0
|
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, Version 1.0 only
|
|
6 |
* (the "License"). You may not use this file except in compliance
|
|
7 |
* with the License.
|
|
8 |
*
|
|
9 |
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
|
|
10 |
* or http://www.opensolaris.org/os/licensing.
|
|
11 |
* See the License for the specific language governing permissions
|
|
12 |
* and limitations under the License.
|
|
13 |
*
|
|
14 |
* When distributing Covered Code, include this CDDL HEADER in each
|
|
15 |
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
|
|
16 |
* If applicable, add the following below this CDDL HEADER, with the
|
|
17 |
* fields enclosed by brackets "[]" replaced with your own identifying
|
|
18 |
* information: Portions Copyright [yyyy] [name of copyright owner]
|
|
19 |
*
|
|
20 |
* CDDL HEADER END
|
|
21 |
*/
|
|
22 |
/*
|
767
|
23 |
* Copyright 2005 Sun Microsystems, Inc. All rights reserved.
|
0
|
24 |
* Use is subject to license terms.
|
|
25 |
*/
|
|
26 |
|
|
27 |
#pragma ident "%Z%%M% %I% %E% SMI"
|
|
28 |
|
|
29 |
/*
|
|
30 |
* This file contains routines to analyze the surface of a disk.
|
|
31 |
*/
|
|
32 |
#include "global.h"
|
|
33 |
#include "analyze.h"
|
|
34 |
#include <stdlib.h>
|
|
35 |
#include <errno.h>
|
|
36 |
#include "misc.h"
|
|
37 |
#include "defect.h"
|
|
38 |
#include "label.h"
|
|
39 |
#include "param.h"
|
767
|
40 |
#include "checkdev.h"
|
0
|
41 |
|
|
42 |
|
|
43 |
/*
|
|
44 |
* These global variables control the surface analysis process. They
|
|
45 |
* are set from a command in the defect menu.
|
|
46 |
*/
|
|
47 |
int scan_entire = 1; /* scan whole disk flag */
|
|
48 |
diskaddr_t scan_lower = 0; /* lower bound */
|
|
49 |
diskaddr_t scan_upper = 0; /* upper bound */
|
|
50 |
int scan_correct = 1; /* correct errors flag */
|
|
51 |
int scan_stop = 0; /* stop after error flag */
|
|
52 |
int scan_loop = 0; /* loop forever flag */
|
|
53 |
int scan_passes = 2; /* number of passes */
|
|
54 |
int scan_random = 0; /* random patterns flag */
|
|
55 |
int scan_size = 0; /* sectors/scan operation */
|
|
56 |
int scan_auto = 1; /* scan after format flag */
|
|
57 |
int scan_restore_defects = 1; /* restore defect list after writing */
|
|
58 |
int scan_restore_label = 1; /* restore label after writing */
|
|
59 |
|
|
60 |
/*
|
|
61 |
* These are summary variables to print out info after analysis.
|
|
62 |
* Values less than 0 imply they are invalid.
|
|
63 |
*/
|
|
64 |
offset_t scan_cur_block = -1; /* current block */
|
|
65 |
int64_t scan_blocks_fixed = -1; /* # blocks repaired */
|
|
66 |
|
|
67 |
/*
|
|
68 |
* This variable is used to tell whether the most recent surface
|
|
69 |
* analysis error was caused by a media defect or some other problem.
|
|
70 |
*/
|
|
71 |
int media_error; /* error was caused by defect */
|
|
72 |
|
|
73 |
int disk_error; /* disk errors during analysis */
|
|
74 |
|
|
75 |
/*
|
|
76 |
* These are the data patterns used if random patterns are not chosen.
|
|
77 |
* They are designed to show pattern dependent errors.
|
|
78 |
*/
|
|
79 |
static unsigned int scan_patterns[] = {
|
|
80 |
0xc6dec6de,
|
|
81 |
0x6db6db6d,
|
|
82 |
0x00000000,
|
|
83 |
0xffffffff,
|
|
84 |
0xaaaaaaaa,
|
|
85 |
};
|
|
86 |
#define NPATTERNS 5 /* number of predefined patterns */
|
|
87 |
|
|
88 |
/*
|
|
89 |
* These are the data patterns from the SunFed requirements document.
