xref: /openbmc/linux/drivers/scsi/sd.c (revision 4e508b25)
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 void sd_start_done_work(struct work_struct *work);
107 static int  sd_probe(struct device *);
108 static int  sd_remove(struct device *);
109 static void sd_shutdown(struct device *);
110 static int sd_suspend_system(struct device *);
111 static int sd_suspend_runtime(struct device *);
112 static int sd_resume_system(struct device *);
113 static int sd_resume_runtime(struct device *);
114 static void sd_rescan(struct device *);
115 static blk_status_t sd_init_command(struct scsi_cmnd *SCpnt);
116 static void sd_uninit_command(struct scsi_cmnd *SCpnt);
117 static int sd_done(struct scsi_cmnd *);
118 static void sd_eh_reset(struct scsi_cmnd *);
119 static int sd_eh_action(struct scsi_cmnd *, int);
120 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
121 static void scsi_disk_release(struct device *cdev);
122 
123 static DEFINE_IDA(sd_index_ida);
124 
125 static struct kmem_cache *sd_cdb_cache;
126 static mempool_t *sd_page_pool;
127 static struct lock_class_key sd_bio_compl_lkclass;
128 
129 static const char *sd_cache_types[] = {
130 	"write through", "none", "write back",
131 	"write back, no read (daft)"
132 };
133 
134 static void sd_set_flush_flag(struct scsi_disk *sdkp)
135 {
136 	bool wc = false, fua = false;
137 
138 	if (sdkp->WCE) {
139 		wc = true;
140 		if (sdkp->DPOFUA)
141 			fua = true;
142 	}
143 
144 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
145 }
146 
147 static ssize_t
148 cache_type_store(struct device *dev, struct device_attribute *attr,
149 		 const char *buf, size_t count)
150 {
151 	int ct, rcd, wce, sp;
152 	struct scsi_disk *sdkp = to_scsi_disk(dev);
153 	struct scsi_device *sdp = sdkp->device;
154 	char buffer[64];
155 	char *buffer_data;
156 	struct scsi_mode_data data;
157 	struct scsi_sense_hdr sshdr;
158 	static const char temp[] = "temporary ";
159 	int len;
160 
161 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
162 		/* no cache control on RBC devices; theoretically they
163 		 * can do it, but there's probably so many exceptions
164 		 * it's not worth the risk */
165 		return -EINVAL;
166 
167 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
168 		buf += sizeof(temp) - 1;
169 		sdkp->cache_override = 1;
170 	} else {
171 		sdkp->cache_override = 0;
172 	}
173 
174 	ct = sysfs_match_string(sd_cache_types, buf);
175 	if (ct < 0)
176 		return -EINVAL;
177 
178 	rcd = ct & 0x01 ? 1 : 0;
179 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
180 
181 	if (sdkp->cache_override) {
182 		sdkp->WCE = wce;
183 		sdkp->RCD = rcd;
184 		sd_set_flush_flag(sdkp);
185 		return count;
186 	}
187 
188 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
189 			    sdkp->max_retries, &data, NULL))
190 		return -EINVAL;
191 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
192 		  data.block_descriptor_length);
193 	buffer_data = buffer + data.header_length +
194 		data.block_descriptor_length;
195 	buffer_data[2] &= ~0x05;
196 	buffer_data[2] |= wce << 2 | rcd;
197 	sp = buffer_data[0] & 0x80 ? 1 : 0;
198 	buffer_data[0] &= ~0x80;
199 
200 	/*
201 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
202 	 * received mode parameter buffer before doing MODE SELECT.
203 	 */
204 	data.device_specific = 0;
205 
206 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
207 			     sdkp->max_retries, &data, &sshdr)) {
208 		if (scsi_sense_valid(&sshdr))
209 			sd_print_sense_hdr(sdkp, &sshdr);
210 		return -EINVAL;
211 	}
212 	sd_revalidate_disk(sdkp->disk);
213 	return count;
214 }
215 
216 static ssize_t
217 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
218 		       char *buf)
219 {
220 	struct scsi_disk *sdkp = to_scsi_disk(dev);
221 	struct scsi_device *sdp = sdkp->device;
222 
223 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
224 }
225 
226 static ssize_t
227 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
228 			const char *buf, size_t count)
229 {
230 	struct scsi_disk *sdkp = to_scsi_disk(dev);
231 	struct scsi_device *sdp = sdkp->device;
232 	bool v;
233 
234 	if (!capable(CAP_SYS_ADMIN))
235 		return -EACCES;
236 
237 	if (kstrtobool(buf, &v))
238 		return -EINVAL;
239 
240 	sdp->manage_start_stop = v;
241 
242 	return count;
243 }
244 static DEVICE_ATTR_RW(manage_start_stop);
245 
246 static ssize_t
247 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
248 {
249 	struct scsi_disk *sdkp = to_scsi_disk(dev);
250 
251 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
252 }
253 
254 static ssize_t
255 allow_restart_store(struct device *dev, struct device_attribute *attr,
256 		    const char *buf, size_t count)
257 {
258 	bool v;
259 	struct scsi_disk *sdkp = to_scsi_disk(dev);
260 	struct scsi_device *sdp = sdkp->device;
261 
262 	if (!capable(CAP_SYS_ADMIN))
263 		return -EACCES;
264 
265 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
266 		return -EINVAL;
267 
268 	if (kstrtobool(buf, &v))
269 		return -EINVAL;
270 
271 	sdp->allow_restart = v;
272 
273 	return count;
274 }
275 static DEVICE_ATTR_RW(allow_restart);
276 
277 static ssize_t
278 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
279 {
280 	struct scsi_disk *sdkp = to_scsi_disk(dev);
281 	int ct = sdkp->RCD + 2*sdkp->WCE;
282 
283 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
284 }
285 static DEVICE_ATTR_RW(cache_type);
286 
287 static ssize_t
288 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
289 {
290 	struct scsi_disk *sdkp = to_scsi_disk(dev);
291 
292 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
293 }
294 static DEVICE_ATTR_RO(FUA);
295 
296 static ssize_t
297 protection_type_show(struct device *dev, struct device_attribute *attr,
298 		     char *buf)
299 {
300 	struct scsi_disk *sdkp = to_scsi_disk(dev);
301 
302 	return sprintf(buf, "%u\n", sdkp->protection_type);
303 }
304 
305 static ssize_t
306 protection_type_store(struct device *dev, struct device_attribute *attr,
307 		      const char *buf, size_t count)
308 {
309 	struct scsi_disk *sdkp = to_scsi_disk(dev);
310 	unsigned int val;
311 	int err;
312 
313 	if (!capable(CAP_SYS_ADMIN))
314 		return -EACCES;
315 
316 	err = kstrtouint(buf, 10, &val);
317 
318 	if (err)
319 		return err;
320 
321 	if (val <= T10_PI_TYPE3_PROTECTION)
322 		sdkp->protection_type = val;
323 
324 	return count;
325 }
326 static DEVICE_ATTR_RW(protection_type);
327 
328 static ssize_t
329 protection_mode_show(struct device *dev, struct device_attribute *attr,
330 		     char *buf)
331 {
332 	struct scsi_disk *sdkp = to_scsi_disk(dev);
333 	struct scsi_device *sdp = sdkp->device;
334 	unsigned int dif, dix;
335 
336 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
337 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
338 
339 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
340 		dif = 0;
341 		dix = 1;
342 	}
343 
344 	if (!dif && !dix)
345 		return sprintf(buf, "none\n");
346 
347 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
348 }
349 static DEVICE_ATTR_RO(protection_mode);
350 
351 static ssize_t
352 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
353 {
354 	struct scsi_disk *sdkp = to_scsi_disk(dev);
355 
356 	return sprintf(buf, "%u\n", sdkp->ATO);
357 }
358 static DEVICE_ATTR_RO(app_tag_own);
359 
360 static ssize_t
361 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
362 		       char *buf)
363 {
364 	struct scsi_disk *sdkp = to_scsi_disk(dev);
365 
366 	return sprintf(buf, "%u\n", sdkp->lbpme);
367 }
368 static DEVICE_ATTR_RO(thin_provisioning);
369 
370 /* sysfs_match_string() requires dense arrays */
371 static const char *lbp_mode[] = {
372 	[SD_LBP_FULL]		= "full",
373 	[SD_LBP_UNMAP]		= "unmap",
374 	[SD_LBP_WS16]		= "writesame_16",
375 	[SD_LBP_WS10]		= "writesame_10",
376 	[SD_LBP_ZERO]		= "writesame_zero",
377 	[SD_LBP_DISABLE]	= "disabled",
378 };
379 
380 static ssize_t
381 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
382 		       char *buf)
383 {
384 	struct scsi_disk *sdkp = to_scsi_disk(dev);
385 
386 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
387 }
388 
389 static ssize_t
390 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
391 			const char *buf, size_t count)
392 {
393 	struct scsi_disk *sdkp = to_scsi_disk(dev);
394 	struct scsi_device *sdp = sdkp->device;
395 	int mode;
396 
397 	if (!capable(CAP_SYS_ADMIN))
398 		return -EACCES;
399 
400 	if (sd_is_zoned(sdkp)) {
401 		sd_config_discard(sdkp, SD_LBP_DISABLE);
402 		return count;
403 	}
404 
405 	if (sdp->type != TYPE_DISK)
406 		return -EINVAL;
407 
408 	mode = sysfs_match_string(lbp_mode, buf);
409 	if (mode < 0)
410 		return -EINVAL;
411 
412 	sd_config_discard(sdkp, mode);
413 
414 	return count;
415 }
416 static DEVICE_ATTR_RW(provisioning_mode);
417 
418 /* sysfs_match_string() requires dense arrays */
419 static const char *zeroing_mode[] = {
420 	[SD_ZERO_WRITE]		= "write",
421 	[SD_ZERO_WS]		= "writesame",
422 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
423 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
424 };
425 
426 static ssize_t
427 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
428 		  char *buf)
429 {
430 	struct scsi_disk *sdkp = to_scsi_disk(dev);
431 
432 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
433 }
434 
435 static ssize_t
436 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
437 		   const char *buf, size_t count)
438 {
439 	struct scsi_disk *sdkp = to_scsi_disk(dev);
440 	int mode;
441 
442 	if (!capable(CAP_SYS_ADMIN))
443 		return -EACCES;
444 
445 	mode = sysfs_match_string(zeroing_mode, buf);
446 	if (mode < 0)
447 		return -EINVAL;
448 
449 	sdkp->zeroing_mode = mode;
450 
451 	return count;
452 }
453 static DEVICE_ATTR_RW(zeroing_mode);
454 
455 static ssize_t
456 max_medium_access_timeouts_show(struct device *dev,
457 				struct device_attribute *attr, char *buf)
458 {
459 	struct scsi_disk *sdkp = to_scsi_disk(dev);
460 
461 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
462 }
463 
464 static ssize_t
465 max_medium_access_timeouts_store(struct device *dev,
466 				 struct device_attribute *attr, const char *buf,
467 				 size_t count)
468 {
469 	struct scsi_disk *sdkp = to_scsi_disk(dev);
470 	int err;
471 
472 	if (!capable(CAP_SYS_ADMIN))
473 		return -EACCES;
474 
475 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
476 
477 	return err ? err : count;
478 }
479 static DEVICE_ATTR_RW(max_medium_access_timeouts);
480 
481 static ssize_t
482 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
483 			   char *buf)
484 {
485 	struct scsi_disk *sdkp = to_scsi_disk(dev);
486 
487 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
488 }
489 
490 static ssize_t
491 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
492 			    const char *buf, size_t count)
493 {
494 	struct scsi_disk *sdkp = to_scsi_disk(dev);
495 	struct scsi_device *sdp = sdkp->device;
496 	unsigned long max;
497 	int err;
498 
499 	if (!capable(CAP_SYS_ADMIN))
500 		return -EACCES;
501 
502 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
503 		return -EINVAL;
504 
505 	err = kstrtoul(buf, 10, &max);
506 
507 	if (err)
508 		return err;
509 
510 	if (max == 0)
511 		sdp->no_write_same = 1;
512 	else if (max <= SD_MAX_WS16_BLOCKS) {
513 		sdp->no_write_same = 0;
514 		sdkp->max_ws_blocks = max;
515 	}
516 
517 	sd_config_write_same(sdkp);
518 
519 	return count;
520 }
521 static DEVICE_ATTR_RW(max_write_same_blocks);
522 
523 static ssize_t
524 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
525 {
526 	struct scsi_disk *sdkp = to_scsi_disk(dev);
527 
528 	if (sdkp->device->type == TYPE_ZBC)
529 		return sprintf(buf, "host-managed\n");
530 	if (sdkp->zoned == 1)
531 		return sprintf(buf, "host-aware\n");
532 	if (sdkp->zoned == 2)
533 		return sprintf(buf, "drive-managed\n");
534 	return sprintf(buf, "none\n");
535 }
536 static DEVICE_ATTR_RO(zoned_cap);
537 
538 static ssize_t
539 max_retries_store(struct device *dev, struct device_attribute *attr,
540 		  const char *buf, size_t count)
541 {
542 	struct scsi_disk *sdkp = to_scsi_disk(dev);
543 	struct scsi_device *sdev = sdkp->device;
544 	int retries, err;
545 
546 	err = kstrtoint(buf, 10, &retries);
547 	if (err)
548 		return err;
549 
550 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
551 		sdkp->max_retries = retries;
552 		return count;
553 	}
554 
555 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
556 		    SD_MAX_RETRIES);
557 	return -EINVAL;
558 }
559 
560 static ssize_t
561 max_retries_show(struct device *dev, struct device_attribute *attr,
562 		 char *buf)
563 {
564 	struct scsi_disk *sdkp = to_scsi_disk(dev);
565 
566 	return sprintf(buf, "%d\n", sdkp->max_retries);
567 }
568 
569 static DEVICE_ATTR_RW(max_retries);
570 
571 static struct attribute *sd_disk_attrs[] = {
572 	&dev_attr_cache_type.attr,
573 	&dev_attr_FUA.attr,
574 	&dev_attr_allow_restart.attr,
575 	&dev_attr_manage_start_stop.attr,
576 	&dev_attr_protection_type.attr,
577 	&dev_attr_protection_mode.attr,
578 	&dev_attr_app_tag_own.attr,
579 	&dev_attr_thin_provisioning.attr,
580 	&dev_attr_provisioning_mode.attr,
581 	&dev_attr_zeroing_mode.attr,
582 	&dev_attr_max_write_same_blocks.attr,
583 	&dev_attr_max_medium_access_timeouts.attr,
584 	&dev_attr_zoned_cap.attr,
585 	&dev_attr_max_retries.attr,
586 	NULL,
587 };
588 ATTRIBUTE_GROUPS(sd_disk);
589 
590 static struct class sd_disk_class = {
591 	.name		= "scsi_disk",
592 	.owner		= THIS_MODULE,
593 	.dev_release	= scsi_disk_release,
594 	.dev_groups	= sd_disk_groups,
595 };
596 
597 static const struct dev_pm_ops sd_pm_ops = {
598 	.suspend		= sd_suspend_system,
599 	.resume			= sd_resume_system,
600 	.poweroff		= sd_suspend_system,
601 	.restore		= sd_resume_system,
602 	.runtime_suspend	= sd_suspend_runtime,
603 	.runtime_resume		= sd_resume_runtime,
604 };
605 
606 static struct scsi_driver sd_template = {
607 	.gendrv = {
608 		.name		= "sd",
609 		.owner		= THIS_MODULE,
610 		.probe		= sd_probe,
611 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
612 		.remove		= sd_remove,
613 		.shutdown	= sd_shutdown,
614 		.pm		= &sd_pm_ops,
615 	},
616 	.rescan			= sd_rescan,
617 	.init_command		= sd_init_command,
618 	.uninit_command		= sd_uninit_command,
619 	.done			= sd_done,
620 	.eh_action		= sd_eh_action,
621 	.eh_reset		= sd_eh_reset,
622 };
623 
624 /*
625  * Don't request a new module, as that could deadlock in multipath
626  * environment.
