xref: /openbmc/linux/drivers/scsi/sd.c (revision 0f4b20ef)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *      sd.c Copyright (C) 1992 Drew Eckhardt
4  *           Copyright (C) 1993, 1994, 1995, 1999 Eric Youngdale
5  *
6  *      Linux scsi disk driver
7  *              Initial versions: Drew Eckhardt
8  *              Subsequent revisions: Eric Youngdale
9  *	Modification history:
10  *       - Drew Eckhardt <drew@colorado.edu> original
11  *       - Eric Youngdale <eric@andante.org> add scatter-gather, multiple
12  *         outstanding request, and other enhancements.
13  *         Support loadable low-level scsi drivers.
14  *       - Jirka Hanika <geo@ff.cuni.cz> support more scsi disks using
15  *         eight major numbers.
16  *       - Richard Gooch <rgooch@atnf.csiro.au> support devfs.
17  *	 - Torben Mathiasen <tmm@image.dk> Resource allocation fixes in
18  *	   sd_init and cleanups.
19  *	 - Alex Davis <letmein@erols.com> Fix problem where partition info
20  *	   not being read in sd_open. Fix problem where removable media
21  *	   could be ejected after sd_open.
22  *	 - Douglas Gilbert <dgilbert@interlog.com> cleanup for lk 2.5.x
23  *	 - Badari Pulavarty <pbadari@us.ibm.com>, Matthew Wilcox
24  *	   <willy@debian.org>, Kurt Garloff <garloff@suse.de>:
25  *	   Support 32k/1M disks.
26  *
27  *	Logging policy (needs CONFIG_SCSI_LOGGING defined):
28  *	 - setting up transfer: SCSI_LOG_HLQUEUE levels 1 and 2
29  *	 - end of transfer (bh + scsi_lib): SCSI_LOG_HLCOMPLETE level 1
30  *	 - entering sd_ioctl: SCSI_LOG_IOCTL level 1
31  *	 - entering other commands: SCSI_LOG_HLQUEUE level 3
32  *	Note: when the logging level is set by the user, it must be greater
33  *	than the level indicated above to trigger output.
34  */
35 
36 #include <linux/module.h>
37 #include <linux/fs.h>
38 #include <linux/kernel.h>
39 #include <linux/mm.h>
40 #include <linux/bio.h>
41 #include <linux/hdreg.h>
42 #include <linux/errno.h>
43 #include <linux/idr.h>
44 #include <linux/interrupt.h>
45 #include <linux/init.h>
46 #include <linux/blkdev.h>
47 #include <linux/blkpg.h>
48 #include <linux/blk-pm.h>
49 #include <linux/delay.h>
50 #include <linux/major.h>
51 #include <linux/mutex.h>
52 #include <linux/string_helpers.h>
53 #include <linux/slab.h>
54 #include <linux/sed-opal.h>
55 #include <linux/pm_runtime.h>
56 #include <linux/pr.h>
57 #include <linux/t10-pi.h>
58 #include <linux/uaccess.h>
59 #include <asm/unaligned.h>
60 
61 #include <scsi/scsi.h>
62 #include <scsi/scsi_cmnd.h>
63 #include <scsi/scsi_dbg.h>
64 #include <scsi/scsi_device.h>
65 #include <scsi/scsi_driver.h>
66 #include <scsi/scsi_eh.h>
67 #include <scsi/scsi_host.h>
68 #include <scsi/scsi_ioctl.h>
69 #include <scsi/scsicam.h>
70 
71 #include "sd.h"
72 #include "scsi_priv.h"
73 #include "scsi_logging.h"
74 
75 MODULE_AUTHOR("Eric Youngdale");
76 MODULE_DESCRIPTION("SCSI disk (sd) driver");
77 MODULE_LICENSE("GPL");
78 
79 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK0_MAJOR);
80 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK1_MAJOR);
81 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK2_MAJOR);
82 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK3_MAJOR);
83 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK4_MAJOR);
84 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK5_MAJOR);
85 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK6_MAJOR);
86 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK7_MAJOR);
87 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK8_MAJOR);
88 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK9_MAJOR);
89 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK10_MAJOR);
90 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK11_MAJOR);
91 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK12_MAJOR);
92 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK13_MAJOR);
93 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK14_MAJOR);
94 MODULE_ALIAS_BLOCKDEV_MAJOR(SCSI_DISK15_MAJOR);
95 MODULE_ALIAS_SCSI_DEVICE(TYPE_DISK);
96 MODULE_ALIAS_SCSI_DEVICE(TYPE_MOD);
97 MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
98 MODULE_ALIAS_SCSI_DEVICE(TYPE_ZBC);
99 
100 #define SD_MINORS	16
101 
102 static void sd_config_discard(struct scsi_disk *, unsigned int);
103 static void sd_config_write_same(struct scsi_disk *);
104 static int  sd_revalidate_disk(struct gendisk *);
105 static void sd_unlock_native_capacity(struct gendisk *disk);
106 static int  sd_probe(struct device *);
107 static int  sd_remove(struct device *);
108 static void sd_shutdown(struct device *);
109 static int sd_suspend_system(struct device *);
110 static int sd_suspend_runtime(struct device *);
111 static int sd_resume_system(struct device *);
112 static int sd_resume_runtime(struct device *);
113 static void sd_rescan(struct device *);
114 static blk_status_t sd_init_command(struct scsi_cmnd *SCpnt);
115 static void sd_uninit_command(struct scsi_cmnd *SCpnt);
116 static int sd_done(struct scsi_cmnd *);
117 static void sd_eh_reset(struct scsi_cmnd *);
118 static int sd_eh_action(struct scsi_cmnd *, int);
119 static void sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer);
120 static void scsi_disk_release(struct device *cdev);
121 
122 static DEFINE_IDA(sd_index_ida);
123 
124 static struct kmem_cache *sd_cdb_cache;
125 static mempool_t *sd_page_pool;
126 static struct lock_class_key sd_bio_compl_lkclass;
127 
128 static const char *sd_cache_types[] = {
129 	"write through", "none", "write back",
130 	"write back, no read (daft)"
131 };
132 
133 static void sd_set_flush_flag(struct scsi_disk *sdkp)
134 {
135 	bool wc = false, fua = false;
136 
137 	if (sdkp->WCE) {
138 		wc = true;
139 		if (sdkp->DPOFUA)
140 			fua = true;
141 	}
142 
143 	blk_queue_write_cache(sdkp->disk->queue, wc, fua);
144 }
145 
146 static ssize_t
147 cache_type_store(struct device *dev, struct device_attribute *attr,
148 		 const char *buf, size_t count)
149 {
150 	int ct, rcd, wce, sp;
151 	struct scsi_disk *sdkp = to_scsi_disk(dev);
152 	struct scsi_device *sdp = sdkp->device;
153 	char buffer[64];
154 	char *buffer_data;
155 	struct scsi_mode_data data;
156 	struct scsi_sense_hdr sshdr;
157 	static const char temp[] = "temporary ";
158 	int len;
159 
160 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
161 		/* no cache control on RBC devices; theoretically they
162 		 * can do it, but there's probably so many exceptions
163 		 * it's not worth the risk */
164 		return -EINVAL;
165 
166 	if (strncmp(buf, temp, sizeof(temp) - 1) == 0) {
167 		buf += sizeof(temp) - 1;
168 		sdkp->cache_override = 1;
169 	} else {
170 		sdkp->cache_override = 0;
171 	}
172 
173 	ct = sysfs_match_string(sd_cache_types, buf);
174 	if (ct < 0)
175 		return -EINVAL;
176 
177 	rcd = ct & 0x01 ? 1 : 0;
178 	wce = (ct & 0x02) && !sdkp->write_prot ? 1 : 0;
179 
180 	if (sdkp->cache_override) {
181 		sdkp->WCE = wce;
182 		sdkp->RCD = rcd;
183 		sd_set_flush_flag(sdkp);
184 		return count;
185 	}
186 
187 	if (scsi_mode_sense(sdp, 0x08, 8, buffer, sizeof(buffer), SD_TIMEOUT,
188 			    sdkp->max_retries, &data, NULL))
189 		return -EINVAL;
190 	len = min_t(size_t, sizeof(buffer), data.length - data.header_length -
191 		  data.block_descriptor_length);
192 	buffer_data = buffer + data.header_length +
193 		data.block_descriptor_length;
194 	buffer_data[2] &= ~0x05;
195 	buffer_data[2] |= wce << 2 | rcd;
196 	sp = buffer_data[0] & 0x80 ? 1 : 0;
197 	buffer_data[0] &= ~0x80;
198 
199 	/*
200 	 * Ensure WP, DPOFUA, and RESERVED fields are cleared in
201 	 * received mode parameter buffer before doing MODE SELECT.
202 	 */
203 	data.device_specific = 0;
204 
205 	if (scsi_mode_select(sdp, 1, sp, buffer_data, len, SD_TIMEOUT,
206 			     sdkp->max_retries, &data, &sshdr)) {
207 		if (scsi_sense_valid(&sshdr))
208 			sd_print_sense_hdr(sdkp, &sshdr);
209 		return -EINVAL;
210 	}
211 	sd_revalidate_disk(sdkp->disk);
212 	return count;
213 }
214 
215 static ssize_t
216 manage_start_stop_show(struct device *dev, struct device_attribute *attr,
217 		       char *buf)
218 {
219 	struct scsi_disk *sdkp = to_scsi_disk(dev);
220 	struct scsi_device *sdp = sdkp->device;
221 
222 	return sprintf(buf, "%u\n", sdp->manage_start_stop);
223 }
224 
225 static ssize_t
226 manage_start_stop_store(struct device *dev, struct device_attribute *attr,
227 			const char *buf, size_t count)
228 {
229 	struct scsi_disk *sdkp = to_scsi_disk(dev);
230 	struct scsi_device *sdp = sdkp->device;
231 	bool v;
232 
233 	if (!capable(CAP_SYS_ADMIN))
234 		return -EACCES;
235 
236 	if (kstrtobool(buf, &v))
237 		return -EINVAL;
238 
239 	sdp->manage_start_stop = v;
240 
241 	return count;
242 }
243 static DEVICE_ATTR_RW(manage_start_stop);
244 
245 static ssize_t
246 allow_restart_show(struct device *dev, struct device_attribute *attr, char *buf)
247 {
248 	struct scsi_disk *sdkp = to_scsi_disk(dev);
249 
250 	return sprintf(buf, "%u\n", sdkp->device->allow_restart);
251 }
252 
253 static ssize_t
254 allow_restart_store(struct device *dev, struct device_attribute *attr,
255 		    const char *buf, size_t count)
256 {
257 	bool v;
258 	struct scsi_disk *sdkp = to_scsi_disk(dev);
259 	struct scsi_device *sdp = sdkp->device;
260 
261 	if (!capable(CAP_SYS_ADMIN))
262 		return -EACCES;
263 
264 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
265 		return -EINVAL;
266 
267 	if (kstrtobool(buf, &v))
268 		return -EINVAL;
269 
270 	sdp->allow_restart = v;
271 
272 	return count;
273 }
274 static DEVICE_ATTR_RW(allow_restart);
275 
276 static ssize_t
277 cache_type_show(struct device *dev, struct device_attribute *attr, char *buf)
278 {
279 	struct scsi_disk *sdkp = to_scsi_disk(dev);
280 	int ct = sdkp->RCD + 2*sdkp->WCE;
281 
282 	return sprintf(buf, "%s\n", sd_cache_types[ct]);
283 }
284 static DEVICE_ATTR_RW(cache_type);
285 
286 static ssize_t
287 FUA_show(struct device *dev, struct device_attribute *attr, char *buf)
288 {
289 	struct scsi_disk *sdkp = to_scsi_disk(dev);
290 
291 	return sprintf(buf, "%u\n", sdkp->DPOFUA);
292 }
293 static DEVICE_ATTR_RO(FUA);
294 
295 static ssize_t
296 protection_type_show(struct device *dev, struct device_attribute *attr,
297 		     char *buf)
298 {
299 	struct scsi_disk *sdkp = to_scsi_disk(dev);
300 
301 	return sprintf(buf, "%u\n", sdkp->protection_type);
302 }
303 
304 static ssize_t
305 protection_type_store(struct device *dev, struct device_attribute *attr,
306 		      const char *buf, size_t count)
307 {
308 	struct scsi_disk *sdkp = to_scsi_disk(dev);
309 	unsigned int val;
310 	int err;
311 
312 	if (!capable(CAP_SYS_ADMIN))
313 		return -EACCES;
314 
315 	err = kstrtouint(buf, 10, &val);
316 
317 	if (err)
318 		return err;
319 
320 	if (val <= T10_PI_TYPE3_PROTECTION)
321 		sdkp->protection_type = val;
322 
323 	return count;
324 }
325 static DEVICE_ATTR_RW(protection_type);
326 
327 static ssize_t
328 protection_mode_show(struct device *dev, struct device_attribute *attr,
329 		     char *buf)
330 {
331 	struct scsi_disk *sdkp = to_scsi_disk(dev);
332 	struct scsi_device *sdp = sdkp->device;
333 	unsigned int dif, dix;
334 
335 	dif = scsi_host_dif_capable(sdp->host, sdkp->protection_type);
336 	dix = scsi_host_dix_capable(sdp->host, sdkp->protection_type);
337 
338 	if (!dix && scsi_host_dix_capable(sdp->host, T10_PI_TYPE0_PROTECTION)) {
339 		dif = 0;
340 		dix = 1;
341 	}
342 
343 	if (!dif && !dix)
344 		return sprintf(buf, "none\n");
345 
346 	return sprintf(buf, "%s%u\n", dix ? "dix" : "dif", dif);
347 }
348 static DEVICE_ATTR_RO(protection_mode);
349 
350 static ssize_t
351 app_tag_own_show(struct device *dev, struct device_attribute *attr, char *buf)
352 {
353 	struct scsi_disk *sdkp = to_scsi_disk(dev);
354 
355 	return sprintf(buf, "%u\n", sdkp->ATO);
356 }
357 static DEVICE_ATTR_RO(app_tag_own);
358 
359 static ssize_t
360 thin_provisioning_show(struct device *dev, struct device_attribute *attr,
361 		       char *buf)
362 {
363 	struct scsi_disk *sdkp = to_scsi_disk(dev);
364 
365 	return sprintf(buf, "%u\n", sdkp->lbpme);
366 }
367 static DEVICE_ATTR_RO(thin_provisioning);
368 
369 /* sysfs_match_string() requires dense arrays */
370 static const char *lbp_mode[] = {
371 	[SD_LBP_FULL]		= "full",
372 	[SD_LBP_UNMAP]		= "unmap",
373 	[SD_LBP_WS16]		= "writesame_16",
374 	[SD_LBP_WS10]		= "writesame_10",
375 	[SD_LBP_ZERO]		= "writesame_zero",
376 	[SD_LBP_DISABLE]	= "disabled",
377 };
378 
379 static ssize_t
380 provisioning_mode_show(struct device *dev, struct device_attribute *attr,
381 		       char *buf)
382 {
383 	struct scsi_disk *sdkp = to_scsi_disk(dev);
384 
385 	return sprintf(buf, "%s\n", lbp_mode[sdkp->provisioning_mode]);
386 }
387 
388 static ssize_t
389 provisioning_mode_store(struct device *dev, struct device_attribute *attr,
390 			const char *buf, size_t count)
391 {
392 	struct scsi_disk *sdkp = to_scsi_disk(dev);
393 	struct scsi_device *sdp = sdkp->device;
394 	int mode;
395 
396 	if (!capable(CAP_SYS_ADMIN))
397 		return -EACCES;
398 
399 	if (sd_is_zoned(sdkp)) {
400 		sd_config_discard(sdkp, SD_LBP_DISABLE);
401 		return count;
402 	}
403 
404 	if (sdp->type != TYPE_DISK)
405 		return -EINVAL;
406 
407 	mode = sysfs_match_string(lbp_mode, buf);
408 	if (mode < 0)
409 		return -EINVAL;
410 
411 	sd_config_discard(sdkp, mode);
412 
413 	return count;
414 }
415 static DEVICE_ATTR_RW(provisioning_mode);
416 
417 /* sysfs_match_string() requires dense arrays */
418 static const char *zeroing_mode[] = {
419 	[SD_ZERO_WRITE]		= "write",
420 	[SD_ZERO_WS]		= "writesame",
421 	[SD_ZERO_WS16_UNMAP]	= "writesame_16_unmap",
422 	[SD_ZERO_WS10_UNMAP]	= "writesame_10_unmap",
423 };
424 
425 static ssize_t
426 zeroing_mode_show(struct device *dev, struct device_attribute *attr,
427 		  char *buf)
428 {
429 	struct scsi_disk *sdkp = to_scsi_disk(dev);
430 
431 	return sprintf(buf, "%s\n", zeroing_mode[sdkp->zeroing_mode]);
432 }
433 
434 static ssize_t
435 zeroing_mode_store(struct device *dev, struct device_attribute *attr,
436 		   const char *buf, size_t count)
437 {
438 	struct scsi_disk *sdkp = to_scsi_disk(dev);
439 	int mode;
440 
441 	if (!capable(CAP_SYS_ADMIN))
442 		return -EACCES;
443 
444 	mode = sysfs_match_string(zeroing_mode, buf);
445 	if (mode < 0)
446 		return -EINVAL;
447 
448 	sdkp->zeroing_mode = mode;
449 
450 	return count;
451 }
452 static DEVICE_ATTR_RW(zeroing_mode);
453 
454 static ssize_t
455 max_medium_access_timeouts_show(struct device *dev,
456 				struct device_attribute *attr, char *buf)
457 {
458 	struct scsi_disk *sdkp = to_scsi_disk(dev);
459 
460 	return sprintf(buf, "%u\n", sdkp->max_medium_access_timeouts);
461 }
462 
463 static ssize_t
464 max_medium_access_timeouts_store(struct device *dev,
465 				 struct device_attribute *attr, const char *buf,
466 				 size_t count)
467 {
468 	struct scsi_disk *sdkp = to_scsi_disk(dev);
469 	int err;
470 
471 	if (!capable(CAP_SYS_ADMIN))
472 		return -EACCES;
473 
474 	err = kstrtouint(buf, 10, &sdkp->max_medium_access_timeouts);
475 
476 	return err ? err : count;
477 }
478 static DEVICE_ATTR_RW(max_medium_access_timeouts);
479 
480 static ssize_t
481 max_write_same_blocks_show(struct device *dev, struct device_attribute *attr,
482 			   char *buf)
483 {
484 	struct scsi_disk *sdkp = to_scsi_disk(dev);
485 
486 	return sprintf(buf, "%u\n", sdkp->max_ws_blocks);
487 }
488 
489 static ssize_t
490 max_write_same_blocks_store(struct device *dev, struct device_attribute *attr,
491 			    const char *buf, size_t count)
492 {
493 	struct scsi_disk *sdkp = to_scsi_disk(dev);
494 	struct scsi_device *sdp = sdkp->device;
495 	unsigned long max;
496 	int err;
497 
498 	if (!capable(CAP_SYS_ADMIN))
499 		return -EACCES;
500 
501 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
502 		return -EINVAL;
503 
504 	err = kstrtoul(buf, 10, &max);
505 
506 	if (err)
507 		return err;
508 
509 	if (max == 0)
510 		sdp->no_write_same = 1;
511 	else if (max <= SD_MAX_WS16_BLOCKS) {
512 		sdp->no_write_same = 0;
513 		sdkp->max_ws_blocks = max;
514 	}
515 
516 	sd_config_write_same(sdkp);
517 
518 	return count;
519 }
520 static DEVICE_ATTR_RW(max_write_same_blocks);
521 
522 static ssize_t
523 zoned_cap_show(struct device *dev, struct device_attribute *attr, char *buf)
524 {
525 	struct scsi_disk *sdkp = to_scsi_disk(dev);
526 
527 	if (sdkp->device->type == TYPE_ZBC)
528 		return sprintf(buf, "host-managed\n");
529 	if (sdkp->zoned == 1)
530 		return sprintf(buf, "host-aware\n");
531 	if (sdkp->zoned == 2)
532 		return sprintf(buf, "drive-managed\n");
533 	return sprintf(buf, "none\n");
534 }
535 static DEVICE_ATTR_RO(zoned_cap);
536 
537 static ssize_t
538 max_retries_store(struct device *dev, struct device_attribute *attr,
539 		  const char *buf, size_t count)
540 {
541 	struct scsi_disk *sdkp = to_scsi_disk(dev);
542 	struct scsi_device *sdev = sdkp->device;
543 	int retries, err;
544 
545 	err = kstrtoint(buf, 10, &retries);
546 	if (err)
547 		return err;
548 
549 	if (retries == SCSI_CMD_RETRIES_NO_LIMIT || retries <= SD_MAX_RETRIES) {
550 		sdkp->max_retries = retries;
551 		return count;
552 	}
553 
554 	sdev_printk(KERN_ERR, sdev, "max_retries must be between -1 and %d\n",
555 		    SD_MAX_RETRIES);
556 	return -EINVAL;
557 }
558 
559 static ssize_t
560 max_retries_show(struct device *dev, struct device_attribute *attr,
561 		 char *buf)
562 {
563 	struct scsi_disk *sdkp = to_scsi_disk(dev);
564 
565 	return sprintf(buf, "%d\n", sdkp->max_retries);
566 }
567 
568 static DEVICE_ATTR_RW(max_retries);
569 
570 static struct attribute *sd_disk_attrs[] = {
571 	&dev_attr_cache_type.attr,
572 	&dev_attr_FUA.attr,
573 	&dev_attr_allow_restart.attr,
574 	&dev_attr_manage_start_stop.attr,
575 	&dev_attr_protection_type.attr,
576 	&dev_attr_protection_mode.attr,
577 	&dev_attr_app_tag_own.attr,
578 	&dev_attr_thin_provisioning.attr,
579 	&dev_attr_provisioning_mode.attr,
580 	&dev_attr_zeroing_mode.attr,
581 	&dev_attr_max_write_same_blocks.attr,
582 	&dev_attr_max_medium_access_timeouts.attr,
583 	&dev_attr_zoned_cap.attr,
584 	&dev_attr_max_retries.attr,
585 	NULL,
586 };
587 ATTRIBUTE_GROUPS(sd_disk);
588 
589 static struct class sd_disk_class = {
590 	.name		= "scsi_disk",
591 	.owner		= THIS_MODULE,
592 	.dev_release	= scsi_disk_release,
593 	.dev_groups	= sd_disk_groups,
594 };
595 
596 static const struct dev_pm_ops sd_pm_ops = {
597 	.suspend		= sd_suspend_system,
598 	.resume			= sd_resume_system,
599 	.poweroff		= sd_suspend_system,
600 	.restore		= sd_resume_system,
601 	.runtime_suspend	= sd_suspend_runtime,
602 	.runtime_resume		= sd_resume_runtime,
603 };
604 
605 static struct scsi_driver sd_template = {
606 	.gendrv = {
607 		.name		= "sd",
608 		.owner		= THIS_MODULE,
609 		.probe		= sd_probe,
610 		.probe_type	= PROBE_PREFER_ASYNCHRONOUS,
611 		.remove		= sd_remove,
612 		.shutdown	= sd_shutdown,
613 		.pm		= &sd_pm_ops,
614 	},
615 	.rescan			= sd_rescan,
616 	.init_command		= sd_init_command,
617 	.uninit_command		= sd_uninit_command,
618 	.done			= sd_done,
619 	.eh_action		= sd_eh_action,
620 	.eh_reset		= sd_eh_reset,
621 };
622 
623 /*
624  * Don't request a new module, as that could deadlock in multipath
625  * environment.