|
|
90 |
*/
|
|
91 |
static unsigned int purge_patterns[] = { /* patterns to be written */
|
|
92 |
0xaaaaaaaa, /* 10101010... */
|
|
93 |
0x55555555, /* 01010101... == UUUU... */
|
|
94 |
0xaaaaaaaa, /* 10101010... */
|
|
95 |
0xaaaaaaaa, /* 10101010... */
|
|
96 |
};
|
|
97 |
|
|
98 |
static unsigned int alpha_pattern = 0x40404040; /* 10000000... == @@@@... */
|
|
99 |
|
|
100 |
/* Function prototypes */
|
|
101 |
#ifdef __STDC__
|
|
102 |
|
|
103 |
static int scan_repair(diskaddr_t bn, int mode);
|
|
104 |
static int analyze_blocks(int flags, diskaddr_t blkno, int blkcnt,
|
|
105 |
unsigned data, int init, int driver_flags, int *xfercntp);
|
|
106 |
static int handle_error_conditions(void);
|
|
107 |
static int verify_blocks(int flags, diskaddr_t blkno, int blkcnt,
|
|
108 |
unsigned data, int driver_flags, int *xfercntp);
|
|
109 |
#else /* __STDC__ */
|
|
110 |
|
|
111 |
static int scan_repair();
|
|
112 |
static int analyze_blocks();
|
|
113 |
static int handle_error_conditions();
|
|
114 |
static int verify_blocks();
|
|
115 |
|
|
116 |
#endif /* __STDC__ */
|
|
117 |
|
|
118 |
/*
|
|
119 |
* This routine performs a surface analysis based upon the global
|
|
120 |
* parameters. It is called from several commands in the defect menu,
|
|
121 |
* and from the format command in the command menu (if post-format
|
|
122 |
* analysis is enable).
|
|
123 |
*/
|
|
124 |
int
|
|
125 |
do_scan(flags, mode)
|
|
126 |
int flags, mode;
|
|
127 |
{
|
|
128 |
diskaddr_t start, end, curnt;
|
|
129 |
int pass, size, needinit, data;
|
|
130 |
int status, founderr, i, j;
|
|
131 |
int error = 0;
|
|
132 |
int pattern = 0;
|
|
133 |
int xfercnt;
|
|
134 |
|
|
135 |
/*
|
|
136 |
* Check to be sure we aren't correcting without a defect list
|
|
137 |
* if the controller can correct the defect.
|
|
138 |
*/
|
|
139 |
if (scan_correct && !EMBEDDED_SCSI && (cur_ops->op_repair != NULL) &&
|
|
140 |
(cur_list.list == NULL)) {
|
|
141 |
err_print("Current Defect List must be initialized ");
|
|
142 |
err_print("to do automatic repair.\n");
|
|
143 |
return (-1);
|
|
144 |
}
|
|
145 |
/*
|
|
146 |
* Define the bounds of the scan.
|
|
147 |
*/
|
|
148 |
if (scan_entire) {
|
|
149 |
start = 0;
|
|
150 |
if (cur_label == L_TYPE_SOLARIS) {
|
|
151 |
if (cur_ctype->ctype_flags & CF_SCSI)
|
|
152 |
end = datasects() - 1;
|
|
153 |
else
|
|
154 |
end = physsects() - 1;
|
|
155 |
} else if (cur_label == L_TYPE_EFI) {
|
|
156 |
end = cur_parts->etoc->efi_last_lba;
|
|
157 |
}
|
|
158 |
} else {
|
|
159 |
start = scan_lower;
|
|
160 |
end = scan_upper;
|
|
161 |
}
|
|
162 |
/*
|
|
163 |
* Make sure the user knows if we are scanning over a mounted
|
|
164 |
* partition.
|
|
165 |
*/
|
|
166 |
if ((flags & (SCAN_PATTERN | SCAN_WRITE)) &&
|
|
167 |
(checkmount(start, end))) {
|
|
168 |
err_print("Cannot do analysis on a mounted partition.\n");
|
|
169 |
return (-1);
|
|
170 |
}
|
|
171 |
|
|
172 |
/*
|
|
173 |
* Make sure the user knows if we are scanning over a
|
|
174 |
* partition being used for swapping.
|
|
175 |
*/
|
|
176 |
if ((flags & (SCAN_PATTERN | SCAN_WRITE)) &&
|
|
177 |
(checkswap(start, end))) {
|
|
178 |
err_print("Cannot do analysis on a partition \
|
|
179 |
which is currently being used for swapping.\n");
|
|
180 |
return (-1);
|
|
181 |
}
|
|
182 |
|
|
183 |
/*
|
|
184 |
* If we are scanning destructively over certain sectors,
|
|
185 |
* we mark the defect list and/or label dirty so it will get rewritten.
