2 * Copyright (C) 1994-1999, Index Data
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.40 2000-10-17 12:37:09 adam
8 * Fixed notification of live-updates. Fixed minor problem with mf_init
9 * where it didn't handle shadow area file names correctly.
11 * Revision 1.39 2000/05/05 13:48:03 adam
12 * Fixed locking for metafiles.
14 * Revision 1.38 2000/03/20 19:08:35 adam
15 * Added remote record import using Z39.50 extended services and Segment
18 * Revision 1.37 2000/03/15 15:00:30 adam
19 * First work on threaded version.
21 * Revision 1.36 1999/12/08 15:03:11 adam
22 * Implemented bf_reset.
24 * Revision 1.35 1999/10/14 14:33:50 adam
25 * Added truncation 5=106.
27 * Revision 1.34 1999/05/26 07:49:12 adam
30 * Revision 1.33 1999/05/12 13:08:06 adam
31 * First version of ISAMS.
33 * Revision 1.32 1999/04/28 14:53:07 adam
34 * Fixed stupid bug regarding split-files.
36 * Revision 1.31 1999/02/18 12:49:33 adam
37 * Changed file naming scheme for register files as well as record
40 * Revision 1.30 1999/02/02 14:50:02 adam
41 * Updated WIN32 code specific sections. Changed header.
43 * Revision 1.29 1998/05/27 14:28:34 adam
44 * Fixed bug in mf_write. 'Cap off' byte written at wrong offset.
46 * Revision 1.28 1998/05/20 10:00:35 adam
47 * Fixed register spec so that colon isn't treated as size separator
48 * unless followed by [0-9+-] in order to allow DOS drive specifications.
50 * Revision 1.27 1998/02/10 11:55:07 adam
53 * Revision 1.26 1997/10/27 14:25:38 adam
56 * Revision 1.25 1997/09/18 08:59:16 adam
57 * Extra generic handle for the character mapping routines.
59 * Revision 1.24 1997/09/17 12:19:06 adam
60 * Zebra version corresponds to YAZ version 1.4.
61 * Changed Zebra server so that it doesn't depend on global common_resource.
63 * Revision 1.23 1997/09/09 13:37:53 adam
64 * Partial port to WIN95/NT.
66 * Revision 1.22 1997/09/04 13:56:39 adam
67 * Added O_BINARY to open calls.
69 * Revision 1.21 1996/10/29 13:56:18 adam
70 * Include of zebrautl.h instead of alexutil.h.
72 * Revision 1.20 1996/05/14 12:10:16 quinn
75 * Revision 1.19 1996/05/01 07:16:30 quinn
78 * Revision 1.18 1996/04/09 06:47:30 adam
79 * Function scan_areadef doesn't use sscanf (%n fails on this Linux).
81 * Revision 1.17 1996/03/20 13:29:11 quinn
84 * Revision 1.16 1995/12/12 15:57:57 adam
85 * Implemented mf_unlink. cf_unlink uses mf_unlink.
87 * Revision 1.15 1995/12/08 16:21:14 adam
88 * Work on commit/update.
90 * Revision 1.14 1995/12/05 13:12:37 quinn
93 * Revision 1.13 1995/11/30 17:00:50 adam
94 * Several bug fixes. Commit system runs now.
96 * Revision 1.12 1995/11/24 17:26:11 quinn
97 * Mostly about making some ISAM stuff in the config file optional.
99 * Revision 1.11 1995/11/13 09:32:43 quinn
102 * Revision 1.10 1995/09/04 12:33:22 adam
103 * Various cleanup. YAZ util used instead.
105 * Revision 1.9 1994/11/04 14:26:39 quinn
108 * Revision 1.8 1994/10/05 16:56:42 quinn
111 * Revision 1.7 1994/09/19 14:12:37 quinn
114 * Revision 1.6 1994/09/14 13:10:15 quinn
115 * Corrected some bugs in the init-phase
117 * Revision 1.5 1994/09/12 08:01:51 quinn
120 * Revision 1.4 1994/09/01 14:51:07 quinn
121 * Allowed mf_write to write beyond eof+1.