627  */
628 static void sd_default_probe(dev_t devt)
629 {
630 }
631 
632 /*
633  * Device no to disk mapping:
634  *
635  *       major         disc2     disc  p1
636  *   |............|.............|....|....| <- dev_t
637  *    31        20 19          8 7  4 3  0
638  *
639  * Inside a major, we have 16k disks, however mapped non-
640  * contiguously. The first 16 disks are for major0, the next
641  * ones with major1, ... Disk 256 is for major0 again, disk 272
642  * for major1, ...
643  * As we stay compatible with our numbering scheme, we can reuse
644  * the well-know SCSI majors 8, 65--71, 136--143.
645  */
646 static int sd_major(int major_idx)
647 {
648 	switch (major_idx) {
649 	case 0:
650 		return SCSI_DISK0_MAJOR;
651 	case 1 ... 7:
652 		return SCSI_DISK1_MAJOR + major_idx - 1;
653 	case 8 ... 15:
654 		return SCSI_DISK8_MAJOR + major_idx - 8;
655 	default:
656 		BUG();
657 		return 0;	/* shut up gcc */
658 	}
659 }
660 
661 #ifdef CONFIG_BLK_SED_OPAL
662 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
663 		size_t len, bool send)
664 {
665 	struct scsi_disk *sdkp = data;
666 	struct scsi_device *sdev = sdkp->device;
667 	u8 cdb[12] = { 0, };
668 	int ret;
669 
670 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
671 	cdb[1] = secp;
672 	put_unaligned_be16(spsp, &cdb[2]);
673 	put_unaligned_be32(len, &cdb[6]);
674 
675 	ret = scsi_execute(sdev, cdb, send ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
676 		buffer, len, NULL, NULL, SD_TIMEOUT, sdkp->max_retries, 0,
677 		RQF_PM, NULL);
678 	return ret <= 0 ? ret : -EIO;
679 }
680 #endif /* CONFIG_BLK_SED_OPAL */
681 
682 /*
683  * Look up the DIX operation based on whether the command is read or
684  * write and whether dix and dif are enabled.
685  */
686 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
687 {
688 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
689 	static const unsigned int ops[] = {	/* wrt dix dif */
690 		SCSI_PROT_NORMAL,		/*  0	0   0  */
691 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
692 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
693 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
694 		SCSI_PROT_NORMAL,		/*  1	0   0  */
695 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
696 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
697 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
698 	};
699 
700 	return ops[write << 2 | dix << 1 | dif];
701 }
702 
703 /*
704  * Returns a mask of the protection flags that are valid for a given DIX
705  * operation.
706  */
707 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
708 {
709 	static const unsigned int flag_mask[] = {
710 		[SCSI_PROT_NORMAL]		= 0,
711 
712 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
713 						  SCSI_PROT_GUARD_CHECK |
714 						  SCSI_PROT_REF_CHECK |
715 						  SCSI_PROT_REF_INCREMENT,
716 
717 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
718 						  SCSI_PROT_IP_CHECKSUM,
719 
720 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
721 						  SCSI_PROT_GUARD_CHECK |
722 						  SCSI_PROT_REF_CHECK |
723 						  SCSI_PROT_REF_INCREMENT |
724 						  SCSI_PROT_IP_CHECKSUM,
725 
726 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
727 						  SCSI_PROT_REF_INCREMENT,
728 
729 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
730 						  SCSI_PROT_REF_CHECK |
731 						  SCSI_PROT_REF_INCREMENT |
732 						  SCSI_PROT_IP_CHECKSUM,
733 
734 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
735 						  SCSI_PROT_GUARD_CHECK |
736 						  SCSI_PROT_REF_CHECK |
737 						  SCSI_PROT_REF_INCREMENT |
738 						  SCSI_PROT_IP_CHECKSUM,
739 	};
740 
741 	return flag_mask[prot_op];
742 }
743 
744 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
745 					   unsigned int dix, unsigned int dif)
746 {
747 	struct request *rq = scsi_cmd_to_rq(scmd);
748 	struct bio *bio = rq->bio;
749 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
750 	unsigned int protect = 0;
751 
752 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
753 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
754 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
755 
756 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
757 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
758 	}
759 
760 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
761 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
762 
763 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
764 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
765 	}
766 
767 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
768 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
769 
770 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
771 			protect = 3 << 5;	/* Disable target PI checking */
772 		else
773 			protect = 1 << 5;	/* Enable target PI checking */
774 	}
775 
776 	scsi_set_prot_op(scmd, prot_op);
777 	scsi_set_prot_type(scmd, dif);
778 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
779 
780 	return protect;
781 }
782 
783 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
784 {
785 	struct request_queue *q = sdkp->disk->queue;
786 	unsigned int logical_block_size = sdkp->device->sector_size;
787 	unsigned int max_blocks = 0;
788 
789 	q->limits.discard_alignment =
790 		sdkp->unmap_alignment * logical_block_size;
791 	q->limits.discard_granularity =
792 		max(sdkp->physical_block_size,
793 		    sdkp->unmap_granularity * logical_block_size);
794 	sdkp->provisioning_mode = mode;
795 
796 	switch (mode) {
797 
798 	case SD_LBP_FULL:
799 	case SD_LBP_DISABLE:
800 		blk_queue_max_discard_sectors(q, 0);
801 		return;
802 
803 	case SD_LBP_UNMAP:
804 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
805 					  (u32)SD_MAX_WS16_BLOCKS);
806 		break;
807 
808 	case SD_LBP_WS16:
809 		if (sdkp->device->unmap_limit_for_ws)
810 			max_blocks = sdkp->max_unmap_blocks;
811 		else
812 			max_blocks = sdkp->max_ws_blocks;
813 
814 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
815 		break;
816 
817 	case SD_LBP_WS10:
818 		if (sdkp->device->unmap_limit_for_ws)
819 			max_blocks = sdkp->max_unmap_blocks;
820 		else
821 			max_blocks = sdkp->max_ws_blocks;
822 
823 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
824 		break;
825 
826 	case SD_LBP_ZERO:
827 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
828 					  (u32)SD_MAX_WS10_BLOCKS);
829 		break;
830 	}
831 
832 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
833 }
834 
835 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
836 {
837 	struct scsi_device *sdp = cmd->device;
838 	struct request *rq = scsi_cmd_to_rq(cmd);
839 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
840 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
841 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
842 	unsigned int data_len = 24;
843 	char *buf;
844 
845 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
846 	if (!rq->special_vec.bv_page)
847 		return BLK_STS_RESOURCE;
848 	clear_highpage(rq->special_vec.bv_page);
849 	rq->special_vec.bv_offset = 0;
850 	rq->special_vec.bv_len = data_len;
851 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
852 
853 	cmd->cmd_len = 10;
854 	cmd->cmnd[0] = UNMAP;
855 	cmd->cmnd[8] = 24;
856 
857 	buf = bvec_virt(&rq->special_vec);
858 	put_unaligned_be16(6 + 16, &buf[0]);
859 	put_unaligned_be16(16, &buf[2]);
860 	put_unaligned_be64(lba, &buf[8]);
861 	put_unaligned_be32(nr_blocks, &buf[16]);
862 
863 	cmd->allowed = sdkp->max_retries;
864 	cmd->transfersize = data_len;
865 	rq->timeout = SD_TIMEOUT;
866 
867 	return scsi_alloc_sgtables(cmd);
868 }
869 
870 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
871 		bool unmap)
872 {
873 	struct scsi_device *sdp = cmd->device;
874 	struct request *rq = scsi_cmd_to_rq(cmd);
875 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
876 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
877 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
878 	u32 data_len = sdp->sector_size;
879 
880 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
881 	if (!rq->special_vec.bv_page)
882 		return BLK_STS_RESOURCE;
883 	clear_highpage(rq->special_vec.bv_page);
884 	rq->special_vec.bv_offset = 0;
885 	rq->special_vec.bv_len = data_len;
886 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
887 
888 	cmd->cmd_len = 16;
889 	cmd->cmnd[0] = WRITE_SAME_16;
890 	if (unmap)
891 		cmd->cmnd[1] = 0x8; /* UNMAP */
892 	put_unaligned_be64(lba, &cmd->cmnd[2]);
893 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
894 
895 	cmd->allowed = sdkp->max_retries;
896 	cmd->transfersize = data_len;
897 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
898 
899 	return scsi_alloc_sgtables(cmd);
900 }
901 
902 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
903 		bool unmap)
904 {
905 	struct scsi_device *sdp = cmd->device;
906 	struct request *rq = scsi_cmd_to_rq(cmd);
907 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
908 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
909 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
910 	u32 data_len = sdp->sector_size;
911 
912 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
913 	if (!rq->special_vec.bv_page)
914 		return BLK_STS_RESOURCE;
915 	clear_highpage(rq->special_vec.bv_page);
916 	rq->special_vec.bv_offset = 0;
917 	rq->special_vec.bv_len = data_len;
918 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
919 
920 	cmd->cmd_len = 10;
921 	cmd->cmnd[0] = WRITE_SAME;
922 	if (unmap)
923 		cmd->cmnd[1] = 0x8; /* UNMAP */
924 	put_unaligned_be32(lba, &cmd->cmnd[2]);
925 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
926 
927 	cmd->allowed = sdkp->max_retries;
928 	cmd->transfersize = data_len;
929 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
930 
931 	return scsi_alloc_sgtables(cmd);
932 }
933 
934 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
935 {
936 	struct request *rq = scsi_cmd_to_rq(cmd);
937 	struct scsi_device *sdp = cmd->device;
938 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
939 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
940 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
941 
942 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
943 		switch (sdkp->zeroing_mode) {
944 		case SD_ZERO_WS16_UNMAP:
945 			return sd_setup_write_same16_cmnd(cmd, true);
946 		case SD_ZERO_WS10_UNMAP:
947 			return sd_setup_write_same10_cmnd(cmd, true);
948 		}
949 	}
950 
951 	if (sdp->no_write_same) {
952 		rq->rq_flags |= RQF_QUIET;
953 		return BLK_STS_TARGET;
954 	}
955 
956 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
957 		return sd_setup_write_same16_cmnd(cmd, false);
958 
959 	return sd_setup_write_same10_cmnd(cmd, false);
960 }
961 
962 static void sd_config_write_same(struct scsi_disk *sdkp)
963 {
964 	struct request_queue *q = sdkp->disk->queue;
965 	unsigned int logical_block_size = sdkp->device->sector_size;
966 
967 	if (sdkp->device->no_write_same) {
968 		sdkp->max_ws_blocks = 0;
969 		goto out;
970 	}
971 
972 	/* Some devices can not handle block counts above 0xffff despite
973 	 * supporting WRITE SAME(16). Consequently we default to 64k
974 	 * blocks per I/O unless the device explicitly advertises a
975 	 * bigger limit.
976 	 */
977 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
978 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
979 						   (u32)SD_MAX_WS16_BLOCKS);
980 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
981 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
982 						   (u32)SD_MAX_WS10_BLOCKS);
983 	else {
984 		sdkp->device->no_write_same = 1;
985 		sdkp->max_ws_blocks = 0;
986 	}
987 
988 	if (sdkp->lbprz && sdkp->lbpws)
989 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
990 	else if (sdkp->lbprz && sdkp->lbpws10)
991 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
992 	else if (sdkp->max_ws_blocks)
993 		sdkp->zeroing_mode = SD_ZERO_WS;
994 	else
995 		sdkp->zeroing_mode = SD_ZERO_WRITE;
996 
997 	if (sdkp->max_ws_blocks &&
998 	    sdkp->physical_block_size > logical_block_size) {
999 		/*
1000 		 * Reporting a maximum number of blocks that is not aligned
1001 		 * on the device physical size would cause a large write same
1002 		 * request to be split into physically unaligned chunks by
1003 		 * __blkdev_issue_write_zeroes() even if the caller of this
1004 		 * functions took care to align the large request. So make sure
1005 		 * the maximum reported is aligned to the device physical block
1006 		 * size. This is only an optional optimization for regular
1007 		 * disks, but this is mandatory to avoid failure of large write
1008 		 * same requests directed at sequential write required zones of
1009 		 * host-managed ZBC disks.
1010 		 */
1011 		sdkp->max_ws_blocks =
1012 			round_down(sdkp->max_ws_blocks,
1013 				   bytes_to_logical(sdkp->device,
1014 						    sdkp->physical_block_size));
1015 	}
1016 
1017 out:
1018 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1019 					 (logical_block_size >> 9));
1020 }
1021 
1022 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1023 {
1024 	struct request *rq = scsi_cmd_to_rq(cmd);
1025 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1026 
1027 	/* flush requests don't perform I/O, zero the S/G table */
1028 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1029 
1030 	cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1031 	cmd->cmd_len = 10;
1032 	cmd->transfersize = 0;
1033 	cmd->allowed = sdkp->max_retries;
1034 
1035 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1036 	return BLK_STS_OK;
1037 }
1038 
1039 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1040 				       sector_t lba, unsigned int nr_blocks,
1041 				       unsigned char flags)
1042 {
1043 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1044 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1045 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1046 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1047 	cmd->cmnd[10] = flags;
1048 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1049 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1050 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1051 
1052 	return BLK_STS_OK;
1053 }
1054 
1055 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1056 				       sector_t lba, unsigned int nr_blocks,
1057 				       unsigned char flags)
1058 {
1059 	cmd->cmd_len  = 16;
1060 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1061 	cmd->cmnd[1]  = flags;
1062 	cmd->cmnd[14] = 0;
1063 	cmd->cmnd[15] = 0;
1064 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1065 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1066 
1067 	return BLK_STS_OK;
1068 }
1069 
1070 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1071 				       sector_t lba, unsigned int nr_blocks,
1072 				       unsigned char flags)
1073 {
1074 	cmd->cmd_len = 10;
1075 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1076 	cmd->cmnd[1] = flags;
1077 	cmd->cmnd[6] = 0;
1078 	cmd->cmnd[9] = 0;
1079 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1080 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1081 
1082 	return BLK_STS_OK;
1083 }
1084 
1085 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1086 				      sector_t lba, unsigned int nr_blocks,
1087 				      unsigned char flags)
1088 {
1089 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1090 	if (WARN_ON_ONCE(nr_blocks == 0))
1091 		return BLK_STS_IOERR;
1092 
1093 	if (unlikely(flags & 0x8)) {
1094 		/*
1095 		 * This happens only if this drive failed 10byte rw
1096 		 * command with ILLEGAL_REQUEST during operation and
1097 		 * thus turned off use_10_for_rw.