626  */
627 static void sd_default_probe(dev_t devt)
628 {
629 }
630 
631 /*
632  * Device no to disk mapping:
633  *
634  *       major         disc2     disc  p1
635  *   |............|.............|....|....| <- dev_t
636  *    31        20 19          8 7  4 3  0
637  *
638  * Inside a major, we have 16k disks, however mapped non-
639  * contiguously. The first 16 disks are for major0, the next
640  * ones with major1, ... Disk 256 is for major0 again, disk 272
641  * for major1, ...
642  * As we stay compatible with our numbering scheme, we can reuse
643  * the well-know SCSI majors 8, 65--71, 136--143.
644  */
645 static int sd_major(int major_idx)
646 {
647 	switch (major_idx) {
648 	case 0:
649 		return SCSI_DISK0_MAJOR;
650 	case 1 ... 7:
651 		return SCSI_DISK1_MAJOR + major_idx - 1;
652 	case 8 ... 15:
653 		return SCSI_DISK8_MAJOR + major_idx - 8;
654 	default:
655 		BUG();
656 		return 0;	/* shut up gcc */
657 	}
658 }
659 
660 #ifdef CONFIG_BLK_SED_OPAL
661 static int sd_sec_submit(void *data, u16 spsp, u8 secp, void *buffer,
662 		size_t len, bool send)
663 {
664 	struct scsi_disk *sdkp = data;
665 	struct scsi_device *sdev = sdkp->device;
666 	u8 cdb[12] = { 0, };
667 	int ret;
668 
669 	cdb[0] = send ? SECURITY_PROTOCOL_OUT : SECURITY_PROTOCOL_IN;
670 	cdb[1] = secp;
671 	put_unaligned_be16(spsp, &cdb[2]);
672 	put_unaligned_be32(len, &cdb[6]);
673 
674 	ret = scsi_execute(sdev, cdb, send ? DMA_TO_DEVICE : DMA_FROM_DEVICE,
675 		buffer, len, NULL, NULL, SD_TIMEOUT, sdkp->max_retries, 0,
676 		RQF_PM, NULL);
677 	return ret <= 0 ? ret : -EIO;
678 }
679 #endif /* CONFIG_BLK_SED_OPAL */
680 
681 /*
682  * Look up the DIX operation based on whether the command is read or
683  * write and whether dix and dif are enabled.
684  */
685 static unsigned int sd_prot_op(bool write, bool dix, bool dif)
686 {
687 	/* Lookup table: bit 2 (write), bit 1 (dix), bit 0 (dif) */
688 	static const unsigned int ops[] = {	/* wrt dix dif */
689 		SCSI_PROT_NORMAL,		/*  0	0   0  */
690 		SCSI_PROT_READ_STRIP,		/*  0	0   1  */
691 		SCSI_PROT_READ_INSERT,		/*  0	1   0  */
692 		SCSI_PROT_READ_PASS,		/*  0	1   1  */
693 		SCSI_PROT_NORMAL,		/*  1	0   0  */
694 		SCSI_PROT_WRITE_INSERT,		/*  1	0   1  */
695 		SCSI_PROT_WRITE_STRIP,		/*  1	1   0  */
696 		SCSI_PROT_WRITE_PASS,		/*  1	1   1  */
697 	};
698 
699 	return ops[write << 2 | dix << 1 | dif];
700 }
701 
702 /*
703  * Returns a mask of the protection flags that are valid for a given DIX
704  * operation.
705  */
706 static unsigned int sd_prot_flag_mask(unsigned int prot_op)
707 {
708 	static const unsigned int flag_mask[] = {
709 		[SCSI_PROT_NORMAL]		= 0,
710 
711 		[SCSI_PROT_READ_STRIP]		= SCSI_PROT_TRANSFER_PI |
712 						  SCSI_PROT_GUARD_CHECK |
713 						  SCSI_PROT_REF_CHECK |
714 						  SCSI_PROT_REF_INCREMENT,
715 
716 		[SCSI_PROT_READ_INSERT]		= SCSI_PROT_REF_INCREMENT |
717 						  SCSI_PROT_IP_CHECKSUM,
718 
719 		[SCSI_PROT_READ_PASS]		= SCSI_PROT_TRANSFER_PI |
720 						  SCSI_PROT_GUARD_CHECK |
721 						  SCSI_PROT_REF_CHECK |
722 						  SCSI_PROT_REF_INCREMENT |
723 						  SCSI_PROT_IP_CHECKSUM,
724 
725 		[SCSI_PROT_WRITE_INSERT]	= SCSI_PROT_TRANSFER_PI |
726 						  SCSI_PROT_REF_INCREMENT,
727 
728 		[SCSI_PROT_WRITE_STRIP]		= SCSI_PROT_GUARD_CHECK |
729 						  SCSI_PROT_REF_CHECK |
730 						  SCSI_PROT_REF_INCREMENT |
731 						  SCSI_PROT_IP_CHECKSUM,
732 
733 		[SCSI_PROT_WRITE_PASS]		= SCSI_PROT_TRANSFER_PI |
734 						  SCSI_PROT_GUARD_CHECK |
735 						  SCSI_PROT_REF_CHECK |
736 						  SCSI_PROT_REF_INCREMENT |
737 						  SCSI_PROT_IP_CHECKSUM,
738 	};
739 
740 	return flag_mask[prot_op];
741 }
742 
743 static unsigned char sd_setup_protect_cmnd(struct scsi_cmnd *scmd,
744 					   unsigned int dix, unsigned int dif)
745 {
746 	struct request *rq = scsi_cmd_to_rq(scmd);
747 	struct bio *bio = rq->bio;
748 	unsigned int prot_op = sd_prot_op(rq_data_dir(rq), dix, dif);
749 	unsigned int protect = 0;
750 
751 	if (dix) {				/* DIX Type 0, 1, 2, 3 */
752 		if (bio_integrity_flagged(bio, BIP_IP_CHECKSUM))
753 			scmd->prot_flags |= SCSI_PROT_IP_CHECKSUM;
754 
755 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
756 			scmd->prot_flags |= SCSI_PROT_GUARD_CHECK;
757 	}
758 
759 	if (dif != T10_PI_TYPE3_PROTECTION) {	/* DIX/DIF Type 0, 1, 2 */
760 		scmd->prot_flags |= SCSI_PROT_REF_INCREMENT;
761 
762 		if (bio_integrity_flagged(bio, BIP_CTRL_NOCHECK) == false)
763 			scmd->prot_flags |= SCSI_PROT_REF_CHECK;
764 	}
765 
766 	if (dif) {				/* DIX/DIF Type 1, 2, 3 */
767 		scmd->prot_flags |= SCSI_PROT_TRANSFER_PI;
768 
769 		if (bio_integrity_flagged(bio, BIP_DISK_NOCHECK))
770 			protect = 3 << 5;	/* Disable target PI checking */
771 		else
772 			protect = 1 << 5;	/* Enable target PI checking */
773 	}
774 
775 	scsi_set_prot_op(scmd, prot_op);
776 	scsi_set_prot_type(scmd, dif);
777 	scmd->prot_flags &= sd_prot_flag_mask(prot_op);
778 
779 	return protect;
780 }
781 
782 static void sd_config_discard(struct scsi_disk *sdkp, unsigned int mode)
783 {
784 	struct request_queue *q = sdkp->disk->queue;
785 	unsigned int logical_block_size = sdkp->device->sector_size;
786 	unsigned int max_blocks = 0;
787 
788 	q->limits.discard_alignment =
789 		sdkp->unmap_alignment * logical_block_size;
790 	q->limits.discard_granularity =
791 		max(sdkp->physical_block_size,
792 		    sdkp->unmap_granularity * logical_block_size);
793 	sdkp->provisioning_mode = mode;
794 
795 	switch (mode) {
796 
797 	case SD_LBP_FULL:
798 	case SD_LBP_DISABLE:
799 		blk_queue_max_discard_sectors(q, 0);
800 		blk_queue_flag_clear(QUEUE_FLAG_DISCARD, q);
801 		return;
802 
803 	case SD_LBP_UNMAP:
804 		max_blocks = min_not_zero(sdkp->max_unmap_blocks,
805 					  (u32)SD_MAX_WS16_BLOCKS);
806 		break;
807 
808 	case SD_LBP_WS16:
809 		if (sdkp->device->unmap_limit_for_ws)
810 			max_blocks = sdkp->max_unmap_blocks;
811 		else
812 			max_blocks = sdkp->max_ws_blocks;
813 
814 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS16_BLOCKS);
815 		break;
816 
817 	case SD_LBP_WS10:
818 		if (sdkp->device->unmap_limit_for_ws)
819 			max_blocks = sdkp->max_unmap_blocks;
820 		else
821 			max_blocks = sdkp->max_ws_blocks;
822 
823 		max_blocks = min_not_zero(max_blocks, (u32)SD_MAX_WS10_BLOCKS);
824 		break;
825 
826 	case SD_LBP_ZERO:
827 		max_blocks = min_not_zero(sdkp->max_ws_blocks,
828 					  (u32)SD_MAX_WS10_BLOCKS);
829 		break;
830 	}
831 
832 	blk_queue_max_discard_sectors(q, max_blocks * (logical_block_size >> 9));
833 	blk_queue_flag_set(QUEUE_FLAG_DISCARD, q);
834 }
835 
836 static blk_status_t sd_setup_unmap_cmnd(struct scsi_cmnd *cmd)
837 {
838 	struct scsi_device *sdp = cmd->device;
839 	struct request *rq = scsi_cmd_to_rq(cmd);
840 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
841 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
842 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
843 	unsigned int data_len = 24;
844 	char *buf;
845 
846 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
847 	if (!rq->special_vec.bv_page)
848 		return BLK_STS_RESOURCE;
849 	clear_highpage(rq->special_vec.bv_page);
850 	rq->special_vec.bv_offset = 0;
851 	rq->special_vec.bv_len = data_len;
852 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
853 
854 	cmd->cmd_len = 10;
855 	cmd->cmnd[0] = UNMAP;
856 	cmd->cmnd[8] = 24;
857 
858 	buf = bvec_virt(&rq->special_vec);
859 	put_unaligned_be16(6 + 16, &buf[0]);
860 	put_unaligned_be16(16, &buf[2]);
861 	put_unaligned_be64(lba, &buf[8]);
862 	put_unaligned_be32(nr_blocks, &buf[16]);
863 
864 	cmd->allowed = sdkp->max_retries;
865 	cmd->transfersize = data_len;
866 	rq->timeout = SD_TIMEOUT;
867 
868 	return scsi_alloc_sgtables(cmd);
869 }
870 
871 static blk_status_t sd_setup_write_same16_cmnd(struct scsi_cmnd *cmd,
872 		bool unmap)
873 {
874 	struct scsi_device *sdp = cmd->device;
875 	struct request *rq = scsi_cmd_to_rq(cmd);
876 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
877 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
878 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
879 	u32 data_len = sdp->sector_size;
880 
881 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
882 	if (!rq->special_vec.bv_page)
883 		return BLK_STS_RESOURCE;
884 	clear_highpage(rq->special_vec.bv_page);
885 	rq->special_vec.bv_offset = 0;
886 	rq->special_vec.bv_len = data_len;
887 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
888 
889 	cmd->cmd_len = 16;
890 	cmd->cmnd[0] = WRITE_SAME_16;
891 	if (unmap)
892 		cmd->cmnd[1] = 0x8; /* UNMAP */
893 	put_unaligned_be64(lba, &cmd->cmnd[2]);
894 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
895 
896 	cmd->allowed = sdkp->max_retries;
897 	cmd->transfersize = data_len;
898 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
899 
900 	return scsi_alloc_sgtables(cmd);
901 }
902 
903 static blk_status_t sd_setup_write_same10_cmnd(struct scsi_cmnd *cmd,
904 		bool unmap)
905 {
906 	struct scsi_device *sdp = cmd->device;
907 	struct request *rq = scsi_cmd_to_rq(cmd);
908 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
909 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
910 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
911 	u32 data_len = sdp->sector_size;
912 
913 	rq->special_vec.bv_page = mempool_alloc(sd_page_pool, GFP_ATOMIC);
914 	if (!rq->special_vec.bv_page)
915 		return BLK_STS_RESOURCE;
916 	clear_highpage(rq->special_vec.bv_page);
917 	rq->special_vec.bv_offset = 0;
918 	rq->special_vec.bv_len = data_len;
919 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
920 
921 	cmd->cmd_len = 10;
922 	cmd->cmnd[0] = WRITE_SAME;
923 	if (unmap)
924 		cmd->cmnd[1] = 0x8; /* UNMAP */
925 	put_unaligned_be32(lba, &cmd->cmnd[2]);
926 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
927 
928 	cmd->allowed = sdkp->max_retries;
929 	cmd->transfersize = data_len;
930 	rq->timeout = unmap ? SD_TIMEOUT : SD_WRITE_SAME_TIMEOUT;
931 
932 	return scsi_alloc_sgtables(cmd);
933 }
934 
935 static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd)
936 {
937 	struct request *rq = scsi_cmd_to_rq(cmd);
938 	struct scsi_device *sdp = cmd->device;
939 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
940 	u64 lba = sectors_to_logical(sdp, blk_rq_pos(rq));
941 	u32 nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
942 
943 	if (!(rq->cmd_flags & REQ_NOUNMAP)) {
944 		switch (sdkp->zeroing_mode) {
945 		case SD_ZERO_WS16_UNMAP:
946 			return sd_setup_write_same16_cmnd(cmd, true);
947 		case SD_ZERO_WS10_UNMAP:
948 			return sd_setup_write_same10_cmnd(cmd, true);
949 		}
950 	}
951 
952 	if (sdp->no_write_same) {
953 		rq->rq_flags |= RQF_QUIET;
954 		return BLK_STS_TARGET;
955 	}
956 
957 	if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff)
958 		return sd_setup_write_same16_cmnd(cmd, false);
959 
960 	return sd_setup_write_same10_cmnd(cmd, false);
961 }
962 
963 static void sd_config_write_same(struct scsi_disk *sdkp)
964 {
965 	struct request_queue *q = sdkp->disk->queue;
966 	unsigned int logical_block_size = sdkp->device->sector_size;
967 
968 	if (sdkp->device->no_write_same) {
969 		sdkp->max_ws_blocks = 0;
970 		goto out;
971 	}
972 
973 	/* Some devices can not handle block counts above 0xffff despite
974 	 * supporting WRITE SAME(16). Consequently we default to 64k
975 	 * blocks per I/O unless the device explicitly advertises a
976 	 * bigger limit.