123 * Revision 1.3 1994/08/24 09:37:17 quinn
124 * Changed reaction to read return values.
126 * Revision 1.2 1994/08/23 14:50:48 quinn
129 * Revision 1.1 1994/08/23 14:41:33 quinn
130 * First functional version.
136 * TODO: The size estimates in init may not be accurate due to
137 * only partially written final blocks.
140 #include <sys/types.h>
154 #include <zebra-lock.h>
155 #include <zebrautl.h>
158 static int scan_areadef(MFile_area ma, const char *name, const char *ad)
161 * If no definition is given, use current directory, unlimited.
163 char dirname[FILENAME_MAX+1];
164 mf_dir **dp = &ma->dirs, *dir = *dp;
170 const char *ad0 = ad;
171 int i = 0, fact = 1, multi;
174 while (*ad == ' ' || *ad == '\t')
180 if (*ad == ':' && strchr ("+-0123456789", ad[1]))
182 if (i < FILENAME_MAX)
189 logf (LOG_WARN, "Missing colon after path: %s", ad0);
194 logf (LOG_WARN, "Empty path: %s", ad0);
197 while (*ad == ' ' || *ad == '\t')
207 if (*ad < '0' || *ad > '9')
209 logf (LOG_FATAL, "Missing size after path: %s", ad0);
213 while (*ad >= '0' && *ad <= '9')
214 size = size*10 + (*ad++ - '0');
217 case 'B': case 'b': multi = 1; break;
218 case 'K': case 'k': multi = 1024; break;
219 case 'M': case 'm': multi = 1048576; break;
220 case 'G': case 'g': multi = 1073741824; break;
222 logf (LOG_FATAL, "Missing unit: %s", ad0);
225 logf (LOG_FATAL, "Illegal unit: %c in %s", *ad, ad0);
229 *dp = dir = (mf_dir *) xmalloc(sizeof(mf_dir));
231 strcpy(dir->name, dirname);
232 dir->max_bytes = dir->avail_bytes = fact * size * multi;
238 static int file_position(MFile mf, int pos, int offset)
240 int off = 0, c = mf->cur_file, ps;
242 if ((c > 0 && pos <= mf->files[c-1].top) ||
243 (c < mf->no_files -1 && pos > mf->files[c].top))
246 while (c + 1 < mf->no_files && mf->files[c].top < pos)
248 off += mf->files[c].blocks;
251 assert(c < mf->no_files);
254 off = c ? (mf->files[c-1].top + 1) : 0;
255 if (mf->files[c].fd < 0 && (mf->files[c].fd = open(mf->files[c].path,
256 mf->wr ? (O_BINARY|O_RDWR|O_CREAT) : (O_BINARY|O_RDONLY), 0666)) < 0)
258 if (!mf->wr && errno == ENOENT && off == 0)
260 logf (LOG_WARN|LOG_ERRNO, "Failed to open %s", mf->files[c].path);
263 if (lseek(mf->files[c].fd, (ps = pos - off) * mf->blocksize + offset,
266 logf (LOG_WARN|LOG_ERRNO, "Failed to seek in %s", mf->files[c].path);
273 static int cmp_part_file(const void *p1, const void *p2)
275 return ((part_file *)p1)->number - ((part_file *)p2)->number;
279 * Create a new area, cotaining metafiles in directories.
280 * Find the part-files in each directory, and inventory the existing metafiles.