|
|
186 |
*/
|
|
187 |
if (flags & (SCAN_PATTERN | SCAN_WRITE)) {
|
|
188 |
if (cur_label == L_TYPE_SOLARIS) {
|
|
189 |
if (start < (daddr_t)totalsects() &&
|
|
190 |
end >= (daddr_t)datasects()) {
|
|
191 |
if (!EMBEDDED_SCSI) {
|
|
192 |
cur_list.flags |= LIST_DIRTY;
|
|
193 |
}
|
|
194 |
if (cur_disk->disk_flags & DSK_LABEL)
|
|
195 |
cur_flags |= LABEL_DIRTY;
|
|
196 |
}
|
|
197 |
}
|
|
198 |
if (start == 0) {
|
|
199 |
if (cur_disk->disk_flags & DSK_LABEL)
|
|
200 |
cur_flags |= LABEL_DIRTY;
|
|
201 |
}
|
|
202 |
}
|
|
203 |
/*
|
|
204 |
* Initialize the summary info on sectors repaired.
|
|
205 |
*/
|
|
206 |
scan_blocks_fixed = 0;
|
|
207 |
/*
|
|
208 |
* Loop through the passes of the scan. If required, loop forever.
|
|
209 |
*/
|
|
210 |
for (pass = 0; pass < scan_passes || scan_loop; pass++) {
|
|
211 |
/*
|
|
212 |
* Determine the data pattern to use if pattern testing
|
|
213 |
* is to be done.
|
|
214 |
*/
|
|
215 |
if (flags & SCAN_PATTERN) {
|
|
216 |
if (scan_random)
|
|
217 |
data = (int)mrand48();
|
|
218 |
else
|
|
219 |
data = scan_patterns[pass % NPPATTERNS];
|
|
220 |
|
|
221 |
if (flags & SCAN_PURGE) {
|
|
222 |
flags &= ~(SCAN_PURGE_READ_PASS
|
|
223 |
| SCAN_PURGE_ALPHA_PASS);
|
|
224 |
switch (pattern % (NPPATTERNS + 1)) {
|
|
225 |
case NPPATTERNS:
|
|
226 |
pattern = 0;
|
|
227 |
if (!error) {
|
|
228 |
fmt_print(
|
|
229 |
"\nThe last %d passes were successful, running alpha pattern pass", NPPATTERNS);
|
|
230 |
flags |= SCAN_PURGE_ALPHA_PASS;
|
|
231 |
data = alpha_pattern;
|
|
232 |
} else {
|
|
233 |
data = purge_patterns[pattern];
|
|
234 |
pattern++;
|
|
235 |
};
|
|
236 |
break;
|
|
237 |
case READPATTERN:
|
|
238 |
flags |= SCAN_PURGE_READ_PASS;
|
|
239 |
default:
|
|
240 |
data = purge_patterns[pattern];
|
|
241 |
pattern++;
|
|
242 |
break;
|
|
243 |
}
|
|
244 |
}
|
|
245 |
fmt_print("\n pass %d", pass);
|
|
246 |
fmt_print(" - pattern = 0x%x", data);
|
|
247 |
} else
|
|
248 |
fmt_print("\n pass %d", pass);
|
|
249 |
|
|
250 |
fmt_print("\n");
|
|
251 |
/*
|
|
252 |
* Mark the pattern buffer as corrupt, since it
|
|
253 |
* hasn't been initialized.
|
|
254 |
*/
|
|
255 |
needinit = 1;
|
|
256 |
/*
|
|
257 |
* Print the first block number to the log file if
|
|
258 |
* logging is on so there is some record of what
|
|
259 |
* analysis was performed.
|
|
260 |
*/
|
|
261 |
if (log_file) {
|
|
262 |
pr_dblock(log_print, start);
|
|
263 |
log_print("\n");
|
|
264 |
}
|
|
265 |
/*
|
|
266 |
* Loop through this pass, each time analyzing an amount
|
|
267 |
* specified by the global parameters.
|
|
268 |
*/
|
|
269 |
xfercnt = 0;
|
|
270 |
for (curnt = start; curnt <= end; curnt += size) {
|
|
271 |
if ((end - curnt) < scan_size)
|
|
272 |
size = end - curnt + 1;
|
|
273 |
else
|
|
274 |
size = scan_size;
|
|
275 |
/*
|
|
276 |
* Print out where we are, so we don't look dead.
|
|
277 |
* Also store it in summary info for logging.
|
|
278 |
*/
|
|
279 |
scan_cur_block = curnt;
|
|
280 |
nolog_print(" ");
|
|
281 |
pr_dblock(nolog_print, curnt);
|
|
282 |
nolog_print(" \015");
|
|
283 |
(void) fflush(stdout);
|
|
284 |
disk_error = 0;
|
|
285 |
/*
|
|
286 |
* Do the actual analysis.
|
|
287 |
*/
|
|
288 |
status = analyze_blocks(flags, (daddr_t)curnt, size,
|
|
289 |
(unsigned)data, needinit, (F_ALLERRS | F_SILENT),
|
|
290 |
&xfercnt);
|
|
291 |
/*
|
|
292 |
* If there were no errors, the pattern buffer is
|
|
293 |
* still initialized, and we just loop to next chunk.