977 	 */
978 	if (sdkp->max_ws_blocks > SD_MAX_WS10_BLOCKS)
979 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
980 						   (u32)SD_MAX_WS16_BLOCKS);
981 	else if (sdkp->ws16 || sdkp->ws10 || sdkp->device->no_report_opcodes)
982 		sdkp->max_ws_blocks = min_not_zero(sdkp->max_ws_blocks,
983 						   (u32)SD_MAX_WS10_BLOCKS);
984 	else {
985 		sdkp->device->no_write_same = 1;
986 		sdkp->max_ws_blocks = 0;
987 	}
988 
989 	if (sdkp->lbprz && sdkp->lbpws)
990 		sdkp->zeroing_mode = SD_ZERO_WS16_UNMAP;
991 	else if (sdkp->lbprz && sdkp->lbpws10)
992 		sdkp->zeroing_mode = SD_ZERO_WS10_UNMAP;
993 	else if (sdkp->max_ws_blocks)
994 		sdkp->zeroing_mode = SD_ZERO_WS;
995 	else
996 		sdkp->zeroing_mode = SD_ZERO_WRITE;
997 
998 	if (sdkp->max_ws_blocks &&
999 	    sdkp->physical_block_size > logical_block_size) {
1000 		/*
1001 		 * Reporting a maximum number of blocks that is not aligned
1002 		 * on the device physical size would cause a large write same
1003 		 * request to be split into physically unaligned chunks by
1004 		 * __blkdev_issue_write_zeroes() even if the caller of this
1005 		 * functions took care to align the large request. So make sure
1006 		 * the maximum reported is aligned to the device physical block
1007 		 * size. This is only an optional optimization for regular
1008 		 * disks, but this is mandatory to avoid failure of large write
1009 		 * same requests directed at sequential write required zones of
1010 		 * host-managed ZBC disks.
1011 		 */
1012 		sdkp->max_ws_blocks =
1013 			round_down(sdkp->max_ws_blocks,
1014 				   bytes_to_logical(sdkp->device,
1015 						    sdkp->physical_block_size));
1016 	}
1017 
1018 out:
1019 	blk_queue_max_write_zeroes_sectors(q, sdkp->max_ws_blocks *
1020 					 (logical_block_size >> 9));
1021 }
1022 
1023 static blk_status_t sd_setup_flush_cmnd(struct scsi_cmnd *cmd)
1024 {
1025 	struct request *rq = scsi_cmd_to_rq(cmd);
1026 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1027 
1028 	/* flush requests don't perform I/O, zero the S/G table */
1029 	memset(&cmd->sdb, 0, sizeof(cmd->sdb));
1030 
1031 	cmd->cmnd[0] = SYNCHRONIZE_CACHE;
1032 	cmd->cmd_len = 10;
1033 	cmd->transfersize = 0;
1034 	cmd->allowed = sdkp->max_retries;
1035 
1036 	rq->timeout = rq->q->rq_timeout * SD_FLUSH_TIMEOUT_MULTIPLIER;
1037 	return BLK_STS_OK;
1038 }
1039 
1040 static blk_status_t sd_setup_rw32_cmnd(struct scsi_cmnd *cmd, bool write,
1041 				       sector_t lba, unsigned int nr_blocks,
1042 				       unsigned char flags)
1043 {
1044 	cmd->cmd_len = SD_EXT_CDB_SIZE;
1045 	cmd->cmnd[0]  = VARIABLE_LENGTH_CMD;
1046 	cmd->cmnd[7]  = 0x18; /* Additional CDB len */
1047 	cmd->cmnd[9]  = write ? WRITE_32 : READ_32;
1048 	cmd->cmnd[10] = flags;
1049 	put_unaligned_be64(lba, &cmd->cmnd[12]);
1050 	put_unaligned_be32(lba, &cmd->cmnd[20]); /* Expected Indirect LBA */
1051 	put_unaligned_be32(nr_blocks, &cmd->cmnd[28]);
1052 
1053 	return BLK_STS_OK;
1054 }
1055 
1056 static blk_status_t sd_setup_rw16_cmnd(struct scsi_cmnd *cmd, bool write,
1057 				       sector_t lba, unsigned int nr_blocks,
1058 				       unsigned char flags)
1059 {
1060 	cmd->cmd_len  = 16;
1061 	cmd->cmnd[0]  = write ? WRITE_16 : READ_16;
1062 	cmd->cmnd[1]  = flags;
1063 	cmd->cmnd[14] = 0;
1064 	cmd->cmnd[15] = 0;
1065 	put_unaligned_be64(lba, &cmd->cmnd[2]);
1066 	put_unaligned_be32(nr_blocks, &cmd->cmnd[10]);
1067 
1068 	return BLK_STS_OK;
1069 }
1070 
1071 static blk_status_t sd_setup_rw10_cmnd(struct scsi_cmnd *cmd, bool write,
1072 				       sector_t lba, unsigned int nr_blocks,
1073 				       unsigned char flags)
1074 {
1075 	cmd->cmd_len = 10;
1076 	cmd->cmnd[0] = write ? WRITE_10 : READ_10;
1077 	cmd->cmnd[1] = flags;
1078 	cmd->cmnd[6] = 0;
1079 	cmd->cmnd[9] = 0;
1080 	put_unaligned_be32(lba, &cmd->cmnd[2]);
1081 	put_unaligned_be16(nr_blocks, &cmd->cmnd[7]);
1082 
1083 	return BLK_STS_OK;
1084 }
1085 
1086 static blk_status_t sd_setup_rw6_cmnd(struct scsi_cmnd *cmd, bool write,
1087 				      sector_t lba, unsigned int nr_blocks,
1088 				      unsigned char flags)
1089 {
1090 	/* Avoid that 0 blocks gets translated into 256 blocks. */
1091 	if (WARN_ON_ONCE(nr_blocks == 0))
1092 		return BLK_STS_IOERR;
1093 
1094 	if (unlikely(flags & 0x8)) {
1095 		/*
1096 		 * This happens only if this drive failed 10byte rw
1097 		 * command with ILLEGAL_REQUEST during operation and
1098 		 * thus turned off use_10_for_rw.
1099 		 */
1100 		scmd_printk(KERN_ERR, cmd, "FUA write on READ/WRITE(6) drive\n");
1101 		return BLK_STS_IOERR;
1102 	}
1103 
1104 	cmd->cmd_len = 6;
1105 	cmd->cmnd[0] = write ? WRITE_6 : READ_6;
1106 	cmd->cmnd[1] = (lba >> 16) & 0x1f;
1107 	cmd->cmnd[2] = (lba >> 8) & 0xff;
1108 	cmd->cmnd[3] = lba & 0xff;
1109 	cmd->cmnd[4] = nr_blocks;
1110 	cmd->cmnd[5] = 0;
1111 
1112 	return BLK_STS_OK;
1113 }
1114 
1115 static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
1116 {
1117 	struct request *rq = scsi_cmd_to_rq(cmd);
1118 	struct scsi_device *sdp = cmd->device;
1119 	struct scsi_disk *sdkp = scsi_disk(rq->q->disk);
1120 	sector_t lba = sectors_to_logical(sdp, blk_rq_pos(rq));
1121 	sector_t threshold;
1122 	unsigned int nr_blocks = sectors_to_logical(sdp, blk_rq_sectors(rq));
1123 	unsigned int mask = logical_to_sectors(sdp, 1) - 1;
1124 	bool write = rq_data_dir(rq) == WRITE;
1125 	unsigned char protect, fua;
1126 	blk_status_t ret;
1127 	unsigned int dif;
1128 	bool dix;
1129 
1130 	ret = scsi_alloc_sgtables(cmd);
1131 	if (ret != BLK_STS_OK)
1132 		return ret;
1133 
1134 	ret = BLK_STS_IOERR;
1135 	if (!scsi_device_online(sdp) || sdp->changed) {
1136 		scmd_printk(KERN_ERR, cmd, "device offline or changed\n");
1137 		goto fail;
1138 	}
1139 
1140 	if (blk_rq_pos(rq) + blk_rq_sectors(rq) > get_capacity(rq->q->disk)) {
1141 		scmd_printk(KERN_ERR, cmd, "access beyond end of device\n");
1142 		goto fail;
1143 	}
1144 
1145 	if ((blk_rq_pos(rq) & mask) || (blk_rq_sectors(rq) & mask)) {
1146 		scmd_printk(KERN_ERR, cmd, "request not aligned to the logical block size\n");
1147 		goto fail;
1148 	}
1149 
1150 	/*
1151 	 * Some SD card readers can't handle accesses which touch the
1152 	 * last one or two logical blocks. Split accesses as needed.
1153 	 */
1154 	threshold = sdkp->capacity - SD_LAST_BUGGY_SECTORS;
1155 
1156 	if (unlikely(sdp->last_sector_bug && lba + nr_blocks > threshold)) {
1157 		if (lba < threshold) {
1158 			/* Access up to the threshold but not beyond */
1159 			nr_blocks = threshold - lba;
1160 		} else {
1161 			/* Access only a single logical block */
1162 			nr_blocks = 1;
1163 		}
1164 	}
1165 
1166 	if (req_op(rq) == REQ_OP_ZONE_APPEND) {
1167 		ret = sd_zbc_prepare_zone_append(cmd, &lba, nr_blocks);
1168 		if (ret)
1169 			goto fail;
1170 	}
1171 
1172 	fua = rq->cmd_flags & REQ_FUA ? 0x8 : 0;
1173 	dix = scsi_prot_sg_count(cmd);
1174 	dif = scsi_host_dif_capable(cmd->device->host, sdkp->protection_type);
1175 
1176 	if (dif || dix)
1177 		protect = sd_setup_protect_cmnd(cmd, dix, dif);
1178 	else
1179 		protect = 0;
1180 
1181 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
1182 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
1183 					 protect | fua);
1184 	} else if (sdp->use_16_for_rw || (nr_blocks > 0xffff)) {
1185 		ret = sd_setup_rw16_cmnd(cmd, write, lba, nr_blocks,
1186 					 protect | fua);
1187 	} else if ((nr_blocks > 0xff) || (lba > 0x1fffff) ||
1188 		   sdp->use_10_for_rw || protect) {
1189 		ret = sd_setup_rw10_cmnd(cmd, write, lba, nr_blocks,
1190 					 protect | fua);
1191 	} else {
1192 		ret = sd_setup_rw6_cmnd(cmd, write, lba, nr_blocks,
1193 					protect | fua);
1194 	}
1195 
1196 	if (unlikely(ret != BLK_STS_OK))
1197 		goto fail;
1198 
1199 	/*
1200 	 * We shouldn't disconnect in the middle of a sector, so with a dumb
1201 	 * host adapter, it's safe to assume that we can at least transfer
1202 	 * this many bytes between each connect / disconnect.
1203 	 */
1204 	cmd->transfersize = sdp->sector_size;
1205 	cmd->underflow = nr_blocks << 9;
1206 	cmd->allowed = sdkp->max_retries;
1207 	cmd->sdb.length = nr_blocks * sdp->sector_size;
1208 
1209 	SCSI_LOG_HLQUEUE(1,
1210 			 scmd_printk(KERN_INFO, cmd,
1211 				     "%s: block=%llu, count=%d\n", __func__,
1212 				     (unsigned long long)blk_rq_pos(rq),
1213 				     blk_rq_sectors(rq)));
1214 	SCSI_LOG_HLQUEUE(2,
1215 			 scmd_printk(KERN_INFO, cmd,
1216 				     "%s %d/%u 512 byte blocks.\n",
1217 				     write ? "writing" : "reading", nr_blocks,
1218 				     blk_rq_sectors(rq)));
1219 
1220 	/*
1221 	 * This indicates that the command is ready from our end to be queued.
1222 	 */
1223 	return BLK_STS_OK;
1224 fail:
1225 	scsi_free_sgtables(cmd);
1226 	return ret;
1227 }
1228 
1229 static blk_status_t sd_init_command(struct scsi_cmnd *cmd)
1230 {
1231 	struct request *rq = scsi_cmd_to_rq(cmd);
1232 
1233 	switch (req_op(rq)) {
1234 	case REQ_OP_DISCARD:
1235 		switch (scsi_disk(rq->q->disk)->provisioning_mode) {
1236 		case SD_LBP_UNMAP:
1237 			return sd_setup_unmap_cmnd(cmd);
1238 		case SD_LBP_WS16:
1239 			return sd_setup_write_same16_cmnd(cmd, true);
1240 		case SD_LBP_WS10:
1241 			return sd_setup_write_same10_cmnd(cmd, true);
1242 		case SD_LBP_ZERO:
1243 			return sd_setup_write_same10_cmnd(cmd, false);
1244 		default:
1245 			return BLK_STS_TARGET;
1246 		}
1247 	case REQ_OP_WRITE_ZEROES:
1248 		return sd_setup_write_zeroes_cmnd(cmd);
1249 	case REQ_OP_FLUSH:
1250 		return sd_setup_flush_cmnd(cmd);
1251 	case REQ_OP_READ:
1252 	case REQ_OP_WRITE:
1253 	case REQ_OP_ZONE_APPEND:
1254 		return sd_setup_read_write_cmnd(cmd);
1255 	case REQ_OP_ZONE_RESET:
1256 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1257 						   false);
1258 	case REQ_OP_ZONE_RESET_ALL:
1259 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_RESET_WRITE_POINTER,
1260 						   true);
1261 	case REQ_OP_ZONE_OPEN:
1262 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_OPEN_ZONE, false);
1263 	case REQ_OP_ZONE_CLOSE:
1264 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_CLOSE_ZONE, false);
1265 	case REQ_OP_ZONE_FINISH:
1266 		return sd_zbc_setup_zone_mgmt_cmnd(cmd, ZO_FINISH_ZONE, false);
1267 	default:
1268 		WARN_ON_ONCE(1);
1269 		return BLK_STS_NOTSUPP;
1270 	}
1271 }
1272 
1273 static void sd_uninit_command(struct scsi_cmnd *SCpnt)
1274 {
1275 	struct request *rq = scsi_cmd_to_rq(SCpnt);
1276 
1277 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
1278 		mempool_free(rq->special_vec.bv_page, sd_page_pool);
1279 }
1280 
1281 static bool sd_need_revalidate(struct block_device *bdev,
1282 		struct scsi_disk *sdkp)
1283 {
1284 	if (sdkp->device->removable || sdkp->write_prot) {
1285 		if (bdev_check_media_change(bdev))
1286 			return true;
1287 	}
1288 
1289 	/*
1290 	 * Force a full rescan after ioctl(BLKRRPART).  While the disk state has
1291 	 * nothing to do with partitions, BLKRRPART is used to force a full
1292 	 * revalidate after things like a format for historical reasons.
1293 	 */
1294 	return test_bit(GD_NEED_PART_SCAN, &bdev->bd_disk->state);
1295 }
1296 
1297 /**
1298  *	sd_open - open a scsi disk device
1299  *	@bdev: Block device of the scsi disk to open
1300  *	@mode: FMODE_* mask
1301  *
1302  *	Returns 0 if successful. Returns a negated errno value in case
1303  *	of error.
1304  *
1305  *	Note: This can be called from a user context (e.g. fsck(1) )
1306  *	or from within the kernel (e.g. as a result of a mount(1) ).
1307  *	In the latter case @inode and @filp carry an abridged amount
1308  *	of information as noted above.
1309  *
1310  *	Locking: called with bdev->bd_disk->open_mutex held.
1311  **/
1312 static int sd_open(struct block_device *bdev, fmode_t mode)
1313 {
1314 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1315 	struct scsi_device *sdev = sdkp->device;
1316 	int retval;
1317 
1318 	if (scsi_device_get(sdev))
1319 		return -ENXIO;
1320 
1321 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_open\n"));
1322 
1323 	/*
1324 	 * If the device is in error recovery, wait until it is done.
1325 	 * If the device is offline, then disallow any access to it.
1326 	 */
1327 	retval = -ENXIO;
1328 	if (!scsi_block_when_processing_errors(sdev))
1329 		goto error_out;
1330 
1331 	if (sd_need_revalidate(bdev, sdkp))
1332 		sd_revalidate_disk(bdev->bd_disk);
1333 
1334 	/*
1335 	 * If the drive is empty, just let the open fail.
1336 	 */
1337 	retval = -ENOMEDIUM;
1338 	if (sdev->removable && !sdkp->media_present && !(mode & FMODE_NDELAY))
1339 		goto error_out;
1340 
1341 	/*
1342 	 * If the device has the write protect tab set, have the open fail
1343 	 * if the user expects to be able to write to the thing.
1344 	 */
1345 	retval = -EROFS;
1346 	if (sdkp->write_prot && (mode & FMODE_WRITE))
1347 		goto error_out;
1348 
1349 	/*
1350 	 * It is possible that the disk changing stuff resulted in
1351 	 * the device being taken offline.  If this is the case,
1352 	 * report this to the user, and don't pretend that the
1353 	 * open actually succeeded.
1354 	 */
1355 	retval = -ENXIO;
1356 	if (!scsi_device_online(sdev))
1357 		goto error_out;
1358 
1359 	if ((atomic_inc_return(&sdkp->openers) == 1) && sdev->removable) {
1360 		if (scsi_block_when_processing_errors(sdev))
1361 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
1362 	}
1363 
1364 	return 0;
1365 
1366 error_out:
1367 	scsi_device_put(sdev);
1368 	return retval;
1369 }
1370 
1371 /**
1372  *	sd_release - invoked when the (last) close(2) is called on this
1373  *	scsi disk.
1374  *	@disk: disk to release
1375  *	@mode: FMODE_* mask
1376  *
1377  *	Returns 0.
1378  *
1379  *	Note: may block (uninterruptible) if error recovery is underway
1380  *	on this disk.
1381  *
1382  *	Locking: called with bdev->bd_disk->open_mutex held.
1383  **/
1384 static void sd_release(struct gendisk *disk, fmode_t mode)
1385 {
1386 	struct scsi_disk *sdkp = scsi_disk(disk);
1387 	struct scsi_device *sdev = sdkp->device;
1388 
1389 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
1390 
1391 	if (atomic_dec_return(&sdkp->openers) == 0 && sdev->removable) {
1392 		if (scsi_block_when_processing_errors(sdev))
1393 			scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
1394 	}
1395 
1396 	scsi_device_put(sdev);
1397 }
1398 
1399 static int sd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
1400 {
1401 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1402 	struct scsi_device *sdp = sdkp->device;
1403 	struct Scsi_Host *host = sdp->host;
1404 	sector_t capacity = logical_to_sectors(sdp, sdkp->capacity);
1405 	int diskinfo[4];
1406 
1407 	/* default to most commonly used values */
1408 	diskinfo[0] = 0x40;	/* 1 << 6 */
1409 	diskinfo[1] = 0x20;	/* 1 << 5 */
1410 	diskinfo[2] = capacity >> 11;
1411 
1412 	/* override with calculated, extended default, or driver values */
1413 	if (host->hostt->bios_param)
1414 		host->hostt->bios_param(sdp, bdev, capacity, diskinfo);
1415 	else
1416 		scsicam_bios_param(bdev, capacity, diskinfo);
1417 
1418 	geo->heads = diskinfo[0];
1419 	geo->sectors = diskinfo[1];
1420 	geo->cylinders = diskinfo[2];
1421 	return 0;
1422 }
1423 
1424 /**
1425  *	sd_ioctl - process an ioctl
1426  *	@bdev: target block device
1427  *	@mode: FMODE_* mask
1428  *	@cmd: ioctl command number
1429  *	@arg: this is third argument given to ioctl(2) system call.
