396 	if (!capable(CAP_SYS_ADMIN))
397 		return -EACCES;
398 
399 	if (sd_is_zoned(sdkp)) {
400 		sd_config_discard(sdkp, SD_LBP_DISABLE);
401 		return count;
402 	}
403 
404 	if (sdp->type != TYPE_DISK)
405 		return -EINVAL;
406 
407 	mode = sysfs_match_string(lbp_mode, buf);
408 	if (mode < 0)
409 		return -EINVAL;
410 
411 	sd_config_discard(sdkp, mode);
412 
413 	return count;
414 }
415 static DEVICE_ATTR_RW(provisioning_mode);
416 
417 /* sysfs_match_string() requires dense arrays */
418 static const char *zeroing_mode[] = {
419 	[SD_ZERO_WRITE]		= "write",
420 	[SD_ZERO_WS]		= "writesame",
421 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
422 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
423 };
424 
425 static ssize_t
426 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
427 		  char *buf)
428 {
429 	struct scsi_disk *sdkp = to_scsi_disk(dev);
430 
431 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
432 }
433 
434 static ssize_t
435 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
436 		   const char *buf, size_t count)
437 {
438 	struct scsi_disk *sdkp = to_scsi_disk(dev);
439 	int mode;
440 
441 	if (!capable(CAP_SYS_ADMIN))
442 		return -EACCES;
443 
444 	mode = sysfs_match_string(zeroing_mode, buf);
445 	if (mode < 0)
446 		return -EINVAL;
447 
448 	sdkp->zeroing_mode = mode;
449 
450 	return count;
451 }
452 static DEVICE_ATTR_RW(zeroing_mode);
453 
454 static ssize_t
455 max_medium_access_timeouts_show(struct device *dev,
456 				struct device_attribute *attr, char *buf)
457 {
458 	struct scsi_disk *sdkp = to_scsi_disk(dev);
459 
460 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
461 }
462 
463 static ssize_t
464 max_medium_access_timeouts_store(struct device *dev,
465 				 struct device_attribute *attr, const char *buf,
466 				 size_t count)
467 {
468 	struct scsi_disk *sdkp = to_scsi_disk(dev);
469 	int err;
470 
471 	if (!capable(CAP_SYS_ADMIN))
472 		return -EACCES;
473 
474 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
475 
476 	return err ? err : count;
477 }
478 static DEVICE_ATTR_RW(max_medium_access_timeouts);
479 
480 static ssize_t
481 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
482 			   char *buf)
483 {
484 	struct scsi_disk *sdkp = to_scsi_disk(dev);
485 
486 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
487 }
488 
489 static ssize_t
490 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
491 			    const char *buf, size_t count)
492 {
493 	struct scsi_disk *sdkp = to_scsi_disk(dev);
494 	struct scsi_device *sdp = sdkp->device;
495 	unsigned long max;
496 	int err;
497 
498 	if (!capable(CAP_SYS_ADMIN))
499 		return -EACCES;
500 
501 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
502 		return -EINVAL;
503 
504 	err = kstrtoul(buf, 10, &max);
505 
506 	if (err)
507 		return err;
508 
509 	if (max == 0)
510 		sdp->no_write_same = 1;
511 	else if (max <= SD_MAX_WS16_BLOCKS) {
512 		sdp->no_write_same = 0;
513 		sdkp->max_ws_blocks = max;
514 	}
515 
516 	sd_config_write_same(sdkp);
517 
518 	return count;
519 }
520 static DEVICE_ATTR_RW(max_write_same_blocks);
521 
522 static ssize_t
523 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
524 {
525 	struct scsi_disk *sdkp = to_scsi_disk(dev);
526 
527 	if (sdkp->device->type == TYPE_ZBC)
528 		return sprintf(buf, "host-managed\n");
529 	if (sdkp->zoned == 1)
530 		return sprintf(buf, "host-aware\n");
531 	if (sdkp->zoned == 2)
532 		return sprintf(buf, "drive-managed\n");
533 	return sprintf(buf, "none\n");
534 }
535 static DEVICE_ATTR_RO(zoned_cap);
536 
537 static ssize_t
538 max_retries_store(struct device *dev, struct device_attribute *attr,
539 		  const char *buf, size_t count)
540 {
541 	struct scsi_disk *sdkp = to_scsi_disk(dev);
542 	struct scsi_device *sdev = sdkp->device;
543 	int retries, err;
544 
545 	err = kstrtoint(buf, 10, &retries);
546 	if (err)
547 		return err;
548 
549 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
550 		sdkp->max_retries = retries;
551 		return count;
552 	}
553 
554 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
555 		    SD_MAX_RETRIES);
556 	return -EINVAL;
557 }
558 
559 static ssize_t
560 max_retries_show(struct device *dev, struct device_attribute *attr,
561 		 char *buf)
562 {
563 	struct scsi_disk *sdkp = to_scsi_disk(dev);
564 
565 	return sprintf(buf, "%d\n", sdkp->max_retries);
566 }
567 
568 static DEVICE_ATTR_RW(max_retries);
569 
570 static struct attribute *sd_disk_attrs[] = {
571 	&dev_attr_cache_type.attr,
572 	&dev_attr_FUA.attr,
573 	&dev_attr_allow_restart.attr,
574 	&dev_attr_manage_start_stop.attr,
575 	&dev_attr_protection_type.attr,
576 	&dev_attr_protection_mode.attr,
577 	&dev_attr_app_tag_own.attr,
578 	&dev_attr_thin_provisioning.attr,
579 	&dev_attr_provisioning_mode.attr,
580 	&dev_attr_zeroing_mode.attr,
581 	&dev_attr_max_write_same_blocks.attr,
582 	&dev_attr_max_medium_access_timeouts.attr,
583 	&dev_attr_zoned_cap.attr,
584 	&dev_attr_max_retries.attr,
585 	NULL,
586 };
587 ATTRIBUTE_GROUPS(sd_disk);
588 
589 static struct class sd_disk_class = {
590 	.name		= "scsi_disk",
591 	.owner		= THIS_MODULE,
592 	.dev_release	= scsi_disk_release,
593 	.dev_groups	= sd_disk_groups,
594 };
595 
596 static const struct dev_pm_ops sd_pm_ops = {
597 	.suspend		= sd_suspend_system,
598 	.resume			= sd_resume_system,
599 	.poweroff		= sd_suspend_system,
600 	.restore		= sd_resume_system,
601 	.runtime_suspend	= sd_suspend_runtime,
602 	.runtime_resume		= sd_resume_runtime,
603 };
604 
605 static struct scsi_driver sd_template = {
606 	.gendrv = {
607 		.name		= "sd",
608 		.owner		= THIS_MODULE,
609 		.probe		= sd_probe,
610 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
611 		.remove		= sd_remove,
612 		.shutdown	= sd_shutdown,
613 		.pm		= &sd_pm_ops,
614 	},
615 	.rescan			= sd_rescan,
616 	.init_command		= sd_init_command,
617 	.uninit_command		= sd_uninit_command,
618 	.done			= sd_done,
619 	.eh_action		= sd_eh_action,
620 	.eh_reset		= sd_eh_reset,
621 };
622 
623 /*
624  * Don't request a new module, as that could deadlock in multipath
625  * environment.
626  */
627 static void sd_default_probe(dev_t devt)
628 {
629 }
630 
631 /*
632  * Device no to disk mapping:
633  *
634  *       major         disc2     disc  p1
635  *   |............|.............|....|....| <- dev_t
636  *    31        20 19          8 7  4 3  0
637  *
638  * Inside a major, we have 16k disks, however mapped non-
639  * contiguously. The first 16 disks are for major0, the next
640  * ones with major1, ... Disk 256 is for major0 again, disk 272
641  * for major1, ...
642  * As we stay compatible with our numbering scheme, we can reuse
643  * the well-know SCSI majors 8, 65--71, 136--143.
644  */
645 static int sd_major(int major_idx)
646 {
647 	switch (major_idx) {
648 	case 0:
649 		return SCSI_DISK0_MAJOR;
650 	case 1 ... 7:
651 		return SCSI_DISK1_MAJOR + major_idx - 1;
652 	case 8 ... 15:
653 		return SCSI_DISK8_MAJOR + major_idx - 8;
654 	default:
655 		BUG();
656 		return 0;	/* shut up gcc */
657 	}
658 }
659 
660 #ifdef CONFIG_BLK_SED_OPAL
661 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
662 		size_t len, bool send)
663 {
664 	struct scsi_disk *sdkp = data;
665 	struct scsi_device *sdev = sdkp->device;
666 	u8 cdb[12] = { 0, };
667 	int ret;
668 
669 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
670 	cdb[1] = secp;
671 	put_unaligned_be16(spsp, &cdb[2]);
672 	put_unaligned_be32(len, &cdb[6]);
673 
674 	ret = scsi_execute(sdev, cdb, send ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
675 		buffer, len, NULL, NULL, SD_TIMEOUT, sdkp->max_retries, 0,
676 		RQF_PM, NULL);
677 	return ret <= 0 ? ret : -EIO;
678 }
679 #endif /* CONFIG_BLK_SED_OPAL */
680 
681 /*
682  * Look up the DIX operation based on whether the command is read or
683  * write and whether dix and dif are enabled.
684  */
685 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
686 {
687 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
688 	static const unsigned int ops[] = {	/* wrt dix dif */
689 		SCSI_PROT_NORMAL,		/*  0	0   0  */
690 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
691 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
692 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
693 		SCSI_PROT_NORMAL,		/*  1	0   0  */
694 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
695 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
696 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
697 	};
698 
699 	return ops[write << 2 | dix << 1 | dif];
700 }
701 
702 /*
703  * Returns a mask of the protection flags that are valid for a given DIX
704  * operation.