|
|
294 |
*/
|
|
295 |
needinit = 0;
|
|
296 |
if (!status)
|
|
297 |
continue;
|
|
298 |
/*
|
|
299 |
* There was an error. Check if surface analysis
|
|
300 |
* can be continued.
|
|
301 |
*/
|
|
302 |
if (handle_error_conditions()) {
|
|
303 |
scan_blocks_fixed = scan_cur_block = -1;
|
|
304 |
return (-1);
|
|
305 |
}
|
|
306 |
/*
|
|
307 |
* There was an error. Mark the pattern buffer
|
|
308 |
* corrupt so it will get reinitialized.
|
|
309 |
*/
|
|
310 |
needinit = 1;
|
|
311 |
/*
|
|
312 |
* If it was not a media error, ignore it.
|
|
313 |
*/
|
|
314 |
if (!media_error)
|
|
315 |
continue;
|
|
316 |
/*
|
|
317 |
* Loop 5 times through each sector of the chunk,
|
|
318 |
* analyzing them individually.
|
|
319 |
*/
|
|
320 |
nolog_print(" ");
|
|
321 |
pr_dblock(nolog_print, curnt);
|
|
322 |
nolog_print(" \015");
|
|
323 |
(void) fflush(stdout);
|
|
324 |
founderr = 0;
|
|
325 |
for (j = 0; j < size * 5; j++) {
|
|
326 |
i = j % size;
|
|
327 |
disk_error = 0;
|
|
328 |
status = analyze_blocks(flags, (daddr_t)
|
|
329 |
(curnt + i), 1, (unsigned)data, needinit,
|
|
330 |
F_ALLERRS, NULL);
|
|
331 |
needinit = 0;
|
|
332 |
if (!status)
|
|
333 |
continue;
|
|
334 |
/*
|
|
335 |
* There was an error. Check if surface analysis
|
|
336 |
* can be continued.
|
|
337 |
*/
|
|
338 |
if (handle_error_conditions()) {
|
|
339 |
scan_blocks_fixed = scan_cur_block = -1;
|
|
340 |
return (-1);
|
|
341 |
}
|
|
342 |
/*
|
|
343 |
* An error occurred. Mark the buffer
|
|
344 |
* corrupt and see if it was media
|
|
345 |
* related.
|
|
346 |
*/
|
|
347 |
needinit = 1;
|
|
348 |
if (!media_error)
|
|
349 |
continue;
|
|
350 |
/*
|
|
351 |
* We found a bad sector. Print out a message
|
|
352 |
* and fix it if required.
|
|
353 |
*/
|
|
354 |
founderr = 1;
|
|
355 |
if (scan_correct && (flags != SCAN_VALID)) {
|
|
356 |
if (scan_repair(curnt+i, mode)) {
|
|
357 |
error = -1;
|
|
358 |
}
|
|
359 |
} else
|
|
360 |
err_print("\n");
|
|
361 |
/*
|
|
362 |
* Stop after the error if required.
|
|
363 |
*/
|
|
364 |
if (scan_stop)
|
|
365 |
goto out;
|
|
366 |
}
|
|
367 |
/*
|
|
368 |
* Mark the pattern buffer corrupt to be safe.
|
|
369 |
*/
|
|
370 |
needinit = 1;
|
|
371 |
/*
|
|
372 |
* We didn't find an individual sector that was bad.
|
|
373 |
* Print out a warning.
|
|
374 |
*/
|
|
375 |
if (!founderr) {
|
|
376 |
err_print("Warning: unable to pinpoint ");
|
|
377 |
err_print("defective block.\n");
|
|
378 |
}
|
|
379 |
}
|
|
380 |
/*
|
|
381 |
* Print the end of each pass to the log file.
|
|
382 |
*/
|
|
383 |
enter_critical();
|
|
384 |
if (log_file) {
|
|
385 |
pr_dblock(log_print, scan_cur_block);
|
|
386 |
log_print("\n");
|
|
387 |
}
|
|
388 |
scan_cur_block = -1;
|
|
389 |
exit_critical();
|
|
390 |
fmt_print("\n");
|
|
391 |
|
|
392 |
/*
|
|
393 |
* alternate the read and write for SCAN_VERIFY test
|
|
394 |
*/
|
|
395 |
if (flags & SCAN_VERIFY) {
|
|
396 |
flags ^= SCAN_VERIFY_READ_PASS;
|
|
397 |
}
|
|
398 |
}
|
|
399 |
out:
|
|
400 |
/*
|
|
401 |
* We got here either by giving up after an error or falling
|
|
402 |
* through after all passes were completed.
|
|
403 |
*/
|
|
404 |
fmt_print("\n");
|
|
405 |
enter_critical();
|
|
406 |
/*
|
|
407 |
* If the defect list is dirty, write it to disk,
|
|
408 |
* if scan_restore_defects (the default) is true.