1430  *	Often contains a pointer.
1431  *
1432  *	Returns 0 if successful (some ioctls return positive numbers on
1433  *	success as well). Returns a negated errno value in case of error.
1434  *
1435  *	Note: most ioctls are forward onto the block subsystem or further
1436  *	down in the scsi subsystem.
1437  **/
1438 static int sd_ioctl(struct block_device *bdev, fmode_t mode,
1439 		    unsigned int cmd, unsigned long arg)
1440 {
1441 	struct gendisk *disk = bdev->bd_disk;
1442 	struct scsi_disk *sdkp = scsi_disk(disk);
1443 	struct scsi_device *sdp = sdkp->device;
1444 	void __user *p = (void __user *)arg;
1445 	int error;
1446 
1447 	SCSI_LOG_IOCTL(1, sd_printk(KERN_INFO, sdkp, "sd_ioctl: disk=%s, "
1448 				    "cmd=0x%x\n", disk->disk_name, cmd));
1449 
1450 	if (bdev_is_partition(bdev) && !capable(CAP_SYS_RAWIO))
1451 		return -ENOIOCTLCMD;
1452 
1453 	/*
1454 	 * If we are in the middle of error recovery, don't let anyone
1455 	 * else try and use this device.  Also, if error recovery fails, it
1456 	 * may try and take the device offline, in which case all further
1457 	 * access to the device is prohibited.
1458 	 */
1459 	error = scsi_ioctl_block_when_processing_errors(sdp, cmd,
1460 			(mode & FMODE_NDELAY) != 0);
1461 	if (error)
1462 		return error;
1463 
1464 	if (is_sed_ioctl(cmd))
1465 		return sed_ioctl(sdkp->opal_dev, cmd, p);
1466 	return scsi_ioctl(sdp, mode, cmd, p);
1467 }
1468 
1469 static void set_media_not_present(struct scsi_disk *sdkp)
1470 {
1471 	if (sdkp->media_present)
1472 		sdkp->device->changed = 1;
1473 
1474 	if (sdkp->device->removable) {
1475 		sdkp->media_present = 0;
1476 		sdkp->capacity = 0;
1477 	}
1478 }
1479 
1480 static int media_not_present(struct scsi_disk *sdkp,
1481 			     struct scsi_sense_hdr *sshdr)
1482 {
1483 	if (!scsi_sense_valid(sshdr))
1484 		return 0;
1485 
1486 	/* not invoked for commands that could return deferred errors */
1487 	switch (sshdr->sense_key) {
1488 	case UNIT_ATTENTION:
1489 	case NOT_READY:
1490 		/* medium not present */
1491 		if (sshdr->asc == 0x3A) {
1492 			set_media_not_present(sdkp);
1493 			return 1;
1494 		}
1495 	}
1496 	return 0;
1497 }
1498 
1499 /**
1500  *	sd_check_events - check media events
1501  *	@disk: kernel device descriptor
1502  *	@clearing: disk events currently being cleared
1503  *
1504  *	Returns mask of DISK_EVENT_*.
1505  *
1506  *	Note: this function is invoked from the block subsystem.
1507  **/
1508 static unsigned int sd_check_events(struct gendisk *disk, unsigned int clearing)
1509 {
1510 	struct scsi_disk *sdkp = disk->private_data;
1511 	struct scsi_device *sdp;
1512 	int retval;
1513 	bool disk_changed;
1514 
1515 	if (!sdkp)
1516 		return 0;
1517 
1518 	sdp = sdkp->device;
1519 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_check_events\n"));
1520 
1521 	/*
1522 	 * If the device is offline, don't send any commands - just pretend as
1523 	 * if the command failed.  If the device ever comes back online, we
1524 	 * can deal with it then.  It is only because of unrecoverable errors
1525 	 * that we would ever take a device offline in the first place.
1526 	 */
1527 	if (!scsi_device_online(sdp)) {
1528 		set_media_not_present(sdkp);
1529 		goto out;
1530 	}
1531 
1532 	/*
1533 	 * Using TEST_UNIT_READY enables differentiation between drive with
1534 	 * no cartridge loaded - NOT READY, drive with changed cartridge -
1535 	 * UNIT ATTENTION, or with same cartridge - GOOD STATUS.
1536 	 *
1537 	 * Drives that auto spin down. eg iomega jaz 1G, will be started
1538 	 * by sd_spinup_disk() from sd_revalidate_disk(), which happens whenever
1539 	 * sd_revalidate() is called.
1540 	 */
1541 	if (scsi_block_when_processing_errors(sdp)) {
1542 		struct scsi_sense_hdr sshdr = { 0, };
1543 
1544 		retval = scsi_test_unit_ready(sdp, SD_TIMEOUT, sdkp->max_retries,
1545 					      &sshdr);
1546 
1547 		/* failed to execute TUR, assume media not present */
1548 		if (retval < 0 || host_byte(retval)) {
1549 			set_media_not_present(sdkp);
1550 			goto out;
1551 		}
1552 
1553 		if (media_not_present(sdkp, &sshdr))
1554 			goto out;
1555 	}
1556 
1557 	/*
1558 	 * For removable scsi disk we have to recognise the presence
1559 	 * of a disk in the drive.
1560 	 */
1561 	if (!sdkp->media_present)
1562 		sdp->changed = 1;
1563 	sdkp->media_present = 1;
1564 out:
1565 	/*
1566 	 * sdp->changed is set under the following conditions:
1567 	 *
1568 	 *	Medium present state has changed in either direction.
1569 	 *	Device has indicated UNIT_ATTENTION.
1570 	 */
1571 	disk_changed = sdp->changed;
1572 	sdp->changed = 0;
1573 	return disk_changed ? DISK_EVENT_MEDIA_CHANGE : 0;
1574 }
1575 
1576 static int sd_sync_cache(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
1577 {
1578 	int retries, res;
1579 	struct scsi_device *sdp = sdkp->device;
1580 	const int timeout = sdp->request_queue->rq_timeout
1581 		* SD_FLUSH_TIMEOUT_MULTIPLIER;
1582 	struct scsi_sense_hdr my_sshdr;
1583 
1584 	if (!scsi_device_online(sdp))
1585 		return -ENODEV;
1586 
1587 	/* caller might not be interested in sense, but we need it */
1588 	if (!sshdr)
1589 		sshdr = &my_sshdr;
1590 
1591 	for (retries = 3; retries > 0; --retries) {
1592 		unsigned char cmd[10] = { 0 };
1593 
1594 		cmd[0] = SYNCHRONIZE_CACHE;
1595 		/*
1596 		 * Leave the rest of the command zero to indicate
1597 		 * flush everything.
1598 		 */
1599 		res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, sshdr,
1600 				timeout, sdkp->max_retries, 0, RQF_PM, NULL);
1601 		if (res == 0)
1602 			break;
1603 	}
1604 
1605 	if (res) {
1606 		sd_print_result(sdkp, "Synchronize Cache(10) failed", res);
1607 
1608 		if (res < 0)
1609 			return res;
1610 
1611 		if (scsi_status_is_check_condition(res) &&
1612 		    scsi_sense_valid(sshdr)) {
1613 			sd_print_sense_hdr(sdkp, sshdr);
1614 
1615 			/* we need to evaluate the error return  */
1616 			if (sshdr->asc == 0x3a ||	/* medium not present */
1617 			    sshdr->asc == 0x20 ||	/* invalid command */
1618 			    (sshdr->asc == 0x74 && sshdr->ascq == 0x71))	/* drive is password locked */
1619 				/* this is no error here */
1620 				return 0;
1621 		}
1622 
1623 		switch (host_byte(res)) {
1624 		/* ignore errors due to racing a disconnection */
1625 		case DID_BAD_TARGET:
1626 		case DID_NO_CONNECT:
1627 			return 0;
1628 		/* signal the upper layer it might try again */
1629 		case DID_BUS_BUSY:
1630 		case DID_IMM_RETRY:
1631 		case DID_REQUEUE:
1632 		case DID_SOFT_ERROR:
1633 			return -EBUSY;
1634 		default:
1635 			return -EIO;
1636 		}
1637 	}
1638 	return 0;
1639 }
1640 
1641 static void sd_rescan(struct device *dev)
1642 {
1643 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
1644 
1645 	sd_revalidate_disk(sdkp->disk);
1646 }
1647 
1648 static int sd_get_unique_id(struct gendisk *disk, u8 id[16],
1649 		enum blk_unique_id type)
1650 {
1651 	struct scsi_device *sdev = scsi_disk(disk)->device;
1652 	const struct scsi_vpd *vpd;
1653 	const unsigned char *d;
1654 	int ret = -ENXIO, len;
1655 
1656 	rcu_read_lock();
1657 	vpd = rcu_dereference(sdev->vpd_pg83);
1658 	if (!vpd)
1659 		goto out_unlock;
1660 
1661 	ret = -EINVAL;
1662 	for (d = vpd->data + 4; d < vpd->data + vpd->len; d += d[3] + 4) {
1663 		/* we only care about designators with LU association */
1664 		if (((d[1] >> 4) & 0x3) != 0x00)
1665 			continue;
1666 		if ((d[1] & 0xf) != type)
1667 			continue;
1668 
1669 		/*
1670 		 * Only exit early if a 16-byte descriptor was found.  Otherwise
1671 		 * keep looking as one with more entropy might still show up.
1672 		 */
1673 		len = d[3];
1674 		if (len != 8 && len != 12 && len != 16)
1675 			continue;
1676 		ret = len;
1677 		memcpy(id, d + 4, len);
1678 		if (len == 16)
1679 			break;
1680 	}
1681 out_unlock:
1682 	rcu_read_unlock();
1683 	return ret;
1684 }
1685 
1686 static char sd_pr_type(enum pr_type type)
1687 {
1688 	switch (type) {
1689 	case PR_WRITE_EXCLUSIVE:
1690 		return 0x01;
1691 	case PR_EXCLUSIVE_ACCESS:
1692 		return 0x03;
1693 	case PR_WRITE_EXCLUSIVE_REG_ONLY:
1694 		return 0x05;
1695 	case PR_EXCLUSIVE_ACCESS_REG_ONLY:
1696 		return 0x06;
1697 	case PR_WRITE_EXCLUSIVE_ALL_REGS:
1698 		return 0x07;
1699 	case PR_EXCLUSIVE_ACCESS_ALL_REGS:
1700 		return 0x08;
1701 	default:
1702 		return 0;
1703 	}
1704 };
1705 
1706 static int sd_pr_command(struct block_device *bdev, u8 sa,
1707 		u64 key, u64 sa_key, u8 type, u8 flags)
1708 {
1709 	struct scsi_disk *sdkp = scsi_disk(bdev->bd_disk);
1710 	struct scsi_device *sdev = sdkp->device;
1711 	struct scsi_sense_hdr sshdr;
1712 	int result;
1713 	u8 cmd[16] = { 0, };
1714 	u8 data[24] = { 0, };
1715 
1716 	cmd[0] = PERSISTENT_RESERVE_OUT;
1717 	cmd[1] = sa;
1718 	cmd[2] = type;
1719 	put_unaligned_be32(sizeof(data), &cmd[5]);
1720 
1721 	put_unaligned_be64(key, &data[0]);
1722 	put_unaligned_be64(sa_key, &data[8]);
1723 	data[20] = flags;
1724 
1725 	result = scsi_execute_req(sdev, cmd, DMA_TO_DEVICE, &data, sizeof(data),
1726 			&sshdr, SD_TIMEOUT, sdkp->max_retries, NULL);
1727 
1728 	if (scsi_status_is_check_condition(result) &&
1729 	    scsi_sense_valid(&sshdr)) {
1730 		sdev_printk(KERN_INFO, sdev, "PR command failed: %d\n", result);
1731 		scsi_print_sense_hdr(sdev, NULL, &sshdr);
1732 	}
1733 
1734 	return result;
1735 }
1736 
1737 static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
1738 		u32 flags)
1739 {
1740 	if (flags & ~PR_FL_IGNORE_KEY)
1741 		return -EOPNOTSUPP;
1742 	return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00,
1743 			old_key, new_key, 0,
1744 			(1 << 0) /* APTPL */);
1745 }
1746 
1747 static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
1748 		u32 flags)
1749 {
1750 	if (flags)
1751 		return -EOPNOTSUPP;
1752 	return sd_pr_command(bdev, 0x01, key, 0, sd_pr_type(type), 0);
1753 }
1754 
1755 static int sd_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
1756 {
1757 	return sd_pr_command(bdev, 0x02, key, 0, sd_pr_type(type), 0);
1758 }
1759 
1760 static int sd_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
1761 		enum pr_type type, bool abort)
1762 {
1763 	return sd_pr_command(bdev, abort ? 0x05 : 0x04, old_key, new_key,
1764 			     sd_pr_type(type), 0);
1765 }
1766 
1767 static int sd_pr_clear(struct block_device *bdev, u64 key)
1768 {
1769 	return sd_pr_command(bdev, 0x03, key, 0, 0, 0);
1770 }
1771 
1772 static const struct pr_ops sd_pr_ops = {
1773 	.pr_register	= sd_pr_register,
1774 	.pr_reserve	= sd_pr_reserve,
1775 	.pr_release	= sd_pr_release,
1776 	.pr_preempt	= sd_pr_preempt,
1777 	.pr_clear	= sd_pr_clear,
1778 };
1779 
1780 static void scsi_disk_free_disk(struct gendisk *disk)
1781 {
1782 	struct scsi_disk *sdkp = scsi_disk(disk);
1783 
1784 	put_device(&sdkp->disk_dev);
1785 }
1786 
1787 static const struct block_device_operations sd_fops = {
1788 	.owner			= THIS_MODULE,
1789 	.open			= sd_open,
1790 	.release		= sd_release,
1791 	.ioctl			= sd_ioctl,
1792 	.getgeo			= sd_getgeo,
1793 	.compat_ioctl		= blkdev_compat_ptr_ioctl,
1794 	.check_events		= sd_check_events,
1795 	.unlock_native_capacity	= sd_unlock_native_capacity,
1796 	.report_zones		= sd_zbc_report_zones,
1797 	.get_unique_id		= sd_get_unique_id,
1798 	.free_disk		= scsi_disk_free_disk,
1799 	.pr_ops			= &sd_pr_ops,
1800 };
1801 
1802 /**
1803  *	sd_eh_reset - reset error handling callback
1804  *	@scmd:		sd-issued command that has failed
1805  *
1806  *	This function is called by the SCSI midlayer before starting
1807  *	SCSI EH. When counting medium access failures we have to be
1808  *	careful to register it only only once per device and SCSI EH run;
1809  *	there might be several timed out commands which will cause the
1810  *	'max_medium_access_timeouts' counter to trigger after the first
1811  *	SCSI EH run already and set the device to offline.
1812  *	So this function resets the internal counter before starting SCSI EH.
1813  **/
1814 static void sd_eh_reset(struct scsi_cmnd *scmd)
1815 {
1816 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1817 
1818 	/* New SCSI EH run, reset gate variable */
1819 	sdkp->ignore_medium_access_errors = false;
1820 }
1821 
1822 /**
1823  *	sd_eh_action - error handling callback
1824  *	@scmd:		sd-issued command that has failed
1825  *	@eh_disp:	The recovery disposition suggested by the midlayer
1826  *
1827  *	This function is called by the SCSI midlayer upon completion of an
1828  *	error test command (currently TEST UNIT READY). The result of sending
1829  *	the eh command is passed in eh_disp.  We're looking for devices that
1830  *	fail medium access commands but are OK with non access commands like
1831  *	test unit ready (so wrongly see the device as having a successful
1832  *	recovery)
1833  **/
1834 static int sd_eh_action(struct scsi_cmnd *scmd, int eh_disp)
1835 {
1836 	struct scsi_disk *sdkp = scsi_disk(scsi_cmd_to_rq(scmd)->q->disk);
1837 	struct scsi_device *sdev = scmd->device;
1838 
1839 	if (!scsi_device_online(sdev) ||
1840 	    !scsi_medium_access_command(scmd) ||
1841 	    host_byte(scmd->result) != DID_TIME_OUT ||
1842 	    eh_disp != SUCCESS)
1843 		return eh_disp;
1844 
1845 	/*
1846 	 * The device has timed out executing a medium access command.
1847 	 * However, the TEST UNIT READY command sent during error
1848 	 * handling completed successfully. Either the device is in the
1849 	 * process of recovering or has it suffered an internal failure
1850 	 * that prevents access to the storage medium.
1851 	 */
1852 	if (!sdkp->ignore_medium_access_errors) {
1853 		sdkp->medium_access_timed_out++;
1854 		sdkp->ignore_medium_access_errors = true;
1855 	}
1856 
1857 	/*
1858 	 * If the device keeps failing read/write commands but TEST UNIT
1859 	 * READY always completes successfully we assume that medium
1860 	 * access is no longer possible and take the device offline.
1861 	 */
1862 	if (sdkp->medium_access_timed_out >= sdkp->max_medium_access_timeouts) {
1863 		scmd_printk(KERN_ERR, scmd,
1864 			    "Medium access timeout failure. Offlining disk!\n");
1865 		mutex_lock(&sdev->state_mutex);
1866 		scsi_device_set_state(sdev, SDEV_OFFLINE);
1867 		mutex_unlock(&sdev->state_mutex);
1868 
1869 		return SUCCESS;
1870 	}
1871 
1872 	return eh_disp;
1873 }
1874 
1875 static unsigned int sd_completed_bytes(struct scsi_cmnd *scmd)
1876 {
1877 	struct request *req = scsi_cmd_to_rq(scmd);
1878 	struct scsi_device *sdev = scmd->device;
1879 	unsigned int transferred, good_bytes;
1880 	u64 start_lba, end_lba, bad_lba;
1881 
1882 	/*
1883 	 * Some commands have a payload smaller than the device logical
1884 	 * block size (e.g. INQUIRY on a 4K disk).
1885 	 */
1886 	if (scsi_bufflen(scmd) <= sdev->sector_size)
1887 		return 0;
1888 
1889 	/* Check if we have a 'bad_lba' information */
1890 	if (!scsi_get_sense_info_fld(scmd->sense_buffer,
1891 				     SCSI_SENSE_BUFFERSIZE,
1892 				     &bad_lba))
1893 		return 0;
1894 
1895 	/*
1896 	 * If the bad lba was reported incorrectly, we have no idea where
1897 	 * the error is.