705  */
706 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
707 {
708 	static const unsigned int flag_mask[] = {
709 		[SCSI_PROT_NORMAL]		= 0,
710 
711 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
712 						  SCSI_PROT_GUARD_CHECK |
713 						  SCSI_PROT_REF_CHECK |
714 						  SCSI_PROT_REF_INCREMENT,
715 
716 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
717 						  SCSI_PROT_IP_CHECKSUM,
718 
719 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
720 						  SCSI_PROT_GUARD_CHECK |
721 						  SCSI_PROT_REF_CHECK |
722 						  SCSI_PROT_REF_INCREMENT |
723 						  SCSI_PROT_IP_CHECKSUM,
724 
725 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
726 						  SCSI_PROT_REF_INCREMENT,
727 
728 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
729 						  SCSI_PROT_REF_CHECK |
730 						  SCSI_PROT_REF_INCREMENT |
731 						  SCSI_PROT_IP_CHECKSUM,
732 
733 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
734 						  SCSI_PROT_GUARD_CHECK |
735 						  SCSI_PROT_REF_CHECK |
736 						  SCSI_PROT_REF_INCREMENT |
737 						  SCSI_PROT_IP_CHECKSUM,
738 	};
739 
740 	return flag_mask[prot_op];
741 }
742 
743 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
744 					   unsigned int dix, unsigned int dif)
745 {
746 	struct request *rq = scsi_cmd_to_rq(scmd);
747 	struct bio *bio = rq->bio;
748 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
749 	unsigned int protect = 0;
750 
751 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
752 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
753 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
754 
755 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
756 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
757 	}
758 
759 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
760 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
761 
762 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
763 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
764 	}
765 
766 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
767 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
768 
769 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
770 			protect = 3 << 5;	/* Disable target PI checking */
771 		else
772 			protect = 1 << 5;	/* Enable target PI checking */
773 	}
774 
775 	scsi_set_prot_op(scmd, prot_op);
776 	scsi_set_prot_type(scmd, dif);
777 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
778 
779 	return protect;
780 }
781 
782 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
783 {
784 	struct request_queue *q = sdkp->disk->queue;
785 	unsigned int logical_block_size = sdkp->device->sector_size;
786 	unsigned int max_blocks = 0;
787 
788 	q->limits.discard_alignment =
789 		sdkp->unmap_alignment * logical_block_size;
790 	q->limits.discard_granularity =
791 		max(sdkp->physical_block_size,
792 		    sdkp->unmap_granularity * logical_block_size);
793 	sdkp->provisioning_mode = mode;
794 
795 	switch (mode) {
796 
797 	case SD_LBP_FULL:
798 	case SD_LBP_DISABLE:
799 		blk_queue_max_discard_sectors(q, 0);
800 		blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
801 		return;
802 
803 	case SD_LBP_UNMAP:
804 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
805 					  (u32)SD_MAX_WS16_BLOCKS);
806 		break;
807 
808 	case SD_LBP_WS16:
809 		if (sdkp->device->unmap_limit_for_ws)
810 			max_blocks = sdkp->max_unmap_blocks;
811 		else
812 			max_blocks = sdkp->max_ws_blocks;
813 
814 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
815 		break;
816 
817 	case SD_LBP_WS10:
818 		if (sdkp->device->unmap_limit_for_ws)
819 			max_blocks = sdkp->max_unmap_blocks;
820 		else
821 			max_blocks = sdkp->max_ws_blocks;
822 
823 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
824 		break;
825 
826 	case SD_LBP_ZERO:
827 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
828 					  (u32)SD_MAX_WS10_BLOCKS);
829 		break;
830 	}
831 
832 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
833 	blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
834 }
835 
836 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
837 {
838 	struct scsi_device *sdp = cmd->device;
839 	struct request *rq = scsi_cmd_to_rq(cmd);
840 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
841 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
842 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
843 	unsigned int data_len = 24;
844 	char *buf;
845 
846 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
847 	if (!rq->special_vec.bv_page)
848 		return BLK_STS_RESOURCE;
849 	clear_highpage(rq->special_vec.bv_page);
850 	rq->special_vec.bv_offset = 0;
851 	rq->special_vec.bv_len = data_len;
852 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
853 
854 	cmd->cmd_len = 10;
855 	cmd->cmnd[0] = UNMAP;
856 	cmd->cmnd[8] = 24;
857 
858 	buf = bvec_virt(&rq->special_vec);
859 	put_unaligned_be16(6 + 16, &buf[0]);
860 	put_unaligned_be16(16, &buf[2]);
861 	put_unaligned_be64(lba, &buf[8]);
862 	put_unaligned_be32(nr_blocks, &buf[16]);
863 
864 	cmd->allowed = sdkp->max_retries;
865 	cmd->transfersize = data_len;
866 	rq->timeout = SD_TIMEOUT;
867 
868 	return scsi_alloc_sgtables(cmd);
869 }
870 
871 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
872 		bool unmap)
873 {
874 	struct scsi_device *sdp = cmd->device;
875 	struct request *rq = scsi_cmd_to_rq(cmd);
876 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
877 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
878 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
879 	u32 data_len = sdp->sector_size;
880 
881 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
882 	if (!rq->special_vec.bv_page)
883 		return BLK_STS_RESOURCE;
884 	clear_highpage(rq->special_vec.bv_page);
885 	rq->special_vec.bv_offset = 0;
886 	rq->special_vec.bv_len = data_len;
887 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
888 
889 	cmd->cmd_len = 16;
890 	cmd->cmnd[0] = WRITE_SAME_16;
891 	if (unmap)
892 		cmd->cmnd[1] = 0x8; /* UNMAP */
893 	put_unaligned_be64(lba, &cmd->cmnd[2]);
894 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
895 
896 	cmd->allowed = sdkp->max_retries;
897 	cmd->transfersize = data_len;
898 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
899 
900 	return scsi_alloc_sgtables(cmd);
901 }
902 
903 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
904 		bool unmap)
905 {
906 	struct scsi_device *sdp = cmd->device;
907 	struct request *rq = scsi_cmd_to_rq(cmd);
908 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
909 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
910 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
911 	u32 data_len = sdp->sector_size;
912 
913 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
914 	if (!rq->special_vec.bv_page)
915 		return BLK_STS_RESOURCE;
916 	clear_highpage(rq->special_vec.bv_page);
917 	rq->special_vec.bv_offset = 0;
918 	rq->special_vec.bv_len = data_len;
919 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
920 
921 	cmd->cmd_len = 10;
922 	cmd->cmnd[0] = WRITE_SAME;
923 	if (unmap)
924 		cmd->cmnd[1] = 0x8; /* UNMAP */
925 	put_unaligned_be32(lba, &cmd->cmnd[2]);
926 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
927 
928 	cmd->allowed = sdkp->max_retries;
929 	cmd->transfersize = data_len;
930 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
931 
932 	return scsi_alloc_sgtables(cmd);
933 }
934 
935 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
936 {
937 	struct request *rq = scsi_cmd_to_rq(cmd);
938 	struct scsi_device *sdp = cmd->device;
939 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
940 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
941 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
942 
943 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
944 		switch (sdkp->zeroing_mode) {
945 		case SD_ZERO_WS16_UNMAP:
946 			return sd_setup_write_same16_cmnd(cmd, true);
947 		case SD_ZERO_WS10_UNMAP:
948 			return sd_setup_write_same10_cmnd(cmd, true);
949 		}
950 	}
951 
952 	if (sdp->no_write_same) {
953 		rq->rq_flags |= RQF_QUIET;
954 		return BLK_STS_TARGET;
955 	}
956 
957 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
958 		return sd_setup_write_same16_cmnd(cmd, false);
959 
960 	return sd_setup_write_same10_cmnd(cmd, false);
961 }
962 
963 static void sd_config_write_same(struct scsi_disk *sdkp)
964 {
965 	struct request_queue *q = sdkp->disk->queue;
966 	unsigned int logical_block_size = sdkp->device->sector_size;
967 
968 	if (sdkp->device->no_write_same) {
969 		sdkp->max_ws_blocks = 0;
970 		goto out;
971 	}
972 
973 	/* Some devices can not handle block counts above 0xffff despite
974 	 * supporting WRITE SAME(16). Consequently we default to 64k
975 	 * blocks per I/O unless the device explicitly advertises a
976 	 * bigger limit.
977 	 */
978 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
979 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
980 						   (u32)SD_MAX_WS16_BLOCKS);
981 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
982 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
983 						   (u32)SD_MAX_WS10_BLOCKS);
984 	else {
985 		sdkp->device->no_write_same = 1;
986 		sdkp->max_ws_blocks = 0;
987 	}
988 
989 	if (sdkp->lbprz && sdkp->lbpws)
990 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
991 	else if (sdkp->lbprz && sdkp->lbpws10)
992 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
993 	else if (sdkp->max_ws_blocks)
994 		sdkp->zeroing_mode = SD_ZERO_WS;
995 	else
996 		sdkp->zeroing_mode = SD_ZERO_WRITE;
997 
998 	if (sdkp->max_ws_blocks &&
999 	    sdkp->physical_block_size > logical_block_size) {
1000 		/*
1001 		 * Reporting a maximum number of blocks that is not aligned
1002 		 * on the device physical size would cause a large write same
1003 		 * request to be split into physically unaligned chunks by
1004 		 * __blkdev_issue_write_zeroes() even if the caller of this
1005 		 * functions took care to align the large request. So make sure
1006 		 * the maximum reported is aligned to the device physical block
1007 		 * size. This is only an optional optimization for regular
1008 		 * disks, but this is mandatory to avoid failure of large write
1009 		 * same requests directed at sequential write required zones of
1010 		 * host-managed ZBC disks.
1011 		 */
1012 		sdkp->max_ws_blocks =
1013 			round_down(sdkp->max_ws_blocks,
1014 				   bytes_to_logical(sdkp->device,
1015 						    sdkp->physical_block_size));
1016 	}
1017 
1018 out:
1019 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1020 					 (logical_block_size >> 9));
1021 }
1022 
1023 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1024 {
1025 	struct request *rq = scsi_cmd_to_rq(cmd);
1026 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1027 
1028 	/* flush requests don't perform I/O, zero the S/G table */
1029 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1030 
1031 	cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1032 	cmd->cmd_len = 10;
1033 	cmd->transfersize = 0;
1034 	cmd->allowed = sdkp->max_retries;
1035 
1036 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1037 	return BLK_STS_OK;
1038 }
1039 
1040 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1041 				       sector_t lba, unsigned int nr_blocks,
1042 				       unsigned char flags)
1043 {
1044 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1045 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1046 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1047 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1048 	cmd->cmnd[10] = flags;
1049 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1050 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1051 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1052 
1053 	return BLK_STS_OK;
1054 }
1055 
1056 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1057 				       sector_t lba, unsigned int nr_blocks,
1058 				       unsigned char flags)
1059 {
1060 	cmd->cmd_len  = 16;
1061 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1062 	cmd->cmnd[1]  = flags;
1063 	cmd->cmnd[14] = 0;
1064 	cmd->cmnd[15] = 0;
1065 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1066 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1067 
1068 	return BLK_STS_OK;
1069 }
1070 
1071 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1072 				       sector_t lba, unsigned int nr_blocks,
1073 				       unsigned char flags)
1074 {
1075 	cmd->cmd_len = 10;
1076 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1077 	cmd->cmnd[1] = flags;
1078 	cmd->cmnd[6] = 0;
1079 	cmd->cmnd[9] = 0;
1080 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1081 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1082 
1083 	return BLK_STS_OK;
1084 }
1085 
1086 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1087 				      sector_t lba, unsigned int nr_blocks,
1088 				      unsigned char flags)
1089 {
1090 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1091 	if (WARN_ON_ONCE(nr_blocks == 0))
1092 		return BLK_STS_IOERR;
1093 
1094 	if (unlikely(flags & 0x8)) {
1095 		/*
1096 		 * This happens only if this drive failed 10byte rw
1097 		 * command with ILLEGAL_REQUEST during operation and
1098 		 * thus turned off use_10_for_rw.
1099 		 */
1100 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1101 		return BLK_STS_IOERR;
1102 	}
1103 
1104 	cmd->cmd_len = 6;
1105 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1106 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1107 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1108 	cmd->cmnd[3] = lba & 0xff;
1109 	cmd->cmnd[4] = nr_blocks;
1110 	cmd->cmnd[5] = 0;
1111 
1112 	return BLK_STS_OK;
1113 }
1114 
1115 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1116 {
1117 	struct request *rq = scsi_cmd_to_rq(cmd);
1118 	struct scsi_device *sdp = cmd->device;
1119 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1120 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1121 	sector_t threshold;
1122 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1123 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1124 	bool write = rq_data_dir(rq) == WRITE;
1125 	unsigned char protect, fua;
1126 	blk_status_t ret;
1127 	unsigned int dif;
1128 	bool dix;
1129 
1130 	ret = scsi_alloc_sgtables(cmd);
1131 	if (ret != BLK_STS_OK)
1132 		return ret;
1133 
1134 	ret = BLK_STS_IOERR;
1135 	if (!scsi_device_online(sdp) || sdp->changed) {
1136 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1137 		goto fail;
1138 	}
1139 
1140 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1141 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1142 		goto fail;
1143 	}
1144 
1145 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1146 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1147 		goto fail;
1148 	}
1149 
1150 	/*
1151 	 * Some SD card readers can't handle accesses which touch the
1152 	 * last one or two logical blocks. Split accesses as needed.
1153 	 */
1154 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1155 
1156 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1157 		if (lba < threshold) {
1158 			/* Access up to the threshold but not beyond */
1159 			nr_blocks = threshold - lba;
1160 		} else {
1161 			/* Access only a single logical block */
1162 			nr_blocks = 1;
1163 		}
1164 	}
1165 
1166 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1167 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1168 		if (ret)
1169 			goto fail;
1170 	}
1171 
1172 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1173 	dix = scsi_prot_sg_count(cmd);
1174 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1175 
1176 	if (dif || dix)
1177 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1178 	else
1179 		protect = 0;
1180 
1181 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1182 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1183 					 protect | fua);
1184 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1185 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1186 					 protect | fua);
1187 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1188 		   sdp->use_10_for_rw || protect) {
1189 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1190 					 protect | fua);
1191 	} else {
1192 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1193 					protect | fua);
1194 	}
1195 
1196 	if (unlikely(ret != BLK_STS_OK))
1197 		goto fail;
1198 
1199 	/*
1200 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1201 	 * host adapter, it's safe to assume that we can at least transfer
1202 	 * this many bytes between each connect / disconnect.
1203 	 */
1204 	cmd->transfersize = sdp->sector_size;
1205 	cmd->underflow = nr_blocks << 9;
1206 	cmd->allowed = sdkp->max_retries;
1207 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1208 
1209 	SCSI_LOG_HLQUEUE(1,
1210 			 scmd_printk(KERN_INFO, cmd,
1211 				     "%s: block=%llu, count=%d\n", __func__,
1212 				     (unsigned long long)blk_rq_pos(rq),
1213 				     blk_rq_sectors(rq)));
1214 	SCSI_LOG_HLQUEUE(2,
1215 			 scmd_printk(KERN_INFO, cmd,
1216 				     "%s %d/%u 512 byte blocks.\n",
1217 				     write ? "writing" : "reading", nr_blocks,
1218 				     blk_rq_sectors(rq)));
1219 
1220 	/*
1221 	 * This indicates that the command is ready from our end to be queued.
1222 	 */
1223 	return BLK_STS_OK;
1224 fail:
1225 	scsi_free_sgtables(cmd);
1226 	return ret;
1227 }
1228 
1229 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1230 {
1231 	struct request *rq = scsi_cmd_to_rq(cmd);
1232 
1233 	switch (req_op(rq)) {
1234 	case REQ_OP_DISCARD:
1235 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1236 		case SD_LBP_UNMAP:
1237 			return sd_setup_unmap_cmnd(cmd);
1238 		case SD_LBP_WS16:
1239 			return sd_setup_write_same16_cmnd(cmd, true);
1240 		case SD_LBP_WS10:
1241 			return sd_setup_write_same10_cmnd(cmd, true);
1242 		case SD_LBP_ZERO:
1243 			return sd_setup_write_same10_cmnd(cmd, false);
1244 		default:
1245 			return BLK_STS_TARGET;
1246 		}
1247 	case REQ_OP_WRITE_ZEROES:
1248 		return sd_setup_write_zeroes_cmnd(cmd);
1249 	case REQ_OP_FLUSH:
1250 		return sd_setup_flush_cmnd(cmd);
1251 	case REQ_OP_READ:
1252 	case REQ_OP_WRITE:
1253 	case REQ_OP_ZONE_APPEND:
1254 		return sd_setup_read_write_cmnd(cmd);
1255 	case REQ_OP_ZONE_RESET:
1256 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1257 						   false);
1258 	case REQ_OP_ZONE_RESET_ALL:
1259 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1260 						   true);
1261 	case REQ_OP_ZONE_OPEN:
1262 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1263 	case REQ_OP_ZONE_CLOSE:
1264 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1265 	case REQ_OP_ZONE_FINISH:
1266 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1267 	default:
1268 		WARN_ON_ONCE(1);
1269 		return BLK_STS_NOTSUPP;
1270 	}
1271 }
1272 
1273 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1274 {
1275 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1276 
1277 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1278 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1279 }
1280 
1281 static bool sd_need_revalidate(struct block_device *bdev,
1282 		struct scsi_disk *sdkp)
1283 {
1284 	if (sdkp->device->removable || sdkp->write_prot) {
1285 		if (bdev_check_media_change(bdev))
1286 			return true;
1287 	}
1288 
1289 	/*
1290 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1291 	 * nothing to do with partitions, BLKRRPART is used to force a full
1292 	 * revalidate after things like a format for historical reasons.