1098 		 */
1099 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1100 		return BLK_STS_IOERR;
1101 	}
1102 
1103 	cmd->cmd_len = 6;
1104 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1105 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1106 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1107 	cmd->cmnd[3] = lba & 0xff;
1108 	cmd->cmnd[4] = nr_blocks;
1109 	cmd->cmnd[5] = 0;
1110 
1111 	return BLK_STS_OK;
1112 }
1113 
1114 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1115 {
1116 	struct request *rq = scsi_cmd_to_rq(cmd);
1117 	struct scsi_device *sdp = cmd->device;
1118 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1119 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1120 	sector_t threshold;
1121 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1122 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1123 	bool write = rq_data_dir(rq) == WRITE;
1124 	unsigned char protect, fua;
1125 	blk_status_t ret;
1126 	unsigned int dif;
1127 	bool dix;
1128 
1129 	ret = scsi_alloc_sgtables(cmd);
1130 	if (ret != BLK_STS_OK)
1131 		return ret;
1132 
1133 	ret = BLK_STS_IOERR;
1134 	if (!scsi_device_online(sdp) || sdp->changed) {
1135 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1136 		goto fail;
1137 	}
1138 
1139 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1140 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1141 		goto fail;
1142 	}
1143 
1144 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1145 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1146 		goto fail;
1147 	}
1148 
1149 	/*
1150 	 * Some SD card readers can't handle accesses which touch the
1151 	 * last one or two logical blocks. Split accesses as needed.
1152 	 */
1153 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1154 
1155 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1156 		if (lba < threshold) {
1157 			/* Access up to the threshold but not beyond */
1158 			nr_blocks = threshold - lba;
1159 		} else {
1160 			/* Access only a single logical block */
1161 			nr_blocks = 1;
1162 		}
1163 	}
1164 
1165 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1166 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1167 		if (ret)
1168 			goto fail;
1169 	}
1170 
1171 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1172 	dix = scsi_prot_sg_count(cmd);
1173 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1174 
1175 	if (dif || dix)
1176 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1177 	else
1178 		protect = 0;
1179 
1180 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1181 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1182 					 protect | fua);
1183 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1184 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1185 					 protect | fua);
1186 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1187 		   sdp->use_10_for_rw || protect) {
1188 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1189 					 protect | fua);
1190 	} else {
1191 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1192 					protect | fua);
1193 	}
1194 
1195 	if (unlikely(ret != BLK_STS_OK))
1196 		goto fail;
1197 
1198 	/*
1199 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1200 	 * host adapter, it's safe to assume that we can at least transfer
1201 	 * this many bytes between each connect / disconnect.
1202 	 */
1203 	cmd->transfersize = sdp->sector_size;
1204 	cmd->underflow = nr_blocks << 9;
1205 	cmd->allowed = sdkp->max_retries;
1206 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1207 
1208 	SCSI_LOG_HLQUEUE(1,
1209 			 scmd_printk(KERN_INFO, cmd,
1210 				     "%s: block=%llu, count=%d\n", __func__,
1211 				     (unsigned long long)blk_rq_pos(rq),
1212 				     blk_rq_sectors(rq)));
1213 	SCSI_LOG_HLQUEUE(2,
1214 			 scmd_printk(KERN_INFO, cmd,
1215 				     "%s %d/%u 512 byte blocks.\n",
1216 				     write ? "writing" : "reading", nr_blocks,
1217 				     blk_rq_sectors(rq)));
1218 
1219 	/*
1220 	 * This indicates that the command is ready from our end to be queued.
1221 	 */
1222 	return BLK_STS_OK;
1223 fail:
1224 	scsi_free_sgtables(cmd);
1225 	return ret;
1226 }
1227 
1228 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1229 {
1230 	struct request *rq = scsi_cmd_to_rq(cmd);
1231 
1232 	switch (req_op(rq)) {
1233 	case REQ_OP_DISCARD:
1234 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1235 		case SD_LBP_UNMAP:
1236 			return sd_setup_unmap_cmnd(cmd);
1237 		case SD_LBP_WS16:
1238 			return sd_setup_write_same16_cmnd(cmd, true);
1239 		case SD_LBP_WS10:
1240 			return sd_setup_write_same10_cmnd(cmd, true);
1241 		case SD_LBP_ZERO:
1242 			return sd_setup_write_same10_cmnd(cmd, false);
1243 		default:
1244 			return BLK_STS_TARGET;
1245 		}
1246 	case REQ_OP_WRITE_ZEROES:
1247 		return sd_setup_write_zeroes_cmnd(cmd);
1248 	case REQ_OP_FLUSH:
1249 		return sd_setup_flush_cmnd(cmd);
1250 	case REQ_OP_READ:
1251 	case REQ_OP_WRITE:
1252 	case REQ_OP_ZONE_APPEND:
1253 		return sd_setup_read_write_cmnd(cmd);
1254 	case REQ_OP_ZONE_RESET:
1255 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1256 						   false);
1257 	case REQ_OP_ZONE_RESET_ALL:
1258 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1259 						   true);
1260 	case REQ_OP_ZONE_OPEN:
1261 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1262 	case REQ_OP_ZONE_CLOSE:
1263 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1264 	case REQ_OP_ZONE_FINISH:
1265 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1266 	default:
1267 		WARN_ON_ONCE(1);
1268 		return BLK_STS_NOTSUPP;
1269 	}
1270 }
1271 
1272 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1273 {
1274 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1275 
1276 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1277 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1278 }
1279 
1280 static bool sd_need_revalidate(struct block_device *bdev,
1281 		struct scsi_disk *sdkp)
1282 {
1283 	if (sdkp->device->removable || sdkp->write_prot) {
1284 		if (bdev_check_media_change(bdev))
1285 			return true;
1286 	}
1287 
1288 	/*
1289 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1290 	 * nothing to do with partitions, BLKRRPART is used to force a full
1291 	 * revalidate after things like a format for historical reasons.
1292 	 */
1293 	return test_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1294 }
1295 
1296 /**
1297  *	sd_open - open a scsi disk device
1298  *	@bdev: Block device of the scsi disk to open
1299  *	@mode: FMODE_* mask
1300  *
1301  *	Returns 0 if successful. Returns a negated errno value in case
1302  *	of error.
1303  *
1304  *	Note: This can be called from a user context (e.g. fsck(1) )
1305  *	or from within the kernel (e.g. as a result of a mount(1) ).
1306  *	In the latter case @inode and @filp carry an abridged amount
1307  *	of information as noted above.
1308  *
1309  *	Locking: called with bdev->bd_disk->open_mutex held.
1310  **/
1311 static int sd_open(struct block_device *bdev, fmode_t mode)
1312 {
1313 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1314 	struct scsi_device *sdev = sdkp->device;
1315 	int retval;
1316 
1317 	if (scsi_device_get(sdev))
1318 		return -ENXIO;
1319 
1320 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1321 
1322 	/*
1323 	 * If the device is in error recovery, wait until it is done.
1324 	 * If the device is offline, then disallow any access to it.
1325 	 */
1326 	retval = -ENXIO;
1327 	if (!scsi_block_when_processing_errors(sdev))
1328 		goto error_out;
1329 
1330 	if (sd_need_revalidate(bdev, sdkp))
1331 		sd_revalidate_disk(bdev->bd_disk);
1332 
1333 	/*
1334 	 * If the drive is empty, just let the open fail.
1335 	 */
1336 	retval = -ENOMEDIUM;
1337 	if (sdev->removable && !sdkp->media_present && !(mode & FMODE_NDELAY))
1338 		goto error_out;
1339 
1340 	/*
1341 	 * If the device has the write protect tab set, have the open fail
1342 	 * if the user expects to be able to write to the thing.
1343 	 */
1344 	retval = -EROFS;
1345 	if (sdkp->write_prot && (mode & FMODE_WRITE))
1346 		goto error_out;
1347 
1348 	/*
1349 	 * It is possible that the disk changing stuff resulted in
1350 	 * the device being taken offline.  If this is the case,
1351 	 * report this to the user, and don't pretend that the
1352 	 * open actually succeeded.
1353 	 */
1354 	retval = -ENXIO;
1355 	if (!scsi_device_online(sdev))
1356 		goto error_out;
1357 
1358 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1359 		if (scsi_block_when_processing_errors(sdev))
1360 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1361 	}
1362 
1363 	return 0;
1364 
1365 error_out:
1366 	scsi_device_put(sdev);
1367 	return retval;
1368 }
1369 
1370 /**
1371  *	sd_release - invoked when the (last) close(2) is called on this
1372  *	scsi disk.
1373  *	@disk: disk to release
1374  *	@mode: FMODE_* mask
1375  *
1376  *	Returns 0.
1377  *
1378  *	Note: may block (uninterruptible) if error recovery is underway
1379  *	on this disk.
1380  *
1381  *	Locking: called with bdev->bd_disk->open_mutex held.
1382  **/
1383 static void sd_release(struct gendisk *disk, fmode_t mode)
1384 {
1385 	struct scsi_disk *sdkp = scsi_disk(disk);
1386 	struct scsi_device *sdev = sdkp->device;
1387 
1388 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1389 
1390 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1391 		if (scsi_block_when_processing_errors(sdev))
1392 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1393 	}
1394 
1395 	scsi_device_put(sdev);
1396 }
1397 
1398 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1399 {
1400 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1401 	struct scsi_device *sdp = sdkp->device;
1402 	struct Scsi_Host *host = sdp->host;
1403 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1404 	int diskinfo[4];
1405 
1406 	/* default to most commonly used values */
1407 	diskinfo[0] = 0x40;	/* 1 << 6 */
1408 	diskinfo[1] = 0x20;	/* 1 << 5 */
1409 	diskinfo[2] = capacity >> 11;
1410 
1411 	/* override with calculated, extended default, or driver values */
1412 	if (host->hostt->bios_param)
1413 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1414 	else
1415 		scsicam_bios_param(bdev, capacity, diskinfo);
1416 
1417 	geo->heads = diskinfo[0];
1418 	geo->sectors = diskinfo[1];
1419 	geo->cylinders = diskinfo[2];
1420 	return 0;
1421 }
1422 
1423 /**
1424  *	sd_ioctl - process an ioctl
1425  *	@bdev: target block device
1426  *	@mode: FMODE_* mask
1427  *	@cmd: ioctl command number
1428  *	@arg: this is third argument given to ioctl(2) system call.
1429  *	Often contains a pointer.
1430  *
1431  *	Returns 0 if successful (some ioctls return positive numbers on
1432  *	success as well). Returns a negated errno value in case of error.
1433  *
1434  *	Note: most ioctls are forward onto the block subsystem or further
1435  *	down in the scsi subsystem.
1436  **/
1437 static int sd_ioctl(struct block_device *bdev, fmode_t mode,
1438 		    unsigned int cmd, unsigned long arg)
1439 {
1440 	struct gendisk *disk = bdev->bd_disk;
1441 	struct scsi_disk *sdkp = scsi_disk(disk);
1442 	struct scsi_device *sdp = sdkp->device;
1443 	void __user *p = (void __user *)arg;
1444 	int error;
1445 
1446 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1447 				    "cmd=0x%x\n", disk->disk_name, cmd));
1448 
1449 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1450 		return -ENOIOCTLCMD;
1451 
1452 	/*
1453 	 * If we are in the middle of error recovery, don't let anyone
1454 	 * else try and use this device.  Also, if error recovery fails, it
1455 	 * may try and take the device offline, in which case all further
1456 	 * access to the device is prohibited.
1457 	 */
1458 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1459 			(mode & FMODE_NDELAY) != 0);
1460 	if (error)
1461 		return error;
1462 
1463 	if (is_sed_ioctl(cmd))
1464 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1465 	return scsi_ioctl(sdp, mode, cmd, p);
1466 }
1467 
1468 static void set_media_not_present(struct scsi_disk *sdkp)
1469 {
1470 	if (sdkp->media_present)
1471 		sdkp->device->changed = 1;
1472 
1473 	if (sdkp->device->removable) {
1474 		sdkp->media_present = 0;
1475 		sdkp->capacity = 0;
1476 	}
1477 }
1478 
1479 static int media_not_present(struct scsi_disk *sdkp,
1480 			     struct scsi_sense_hdr *sshdr)
1481 {
1482 	if (!scsi_sense_valid(sshdr))
1483 		return 0;
1484 
1485 	/* not invoked for commands that could return deferred errors */
1486 	switch (sshdr->sense_key) {
1487 	case UNIT_ATTENTION:
1488 	case NOT_READY:
1489 		/* medium not present */
1490 		if (sshdr->asc == 0x3A) {
1491 			set_media_not_present(sdkp);
1492 			return 1;
1493 		}
1494 	}
1495 	return 0;
1496 }
1497 
1498 /**
1499  *	sd_check_events - check media events
1500  *	@disk: kernel device descriptor
1501  *	@clearing: disk events currently being cleared
1502  *
1503  *	Returns mask of DISK_EVENT_*.
1504  *
1505  *	Note: this function is invoked from the block subsystem.
1506  **/
1507 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1508 {
1509 	struct scsi_disk *sdkp = disk->private_data;
1510 	struct scsi_device *sdp;
1511 	int retval;
1512 	bool disk_changed;
1513 
1514 	if (!sdkp)
1515 		return 0;
1516 
1517 	sdp = sdkp->device;
1518 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1519 
1520 	/*
1521 	 * If the device is offline, don't send any commands - just pretend as
1522 	 * if the command failed.  If the device ever comes back online, we
1523 	 * can deal with it then.  It is only because of unrecoverable errors
1524 	 * that we would ever take a device offline in the first place.
1525 	 */
1526 	if (!scsi_device_online(sdp)) {
1527 		set_media_not_present(sdkp);
1528 		goto out;
1529 	}
1530 
1531 	/*
1532 	 * Using TEST_UNIT_READY enables differentiation between drive with
1533 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1534 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1535 	 *
1536 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1537 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1538 	 * sd_revalidate() is called.
1539 	 */
1540 	if (scsi_block_when_processing_errors(sdp)) {
1541 		struct scsi_sense_hdr sshdr = { 0, };
1542 
1543 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1544 					      &sshdr);
1545 
1546 		/* failed to execute TUR, assume media not present */
1547 		if (retval < 0 || host_byte(retval)) {
1548 			set_media_not_present(sdkp);
1549 			goto out;
1550 		}
1551 
1552 		if (media_not_present(sdkp, &sshdr))
1553 			goto out;
1554 	}
1555 
1556 	/*
1557 	 * For removable scsi disk we have to recognise the presence
1558 	 * of a disk in the drive.
1559 	 */
1560 	if (!sdkp->media_present)
1561 		sdp->changed = 1;
1562 	sdkp->media_present = 1;
1563 out:
1564 	/*
1565 	 * sdp->changed is set under the following conditions:
1566 	 *
1567 	 *	Medium present state has changed in either direction.
1568 	 *	Device has indicated UNIT_ATTENTION.