1898 	 */
1899 	start_lba = sectors_to_logical(sdev, blk_rq_pos(req));
1900 	end_lba = start_lba + bytes_to_logical(sdev, scsi_bufflen(scmd));
1901 	if (bad_lba < start_lba || bad_lba >= end_lba)
1902 		return 0;
1903 
1904 	/*
1905 	 * resid is optional but mostly filled in.  When it's unused,
1906 	 * its value is zero, so we assume the whole buffer transferred
1907 	 */
1908 	transferred = scsi_bufflen(scmd) - scsi_get_resid(scmd);
1909 
1910 	/* This computation should always be done in terms of the
1911 	 * resolution of the device's medium.
1912 	 */
1913 	good_bytes = logical_to_bytes(sdev, bad_lba - start_lba);
1914 
1915 	return min(good_bytes, transferred);
1916 }
1917 
1918 /**
1919  *	sd_done - bottom half handler: called when the lower level
1920  *	driver has completed (successfully or otherwise) a scsi command.
1921  *	@SCpnt: mid-level's per command structure.
1922  *
1923  *	Note: potentially run from within an ISR. Must not block.
1924  **/
1925 static int sd_done(struct scsi_cmnd *SCpnt)
1926 {
1927 	int result = SCpnt->result;
1928 	unsigned int good_bytes = result ? 0 : scsi_bufflen(SCpnt);
1929 	unsigned int sector_size = SCpnt->device->sector_size;
1930 	unsigned int resid;
1931 	struct scsi_sense_hdr sshdr;
1932 	struct request *req = scsi_cmd_to_rq(SCpnt);
1933 	struct scsi_disk *sdkp = scsi_disk(req->q->disk);
1934 	int sense_valid = 0;
1935 	int sense_deferred = 0;
1936 
1937 	switch (req_op(req)) {
1938 	case REQ_OP_DISCARD:
1939 	case REQ_OP_WRITE_ZEROES:
1940 	case REQ_OP_ZONE_RESET:
1941 	case REQ_OP_ZONE_RESET_ALL:
1942 	case REQ_OP_ZONE_OPEN:
1943 	case REQ_OP_ZONE_CLOSE:
1944 	case REQ_OP_ZONE_FINISH:
1945 		if (!result) {
1946 			good_bytes = blk_rq_bytes(req);
1947 			scsi_set_resid(SCpnt, 0);
1948 		} else {
1949 			good_bytes = 0;
1950 			scsi_set_resid(SCpnt, blk_rq_bytes(req));
1951 		}
1952 		break;
1953 	default:
1954 		/*
1955 		 * In case of bogus fw or device, we could end up having
1956 		 * an unaligned partial completion. Check this here and force
1957 		 * alignment.
1958 		 */
1959 		resid = scsi_get_resid(SCpnt);
1960 		if (resid & (sector_size - 1)) {
1961 			sd_printk(KERN_INFO, sdkp,
1962 				"Unaligned partial completion (resid=%u, sector_sz=%u)\n",
1963 				resid, sector_size);
1964 			scsi_print_command(SCpnt);
1965 			resid = min(scsi_bufflen(SCpnt),
1966 				    round_up(resid, sector_size));
1967 			scsi_set_resid(SCpnt, resid);
1968 		}
1969 	}
1970 
1971 	if (result) {
1972 		sense_valid = scsi_command_normalize_sense(SCpnt, &sshdr);
1973 		if (sense_valid)
1974 			sense_deferred = scsi_sense_is_deferred(&sshdr);
1975 	}
1976 	sdkp->medium_access_timed_out = 0;
1977 
1978 	if (!scsi_status_is_check_condition(result) &&
1979 	    (!sense_valid || sense_deferred))
1980 		goto out;
1981 
1982 	switch (sshdr.sense_key) {
1983 	case HARDWARE_ERROR:
1984 	case MEDIUM_ERROR:
1985 		good_bytes = sd_completed_bytes(SCpnt);
1986 		break;
1987 	case RECOVERED_ERROR:
1988 		good_bytes = scsi_bufflen(SCpnt);
1989 		break;
1990 	case NO_SENSE:
1991 		/* This indicates a false check condition, so ignore it.  An
1992 		 * unknown amount of data was transferred so treat it as an
1993 		 * error.
1994 		 */
1995 		SCpnt->result = 0;
1996 		memset(SCpnt->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
1997 		break;
1998 	case ABORTED_COMMAND:
1999 		if (sshdr.asc == 0x10)  /* DIF: Target detected corruption */
2000 			good_bytes = sd_completed_bytes(SCpnt);
2001 		break;
2002 	case ILLEGAL_REQUEST:
2003 		switch (sshdr.asc) {
2004 		case 0x10:	/* DIX: Host detected corruption */
2005 			good_bytes = sd_completed_bytes(SCpnt);
2006 			break;
2007 		case 0x20:	/* INVALID COMMAND OPCODE */
2008 		case 0x24:	/* INVALID FIELD IN CDB */
2009 			switch (SCpnt->cmnd[0]) {
2010 			case UNMAP:
2011 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2012 				break;
2013 			case WRITE_SAME_16:
2014 			case WRITE_SAME:
2015 				if (SCpnt->cmnd[1] & 8) { /* UNMAP */
2016 					sd_config_discard(sdkp, SD_LBP_DISABLE);
2017 				} else {
2018 					sdkp->device->no_write_same = 1;
2019 					sd_config_write_same(sdkp);
2020 					req->rq_flags |= RQF_QUIET;
2021 				}
2022 				break;
2023 			}
2024 		}
2025 		break;
2026 	default:
2027 		break;
2028 	}
2029 
2030  out:
2031 	if (sd_is_zoned(sdkp))
2032 		good_bytes = sd_zbc_complete(SCpnt, good_bytes, &sshdr);
2033 
2034 	SCSI_LOG_HLCOMPLETE(1, scmd_printk(KERN_INFO, SCpnt,
2035 					   "sd_done: completed %d of %d bytes\n",
2036 					   good_bytes, scsi_bufflen(SCpnt)));
2037 
2038 	return good_bytes;
2039 }
2040 
2041 /*
2042  * spinup disk - called only in sd_revalidate_disk()
2043  */
2044 static void
2045 sd_spinup_disk(struct scsi_disk *sdkp)
2046 {
2047 	unsigned char cmd[10];
2048 	unsigned long spintime_expire = 0;
2049 	int retries, spintime;
2050 	unsigned int the_result;
2051 	struct scsi_sense_hdr sshdr;
2052 	int sense_valid = 0;
2053 
2054 	spintime = 0;
2055 
2056 	/* Spin up drives, as required.  Only do this at boot time */
2057 	/* Spinup needs to be done for module loads too. */
2058 	do {
2059 		retries = 0;
2060 
2061 		do {
2062 			bool media_was_present = sdkp->media_present;
2063 
2064 			cmd[0] = TEST_UNIT_READY;
2065 			memset((void *) &cmd[1], 0, 9);
2066 
2067 			the_result = scsi_execute_req(sdkp->device, cmd,
2068 						      DMA_NONE, NULL, 0,
2069 						      &sshdr, SD_TIMEOUT,
2070 						      sdkp->max_retries, NULL);
2071 
2072 			/*
2073 			 * If the drive has indicated to us that it
2074 			 * doesn't have any media in it, don't bother
2075 			 * with any more polling.
2076 			 */
2077 			if (media_not_present(sdkp, &sshdr)) {
2078 				if (media_was_present)
2079 					sd_printk(KERN_NOTICE, sdkp, "Media removed, stopped polling\n");
2080 				return;
2081 			}
2082 
2083 			if (the_result)
2084 				sense_valid = scsi_sense_valid(&sshdr);
2085 			retries++;
2086 		} while (retries < 3 &&
2087 			 (!scsi_status_is_good(the_result) ||
2088 			  (scsi_status_is_check_condition(the_result) &&
2089 			  sense_valid && sshdr.sense_key == UNIT_ATTENTION)));
2090 
2091 		if (!scsi_status_is_check_condition(the_result)) {
2092 			/* no sense, TUR either succeeded or failed
2093 			 * with a status error */
2094 			if(!spintime && !scsi_status_is_good(the_result)) {
2095 				sd_print_result(sdkp, "Test Unit Ready failed",
2096 						the_result);
2097 			}
2098 			break;
2099 		}
2100 
2101 		/*
2102 		 * The device does not want the automatic start to be issued.
2103 		 */
2104 		if (sdkp->device->no_start_on_add)
2105 			break;
2106 
2107 		if (sense_valid && sshdr.sense_key == NOT_READY) {
2108 			if (sshdr.asc == 4 && sshdr.ascq == 3)
2109 				break;	/* manual intervention required */
2110 			if (sshdr.asc == 4 && sshdr.ascq == 0xb)
2111 				break;	/* standby */
2112 			if (sshdr.asc == 4 && sshdr.ascq == 0xc)
2113 				break;	/* unavailable */
2114 			if (sshdr.asc == 4 && sshdr.ascq == 0x1b)
2115 				break;	/* sanitize in progress */
2116 			/*
2117 			 * Issue command to spin up drive when not ready
2118 			 */
2119 			if (!spintime) {
2120 				sd_printk(KERN_NOTICE, sdkp, "Spinning up disk...");
2121 				cmd[0] = START_STOP;
2122 				cmd[1] = 1;	/* Return immediately */
2123 				memset((void *) &cmd[2], 0, 8);
2124 				cmd[4] = 1;	/* Start spin cycle */
2125 				if (sdkp->device->start_stop_pwr_cond)
2126 					cmd[4] |= 1 << 4;
2127 				scsi_execute_req(sdkp->device, cmd, DMA_NONE,
2128 						 NULL, 0, &sshdr,
2129 						 SD_TIMEOUT, sdkp->max_retries,
2130 						 NULL);
2131 				spintime_expire = jiffies + 100 * HZ;
2132 				spintime = 1;
2133 			}
2134 			/* Wait 1 second for next try */
2135 			msleep(1000);
2136 			printk(KERN_CONT ".");
2137 
2138 		/*
2139 		 * Wait for USB flash devices with slow firmware.
2140 		 * Yes, this sense key/ASC combination shouldn't
2141 		 * occur here.  It's characteristic of these devices.
2142 		 */
2143 		} else if (sense_valid &&
2144 				sshdr.sense_key == UNIT_ATTENTION &&
2145 				sshdr.asc == 0x28) {
2146 			if (!spintime) {
2147 				spintime_expire = jiffies + 5 * HZ;
2148 				spintime = 1;
2149 			}
2150 			/* Wait 1 second for next try */
2151 			msleep(1000);
2152 		} else {
2153 			/* we don't understand the sense code, so it's
2154 			 * probably pointless to loop */
2155 			if(!spintime) {
2156 				sd_printk(KERN_NOTICE, sdkp, "Unit Not Ready\n");
2157 				sd_print_sense_hdr(sdkp, &sshdr);
2158 			}
2159 			break;
2160 		}
2161 
2162 	} while (spintime && time_before_eq(jiffies, spintime_expire));
2163 
2164 	if (spintime) {
2165 		if (scsi_status_is_good(the_result))
2166 			printk(KERN_CONT "ready\n");
2167 		else
2168 			printk(KERN_CONT "not responding...\n");
2169 	}
2170 }
2171 
2172 /*
2173  * Determine whether disk supports Data Integrity Field.
2174  */
2175 static int sd_read_protection_type(struct scsi_disk *sdkp, unsigned char *buffer)
2176 {
2177 	struct scsi_device *sdp = sdkp->device;
2178 	u8 type;
2179 	int ret = 0;
2180 
2181 	if (scsi_device_protection(sdp) == 0 || (buffer[12] & 1) == 0) {
2182 		sdkp->protection_type = 0;
2183 		return ret;
2184 	}
2185 
2186 	type = ((buffer[12] >> 1) & 7) + 1; /* P_TYPE 0 = Type 1 */
2187 
2188 	if (type > T10_PI_TYPE3_PROTECTION)
2189 		ret = -ENODEV;
2190 	else if (scsi_host_dif_capable(sdp->host, type))
2191 		ret = 1;
2192 
2193 	if (sdkp->first_scan || type != sdkp->protection_type)
2194 		switch (ret) {
2195 		case -ENODEV:
2196 			sd_printk(KERN_ERR, sdkp, "formatted with unsupported" \
2197 				  " protection type %u. Disabling disk!\n",
2198 				  type);
2199 			break;
2200 		case 1:
2201 			sd_printk(KERN_NOTICE, sdkp,
2202 				  "Enabling DIF Type %u protection\n", type);
2203 			break;
2204 		case 0:
2205 			sd_printk(KERN_NOTICE, sdkp,
2206 				  "Disabling DIF Type %u protection\n", type);
2207 			break;
2208 		}
2209 
2210 	sdkp->protection_type = type;
2211 
2212 	return ret;
2213 }
2214 
2215 static void read_capacity_error(struct scsi_disk *sdkp, struct scsi_device *sdp,
2216 			struct scsi_sense_hdr *sshdr, int sense_valid,
2217 			int the_result)
2218 {
2219 	if (sense_valid)
2220 		sd_print_sense_hdr(sdkp, sshdr);
2221 	else
2222 		sd_printk(KERN_NOTICE, sdkp, "Sense not available.\n");
2223 
2224 	/*
2225 	 * Set dirty bit for removable devices if not ready -
2226 	 * sometimes drives will not report this properly.
2227 	 */
2228 	if (sdp->removable &&
2229 	    sense_valid && sshdr->sense_key == NOT_READY)
2230 		set_media_not_present(sdkp);
2231 
2232 	/*
2233 	 * We used to set media_present to 0 here to indicate no media
2234 	 * in the drive, but some drives fail read capacity even with
2235 	 * media present, so we can't do that.
2236 	 */
2237 	sdkp->capacity = 0; /* unknown mapped to zero - as usual */
2238 }
2239 
2240 #define RC16_LEN 32
2241 #if RC16_LEN > SD_BUF_SIZE
2242 #error RC16_LEN must not be more than SD_BUF_SIZE
2243 #endif
2244 
2245 #define READ_CAPACITY_RETRIES_ON_RESET	10
2246 
2247 static int read_capacity_16(struct scsi_disk *sdkp, struct scsi_device *sdp,
2248 						unsigned char *buffer)
2249 {
2250 	unsigned char cmd[16];
2251 	struct scsi_sense_hdr sshdr;
2252 	int sense_valid = 0;
2253 	int the_result;
2254 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2255 	unsigned int alignment;
2256 	unsigned long long lba;
2257 	unsigned sector_size;
2258 
2259 	if (sdp->no_read_capacity_16)
2260 		return -EINVAL;
2261 
2262 	do {
2263 		memset(cmd, 0, 16);
2264 		cmd[0] = SERVICE_ACTION_IN_16;
2265 		cmd[1] = SAI_READ_CAPACITY_16;
2266 		cmd[13] = RC16_LEN;
2267 		memset(buffer, 0, RC16_LEN);
2268 
2269 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2270 					buffer, RC16_LEN, &sshdr,
2271 					SD_TIMEOUT, sdkp->max_retries, NULL);
2272 
2273 		if (media_not_present(sdkp, &sshdr))
2274 			return -ENODEV;
2275 
2276 		if (the_result > 0) {
2277 			sense_valid = scsi_sense_valid(&sshdr);
2278 			if (sense_valid &&
2279 			    sshdr.sense_key == ILLEGAL_REQUEST &&
2280 			    (sshdr.asc == 0x20 || sshdr.asc == 0x24) &&
2281 			    sshdr.ascq == 0x00)
2282 				/* Invalid Command Operation Code or
2283 				 * Invalid Field in CDB, just retry
2284 				 * silently with RC10 */
2285 				return -EINVAL;
2286 			if (sense_valid &&
2287 			    sshdr.sense_key == UNIT_ATTENTION &&
2288 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2289 				/* Device reset might occur several times,
2290 				 * give it one more chance */
2291 				if (--reset_retries > 0)
2292 					continue;
2293 		}
2294 		retries--;
2295 
2296 	} while (the_result && retries);
2297 
2298 	if (the_result) {
2299 		sd_print_result(sdkp, "Read Capacity(16) failed", the_result);
2300 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2301 		return -EINVAL;
2302 	}
2303 
2304 	sector_size = get_unaligned_be32(&buffer[8]);
2305 	lba = get_unaligned_be64(&buffer[0]);
2306 
2307 	if (sd_read_protection_type(sdkp, buffer) < 0) {
2308 		sdkp->capacity = 0;
2309 		return -ENODEV;
2310 	}
2311 
2312 	/* Logical blocks per physical block exponent */
2313 	sdkp->physical_block_size = (1 << (buffer[13] & 0xf)) * sector_size;
2314 
2315 	/* RC basis */
2316 	sdkp->rc_basis = (buffer[12] >> 4) & 0x3;
2317 
2318 	/* Lowest aligned logical block */
2319 	alignment = ((buffer[14] & 0x3f) << 8 | buffer[15]) * sector_size;
2320 	blk_queue_alignment_offset(sdp->request_queue, alignment);
2321 	if (alignment && sdkp->first_scan)
2322 		sd_printk(KERN_NOTICE, sdkp,
2323 			  "physical block alignment offset: %u\n", alignment);
2324 
2325 	if (buffer[14] & 0x80) { /* LBPME */
2326 		sdkp->lbpme = 1;
2327 
2328 		if (buffer[14] & 0x40) /* LBPRZ */
2329 			sdkp->lbprz = 1;
2330 
2331 		sd_config_discard(sdkp, SD_LBP_WS16);
2332 	}
2333 
2334 	sdkp->capacity = lba + 1;
2335 	return sector_size;
2336 }
2337 
2338 static int read_capacity_10(struct scsi_disk *sdkp, struct scsi_device *sdp,
2339 						unsigned char *buffer)
2340 {
2341 	unsigned char cmd[16];
2342 	struct scsi_sense_hdr sshdr;
2343 	int sense_valid = 0;
2344 	int the_result;
2345 	int retries = 3, reset_retries = READ_CAPACITY_RETRIES_ON_RESET;
2346 	sector_t lba;
2347 	unsigned sector_size;
2348 
2349 	do {
2350 		cmd[0] = READ_CAPACITY;
2351 		memset(&cmd[1], 0, 9);
2352 		memset(buffer, 0, 8);
2353 
2354 		the_result = scsi_execute_req(sdp, cmd, DMA_FROM_DEVICE,
2355 					buffer, 8, &sshdr,
2356 					SD_TIMEOUT, sdkp->max_retries, NULL);
2357 
2358 		if (media_not_present(sdkp, &sshdr))
2359 			return -ENODEV;
2360 
2361 		if (the_result > 0) {
2362 			sense_valid = scsi_sense_valid(&sshdr);
2363 			if (sense_valid &&
2364 			    sshdr.sense_key == UNIT_ATTENTION &&
2365 			    sshdr.asc == 0x29 && sshdr.ascq == 0x00)
2366 				/* Device reset might occur several times,
2367 				 * give it one more chance */
2368 				if (--reset_retries > 0)
2369 					continue;
2370 		}
2371 		retries--;
2372 
2373 	} while (the_result && retries);
2374 
2375 	if (the_result) {
2376 		sd_print_result(sdkp, "Read Capacity(10) failed", the_result);
2377 		read_capacity_error(sdkp, sdp, &sshdr, sense_valid, the_result);
2378 		return -EINVAL;
2379 	}
2380 
2381 	sector_size = get_unaligned_be32(&buffer[4]);
2382 	lba = get_unaligned_be32(&buffer[0]);
2383 
2384 	if (sdp->no_read_capacity_16 && (lba == 0xffffffff)) {
2385 		/* Some buggy (usb cardreader) devices return an lba of
2386 		   0xffffffff when the want to report a size of 0 (with
2387 		   which they really mean no media is present) */
2388 		sdkp->capacity = 0;
2389 		sdkp->physical_block_size = sector_size;
2390 		return sector_size;
2391 	}
2392 
2393 	sdkp->capacity = lba + 1;
2394 	sdkp->physical_block_size = sector_size;
2395 	return sector_size;
2396 }
2397 
2398 static int sd_try_rc16_first(struct scsi_device *sdp)
2399 {
2400 	if (sdp->host->max_cmd_len < 16)
2401 		return 0;
2402 	if (sdp->try_rc_10_first)
2403 		return 0;
2404 	if (sdp->scsi_level > SCSI_SPC_2)
2405 		return 1;
2406 	if (scsi_device_protection(sdp))
2407 		return 1;
2408 	return 0;
2409 }
2410 
2411 /*
2412  * read disk capacity
2413  */
2414 static void
2415 sd_read_capacity(struct scsi_disk *sdkp, unsigned char *buffer)
2416 {
2417 	int sector_size;
2418 	struct scsi_device *sdp = sdkp->device;
2419 
2420 	if (sd_try_rc16_first(sdp)) {
2421 		sector_size = read_capacity_16(sdkp, sdp, buffer);
2422 		if (sector_size == -EOVERFLOW)
2423 			goto got_data;
2424 		if (sector_size == -ENODEV)
2425 			return;
2426 		if (sector_size < 0)
2427 			sector_size = read_capacity_10(sdkp, sdp, buffer);
2428 		if (sector_size < 0)
2429 			return;
2430 	} else {
2431 		sector_size = read_capacity_10(sdkp, sdp, buffer);
2432 		if (sector_size == -EOVERFLOW)
2433 			goto got_data;
2434 		if (sector_size < 0)
2435 			return;
2436 		if ((sizeof(sdkp->capacity) > 4) &&
2437 		    (sdkp->capacity > 0xffffffffULL)) {
2438 			int old_sector_size = sector_size;
2439 			sd_printk(KERN_NOTICE, sdkp, "Very big device. "
2440 					"Trying to use READ CAPACITY(16).\n");
2441 			sector_size = read_capacity_16(sdkp, sdp, buffer);
2442 			if (sector_size < 0) {
2443 				sd_printk(KERN_NOTICE, sdkp,
2444 					"Using 0xffffffff as device size\n");
2445 				sdkp->capacity = 1 + (sector_t) 0xffffffff;
2446 				sector_size = old_sector_size;
2447 				goto got_data;
2448 			}
2449 			/* Remember that READ CAPACITY(16) succeeded */
2450 			sdp->try_rc_10_first = 0;
2451 		}
2452 	}
2453 
2454 	/* Some devices are known to return the total number of blocks,
2455 	 * not the highest block number.  Some devices have versions
2456 	 * which do this and others which do not.  Some devices we might
2457 	 * suspect of doing this but we don't know for certain.
