xref: /illumos-gate/usr/src/cmd/format/analyze.c (revision d2a70789f056fc6c9ce3ab047b52126d80b0e3da)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 /*
27  * This file contains routines to analyze the surface of a disk.
28  */
29 #include "global.h"
30 #include "analyze.h"
31 #include <stdlib.h>
32 #include <errno.h>
33 #include "misc.h"
34 #include "defect.h"
35 #include "label.h"
36 #include "param.h"
37 #include "checkdev.h"
38 
39 
40 /*
41  * These global variables control the surface analysis process.  They
42  * are set from a command in the defect menu.
43  */
44 int	scan_entire = 1;		/* scan whole disk flag */
45 diskaddr_t	scan_lower = 0;			/* lower bound */
46 diskaddr_t	scan_upper = 0;			/* upper bound */
47 int	scan_correct = 1;		/* correct errors flag */
48 int	scan_stop = 0;			/* stop after error flag */
49 int	scan_loop = 0;			/* loop forever flag */
50 int	scan_passes = 2;		/* number of passes */
51 int	scan_random = 0;		/* random patterns flag */
52 uint_t	scan_size = 0;			/* sectors/scan operation */
53 int	scan_auto = 1;			/* scan after format flag */
54 int	scan_restore_defects = 1;	/* restore defect list after writing */
55 int	scan_restore_label = 1;		/* restore label after writing */
56 
57 /*
58  * These are summary variables to print out info after analysis.
59  * Values less than 0 imply they are invalid.
60  */
61 offset_t	scan_cur_block = -1;		/* current block */
62 int64_t		scan_blocks_fixed = -1;		/* # blocks repaired */
63 
64 /*
65  * This variable is used to tell whether the most recent surface
66  * analysis error was caused by a media defect or some other problem.
67  */
68 int	media_error;			/* error was caused by defect */
69 
70 int	disk_error;			/* disk errors during analysis */
71 
72 /*
73  * These are the data patterns used if random patterns are not chosen.
74  * They are designed to show pattern dependent errors.
75  */
76 static unsigned int	scan_patterns[] = {
77 	0xc6dec6de,
78 	0x6db6db6d,
79 	0x00000000,
80 	0xffffffff,
81 	0xaaaaaaaa,
82 };
83 #define	NPATTERNS	5		/* number of predefined patterns */
84 
85 /*
86  * These are the data patterns from the SunFed requirements document.
87  */
88 static unsigned int purge_patterns[] = {	/* patterns to be written */
89 	0xaaaaaaaa,		/* 10101010... */
90 	0x55555555,		/* 01010101...  == UUUU... */
91 	0xaaaaaaaa,		/* 10101010... */
92 	0xaaaaaaaa,		/* 10101010... */
93 };
94 
95 static unsigned int alpha_pattern =  0x40404040;   /* 10000000...  == @@@@... */
96 
97 /* Function prototypes */
98 #ifdef	__STDC__
99 
100 static int	scan_repair(diskaddr_t bn, int mode);
101 static int	analyze_blocks(int flags, diskaddr_t blkno, uint_t blkcnt,
102 		unsigned data, int init, int driver_flags, int *xfercntp);
103 static int	handle_error_conditions(void);
104 static int	verify_blocks(int flags, diskaddr_t blkno, uint_t blkcnt,
105 		unsigned data, int driver_flags, int *xfercntp);
106 #else	/* __STDC__ */
107 
108 static int	scan_repair();
109 static int	analyze_blocks();
110 static int	handle_error_conditions();
111 static int	verify_blocks();
112 
113 #endif	/* __STDC__ */
114 
115 /*
116  * This routine performs a surface analysis based upon the global
117  * parameters.  It is called from several commands in the defect menu,
118  * and from the format command in the command menu (if post-format
119  * analysis is enable).
120  */
121 int
122 do_scan(flags, mode)
123 	int	flags, mode;
124 {
125 	diskaddr_t	start, end, curnt;
126 	int	pass, needinit, data;
127 	uint_t	size;
128 	int	status, founderr, i, j;
129 	int	error = 0;
130 	int	pattern = 0;
131 	int	xfercnt;
132 
133 	/*
134 	 * Check to be sure we aren't correcting without a defect list
135 	 * if the controller can correct the defect.