|
|
409 |
*/
|
|
410 |
if (!EMBEDDED_SCSI && (cur_list.flags & LIST_DIRTY) &&
|
|
411 |
(scan_restore_defects)) {
|
|
412 |
cur_list.flags = 0;
|
|
413 |
write_deflist(&cur_list);
|
|
414 |
}
|
|
415 |
/*
|
|
416 |
* If the label is dirty, write it to disk.
|
|
417 |
* if scan_restore_label (the default) is true.
|
|
418 |
*/
|
|
419 |
if ((cur_flags & LABEL_DIRTY) && (scan_restore_label)) {
|
|
420 |
cur_flags &= ~LABEL_DIRTY;
|
|
421 |
(void) write_label();
|
|
422 |
}
|
|
423 |
/*
|
|
424 |
* If we dropped down to here after an error, we need to write
|
|
425 |
* the final block number to the log file for record keeping.
|
|
426 |
*/
|
|
427 |
if (log_file && scan_cur_block >= 0) {
|
|
428 |
pr_dblock(log_print, scan_cur_block);
|
|
429 |
log_print("\n");
|
|
430 |
}
|
|
431 |
fmt_print("Total of %lld defective blocks repaired.\n",
|
|
432 |
scan_blocks_fixed);
|
|
433 |
/*
|
|
434 |
* Reinitialize the logging variables so they don't get used
|
|
435 |
* when they are not really valid.
|
|
436 |
*/
|
|
437 |
scan_blocks_fixed = scan_cur_block = -1;
|
|
438 |
exit_critical();
|
|
439 |
return (error);
|
|
440 |
}
|
|
441 |
|
|
442 |
|
|
443 |
/*
|
|
444 |
* This routine is called to repair a bad block discovered
|
|
445 |
* during a scan operation. Return 0 for success, 1 for failure.
|
|
446 |
* (This has been extracted out of do_scan(), to simplify it.)
|
|
447 |
*/
|
|
448 |
static int
|
|
449 |
scan_repair(bn, mode)
|
|
450 |
diskaddr_t bn;
|
|
451 |
int mode;
|
|
452 |
{
|
|
453 |
int status;
|
|
454 |
int result = 1;
|
|
455 |
char buf[SECSIZE];
|
|
456 |
int buf_is_good;
|
|
457 |
int i;
|
|
458 |
|
|
459 |
if (cur_ops->op_repair == NULL) {
|
|
460 |
err_print("Warning: Controller does ");
|
|
461 |
err_print("not support repairing.\n\n");
|
|
462 |
return (result);
|
|
463 |
}
|
|
464 |
|
|
465 |
enter_critical();
|
|
466 |
|
|
467 |
/*
|
|
468 |
* Determine if the error appears to be hard or soft. We
|
|
469 |
* already assume there's an error. If we can get any
|
|
470 |
* good data out of the sector, write that data back
|
|
471 |
* after the repair.
|
|
472 |
*/
|
|
473 |
buf_is_good = 0;
|
|
474 |
for (i = 0; i < 5; i++) {
|
|
475 |
status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, bn, 1,
|
|
476 |
buf, F_SILENT, NULL);
|
|
477 |
if (status == 0) {
|
|
478 |
buf_is_good = 1;
|
|
479 |
break;
|
|
480 |
}
|
|
481 |
}
|
|
482 |
|
|
483 |
fmt_print("Repairing %s error on %llu (",
|
|
484 |
buf_is_good ? "soft" : "hard", bn);
|
|
485 |
pr_dblock(fmt_print, bn);
|
|
486 |
fmt_print(")...");
|
|
487 |
|
|
488 |
status = (*cur_ops->op_repair)(bn, mode);
|
|
489 |
if (status) {
|
|
490 |
/*
|
|
491 |
* If the repair failed, we note it and will return the
|
|
492 |
* failure. However, the analysis goes on.
|
|
493 |
*/
|
|
494 |
fmt_print("failed.\n\n");
|
|
495 |
} else {
|
|
496 |
/*
|
|
497 |
* The repair worked. Write the good data we could
|
|
498 |
* recover from the failed block, if possible.
|
|
499 |
* If not, zero the block. In doing so, try to
|
|
500 |
* determine if the new block appears ok.