2458 	 *
2459 	 * If we know the reported capacity is wrong, decrement it.  If
2460 	 * we can only guess, then assume the number of blocks is even
2461 	 * (usually true but not always) and err on the side of lowering
2462 	 * the capacity.
2463 	 */
2464 	if (sdp->fix_capacity ||
2465 	    (sdp->guess_capacity && (sdkp->capacity & 0x01))) {
2466 		sd_printk(KERN_INFO, sdkp, "Adjusting the sector count "
2467 				"from its reported value: %llu\n",
2468 				(unsigned long long) sdkp->capacity);
2469 		--sdkp->capacity;
2470 	}
2471 
2472 got_data:
2473 	if (sector_size == 0) {
2474 		sector_size = 512;
2475 		sd_printk(KERN_NOTICE, sdkp, "Sector size 0 reported, "
2476 			  "assuming 512.\n");
2477 	}
2478 
2479 	if (sector_size != 512 &&
2480 	    sector_size != 1024 &&
2481 	    sector_size != 2048 &&
2482 	    sector_size != 4096) {
2483 		sd_printk(KERN_NOTICE, sdkp, "Unsupported sector size %d.\n",
2484 			  sector_size);
2485 		/*
2486 		 * The user might want to re-format the drive with
2487 		 * a supported sectorsize.  Once this happens, it
2488 		 * would be relatively trivial to set the thing up.
2489 		 * For this reason, we leave the thing in the table.
2490 		 */
2491 		sdkp->capacity = 0;
2492 		/*
2493 		 * set a bogus sector size so the normal read/write
2494 		 * logic in the block layer will eventually refuse any
2495 		 * request on this device without tripping over power
2496 		 * of two sector size assumptions
2497 		 */
2498 		sector_size = 512;
2499 	}
2500 	blk_queue_logical_block_size(sdp->request_queue, sector_size);
2501 	blk_queue_physical_block_size(sdp->request_queue,
2502 				      sdkp->physical_block_size);
2503 	sdkp->device->sector_size = sector_size;
2504 
2505 	if (sdkp->capacity > 0xffffffff)
2506 		sdp->use_16_for_rw = 1;
2507 
2508 }
2509 
2510 /*
2511  * Print disk capacity
2512  */
2513 static void
2514 sd_print_capacity(struct scsi_disk *sdkp,
2515 		  sector_t old_capacity)
2516 {
2517 	int sector_size = sdkp->device->sector_size;
2518 	char cap_str_2[10], cap_str_10[10];
2519 
2520 	if (!sdkp->first_scan && old_capacity == sdkp->capacity)
2521 		return;
2522 
2523 	string_get_size(sdkp->capacity, sector_size,
2524 			STRING_UNITS_2, cap_str_2, sizeof(cap_str_2));
2525 	string_get_size(sdkp->capacity, sector_size,
2526 			STRING_UNITS_10, cap_str_10, sizeof(cap_str_10));
2527 
2528 	sd_printk(KERN_NOTICE, sdkp,
2529 		  "%llu %d-byte logical blocks: (%s/%s)\n",
2530 		  (unsigned long long)sdkp->capacity,
2531 		  sector_size, cap_str_10, cap_str_2);
2532 
2533 	if (sdkp->physical_block_size != sector_size)
2534 		sd_printk(KERN_NOTICE, sdkp,
2535 			  "%u-byte physical blocks\n",
2536 			  sdkp->physical_block_size);
2537 }
2538 
2539 /* called with buffer of length 512 */
2540 static inline int
2541 sd_do_mode_sense(struct scsi_disk *sdkp, int dbd, int modepage,
2542 		 unsigned char *buffer, int len, struct scsi_mode_data *data,
2543 		 struct scsi_sense_hdr *sshdr)
2544 {
2545 	/*
2546 	 * If we must use MODE SENSE(10), make sure that the buffer length
2547 	 * is at least 8 bytes so that the mode sense header fits.
2548 	 */
2549 	if (sdkp->device->use_10_for_ms && len < 8)
2550 		len = 8;
2551 
2552 	return scsi_mode_sense(sdkp->device, dbd, modepage, buffer, len,
2553 			       SD_TIMEOUT, sdkp->max_retries, data,
2554 			       sshdr);
2555 }
2556 
2557 /*
2558  * read write protect setting, if possible - called only in sd_revalidate_disk()
2559  * called with buffer of length SD_BUF_SIZE
2560  */
2561 static void
2562 sd_read_write_protect_flag(struct scsi_disk *sdkp, unsigned char *buffer)
2563 {
2564 	int res;
2565 	struct scsi_device *sdp = sdkp->device;
2566 	struct scsi_mode_data data;
2567 	int old_wp = sdkp->write_prot;
2568 
2569 	set_disk_ro(sdkp->disk, 0);
2570 	if (sdp->skip_ms_page_3f) {
2571 		sd_first_printk(KERN_NOTICE, sdkp, "Assuming Write Enabled\n");
2572 		return;
2573 	}
2574 
2575 	if (sdp->use_192_bytes_for_3f) {
2576 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 192, &data, NULL);
2577 	} else {
2578 		/*
2579 		 * First attempt: ask for all pages (0x3F), but only 4 bytes.
2580 		 * We have to start carefully: some devices hang if we ask
2581 		 * for more than is available.
2582 		 */
2583 		res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 4, &data, NULL);
2584 
2585 		/*
2586 		 * Second attempt: ask for page 0 When only page 0 is
2587 		 * implemented, a request for page 3F may return Sense Key
2588 		 * 5: Illegal Request, Sense Code 24: Invalid field in
2589 		 * CDB.
2590 		 */
2591 		if (res < 0)
2592 			res = sd_do_mode_sense(sdkp, 0, 0, buffer, 4, &data, NULL);
2593 
2594 		/*
2595 		 * Third attempt: ask 255 bytes, as we did earlier.
2596 		 */
2597 		if (res < 0)
2598 			res = sd_do_mode_sense(sdkp, 0, 0x3F, buffer, 255,
2599 					       &data, NULL);
2600 	}
2601 
2602 	if (res < 0) {
2603 		sd_first_printk(KERN_WARNING, sdkp,
2604 			  "Test WP failed, assume Write Enabled\n");
2605 	} else {
2606 		sdkp->write_prot = ((data.device_specific & 0x80) != 0);
2607 		set_disk_ro(sdkp->disk, sdkp->write_prot);
2608 		if (sdkp->first_scan || old_wp != sdkp->write_prot) {
2609 			sd_printk(KERN_NOTICE, sdkp, "Write Protect is %s\n",
2610 				  sdkp->write_prot ? "on" : "off");
2611 			sd_printk(KERN_DEBUG, sdkp, "Mode Sense: %4ph\n", buffer);
2612 		}
2613 	}
2614 }
2615 
2616 /*
2617  * sd_read_cache_type - called only from sd_revalidate_disk()
2618  * called with buffer of length SD_BUF_SIZE
2619  */
2620 static void
2621 sd_read_cache_type(struct scsi_disk *sdkp, unsigned char *buffer)
2622 {
2623 	int len = 0, res;
2624 	struct scsi_device *sdp = sdkp->device;
2625 
2626 	int dbd;
2627 	int modepage;
2628 	int first_len;
2629 	struct scsi_mode_data data;
2630 	struct scsi_sense_hdr sshdr;
2631 	int old_wce = sdkp->WCE;
2632 	int old_rcd = sdkp->RCD;
2633 	int old_dpofua = sdkp->DPOFUA;
2634 
2635 
2636 	if (sdkp->cache_override)
2637 		return;
2638 
2639 	first_len = 4;
2640 	if (sdp->skip_ms_page_8) {
2641 		if (sdp->type == TYPE_RBC)
2642 			goto defaults;
2643 		else {
2644 			if (sdp->skip_ms_page_3f)
2645 				goto defaults;
2646 			modepage = 0x3F;
2647 			if (sdp->use_192_bytes_for_3f)
2648 				first_len = 192;
2649 			dbd = 0;
2650 		}
2651 	} else if (sdp->type == TYPE_RBC) {
2652 		modepage = 6;
2653 		dbd = 8;
2654 	} else {
2655 		modepage = 8;
2656 		dbd = 0;
2657 	}
2658 
2659 	/* cautiously ask */
2660 	res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, first_len,
2661 			&data, &sshdr);
2662 
2663 	if (res < 0)
2664 		goto bad_sense;
2665 
2666 	if (!data.header_length) {
2667 		modepage = 6;
2668 		first_len = 0;
2669 		sd_first_printk(KERN_ERR, sdkp,
2670 				"Missing header in MODE_SENSE response\n");
2671 	}
2672 
2673 	/* that went OK, now ask for the proper length */
2674 	len = data.length;
2675 
2676 	/*
2677 	 * We're only interested in the first three bytes, actually.
2678 	 * But the data cache page is defined for the first 20.
2679 	 */
2680 	if (len < 3)
2681 		goto bad_sense;
2682 	else if (len > SD_BUF_SIZE) {
2683 		sd_first_printk(KERN_NOTICE, sdkp, "Truncating mode parameter "
2684 			  "data from %d to %d bytes\n", len, SD_BUF_SIZE);
2685 		len = SD_BUF_SIZE;
2686 	}
2687 	if (modepage == 0x3F && sdp->use_192_bytes_for_3f)
2688 		len = 192;
2689 
2690 	/* Get the data */
2691 	if (len > first_len)
2692 		res = sd_do_mode_sense(sdkp, dbd, modepage, buffer, len,
2693 				&data, &sshdr);
2694 
2695 	if (!res) {
2696 		int offset = data.header_length + data.block_descriptor_length;
2697 
2698 		while (offset < len) {
2699 			u8 page_code = buffer[offset] & 0x3F;
2700 			u8 spf       = buffer[offset] & 0x40;
2701 
2702 			if (page_code == 8 || page_code == 6) {
2703 				/* We're interested only in the first 3 bytes.
2704 				 */
2705 				if (len - offset <= 2) {
2706 					sd_first_printk(KERN_ERR, sdkp,
2707 						"Incomplete mode parameter "
2708 							"data\n");
2709 					goto defaults;
2710 				} else {
2711 					modepage = page_code;
2712 					goto Page_found;
2713 				}
2714 			} else {
2715 				/* Go to the next page */
2716 				if (spf && len - offset > 3)
2717 					offset += 4 + (buffer[offset+2] << 8) +
2718 						buffer[offset+3];
2719 				else if (!spf && len - offset > 1)
2720 					offset += 2 + buffer[offset+1];
2721 				else {
2722 					sd_first_printk(KERN_ERR, sdkp,
2723 							"Incomplete mode "
2724 							"parameter data\n");
2725 					goto defaults;
2726 				}
2727 			}
2728 		}
2729 
2730 		sd_first_printk(KERN_WARNING, sdkp,
2731 				"No Caching mode page found\n");
2732 		goto defaults;
2733 
2734 	Page_found:
2735 		if (modepage == 8) {
2736 			sdkp->WCE = ((buffer[offset + 2] & 0x04) != 0);
2737 			sdkp->RCD = ((buffer[offset + 2] & 0x01) != 0);
2738 		} else {
2739 			sdkp->WCE = ((buffer[offset + 2] & 0x01) == 0);
2740 			sdkp->RCD = 0;
2741 		}
2742 
2743 		sdkp->DPOFUA = (data.device_specific & 0x10) != 0;
2744 		if (sdp->broken_fua) {
2745 			sd_first_printk(KERN_NOTICE, sdkp, "Disabling FUA\n");
2746 			sdkp->DPOFUA = 0;
2747 		} else if (sdkp->DPOFUA && !sdkp->device->use_10_for_rw &&
2748 			   !sdkp->device->use_16_for_rw) {
2749 			sd_first_printk(KERN_NOTICE, sdkp,
2750 				  "Uses READ/WRITE(6), disabling FUA\n");
2751 			sdkp->DPOFUA = 0;
2752 		}
2753 
2754 		/* No cache flush allowed for write protected devices */
2755 		if (sdkp->WCE && sdkp->write_prot)
2756 			sdkp->WCE = 0;
2757 
2758 		if (sdkp->first_scan || old_wce != sdkp->WCE ||
2759 		    old_rcd != sdkp->RCD || old_dpofua != sdkp->DPOFUA)
2760 			sd_printk(KERN_NOTICE, sdkp,
2761 				  "Write cache: %s, read cache: %s, %s\n",
2762 				  sdkp->WCE ? "enabled" : "disabled",
2763 				  sdkp->RCD ? "disabled" : "enabled",
2764 				  sdkp->DPOFUA ? "supports DPO and FUA"
2765 				  : "doesn't support DPO or FUA");
2766 
2767 		return;
2768 	}
2769 
2770 bad_sense:
2771 	if (scsi_sense_valid(&sshdr) &&
2772 	    sshdr.sense_key == ILLEGAL_REQUEST &&
2773 	    sshdr.asc == 0x24 && sshdr.ascq == 0x0)
2774 		/* Invalid field in CDB */
2775 		sd_first_printk(KERN_NOTICE, sdkp, "Cache data unavailable\n");
2776 	else
2777 		sd_first_printk(KERN_ERR, sdkp,
2778 				"Asking for cache data failed\n");
2779 
2780 defaults:
2781 	if (sdp->wce_default_on) {
2782 		sd_first_printk(KERN_NOTICE, sdkp,
2783 				"Assuming drive cache: write back\n");
2784 		sdkp->WCE = 1;
2785 	} else {
2786 		sd_first_printk(KERN_WARNING, sdkp,
2787 				"Assuming drive cache: write through\n");
2788 		sdkp->WCE = 0;
2789 	}
2790 	sdkp->RCD = 0;
2791 	sdkp->DPOFUA = 0;
2792 }
2793 
2794 /*
2795  * The ATO bit indicates whether the DIF application tag is available
2796  * for use by the operating system.
2797  */
2798 static void sd_read_app_tag_own(struct scsi_disk *sdkp, unsigned char *buffer)
2799 {
2800 	int res, offset;
2801 	struct scsi_device *sdp = sdkp->device;
2802 	struct scsi_mode_data data;
2803 	struct scsi_sense_hdr sshdr;
2804 
2805 	if (sdp->type != TYPE_DISK && sdp->type != TYPE_ZBC)
2806 		return;
2807 
2808 	if (sdkp->protection_type == 0)
2809 		return;
2810 
2811 	res = scsi_mode_sense(sdp, 1, 0x0a, buffer, 36, SD_TIMEOUT,
2812 			      sdkp->max_retries, &data, &sshdr);
2813 
2814 	if (res < 0 || !data.header_length ||
2815 	    data.length < 6) {
2816 		sd_first_printk(KERN_WARNING, sdkp,
2817 			  "getting Control mode page failed, assume no ATO\n");
2818 
2819 		if (scsi_sense_valid(&sshdr))
2820 			sd_print_sense_hdr(sdkp, &sshdr);
2821 
2822 		return;
2823 	}
2824 
2825 	offset = data.header_length + data.block_descriptor_length;
2826 
2827 	if ((buffer[offset] & 0x3f) != 0x0a) {
2828 		sd_first_printk(KERN_ERR, sdkp, "ATO Got wrong page\n");
2829 		return;
2830 	}
2831 
2832 	if ((buffer[offset + 5] & 0x80) == 0)
2833 		return;
2834 
2835 	sdkp->ATO = 1;
2836 
2837 	return;
2838 }
2839 
2840 /**
2841  * sd_read_block_limits - Query disk device for preferred I/O sizes.