136 	 */
137 	if (scan_correct && !EMBEDDED_SCSI && (cur_ops->op_repair != NULL) &&
138 			(cur_list.list == NULL)) {
139 		err_print("Current Defect List must be initialized ");
140 		err_print("to do automatic repair.\n");
141 		return (-1);
142 	}
143 	/*
144 	 * Define the bounds of the scan.
145 	 */
146 	if (scan_entire) {
147 		start = 0;
148 	    if (cur_label == L_TYPE_SOLARIS) {
149 		if (cur_ctype->ctype_flags & CF_SCSI)
150 			end = datasects() - 1;
151 		else
152 			end = physsects() - 1;
153 	    } else if (cur_label == L_TYPE_EFI) {
154 		end = cur_parts->etoc->efi_last_lba;
155 	    }
156 	} else {
157 		start = scan_lower;
158 		end = scan_upper;
159 	}
160 	/*
161 	 * Make sure the user knows if we are scanning over a mounted
162 	 * partition.
163 	 */
164 	if ((flags & (SCAN_PATTERN | SCAN_WRITE)) &&
165 	    (checkmount(start, end))) {
166 		err_print("Cannot do analysis on a mounted partition.\n");
167 		return (-1);
168 	}
169 
170 	/*
171 	 * Make sure the user knows if we are scanning over a
172 	 * partition being used for swapping.
173 	 */
174 	if ((flags & (SCAN_PATTERN | SCAN_WRITE)) &&
175 	    (checkswap(start, end))) {
176 		err_print("Cannot do analysis on a partition \
177 		    which is currently being used for swapping.\n");
178 		return (-1);
179 	}
180 
181 	/*
182 	 * Check to see if any partitions used for svm, vxvm, ZFS zpool
183 	 * or live upgrade are on the disk.
184 	 */
185 	if ((flags & (SCAN_PATTERN | SCAN_WRITE)) &&
186 	    (checkdevinuse(cur_disk->disk_name, (diskaddr_t)-1,
187 	    (diskaddr_t)-1, 0, 0))) {
188 		err_print("Cannot do analysis on a partition "
189 		    "while it in use as described above.\n");
190 		return (-1);
191 	}
192 
193 	/*
194 	 * If we are scanning destructively over certain sectors,
195 	 * we mark the defect list and/or label dirty so it will get rewritten.
196 	 */
197 	if (flags & (SCAN_PATTERN | SCAN_WRITE)) {
198 	    if (cur_label == L_TYPE_SOLARIS) {
199 		if (start < (diskaddr_t)totalsects() &&
200 				end >= (diskaddr_t)datasects()) {
201 			if (!EMBEDDED_SCSI) {
202 				cur_list.flags |= LIST_DIRTY;
203 			}
204 			if (cur_disk->disk_flags & DSK_LABEL)
205 				cur_flags |= LABEL_DIRTY;
206 		}
207 	    }
208 	    if (start == 0) {
209 		if (cur_disk->disk_flags & DSK_LABEL)
210 			cur_flags |= LABEL_DIRTY;
211 	    }
212 	}
213 	/*
214 	 * Initialize the summary info on sectors repaired.
215 	 */
216 	scan_blocks_fixed = 0;
217 	/*
218 	 * Loop through the passes of the scan. If required, loop forever.
219 	 */
220 	for (pass = 0; pass < scan_passes || scan_loop; pass++) {
221 		/*
222 		 * Determine the data pattern to use if pattern testing
223 		 * is to be done.