|
|
501 |
*/
|
|
502 |
if (!buf_is_good) {
|
|
503 |
bzero(buf, SECSIZE);
|
|
504 |
fmt_print("Warning: Block %llu zero-filled.\n", bn);
|
|
505 |
} else {
|
|
506 |
fmt_print("ok.\n");
|
|
507 |
}
|
|
508 |
status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, bn,
|
|
509 |
1, buf, (F_SILENT | F_ALLERRS), NULL);
|
|
510 |
if (status == 0) {
|
|
511 |
status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, bn,
|
|
512 |
1, buf, (F_SILENT | F_ALLERRS), NULL);
|
|
513 |
}
|
|
514 |
if (status) {
|
|
515 |
fmt_print("The new block also appears defective.\n");
|
|
516 |
}
|
|
517 |
fmt_print("\n");
|
|
518 |
/*
|
|
519 |
* add the defect to the list and write the list out.
|
|
520 |
* Also, kill the working list so it will get resynced
|
|
521 |
* with the current list.
|
|
522 |
*
|
|
523 |
* For embedded scsi, we don't require a defect list.
|
|
524 |
* However, if we have one, add the defect if the
|
|
525 |
* list includes the grown list. If not, kill it
|
|
526 |
* to force a resync if we need the list later.
|
|
527 |
*/
|
|
528 |
if (EMBEDDED_SCSI) {
|
|
529 |
if (cur_list.list != NULL) {
|
|
530 |
if (cur_list.flags & LIST_PGLIST) {
|
|
531 |
add_ldef(bn, &cur_list);
|
|
532 |
} else {
|
|
533 |
kill_deflist(&cur_list);
|
|
534 |
}
|
|
535 |
}
|
|
536 |
/*
|
|
537 |
* The next "if" statement reflects the fix for
|
|
538 |
* bug id 1026096 where format keeps adding the
|
|
539 |
* same defect to the defect list.
|
|
540 |
*/
|
|
541 |
} else if (cur_ctype->ctype_flags & CF_WLIST) {
|
|
542 |
kill_deflist(&cur_list);
|
|
543 |
(*cur_ops->op_ex_cur)(&cur_list);
|
|
544 |
fmt_print("Current list updated\n");
|
|
545 |
} else {
|
|
546 |
add_ldef(bn, &cur_list);
|
|
547 |
write_deflist(&cur_list);
|
|
548 |
}
|
|
549 |
kill_deflist(&work_list);
|
|
550 |
|
|
551 |
/* Log the repair. */
|
|
552 |
scan_blocks_fixed++;
|
|
553 |
|
|
554 |
/* return ok */
|
|
555 |
result = 0;
|
|
556 |
}
|
|
557 |
|
|
558 |
exit_critical();
|
|
559 |
|
|
560 |
return (result);
|
|
561 |
}
|
|
562 |
|
|
563 |
|
|
564 |
/*
|
|
565 |
* This routine analyzes a set of sectors on the disk. It simply returns
|
|
566 |
* an error if a defect is found. It is called by do_scan().
|
|
567 |
*/
|
|
568 |
static int
|
|
569 |
analyze_blocks(flags, blkno, blkcnt, data, init, driver_flags, xfercntp)
|
|
570 |
int flags, driver_flags, blkcnt, init;
|
|
571 |
register unsigned data;
|
|
572 |
diskaddr_t blkno;
|
|
573 |
int *xfercntp;
|
|
574 |
{
|
|
575 |
int corrupt = 0;
|
|
576 |
register int status, i, nints;
|
|
577 |
register unsigned *ptr = (uint_t *)pattern_buf;
|
|
578 |
|
|
579 |
media_error = 0;
|
|
580 |
if (flags & SCAN_VERIFY) {
|
|
581 |
return (verify_blocks(flags, blkno, blkcnt, data,
|
|
582 |
driver_flags, xfercntp));
|
|
583 |
}
|
|
584 |
|
|
585 |
/*
|
|
586 |
* Initialize the pattern buffer if necessary.
|
|
587 |
*/
|
|
588 |
nints = blkcnt * SECSIZE / sizeof (int);
|
|
589 |
if ((flags & SCAN_PATTERN) && init) {
|
|
590 |
for (i = 0; i < nints; i++)
|
|
591 |
*((int *)((int *)pattern_buf + i)) = data;
|
|
592 |
}
|
|
593 |
/*
|
|
594 |
* Lock out interrupts so we can insure valid data will get
|
|
595 |
* restored. This is necessary because there are modes
|
|
596 |
* of scanning that corrupt the disk data then restore it at
|
|
597 |
* the end of the analysis.
|
|
598 |
*/
|
|
599 |
enter_critical();
|
|
600 |
/*
|
|
601 |
* If the disk data is valid, read it into the data buffer.
|
|
602 |
*/
|
|
603 |
if (flags & SCAN_VALID) {
|
|
604 |
status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, blkno,
|
|
605 |
blkcnt, (caddr_t)cur_buf, driver_flags, xfercntp);
|
|
606 |
if (status)
|
|
607 |
goto bad;
|
|
608 |
}
|
|
609 |
/*
|
|
610 |
* If we are doing pattern testing, write and read the pattern
|
|
611 |
* from the pattern buffer.