2842  * @sdkp: disk to query
2843  */
2844 static void sd_read_block_limits(struct scsi_disk *sdkp)
2845 {
2846 	unsigned int sector_sz = sdkp->device->sector_size;
2847 	const int vpd_len = 64;
2848 	unsigned char *buffer = kmalloc(vpd_len, GFP_KERNEL);
2849 
2850 	if (!buffer ||
2851 	    /* Block Limits VPD */
2852 	    scsi_get_vpd_page(sdkp->device, 0xb0, buffer, vpd_len))
2853 		goto out;
2854 
2855 	blk_queue_io_min(sdkp->disk->queue,
2856 			 get_unaligned_be16(&buffer[6]) * sector_sz);
2857 
2858 	sdkp->max_xfer_blocks = get_unaligned_be32(&buffer[8]);
2859 	sdkp->opt_xfer_blocks = get_unaligned_be32(&buffer[12]);
2860 
2861 	if (buffer[3] == 0x3c) {
2862 		unsigned int lba_count, desc_count;
2863 
2864 		sdkp->max_ws_blocks = (u32)get_unaligned_be64(&buffer[36]);
2865 
2866 		if (!sdkp->lbpme)
2867 			goto out;
2868 
2869 		lba_count = get_unaligned_be32(&buffer[20]);
2870 		desc_count = get_unaligned_be32(&buffer[24]);
2871 
2872 		if (lba_count && desc_count)
2873 			sdkp->max_unmap_blocks = lba_count;
2874 
2875 		sdkp->unmap_granularity = get_unaligned_be32(&buffer[28]);
2876 
2877 		if (buffer[32] & 0x80)
2878 			sdkp->unmap_alignment =
2879 				get_unaligned_be32(&buffer[32]) & ~(1 << 31);
2880 
2881 		if (!sdkp->lbpvpd) { /* LBP VPD page not provided */
2882 
2883 			if (sdkp->max_unmap_blocks)
2884 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2885 			else
2886 				sd_config_discard(sdkp, SD_LBP_WS16);
2887 
2888 		} else {	/* LBP VPD page tells us what to use */
2889 			if (sdkp->lbpu && sdkp->max_unmap_blocks)
2890 				sd_config_discard(sdkp, SD_LBP_UNMAP);
2891 			else if (sdkp->lbpws)
2892 				sd_config_discard(sdkp, SD_LBP_WS16);
2893 			else if (sdkp->lbpws10)
2894 				sd_config_discard(sdkp, SD_LBP_WS10);
2895 			else
2896 				sd_config_discard(sdkp, SD_LBP_DISABLE);
2897 		}
2898 	}
2899 
2900  out:
2901 	kfree(buffer);
2902 }
2903 
2904 /**
2905  * sd_read_block_characteristics - Query block dev. characteristics
2906  * @sdkp: disk to query
2907  */
2908 static void sd_read_block_characteristics(struct scsi_disk *sdkp)
2909 {
2910 	struct request_queue *q = sdkp->disk->queue;
2911 	unsigned char *buffer;
2912 	u16 rot;
2913 	const int vpd_len = 64;
2914 
2915 	buffer = kmalloc(vpd_len, GFP_KERNEL);
2916 
2917 	if (!buffer ||
2918 	    /* Block Device Characteristics VPD */
2919 	    scsi_get_vpd_page(sdkp->device, 0xb1, buffer, vpd_len))
2920 		goto out;
2921 
2922 	rot = get_unaligned_be16(&buffer[4]);
2923 
2924 	if (rot == 1) {
2925 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
2926 		blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, q);
2927 	}
2928 
2929 	if (sdkp->device->type == TYPE_ZBC) {
2930 		/* Host-managed */
2931 		blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HM);
2932 	} else {
2933 		sdkp->zoned = (buffer[8] >> 4) & 3;
2934 		if (sdkp->zoned == 1) {
2935 			/* Host-aware */
2936 			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_HA);
2937 		} else {
2938 			/* Regular disk or drive managed disk */
2939 			blk_queue_set_zoned(sdkp->disk, BLK_ZONED_NONE);
2940 		}
2941 	}
2942 
2943 	if (!sdkp->first_scan)
2944 		goto out;
2945 
2946 	if (blk_queue_is_zoned(q)) {
2947 		sd_printk(KERN_NOTICE, sdkp, "Host-%s zoned block device\n",
2948 		      q->limits.zoned == BLK_ZONED_HM ? "managed" : "aware");
2949 	} else {
2950 		if (sdkp->zoned == 1)
2951 			sd_printk(KERN_NOTICE, sdkp,
2952 				  "Host-aware SMR disk used as regular disk\n");
2953 		else if (sdkp->zoned == 2)
2954 			sd_printk(KERN_NOTICE, sdkp,
2955 				  "Drive-managed SMR disk\n");
2956 	}
2957 
2958  out:
2959 	kfree(buffer);
2960 }
2961 
2962 /**
2963  * sd_read_block_provisioning - Query provisioning VPD page
2964  * @sdkp: disk to query
2965  */
2966 static void sd_read_block_provisioning(struct scsi_disk *sdkp)
2967 {
2968 	unsigned char *buffer;
2969 	const int vpd_len = 8;
2970 
2971 	if (sdkp->lbpme == 0)
2972 		return;
2973 
2974 	buffer = kmalloc(vpd_len, GFP_KERNEL);
2975 
2976 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb2, buffer, vpd_len))
2977 		goto out;
2978 
2979 	sdkp->lbpvpd	= 1;
2980 	sdkp->lbpu	= (buffer[5] >> 7) & 1;	/* UNMAP */
2981 	sdkp->lbpws	= (buffer[5] >> 6) & 1;	/* WRITE SAME(16) with UNMAP */
2982 	sdkp->lbpws10	= (buffer[5] >> 5) & 1;	/* WRITE SAME(10) with UNMAP */
2983 
2984  out:
2985 	kfree(buffer);
2986 }
2987 
2988 static void sd_read_write_same(struct scsi_disk *sdkp, unsigned char *buffer)
2989 {
2990 	struct scsi_device *sdev = sdkp->device;
2991 
2992 	if (sdev->host->no_write_same) {
2993 		sdev->no_write_same = 1;
2994 
2995 		return;
2996 	}
2997 
2998 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, INQUIRY) < 0) {
2999 		/* too large values might cause issues with arcmsr */
3000 		int vpd_buf_len = 64;
3001 
3002 		sdev->no_report_opcodes = 1;
3003 
3004 		/* Disable WRITE SAME if REPORT SUPPORTED OPERATION
3005 		 * CODES is unsupported and the device has an ATA
3006 		 * Information VPD page (SAT).
3007 		 */
3008 		if (!scsi_get_vpd_page(sdev, 0x89, buffer, vpd_buf_len))
3009 			sdev->no_write_same = 1;
3010 	}
3011 
3012 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME_16) == 1)
3013 		sdkp->ws16 = 1;
3014 
3015 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE, WRITE_SAME) == 1)
3016 		sdkp->ws10 = 1;
3017 }
3018 
3019 static void sd_read_security(struct scsi_disk *sdkp, unsigned char *buffer)
3020 {
3021 	struct scsi_device *sdev = sdkp->device;
3022 
3023 	if (!sdev->security_supported)
3024 		return;
3025 
3026 	if (scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3027 			SECURITY_PROTOCOL_IN) == 1 &&
3028 	    scsi_report_opcode(sdev, buffer, SD_BUF_SIZE,
3029 			SECURITY_PROTOCOL_OUT) == 1)
3030 		sdkp->security = 1;
3031 }
3032 
3033 static inline sector_t sd64_to_sectors(struct scsi_disk *sdkp, u8 *buf)
3034 {
3035 	return logical_to_sectors(sdkp->device, get_unaligned_be64(buf));
3036 }
3037 
3038 /**
3039  * sd_read_cpr - Query concurrent positioning ranges
3040  * @sdkp:	disk to query
3041  */
3042 static void sd_read_cpr(struct scsi_disk *sdkp)
3043 {
3044 	struct blk_independent_access_ranges *iars = NULL;
3045 	unsigned char *buffer = NULL;
3046 	unsigned int nr_cpr = 0;
3047 	int i, vpd_len, buf_len = SD_BUF_SIZE;
3048 	u8 *desc;
3049 
3050 	/*
3051 	 * We need to have the capacity set first for the block layer to be
3052 	 * able to check the ranges.
3053 	 */
3054 	if (sdkp->first_scan)
3055 		return;
3056 
3057 	if (!sdkp->capacity)
3058 		goto out;
3059 
3060 	/*
3061 	 * Concurrent Positioning Ranges VPD: there can be at most 256 ranges,
3062 	 * leading to a maximum page size of 64 + 256*32 bytes.
3063 	 */
3064 	buf_len = 64 + 256*32;
3065 	buffer = kmalloc(buf_len, GFP_KERNEL);
3066 	if (!buffer || scsi_get_vpd_page(sdkp->device, 0xb9, buffer, buf_len))
3067 		goto out;
3068 
3069 	/* We must have at least a 64B header and one 32B range descriptor */
3070 	vpd_len = get_unaligned_be16(&buffer[2]) + 3;
3071 	if (vpd_len > buf_len || vpd_len < 64 + 32 || (vpd_len & 31)) {
3072 		sd_printk(KERN_ERR, sdkp,
3073 			  "Invalid Concurrent Positioning Ranges VPD page\n");
3074 		goto out;
3075 	}
3076 
3077 	nr_cpr = (vpd_len - 64) / 32;
3078 	if (nr_cpr == 1) {
3079 		nr_cpr = 0;
3080 		goto out;
3081 	}
3082 
3083 	iars = disk_alloc_independent_access_ranges(sdkp->disk, nr_cpr);
3084 	if (!iars) {
3085 		nr_cpr = 0;
3086 		goto out;
3087 	}
3088 
3089 	desc = &buffer[64];
3090 	for (i = 0; i < nr_cpr; i++, desc += 32) {
3091 		if (desc[0] != i) {
3092 			sd_printk(KERN_ERR, sdkp,
3093 				"Invalid Concurrent Positioning Range number\n");
3094 			nr_cpr = 0;
3095 			break;
3096 		}
3097 
3098 		iars->ia_range[i].sector = sd64_to_sectors(sdkp, desc + 8);
3099 		iars->ia_range[i].nr_sectors = sd64_to_sectors(sdkp, desc + 16);
3100 	}
3101 
3102 out:
3103 	disk_set_independent_access_ranges(sdkp->disk, iars);
3104 	if (nr_cpr && sdkp->nr_actuators != nr_cpr) {
3105 		sd_printk(KERN_NOTICE, sdkp,
3106 			  "%u concurrent positioning ranges\n", nr_cpr);
3107 		sdkp->nr_actuators = nr_cpr;
3108 	}
3109 
3110 	kfree(buffer);
3111 }
3112 
3113 /*
3114  * Determine the device's preferred I/O size for reads and writes
3115  * unless the reported value is unreasonably small, large, not a
3116  * multiple of the physical block size, or simply garbage.
3117  */
3118 static bool sd_validate_opt_xfer_size(struct scsi_disk *sdkp,
3119 				      unsigned int dev_max)
3120 {
3121 	struct scsi_device *sdp = sdkp->device;
3122 	unsigned int opt_xfer_bytes =
3123 		logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3124 
3125 	if (sdkp->opt_xfer_blocks == 0)
3126 		return false;
3127 
3128 	if (sdkp->opt_xfer_blocks > dev_max) {
3129 		sd_first_printk(KERN_WARNING, sdkp,
3130 				"Optimal transfer size %u logical blocks " \
3131 				"> dev_max (%u logical blocks)\n",
3132 				sdkp->opt_xfer_blocks, dev_max);
3133 		return false;
3134 	}
3135 
3136 	if (sdkp->opt_xfer_blocks > SD_DEF_XFER_BLOCKS) {
3137 		sd_first_printk(KERN_WARNING, sdkp,
3138 				"Optimal transfer size %u logical blocks " \
3139 				"> sd driver limit (%u logical blocks)\n",
3140 				sdkp->opt_xfer_blocks, SD_DEF_XFER_BLOCKS);
3141 		return false;
3142 	}
3143 
3144 	if (opt_xfer_bytes < PAGE_SIZE) {
3145 		sd_first_printk(KERN_WARNING, sdkp,
3146 				"Optimal transfer size %u bytes < " \
3147 				"PAGE_SIZE (%u bytes)\n",
3148 				opt_xfer_bytes, (unsigned int)PAGE_SIZE);
3149 		return false;
3150 	}
3151 
3152 	if (opt_xfer_bytes & (sdkp->physical_block_size - 1)) {
3153 		sd_first_printk(KERN_WARNING, sdkp,
3154 				"Optimal transfer size %u bytes not a " \
3155 				"multiple of physical block size (%u bytes)\n",
3156 				opt_xfer_bytes, sdkp->physical_block_size);
3157 		return false;
3158 	}
3159 
3160 	sd_first_printk(KERN_INFO, sdkp, "Optimal transfer size %u bytes\n",
3161 			opt_xfer_bytes);
3162 	return true;
3163 }
3164 
3165 /**
3166  *	sd_revalidate_disk - called the first time a new disk is seen,
3167  *	performs disk spin up, read_capacity, etc.
3168  *	@disk: struct gendisk we care about
3169  **/
3170 static int sd_revalidate_disk(struct gendisk *disk)
3171 {
3172 	struct scsi_disk *sdkp = scsi_disk(disk);
3173 	struct scsi_device *sdp = sdkp->device;
3174 	struct request_queue *q = sdkp->disk->queue;
3175 	sector_t old_capacity = sdkp->capacity;
3176 	unsigned char *buffer;
3177 	unsigned int dev_max, rw_max;
3178 
3179 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
3180 				      "sd_revalidate_disk\n"));
3181 
3182 	/*
3183 	 * If the device is offline, don't try and read capacity or any
3184 	 * of the other niceties.
3185 	 */
3186 	if (!scsi_device_online(sdp))
3187 		goto out;
3188 
3189 	buffer = kmalloc(SD_BUF_SIZE, GFP_KERNEL);
3190 	if (!buffer) {
3191 		sd_printk(KERN_WARNING, sdkp, "sd_revalidate_disk: Memory "
3192 			  "allocation failure.\n");
3193 		goto out;
3194 	}
3195 
3196 	sd_spinup_disk(sdkp);
3197 
3198 	/*
3199 	 * Without media there is no reason to ask; moreover, some devices
3200 	 * react badly if we do.
3201 	 */
3202 	if (sdkp->media_present) {
3203 		sd_read_capacity(sdkp, buffer);
3204 
3205 		/*
3206 		 * set the default to rotational.  All non-rotational devices
3207 		 * support the block characteristics VPD page, which will
3208 		 * cause this to be updated correctly and any device which
3209 		 * doesn't support it should be treated as rotational.
3210 		 */
3211 		blk_queue_flag_clear(QUEUE_FLAG_NONROT, q);
3212 		blk_queue_flag_set(QUEUE_FLAG_ADD_RANDOM, q);
3213 
3214 		if (scsi_device_supports_vpd(sdp)) {
3215 			sd_read_block_provisioning(sdkp);
3216 			sd_read_block_limits(sdkp);
3217 			sd_read_block_characteristics(sdkp);
3218 			sd_zbc_read_zones(sdkp, buffer);
3219 			sd_read_cpr(sdkp);
3220 		}
3221 
3222 		sd_print_capacity(sdkp, old_capacity);
3223 
3224 		sd_read_write_protect_flag(sdkp, buffer);
3225 		sd_read_cache_type(sdkp, buffer);
3226 		sd_read_app_tag_own(sdkp, buffer);
3227 		sd_read_write_same(sdkp, buffer);
3228 		sd_read_security(sdkp, buffer);
3229 	}
3230 
3231 	/*
3232 	 * We now have all cache related info, determine how we deal
3233 	 * with flush requests.
3234 	 */
3235 	sd_set_flush_flag(sdkp);
3236 
3237 	/* Initial block count limit based on CDB TRANSFER LENGTH field size. */
3238 	dev_max = sdp->use_16_for_rw ? SD_MAX_XFER_BLOCKS : SD_DEF_XFER_BLOCKS;
3239 
3240 	/* Some devices report a maximum block count for READ/WRITE requests. */
3241 	dev_max = min_not_zero(dev_max, sdkp->max_xfer_blocks);
3242 	q->limits.max_dev_sectors = logical_to_sectors(sdp, dev_max);
3243 
3244 	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
3245 		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
3246 		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
3247 	} else {
3248 		q->limits.io_opt = 0;
3249 		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
3250 				      (sector_t)BLK_DEF_MAX_SECTORS);
3251 	}
3252 
3253 	/* Do not exceed controller limit */
3254 	rw_max = min(rw_max, queue_max_hw_sectors(q));
3255 
3256 	/*
3257 	 * Only update max_sectors if previously unset or if the current value
3258 	 * exceeds the capabilities of the hardware.
3259 	 */
3260 	if (sdkp->first_scan ||
3261 	    q->limits.max_sectors > q->limits.max_dev_sectors ||
3262 	    q->limits.max_sectors > q->limits.max_hw_sectors)
3263 		q->limits.max_sectors = rw_max;
3264 
3265 	sdkp->first_scan = 0;
3266 
3267 	set_capacity_and_notify(disk, logical_to_sectors(sdp, sdkp->capacity));
3268 	sd_config_write_same(sdkp);
3269 	kfree(buffer);
3270 
3271 	/*
3272 	 * For a zoned drive, revalidating the zones can be done only once
3273 	 * the gendisk capacity is set. So if this fails, set back the gendisk
3274 	 * capacity to 0.
3275 	 */
3276 	if (sd_zbc_revalidate_zones(sdkp))
3277 		set_capacity_and_notify(disk, 0);
3278 
3279  out:
3280 	return 0;
3281 }
3282 
3283 /**
3284  *	sd_unlock_native_capacity - unlock native capacity
3285  *	@disk: struct gendisk to set capacity for
3286  *
3287  *	Block layer calls this function if it detects that partitions
3288  *	on @disk reach beyond the end of the device.  If the SCSI host
3289  *	implements ->unlock_native_capacity() method, it's invoked to
3290  *	give it a chance to adjust the device capacity.
3291  *
3292  *	CONTEXT:
3293  *	Defined by block layer.  Might sleep.
3294  */
3295 static void sd_unlock_native_capacity(struct gendisk *disk)
3296 {
3297 	struct scsi_device *sdev = scsi_disk(disk)->device;
3298 
3299 	if (sdev->host->hostt->unlock_native_capacity)
3300 		sdev->host->hostt->unlock_native_capacity(sdev);
3301 }
3302 
3303 /**
3304  *	sd_format_disk_name - format disk name
3305  *	@prefix: name prefix - ie. "sd" for SCSI disks
3306  *	@index: index of the disk to format name for
3307  *	@buf: output buffer
3308  *	@buflen: length of the output buffer
3309  *
3310  *	SCSI disk names starts at sda.  The 26th device is sdz and the
3311  *	27th is sdaa.  The last one for two lettered suffix is sdzz
3312  *	which is followed by sdaaa.