1293 	 */
1294 	return test_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1295 }
1296 
1297 /**
1298  *	sd_open - open a scsi disk device
1299  *	@bdev: Block device of the scsi disk to open
1300  *	@mode: FMODE_* mask
1301  *
1302  *	Returns 0 if successful. Returns a negated errno value in case
1303  *	of error.
1304  *
1305  *	Note: This can be called from a user context (e.g. fsck(1) )
1306  *	or from within the kernel (e.g. as a result of a mount(1) ).
1307  *	In the latter case @inode and @filp carry an abridged amount
1308  *	of information as noted above.
1309  *
1310  *	Locking: called with bdev->bd_disk->open_mutex held.
1311  **/
1312 static int sd_open(struct block_device *bdev, fmode_t mode)
1313 {
1314 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1315 	struct scsi_device *sdev = sdkp->device;
1316 	int retval;
1317 
1318 	if (scsi_device_get(sdev))
1319 		return -ENXIO;
1320 
1321 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1322 
1323 	/*
1324 	 * If the device is in error recovery, wait until it is done.
1325 	 * If the device is offline, then disallow any access to it.
1326 	 */
1327 	retval = -ENXIO;
1328 	if (!scsi_block_when_processing_errors(sdev))
1329 		goto error_out;
1330 
1331 	if (sd_need_revalidate(bdev, sdkp))
1332 		sd_revalidate_disk(bdev->bd_disk);
1333 
1334 	/*
1335 	 * If the drive is empty, just let the open fail.
1336 	 */
1337 	retval = -ENOMEDIUM;
1338 	if (sdev->removable && !sdkp->media_present && !(mode & FMODE_NDELAY))
1339 		goto error_out;
1340 
1341 	/*
1342 	 * If the device has the write protect tab set, have the open fail
1343 	 * if the user expects to be able to write to the thing.
1344 	 */
1345 	retval = -EROFS;
1346 	if (sdkp->write_prot && (mode & FMODE_WRITE))
1347 		goto error_out;
1348 
1349 	/*
1350 	 * It is possible that the disk changing stuff resulted in
1351 	 * the device being taken offline.  If this is the case,
1352 	 * report this to the user, and don't pretend that the
1353 	 * open actually succeeded.
1354 	 */
1355 	retval = -ENXIO;
1356 	if (!scsi_device_online(sdev))
1357 		goto error_out;
1358 
1359 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1360 		if (scsi_block_when_processing_errors(sdev))
1361 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1362 	}
1363 
1364 	return 0;
1365 
1366 error_out:
1367 	scsi_device_put(sdev);
1368 	return retval;
1369 }
1370 
1371 /**
1372  *	sd_release - invoked when the (last) close(2) is called on this
1373  *	scsi disk.
1374  *	@disk: disk to release
1375  *	@mode: FMODE_* mask
1376  *
1377  *	Returns 0.
1378  *
1379  *	Note: may block (uninterruptible) if error recovery is underway
1380  *	on this disk.
1381  *
1382  *	Locking: called with bdev->bd_disk->open_mutex held.
1383  **/
1384 static void sd_release(struct gendisk *disk, fmode_t mode)
1385 {
1386 	struct scsi_disk *sdkp = scsi_disk(disk);
1387 	struct scsi_device *sdev = sdkp->device;
1388 
1389 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1390 
1391 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1392 		if (scsi_block_when_processing_errors(sdev))
1393 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1394 	}
1395 
1396 	scsi_device_put(sdev);
1397 }
1398 
1399 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1400 {
1401 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1402 	struct scsi_device *sdp = sdkp->device;
1403 	struct Scsi_Host *host = sdp->host;
1404 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1405 	int diskinfo[4];
1406 
1407 	/* default to most commonly used values */
1408 	diskinfo[0] = 0x40;	/* 1 << 6 */
1409 	diskinfo[1] = 0x20;	/* 1 << 5 */
1410 	diskinfo[2] = capacity >> 11;
1411 
1412 	/* override with calculated, extended default, or driver values */
1413 	if (host->hostt->bios_param)
1414 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1415 	else
1416 		scsicam_bios_param(bdev, capacity, diskinfo);
1417 
1418 	geo->heads = diskinfo[0];
1419 	geo->sectors = diskinfo[1];
1420 	geo->cylinders = diskinfo[2];
1421 	return 0;
1422 }
1423 
1424 /**
1425  *	sd_ioctl - process an ioctl
1426  *	@bdev: target block device
1427  *	@mode: FMODE_* mask
1428  *	@cmd: ioctl command number
1429  *	@arg: this is third argument given to ioctl(2) system call.
1430  *	Often contains a pointer.
1431  *
1432  *	Returns 0 if successful (some ioctls return positive numbers on
1433  *	success as well). Returns a negated errno value in case of error.
1434  *
1435  *	Note: most ioctls are forward onto the block subsystem or further
1436  *	down in the scsi subsystem.
1437  **/
1438 static int sd_ioctl(struct block_device *bdev, fmode_t mode,
1439 		    unsigned int cmd, unsigned long arg)
1440 {
1441 	struct gendisk *disk = bdev->bd_disk;
1442 	struct scsi_disk *sdkp = scsi_disk(disk);
1443 	struct scsi_device *sdp = sdkp->device;
1444 	void __user *p = (void __user *)arg;
1445 	int error;
1446 
1447 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1448 				    "cmd=0x%x\n", disk->disk_name, cmd));
1449 
1450 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1451 		return -ENOIOCTLCMD;
1452 
1453 	/*
1454 	 * If we are in the middle of error recovery, don't let anyone
1455 	 * else try and use this device.  Also, if error recovery fails, it
1456 	 * may try and take the device offline, in which case all further
1457 	 * access to the device is prohibited.
1458 	 */
1459 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1460 			(mode & FMODE_NDELAY) != 0);
1461 	if (error)
1462 		return error;
1463 
1464 	if (is_sed_ioctl(cmd))
1465 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1466 	return scsi_ioctl(sdp, mode, cmd, p);
1467 }
1468 
1469 static void set_media_not_present(struct scsi_disk *sdkp)
1470 {
1471 	if (sdkp->media_present)
1472 		sdkp->device->changed = 1;
1473 
1474 	if (sdkp->device->removable) {
1475 		sdkp->media_present = 0;
1476 		sdkp->capacity = 0;
1477 	}
1478 }
1479 
1480 static int media_not_present(struct scsi_disk *sdkp,
1481 			     struct scsi_sense_hdr *sshdr)
1482 {
1483 	if (!scsi_sense_valid(sshdr))
1484 		return 0;
1485 
1486 	/* not invoked for commands that could return deferred errors */
1487 	switch (sshdr->sense_key) {
1488 	case UNIT_ATTENTION:
1489 	case NOT_READY:
1490 		/* medium not present */
1491 		if (sshdr->asc == 0x3A) {
1492 			set_media_not_present(sdkp);
1493 			return 1;
1494 		}
1495 	}
1496 	return 0;
1497 }
1498 
1499 /**
1500  *	sd_check_events - check media events
1501  *	@disk: kernel device descriptor
1502  *	@clearing: disk events currently being cleared
1503  *
1504  *	Returns mask of DISK_EVENT_*.
1505  *
1506  *	Note: this function is invoked from the block subsystem.
1507  **/
1508 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1509 {
1510 	struct scsi_disk *sdkp = disk->private_data;
1511 	struct scsi_device *sdp;
1512 	int retval;
1513 	bool disk_changed;
1514 
1515 	if (!sdkp)
1516 		return 0;
1517 
1518 	sdp = sdkp->device;
1519 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1520 
1521 	/*
1522 	 * If the device is offline, don't send any commands - just pretend as
1523 	 * if the command failed.  If the device ever comes back online, we
1524 	 * can deal with it then.  It is only because of unrecoverable errors
1525 	 * that we would ever take a device offline in the first place.
1526 	 */
1527 	if (!scsi_device_online(sdp)) {
1528 		set_media_not_present(sdkp);
1529 		goto out;
1530 	}
1531 
1532 	/*
1533 	 * Using TEST_UNIT_READY enables differentiation between drive with
1534 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1535 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1536 	 *
1537 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1538 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1539 	 * sd_revalidate() is called.
1540 	 */
1541 	if (scsi_block_when_processing_errors(sdp)) {
1542 		struct scsi_sense_hdr sshdr = { 0, };
1543 
1544 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1545 					      &sshdr);
1546 
1547 		/* failed to execute TUR, assume media not present */
1548 		if (retval < 0 || host_byte(retval)) {
1549 			set_media_not_present(sdkp);
1550 			goto out;
1551 		}
1552 
1553 		if (media_not_present(sdkp, &sshdr))
1554 			goto out;
1555 	}
1556 
1557 	/*
1558 	 * For removable scsi disk we have to recognise the presence
1559 	 * of a disk in the drive.
1560 	 */
1561 	if (!sdkp->media_present)
1562 		sdp->changed = 1;
1563 	sdkp->media_present = 1;
1564 out:
1565 	/*
1566 	 * sdp->changed is set under the following conditions:
1567 	 *
1568 	 *	Medium present state has changed in either direction.
1569 	 *	Device has indicated UNIT_ATTENTION.
1570 	 */
1571 	disk_changed = sdp->changed;
1572 	sdp->changed = 0;
1573 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1574 }
1575 
1576 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1577 {
1578 	int retries, res;
1579 	struct scsi_device *sdp = sdkp->device;
1580 	const int timeout = sdp->request_queue->rq_timeout
1581 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1582 	struct scsi_sense_hdr my_sshdr;
1583 
1584 	if (!scsi_device_online(sdp))
1585 		return -ENODEV;
1586 
1587 	/* caller might not be interested in sense, but we need it */
1588 	if (!sshdr)
1589 		sshdr = &my_sshdr;
1590 
1591 	for (retries = 3; retries > 0; --retries) {
1592 		unsigned char cmd[10] = { 0 };
1593 
1594 		cmd[0] = SYNCHRONIZE_CACHE;
1595 		/*
1596 		 * Leave the rest of the command zero to indicate
1597 		 * flush everything.
1598 		 */
1599 		res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, sshdr,
1600 				timeout, sdkp->max_retries, 0, RQF_PM, NULL);
1601 		if (res == 0)
1602 			break;
1603 	}
1604 
1605 	if (res) {
1606 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1607 
1608 		if (res < 0)
1609 			return res;
1610 
1611 		if (scsi_status_is_check_condition(res) &&
1612 		    scsi_sense_valid(sshdr)) {
1613 			sd_print_sense_hdr(sdkp, sshdr);
1614 
1615 			/* we need to evaluate the error return  */
1616 			if (sshdr->asc == 0x3a ||	/* medium not present */
1617 			    sshdr->asc == 0x20 ||	/* invalid command */
1618 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1619 				/* this is no error here */
1620 				return 0;
1621 		}
1622 
1623 		switch (host_byte(res)) {
1624 		/* ignore errors due to racing a disconnection */
1625 		case DID_BAD_TARGET:
1626 		case DID_NO_CONNECT:
1627 			return 0;
1628 		/* signal the upper layer it might try again */
1629 		case DID_BUS_BUSY:
1630 		case DID_IMM_RETRY:
1631 		case DID_REQUEUE:
1632 		case DID_SOFT_ERROR:
1633 			return -EBUSY;
1634 		default:
1635 			return -EIO;
1636 		}
1637 	}
1638 	return 0;
1639 }
1640 
1641 static void sd_rescan(struct device *dev)
1642 {
1643 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1644 
1645 	sd_revalidate_disk(sdkp->disk);
1646 }
1647 
1648 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1649 		enum blk_unique_id type)
1650 {
1651 	struct scsi_device *sdev = scsi_disk(disk)->device;
1652 	const struct scsi_vpd *vpd;
1653 	const unsigned char *d;
1654 	int ret = -ENXIO, len;
1655 
1656 	rcu_read_lock();
1657 	vpd = rcu_dereference(sdev->vpd_pg83);
1658 	if (!vpd)
1659 		goto out_unlock;
1660 
1661 	ret = -EINVAL;
1662 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1663 		/* we only care about designators with LU association */
1664 		if (((d[1] >> 4) & 0x3) != 0x00)
1665 			continue;
1666 		if ((d[1] & 0xf) != type)
1667 			continue;
1668 
1669 		/*
1670 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1671 		 * keep looking as one with more entropy might still show up.