|
|
612 |
*/
|
|
613 |
if (flags & SCAN_PATTERN) {
|
|
614 |
/*
|
|
615 |
* If the disk data was valid, mark it corrupt so we know
|
|
616 |
* to restore it later.
|
|
617 |
*/
|
|
618 |
if (flags & SCAN_VALID)
|
|
619 |
corrupt++;
|
|
620 |
/*
|
|
621 |
* Only write if we're not on the read pass of SCAN_PURGE.
|
|
622 |
*/
|
|
623 |
if (!(flags & SCAN_PURGE_READ_PASS)) {
|
|
624 |
status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
|
|
625 |
blkcnt, (caddr_t)pattern_buf, driver_flags,
|
|
626 |
xfercntp);
|
|
627 |
if (status)
|
|
628 |
goto bad;
|
|
629 |
}
|
|
630 |
/*
|
|
631 |
* Only read if we are on the read pass of SCAN_PURGE, if we
|
|
632 |
* are purging.
|
|
633 |
*/
|
|
634 |
if ((!(flags & SCAN_PURGE)) || (flags & SCAN_PURGE_READ_PASS)) {
|
|
635 |
status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, blkno,
|
|
636 |
blkcnt, (caddr_t)pattern_buf, driver_flags,
|
|
637 |
xfercntp);
|
|
638 |
if (status)
|
|
639 |
goto bad;
|
|
640 |
}
|
|
641 |
}
|
|
642 |
/*
|
|
643 |
* If we are doing a data compare, make sure the pattern
|
|
644 |
* came back intact.
|
|
645 |
* Only compare if we are on the read pass of SCAN_PURGE, or
|
|
646 |
* we wrote random data instead of the expected data pattern.
|
|
647 |
*/
|
|
648 |
if ((flags & SCAN_COMPARE) || (flags & SCAN_PURGE_READ_PASS)) {
|
|
649 |
for (i = nints, ptr = (uint_t *)pattern_buf; i; i--)
|
|
650 |
if (*ptr++ != data) {
|
|
651 |
err_print("Data miscompare error (expecting ");
|
|
652 |
err_print("0x%x, got 0x%x) at ", data,
|
|
653 |
*((int *)((int *)pattern_buf +
|
|
654 |
(nints - i))));
|
|
655 |
pr_dblock(err_print, blkno);
|
|
656 |
err_print(", offset = 0x%x.\n",
|
|
657 |
(nints - i) * sizeof (int));
|
|
658 |
goto bad;
|
|
659 |
}
|
|
660 |
}
|
|
661 |
/*
|
|
662 |
* If we are supposed to write data out, do so.
|
|
663 |
*/
|
|
664 |
if (flags & SCAN_WRITE) {
|
|
665 |
status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
|
|
666 |
blkcnt, (caddr_t)cur_buf, driver_flags, xfercntp);
|
|
667 |
if (status)
|
|
668 |
goto bad;
|
|
669 |
}
|
|
670 |
exit_critical();
|
|
671 |
/*
|
|
672 |
* No errors occurred, return ok.
|
|
673 |
*/
|
|
674 |
return (0);
|
|
675 |
bad:
|
|
676 |
/*
|
|
677 |
* There was an error. If the data was corrupted, we write it
|
|
678 |
* out from the data buffer to restore it.
|
|
679 |
*/
|
|
680 |
if (corrupt) {
|
|
681 |
if ((*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
|
|
682 |
blkcnt, (caddr_t)cur_buf, F_NORMAL, xfercntp))
|
|
683 |
err_print("Warning: unable to restore original data.\n");
|
|
684 |
}
|
|
685 |
exit_critical();
|
|
686 |
/*
|
|
687 |
* Return the error.
|
|
688 |
*/
|
|
689 |
return (-1);
|
|
690 |
}
|
|
691 |
|
|
692 |
|
|
693 |
/*
|
|
694 |
* This routine analyzes a set of sectors on the disk. It simply returns
|
|
695 |
* an error if a defect is found. It is called by analyze_blocks().
|
|
696 |
* For simplicity, this is done as a separate function instead of
|
|
697 |
* making the analyze_block routine complex.
|
|
698 |
*
|
|
699 |
* This routine implements the 'verify' command. It writes the disk
|
|
700 |
* by writing unique data for each block; after the write pass, it
|
|
701 |
* reads the data and verifies for correctness. Note that the entire
|
|
702 |
* disk (or the range of disk) is fully written first and then read.
|
|
703 |
* This should eliminate any caching effect on the drives.