224 		 */
225 		if (flags & SCAN_PATTERN) {
226 			if (scan_random)
227 				data = (int)mrand48();
228 			else
229 				data = scan_patterns[pass % NPPATTERNS];
230 
231 			if (flags & SCAN_PURGE) {
232 				flags &= ~(SCAN_PURGE_READ_PASS
233 						| SCAN_PURGE_ALPHA_PASS);
234 				switch (pattern % (NPPATTERNS + 1)) {
235 				case NPPATTERNS:
236 					pattern = 0;
237 					if (!error) {
238 					    fmt_print(
239 "\nThe last %d passes were successful, running alpha pattern pass", NPPATTERNS);
240 					    flags |= SCAN_PURGE_ALPHA_PASS;
241 					    data = alpha_pattern;
242 					} else {
243 					    data = purge_patterns[pattern];
244 					    pattern++;
245 					};
246 					break;
247 				case READPATTERN:
248 					flags |=  SCAN_PURGE_READ_PASS;
249 				default:
250 					data = purge_patterns[pattern];
251 					pattern++;
252 					break;
253 				}
254 			}
255 			fmt_print("\n        pass %d", pass);
256 			fmt_print(" - pattern = 0x%x", data);
257 		} else
258 			fmt_print("\n        pass %d", pass);
259 
260 		fmt_print("\n");
261 		/*
262 		 * Mark the pattern buffer as corrupt, since it
263 		 * hasn't been initialized.
264 		 */
265 		needinit = 1;
266 		/*
267 		 * Print the first block number to the log file if
268 		 * logging is on so there is some record of what
269 		 * analysis was performed.
270 		 */
271 		if (log_file) {
272 			pr_dblock(log_print, start);
273 			log_print("\n");
274 		}
275 		/*
276 		 * Loop through this pass, each time analyzing an amount
277 		 * specified by the global parameters.
278 		 */
279 		xfercnt = 0;
280 		for (curnt = start; curnt <= end; curnt += size) {
281 			if ((end - curnt) < scan_size)
282 				size = end - curnt + 1;
283 			else
284 				size = scan_size;
285 			/*
286 			 * Print out where we are, so we don't look dead.
287 			 * Also store it in summary info for logging.
288 			 */
289 			scan_cur_block = curnt;
290 			nolog_print("   ");
291 			pr_dblock(nolog_print, curnt);
292 			nolog_print("  \015");
293 			(void) fflush(stdout);
294 			disk_error = 0;
295 			/*
296 			 * Do the actual analysis.
297 			 */
298 			status = analyze_blocks(flags, curnt, size,
299 			    (unsigned)data, needinit, (F_ALLERRS | F_SILENT),
300 			    &xfercnt);
301 			/*
302 			 * If there were no errors, the pattern buffer is
303 			 * still initialized, and we just loop to next chunk.
304 			 */
305 			needinit = 0;
306 			if (!status)
307 				continue;
308 			/*
309 			 * There was an error. Check if surface analysis
310 			 * can be continued.
311 			 */
312 			if (handle_error_conditions()) {
313 				scan_blocks_fixed = scan_cur_block = -1;
314 				return (-1);
315 			}
316 			/*
317 			 * There was an error. Mark the pattern buffer
318 			 * corrupt so it will get reinitialized.
319 			 */
320 			needinit = 1;
321 			/*
322 			 * If it was not a media error, ignore it.
323 			 */
324 			if (!media_error)
325 				continue;
326 			/*
327 			 * Loop 5 times through each sector of the chunk,
328 			 * analyzing them individually.
329 			 */
330 			nolog_print("   ");
331 			pr_dblock(nolog_print, curnt);
332 			nolog_print("  \015");
333 			(void) fflush(stdout);
334 			founderr = 0;
335 			for (j = 0; j < size * 5; j++) {
336 				i = j % size;
337 				disk_error = 0;
338 				status = analyze_blocks(flags, (curnt + i), 1,
339 				    (unsigned)data, needinit, F_ALLERRS, NULL);
340 				needinit = 0;
341 				if (!status)
342 					continue;
343 				/*
344 				 * There was an error. Check if surface analysis
345 				 * can be continued.
346 				 */
347 				if (handle_error_conditions()) {
348 					scan_blocks_fixed = scan_cur_block = -1;
349 					return (-1);
350 				}
351 				/*
352 				 * An error occurred.  Mark the buffer
353 				 * corrupt and see if it was media
354 				 * related.