1672 		 */
1673 		len = d[3];
1674 		if (len != 8 && len != 12 && len != 16)
1675 			continue;
1676 		ret = len;
1677 		memcpy(id, d + 4, len);
1678 		if (len == 16)
1679 			break;
1680 	}
1681 out_unlock:
1682 	rcu_read_unlock();
1683 	return ret;
1684 }
1685 
1686 static char sd_pr_type(enum pr_type type)
1687 {
1688 	switch (type) {
1689 	case PR_WRITE_EXCLUSIVE:
1690 		return 0x01;
1691 	case PR_EXCLUSIVE_ACCESS:
1692 		return 0x03;
1693 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1694 		return 0x05;
1695 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1696 		return 0x06;
1697 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1698 		return 0x07;
1699 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1700 		return 0x08;
1701 	default:
1702 		return 0;
1703 	}
1704 };
1705 
1706 static int sd_pr_command(struct block_device *bdev, u8 sa,
1707 		u64 key, u64 sa_key, u8 type, u8 flags)
1708 {
1709 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1710 	struct scsi_device *sdev = sdkp->device;
1711 	struct scsi_sense_hdr sshdr;
1712 	int result;
1713 	u8 cmd[16] = { 0, };
1714 	u8 data[24] = { 0, };
1715 
1716 	cmd[0] = PERSISTENT_RESERVE_OUT;
1717 	cmd[1] = sa;
1718 	cmd[2] = type;
1719 	put_unaligned_be32(sizeof(data), &cmd[5]);
1720 
1721 	put_unaligned_be64(key, &data[0]);
1722 	put_unaligned_be64(sa_key, &data[8]);
1723 	data[20] = flags;
1724 
1725 	result = scsi_execute_req(sdev, cmd, DMA_TO_DEVICE, &data, sizeof(data),
1726 			&sshdr, SD_TIMEOUT, sdkp->max_retries, NULL);
1727 
1728 	if (scsi_status_is_check_condition(result) &&
1729 	    scsi_sense_valid(&sshdr)) {
1730 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1731 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1732 	}
1733 
1734 	return result;
1735 }
1736 
1737 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1738 		u32 flags)
1739 {
1740 	if (flags & ~PR_FL_IGNORE_KEY)
1741 		return -EOPNOTSUPP;
1742 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1743 			old_key, new_key, 0,
1744 			(1 << 0) /* APTPL */);
1745 }
1746 
1747 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1748 		u32 flags)
1749 {
1750 	if (flags)
1751 		return -EOPNOTSUPP;
1752 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1753 }
1754 
1755 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1756 {
1757 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1758 }
1759 
1760 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1761 		enum pr_type type, bool abort)
1762 {
1763 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1764 			     sd_pr_type(type), 0);
1765 }
1766 
1767 static int sd_pr_clear(struct block_device *bdev, u64 key)
1768 {
1769 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1770 }
1771 
1772 static const struct pr_ops sd_pr_ops = {
1773 	.pr_register	= sd_pr_register,
1774 	.pr_reserve	= sd_pr_reserve,
1775 	.pr_release	= sd_pr_release,
1776 	.pr_preempt	= sd_pr_preempt,
1777 	.pr_clear	= sd_pr_clear,
1778 };
1779 
1780 static void scsi_disk_free_disk(struct gendisk *disk)
1781 {
1782 	struct scsi_disk *sdkp = scsi_disk(disk);
1783 
1784 	put_device(&sdkp->disk_dev);
1785 }
1786 
1787 static const struct block_device_operations sd_fops = {
1788 	.owner			= THIS_MODULE,
1789 	.open			= sd_open,
1790 	.release		= sd_release,
1791 	.ioctl			= sd_ioctl,
1792 	.getgeo			= sd_getgeo,
1793 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1794 	.check_events		= sd_check_events,
1795 	.unlock_native_capacity	= sd_unlock_native_capacity,
1796 	.report_zones		= sd_zbc_report_zones,
1797 	.get_unique_id		= sd_get_unique_id,
1798 	.free_disk		= scsi_disk_free_disk,
1799 	.pr_ops			= &sd_pr_ops,
1800 };
1801 
1802 /**
1803  *	sd_eh_reset - reset error handling callback
1804  *	@scmd:		sd-issued command that has failed
1805  *
1806  *	This function is called by the SCSI midlayer before starting
1807  *	SCSI EH. When counting medium access failures we have to be
1808  *	careful to register it only only once per device and SCSI EH run;
1809  *	there might be several timed out commands which will cause the
1810  *	'max_medium_access_timeouts' counter to trigger after the first
1811  *	SCSI EH run already and set the device to offline.
1812  *	So this function resets the internal counter before starting SCSI EH.
1813  **/
1814 static void sd_eh_reset(struct scsi_cmnd *scmd)
1815 {
1816 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1817 
1818 	/* New SCSI EH run, reset gate variable */
1819 	sdkp->ignore_medium_access_errors = false;
1820 }
1821 
1822 /**
1823  *	sd_eh_action - error handling callback
1824  *	@scmd:		sd-issued command that has failed
1825  *	@eh_disp:	The recovery disposition suggested by the midlayer
1826  *
1827  *	This function is called by the SCSI midlayer upon completion of an
1828  *	error test command (currently TEST UNIT READY). The result of sending
1829  *	the eh command is passed in eh_disp.  We're looking for devices that
1830  *	fail medium access commands but are OK with non access commands like
1831  *	test unit ready (so wrongly see the device as having a successful
1832  *	recovery)
1833  **/
1834 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1835 {
1836 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1837 	struct scsi_device *sdev = scmd->device;
1838 
1839 	if (!scsi_device_online(sdev) ||
1840 	    !scsi_medium_access_command(scmd) ||
1841 	    host_byte(scmd->result) != DID_TIME_OUT ||
1842 	    eh_disp != SUCCESS)
1843 		return eh_disp;
1844 
1845 	/*
1846 	 * The device has timed out executing a medium access command.
1847 	 * However, the TEST UNIT READY command sent during error
1848 	 * handling completed successfully. Either the device is in the
1849 	 * process of recovering or has it suffered an internal failure
1850 	 * that prevents access to the storage medium.
1851 	 */
1852 	if (!sdkp->ignore_medium_access_errors) {
1853 		sdkp->medium_access_timed_out++;
1854 		sdkp->ignore_medium_access_errors = true;
1855 	}
1856 
1857 	/*
1858 	 * If the device keeps failing read/write commands but TEST UNIT
1859 	 * READY always completes successfully we assume that medium
1860 	 * access is no longer possible and take the device offline.
1861 	 */
1862 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1863 		scmd_printk(KERN_ERR, scmd,
1864 			    "Medium access timeout failure. Offlining disk!\n");
1865 		mutex_lock(&sdev->state_mutex);
1866 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1867 		mutex_unlock(&sdev->state_mutex);
1868 
1869 		return SUCCESS;
1870 	}
1871 
1872 	return eh_disp;
1873 }
1874 
1875 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1876 {
1877 	struct request *req = scsi_cmd_to_rq(scmd);
1878 	struct scsi_device *sdev = scmd->device;
1879 	unsigned int transferred, good_bytes;
1880 	u64 start_lba, end_lba, bad_lba;
1881 
1882 	/*
1883 	 * Some commands have a payload smaller than the device logical
1884 	 * block size (e.g. INQUIRY on a 4K disk).
1885 	 */
1886 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1887 		return 0;
1888 
1889 	/* Check if we have a 'bad_lba' information */
1890 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1891 				     SCSI_SENSE_BUFFERSIZE,
1892 				     &bad_lba))
1893 		return 0;
1894 
1895 	/*
1896 	 * If the bad lba was reported incorrectly, we have no idea where
1897 	 * the error is.
1898 	 */
1899 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1900 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1901 	if (bad_lba < start_lba || bad_lba >= end_lba)
1902 		return 0;
1903 
1904 	/*
1905 	 * resid is optional but mostly filled in.  When it's unused,
1906 	 * its value is zero, so we assume the whole buffer transferred
1907 	 */
1908 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1909 
1910 	/* This computation should always be done in terms of the
1911 	 * resolution of the device's medium.
1912 	 */
1913 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1914 
1915 	return min(good_bytes, transferred);
1916 }
1917 
1918 /**
1919  *	sd_done - bottom half handler: called when the lower level
1920  *	driver has completed (successfully or otherwise) a scsi command.
1921  *	@SCpnt: mid-level's per command structure.
1922  *
1923  *	Note: potentially run from within an ISR. Must not block.
1924  **/
1925 static int sd_done(struct scsi_cmnd *SCpnt)
1926 {
1927 	int result = SCpnt->result;
1928 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1929 	unsigned int sector_size = SCpnt->device->sector_size;
1930 	unsigned int resid;
1931 	struct scsi_sense_hdr sshdr;
1932 	struct request *req = scsi_cmd_to_rq(SCpnt);
1933 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1934 	int sense_valid = 0;
1935 	int sense_deferred = 0;
1936 
1937 	switch (req_op(req)) {
1938 	case REQ_OP_DISCARD:
1939 	case REQ_OP_WRITE_ZEROES:
1940 	case REQ_OP_ZONE_RESET:
1941 	case REQ_OP_ZONE_RESET_ALL:
1942 	case REQ_OP_ZONE_OPEN:
1943 	case REQ_OP_ZONE_CLOSE:
1944 	case REQ_OP_ZONE_FINISH:
1945 		if (!result) {
1946 			good_bytes = blk_rq_bytes(req);
1947 			scsi_set_resid(SCpnt, 0);
1948 		} else {
1949 			good_bytes = 0;
1950 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1951 		}
1952 		break;
1953 	default:
1954 		/*
1955 		 * In case of bogus fw or device, we could end up having
1956 		 * an unaligned partial completion. Check this here and force
1957 		 * alignment.
1958 		 */
1959 		resid = scsi_get_resid(SCpnt);
1960 		if (resid & (sector_size - 1)) {
1961 			sd_printk(KERN_INFO, sdkp,
1962 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
1963 				resid, sector_size);
1964 			scsi_print_command(SCpnt);
1965 			resid = min(scsi_bufflen(SCpnt),
1966 				    round_up(resid, sector_size));
1967 			scsi_set_resid(SCpnt, resid);
1968 		}
1969 	}
1970 
1971 	if (result) {
1972 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
1973 		if (sense_valid)
1974 			sense_deferred = scsi_sense_is_deferred(&sshdr);
1975 	}
1976 	sdkp->medium_access_timed_out = 0;
1977 
1978 	if (!scsi_status_is_check_condition(result) &&
1979 	    (!sense_valid || sense_deferred))
1980 		goto out;
1981 
1982 	switch (sshdr.sense_key) {
1983 	case HARDWARE_ERROR:
1984 	case MEDIUM_ERROR:
1985 		good_bytes = sd_completed_bytes(SCpnt);
1986 		break;
1987 	case RECOVERED_ERROR:
1988 		good_bytes = scsi_bufflen(SCpnt);
1989 		break;
1990 	case NO_SENSE:
1991 		/* This indicates a false check condition, so ignore it.  An
1992 		 * unknown amount of data was transferred so treat it as an
1993 		 * error.
1994 		 */
1995 		SCpnt->result = 0;
1996 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
1997 		break;
1998 	case ABORTED_COMMAND:
1999 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2000 			good_bytes = sd_completed_bytes(SCpnt);
2001 		break;
2002 	case ILLEGAL_REQUEST:
2003 		switch (sshdr.asc) {
2004 		case 0x10:	/* DIX: Host detected corruption */
2005 			good_bytes = sd_completed_bytes(SCpnt);
2006 			break;
2007 		case 0x20:	/* INVALID COMMAND OPCODE */
2008 		case 0x24:	/* INVALID FIELD IN CDB */
2009 			switch (SCpnt->cmnd[0]) {
2010 			case UNMAP:
2011 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2012 				break;
2013 			case WRITE_SAME_16:
2014 			case WRITE_SAME:
2015 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2016 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2017 				} else {
2018 					sdkp->device->no_write_same = 1;
2019 					sd_config_write_same(sdkp);
2020 					req->rq_flags |= RQF_QUIET;
2021 				}
2022 				break;
2023 			}
2024 		}
2025 		break;
2026 	default:
2027 		break;
2028 	}
2029 
2030  out:
2031 	if (sd_is_zoned(sdkp))
2032 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2033 
2034 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2035 					   "sd_done: completed %d of %d bytes\n",
2036 					   good_bytes, scsi_bufflen(SCpnt)));
2037 
2038 	return good_bytes;
2039 }
2040 
2041 /*
2042  * spinup disk - called only in sd_revalidate_disk()
2043  */
2044 static void
2045 sd_spinup_disk(struct scsi_disk *sdkp)
2046 {
2047 	unsigned char cmd[10];
2048 	unsigned long spintime_expire = 0;
2049 	int retries, spintime;
2050 	unsigned int the_result;
2051 	struct scsi_sense_hdr sshdr;
2052 	int sense_valid = 0;
2053 
2054 	spintime = 0;
2055 
2056 	/* Spin up drives, as required.  Only do this at boot time */
2057 	/* Spinup needs to be done for module loads too. */
2058 	do {
2059 		retries = 0;
2060 
2061 		do {
2062 			bool media_was_present = sdkp->media_present;
2063 
2064 			cmd[0] = TEST_UNIT_READY;
2065 			memset((void *) &cmd[1], 0, 9);
2066 
2067 			the_result = scsi_execute_req(sdkp->device, cmd,
2068 						      DMA_NONE, NULL, 0,
2069 						      &sshdr, SD_TIMEOUT,
2070 						      sdkp->max_retries, NULL);
2071 
2072 			/*
2073 			 * If the drive has indicated to us that it
2074 			 * doesn't have any media in it, don't bother
2075 			 * with any more polling.
2076 			 */
2077 			if (media_not_present(sdkp, &sshdr)) {
2078 				if (media_was_present)
2079 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2080 				return;
2081 			}
2082 
2083 			if (the_result)
2084 				sense_valid = scsi_sense_valid(&sshdr);
2085 			retries++;
2086 		} while (retries < 3 &&
2087 			 (!scsi_status_is_good(the_result) ||
2088 			  (scsi_status_is_check_condition(the_result) &&
2089 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2090 
2091 		if (!scsi_status_is_check_condition(the_result)) {
2092 			/* no sense, TUR either succeeded or failed
2093 			 * with a status error */
2094 			if(!spintime && !scsi_status_is_good(the_result)) {
2095 				sd_print_result(sdkp, "Test Unit Ready failed",
2096 						the_result);
2097 			}
2098 			break;
2099 		}
2100 
2101 		/*
2102 		 * The device does not want the automatic start to be issued.