3313  *
3314  *	This is basically 26 base counting with one extra 'nil' entry
3315  *	at the beginning from the second digit on and can be
3316  *	determined using similar method as 26 base conversion with the
3317  *	index shifted -1 after each digit is computed.
3318  *
3319  *	CONTEXT:
3320  *	Don't care.
3321  *
3322  *	RETURNS:
3323  *	0 on success, -errno on failure.
3324  */
3325 static int sd_format_disk_name(char *prefix, int index, char *buf, int buflen)
3326 {
3327 	const int base = 'z' - 'a' + 1;
3328 	char *begin = buf + strlen(prefix);
3329 	char *end = buf + buflen;
3330 	char *p;
3331 	int unit;
3332 
3333 	p = end - 1;
3334 	*p = '\0';
3335 	unit = base;
3336 	do {
3337 		if (p == begin)
3338 			return -EINVAL;
3339 		*--p = 'a' + (index % unit);
3340 		index = (index / unit) - 1;
3341 	} while (index >= 0);
3342 
3343 	memmove(begin, p, end - p);
3344 	memcpy(buf, prefix, strlen(prefix));
3345 
3346 	return 0;
3347 }
3348 
3349 /**
3350  *	sd_probe - called during driver initialization and whenever a
3351  *	new scsi device is attached to the system. It is called once
3352  *	for each scsi device (not just disks) present.
3353  *	@dev: pointer to device object
3354  *
3355  *	Returns 0 if successful (or not interested in this scsi device
3356  *	(e.g. scanner)); 1 when there is an error.
3357  *
3358  *	Note: this function is invoked from the scsi mid-level.
3359  *	This function sets up the mapping between a given
3360  *	<host,channel,id,lun> (found in sdp) and new device name
3361  *	(e.g. /dev/sda). More precisely it is the block device major
3362  *	and minor number that is chosen here.
3363  *
3364  *	Assume sd_probe is not re-entrant (for time being)
3365  *	Also think about sd_probe() and sd_remove() running coincidentally.
3366  **/
3367 static int sd_probe(struct device *dev)
3368 {
3369 	struct scsi_device *sdp = to_scsi_device(dev);
3370 	struct scsi_disk *sdkp;
3371 	struct gendisk *gd;
3372 	int index;
3373 	int error;
3374 
3375 	scsi_autopm_get_device(sdp);
3376 	error = -ENODEV;
3377 	if (sdp->type != TYPE_DISK &&
3378 	    sdp->type != TYPE_ZBC &&
3379 	    sdp->type != TYPE_MOD &&
3380 	    sdp->type != TYPE_RBC)
3381 		goto out;
3382 
3383 	if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED) && sdp->type == TYPE_ZBC) {
3384 		sdev_printk(KERN_WARNING, sdp,
3385 			    "Unsupported ZBC host-managed device.\n");
3386 		goto out;
3387 	}
3388 
3389 	SCSI_LOG_HLQUEUE(3, sdev_printk(KERN_INFO, sdp,
3390 					"sd_probe\n"));
3391 
3392 	error = -ENOMEM;
3393 	sdkp = kzalloc(sizeof(*sdkp), GFP_KERNEL);
3394 	if (!sdkp)
3395 		goto out;
3396 
3397 	gd = __alloc_disk_node(sdp->request_queue, NUMA_NO_NODE,
3398 			       &sd_bio_compl_lkclass);
3399 	if (!gd)
3400 		goto out_free;
3401 
3402 	index = ida_alloc(&sd_index_ida, GFP_KERNEL);
3403 	if (index < 0) {
3404 		sdev_printk(KERN_WARNING, sdp, "sd_probe: memory exhausted.\n");
3405 		goto out_put;
3406 	}
3407 
3408 	error = sd_format_disk_name("sd", index, gd->disk_name, DISK_NAME_LEN);
3409 	if (error) {
3410 		sdev_printk(KERN_WARNING, sdp, "SCSI disk (sd) name length exceeded.\n");
3411 		goto out_free_index;
3412 	}
3413 
3414 	sdkp->device = sdp;
3415 	sdkp->disk = gd;
3416 	sdkp->index = index;
3417 	sdkp->max_retries = SD_MAX_RETRIES;
3418 	atomic_set(&sdkp->openers, 0);
3419 	atomic_set(&sdkp->device->ioerr_cnt, 0);
3420 
3421 	if (!sdp->request_queue->rq_timeout) {
3422 		if (sdp->type != TYPE_MOD)
3423 			blk_queue_rq_timeout(sdp->request_queue, SD_TIMEOUT);
3424 		else
3425 			blk_queue_rq_timeout(sdp->request_queue,
3426 					     SD_MOD_TIMEOUT);
3427 	}
3428 
3429 	device_initialize(&sdkp->disk_dev);
3430 	sdkp->disk_dev.parent = get_device(dev);
3431 	sdkp->disk_dev.class = &sd_disk_class;
3432 	dev_set_name(&sdkp->disk_dev, "%s", dev_name(dev));
3433 
3434 	error = device_add(&sdkp->disk_dev);
3435 	if (error) {
3436 		put_device(&sdkp->disk_dev);
3437 		goto out;
3438 	}
3439 
3440 	dev_set_drvdata(dev, sdkp);
3441 
3442 	gd->major = sd_major((index & 0xf0) >> 4);
3443 	gd->first_minor = ((index & 0xf) << 4) | (index & 0xfff00);
3444 	gd->minors = SD_MINORS;
3445 
3446 	gd->fops = &sd_fops;
3447 	gd->private_data = sdkp;
3448 
3449 	/* defaults, until the device tells us otherwise */
3450 	sdp->sector_size = 512;
3451 	sdkp->capacity = 0;
3452 	sdkp->media_present = 1;
3453 	sdkp->write_prot = 0;
3454 	sdkp->cache_override = 0;
3455 	sdkp->WCE = 0;
3456 	sdkp->RCD = 0;
3457 	sdkp->ATO = 0;
3458 	sdkp->first_scan = 1;
3459 	sdkp->max_medium_access_timeouts = SD_MAX_MEDIUM_TIMEOUTS;
3460 
3461 	sd_revalidate_disk(gd);
3462 
3463 	if (sdp->removable) {
3464 		gd->flags |= GENHD_FL_REMOVABLE;
3465 		gd->events |= DISK_EVENT_MEDIA_CHANGE;
3466 		gd->event_flags = DISK_EVENT_FLAG_POLL | DISK_EVENT_FLAG_UEVENT;
3467 	}
3468 
3469 	blk_pm_runtime_init(sdp->request_queue, dev);
3470 	if (sdp->rpm_autosuspend) {
3471 		pm_runtime_set_autosuspend_delay(dev,
3472 			sdp->host->hostt->rpm_autosuspend_delay);
3473 	}
3474 
3475 	error = device_add_disk(dev, gd, NULL);
3476 	if (error) {
3477 		put_device(&sdkp->disk_dev);
3478 		blk_cleanup_disk(gd);
3479 		goto out;
3480 	}
3481 
3482 	if (sdkp->capacity)
3483 		sd_dif_config_host(sdkp);
3484 
3485 	sd_revalidate_disk(gd);
3486 
3487 	if (sdkp->security) {
3488 		sdkp->opal_dev = init_opal_dev(sdkp, &sd_sec_submit);
3489 		if (sdkp->opal_dev)
3490 			sd_printk(KERN_NOTICE, sdkp, "supports TCG Opal\n");
3491 	}
3492 
3493 	sd_printk(KERN_NOTICE, sdkp, "Attached SCSI %sdisk\n",
3494 		  sdp->removable ? "removable " : "");
3495 	scsi_autopm_put_device(sdp);
3496 
3497 	return 0;
3498 
3499  out_free_index:
3500 	ida_free(&sd_index_ida, index);
3501  out_put:
3502 	put_disk(gd);
3503  out_free:
3504 	sd_zbc_release_disk(sdkp);
3505 	kfree(sdkp);
3506  out:
3507 	scsi_autopm_put_device(sdp);
3508 	return error;
3509 }
3510 
3511 /**
3512  *	sd_remove - called whenever a scsi disk (previously recognized by
3513  *	sd_probe) is detached from the system. It is called (potentially
3514  *	multiple times) during sd module unload.
3515  *	@dev: pointer to device object
3516  *
3517  *	Note: this function is invoked from the scsi mid-level.
3518  *	This function potentially frees up a device name (e.g. /dev/sdc)
3519  *	that could be re-used by a subsequent sd_probe().
3520  *	This function is not called when the built-in sd driver is "exit-ed".
3521  **/
3522 static int sd_remove(struct device *dev)
3523 {
3524 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3525 
3526 	scsi_autopm_get_device(sdkp->device);
3527 
3528 	device_del(&sdkp->disk_dev);
3529 	del_gendisk(sdkp->disk);
3530 	sd_shutdown(dev);
3531 
3532 	put_disk(sdkp->disk);
3533 	return 0;
3534 }
3535 
3536 static void scsi_disk_release(struct device *dev)
3537 {
3538 	struct scsi_disk *sdkp = to_scsi_disk(dev);
3539 
3540 	ida_free(&sd_index_ida, sdkp->index);
3541 	sd_zbc_release_disk(sdkp);
3542 	put_device(&sdkp->device->sdev_gendev);
3543 	free_opal_dev(sdkp->opal_dev);
3544 
3545 	kfree(sdkp);
3546 }
3547 
3548 static int sd_start_stop_device(struct scsi_disk *sdkp, int start)
3549 {
3550 	unsigned char cmd[6] = { START_STOP };	/* START_VALID */
3551 	struct scsi_sense_hdr sshdr;
3552 	struct scsi_device *sdp = sdkp->device;
3553 	int res;
3554 
3555 	if (start)
3556 		cmd[4] |= 1;	/* START */
3557 
3558 	if (sdp->start_stop_pwr_cond)
3559 		cmd[4] |= start ? 1 << 4 : 3 << 4;	/* Active or Standby */
3560 
3561 	if (!scsi_device_online(sdp))
3562 		return -ENODEV;
3563 
3564 	res = scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL, &sshdr,
3565 			SD_TIMEOUT, sdkp->max_retries, 0, RQF_PM, NULL);
3566 	if (res) {
3567 		sd_print_result(sdkp, "Start/Stop Unit failed", res);
3568 		if (res > 0 && scsi_sense_valid(&sshdr)) {
3569 			sd_print_sense_hdr(sdkp, &sshdr);
3570 			/* 0x3a is medium not present */
3571 			if (sshdr.asc == 0x3a)
3572 				res = 0;
3573 		}
3574 	}
3575 
3576 	/* SCSI error codes must not go to the generic layer */
3577 	if (res)
3578 		return -EIO;
3579 
3580 	return 0;
3581 }
3582 
3583 /*
3584  * Send a SYNCHRONIZE CACHE instruction down to the device through
3585  * the normal SCSI command structure.  Wait for the command to
3586  * complete.
3587  */
3588 static void sd_shutdown(struct device *dev)
3589 {
3590 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3591 
3592 	if (!sdkp)
3593 		return;         /* this can happen */
3594 
3595 	if (pm_runtime_suspended(dev))
3596 		return;
3597 
3598 	if (sdkp->WCE && sdkp->media_present) {
3599 		sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3600 		sd_sync_cache(sdkp, NULL);
3601 	}
3602 
3603 	if (system_state != SYSTEM_RESTART && sdkp->device->manage_start_stop) {
3604 		sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3605 		sd_start_stop_device(sdkp, 0);
3606 	}
3607 }
3608 
3609 static int sd_suspend_common(struct device *dev, bool ignore_stop_errors)
3610 {
3611 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3612 	struct scsi_sense_hdr sshdr;
3613 	int ret = 0;
3614 
3615 	if (!sdkp)	/* E.g.: runtime suspend following sd_remove() */
3616 		return 0;
3617 
3618 	if (sdkp->WCE && sdkp->media_present) {
3619 		if (!sdkp->device->silence_suspend)
3620 			sd_printk(KERN_NOTICE, sdkp, "Synchronizing SCSI cache\n");
3621 		ret = sd_sync_cache(sdkp, &sshdr);
3622 
3623 		if (ret) {
3624 			/* ignore OFFLINE device */
3625 			if (ret == -ENODEV)
3626 				return 0;
3627 
3628 			if (!scsi_sense_valid(&sshdr) ||
3629 			    sshdr.sense_key != ILLEGAL_REQUEST)
3630 				return ret;
3631 
3632 			/*
3633 			 * sshdr.sense_key == ILLEGAL_REQUEST means this drive
3634 			 * doesn't support sync. There's not much to do and
3635 			 * suspend shouldn't fail.
3636 			 */
3637 			ret = 0;
3638 		}
3639 	}
3640 
3641 	if (sdkp->device->manage_start_stop) {
3642 		if (!sdkp->device->silence_suspend)
3643 			sd_printk(KERN_NOTICE, sdkp, "Stopping disk\n");
3644 		/* an error is not worth aborting a system sleep */
3645 		ret = sd_start_stop_device(sdkp, 0);
3646 		if (ignore_stop_errors)
3647 			ret = 0;
3648 	}
3649 
3650 	return ret;
3651 }
3652 
3653 static int sd_suspend_system(struct device *dev)
3654 {
3655 	if (pm_runtime_suspended(dev))
3656 		return 0;
3657 
3658 	return sd_suspend_common(dev, true);
3659 }
3660 
3661 static int sd_suspend_runtime(struct device *dev)
3662 {
3663 	return sd_suspend_common(dev, false);
3664 }
3665 
3666 static int sd_resume(struct device *dev)
3667 {
3668 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3669 	int ret;
3670 
3671 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3672 		return 0;
3673 
3674 	if (!sdkp->device->manage_start_stop)
3675 		return 0;
3676 
3677 	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
3678 	ret = sd_start_stop_device(sdkp, 1);
3679 	if (!ret)
3680 		opal_unlock_from_suspend(sdkp->opal_dev);
3681 	return ret;
3682 }
3683 
3684 static int sd_resume_system(struct device *dev)
3685 {
3686 	if (pm_runtime_suspended(dev))
3687 		return 0;
3688 
3689 	return sd_resume(dev);
3690 }
3691 
3692 static int sd_resume_runtime(struct device *dev)
3693 {
3694 	struct scsi_disk *sdkp = dev_get_drvdata(dev);
3695 	struct scsi_device *sdp;
3696 
3697 	if (!sdkp)	/* E.g.: runtime resume at the start of sd_probe() */
3698 		return 0;
3699 
3700 	sdp = sdkp->device;
3701 
3702 	if (sdp->ignore_media_change) {
3703 		/* clear the device's sense data */
3704 		static const u8 cmd[10] = { REQUEST_SENSE };
3705 
3706 		if (scsi_execute(sdp, cmd, DMA_NONE, NULL, 0, NULL,
3707 				 NULL, sdp->request_queue->rq_timeout, 1, 0,
3708 				 RQF_PM, NULL))
3709 			sd_printk(KERN_NOTICE, sdkp,
3710 				  "Failed to clear sense data\n");
3711 	}
3712 
3713 	return sd_resume(dev);
3714 }
3715 
3716 /**
3717  *	init_sd - entry point for this driver (both when built in or when
3718  *	a module).
3719  *
3720  *	Note: this function registers this driver with the scsi mid-level.
3721  **/
3722 static int __init init_sd(void)
3723 {
3724 	int majors = 0, i, err;
3725 
3726 	SCSI_LOG_HLQUEUE(3, printk("init_sd: sd driver entry point\n"));
3727 
3728 	for (i = 0; i < SD_MAJORS; i++) {
3729 		if (__register_blkdev(sd_major(i), "sd", sd_default_probe))
3730 			continue;
3731 		majors++;
3732 	}
3733 
3734 	if (!majors)
3735 		return -ENODEV;
3736 
3737 	err = class_register(&sd_disk_class);
3738 	if (err)
3739 		goto err_out;
3740 
3741 	sd_cdb_cache = kmem_cache_create("sd_ext_cdb", SD_EXT_CDB_SIZE,
3742 					 0, 0, NULL);
3743 	if (!sd_cdb_cache) {
3744 		printk(KERN_ERR "sd: can't init extended cdb cache\n");
3745 		err = -ENOMEM;
3746 		goto err_out_class;
3747 	}
3748 
3749 	sd_page_pool = mempool_create_page_pool(SD_MEMPOOL_SIZE, 0);
3750 	if (!sd_page_pool) {
3751 		printk(KERN_ERR "sd: can't init discard page pool\n");
3752 		err = -ENOMEM;
3753 		goto err_out_cache;
3754 	}
3755 
3756 	err = scsi_register_driver(&sd_template.gendrv);
3757 	if (err)
3758 		goto err_out_driver;
3759 
3760 	return 0;
3761 
3762 err_out_driver:
3763 	mempool_destroy(sd_page_pool);
3764 
3765 err_out_cache:
3766 	kmem_cache_destroy(sd_cdb_cache);
3767 
3768 err_out_class:
3769 	class_unregister(&sd_disk_class);
3770 err_out:
3771 	for (i = 0; i < SD_MAJORS; i++)
3772 		unregister_blkdev(sd_major(i), "sd");
3773 	return err;
3774 }
3775 
3776 /**
3777  *	exit_sd - exit point for this driver (when it is a module).
3778  *
3779  *	Note: this function unregisters this driver from the scsi mid-level.
3780  **/
3781 static void __exit exit_sd(void)
3782 {
3783 	int i;
3784 
3785 	SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
3786 
3787 	scsi_unregister_driver(&sd_template.gendrv);
3788 	mempool_destroy(sd_page_pool);
3789 	kmem_cache_destroy(sd_cdb_cache);
3790 
3791 	class_unregister(&sd_disk_class);
3792 
3793 	for (i = 0; i < SD_MAJORS; i++)
3794 		unregister_blkdev(sd_major(i), "sd");
3795 }
3796 
3797 module_init(init_sd);
3798 module_exit(exit_sd);
3799 
3800 void sd_print_sense_hdr(struct scsi_disk *sdkp, struct scsi_sense_hdr *sshdr)
3801 {
3802 	scsi_print_sense_hdr(sdkp->device,
3803 			     sdkp->disk ? sdkp->disk->disk_name : NULL, sshdr);
3804 }
3805 
3806 void sd_print_result(const struct scsi_disk *sdkp, const char *msg, int result)
3807 {
3808 	const char *hb_string = scsi_hostbyte_string(result);
3809 
3810 	if (hb_string)
3811 		sd_printk(KERN_INFO, sdkp,
3812 			  "%s: Result: hostbyte=%s driverbyte=%s\n", msg,
3813 			  hb_string ? hb_string : "invalid",
3814 			  "DRIVER_OK");
3815 	else
3816 		sd_printk(KERN_INFO, sdkp,
3817 			  "%s: Result: hostbyte=0x%02x driverbyte=%s\n",
3818 			  msg, host_byte(result), "DRIVER_OK");
3819 }
3820