355 				 */
356 				needinit = 1;
357 				if (!media_error)
358 					continue;
359 				/*
360 				 * We found a bad sector. Print out a message
361 				 * and fix it if required.
362 				 */
363 				founderr = 1;
364 				if (scan_correct && (flags != SCAN_VALID)) {
365 					if (scan_repair(curnt+i, mode)) {
366 						error = -1;
367 					}
368 				} else
369 					err_print("\n");
370 				/*
371 				 * Stop after the error if required.
372 				 */
373 				if (scan_stop)
374 					goto out;
375 			}
376 			/*
377 			 * Mark the pattern buffer corrupt to be safe.
378 			 */
379 			needinit = 1;
380 			/*
381 			 * We didn't find an individual sector that was bad.
382 			 * Print out a warning.
383 			 */
384 			if (!founderr) {
385 				err_print("Warning: unable to pinpoint ");
386 				err_print("defective block.\n");
387 			}
388 		}
389 		/*
390 		 * Print the end of each pass to the log file.
391 		 */
392 		enter_critical();
393 		if (log_file) {
394 			pr_dblock(log_print, scan_cur_block);
395 			log_print("\n");
396 		}
397 		scan_cur_block = -1;
398 		exit_critical();
399 		fmt_print("\n");
400 
401 		/*
402 		 * alternate the read and write for SCAN_VERIFY test
403 		 */
404 		if (flags & SCAN_VERIFY) {
405 			flags ^= SCAN_VERIFY_READ_PASS;
406 		}
407 	}
408 out:
409 	/*
410 	 * We got here either by giving up after an error or falling
411 	 * through after all passes were completed.
412 	 */
413 	fmt_print("\n");
414 	enter_critical();
415 	/*
416 	 * If the defect list is dirty, write it to disk,
417 	 * if scan_restore_defects (the default) is true.
418 	 */
419 	if (!EMBEDDED_SCSI && (cur_list.flags & LIST_DIRTY) &&
420 				(scan_restore_defects)) {
421 		cur_list.flags = 0;
422 		write_deflist(&cur_list);
423 		}
424 	/*
425 	 * If the label is dirty, write it to disk.
426 	 * if scan_restore_label (the default) is true.
427 	 */
428 	if ((cur_flags & LABEL_DIRTY) && (scan_restore_label)) {
429 		cur_flags &= ~LABEL_DIRTY;
430 		(void) write_label();
431 	}
432 	/*
433 	 * If we dropped down to here after an error, we need to write
434 	 * the final block number to the log file for record keeping.
435 	 */
436 	if (log_file && scan_cur_block >= 0) {
437 		pr_dblock(log_print, scan_cur_block);
438 		log_print("\n");
439 	}
440 	fmt_print("Total of %lld defective blocks repaired.\n",
441 		scan_blocks_fixed);
442 	/*
443 	 * Reinitialize the logging variables so they don't get used
444 	 * when they are not really valid.
445 	 */
446 	scan_blocks_fixed = scan_cur_block = -1;
447 	exit_critical();
448 	return (error);
449 }
450 
451 
452 /*
453  * This routine is called to repair a bad block discovered
454  * during a scan operation.  Return 0 for success, 1 for failure.
455  * (This has been extracted out of do_scan(), to simplify it.)
456  */
457 static int
458 scan_repair(bn, mode)
459 	diskaddr_t	bn;
460 	int	mode;
461 {
462 	int	status;
463 	int	result = 1;
464 	char	*buf;
465 	int	buf_is_good;
466 	int	i;
467 
468 	if (cur_ops->op_repair == NULL) {
469 		err_print("Warning: Controller does ");
470 		err_print("not support repairing.\n\n");
471 		return (result);
472 	}
473 
474 	buf = malloc(cur_blksz);
475 	if (buf == NULL) {
476 		err_print("Warning: no memory.\n\n");
477 		return (result);
478 	}
479 	enter_critical();
480 
481 	/*
482 	 * Determine if the error appears to be hard or soft.  We
483 	 * already assume there's an error.  If we can get any
484 	 * good data out of the sector, write that data back
485 	 * after the repair.