2103 		 */
2104 		if (sdkp->device->no_start_on_add)
2105 			break;
2106 
2107 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2108 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2109 				break;	/* manual intervention required */
2110 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2111 				break;	/* standby */
2112 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2113 				break;	/* unavailable */
2114 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2115 				break;	/* sanitize in progress */
2116 			/*
2117 			 * Issue command to spin up drive when not ready
2118 			 */
2119 			if (!spintime) {
2120 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2121 				cmd[0] = START_STOP;
2122 				cmd[1] = 1;	/* Return immediately */
2123 				memset((void *) &cmd[2], 0, 8);
2124 				cmd[4] = 1;	/* Start spin cycle */
2125 				if (sdkp->device->start_stop_pwr_cond)
2126 					cmd[4] |= 1 << 4;
2127 				scsi_execute_req(sdkp->device, cmd, DMA_NONE,
2128 						 NULL, 0, &sshdr,
2129 						 SD_TIMEOUT, sdkp->max_retries,
2130 						 NULL);
2131 				spintime_expire = jiffies + 100 * HZ;
2132 				spintime = 1;
2133 			}
2134 			/* Wait 1 second for next try */
2135 			msleep(1000);
2136 			printk(KERN_CONT ".");
2137 
2138 		/*
2139 		 * Wait for USB flash devices with slow firmware.
2140 		 * Yes, this sense key/ASC combination shouldn't
2141 		 * occur here.  It's characteristic of these devices.
2142 		 */
2143 		} else if (sense_valid &&
2144 				sshdr.sense_key == UNIT_ATTENTION &&
2145 				sshdr.asc == 0x28) {
2146 			if (!spintime) {
2147 				spintime_expire = jiffies + 5 * HZ;
2148 				spintime = 1;
2149 			}
2150 			/* Wait 1 second for next try */
2151 			msleep(1000);
2152 		} else {
2153 			/* we don't understand the sense code, so it's
2154 			 * probably pointless to loop */
2155 			if(!spintime) {
2156 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2157 				sd_print_sense_hdr(sdkp, &sshdr);
2158 			}
2159 			break;
2160 		}
2161 
2162 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2163 
2164 	if (spintime) {
2165 		if (scsi_status_is_good(the_result))
2166 			printk(KERN_CONT "ready\n");
2167 		else
2168 			printk(KERN_CONT "not responding...\n");
2169 	}
2170 }
2171 
2172 /*
2173  * Determine whether disk supports Data Integrity Field.
2174  */
2175 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2176 {
2177 	struct scsi_device *sdp = sdkp->device;
2178 	u8 type;
2179 
2180 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2181 		sdkp->protection_type = 0;
2182 		return 0;
2183 	}
2184 
2185 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2186 
2187 	if (type > T10_PI_TYPE3_PROTECTION) {
2188 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2189 			  " protection type %u. Disabling disk!\n",
2190 			  type);
2191 		sdkp->protection_type = 0;
2192 		return -ENODEV;
2193 	}
2194 
2195 	sdkp->protection_type = type;
2196 
2197 	return 0;
2198 }
2199 
2200 static void sd_config_protection(struct scsi_disk *sdkp)
2201 {
2202 	struct scsi_device *sdp = sdkp->device;
2203 
2204 	if (!sdkp->first_scan)
2205 		return;
2206 
2207 	sd_dif_config_host(sdkp);
2208 
2209 	if (!sdkp->protection_type)
2210 		return;
2211 
2212 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2213 		sd_printk(KERN_NOTICE, sdkp,
2214 			  "Disabling DIF Type %u protection\n",
2215 			  sdkp->protection_type);
2216 		sdkp->protection_type = 0;
2217 	}
2218 
2219 	sd_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2220 		  sdkp->protection_type);
2221 }
2222 
2223 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2224 			struct scsi_sense_hdr *sshdr, int sense_valid,
2225 			int the_result)
2226 {
2227 	if (sense_valid)
2228 		sd_print_sense_hdr(sdkp, sshdr);
2229 	else
2230 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2231 
2232 	/*
2233 	 * Set dirty bit for removable devices if not ready -
2234 	 * sometimes drives will not report this properly.
2235 	 */
2236 	if (sdp->removable &&
2237 	    sense_valid && sshdr->sense_key == NOT_READY)
2238 		set_media_not_present(sdkp);
2239 
2240 	/*
2241 	 * We used to set media_present to 0 here to indicate no media
2242 	 * in the drive, but some drives fail read capacity even with
2243 	 * media present, so we can't do that.
2244 	 */
2245 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2246 }
2247 
2248 #define RC16_LEN 32
2249 #if RC16_LEN > SD_BUF_SIZE
2250 #error RC16_LEN must not be more than SD_BUF_SIZE
2251 #endif
2252 
2253 #define READ_CAPACITY_RETRIES_ON_RESET	10
2254 
2255 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2256 						unsigned char *buffer)
2257 {
2258 	unsigned char cmd[16];
2259 	struct scsi_sense_hdr sshdr;
2260 	int sense_valid = 0;
2261 	int the_result;
2262 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2263 	unsigned int alignment;
2264 	unsigned long long lba;
2265 	unsigned sector_size;
2266 
2267 	if (sdp->no_read_capacity_16)
2268 		return -EINVAL;
2269 
2270 	do {
2271 		memset(cmd, 0, 16);
2272 		cmd[0] = SERVICE_ACTION_IN_16;
2273 		cmd[1] = SAI_READ_CAPACITY_16;
2274 		cmd[13] = RC16_LEN;
2275 		memset(buffer, 0, RC16_LEN);
2276 
2277 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2278 					buffer, RC16_LEN, &sshdr,
2279 					SD_TIMEOUT, sdkp->max_retries, NULL);
2280 
2281 		if (media_not_present(sdkp, &sshdr))
2282 			return -ENODEV;
2283 
2284 		if (the_result > 0) {
2285 			sense_valid = scsi_sense_valid(&sshdr);
2286 			if (sense_valid &&
2287 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2288 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2289 			    sshdr.ascq == 0x00)
2290 				/* Invalid Command Operation Code or
2291 				 * Invalid Field in CDB, just retry
2292 				 * silently with RC10 */
2293 				return -EINVAL;
2294 			if (sense_valid &&
2295 			    sshdr.sense_key == UNIT_ATTENTION &&
2296 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2297 				/* Device reset might occur several times,
2298 				 * give it one more chance */
2299 				if (--reset_retries > 0)
2300 					continue;
2301 		}
2302 		retries--;
2303 
2304 	} while (the_result && retries);
2305 
2306 	if (the_result) {
2307 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2308 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2309 		return -EINVAL;
2310 	}
2311 
2312 	sector_size = get_unaligned_be32(&buffer[8]);
2313 	lba = get_unaligned_be64(&buffer[0]);
2314 
2315 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2316 		sdkp->capacity = 0;
2317 		return -ENODEV;
2318 	}
2319 
2320 	/* Logical blocks per physical block exponent */
2321 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2322 
2323 	/* RC basis */
2324 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2325 
2326 	/* Lowest aligned logical block */
2327 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2328 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2329 	if (alignment && sdkp->first_scan)
2330 		sd_printk(KERN_NOTICE, sdkp,
2331 			  "physical block alignment offset: %u\n", alignment);
2332 
2333 	if (buffer[14] & 0x80) { /* LBPME */
2334 		sdkp->lbpme = 1;
2335 
2336 		if (buffer[14] & 0x40) /* LBPRZ */
2337 			sdkp->lbprz = 1;
2338 
2339 		sd_config_discard(sdkp, SD_LBP_WS16);
2340 	}
2341 
2342 	sdkp->capacity = lba + 1;
2343 	return sector_size;
2344 }
2345 
2346 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2347 						unsigned char *buffer)
2348 {
2349 	unsigned char cmd[16];
2350 	struct scsi_sense_hdr sshdr;
2351 	int sense_valid = 0;
2352 	int the_result;
2353 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2354 	sector_t lba;
2355 	unsigned sector_size;
2356 
2357 	do {
2358 		cmd[0] = READ_CAPACITY;
2359 		memset(&cmd[1], 0, 9);
2360 		memset(buffer, 0, 8);
2361 
2362 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2363 					buffer, 8, &sshdr,
2364 					SD_TIMEOUT, sdkp->max_retries, NULL);
2365 
2366 		if (media_not_present(sdkp, &sshdr))
2367 			return -ENODEV;
2368 
2369 		if (the_result > 0) {
2370 			sense_valid = scsi_sense_valid(&sshdr);
2371 			if (sense_valid &&
2372 			    sshdr.sense_key == UNIT_ATTENTION &&
2373 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2374 				/* Device reset might occur several times,
2375 				 * give it one more chance */
2376 				if (--reset_retries > 0)
2377 					continue;
2378 		}
2379 		retries--;
2380 
2381 	} while (the_result && retries);
2382 
2383 	if (the_result) {
2384 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2385 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2386 		return -EINVAL;
2387 	}
2388 
2389 	sector_size = get_unaligned_be32(&buffer[4]);
2390 	lba = get_unaligned_be32(&buffer[0]);
2391 
2392 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2393 		/* Some buggy (usb cardreader) devices return an lba of
2394 		   0xffffffff when the want to report a size of 0 (with
2395 		   which they really mean no media is present) */
2396 		sdkp->capacity = 0;
2397 		sdkp->physical_block_size = sector_size;
2398 		return sector_size;
2399 	}
2400 
2401 	sdkp->capacity = lba + 1;
2402 	sdkp->physical_block_size = sector_size;
2403 	return sector_size;
2404 }
2405 
2406 static int sd_try_rc16_first(struct scsi_device *sdp)
2407 {
2408 	if (sdp->host->max_cmd_len < 16)
2409 		return 0;
2410 	if (sdp->try_rc_10_first)
2411 		return 0;
2412 	if (sdp->scsi_level > SCSI_SPC_2)
2413 		return 1;
2414 	if (scsi_device_protection(sdp))
2415 		return 1;
2416 	return 0;
2417 }
2418 
2419 /*
2420  * read disk capacity
2421  */
2422 static void
2423 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2424 {
2425 	int sector_size;
2426 	struct scsi_device *sdp = sdkp->device;
2427 
2428 	if (sd_try_rc16_first(sdp)) {
2429 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2430 		if (sector_size == -EOVERFLOW)
2431 			goto got_data;
2432 		if (sector_size == -ENODEV)
2433 			return;
2434 		if (sector_size < 0)
2435 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2436 		if (sector_size < 0)
2437 			return;
2438 	} else {
2439 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2440 		if (sector_size == -EOVERFLOW)
2441 			goto got_data;
2442 		if (sector_size < 0)
2443 			return;
2444 		if ((sizeof(sdkp->capacity) > 4) &&
2445 		    (sdkp->capacity > 0xffffffffULL)) {
2446 			int old_sector_size = sector_size;
2447 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2448 					"Trying to use READ CAPACITY(16).\n");
2449 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2450 			if (sector_size < 0) {
2451 				sd_printk(KERN_NOTICE, sdkp,
2452 					"Using 0xffffffff as device size\n");
2453 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2454 				sector_size = old_sector_size;
2455 				goto got_data;
2456 			}
2457 			/* Remember that READ CAPACITY(16) succeeded */
2458 			sdp->try_rc_10_first = 0;
2459 		}
2460 	}
2461 
2462 	/* Some devices are known to return the total number of blocks,
2463 	 * not the highest block number.  Some devices have versions
2464 	 * which do this and others which do not.  Some devices we might
2465 	 * suspect of doing this but we don't know for certain.
2466 	 *
2467 	 * If we know the reported capacity is wrong, decrement it.  If
2468 	 * we can only guess, then assume the number of blocks is even
2469 	 * (usually true but not always) and err on the side of lowering
2470 	 * the capacity.
2471 	 */
2472 	if (sdp->fix_capacity ||
2473 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2474 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2475 				"from its reported value: %llu\n",
2476 				(unsigned long long) sdkp->capacity);
2477 		--sdkp->capacity;
2478 	}
2479 
2480 got_data:
2481 	if (sector_size == 0) {
2482 		sector_size = 512;
2483 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2484 			  "assuming 512.\n");
2485 	}
2486 
2487 	if (sector_size != 512 &&
2488 	    sector_size != 1024 &&
2489 	    sector_size != 2048 &&
2490 	    sector_size != 4096) {
2491 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2492 			  sector_size);
2493 		/*
2494 		 * The user might want to re-format the drive with
2495 		 * a supported sectorsize.  Once this happens, it
2496 		 * would be relatively trivial to set the thing up.
2497 		 * For this reason, we leave the thing in the table.
2498 		 */
2499 		sdkp->capacity = 0;
2500 		/*
2501 		 * set a bogus sector size so the normal read/write
2502 		 * logic in the block layer will eventually refuse any
2503 		 * request on this device without tripping over power
2504 		 * of two sector size assumptions
2505 		 */
2506 		sector_size = 512;
2507 	}
2508 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2509 	blk_queue_physical_block_size(sdp->request_queue,
2510 				      sdkp->physical_block_size);
2511 	sdkp->device->sector_size = sector_size;
2512 
2513 	if (sdkp->capacity > 0xffffffff)
2514 		sdp->use_16_for_rw = 1;
2515 
2516 }
2517 
2518 /*
2519  * Print disk capacity
2520  */
2521 static void
2522 sd_print_capacity(struct scsi_disk *sdkp,
2523 		  sector_t old_capacity)
2524 {
2525 	int sector_size = sdkp->device->sector_size;
2526 	char cap_str_2[10], cap_str_10[10];
2527 
2528 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2529 		return;
2530 
2531 	string_get_size(sdkp->capacity, sector_size,
2532 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2533 	string_get_size(sdkp->capacity, sector_size,
2534 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2535 
2536 	sd_printk(KERN_NOTICE, sdkp,
2537 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2538 		  (unsigned long long)sdkp->capacity,
2539 		  sector_size, cap_str_10, cap_str_2);
2540 
2541 	if (sdkp->physical_block_size != sector_size)
2542 		sd_printk(KERN_NOTICE, sdkp,
2543 			  "%u-byte physical blocks\n",
2544 			  sdkp->physical_block_size);
2545 }
2546 
2547 /* called with buffer of length 512 */
2548 static inline int
2549 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2550 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2551 		 struct scsi_sense_hdr *sshdr)
2552 {
2553 	/*
2554 	 * If we must use MODE SENSE(10), make sure that the buffer length
2555 	 * is at least 8 bytes so that the mode sense header fits.