282 MFile_area mf_init(const char *name, const char *spec)
284 MFile_area ma = (MFile_area) xmalloc(sizeof(*ma));
287 part_file *part_f = 0;
291 char metaname[FILENAME_MAX+1], tmpnam[FILENAME_MAX+1];
293 logf (LOG_DEBUG, "mf_init(%s)", name);
294 strcpy(ma->name, name);
297 if (scan_areadef(ma, name, spec) < 0)
299 logf (LOG_WARN, "Failed to access description of '%s'", name);
302 /* look at each directory */
303 for (dirp = ma->dirs; dirp; dirp = dirp->next)
305 if (!(dd = opendir(dirp->name)))
307 logf (LOG_WARN|LOG_ERRNO, "Failed to open directory %s",
311 /* look at each file */
312 while ((dent = readdir(dd)))
314 int len = strlen(dent->d_name);
315 const char *cp = strrchr (dent->d_name, '-');
316 if (strchr (".-", *dent->d_name))
318 if (len < 5 || !cp || strcmp (dent->d_name + len - 3, ".mf"))
320 logf (LOG_WARN, "bf: %s is not a part-file.", dent->d_name);
324 memcpy (metaname, dent->d_name, cp - dent->d_name);
325 metaname[ cp - dent->d_name] = '\0';
327 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
330 if (!strcmp(meta_f->name, metaname))
332 part_f = &meta_f->files[meta_f->no_files++];
339 meta_f = (meta_file *) xmalloc(sizeof(*meta_f));
340 zebra_mutex_init (&meta_f->mutex);
342 meta_f->next = ma->mfiles;
344 meta_f->cur_file = -1;
346 strcpy(meta_f->name, metaname);
347 part_f = &meta_f->files[0];
348 meta_f->no_files = 1;
350 part_f->number = number;
353 sprintf(tmpnam, "%s/%s", dirp->name, dent->d_name);
354 part_f->path = xstrdup(tmpnam);
356 if ((fd = open(part_f->path, O_BINARY|O_RDONLY)) < 0)
358 logf (LOG_FATAL|LOG_ERRNO, "Failed to access %s",
362 if ((part_f->bytes = lseek(fd, 0, SEEK_END)) < 0)
364 logf (LOG_FATAL|LOG_ERRNO, "Failed to seek in %s",
369 if (dirp->max_bytes >= 0)
370 dirp->avail_bytes -= part_f->bytes;
374 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
376 logf (LOG_DEBUG, "mf_init: %s consists of %d part(s)", meta_f->name,
378 qsort(meta_f->files, meta_f->no_files, sizeof(part_file),
384 void mf_destroy(MFile_area ma)
402 meta_file *m = meta_f;
404 for (i = 0; i<m->no_files; i++)
406 xfree (m->files[i].path);
408 zebra_mutex_destroy (&meta_f->mutex);
409 meta_f = meta_f->next;
415 void mf_reset(MFile_area ma)
425 meta_file *m = meta_f;
428 for (i = 0; i<m->no_files; i++)
430 unlink (m->files[i].path);
431 xfree (m->files[i].path);
433 meta_f = meta_f->next;
441 * If !ma, Use MF_DEFAULT_AREA.
443 MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
447 char tmp[FILENAME_MAX+1];
450 logf(LOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
451 wflag ? "RW" : "RDONLY");
453 for (mnew = ma->mfiles; mnew; mnew = mnew->next)
454 if (!strcmp(name, mnew->name))
463 mnew = (meta_file *) xmalloc(sizeof(*mnew));
464 strcpy(mnew->name, name);
465 /* allocate one, empty file */
466 zebra_mutex_init (&mnew->mutex);
468 mnew->files[0].bytes = 0;
469 mnew->files[0].blocks = 0;
470 mnew->files[0].top = -1;
471 mnew->files[0].number = 0;
472 mnew->files[0].fd = -1;
473 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
474 for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
475 mnew->min_bytes_creat; dp = dp->next);
478 logf (LOG_FATAL, "Insufficient space for new mfile.");
481 mnew->files[0].dir = dp;
482 sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
483 mnew->files[0].path = xstrdup(tmp);
485 mnew->next = ma->mfiles;
490 for (i = 0; i < mnew->no_files; i++)
492 if (mnew->files[i].bytes % block_size)
493 mnew->files[i].bytes += block_size - mnew->files[i].bytes %
495 mnew->files[i].blocks = mnew->files[i].bytes / block_size;
499 mnew->blocksize = block_size;
500 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
505 for (i = 0; i < mnew->no_files; i++)
507 mnew->files[i].blocks = mnew->files[i].bytes / mnew->blocksize;
508 if (i == mnew->no_files - 1)
509 mnew->files[i].top = -1;
512 i ? (mnew->files[i-1].top + mnew->files[i].blocks)
513 : (mnew->files[i].blocks - 1);
521 int mf_close(MFile mf)
525 logf (LOG_DEBUG, "mf_close(%s)", mf->name);
527 for (i = 0; i < mf->no_files; i++)
528 if (mf->files[i].fd >= 0)
530 close(mf->files[i].fd);
531 mf->files[i].fd = -1;
538 * Read one block from a metafile. Interface mirrors bfile.