486 	 */
487 	buf_is_good = 0;
488 	for (i = 0; i < 5; i++) {
489 		status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, bn, 1,
490 				buf, F_SILENT, NULL);
491 		if (status == 0) {
492 			buf_is_good = 1;
493 			break;
494 		}
495 	}
496 
497 	fmt_print("Repairing %s error on %llu (",
498 				buf_is_good ? "soft" : "hard", bn);
499 	pr_dblock(fmt_print, bn);
500 	fmt_print(")...");
501 
502 	status = (*cur_ops->op_repair)(bn, mode);
503 	if (status) {
504 		/*
505 		 * If the repair failed, we note it and will return the
506 		 * failure. However, the analysis goes on.
507 		 */
508 		fmt_print("failed.\n\n");
509 	} else {
510 		/*
511 		 * The repair worked.  Write the good data we could
512 		 * recover from the failed block, if possible.
513 		 * If not, zero the block.  In doing so, try to
514 		 * determine if the new block appears ok.
515 		 */
516 		if (!buf_is_good) {
517 			bzero(buf, cur_blksz);
518 			fmt_print("Warning: Block %llu zero-filled.\n", bn);
519 		} else {
520 			fmt_print("ok.\n");
521 		}
522 		status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, bn,
523 					1, buf, (F_SILENT | F_ALLERRS), NULL);
524 		if (status == 0) {
525 			status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, bn,
526 					1, buf, (F_SILENT | F_ALLERRS), NULL);
527 		}
528 		if (status) {
529 			fmt_print("The new block also appears defective.\n");
530 		}
531 		fmt_print("\n");
532 		/*
533 		 * add the defect to the list and write the list out.
534 		 * Also, kill the working list so it will get resynced
535 		 * with the current list.
536 		 *
537 		 * For embedded scsi, we don't require a defect list.
538 		 * However, if we have one, add the defect if the
539 		 * list includes the grown list.  If not, kill it
540 		 * to force a resync if we need the list later.
541 		 */
542 		if (EMBEDDED_SCSI) {
543 			if (cur_list.list != NULL) {
544 				if (cur_list.flags & LIST_PGLIST) {
545 					add_ldef(bn, &cur_list);
546 				} else {
547 					kill_deflist(&cur_list);
548 				}
549 			}
550 		/*
551 		 * The next "if" statement reflects the fix for
552 		 * bug id 1026096 where format keeps adding the
553 		 * same defect to the defect list.
554 		 */
555 		} else if (cur_ctype->ctype_flags & CF_WLIST) {
556 			kill_deflist(&cur_list);
557 			(*cur_ops->op_ex_cur)(&cur_list);
558 			fmt_print("Current list updated\n");
559 		} else {
560 			add_ldef(bn, &cur_list);
561 			write_deflist(&cur_list);
562 		}
563 		kill_deflist(&work_list);
564 
565 		/* Log the repair.  */
566 		scan_blocks_fixed++;
567 
568 		/* return ok */
569 		result = 0;
570 	}
571 
572 	exit_critical();
573 	free(buf);
574 	return (result);
575 }
576 
577 
578 /*
579  * This routine analyzes a set of sectors on the disk.  It simply returns
580  * an error if a defect is found.  It is called by do_scan().
581  */
582 static int
583 analyze_blocks(flags, blkno, blkcnt, data, init, driver_flags, xfercntp)
584 	int	flags, driver_flags, init;
585 	uint_t	blkcnt;
586 	register unsigned data;
587 	diskaddr_t	blkno;
588 	int	*xfercntp;
589 {
590 	int		corrupt = 0;
591 	int		status;
592 	register diskaddr_t	i, nints;
593 	register unsigned *ptr = (uint_t *)pattern_buf;
594 
595 	media_error = 0;
596 	if (flags & SCAN_VERIFY) {
597 		return (verify_blocks(flags, blkno, blkcnt, data,
598 		    driver_flags, xfercntp));
599 	}
600 
601 	/*
602 	 * Initialize the pattern buffer if necessary.