1569 	 */
1570 	disk_changed = sdp->changed;
1571 	sdp->changed = 0;
1572 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1573 }
1574 
1575 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1576 {
1577 	int retries, res;
1578 	struct scsi_device *sdp = sdkp->device;
1579 	const int timeout = sdp->request_queue->rq_timeout
1580 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1581 	struct scsi_sense_hdr my_sshdr;
1582 
1583 	if (!scsi_device_online(sdp))
1584 		return -ENODEV;
1585 
1586 	/* caller might not be interested in sense, but we need it */
1587 	if (!sshdr)
1588 		sshdr = &my_sshdr;
1589 
1590 	for (retries = 3; retries > 0; --retries) {
1591 		unsigned char cmd[10] = { 0 };
1592 
1593 		cmd[0] = SYNCHRONIZE_CACHE;
1594 		/*
1595 		 * Leave the rest of the command zero to indicate
1596 		 * flush everything.
1597 		 */
1598 		res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, sshdr,
1599 				timeout, sdkp->max_retries, 0, RQF_PM, NULL);
1600 		if (res == 0)
1601 			break;
1602 	}
1603 
1604 	if (res) {
1605 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1606 
1607 		if (res < 0)
1608 			return res;
1609 
1610 		if (scsi_status_is_check_condition(res) &&
1611 		    scsi_sense_valid(sshdr)) {
1612 			sd_print_sense_hdr(sdkp, sshdr);
1613 
1614 			/* we need to evaluate the error return  */
1615 			if (sshdr->asc == 0x3a ||	/* medium not present */
1616 			    sshdr->asc == 0x20 ||	/* invalid command */
1617 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1618 				/* this is no error here */
1619 				return 0;
1620 		}
1621 
1622 		switch (host_byte(res)) {
1623 		/* ignore errors due to racing a disconnection */
1624 		case DID_BAD_TARGET:
1625 		case DID_NO_CONNECT:
1626 			return 0;
1627 		/* signal the upper layer it might try again */
1628 		case DID_BUS_BUSY:
1629 		case DID_IMM_RETRY:
1630 		case DID_REQUEUE:
1631 		case DID_SOFT_ERROR:
1632 			return -EBUSY;
1633 		default:
1634 			return -EIO;
1635 		}
1636 	}
1637 	return 0;
1638 }
1639 
1640 static void sd_rescan(struct device *dev)
1641 {
1642 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1643 
1644 	sd_revalidate_disk(sdkp->disk);
1645 }
1646 
1647 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1648 		enum blk_unique_id type)
1649 {
1650 	struct scsi_device *sdev = scsi_disk(disk)->device;
1651 	const struct scsi_vpd *vpd;
1652 	const unsigned char *d;
1653 	int ret = -ENXIO, len;
1654 
1655 	rcu_read_lock();
1656 	vpd = rcu_dereference(sdev->vpd_pg83);
1657 	if (!vpd)
1658 		goto out_unlock;
1659 
1660 	ret = -EINVAL;
1661 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1662 		/* we only care about designators with LU association */
1663 		if (((d[1] >> 4) & 0x3) != 0x00)
1664 			continue;
1665 		if ((d[1] & 0xf) != type)
1666 			continue;
1667 
1668 		/*
1669 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1670 		 * keep looking as one with more entropy might still show up.
1671 		 */
1672 		len = d[3];
1673 		if (len != 8 && len != 12 && len != 16)
1674 			continue;
1675 		ret = len;
1676 		memcpy(id, d + 4, len);
1677 		if (len == 16)
1678 			break;
1679 	}
1680 out_unlock:
1681 	rcu_read_unlock();
1682 	return ret;
1683 }
1684 
1685 static char sd_pr_type(enum pr_type type)
1686 {
1687 	switch (type) {
1688 	case PR_WRITE_EXCLUSIVE:
1689 		return 0x01;
1690 	case PR_EXCLUSIVE_ACCESS:
1691 		return 0x03;
1692 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1693 		return 0x05;
1694 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1695 		return 0x06;
1696 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1697 		return 0x07;
1698 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1699 		return 0x08;
1700 	default:
1701 		return 0;
1702 	}
1703 };
1704 
1705 static int sd_pr_command(struct block_device *bdev, u8 sa,
1706 		u64 key, u64 sa_key, u8 type, u8 flags)
1707 {
1708 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1709 	struct scsi_device *sdev = sdkp->device;
1710 	struct scsi_sense_hdr sshdr;
1711 	int result;
1712 	u8 cmd[16] = { 0, };
1713 	u8 data[24] = { 0, };
1714 
1715 	cmd[0] = PERSISTENT_RESERVE_OUT;
1716 	cmd[1] = sa;
1717 	cmd[2] = type;
1718 	put_unaligned_be32(sizeof(data), &cmd[5]);
1719 
1720 	put_unaligned_be64(key, &data[0]);
1721 	put_unaligned_be64(sa_key, &data[8]);
1722 	data[20] = flags;
1723 
1724 	result = scsi_execute_req(sdev, cmd, DMA_TO_DEVICE, &data, sizeof(data),
1725 			&sshdr, SD_TIMEOUT, sdkp->max_retries, NULL);
1726 
1727 	if (scsi_status_is_check_condition(result) &&
1728 	    scsi_sense_valid(&sshdr)) {
1729 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1730 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1731 	}
1732 
1733 	return result;
1734 }
1735 
1736 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1737 		u32 flags)
1738 {
1739 	if (flags & ~PR_FL_IGNORE_KEY)
1740 		return -EOPNOTSUPP;
1741 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1742 			old_key, new_key, 0,
1743 			(1 << 0) /* APTPL */);
1744 }
1745 
1746 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1747 		u32 flags)
1748 {
1749 	if (flags)
1750 		return -EOPNOTSUPP;
1751 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1752 }
1753 
1754 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1755 {
1756 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1757 }
1758 
1759 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1760 		enum pr_type type, bool abort)
1761 {
1762 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1763 			     sd_pr_type(type), 0);
1764 }
1765 
1766 static int sd_pr_clear(struct block_device *bdev, u64 key)
1767 {
1768 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1769 }
1770 
1771 static const struct pr_ops sd_pr_ops = {
1772 	.pr_register	= sd_pr_register,
1773 	.pr_reserve	= sd_pr_reserve,
1774 	.pr_release	= sd_pr_release,
1775 	.pr_preempt	= sd_pr_preempt,
1776 	.pr_clear	= sd_pr_clear,
1777 };
1778 
1779 static void scsi_disk_free_disk(struct gendisk *disk)
1780 {
1781 	struct scsi_disk *sdkp = scsi_disk(disk);
1782 
1783 	put_device(&sdkp->disk_dev);
1784 }
1785 
1786 static const struct block_device_operations sd_fops = {
1787 	.owner			= THIS_MODULE,
1788 	.open			= sd_open,
1789 	.release		= sd_release,
1790 	.ioctl			= sd_ioctl,
1791 	.getgeo			= sd_getgeo,
1792 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1793 	.check_events		= sd_check_events,
1794 	.unlock_native_capacity	= sd_unlock_native_capacity,
1795 	.report_zones		= sd_zbc_report_zones,
1796 	.get_unique_id		= sd_get_unique_id,
1797 	.free_disk		= scsi_disk_free_disk,
1798 	.pr_ops			= &sd_pr_ops,
1799 };
1800 
1801 /**
1802  *	sd_eh_reset - reset error handling callback
1803  *	@scmd:		sd-issued command that has failed
1804  *
1805  *	This function is called by the SCSI midlayer before starting
1806  *	SCSI EH. When counting medium access failures we have to be
1807  *	careful to register it only only once per device and SCSI EH run;
1808  *	there might be several timed out commands which will cause the
1809  *	'max_medium_access_timeouts' counter to trigger after the first
1810  *	SCSI EH run already and set the device to offline.
1811  *	So this function resets the internal counter before starting SCSI EH.
1812  **/
1813 static void sd_eh_reset(struct scsi_cmnd *scmd)
1814 {
1815 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1816 
1817 	/* New SCSI EH run, reset gate variable */
1818 	sdkp->ignore_medium_access_errors = false;
1819 }
1820 
1821 /**
1822  *	sd_eh_action - error handling callback
1823  *	@scmd:		sd-issued command that has failed
1824  *	@eh_disp:	The recovery disposition suggested by the midlayer
1825  *
1826  *	This function is called by the SCSI midlayer upon completion of an
1827  *	error test command (currently TEST UNIT READY). The result of sending
1828  *	the eh command is passed in eh_disp.  We're looking for devices that
1829  *	fail medium access commands but are OK with non access commands like
1830  *	test unit ready (so wrongly see the device as having a successful
1831  *	recovery)
1832  **/
1833 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1834 {
1835 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1836 	struct scsi_device *sdev = scmd->device;
1837 
1838 	if (!scsi_device_online(sdev) ||
1839 	    !scsi_medium_access_command(scmd) ||
1840 	    host_byte(scmd->result) != DID_TIME_OUT ||
1841 	    eh_disp != SUCCESS)
1842 		return eh_disp;
1843 
1844 	/*
1845 	 * The device has timed out executing a medium access command.
1846 	 * However, the TEST UNIT READY command sent during error
1847 	 * handling completed successfully. Either the device is in the
1848 	 * process of recovering or has it suffered an internal failure
1849 	 * that prevents access to the storage medium.
1850 	 */
1851 	if (!sdkp->ignore_medium_access_errors) {
1852 		sdkp->medium_access_timed_out++;
1853 		sdkp->ignore_medium_access_errors = true;
1854 	}
1855 
1856 	/*
1857 	 * If the device keeps failing read/write commands but TEST UNIT
1858 	 * READY always completes successfully we assume that medium
1859 	 * access is no longer possible and take the device offline.
1860 	 */
1861 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1862 		scmd_printk(KERN_ERR, scmd,
1863 			    "Medium access timeout failure. Offlining disk!\n");
1864 		mutex_lock(&sdev->state_mutex);
1865 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1866 		mutex_unlock(&sdev->state_mutex);
1867 
1868 		return SUCCESS;
1869 	}
1870 
1871 	return eh_disp;
1872 }
1873 
1874 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1875 {
1876 	struct request *req = scsi_cmd_to_rq(scmd);
1877 	struct scsi_device *sdev = scmd->device;
1878 	unsigned int transferred, good_bytes;
1879 	u64 start_lba, end_lba, bad_lba;
1880 
1881 	/*
1882 	 * Some commands have a payload smaller than the device logical
1883 	 * block size (e.g. INQUIRY on a 4K disk).
1884 	 */
1885 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1886 		return 0;
1887 
1888 	/* Check if we have a 'bad_lba' information */
1889 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1890 				     SCSI_SENSE_BUFFERSIZE,
1891 				     &bad_lba))
1892 		return 0;
1893 
1894 	/*
1895 	 * If the bad lba was reported incorrectly, we have no idea where
1896 	 * the error is.
1897 	 */
1898 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1899 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1900 	if (bad_lba < start_lba || bad_lba >= end_lba)
1901 		return 0;
1902 
1903 	/*
1904 	 * resid is optional but mostly filled in.  When it's unused,
1905 	 * its value is zero, so we assume the whole buffer transferred
1906 	 */
1907 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1908 
1909 	/* This computation should always be done in terms of the
1910 	 * resolution of the device's medium.
1911 	 */
1912 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1913 
1914 	return min(good_bytes, transferred);
1915 }
1916 
1917 /**
1918  *	sd_done - bottom half handler: called when the lower level
1919  *	driver has completed (successfully or otherwise) a scsi command.
1920  *	@SCpnt: mid-level's per command structure.
1921  *
1922  *	Note: potentially run from within an ISR. Must not block.
1923  **/
1924 static int sd_done(struct scsi_cmnd *SCpnt)
1925 {
1926 	int result = SCpnt->result;
1927 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1928 	unsigned int sector_size = SCpnt->device->sector_size;
1929 	unsigned int resid;
1930 	struct scsi_sense_hdr sshdr;
1931 	struct request *req = scsi_cmd_to_rq(SCpnt);
1932 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1933 	int sense_valid = 0;
1934 	int sense_deferred = 0;
1935 
1936 	switch (req_op(req)) {
1937 	case REQ_OP_DISCARD:
1938 	case REQ_OP_WRITE_ZEROES:
1939 	case REQ_OP_ZONE_RESET:
1940 	case REQ_OP_ZONE_RESET_ALL:
1941 	case REQ_OP_ZONE_OPEN:
1942 	case REQ_OP_ZONE_CLOSE:
1943 	case REQ_OP_ZONE_FINISH:
1944 		if (!result) {
1945 			good_bytes = blk_rq_bytes(req);
1946 			scsi_set_resid(SCpnt, 0);
1947 		} else {
1948 			good_bytes = 0;
1949 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1950 		}
1951 		break;
1952 	default:
1953 		/*
1954 		 * In case of bogus fw or device, we could end up having
1955 		 * an unaligned partial completion. Check this here and force
1956 		 * alignment.
1957 		 */
1958 		resid = scsi_get_resid(SCpnt);
1959 		if (resid & (sector_size - 1)) {
1960 			sd_printk(KERN_INFO, sdkp,
1961 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
1962 				resid, sector_size);
1963 			scsi_print_command(SCpnt);
1964 			resid = min(scsi_bufflen(SCpnt),
1965 				    round_up(resid, sector_size));
1966 			scsi_set_resid(SCpnt, resid);
1967 		}
1968 	}
1969 
1970 	if (result) {
1971 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
1972 		if (sense_valid)
1973 			sense_deferred = scsi_sense_is_deferred(&sshdr);
1974 	}
1975 	sdkp->medium_access_timed_out = 0;
1976 
1977 	if (!scsi_status_is_check_condition(result) &&
1978 	    (!sense_valid || sense_deferred))
1979 		goto out;
1980 
1981 	switch (sshdr.sense_key) {
1982 	case HARDWARE_ERROR:
1983 	case MEDIUM_ERROR:
1984 		good_bytes = sd_completed_bytes(SCpnt);
1985 		break;
1986 	case RECOVERED_ERROR:
1987 		good_bytes = scsi_bufflen(SCpnt);
1988 		break;
1989 	case NO_SENSE:
1990 		/* This indicates a false check condition, so ignore it.  An
1991 		 * unknown amount of data was transferred so treat it as an
1992 		 * error.