2556 	 */
2557 	if (sdkp->device->use_10_for_ms && len < 8)
2558 		len = 8;
2559 
2560 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2561 			       SD_TIMEOUT, sdkp->max_retries, data,
2562 			       sshdr);
2563 }
2564 
2565 /*
2566  * read write protect setting, if possible - called only in sd_revalidate_disk()
2567  * called with buffer of length SD_BUF_SIZE
2568  */
2569 static void
2570 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2571 {
2572 	int res;
2573 	struct scsi_device *sdp = sdkp->device;
2574 	struct scsi_mode_data data;
2575 	int old_wp = sdkp->write_prot;
2576 
2577 	set_disk_ro(sdkp->disk, 0);
2578 	if (sdp->skip_ms_page_3f) {
2579 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2580 		return;
2581 	}
2582 
2583 	if (sdp->use_192_bytes_for_3f) {
2584 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2585 	} else {
2586 		/*
2587 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2588 		 * We have to start carefully: some devices hang if we ask
2589 		 * for more than is available.
2590 		 */
2591 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2592 
2593 		/*
2594 		 * Second attempt: ask for page 0 When only page 0 is
2595 		 * implemented, a request for page 3F may return Sense Key
2596 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2597 		 * CDB.
2598 		 */
2599 		if (res < 0)
2600 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2601 
2602 		/*
2603 		 * Third attempt: ask 255 bytes, as we did earlier.
2604 		 */
2605 		if (res < 0)
2606 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2607 					       &data, NULL);
2608 	}
2609 
2610 	if (res < 0) {
2611 		sd_first_printk(KERN_WARNING, sdkp,
2612 			  "Test WP failed, assume Write Enabled\n");
2613 	} else {
2614 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2615 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2616 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2617 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2618 				  sdkp->write_prot ? "on" : "off");
2619 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2620 		}
2621 	}
2622 }
2623 
2624 /*
2625  * sd_read_cache_type - called only from sd_revalidate_disk()
2626  * called with buffer of length SD_BUF_SIZE
2627  */
2628 static void
2629 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2630 {
2631 	int len = 0, res;
2632 	struct scsi_device *sdp = sdkp->device;
2633 
2634 	int dbd;
2635 	int modepage;
2636 	int first_len;
2637 	struct scsi_mode_data data;
2638 	struct scsi_sense_hdr sshdr;
2639 	int old_wce = sdkp->WCE;
2640 	int old_rcd = sdkp->RCD;
2641 	int old_dpofua = sdkp->DPOFUA;
2642 
2643 
2644 	if (sdkp->cache_override)
2645 		return;
2646 
2647 	first_len = 4;
2648 	if (sdp->skip_ms_page_8) {
2649 		if (sdp->type == TYPE_RBC)
2650 			goto defaults;
2651 		else {
2652 			if (sdp->skip_ms_page_3f)
2653 				goto defaults;
2654 			modepage = 0x3F;
2655 			if (sdp->use_192_bytes_for_3f)
2656 				first_len = 192;
2657 			dbd = 0;
2658 		}
2659 	} else if (sdp->type == TYPE_RBC) {
2660 		modepage = 6;
2661 		dbd = 8;
2662 	} else {
2663 		modepage = 8;
2664 		dbd = 0;
2665 	}
2666 
2667 	/* cautiously ask */
2668 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2669 			&data, &sshdr);
2670 
2671 	if (res < 0)
2672 		goto bad_sense;
2673 
2674 	if (!data.header_length) {
2675 		modepage = 6;
2676 		first_len = 0;
2677 		sd_first_printk(KERN_ERR, sdkp,
2678 				"Missing header in MODE_SENSE response\n");
2679 	}
2680 
2681 	/* that went OK, now ask for the proper length */
2682 	len = data.length;
2683 
2684 	/*
2685 	 * We're only interested in the first three bytes, actually.
2686 	 * But the data cache page is defined for the first 20.
2687 	 */
2688 	if (len < 3)
2689 		goto bad_sense;
2690 	else if (len > SD_BUF_SIZE) {
2691 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2692 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2693 		len = SD_BUF_SIZE;
2694 	}
2695 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2696 		len = 192;
2697 
2698 	/* Get the data */
2699 	if (len > first_len)
2700 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2701 				&data, &sshdr);
2702 
2703 	if (!res) {
2704 		int offset = data.header_length + data.block_descriptor_length;
2705 
2706 		while (offset < len) {
2707 			u8 page_code = buffer[offset] & 0x3F;
2708 			u8 spf       = buffer[offset] & 0x40;
2709 
2710 			if (page_code == 8 || page_code == 6) {
2711 				/* We're interested only in the first 3 bytes.
2712 				 */
2713 				if (len - offset <= 2) {
2714 					sd_first_printk(KERN_ERR, sdkp,
2715 						"Incomplete mode parameter "
2716 							"data\n");
2717 					goto defaults;
2718 				} else {
2719 					modepage = page_code;
2720 					goto Page_found;
2721 				}
2722 			} else {
2723 				/* Go to the next page */
2724 				if (spf && len - offset > 3)
2725 					offset += 4 + (buffer[offset+2] << 8) +
2726 						buffer[offset+3];
2727 				else if (!spf && len - offset > 1)
2728 					offset += 2 + buffer[offset+1];
2729 				else {
2730 					sd_first_printk(KERN_ERR, sdkp,
2731 							"Incomplete mode "
2732 							"parameter data\n");
2733 					goto defaults;
2734 				}
2735 			}
2736 		}
2737 
2738 		sd_first_printk(KERN_WARNING, sdkp,
2739 				"No Caching mode page found\n");
2740 		goto defaults;
2741 
2742 	Page_found:
2743 		if (modepage == 8) {
2744 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2745 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2746 		} else {
2747 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2748 			sdkp->RCD = 0;
2749 		}
2750 
2751 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2752 		if (sdp->broken_fua) {
2753 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2754 			sdkp->DPOFUA = 0;
2755 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2756 			   !sdkp->device->use_16_for_rw) {
2757 			sd_first_printk(KERN_NOTICE, sdkp,
2758 				  "Uses READ/WRITE(6), disabling FUA\n");
2759 			sdkp->DPOFUA = 0;
2760 		}
2761 
2762 		/* No cache flush allowed for write protected devices */
2763 		if (sdkp->WCE && sdkp->write_prot)
2764 			sdkp->WCE = 0;
2765 
2766 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2767 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2768 			sd_printk(KERN_NOTICE, sdkp,
2769 				  "Write cache: %s, read cache: %s, %s\n",
2770 				  sdkp->WCE ? "enabled" : "disabled",
2771 				  sdkp->RCD ? "disabled" : "enabled",
2772 				  sdkp->DPOFUA ? "supports DPO and FUA"
2773 				  : "doesn't support DPO or FUA");
2774 
2775 		return;
2776 	}
2777 
2778 bad_sense:
2779 	if (scsi_sense_valid(&sshdr) &&
2780 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2781 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2782 		/* Invalid field in CDB */
2783 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2784 	else
2785 		sd_first_printk(KERN_ERR, sdkp,
2786 				"Asking for cache data failed\n");
2787 
2788 defaults:
2789 	if (sdp->wce_default_on) {
2790 		sd_first_printk(KERN_NOTICE, sdkp,
2791 				"Assuming drive cache: write back\n");
2792 		sdkp->WCE = 1;
2793 	} else {
2794 		sd_first_printk(KERN_WARNING, sdkp,
2795 				"Assuming drive cache: write through\n");
2796 		sdkp->WCE = 0;
2797 	}
2798 	sdkp->RCD = 0;
2799 	sdkp->DPOFUA = 0;
2800 }
2801 
2802 /*
2803  * The ATO bit indicates whether the DIF application tag is available
2804  * for use by the operating system.
2805  */
2806 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2807 {
2808 	int res, offset;
2809 	struct scsi_device *sdp = sdkp->device;
2810 	struct scsi_mode_data data;
2811 	struct scsi_sense_hdr sshdr;
2812 
2813 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2814 		return;
2815 
2816 	if (sdkp->protection_type == 0)
2817 		return;
2818 
2819 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2820 			      sdkp->max_retries, &data, &sshdr);
2821 
2822 	if (res < 0 || !data.header_length ||
2823 	    data.length < 6) {
2824 		sd_first_printk(KERN_WARNING, sdkp,
2825 			  "getting Control mode page failed, assume no ATO\n");
2826 
2827 		if (scsi_sense_valid(&sshdr))
2828 			sd_print_sense_hdr(sdkp, &sshdr);
2829 
2830 		return;
2831 	}
2832 
2833 	offset = data.header_length + data.block_descriptor_length;
2834 
2835 	if ((buffer[offset] & 0x3f) != 0x0a) {
2836 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2837 		return;
2838 	}
2839 
2840 	if ((buffer[offset + 5] & 0x80) == 0)
2841 		return;
2842 
2843 	sdkp->ATO = 1;
2844 
2845 	return;
2846 }
2847 
2848 /**
2849  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2850  * @sdkp: disk to query
2851  */
2852 static void sd_read_block_limits(struct scsi_disk *sdkp)
2853 {
2854 	struct scsi_vpd *vpd;
2855 
2856 	rcu_read_lock();
2857 
2858 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2859 	if (!vpd || vpd->len < 16)
2860 		goto out;
2861 
2862 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2863 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2864 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2865 
2866 	if (vpd->len >= 64) {
2867 		unsigned int lba_count, desc_count;
2868 
2869 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2870 
2871 		if (!sdkp->lbpme)
2872 			goto out;
2873 
2874 		lba_count = get_unaligned_be32(&vpd->data[20]);
2875 		desc_count = get_unaligned_be32(&vpd->data[24]);
2876 
2877 		if (lba_count && desc_count)
2878 			sdkp->max_unmap_blocks = lba_count;
2879 
2880 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2881 
2882 		if (vpd->data[32] & 0x80)
2883 			sdkp->unmap_alignment =
2884 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2885 
2886 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2887 
2888 			if (sdkp->max_unmap_blocks)
2889 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2890 			else
2891 				sd_config_discard(sdkp, SD_LBP_WS16);
2892 
2893 		} else {	/* LBP VPD page tells us what to use */
2894 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2895 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2896 			else if (sdkp->lbpws)
2897 				sd_config_discard(sdkp, SD_LBP_WS16);
2898 			else if (sdkp->lbpws10)
2899 				sd_config_discard(sdkp, SD_LBP_WS10);
2900 			else
2901 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2902 		}
2903 	}
2904 
2905  out:
2906 	rcu_read_unlock();
2907 }
2908 
2909 /**
2910  * sd_read_block_characteristics - Query block dev. characteristics
2911  * @sdkp: disk to query
2912  */
2913 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2914 {
2915 	struct request_queue *q = sdkp->disk->queue;
2916 	struct scsi_vpd *vpd;
2917 	u16 rot;
2918 	u8 zoned;
2919 
2920 	rcu_read_lock();
2921 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
2922 
2923 	if (!vpd || vpd->len < 8) {
2924 		rcu_read_unlock();
2925 	        return;
2926 	}
2927 
2928 	rot = get_unaligned_be16(&vpd->data[4]);
2929 	zoned = (vpd->data[8] >> 4) & 3;
2930 	rcu_read_unlock();
2931 
2932 	if (rot == 1) {
2933 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2934 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2935 	}
2936 
2937 	if (sdkp->device->type == TYPE_ZBC) {
2938 		/* Host-managed */
2939 		blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HM);
2940 	} else {
2941 		sdkp->zoned = zoned;
2942 		if (sdkp->zoned == 1) {
2943 			/* Host-aware */
2944 			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HA);
2945 		} else {
2946 			/* Regular disk or drive managed disk */
2947 			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_NONE);
2948 		}
2949 	}
2950 
2951 	if (!sdkp->first_scan)
2952 		return;
2953 
2954 	if (blk_queue_is_zoned(q)) {
2955 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
2956 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
2957 	} else {
2958 		if (sdkp->zoned == 1)
2959 			sd_printk(KERN_NOTICE, sdkp,
2960 				  "Host-aware SMR disk used as regular disk\n");
2961 		else if (sdkp->zoned == 2)
2962 			sd_printk(KERN_NOTICE, sdkp,
2963 				  "Drive-managed SMR disk\n");
2964 	}
2965 }
2966 
2967 /**
2968  * sd_read_block_provisioning - Query provisioning VPD page
2969  * @sdkp: disk to query
2970  */
2971 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
2972 {
2973 	struct scsi_vpd *vpd;
2974 
2975 	if (sdkp->lbpme == 0)
2976 		return;
2977 
2978 	rcu_read_lock();
2979 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
2980 
2981 	if (!vpd || vpd->len < 8) {
2982 		rcu_read_unlock();
2983 		return;
2984 	}
2985 
2986 	sdkp->lbpvpd	= 1;
2987 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
2988 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
2989 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
2990 	rcu_read_unlock();
2991 }
2992 
2993 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
2994 {
2995 	struct scsi_device *sdev = sdkp->device;
2996 
2997 	if (sdev->host->no_write_same) {
2998 		sdev->no_write_same = 1;
2999 
3000 		return;
3001 	}
3002 
3003 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
3004 		struct scsi_vpd *vpd;
3005 
3006 		sdev->no_report_opcodes = 1;
3007 
3008 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3009 		 * CODES is unsupported and the device has an ATA
3010 		 * Information VPD page (SAT).