|
|
704 |
*/
|
|
705 |
static int
|
|
706 |
verify_blocks(int flags,
|
|
707 |
diskaddr_t blkno,
|
|
708 |
int blkcnt,
|
|
709 |
unsigned data,
|
|
710 |
int driver_flags,
|
|
711 |
int *xfercntp)
|
|
712 |
{
|
|
713 |
int status, i, nints;
|
|
714 |
unsigned *ptr = (uint_t *)pattern_buf;
|
|
715 |
|
|
716 |
nints = SECSIZE / sizeof (int);
|
|
717 |
|
|
718 |
/*
|
|
719 |
* Initialize the pattern buffer if we are in write pass.
|
|
720 |
* Use the block number itself as data, each block has unique
|
|
721 |
* buffer data that way.
|
|
722 |
*/
|
|
723 |
if (!(flags & SCAN_VERIFY_READ_PASS)) {
|
|
724 |
for (data = blkno; data < blkno + blkcnt; data++) {
|
|
725 |
for (i = 0; i < nints; i++) {
|
|
726 |
*ptr++ = data;
|
|
727 |
}
|
|
728 |
}
|
|
729 |
ptr = (uint_t *)pattern_buf;
|
|
730 |
}
|
|
731 |
|
|
732 |
/*
|
|
733 |
* Only write if we're not on the read pass of SCAN_VERIFY.
|
|
734 |
*/
|
|
735 |
if (!(flags & SCAN_VERIFY_READ_PASS)) {
|
|
736 |
status = (*cur_ops->op_rdwr)(DIR_WRITE, cur_file, blkno,
|
|
737 |
blkcnt, (caddr_t)pattern_buf, driver_flags, xfercntp);
|
|
738 |
if (status)
|
|
739 |
goto bad;
|
|
740 |
} else {
|
|
741 |
/*
|
|
742 |
* Only read if we are on the read pass of SCAN_VERIFY
|
|
743 |
*/
|
|
744 |
status = (*cur_ops->op_rdwr)(DIR_READ, cur_file, blkno,
|
|
745 |
blkcnt, (caddr_t)pattern_buf, driver_flags, xfercntp);
|
|
746 |
if (status)
|
|
747 |
goto bad;
|
|
748 |
/*
|
|
749 |
* compare and make sure the pattern came back intact.
|
|
750 |
*/
|
|
751 |
for (data = blkno; data < blkno + blkcnt; data++) {
|
|
752 |
for (i = 0; i < nints; i++) {
|
|
753 |
if (*ptr++ != data) {
|
|
754 |
ptr--;
|
|
755 |
err_print("Data miscompare error (expecting "
|
|
756 |
"0x%x, got 0x%x) at ", data, *ptr);
|
|
757 |
pr_dblock(err_print, blkno);
|
|
758 |
err_print(", offset = 0x%x.\n", (ptr -
|
|
759 |
(uint_t *)pattern_buf) * sizeof (int));
|
|
760 |
goto bad;
|
|
761 |
}
|
|
762 |
}
|
|
763 |
}
|
|
764 |
}
|
|
765 |
/*
|
|
766 |
* No errors occurred, return ok.
|
|
767 |
*/
|
|
768 |
return (0);
|
|
769 |
bad:
|
|
770 |
return (-1);
|
|
771 |
}
|
|
772 |
|
|
773 |
|
|
774 |
static int
|
|
775 |
handle_error_conditions()
|
|
776 |
{
|
|
777 |
|
|
778 |
/*
|
|
779 |
* Check if the errno is ENXIO.
|
|
780 |
*/
|
|
781 |
if (errno == ENXIO) {
|
|
782 |
fmt_print("\n\nWarning:Cannot access drive, ");
|
|
783 |
fmt_print("aborting surface analysis.\n");
|
|
784 |
return (-1);
|
|
785 |
}
|
|
786 |
/*
|
|
787 |
* check for disk errors
|
|
788 |
*/
|
|
789 |
switch (disk_error) {
|
|
790 |
case DISK_STAT_RESERVED:
|
|
791 |
case DISK_STAT_UNAVAILABLE:
|
|
792 |
fmt_print("\n\nWarning:Drive may be reserved ");
|
|
793 |
fmt_print("or has been removed, ");
|
|
794 |
fmt_print("aborting surface analysis.\n");
|
|
795 |
return (-1);
|
|
796 |
case DISK_STAT_NOTREADY:
|
|
797 |
fmt_print("\n\nWarning: Drive not ready, ");
|
|
798 |
fmt_print("aborting surface analysis.\n");
|
|
799 |
return (-1);
|
|
800 |
case DISK_STAT_DATA_PROTECT:
|
|
801 |
fmt_print("\n\nWarning: Drive is write protected, ");
|
|
802 |
fmt_print("aborting surface analysis.\n");
|
|
803 |
return (-1);
|
|
804 |
default:
|
|
805 |
break;
|
|
806 |
}
|
|
807 |
return (0);
|
|
808 |
}
|