540 int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
544 zebra_mutex_lock (&mf->mutex);
545 if ((rd = file_position(mf, no, offset)) < 0)
549 zebra_mutex_unlock (&mf->mutex);
555 toread = nbytes ? nbytes : mf->blocksize;
556 if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
558 logf (LOG_FATAL|LOG_ERRNO, "mf_read: Read failed (%s)",
559 mf->files[mf->cur_file].path);
562 zebra_mutex_unlock (&mf->mutex);
572 int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
574 int ps, nblocks, towrite;
576 char tmp[FILENAME_MAX+1];
577 unsigned char dummych = '\xff';
579 zebra_mutex_lock (&mf->mutex);
580 if ((ps = file_position(mf, no, offset)) < 0)
582 /* file needs to grow */
583 while (ps >= mf->files[mf->cur_file].blocks)
585 /* file overflow - allocate new file */
586 if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
587 (ps - mf->files[mf->cur_file].blocks + 1) * mf->blocksize >
588 mf->files[mf->cur_file].dir->avail_bytes)
591 if ((nblocks = mf->files[mf->cur_file].dir->avail_bytes /
594 logf (LOG_DEBUG, "Capping off file %s at pos %d",
595 mf->files[mf->cur_file].path, nblocks);
596 if ((ps = file_position(mf,
597 (mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
598 mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
600 logf (LOG_DEBUG, "ps = %d", ps);
601 if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
603 logf (LOG_ERRNO|LOG_FATAL, "write dummy");
606 mf->files[mf->cur_file].blocks += nblocks;
607 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
608 mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
612 logf (LOG_DEBUG, "Creating new file.");
613 for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
614 dp->avail_bytes < mf->min_bytes_creat; dp = dp->next);
617 logf (LOG_FATAL, "Cannot allocate more space for %s",
621 mf->files[mf->cur_file].top = (mf->cur_file ?
622 mf->files[mf->cur_file-1].top : -1) +
623 mf->files[mf->cur_file].blocks;
624 mf->files[++(mf->cur_file)].top = -1;
625 mf->files[mf->cur_file].dir = dp;
626 mf->files[mf->cur_file].number =
627 mf->files[mf->cur_file-1].number + 1;
628 mf->files[mf->cur_file].blocks =
629 mf->files[mf->cur_file].bytes = 0;
630 mf->files[mf->cur_file].fd = -1;
631 sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
632 mf->files[mf->cur_file].number);
633 mf->files[mf->cur_file].path = xstrdup(tmp);
635 /* open new file and position at beginning */
636 if ((ps = file_position(mf, no, offset)) < 0)
641 nblocks = ps - mf->files[mf->cur_file].blocks + 1;
642 mf->files[mf->cur_file].blocks += nblocks;
643 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
644 if (mf->files[mf->cur_file].dir->max_bytes >= 0)
645 mf->files[mf->cur_file].dir->avail_bytes -=
646 nblocks * mf->blocksize;
649 towrite = nbytes ? nbytes : mf->blocksize;
650 if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
652 logf (LOG_FATAL|LOG_ERRNO, "Write failed for file %s part %d",
653 mf->name, mf->cur_file);
656 zebra_mutex_unlock (&mf->mutex);
661 * Destroy a metafile, unlinking component files. File must be open.
663 int mf_unlink(MFile mf)
667 for (i = 0; i < mf->no_files; i++)
668 unlink (mf->files[i].path);
673 * Unlink the file by name, rather than MFile-handle. File should be closed.
675 int mf_unlink_name(MFile_area ma, const char *name)