1993 		 */
1994 		SCpnt->result = 0;
1995 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
1996 		break;
1997 	case ABORTED_COMMAND:
1998 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
1999 			good_bytes = sd_completed_bytes(SCpnt);
2000 		break;
2001 	case ILLEGAL_REQUEST:
2002 		switch (sshdr.asc) {
2003 		case 0x10:	/* DIX: Host detected corruption */
2004 			good_bytes = sd_completed_bytes(SCpnt);
2005 			break;
2006 		case 0x20:	/* INVALID COMMAND OPCODE */
2007 		case 0x24:	/* INVALID FIELD IN CDB */
2008 			switch (SCpnt->cmnd[0]) {
2009 			case UNMAP:
2010 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2011 				break;
2012 			case WRITE_SAME_16:
2013 			case WRITE_SAME:
2014 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2015 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2016 				} else {
2017 					sdkp->device->no_write_same = 1;
2018 					sd_config_write_same(sdkp);
2019 					req->rq_flags |= RQF_QUIET;
2020 				}
2021 				break;
2022 			}
2023 		}
2024 		break;
2025 	default:
2026 		break;
2027 	}
2028 
2029  out:
2030 	if (sd_is_zoned(sdkp))
2031 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2032 
2033 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2034 					   "sd_done: completed %d of %d bytes\n",
2035 					   good_bytes, scsi_bufflen(SCpnt)));
2036 
2037 	return good_bytes;
2038 }
2039 
2040 /*
2041  * spinup disk - called only in sd_revalidate_disk()
2042  */
2043 static void
2044 sd_spinup_disk(struct scsi_disk *sdkp)
2045 {
2046 	unsigned char cmd[10];
2047 	unsigned long spintime_expire = 0;
2048 	int retries, spintime;
2049 	unsigned int the_result;
2050 	struct scsi_sense_hdr sshdr;
2051 	int sense_valid = 0;
2052 
2053 	spintime = 0;
2054 
2055 	/* Spin up drives, as required.  Only do this at boot time */
2056 	/* Spinup needs to be done for module loads too. */
2057 	do {
2058 		retries = 0;
2059 
2060 		do {
2061 			bool media_was_present = sdkp->media_present;
2062 
2063 			cmd[0] = TEST_UNIT_READY;
2064 			memset((void *) &cmd[1], 0, 9);
2065 
2066 			the_result = scsi_execute_req(sdkp->device, cmd,
2067 						      DMA_NONE, NULL, 0,
2068 						      &sshdr, SD_TIMEOUT,
2069 						      sdkp->max_retries, NULL);
2070 
2071 			/*
2072 			 * If the drive has indicated to us that it
2073 			 * doesn't have any media in it, don't bother
2074 			 * with any more polling.
2075 			 */
2076 			if (media_not_present(sdkp, &sshdr)) {
2077 				if (media_was_present)
2078 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2079 				return;
2080 			}
2081 
2082 			if (the_result)
2083 				sense_valid = scsi_sense_valid(&sshdr);
2084 			retries++;
2085 		} while (retries < 3 &&
2086 			 (!scsi_status_is_good(the_result) ||
2087 			  (scsi_status_is_check_condition(the_result) &&
2088 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2089 
2090 		if (!scsi_status_is_check_condition(the_result)) {
2091 			/* no sense, TUR either succeeded or failed
2092 			 * with a status error */
2093 			if(!spintime && !scsi_status_is_good(the_result)) {
2094 				sd_print_result(sdkp, "Test Unit Ready failed",
2095 						the_result);
2096 			}
2097 			break;
2098 		}
2099 
2100 		/*
2101 		 * The device does not want the automatic start to be issued.
2102 		 */
2103 		if (sdkp->device->no_start_on_add)
2104 			break;
2105 
2106 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2107 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2108 				break;	/* manual intervention required */
2109 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2110 				break;	/* standby */
2111 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2112 				break;	/* unavailable */
2113 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2114 				break;	/* sanitize in progress */
2115 			/*
2116 			 * Issue command to spin up drive when not ready
2117 			 */
2118 			if (!spintime) {
2119 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2120 				cmd[0] = START_STOP;
2121 				cmd[1] = 1;	/* Return immediately */
2122 				memset((void *) &cmd[2], 0, 8);
2123 				cmd[4] = 1;	/* Start spin cycle */
2124 				if (sdkp->device->start_stop_pwr_cond)
2125 					cmd[4] |= 1 << 4;
2126 				scsi_execute_req(sdkp->device, cmd, DMA_NONE,
2127 						 NULL, 0, &sshdr,
2128 						 SD_TIMEOUT, sdkp->max_retries,
2129 						 NULL);
2130 				spintime_expire = jiffies + 100 * HZ;
2131 				spintime = 1;
2132 			}
2133 			/* Wait 1 second for next try */
2134 			msleep(1000);
2135 			printk(KERN_CONT ".");
2136 
2137 		/*
2138 		 * Wait for USB flash devices with slow firmware.
2139 		 * Yes, this sense key/ASC combination shouldn't
2140 		 * occur here.  It's characteristic of these devices.
2141 		 */
2142 		} else if (sense_valid &&
2143 				sshdr.sense_key == UNIT_ATTENTION &&
2144 				sshdr.asc == 0x28) {
2145 			if (!spintime) {
2146 				spintime_expire = jiffies + 5 * HZ;
2147 				spintime = 1;
2148 			}
2149 			/* Wait 1 second for next try */
2150 			msleep(1000);
2151 		} else {
2152 			/* we don't understand the sense code, so it's
2153 			 * probably pointless to loop */
2154 			if(!spintime) {
2155 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2156 				sd_print_sense_hdr(sdkp, &sshdr);
2157 			}
2158 			break;
2159 		}
2160 
2161 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2162 
2163 	if (spintime) {
2164 		if (scsi_status_is_good(the_result))
2165 			printk(KERN_CONT "ready\n");
2166 		else
2167 			printk(KERN_CONT "not responding...\n");
2168 	}
2169 }
2170 
2171 /*
2172  * Determine whether disk supports Data Integrity Field.
2173  */
2174 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2175 {
2176 	struct scsi_device *sdp = sdkp->device;
2177 	u8 type;
2178 
2179 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2180 		sdkp->protection_type = 0;
2181 		return 0;
2182 	}
2183 
2184 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2185 
2186 	if (type > T10_PI_TYPE3_PROTECTION) {
2187 		sd_printk(KERN_ERR, sdkp, "formatted with unsupported"	\
2188 			  " protection type %u. Disabling disk!\n",
2189 			  type);
2190 		sdkp->protection_type = 0;
2191 		return -ENODEV;
2192 	}
2193 
2194 	sdkp->protection_type = type;
2195 
2196 	return 0;
2197 }
2198 
2199 static void sd_config_protection(struct scsi_disk *sdkp)
2200 {
2201 	struct scsi_device *sdp = sdkp->device;
2202 
2203 	if (!sdkp->first_scan)
2204 		return;
2205 
2206 	sd_dif_config_host(sdkp);
2207 
2208 	if (!sdkp->protection_type)
2209 		return;
2210 
2211 	if (!scsi_host_dif_capable(sdp->host, sdkp->protection_type)) {
2212 		sd_printk(KERN_NOTICE, sdkp,
2213 			  "Disabling DIF Type %u protection\n",
2214 			  sdkp->protection_type);
2215 		sdkp->protection_type = 0;
2216 	}
2217 
2218 	sd_printk(KERN_NOTICE, sdkp, "Enabling DIF Type %u protection\n",
2219 		  sdkp->protection_type);
2220 }
2221 
2222 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2223 			struct scsi_sense_hdr *sshdr, int sense_valid,
2224 			int the_result)
2225 {
2226 	if (sense_valid)
2227 		sd_print_sense_hdr(sdkp, sshdr);
2228 	else
2229 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2230 
2231 	/*
2232 	 * Set dirty bit for removable devices if not ready -
2233 	 * sometimes drives will not report this properly.
2234 	 */
2235 	if (sdp->removable &&
2236 	    sense_valid && sshdr->sense_key == NOT_READY)
2237 		set_media_not_present(sdkp);
2238 
2239 	/*
2240 	 * We used to set media_present to 0 here to indicate no media
2241 	 * in the drive, but some drives fail read capacity even with
2242 	 * media present, so we can't do that.
2243 	 */
2244 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2245 }
2246 
2247 #define RC16_LEN 32
2248 #if RC16_LEN > SD_BUF_SIZE
2249 #error RC16_LEN must not be more than SD_BUF_SIZE
2250 #endif
2251 
2252 #define READ_CAPACITY_RETRIES_ON_RESET	10
2253 
2254 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2255 						unsigned char *buffer)
2256 {
2257 	unsigned char cmd[16];
2258 	struct scsi_sense_hdr sshdr;
2259 	int sense_valid = 0;
2260 	int the_result;
2261 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2262 	unsigned int alignment;
2263 	unsigned long long lba;
2264 	unsigned sector_size;
2265 
2266 	if (sdp->no_read_capacity_16)
2267 		return -EINVAL;
2268 
2269 	do {
2270 		memset(cmd, 0, 16);
2271 		cmd[0] = SERVICE_ACTION_IN_16;
2272 		cmd[1] = SAI_READ_CAPACITY_16;
2273 		cmd[13] = RC16_LEN;
2274 		memset(buffer, 0, RC16_LEN);
2275 
2276 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2277 					buffer, RC16_LEN, &sshdr,
2278 					SD_TIMEOUT, sdkp->max_retries, NULL);
2279 
2280 		if (media_not_present(sdkp, &sshdr))
2281 			return -ENODEV;
2282 
2283 		if (the_result > 0) {
2284 			sense_valid = scsi_sense_valid(&sshdr);
2285 			if (sense_valid &&
2286 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2287 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2288 			    sshdr.ascq == 0x00)
2289 				/* Invalid Command Operation Code or
2290 				 * Invalid Field in CDB, just retry
2291 				 * silently with RC10 */
2292 				return -EINVAL;
2293 			if (sense_valid &&
2294 			    sshdr.sense_key == UNIT_ATTENTION &&
2295 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2296 				/* Device reset might occur several times,
2297 				 * give it one more chance */
2298 				if (--reset_retries > 0)
2299 					continue;
2300 		}
2301 		retries--;
2302 
2303 	} while (the_result && retries);
2304 
2305 	if (the_result) {
2306 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2307 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2308 		return -EINVAL;
2309 	}
2310 
2311 	sector_size = get_unaligned_be32(&buffer[8]);
2312 	lba = get_unaligned_be64(&buffer[0]);
2313 
2314 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2315 		sdkp->capacity = 0;
2316 		return -ENODEV;
2317 	}
2318 
2319 	/* Logical blocks per physical block exponent */
2320 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2321 
2322 	/* RC basis */
2323 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2324 
2325 	/* Lowest aligned logical block */
2326 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2327 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2328 	if (alignment && sdkp->first_scan)
2329 		sd_printk(KERN_NOTICE, sdkp,
2330 			  "physical block alignment offset: %u\n", alignment);
2331 
2332 	if (buffer[14] & 0x80) { /* LBPME */
2333 		sdkp->lbpme = 1;
2334 
2335 		if (buffer[14] & 0x40) /* LBPRZ */
2336 			sdkp->lbprz = 1;
2337 
2338 		sd_config_discard(sdkp, SD_LBP_WS16);
2339 	}
2340 
2341 	sdkp->capacity = lba + 1;
2342 	return sector_size;
2343 }
2344 
2345 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2346 						unsigned char *buffer)
2347 {
2348 	unsigned char cmd[16];
2349 	struct scsi_sense_hdr sshdr;
2350 	int sense_valid = 0;
2351 	int the_result;
2352 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2353 	sector_t lba;
2354 	unsigned sector_size;
2355 
2356 	do {
2357 		cmd[0] = READ_CAPACITY;
2358 		memset(&cmd[1], 0, 9);
2359 		memset(buffer, 0, 8);
2360 
2361 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2362 					buffer, 8, &sshdr,
2363 					SD_TIMEOUT, sdkp->max_retries, NULL);
2364 
2365 		if (media_not_present(sdkp, &sshdr))
2366 			return -ENODEV;
2367 
2368 		if (the_result > 0) {
2369 			sense_valid = scsi_sense_valid(&sshdr);
2370 			if (sense_valid &&
2371 			    sshdr.sense_key == UNIT_ATTENTION &&
2372 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2373 				/* Device reset might occur several times,
2374 				 * give it one more chance */
2375 				if (--reset_retries > 0)
2376 					continue;
2377 		}
2378 		retries--;
2379 
2380 	} while (the_result && retries);
2381 
2382 	if (the_result) {
2383 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2384 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2385 		return -EINVAL;
2386 	}
2387 
2388 	sector_size = get_unaligned_be32(&buffer[4]);
2389 	lba = get_unaligned_be32(&buffer[0]);
2390 
2391 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2392 		/* Some buggy (usb cardreader) devices return an lba of
2393 		   0xffffffff when the want to report a size of 0 (with
2394 		   which they really mean no media is present) */
2395 		sdkp->capacity = 0;
2396 		sdkp->physical_block_size = sector_size;
2397 		return sector_size;
2398 	}
2399 
2400 	sdkp->capacity = lba + 1;
2401 	sdkp->physical_block_size = sector_size;
2402 	return sector_size;
2403 }
2404 
2405 static int sd_try_rc16_first(struct scsi_device *sdp)
2406 {
2407 	if (sdp->host->max_cmd_len < 16)
2408 		return 0;
2409 	if (sdp->try_rc_10_first)
2410 		return 0;
2411 	if (sdp->scsi_level > SCSI_SPC_2)
2412 		return 1;
2413 	if (scsi_device_protection(sdp))
2414 		return 1;
2415 	return 0;
2416 }
2417 
2418 /*
2419  * read disk capacity
2420  */
2421 static void
2422 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2423 {
2424 	int sector_size;
2425 	struct scsi_device *sdp = sdkp->device;
2426 
2427 	if (sd_try_rc16_first(sdp)) {
2428 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2429 		if (sector_size == -EOVERFLOW)
2430 			goto got_data;
2431 		if (sector_size == -ENODEV)
2432 			return;
2433 		if (sector_size < 0)
2434 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2435 		if (sector_size < 0)
2436 			return;
2437 	} else {
2438 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2439 		if (sector_size == -EOVERFLOW)
2440 			goto got_data;
2441 		if (sector_size < 0)
2442 			return;
2443 		if ((sizeof(sdkp->capacity) > 4) &&
2444 		    (sdkp->capacity > 0xffffffffULL)) {
2445 			int old_sector_size = sector_size;
2446 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2447 					"Trying to use READ CAPACITY(16).\n");
2448 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2449 			if (sector_size < 0) {
2450 				sd_printk(KERN_NOTICE, sdkp,
2451 					"Using 0xffffffff as device size\n");
2452 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2453 				sector_size = old_sector_size;
2454 				goto got_data;
2455 			}
2456 			/* Remember that READ CAPACITY(16) succeeded */
2457 			sdp->try_rc_10_first = 0;
2458 		}
2459 	}
2460 
2461 	/* Some devices are known to return the total number of blocks,
2462 	 * not the highest block number.  Some devices have versions
2463 	 * which do this and others which do not.  Some devices we might
2464 	 * suspect of doing this but we don't know for certain.
2465 	 *
2466 	 * If we know the reported capacity is wrong, decrement it.  If
2467 	 * we can only guess, then assume the number of blocks is even
2468 	 * (usually true but not always) and err on the side of lowering
2469 	 * the capacity.
2470 	 */
2471 	if (sdp->fix_capacity ||
2472 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2473 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2474 				"from its reported value: %llu\n",
2475 				(unsigned long long) sdkp->capacity);
2476 		--sdkp->capacity;
2477 	}
2478 
2479 got_data:
2480 	if (sector_size == 0) {
2481 		sector_size = 512;
2482 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2483 			  "assuming 512.\n");
2484 	}
2485 
2486 	if (sector_size != 512 &&
2487 	    sector_size != 1024 &&
2488 	    sector_size != 2048 &&
2489 	    sector_size != 4096) {
2490 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2491 			  sector_size);
2492 		/*
2493 		 * The user might want to re-format the drive with
2494 		 * a supported sectorsize.  Once this happens, it
2495 		 * would be relatively trivial to set the thing up.
2496 		 * For this reason, we leave the thing in the table.