3011 		 */
3012 		rcu_read_lock();
3013 		vpd = rcu_dereference(sdev->vpd_pg89);
3014 		if (vpd)
3015 			sdev->no_write_same = 1;
3016 		rcu_read_unlock();
3017 	}
3018 
3019 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3020 		sdkp->ws16 = 1;
3021 
3022 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3023 		sdkp->ws10 = 1;
3024 }
3025 
3026 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3027 {
3028 	struct scsi_device *sdev = sdkp->device;
3029 
3030 	if (!sdev->security_supported)
3031 		return;
3032 
3033 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3034 			SECURITY_PROTOCOL_IN) == 1 &&
3035 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3036 			SECURITY_PROTOCOL_OUT) == 1)
3037 		sdkp->security = 1;
3038 }
3039 
3040 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3041 {
3042 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3043 }
3044 
3045 /**
3046  * sd_read_cpr - Query concurrent positioning ranges
3047  * @sdkp:	disk to query
3048  */
3049 static void sd_read_cpr(struct scsi_disk *sdkp)
3050 {
3051 	struct blk_independent_access_ranges *iars = NULL;
3052 	unsigned char *buffer = NULL;
3053 	unsigned int nr_cpr = 0;
3054 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3055 	u8 *desc;
3056 
3057 	/*
3058 	 * We need to have the capacity set first for the block layer to be
3059 	 * able to check the ranges.
3060 	 */
3061 	if (sdkp->first_scan)
3062 		return;
3063 
3064 	if (!sdkp->capacity)
3065 		goto out;
3066 
3067 	/*
3068 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3069 	 * leading to a maximum page size of 64 + 256*32 bytes.
3070 	 */
3071 	buf_len = 64 + 256*32;
3072 	buffer = kmalloc(buf_len, GFP_KERNEL);
3073 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3074 		goto out;
3075 
3076 	/* We must have at least a 64B header and one 32B range descriptor */
3077 	vpd_len = get_unaligned_be16(&buffer[2]) + 3;
3078 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3079 		sd_printk(KERN_ERR, sdkp,
3080 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3081 		goto out;
3082 	}
3083 
3084 	nr_cpr = (vpd_len - 64) / 32;
3085 	if (nr_cpr == 1) {
3086 		nr_cpr = 0;
3087 		goto out;
3088 	}
3089 
3090 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3091 	if (!iars) {
3092 		nr_cpr = 0;
3093 		goto out;
3094 	}
3095 
3096 	desc = &buffer[64];
3097 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3098 		if (desc[0] != i) {
3099 			sd_printk(KERN_ERR, sdkp,
3100 				"Invalid Concurrent Positioning Range number\n");
3101 			nr_cpr = 0;
3102 			break;
3103 		}
3104 
3105 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3106 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3107 	}
3108 
3109 out:
3110 	disk_set_independent_access_ranges(sdkp->disk, iars);
3111 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3112 		sd_printk(KERN_NOTICE, sdkp,
3113 			  "%u concurrent positioning ranges\n", nr_cpr);
3114 		sdkp->nr_actuators = nr_cpr;
3115 	}
3116 
3117 	kfree(buffer);
3118 }
3119 
3120 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3121 {
3122 	struct scsi_device *sdp = sdkp->device;
3123 	unsigned int min_xfer_bytes =
3124 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3125 
3126 	if (sdkp->min_xfer_blocks == 0)
3127 		return false;
3128 
3129 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3130 		sd_first_printk(KERN_WARNING, sdkp,
3131 				"Preferred minimum I/O size %u bytes not a " \
3132 				"multiple of physical block size (%u bytes)\n",
3133 				min_xfer_bytes, sdkp->physical_block_size);
3134 		sdkp->min_xfer_blocks = 0;
3135 		return false;
3136 	}
3137 
3138 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3139 			min_xfer_bytes);
3140 	return true;
3141 }
3142 
3143 /*
3144  * Determine the device's preferred I/O size for reads and writes
3145  * unless the reported value is unreasonably small, large, not a
3146  * multiple of the physical block size, or simply garbage.
3147  */
3148 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3149 				      unsigned int dev_max)
3150 {
3151 	struct scsi_device *sdp = sdkp->device;
3152 	unsigned int opt_xfer_bytes =
3153 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3154 	unsigned int min_xfer_bytes =
3155 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3156 
3157 	if (sdkp->opt_xfer_blocks == 0)
3158 		return false;
3159 
3160 	if (sdkp->opt_xfer_blocks > dev_max) {
3161 		sd_first_printk(KERN_WARNING, sdkp,
3162 				"Optimal transfer size %u logical blocks " \
3163 				"> dev_max (%u logical blocks)\n",
3164 				sdkp->opt_xfer_blocks, dev_max);
3165 		return false;
3166 	}
3167 
3168 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3169 		sd_first_printk(KERN_WARNING, sdkp,
3170 				"Optimal transfer size %u logical blocks " \
3171 				"> sd driver limit (%u logical blocks)\n",
3172 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3173 		return false;
3174 	}
3175 
3176 	if (opt_xfer_bytes < PAGE_SIZE) {
3177 		sd_first_printk(KERN_WARNING, sdkp,
3178 				"Optimal transfer size %u bytes < " \
3179 				"PAGE_SIZE (%u bytes)\n",
3180 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3181 		return false;
3182 	}
3183 
3184 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3185 		sd_first_printk(KERN_WARNING, sdkp,
3186 				"Optimal transfer size %u bytes not a " \
3187 				"multiple of preferred minimum block " \
3188 				"size (%u bytes)\n",
3189 				opt_xfer_bytes, min_xfer_bytes);
3190 		return false;
3191 	}
3192 
3193 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3194 		sd_first_printk(KERN_WARNING, sdkp,
3195 				"Optimal transfer size %u bytes not a " \
3196 				"multiple of physical block size (%u bytes)\n",
3197 				opt_xfer_bytes, sdkp->physical_block_size);
3198 		return false;
3199 	}
3200 
3201 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3202 			opt_xfer_bytes);
3203 	return true;
3204 }
3205 
3206 /**
3207  *	sd_revalidate_disk - called the first time a new disk is seen,
3208  *	performs disk spin up, read_capacity, etc.
3209  *	@disk: struct gendisk we care about
3210  **/
3211 static int sd_revalidate_disk(struct gendisk *disk)
3212 {
3213 	struct scsi_disk *sdkp = scsi_disk(disk);
3214 	struct scsi_device *sdp = sdkp->device;
3215 	struct request_queue *q = sdkp->disk->queue;
3216 	sector_t old_capacity = sdkp->capacity;
3217 	unsigned char *buffer;
3218 	unsigned int dev_max, rw_max;
3219 
3220 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3221 				      "sd_revalidate_disk\n"));
3222 
3223 	/*
3224 	 * If the device is offline, don't try and read capacity or any
3225 	 * of the other niceties.
3226 	 */
3227 	if (!scsi_device_online(sdp))
3228 		goto out;
3229 
3230 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3231 	if (!buffer) {
3232 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3233 			  "allocation failure.\n");
3234 		goto out;
3235 	}
3236 
3237 	sd_spinup_disk(sdkp);
3238 
3239 	/*
3240 	 * Without media there is no reason to ask; moreover, some devices
3241 	 * react badly if we do.
3242 	 */
3243 	if (sdkp->media_present) {
3244 		sd_read_capacity(sdkp, buffer);
3245 
3246 		/*
3247 		 * set the default to rotational.  All non-rotational devices
3248 		 * support the block characteristics VPD page, which will
3249 		 * cause this to be updated correctly and any device which
3250 		 * doesn't support it should be treated as rotational.
3251 		 */
3252 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3253 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3254 
3255 		if (scsi_device_supports_vpd(sdp)) {
3256 			sd_read_block_provisioning(sdkp);
3257 			sd_read_block_limits(sdkp);
3258 			sd_read_block_characteristics(sdkp);
3259 			sd_zbc_read_zones(sdkp, buffer);
3260 			sd_read_cpr(sdkp);
3261 		}
3262 
3263 		sd_print_capacity(sdkp, old_capacity);
3264 
3265 		sd_read_write_protect_flag(sdkp, buffer);
3266 		sd_read_cache_type(sdkp, buffer);
3267 		sd_read_app_tag_own(sdkp, buffer);
3268 		sd_read_write_same(sdkp, buffer);
3269 		sd_read_security(sdkp, buffer);
3270 		sd_config_protection(sdkp);
3271 	}
3272 
3273 	/*
3274 	 * We now have all cache related info, determine how we deal
3275 	 * with flush requests.
3276 	 */
3277 	sd_set_flush_flag(sdkp);
3278 
3279 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3280 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3281 
3282 	/* Some devices report a maximum block count for READ/WRITE requests. */
3283 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3284 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3285 
3286 	if (sd_validate_min_xfer_size(sdkp))
3287 		blk_queue_io_min(sdkp->disk->queue,
3288 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3289 	else
3290 		blk_queue_io_min(sdkp->disk->queue, 0);
3291 
3292 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3293 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3294 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3295 	} else {
3296 		q->limits.io_opt = 0;
3297 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3298 				      (sector_t)BLK_DEF_MAX_SECTORS);
3299 	}
3300 
3301 	/* Do not exceed controller limit */
3302 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3303 
3304 	/*
3305 	 * Only update max_sectors if previously unset or if the current value
3306 	 * exceeds the capabilities of the hardware.
3307 	 */
3308 	if (sdkp->first_scan ||
3309 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3310 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3311 		q->limits.max_sectors = rw_max;
3312 
3313 	sdkp->first_scan = 0;
3314 
3315 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3316 	sd_config_write_same(sdkp);
3317 	kfree(buffer);
3318 
3319 	/*
3320 	 * For a zoned drive, revalidating the zones can be done only once
3321 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3322 	 * capacity to 0.
3323 	 */
3324 	if (sd_zbc_revalidate_zones(sdkp))
3325 		set_capacity_and_notify(disk, 0);
3326 
3327  out:
3328 	return 0;
3329 }
3330 
3331 /**
3332  *	sd_unlock_native_capacity - unlock native capacity
3333  *	@disk: struct gendisk to set capacity for
3334  *
3335  *	Block layer calls this function if it detects that partitions
3336  *	on @disk reach beyond the end of the device.  If the SCSI host
3337  *	implements ->unlock_native_capacity() method, it's invoked to
3338  *	give it a chance to adjust the device capacity.
3339  *
3340  *	CONTEXT:
3341  *	Defined by block layer.  Might sleep.
3342  */
3343 static void sd_unlock_native_capacity(struct gendisk *disk)
3344 {
3345 	struct scsi_device *sdev = scsi_disk(disk)->device;
3346 
3347 	if (sdev->host->hostt->unlock_native_capacity)
3348 		sdev->host->hostt->unlock_native_capacity(sdev);
3349 }
3350 
3351 /**
3352  *	sd_format_disk_name - format disk name
3353  *	@prefix: name prefix - ie. "sd" for SCSI disks
3354  *	@index: index of the disk to format name for
3355  *	@buf: output buffer
3356  *	@buflen: length of the output buffer
3357  *
3358  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3359  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3360  *	which is followed by sdaaa.
3361  *
3362  *	This is basically 26 base counting with one extra 'nil' entry
3363  *	at the beginning from the second digit on and can be
3364  *	determined using similar method as 26 base conversion with the
3365  *	index shifted -1 after each digit is computed.
3366  *
3367  *	CONTEXT:
3368  *	Don't care.
3369  *
3370  *	RETURNS:
3371  *	0 on success, -errno on failure.
3372  */
3373 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3374 {
3375 	const int base = 'z' - 'a' + 1;
3376 	char *begin = buf + strlen(prefix);
3377 	char *end = buf + buflen;
3378 	char *p;
3379 	int unit;
3380 
3381 	p = end - 1;
3382 	*p = '\0';
3383 	unit = base;
3384 	do {
3385 		if (p == begin)
3386 			return -EINVAL;
3387 		*--p = 'a' + (index % unit);
3388 		index = (index / unit) - 1;
3389 	} while (index >= 0);
3390 
3391 	memmove(begin, p, end - p);
3392 	memcpy(buf, prefix, strlen(prefix));
3393 
3394 	return 0;
3395 }
3396 
3397 /**
3398  *	sd_probe - called during driver initialization and whenever a
3399  *	new scsi device is attached to the system. It is called once
3400  *	for each scsi device (not just disks) present.
3401  *	@dev: pointer to device object
3402  *
3403  *	Returns 0 if successful (or not interested in this scsi device
3404  *	(e.g. scanner)); 1 when there is an error.
3405  *
3406  *	Note: this function is invoked from the scsi mid-level.
3407  *	This function sets up the mapping between a given
3408  *	<host,channel,id,lun> (found in sdp) and new device name
3409  *	(e.g. /dev/sda). More precisely it is the block device major
3410  *	and minor number that is chosen here.
3411  *
3412  *	Assume sd_probe is not re-entrant (for time being)
3413  *	Also think about sd_probe() and sd_remove() running coincidentally.