603 	 */
604 	nints = (diskaddr_t)blkcnt * cur_blksz / sizeof (int);
605 	if ((flags & SCAN_PATTERN) && init) {
606 		for (i = 0; i < nints; i++)
607 			*((int *)((int *)pattern_buf + i)) = data;
608 	}
609 	/*
610 	 * Lock out interrupts so we can insure valid data will get
611 	 * restored. This is necessary because there are modes
612 	 * of scanning that corrupt the disk data then restore it at
613 	 * the end of the analysis.
614 	 */
615 	enter_critical();
616 	/*
617 	 * If the disk data is valid, read it into the data buffer.
618 	 */
619 	if (flags & SCAN_VALID) {
620 		status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, blkno,
621 		    blkcnt, (caddr_t)cur_buf, driver_flags, xfercntp);
622 		if (status)
623 			goto bad;
624 	}
625 	/*
626 	 * If we are doing pattern testing, write and read the pattern
627 	 * from the pattern buffer.
628 	 */
629 	if (flags & SCAN_PATTERN) {
630 		/*
631 		 * If the disk data was valid, mark it corrupt so we know
632 		 * to restore it later.
633 		 */
634 		if (flags & SCAN_VALID)
635 			corrupt++;
636 		/*
637 		 * Only write if we're not on the read pass of SCAN_PURGE.
638 		 */
639 		if (!(flags & SCAN_PURGE_READ_PASS)) {
640 			status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
641 			    blkcnt, (caddr_t)pattern_buf, driver_flags,
642 			    xfercntp);
643 			if (status)
644 			    goto bad;
645 		}
646 		/*
647 		 * Only read if we are on the read pass of SCAN_PURGE, if we
648 		 * are purging.
649 		 */
650 		if ((!(flags & SCAN_PURGE)) || (flags & SCAN_PURGE_READ_PASS)) {
651 			status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, blkno,
652 			    blkcnt, (caddr_t)pattern_buf, driver_flags,
653 			    xfercntp);
654 			if (status)
655 			    goto bad;
656 		}
657 	}
658 	/*
659 	 * If we are doing a data compare, make sure the pattern
660 	 * came back intact.
661 	 * Only compare if we are on the read pass of SCAN_PURGE, or
662 	 * we wrote random data instead of the expected data pattern.
663 	 */
664 	if ((flags & SCAN_COMPARE) || (flags & SCAN_PURGE_READ_PASS)) {
665 		for (i = nints, ptr = (uint_t *)pattern_buf; i; i--)
666 			if (*ptr++ != data) {
667 				err_print("Data miscompare error (expecting ");
668 				err_print("0x%x, got 0x%x) at ", data,
669 					*((int *)((int *)pattern_buf +
670 					(nints - i))));
671 				pr_dblock(err_print, blkno);
672 				err_print(", offset = 0x%llx.\n",
673 					(nints - i) * sizeof (int));
674 				goto bad;
675 			}
676 	}
677 	/*
678 	 * If we are supposed to write data out, do so.
679 	 */
680 	if (flags & SCAN_WRITE) {
681 		status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
682 		    blkcnt, (caddr_t)cur_buf, driver_flags, xfercntp);
683 		if (status)
684 			goto bad;
685 	}
686 	exit_critical();
687 	/*
688 	 * No errors occurred, return ok.
689 	 */
690 	return (0);
691 bad:
692 	/*
693 	 * There was an error.  If the data was corrupted, we write it
694 	 * out from the data buffer to restore it.
695 	 */
696 	if (corrupt) {
697 		if ((*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
698 				blkcnt, (caddr_t)cur_buf, F_NORMAL, xfercntp))
699 		err_print("Warning: unable to restore original data.\n");
700 	}
701 	exit_critical();
702 	/*
703 	 * Return the error.
704 	 */
705 	return (-1);
706 }
707 
708 
709 /*
710  * This routine analyzes a set of sectors on the disk. It simply returns
711  * an error if a defect is found.  It is called by analyze_blocks().