2497 		 */
2498 		sdkp->capacity = 0;
2499 		/*
2500 		 * set a bogus sector size so the normal read/write
2501 		 * logic in the block layer will eventually refuse any
2502 		 * request on this device without tripping over power
2503 		 * of two sector size assumptions
2504 		 */
2505 		sector_size = 512;
2506 	}
2507 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2508 	blk_queue_physical_block_size(sdp->request_queue,
2509 				      sdkp->physical_block_size);
2510 	sdkp->device->sector_size = sector_size;
2511 
2512 	if (sdkp->capacity > 0xffffffff)
2513 		sdp->use_16_for_rw = 1;
2514 
2515 }
2516 
2517 /*
2518  * Print disk capacity
2519  */
2520 static void
2521 sd_print_capacity(struct scsi_disk *sdkp,
2522 		  sector_t old_capacity)
2523 {
2524 	int sector_size = sdkp->device->sector_size;
2525 	char cap_str_2[10], cap_str_10[10];
2526 
2527 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2528 		return;
2529 
2530 	string_get_size(sdkp->capacity, sector_size,
2531 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2532 	string_get_size(sdkp->capacity, sector_size,
2533 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2534 
2535 	sd_printk(KERN_NOTICE, sdkp,
2536 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2537 		  (unsigned long long)sdkp->capacity,
2538 		  sector_size, cap_str_10, cap_str_2);
2539 
2540 	if (sdkp->physical_block_size != sector_size)
2541 		sd_printk(KERN_NOTICE, sdkp,
2542 			  "%u-byte physical blocks\n",
2543 			  sdkp->physical_block_size);
2544 }
2545 
2546 /* called with buffer of length 512 */
2547 static inline int
2548 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2549 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2550 		 struct scsi_sense_hdr *sshdr)
2551 {
2552 	/*
2553 	 * If we must use MODE SENSE(10), make sure that the buffer length
2554 	 * is at least 8 bytes so that the mode sense header fits.
2555 	 */
2556 	if (sdkp->device->use_10_for_ms && len < 8)
2557 		len = 8;
2558 
2559 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2560 			       SD_TIMEOUT, sdkp->max_retries, data,
2561 			       sshdr);
2562 }
2563 
2564 /*
2565  * read write protect setting, if possible - called only in sd_revalidate_disk()
2566  * called with buffer of length SD_BUF_SIZE
2567  */
2568 static void
2569 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2570 {
2571 	int res;
2572 	struct scsi_device *sdp = sdkp->device;
2573 	struct scsi_mode_data data;
2574 	int old_wp = sdkp->write_prot;
2575 
2576 	set_disk_ro(sdkp->disk, 0);
2577 	if (sdp->skip_ms_page_3f) {
2578 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2579 		return;
2580 	}
2581 
2582 	if (sdp->use_192_bytes_for_3f) {
2583 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2584 	} else {
2585 		/*
2586 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2587 		 * We have to start carefully: some devices hang if we ask
2588 		 * for more than is available.
2589 		 */
2590 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2591 
2592 		/*
2593 		 * Second attempt: ask for page 0 When only page 0 is
2594 		 * implemented, a request for page 3F may return Sense Key
2595 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2596 		 * CDB.
2597 		 */
2598 		if (res < 0)
2599 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2600 
2601 		/*
2602 		 * Third attempt: ask 255 bytes, as we did earlier.
2603 		 */
2604 		if (res < 0)
2605 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2606 					       &data, NULL);
2607 	}
2608 
2609 	if (res < 0) {
2610 		sd_first_printk(KERN_WARNING, sdkp,
2611 			  "Test WP failed, assume Write Enabled\n");
2612 	} else {
2613 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2614 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2615 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2616 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2617 				  sdkp->write_prot ? "on" : "off");
2618 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2619 		}
2620 	}
2621 }
2622 
2623 /*
2624  * sd_read_cache_type - called only from sd_revalidate_disk()
2625  * called with buffer of length SD_BUF_SIZE
2626  */
2627 static void
2628 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2629 {
2630 	int len = 0, res;
2631 	struct scsi_device *sdp = sdkp->device;
2632 
2633 	int dbd;
2634 	int modepage;
2635 	int first_len;
2636 	struct scsi_mode_data data;
2637 	struct scsi_sense_hdr sshdr;
2638 	int old_wce = sdkp->WCE;
2639 	int old_rcd = sdkp->RCD;
2640 	int old_dpofua = sdkp->DPOFUA;
2641 
2642 
2643 	if (sdkp->cache_override)
2644 		return;
2645 
2646 	first_len = 4;
2647 	if (sdp->skip_ms_page_8) {
2648 		if (sdp->type == TYPE_RBC)
2649 			goto defaults;
2650 		else {
2651 			if (sdp->skip_ms_page_3f)
2652 				goto defaults;
2653 			modepage = 0x3F;
2654 			if (sdp->use_192_bytes_for_3f)
2655 				first_len = 192;
2656 			dbd = 0;
2657 		}
2658 	} else if (sdp->type == TYPE_RBC) {
2659 		modepage = 6;
2660 		dbd = 8;
2661 	} else {
2662 		modepage = 8;
2663 		dbd = 0;
2664 	}
2665 
2666 	/* cautiously ask */
2667 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2668 			&data, &sshdr);
2669 
2670 	if (res < 0)
2671 		goto bad_sense;
2672 
2673 	if (!data.header_length) {
2674 		modepage = 6;
2675 		first_len = 0;
2676 		sd_first_printk(KERN_ERR, sdkp,
2677 				"Missing header in MODE_SENSE response\n");
2678 	}
2679 
2680 	/* that went OK, now ask for the proper length */
2681 	len = data.length;
2682 
2683 	/*
2684 	 * We're only interested in the first three bytes, actually.
2685 	 * But the data cache page is defined for the first 20.
2686 	 */
2687 	if (len < 3)
2688 		goto bad_sense;
2689 	else if (len > SD_BUF_SIZE) {
2690 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2691 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2692 		len = SD_BUF_SIZE;
2693 	}
2694 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2695 		len = 192;
2696 
2697 	/* Get the data */
2698 	if (len > first_len)
2699 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2700 				&data, &sshdr);
2701 
2702 	if (!res) {
2703 		int offset = data.header_length + data.block_descriptor_length;
2704 
2705 		while (offset < len) {
2706 			u8 page_code = buffer[offset] & 0x3F;
2707 			u8 spf       = buffer[offset] & 0x40;
2708 
2709 			if (page_code == 8 || page_code == 6) {
2710 				/* We're interested only in the first 3 bytes.
2711 				 */
2712 				if (len - offset <= 2) {
2713 					sd_first_printk(KERN_ERR, sdkp,
2714 						"Incomplete mode parameter "
2715 							"data\n");
2716 					goto defaults;
2717 				} else {
2718 					modepage = page_code;
2719 					goto Page_found;
2720 				}
2721 			} else {
2722 				/* Go to the next page */
2723 				if (spf && len - offset > 3)
2724 					offset += 4 + (buffer[offset+2] << 8) +
2725 						buffer[offset+3];
2726 				else if (!spf && len - offset > 1)
2727 					offset += 2 + buffer[offset+1];
2728 				else {
2729 					sd_first_printk(KERN_ERR, sdkp,
2730 							"Incomplete mode "
2731 							"parameter data\n");
2732 					goto defaults;
2733 				}
2734 			}
2735 		}
2736 
2737 		sd_first_printk(KERN_WARNING, sdkp,
2738 				"No Caching mode page found\n");
2739 		goto defaults;
2740 
2741 	Page_found:
2742 		if (modepage == 8) {
2743 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2744 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2745 		} else {
2746 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2747 			sdkp->RCD = 0;
2748 		}
2749 
2750 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2751 		if (sdp->broken_fua) {
2752 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2753 			sdkp->DPOFUA = 0;
2754 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2755 			   !sdkp->device->use_16_for_rw) {
2756 			sd_first_printk(KERN_NOTICE, sdkp,
2757 				  "Uses READ/WRITE(6), disabling FUA\n");
2758 			sdkp->DPOFUA = 0;
2759 		}
2760 
2761 		/* No cache flush allowed for write protected devices */
2762 		if (sdkp->WCE && sdkp->write_prot)
2763 			sdkp->WCE = 0;
2764 
2765 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2766 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2767 			sd_printk(KERN_NOTICE, sdkp,
2768 				  "Write cache: %s, read cache: %s, %s\n",
2769 				  sdkp->WCE ? "enabled" : "disabled",
2770 				  sdkp->RCD ? "disabled" : "enabled",
2771 				  sdkp->DPOFUA ? "supports DPO and FUA"
2772 				  : "doesn't support DPO or FUA");
2773 
2774 		return;
2775 	}
2776 
2777 bad_sense:
2778 	if (scsi_sense_valid(&sshdr) &&
2779 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2780 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2781 		/* Invalid field in CDB */
2782 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2783 	else
2784 		sd_first_printk(KERN_ERR, sdkp,
2785 				"Asking for cache data failed\n");
2786 
2787 defaults:
2788 	if (sdp->wce_default_on) {
2789 		sd_first_printk(KERN_NOTICE, sdkp,
2790 				"Assuming drive cache: write back\n");
2791 		sdkp->WCE = 1;
2792 	} else {
2793 		sd_first_printk(KERN_WARNING, sdkp,
2794 				"Assuming drive cache: write through\n");
2795 		sdkp->WCE = 0;
2796 	}
2797 	sdkp->RCD = 0;
2798 	sdkp->DPOFUA = 0;
2799 }
2800 
2801 /*
2802  * The ATO bit indicates whether the DIF application tag is available
2803  * for use by the operating system.
2804  */
2805 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2806 {
2807 	int res, offset;
2808 	struct scsi_device *sdp = sdkp->device;
2809 	struct scsi_mode_data data;
2810 	struct scsi_sense_hdr sshdr;
2811 
2812 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2813 		return;
2814 
2815 	if (sdkp->protection_type == 0)
2816 		return;
2817 
2818 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2819 			      sdkp->max_retries, &data, &sshdr);
2820 
2821 	if (res < 0 || !data.header_length ||
2822 	    data.length < 6) {
2823 		sd_first_printk(KERN_WARNING, sdkp,
2824 			  "getting Control mode page failed, assume no ATO\n");
2825 
2826 		if (scsi_sense_valid(&sshdr))
2827 			sd_print_sense_hdr(sdkp, &sshdr);
2828 
2829 		return;
2830 	}
2831 
2832 	offset = data.header_length + data.block_descriptor_length;
2833 
2834 	if ((buffer[offset] & 0x3f) != 0x0a) {
2835 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2836 		return;
2837 	}
2838 
2839 	if ((buffer[offset + 5] & 0x80) == 0)
2840 		return;
2841 
2842 	sdkp->ATO = 1;
2843 
2844 	return;
2845 }
2846 
2847 /**
2848  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2849  * @sdkp: disk to query
2850  */
2851 static void sd_read_block_limits(struct scsi_disk *sdkp)
2852 {
2853 	struct scsi_vpd *vpd;
2854 
2855 	rcu_read_lock();
2856 
2857 	vpd = rcu_dereference(sdkp->device->vpd_pgb0);
2858 	if (!vpd || vpd->len < 16)
2859 		goto out;
2860 
2861 	sdkp->min_xfer_blocks = get_unaligned_be16(&vpd->data[6]);
2862 	sdkp->max_xfer_blocks = get_unaligned_be32(&vpd->data[8]);
2863 	sdkp->opt_xfer_blocks = get_unaligned_be32(&vpd->data[12]);
2864 
2865 	if (vpd->len >= 64) {
2866 		unsigned int lba_count, desc_count;
2867 
2868 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&vpd->data[36]);
2869 
2870 		if (!sdkp->lbpme)
2871 			goto out;
2872 
2873 		lba_count = get_unaligned_be32(&vpd->data[20]);
2874 		desc_count = get_unaligned_be32(&vpd->data[24]);
2875 
2876 		if (lba_count && desc_count)
2877 			sdkp->max_unmap_blocks = lba_count;
2878 
2879 		sdkp->unmap_granularity = get_unaligned_be32(&vpd->data[28]);
2880 
2881 		if (vpd->data[32] & 0x80)
2882 			sdkp->unmap_alignment =
2883 				get_unaligned_be32(&vpd->data[32]) & ~(1 << 31);
2884 
2885 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2886 
2887 			if (sdkp->max_unmap_blocks)
2888 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2889 			else
2890 				sd_config_discard(sdkp, SD_LBP_WS16);
2891 
2892 		} else {	/* LBP VPD page tells us what to use */
2893 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2894 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2895 			else if (sdkp->lbpws)
2896 				sd_config_discard(sdkp, SD_LBP_WS16);
2897 			else if (sdkp->lbpws10)
2898 				sd_config_discard(sdkp, SD_LBP_WS10);
2899 			else
2900 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2901 		}
2902 	}
2903 
2904  out:
2905 	rcu_read_unlock();
2906 }
2907 
2908 /**
2909  * sd_read_block_characteristics - Query block dev. characteristics
2910  * @sdkp: disk to query
2911  */
2912 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2913 {
2914 	struct request_queue *q = sdkp->disk->queue;
2915 	struct scsi_vpd *vpd;
2916 	u16 rot;
2917 	u8 zoned;
2918 
2919 	rcu_read_lock();
2920 	vpd = rcu_dereference(sdkp->device->vpd_pgb1);
2921 
2922 	if (!vpd || vpd->len < 8) {
2923 		rcu_read_unlock();
2924 	        return;
2925 	}
2926 
2927 	rot = get_unaligned_be16(&vpd->data[4]);
2928 	zoned = (vpd->data[8] >> 4) & 3;
2929 	rcu_read_unlock();
2930 
2931 	if (rot == 1) {
2932 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2933 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2934 	}
2935 
2936 	if (sdkp->device->type == TYPE_ZBC) {
2937 		/* Host-managed */
2938 		disk_set_zoned(sdkp->disk, BLK_ZONED_HM);
2939 	} else {
2940 		sdkp->zoned = zoned;
2941 		if (sdkp->zoned == 1) {
2942 			/* Host-aware */
2943 			disk_set_zoned(sdkp->disk, BLK_ZONED_HA);
2944 		} else {
2945 			/* Regular disk or drive managed disk */
2946 			disk_set_zoned(sdkp->disk, BLK_ZONED_NONE);
2947 		}
2948 	}
2949 
2950 	if (!sdkp->first_scan)
2951 		return;
2952 
2953 	if (blk_queue_is_zoned(q)) {
2954 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
2955 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
2956 	} else {
2957 		if (sdkp->zoned == 1)
2958 			sd_printk(KERN_NOTICE, sdkp,
2959 				  "Host-aware SMR disk used as regular disk\n");
2960 		else if (sdkp->zoned == 2)
2961 			sd_printk(KERN_NOTICE, sdkp,
2962 				  "Drive-managed SMR disk\n");
2963 	}
2964 }
2965 
2966 /**
2967  * sd_read_block_provisioning - Query provisioning VPD page
2968  * @sdkp: disk to query
2969  */
2970 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
2971 {
2972 	struct scsi_vpd *vpd;
2973 
2974 	if (sdkp->lbpme == 0)
2975 		return;
2976 
2977 	rcu_read_lock();
2978 	vpd = rcu_dereference(sdkp->device->vpd_pgb2);
2979 
2980 	if (!vpd || vpd->len < 8) {
2981 		rcu_read_unlock();
2982 		return;
2983 	}
2984 
2985 	sdkp->lbpvpd	= 1;
2986 	sdkp->lbpu	= (vpd->data[5] >> 7) & 1; /* UNMAP */
2987 	sdkp->lbpws	= (vpd->data[5] >> 6) & 1; /* WRITE SAME(16) w/ UNMAP */
2988 	sdkp->lbpws10	= (vpd->data[5] >> 5) & 1; /* WRITE SAME(10) w/ UNMAP */
2989 	rcu_read_unlock();
2990 }
2991 
2992 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
2993 {
2994 	struct scsi_device *sdev = sdkp->device;
2995 
2996 	if (sdev->host->no_write_same) {
2997 		sdev->no_write_same = 1;
2998 
2999 		return;
3000 	}
3001 
3002 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
3003 		struct scsi_vpd *vpd;
3004 
3005 		sdev->no_report_opcodes = 1;
3006 
3007 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3008 		 * CODES is unsupported and the device has an ATA
3009 		 * Information VPD page (SAT).