3414  **/
3415 static int sd_probe(struct device *dev)
3416 {
3417 	struct scsi_device *sdp = to_scsi_device(dev);
3418 	struct scsi_disk *sdkp;
3419 	struct gendisk *gd;
3420 	int index;
3421 	int error;
3422 
3423 	scsi_autopm_get_device(sdp);
3424 	error = -ENODEV;
3425 	if (sdp->type != TYPE_DISK &&
3426 	    sdp->type != TYPE_ZBC &&
3427 	    sdp->type != TYPE_MOD &&
3428 	    sdp->type != TYPE_RBC)
3429 		goto out;
3430 
3431 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3432 		sdev_printk(KERN_WARNING, sdp,
3433 			    "Unsupported ZBC host-managed device.\n");
3434 		goto out;
3435 	}
3436 
3437 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3438 					"sd_probe\n"));
3439 
3440 	error = -ENOMEM;
3441 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3442 	if (!sdkp)
3443 		goto out;
3444 
3445 	gd = __alloc_disk_node(sdp->request_queue, NUMA_NO_NODE,
3446 			       &sd_bio_compl_lkclass);
3447 	if (!gd)
3448 		goto out_free;
3449 
3450 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3451 	if (index < 0) {
3452 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3453 		goto out_put;
3454 	}
3455 
3456 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3457 	if (error) {
3458 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3459 		goto out_free_index;
3460 	}
3461 
3462 	sdkp->device = sdp;
3463 	sdkp->disk = gd;
3464 	sdkp->index = index;
3465 	sdkp->max_retries = SD_MAX_RETRIES;
3466 	atomic_set(&sdkp->openers, 0);
3467 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3468 
3469 	if (!sdp->request_queue->rq_timeout) {
3470 		if (sdp->type != TYPE_MOD)
3471 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3472 		else
3473 			blk_queue_rq_timeout(sdp->request_queue,
3474 					     SD_MOD_TIMEOUT);
3475 	}
3476 
3477 	device_initialize(&sdkp->disk_dev);
3478 	sdkp->disk_dev.parent = get_device(dev);
3479 	sdkp->disk_dev.class = &sd_disk_class;
3480 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3481 
3482 	error = device_add(&sdkp->disk_dev);
3483 	if (error) {
3484 		put_device(&sdkp->disk_dev);
3485 		goto out;
3486 	}
3487 
3488 	dev_set_drvdata(dev, sdkp);
3489 
3490 	gd->major = sd_major((index & 0xf0) >> 4);
3491 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3492 	gd->minors = SD_MINORS;
3493 
3494 	gd->fops = &sd_fops;
3495 	gd->private_data = sdkp;
3496 
3497 	/* defaults, until the device tells us otherwise */
3498 	sdp->sector_size = 512;
3499 	sdkp->capacity = 0;
3500 	sdkp->media_present = 1;
3501 	sdkp->write_prot = 0;
3502 	sdkp->cache_override = 0;
3503 	sdkp->WCE = 0;
3504 	sdkp->RCD = 0;
3505 	sdkp->ATO = 0;
3506 	sdkp->first_scan = 1;
3507 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3508 
3509 	sd_revalidate_disk(gd);
3510 
3511 	if (sdp->removable) {
3512 		gd->flags |= GENHD_FL_REMOVABLE;
3513 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3514 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3515 	}
3516 
3517 	blk_pm_runtime_init(sdp->request_queue, dev);
3518 	if (sdp->rpm_autosuspend) {
3519 		pm_runtime_set_autosuspend_delay(dev,
3520 			sdp->host->hostt->rpm_autosuspend_delay);
3521 	}
3522 
3523 	error = device_add_disk(dev, gd, NULL);
3524 	if (error) {
3525 		put_device(&sdkp->disk_dev);
3526 		put_disk(gd);
3527 		goto out;
3528 	}
3529 
3530 	if (sdkp->security) {
3531 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3532 		if (sdkp->opal_dev)
3533 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3534 	}
3535 
3536 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3537 		  sdp->removable ? "removable " : "");
3538 	scsi_autopm_put_device(sdp);
3539 
3540 	return 0;
3541 
3542  out_free_index:
3543 	ida_free(&sd_index_ida, index);
3544  out_put:
3545 	put_disk(gd);
3546  out_free:
3547 	sd_zbc_release_disk(sdkp);
3548 	kfree(sdkp);
3549  out:
3550 	scsi_autopm_put_device(sdp);
3551 	return error;
3552 }
3553 
3554 /**
3555  *	sd_remove - called whenever a scsi disk (previously recognized by
3556  *	sd_probe) is detached from the system. It is called (potentially
3557  *	multiple times) during sd module unload.
3558  *	@dev: pointer to device object
3559  *
3560  *	Note: this function is invoked from the scsi mid-level.
3561  *	This function potentially frees up a device name (e.g. /dev/sdc)
3562  *	that could be re-used by a subsequent sd_probe().
3563  *	This function is not called when the built-in sd driver is "exit-ed".
3564  **/
3565 static int sd_remove(struct device *dev)
3566 {
3567 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3568 
3569 	scsi_autopm_get_device(sdkp->device);
3570 
3571 	device_del(&sdkp->disk_dev);
3572 	del_gendisk(sdkp->disk);
3573 	sd_shutdown(dev);
3574 
3575 	put_disk(sdkp->disk);
3576 	return 0;
3577 }
3578 
3579 static void scsi_disk_release(struct device *dev)
3580 {
3581 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3582 
3583 	ida_free(&sd_index_ida, sdkp->index);
3584 	sd_zbc_release_disk(sdkp);
3585 	put_device(&sdkp->device->sdev_gendev);
3586 	free_opal_dev(sdkp->opal_dev);
3587 
3588 	kfree(sdkp);
3589 }
3590 
3591 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3592 {
3593 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3594 	struct scsi_sense_hdr sshdr;
3595 	struct scsi_device *sdp = sdkp->device;
3596 	int res;
3597 
3598 	if (start)
3599 		cmd[4] |= 1;	/* START */
3600 
3601 	if (sdp->start_stop_pwr_cond)
3602 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3603 
3604 	if (!scsi_device_online(sdp))
3605 		return -ENODEV;
3606 
3607 	res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, &sshdr,
3608 			SD_TIMEOUT, sdkp->max_retries, 0, RQF_PM, NULL);
3609 	if (res) {
3610 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3611 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3612 			sd_print_sense_hdr(sdkp, &sshdr);
3613 			/* 0x3a is medium not present */
3614 			if (sshdr.asc == 0x3a)
3615 				res = 0;
3616 		}
3617 	}
3618 
3619 	/* SCSI error codes must not go to the generic layer */
3620 	if (res)
3621 		return -EIO;
3622 
3623 	return 0;
3624 }
3625 
3626 /*
3627  * Send a SYNCHRONIZE CACHE instruction down to the device through
3628  * the normal SCSI command structure.  Wait for the command to
3629  * complete.
3630  */
3631 static void sd_shutdown(struct device *dev)
3632 {
3633 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3634 
3635 	if (!sdkp)
3636 		return;         /* this can happen */
3637 
3638 	if (pm_runtime_suspended(dev))
3639 		return;
3640 
3641 	if (sdkp->WCE && sdkp->media_present) {
3642 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3643 		sd_sync_cache(sdkp, NULL);
3644 	}
3645 
3646 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3647 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3648 		sd_start_stop_device(sdkp, 0);
3649 	}
3650 }
3651 
3652 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3653 {
3654 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3655 	struct scsi_sense_hdr sshdr;
3656 	int ret = 0;
3657 
3658 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3659 		return 0;
3660 
3661 	if (sdkp->WCE && sdkp->media_present) {
3662 		if (!sdkp->device->silence_suspend)
3663 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3664 		ret = sd_sync_cache(sdkp, &sshdr);
3665 
3666 		if (ret) {
3667 			/* ignore OFFLINE device */
3668 			if (ret == -ENODEV)
3669 				return 0;
3670 
3671 			if (!scsi_sense_valid(&sshdr) ||
3672 			    sshdr.sense_key != ILLEGAL_REQUEST)
3673 				return ret;
3674 
3675 			/*
3676 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3677 			 * doesn't support sync. There's not much to do and
3678 			 * suspend shouldn't fail.
3679 			 */
3680 			ret = 0;
3681 		}
3682 	}
3683 
3684 	if (sdkp->device->manage_start_stop) {
3685 		if (!sdkp->device->silence_suspend)
3686 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3687 		/* an error is not worth aborting a system sleep */
3688 		ret = sd_start_stop_device(sdkp, 0);
3689 		if (ignore_stop_errors)
3690 			ret = 0;
3691 	}
3692 
3693 	return ret;
3694 }
3695 
3696 static int sd_suspend_system(struct device *dev)
3697 {
3698 	if (pm_runtime_suspended(dev))
3699 		return 0;
3700 
3701 	return sd_suspend_common(dev, true);
3702 }
3703 
3704 static int sd_suspend_runtime(struct device *dev)
3705 {
3706 	return sd_suspend_common(dev, false);
3707 }
3708 
3709 static int sd_resume(struct device *dev)
3710 {
3711 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3712 	int ret;
3713 
3714 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3715 		return 0;
3716 
3717 	if (!sdkp->device->manage_start_stop)
3718 		return 0;
3719 
3720 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3721 	ret = sd_start_stop_device(sdkp, 1);
3722 	if (!ret)
3723 		opal_unlock_from_suspend(sdkp->opal_dev);
3724 	return ret;
3725 }
3726 
3727 static int sd_resume_system(struct device *dev)
3728 {
3729 	if (pm_runtime_suspended(dev))
3730 		return 0;
3731 
3732 	return sd_resume(dev);
3733 }
3734 
3735 static int sd_resume_runtime(struct device *dev)
3736 {
3737 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3738 	struct scsi_device *sdp;
3739 
3740 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3741 		return 0;
3742 
3743 	sdp = sdkp->device;
3744 
3745 	if (sdp->ignore_media_change) {
3746 		/* clear the device's sense data */
3747 		static const u8 cmd[10] = { REQUEST_SENSE };
3748 
3749 		if (scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL,
3750 				 NULL, sdp->request_queue->rq_timeout, 1, 0,
3751 				 RQF_PM, NULL))
3752 			sd_printk(KERN_NOTICE, sdkp,
3753 				  "Failed to clear sense data\n");
3754 	}
3755 
3756 	return sd_resume(dev);
3757 }
3758 
3759 /**
3760  *	init_sd - entry point for this driver (both when built in or when
3761  *	a module).
3762  *
3763  *	Note: this function registers this driver with the scsi mid-level.
3764  **/
3765 static int __init init_sd(void)
3766 {
3767 	int majors = 0, i, err;
3768 
3769 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3770 
3771 	for (i = 0; i < SD_MAJORS; i++) {
3772 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3773 			continue;
3774 		majors++;
3775 	}
3776 
3777 	if (!majors)
3778 		return -ENODEV;
3779 
3780 	err = class_register(&sd_disk_class);
3781 	if (err)
3782 		goto err_out;
3783 
3784 	sd_cdb_cache = kmem_cache_create("sd_ext_cdb", SD_EXT_CDB_SIZE,
3785 					 0, 0, NULL);
3786 	if (!sd_cdb_cache) {
3787 		printk(KERN_ERR "sd: can't init extended cdb cache\n");
3788 		err = -ENOMEM;
3789 		goto err_out_class;
3790 	}
3791 
3792 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3793 	if (!sd_page_pool) {
3794 		printk(KERN_ERR "sd: can't init discard page pool\n");
3795 		err = -ENOMEM;
3796 		goto err_out_cache;
3797 	}
3798 
3799 	err = scsi_register_driver(&sd_template.gendrv);
3800 	if (err)
3801 		goto err_out_driver;
3802 
3803 	return 0;
3804 
3805 err_out_driver:
3806 	mempool_destroy(sd_page_pool);
3807 
3808 err_out_cache:
3809 	kmem_cache_destroy(sd_cdb_cache);
3810 
3811 err_out_class:
3812 	class_unregister(&sd_disk_class);
3813 err_out:
3814 	for (i = 0; i < SD_MAJORS; i++)
3815 		unregister_blkdev(sd_major(i), "sd");
3816 	return err;
3817 }
3818 
3819 /**
3820  *	exit_sd - exit point for this driver (when it is a module).
3821  *
3822  *	Note: this function unregisters this driver from the scsi mid-level.
3823  **/
3824 static void __exit exit_sd(void)
3825 {
3826 	int i;
3827 
3828 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3829 
3830 	scsi_unregister_driver(&sd_template.gendrv);
3831 	mempool_destroy(sd_page_pool);
3832 	kmem_cache_destroy(sd_cdb_cache);
3833 
3834 	class_unregister(&sd_disk_class);
3835 
3836 	for (i = 0; i < SD_MAJORS; i++)
3837 		unregister_blkdev(sd_major(i), "sd");
3838 }
3839 
3840 module_init(init_sd);
3841 module_exit(exit_sd);
3842 
3843 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3844 {
3845 	scsi_print_sense_hdr(sdkp->device,
3846 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3847 }
3848 
3849 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3850 {
3851 	const char *hb_string = scsi_hostbyte_string(result);
3852 
3853 	if (hb_string)
3854 		sd_printk(KERN_INFO, sdkp,
3855 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3856 			  hb_string ? hb_string : "invalid",
3857 			  "DRIVER_OK");
3858 	else
3859 		sd_printk(KERN_INFO, sdkp,
3860 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3861 			  msg, host_byte(result), "DRIVER_OK");
3862 }
3863