712  * For simplicity, this is done as a separate function instead of
713  * making the analyze_block routine complex.
714  *
715  * This routine implements the 'verify' command.  It writes the disk
716  * by writing unique data for each block; after the write pass, it
717  * reads the data and verifies for correctness. Note that the entire
718  * disk (or the range of disk) is fully written first and then read.
719  * This should eliminate any caching effect on the drives.
720  */
721 static int
722 verify_blocks(int flags,
723 		diskaddr_t blkno,
724 		uint_t blkcnt,
725 		unsigned data,
726 		int driver_flags,
727 		int *xfercntp)
728 {
729 	int		status, i, nints;
730 	unsigned	*ptr = (uint_t *)pattern_buf;
731 
732 	nints = cur_blksz / sizeof (int);
733 
734 	/*
735 	 * Initialize the pattern buffer if we are in write pass.
736 	 * Use the block number itself as data, each block has unique
737 	 * buffer data that way.
738 	 */
739 	if (!(flags & SCAN_VERIFY_READ_PASS)) {
740 		for (data = blkno; data < blkno + blkcnt; data++) {
741 			for (i = 0; i < nints; i++) {
742 				*ptr++ = data;
743 			}
744 		}
745 		ptr = (uint_t *)pattern_buf;
746 	}
747 
748 	/*
749 	 * Only write if we're not on the read pass of SCAN_VERIFY.
750 	 */
751 	if (!(flags & SCAN_VERIFY_READ_PASS)) {
752 		status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
753 		    blkcnt, (caddr_t)pattern_buf, driver_flags, xfercntp);
754 		if (status)
755 			goto bad;
756 	} else {
757 		/*
758 		 * Only read if we are on the read pass of SCAN_VERIFY
759 		 */
760 		status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, blkno,
761 		    blkcnt, (caddr_t)pattern_buf, driver_flags, xfercntp);
762 		if (status)
763 			goto bad;
764 		/*
765 		 * compare and make sure the pattern came back intact.
766 		 */
767 		for (data = blkno; data < blkno + blkcnt; data++) {
768 			for (i = 0; i < nints; i++) {
769 				if (*ptr++ != data) {
770 					ptr--;
771 					err_print("Data miscompare error "
772 					    "(expecting 0x%x, got 0x%x) at ",
773 					    data, *ptr);
774 					pr_dblock(err_print, blkno);
775 					err_print(", offset = 0x%x.\n",
776 					    (ptr - (uint_t *)pattern_buf) *
777 					    sizeof (int));
778 					goto bad;
779 				}
780 			}
781 		}
782 	}
783 	/*
784 	 * No errors occurred, return ok.
785 	 */
786 	return (0);
787 bad:
788 	return (-1);
789 }
790 
791 
792 static int
793 handle_error_conditions()
794 {
795 
796 	/*
797 	 * Check if the errno is ENXIO.
798 	 */
799 	if (errno == ENXIO) {
800 		fmt_print("\n\nWarning:Cannot access drive, ");
801 		fmt_print("aborting surface analysis.\n");
802 		return (-1);
803 	}
804 	/*
805 	 * check for disk errors
806 	 */
807 	switch (disk_error) {
808 	case DISK_STAT_RESERVED:
809 	case DISK_STAT_UNAVAILABLE:
810 		fmt_print("\n\nWarning:Drive may be reserved ");
811 		fmt_print("or has been removed, ");
812 		fmt_print("aborting surface analysis.\n");
813 		return (-1);
814 	case DISK_STAT_NOTREADY:
815 		fmt_print("\n\nWarning: Drive not ready, ");
816 		fmt_print("aborting surface analysis.\n");
817 		return (-1);
818 	case DISK_STAT_DATA_PROTECT:
819 		fmt_print("\n\nWarning: Drive is write protected, ");
820 		fmt_print("aborting surface analysis.\n");
821 		return (-1);
822 	default:
823 		break;
824 	}
825 	return (0);
826 }
827