3010 		 */
3011 		rcu_read_lock();
3012 		vpd = rcu_dereference(sdev->vpd_pg89);
3013 		if (vpd)
3014 			sdev->no_write_same = 1;
3015 		rcu_read_unlock();
3016 	}
3017 
3018 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3019 		sdkp->ws16 = 1;
3020 
3021 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3022 		sdkp->ws10 = 1;
3023 }
3024 
3025 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3026 {
3027 	struct scsi_device *sdev = sdkp->device;
3028 
3029 	if (!sdev->security_supported)
3030 		return;
3031 
3032 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3033 			SECURITY_PROTOCOL_IN) == 1 &&
3034 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3035 			SECURITY_PROTOCOL_OUT) == 1)
3036 		sdkp->security = 1;
3037 }
3038 
3039 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3040 {
3041 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3042 }
3043 
3044 /**
3045  * sd_read_cpr - Query concurrent positioning ranges
3046  * @sdkp:	disk to query
3047  */
3048 static void sd_read_cpr(struct scsi_disk *sdkp)
3049 {
3050 	struct blk_independent_access_ranges *iars = NULL;
3051 	unsigned char *buffer = NULL;
3052 	unsigned int nr_cpr = 0;
3053 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3054 	u8 *desc;
3055 
3056 	/*
3057 	 * We need to have the capacity set first for the block layer to be
3058 	 * able to check the ranges.
3059 	 */
3060 	if (sdkp->first_scan)
3061 		return;
3062 
3063 	if (!sdkp->capacity)
3064 		goto out;
3065 
3066 	/*
3067 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3068 	 * leading to a maximum page size of 64 + 256*32 bytes.
3069 	 */
3070 	buf_len = 64 + 256*32;
3071 	buffer = kmalloc(buf_len, GFP_KERNEL);
3072 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3073 		goto out;
3074 
3075 	/* We must have at least a 64B header and one 32B range descriptor */
3076 	vpd_len = get_unaligned_be16(&buffer[2]) + 4;
3077 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3078 		sd_printk(KERN_ERR, sdkp,
3079 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3080 		goto out;
3081 	}
3082 
3083 	nr_cpr = (vpd_len - 64) / 32;
3084 	if (nr_cpr == 1) {
3085 		nr_cpr = 0;
3086 		goto out;
3087 	}
3088 
3089 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3090 	if (!iars) {
3091 		nr_cpr = 0;
3092 		goto out;
3093 	}
3094 
3095 	desc = &buffer[64];
3096 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3097 		if (desc[0] != i) {
3098 			sd_printk(KERN_ERR, sdkp,
3099 				"Invalid Concurrent Positioning Range number\n");
3100 			nr_cpr = 0;
3101 			break;
3102 		}
3103 
3104 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3105 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3106 	}
3107 
3108 out:
3109 	disk_set_independent_access_ranges(sdkp->disk, iars);
3110 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3111 		sd_printk(KERN_NOTICE, sdkp,
3112 			  "%u concurrent positioning ranges\n", nr_cpr);
3113 		sdkp->nr_actuators = nr_cpr;
3114 	}
3115 
3116 	kfree(buffer);
3117 }
3118 
3119 static bool sd_validate_min_xfer_size(struct scsi_disk *sdkp)
3120 {
3121 	struct scsi_device *sdp = sdkp->device;
3122 	unsigned int min_xfer_bytes =
3123 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3124 
3125 	if (sdkp->min_xfer_blocks == 0)
3126 		return false;
3127 
3128 	if (min_xfer_bytes & (sdkp->physical_block_size - 1)) {
3129 		sd_first_printk(KERN_WARNING, sdkp,
3130 				"Preferred minimum I/O size %u bytes not a " \
3131 				"multiple of physical block size (%u bytes)\n",
3132 				min_xfer_bytes, sdkp->physical_block_size);
3133 		sdkp->min_xfer_blocks = 0;
3134 		return false;
3135 	}
3136 
3137 	sd_first_printk(KERN_INFO, sdkp, "Preferred minimum I/O size %u bytes\n",
3138 			min_xfer_bytes);
3139 	return true;
3140 }
3141 
3142 /*
3143  * Determine the device's preferred I/O size for reads and writes
3144  * unless the reported value is unreasonably small, large, not a
3145  * multiple of the physical block size, or simply garbage.
3146  */
3147 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3148 				      unsigned int dev_max)
3149 {
3150 	struct scsi_device *sdp = sdkp->device;
3151 	unsigned int opt_xfer_bytes =
3152 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3153 	unsigned int min_xfer_bytes =
3154 		logical_to_bytes(sdp, sdkp->min_xfer_blocks);
3155 
3156 	if (sdkp->opt_xfer_blocks == 0)
3157 		return false;
3158 
3159 	if (sdkp->opt_xfer_blocks > dev_max) {
3160 		sd_first_printk(KERN_WARNING, sdkp,
3161 				"Optimal transfer size %u logical blocks " \
3162 				"> dev_max (%u logical blocks)\n",
3163 				sdkp->opt_xfer_blocks, dev_max);
3164 		return false;
3165 	}
3166 
3167 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3168 		sd_first_printk(KERN_WARNING, sdkp,
3169 				"Optimal transfer size %u logical blocks " \
3170 				"> sd driver limit (%u logical blocks)\n",
3171 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3172 		return false;
3173 	}
3174 
3175 	if (opt_xfer_bytes < PAGE_SIZE) {
3176 		sd_first_printk(KERN_WARNING, sdkp,
3177 				"Optimal transfer size %u bytes < " \
3178 				"PAGE_SIZE (%u bytes)\n",
3179 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3180 		return false;
3181 	}
3182 
3183 	if (min_xfer_bytes && opt_xfer_bytes % min_xfer_bytes) {
3184 		sd_first_printk(KERN_WARNING, sdkp,
3185 				"Optimal transfer size %u bytes not a " \
3186 				"multiple of preferred minimum block " \
3187 				"size (%u bytes)\n",
3188 				opt_xfer_bytes, min_xfer_bytes);
3189 		return false;
3190 	}
3191 
3192 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3193 		sd_first_printk(KERN_WARNING, sdkp,
3194 				"Optimal transfer size %u bytes not a " \
3195 				"multiple of physical block size (%u bytes)\n",
3196 				opt_xfer_bytes, sdkp->physical_block_size);
3197 		return false;
3198 	}
3199 
3200 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3201 			opt_xfer_bytes);
3202 	return true;
3203 }
3204 
3205 /**
3206  *	sd_revalidate_disk - called the first time a new disk is seen,
3207  *	performs disk spin up, read_capacity, etc.
3208  *	@disk: struct gendisk we care about
3209  **/
3210 static int sd_revalidate_disk(struct gendisk *disk)
3211 {
3212 	struct scsi_disk *sdkp = scsi_disk(disk);
3213 	struct scsi_device *sdp = sdkp->device;
3214 	struct request_queue *q = sdkp->disk->queue;
3215 	sector_t old_capacity = sdkp->capacity;
3216 	unsigned char *buffer;
3217 	unsigned int dev_max, rw_max;
3218 
3219 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3220 				      "sd_revalidate_disk\n"));
3221 
3222 	/*
3223 	 * If the device is offline, don't try and read capacity or any
3224 	 * of the other niceties.
3225 	 */
3226 	if (!scsi_device_online(sdp))
3227 		goto out;
3228 
3229 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3230 	if (!buffer) {
3231 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3232 			  "allocation failure.\n");
3233 		goto out;
3234 	}
3235 
3236 	sd_spinup_disk(sdkp);
3237 
3238 	/*
3239 	 * Without media there is no reason to ask; moreover, some devices
3240 	 * react badly if we do.
3241 	 */
3242 	if (sdkp->media_present) {
3243 		sd_read_capacity(sdkp, buffer);
3244 
3245 		/*
3246 		 * set the default to rotational.  All non-rotational devices
3247 		 * support the block characteristics VPD page, which will
3248 		 * cause this to be updated correctly and any device which
3249 		 * doesn't support it should be treated as rotational.
3250 		 */
3251 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3252 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3253 
3254 		if (scsi_device_supports_vpd(sdp)) {
3255 			sd_read_block_provisioning(sdkp);
3256 			sd_read_block_limits(sdkp);
3257 			sd_read_block_characteristics(sdkp);
3258 			sd_zbc_read_zones(sdkp, buffer);
3259 			sd_read_cpr(sdkp);
3260 		}
3261 
3262 		sd_print_capacity(sdkp, old_capacity);
3263 
3264 		sd_read_write_protect_flag(sdkp, buffer);
3265 		sd_read_cache_type(sdkp, buffer);
3266 		sd_read_app_tag_own(sdkp, buffer);
3267 		sd_read_write_same(sdkp, buffer);
3268 		sd_read_security(sdkp, buffer);
3269 		sd_config_protection(sdkp);
3270 	}
3271 
3272 	/*
3273 	 * We now have all cache related info, determine how we deal
3274 	 * with flush requests.
3275 	 */
3276 	sd_set_flush_flag(sdkp);
3277 
3278 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3279 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3280 
3281 	/* Some devices report a maximum block count for READ/WRITE requests. */
3282 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3283 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3284 
3285 	if (sd_validate_min_xfer_size(sdkp))
3286 		blk_queue_io_min(sdkp->disk->queue,
3287 				 logical_to_bytes(sdp, sdkp->min_xfer_blocks));
3288 	else
3289 		blk_queue_io_min(sdkp->disk->queue, 0);
3290 
3291 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3292 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3293 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3294 	} else {
3295 		q->limits.io_opt = 0;
3296 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3297 				      (sector_t)BLK_DEF_MAX_SECTORS);
3298 	}
3299 
3300 	/*
3301 	 * Limit default to SCSI host optimal sector limit if set. There may be
3302 	 * an impact on performance for when the size of a request exceeds this
3303 	 * host limit.
3304 	 */
3305 	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
3306 
3307 	/* Do not exceed controller limit */
3308 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3309 
3310 	/*
3311 	 * Only update max_sectors if previously unset or if the current value
3312 	 * exceeds the capabilities of the hardware.
3313 	 */
3314 	if (sdkp->first_scan ||
3315 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3316 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3317 		q->limits.max_sectors = rw_max;
3318 
3319 	sdkp->first_scan = 0;
3320 
3321 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3322 	sd_config_write_same(sdkp);
3323 	kfree(buffer);
3324 
3325 	/*
3326 	 * For a zoned drive, revalidating the zones can be done only once
3327 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3328 	 * capacity to 0.
3329 	 */
3330 	if (sd_zbc_revalidate_zones(sdkp))
3331 		set_capacity_and_notify(disk, 0);
3332 
3333  out:
3334 	return 0;
3335 }
3336 
3337 /**
3338  *	sd_unlock_native_capacity - unlock native capacity
3339  *	@disk: struct gendisk to set capacity for
3340  *
3341  *	Block layer calls this function if it detects that partitions
3342  *	on @disk reach beyond the end of the device.  If the SCSI host
3343  *	implements ->unlock_native_capacity() method, it's invoked to
3344  *	give it a chance to adjust the device capacity.
3345  *
3346  *	CONTEXT:
3347  *	Defined by block layer.  Might sleep.
3348  */
3349 static void sd_unlock_native_capacity(struct gendisk *disk)
3350 {
3351 	struct scsi_device *sdev = scsi_disk(disk)->device;
3352 
3353 	if (sdev->host->hostt->unlock_native_capacity)
3354 		sdev->host->hostt->unlock_native_capacity(sdev);
3355 }
3356 
3357 /**
3358  *	sd_format_disk_name - format disk name
3359  *	@prefix: name prefix - ie. "sd" for SCSI disks
3360  *	@index: index of the disk to format name for
3361  *	@buf: output buffer
3362  *	@buflen: length of the output buffer
3363  *
3364  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3365  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3366  *	which is followed by sdaaa.
3367  *
3368  *	This is basically 26 base counting with one extra 'nil' entry
3369  *	at the beginning from the second digit on and can be
3370  *	determined using similar method as 26 base conversion with the
3371  *	index shifted -1 after each digit is computed.
3372  *
3373  *	CONTEXT:
3374  *	Don't care.
3375  *
3376  *	RETURNS:
3377  *	0 on success, -errno on failure.
3378  */
3379 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3380 {
3381 	const int base = 'z' - 'a' + 1;
3382 	char *begin = buf + strlen(prefix);
3383 	char *end = buf + buflen;
3384 	char *p;
3385 	int unit;
3386 
3387 	p = end - 1;
3388 	*p = '\0';
3389 	unit = base;
3390 	do {
3391 		if (p == begin)
3392 			return -EINVAL;
3393 		*--p = 'a' + (index % unit);
3394 		index = (index / unit) - 1;
3395 	} while (index >= 0);
3396 
3397 	memmove(begin, p, end - p);
3398 	memcpy(buf, prefix, strlen(prefix));
3399 
3400 	return 0;
3401 }
3402 
3403 /**
3404  *	sd_probe - called during driver initialization and whenever a
3405  *	new scsi device is attached to the system. It is called once
3406  *	for each scsi device (not just disks) present.
3407  *	@dev: pointer to device object
3408  *
3409  *	Returns 0 if successful (or not interested in this scsi device
3410  *	(e.g. scanner)); 1 when there is an error.
3411  *
3412  *	Note: this function is invoked from the scsi mid-level.
3413  *	This function sets up the mapping between a given
3414  *	<host,channel,id,lun> (found in sdp) and new device name
3415  *	(e.g. /dev/sda). More precisely it is the block device major
3416  *	and minor number that is chosen here.
3417  *
3418  *	Assume sd_probe is not re-entrant (for time being)
3419  *	Also think about sd_probe() and sd_remove() running coincidentally.
3420  **/
3421 static int sd_probe(struct device *dev)
3422 {
3423 	struct scsi_device *sdp = to_scsi_device(dev);
3424 	struct scsi_disk *sdkp;
3425 	struct gendisk *gd;
3426 	int index;
3427 	int error;
3428 
3429 	scsi_autopm_get_device(sdp);
3430 	error = -ENODEV;
3431 	if (sdp->type != TYPE_DISK &&
3432 	    sdp->type != TYPE_ZBC &&
3433 	    sdp->type != TYPE_MOD &&
3434 	    sdp->type != TYPE_RBC)
3435 		goto out;
3436 
3437 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3438 		sdev_printk(KERN_WARNING, sdp,
3439 			    "Unsupported ZBC host-managed device.\n");
3440 		goto out;
3441 	}
3442 
3443 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3444 					"sd_probe\n"));
3445 
3446 	error = -ENOMEM;
3447 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3448 	if (!sdkp)
3449 		goto out;
3450 
3451 	gd = blk_mq_alloc_disk_for_queue(sdp->request_queue,
3452 					 &sd_bio_compl_lkclass);
3453 	if (!gd)
3454 		goto out_free;
3455 
3456 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3457 	if (index < 0) {
3458 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3459 		goto out_put;
3460 	}
3461 
3462 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3463 	if (error) {
3464 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3465 		goto out_free_index;
3466 	}
3467 
3468 	sdkp->device = sdp;
3469 	sdkp->disk = gd;
3470 	sdkp->index = index;
3471 	sdkp->max_retries = SD_MAX_RETRIES;
3472 	atomic_set(&sdkp->openers, 0);
3473 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3474 	INIT_WORK(&sdkp->start_done_work, sd_start_done_work);
3475 
3476 	if (!sdp->request_queue->rq_timeout) {
3477 		if (sdp->type != TYPE_MOD)
3478 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3479 		else
3480 			blk_queue_rq_timeout(sdp->request_queue,
3481 					     SD_MOD_TIMEOUT);
3482 	}
3483 
3484 	device_initialize(&sdkp->disk_dev);
3485 	sdkp->disk_dev.parent = get_device(dev);
3486 	sdkp->disk_dev.class = &sd_disk_class;
3487 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3488 
3489 	error = device_add(&sdkp->disk_dev);
3490 	if (error) {
3491 		put_device(&sdkp->disk_dev);
3492 		goto out;
3493 	}
3494 
3495 	dev_set_drvdata(dev, sdkp);
3496 
3497 	gd->major = sd_major((index & 0xf0) >> 4);
3498 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3499 	gd->minors = SD_MINORS;
3500 
3501 	gd->fops = &sd_fops;
3502 	gd->private_data = sdkp;
3503 
3504 	/* defaults, until the device tells us otherwise */
3505 	sdp->sector_size = 512;
3506 	sdkp->capacity = 0;
3507 	sdkp->media_present = 1;
3508 	sdkp->write_prot = 0;
3509 	sdkp->cache_override = 0;
3510 	sdkp->WCE = 0;
3511 	sdkp->RCD = 0;
3512 	sdkp->ATO = 0;
3513 	sdkp->first_scan = 1;
3514 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3515 
3516 	sd_revalidate_disk(gd);
3517 
3518 	if (sdp->removable) {
3519 		gd->flags |= GENHD_FL_REMOVABLE;
3520 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3521 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3522 	}
3523 
3524 	blk_pm_runtime_init(sdp->request_queue, dev);
3525 	if (sdp->rpm_autosuspend) {
3526 		pm_runtime_set_autosuspend_delay(dev,
3527 			sdp->host->hostt->rpm_autosuspend_delay);
3528 	}
3529 
3530 	error = device_add_disk(dev, gd, NULL);
3531 	if (error) {
3532 		put_device(&sdkp->disk_dev);
3533 		put_disk(gd);
3534 		goto out;
3535 	}
3536 
3537 	if (sdkp->security) {
3538 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3539 		if (sdkp->opal_dev)
3540 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3541 	}
3542 
3543 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3544 		  sdp->removable ? "removable " : "");
3545 	scsi_autopm_put_device(sdp);
3546 
3547 	return 0;
3548 
3549  out_free_index:
3550 	ida_free(&sd_index_ida, index);
3551  out_put:
3552 	put_disk(gd);
3553  out_free:
3554 	kfree(sdkp);
3555  out:
3556 	scsi_autopm_put_device(sdp);
3557 	return error;
3558 }
3559 
3560 /**
3561  *	sd_remove - called whenever a scsi disk (previously recognized by
3562  *	sd_probe) is detached from the system. It is called (potentially
3563  *	multiple times) during sd module unload.
3564  *	@dev: pointer to device object
3565  *
3566  *	Note: this function is invoked from the scsi mid-level.
3567  *	This function potentially frees up a device name (e.g. /dev/sdc)
3568  *	that could be re-used by a subsequent sd_probe().
3569  *	This function is not called when the built-in sd driver is "exit-ed".
3570  **/
3571 static int sd_remove(struct device *dev)
3572 {
3573 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3574 
3575 	scsi_autopm_get_device(sdkp->device);
3576 
3577 	device_del(&sdkp->disk_dev);
3578 	del_gendisk(sdkp->disk);
3579 	sd_shutdown(dev);
3580 
3581 	put_disk(sdkp->disk);
3582 	return 0;
3583 }
3584 
3585 static void scsi_disk_release(struct device *dev)
3586 {
3587 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3588 
3589 	ida_free(&sd_index_ida, sdkp->index);
3590 	sd_zbc_free_zone_info(sdkp);
3591 	put_device(&sdkp->device->sdev_gendev);
3592 	free_opal_dev(sdkp->opal_dev);
3593 
3594 	kfree(sdkp);
3595 }
3596 
3597 /* Process sense data after a START command finished. */
3598 static void sd_start_done_work(struct work_struct *work)
3599 {
3600 	struct scsi_disk *sdkp = container_of(work, typeof(*sdkp),
3601 					      start_done_work);
3602 	struct scsi_sense_hdr sshdr;
3603 	int res = sdkp->start_result;
3604 
3605 	if (res == 0)
3606 		return;
3607 
3608 	sd_print_result(sdkp, "Start/Stop Unit failed", res);
3609 
3610 	if (res < 0)
3611 		return;
3612 
3613 	if (scsi_normalize_sense(sdkp->start_sense_buffer,
3614 				 sdkp->start_sense_len, &sshdr))
3615 		sd_print_sense_hdr(sdkp, &sshdr);
3616 }
3617 
3618 /* A START command finished. May be called from interrupt context. */
3619 static void sd_start_done(struct request *req, blk_status_t status)
3620 {
3621 	const struct scsi_cmnd *scmd = blk_mq_rq_to_pdu(req);
3622 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
3623 
3624 	sdkp->start_result = scmd->result;
3625 	WARN_ON_ONCE(scmd->sense_len > SCSI_SENSE_BUFFERSIZE);
3626 	sdkp->start_sense_len = scmd->sense_len;
3627 	memcpy(sdkp->start_sense_buffer, scmd->sense_buffer,
3628 	       ARRAY_SIZE(sdkp->start_sense_buffer));
3629 	WARN_ON_ONCE(!schedule_work(&sdkp->start_done_work));
3630 }
3631 
3632 /* Submit a START command asynchronously. */
3633 static int sd_submit_start(struct scsi_disk *sdkp, u8 cmd[], u8 cmd_len)
3634 {
3635 	struct scsi_device *sdev = sdkp->device;
3636 	struct request_queue *q = sdev->request_queue;
3637 	struct request *req;
3638 	struct scsi_cmnd *scmd;
3639 
3640 	req = scsi_alloc_request(q, REQ_OP_DRV_IN, BLK_MQ_REQ_PM);
3641 	if (IS_ERR(req))
3642 		return PTR_ERR(req);
3643 
3644 	scmd = blk_mq_rq_to_pdu(req);
3645 	scmd->cmd_len = cmd_len;
3646 	memcpy(scmd->cmnd, cmd, cmd_len);
3647 	scmd->allowed = sdkp->max_retries;
3648 	req->timeout = SD_TIMEOUT;
3649 	req->rq_flags |= RQF_PM | RQF_QUIET;
3650 	req->end_io = sd_start_done;
3651 	blk_execute_rq_nowait(req, /*at_head=*/true);
3652 
3653 	return 0;
3654 }
3655 
3656 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3657 {
3658 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3659 	struct scsi_device *sdp = sdkp->device;
3660 
3661 	if (start)
3662 		cmd[4] |= 1;	/* START */
3663 
3664 	if (sdp->start_stop_pwr_cond)
3665 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3666 
3667 	if (!scsi_device_online(sdp))
3668 		return -ENODEV;
3669 
3670 	/* Wait until processing of sense data has finished. */
3671 	flush_work(&sdkp->start_done_work);
3672 
3673 	return sd_submit_start(sdkp, cmd, sizeof(cmd));
3674 }
3675 
3676 /*
3677  * Send a SYNCHRONIZE CACHE instruction down to the device through
3678  * the normal SCSI command structure.  Wait for the command to
3679  * complete.
3680  */
3681 static void sd_shutdown(struct device *dev)
3682 {
3683 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3684 
3685 	if (!sdkp)
3686 		return;         /* this can happen */
3687 
3688 	if (pm_runtime_suspended(dev))
3689 		return;
3690 
3691 	if (sdkp->WCE && sdkp->media_present) {
3692 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3693 		sd_sync_cache(sdkp, NULL);
3694 	}
3695 
3696 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3697 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3698 		sd_start_stop_device(sdkp, 0);
3699 	}
3700 
3701 	flush_work(&sdkp->start_done_work);
3702 }
3703 
3704 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3705 {
3706 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3707 	struct scsi_sense_hdr sshdr;
3708 	int ret = 0;
3709 
3710 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3711 		return 0;
3712 
3713 	if (sdkp->WCE && sdkp->media_present) {
3714 		if (!sdkp->device->silence_suspend)
3715 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3716 		ret = sd_sync_cache(sdkp, &sshdr);
3717 
3718 		if (ret) {
3719 			/* ignore OFFLINE device */
3720 			if (ret == -ENODEV)
3721 				return 0;
3722 
3723 			if (!scsi_sense_valid(&sshdr) ||
3724 			    sshdr.sense_key != ILLEGAL_REQUEST)
3725 				return ret;
3726 
3727 			/*
3728 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3729 			 * doesn't support sync. There's not much to do and
3730 			 * suspend shouldn't fail.
3731 			 */
3732 			ret = 0;
3733 		}
3734 	}
3735 
3736 	if (sdkp->device->manage_start_stop) {
3737 		if (!sdkp->device->silence_suspend)
3738 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3739 		/* an error is not worth aborting a system sleep */
3740 		ret = sd_start_stop_device(sdkp, 0);
3741 		if (ignore_stop_errors)
3742 			ret = 0;
3743 	}
3744 
3745 	return ret;
3746 }
3747 
3748 static int sd_suspend_system(struct device *dev)
3749 {
3750 	if (pm_runtime_suspended(dev))
3751 		return 0;
3752 
3753 	return sd_suspend_common(dev, true);
3754 }
3755 
3756 static int sd_suspend_runtime(struct device *dev)
3757 {
3758 	return sd_suspend_common(dev, false);
3759 }
3760 
3761 static int sd_resume(struct device *dev)
3762 {
3763 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3764 	int ret;
3765 
3766 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3767 		return 0;
3768 
3769 	if (!sdkp->device->manage_start_stop)
3770 		return 0;
3771 
3772 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3773 	ret = sd_start_stop_device(sdkp, 1);
3774 	if (!ret)
3775 		opal_unlock_from_suspend(sdkp->opal_dev);
3776 	return ret;
3777 }
3778 
3779 static int sd_resume_system(struct device *dev)
3780 {
3781 	if (pm_runtime_suspended(dev))
3782 		return 0;
3783 
3784 	return sd_resume(dev);
3785 }
3786 
3787 static int sd_resume_runtime(struct device *dev)
3788 {
3789 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3790 	struct scsi_device *sdp;
3791 
3792 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3793 		return 0;
3794 
3795 	sdp = sdkp->device;
3796 
3797 	if (sdp->ignore_media_change) {
3798 		/* clear the device's sense data */
3799 		static const u8 cmd[10] = { REQUEST_SENSE };
3800 
3801 		if (scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL,
3802 				 NULL, sdp->request_queue->rq_timeout, 1, 0,
3803 				 RQF_PM, NULL))
3804 			sd_printk(KERN_NOTICE, sdkp,
3805 				  "Failed to clear sense data\n");
3806 	}
3807 
3808 	return sd_resume(dev);
3809 }
3810 
3811 /**
3812  *	init_sd - entry point for this driver (both when built in or when
3813  *	a module).
3814  *
3815  *	Note: this function registers this driver with the scsi mid-level.
3816  **/
3817 static int __init init_sd(void)
3818 {
3819 	int majors = 0, i, err;
3820 
3821 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3822 
3823 	for (i = 0; i < SD_MAJORS; i++) {
3824 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3825 			continue;
3826 		majors++;
3827 	}
3828 
3829 	if (!majors)
3830 		return -ENODEV;
3831 
3832 	err = class_register(&sd_disk_class);
3833 	if (err)
3834 		goto err_out;
3835 
3836 	sd_cdb_cache = kmem_cache_create("sd_ext_cdb", SD_EXT_CDB_SIZE,
3837 					 0, 0, NULL);
3838 	if (!sd_cdb_cache) {
3839 		printk(KERN_ERR "sd: can't init extended cdb cache\n");
3840 		err = -ENOMEM;
3841 		goto err_out_class;
3842 	}
3843 
3844 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3845 	if (!sd_page_pool) {
3846 		printk(KERN_ERR "sd: can't init discard page pool\n");
3847 		err = -ENOMEM;
3848 		goto err_out_cache;
3849 	}
3850 
3851 	err = scsi_register_driver(&sd_template.gendrv);
3852 	if (err)
3853 		goto err_out_driver;
3854 
3855 	return 0;
3856 
3857 err_out_driver:
3858 	mempool_destroy(sd_page_pool);
3859 
3860 err_out_cache:
3861 	kmem_cache_destroy(sd_cdb_cache);
3862 
3863 err_out_class:
3864 	class_unregister(&sd_disk_class);
3865 err_out:
3866 	for (i = 0; i < SD_MAJORS; i++)
3867 		unregister_blkdev(sd_major(i), "sd");
3868 	return err;
3869 }
3870 
3871 /**
3872  *	exit_sd - exit point for this driver (when it is a module).
3873  *
3874  *	Note: this function unregisters this driver from the scsi mid-level.
3875  **/
3876 static void __exit exit_sd(void)
3877 {
3878 	int i;
3879 
3880 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3881 
3882 	scsi_unregister_driver(&sd_template.gendrv);
3883 	mempool_destroy(sd_page_pool);
3884 	kmem_cache_destroy(sd_cdb_cache);
3885 
3886 	class_unregister(&sd_disk_class);
3887 
3888 	for (i = 0; i < SD_MAJORS; i++)
3889 		unregister_blkdev(sd_major(i), "sd");
3890 }
3891 
3892 module_init(init_sd);
3893 module_exit(exit_sd);
3894 
3895 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3896 {
3897 	scsi_print_sense_hdr(sdkp->device,
3898 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3899 }
3900 
3901 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3902 {
3903 	const char *hb_string = scsi_hostbyte_string(result);
3904 
3905 	if (hb_string)
3906 		sd_printk(KERN_INFO, sdkp,
3907 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3908 			  hb_string ? hb_string : "invalid",
3909 			  "DRIVER_OK");
3910 	else
3911 		sd_printk(KERN_INFO, sdkp,
3912 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3913 			  msg, host_byte(result), "DRIVER_OK");
3914 }
3915