2 * Copyright (C) 1994-1999, Index Data
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.39 2000-05-05 13:48:03 adam
8 * Fixed locking for metafiles.
10 * Revision 1.38 2000/03/20 19:08:35 adam
11 * Added remote record import using Z39.50 extended services and Segment
14 * Revision 1.37 2000/03/15 15:00:30 adam
15 * First work on threaded version.
17 * Revision 1.36 1999/12/08 15:03:11 adam
18 * Implemented bf_reset.
20 * Revision 1.35 1999/10/14 14:33:50 adam
21 * Added truncation 5=106.
23 * Revision 1.34 1999/05/26 07:49:12 adam
26 * Revision 1.33 1999/05/12 13:08:06 adam
27 * First version of ISAMS.
29 * Revision 1.32 1999/04/28 14:53:07 adam
30 * Fixed stupid bug regarding split-files.
32 * Revision 1.31 1999/02/18 12:49:33 adam
33 * Changed file naming scheme for register files as well as record
36 * Revision 1.30 1999/02/02 14:50:02 adam
37 * Updated WIN32 code specific sections. Changed header.
39 * Revision 1.29 1998/05/27 14:28:34 adam
40 * Fixed bug in mf_write. 'Cap off' byte written at wrong offset.
42 * Revision 1.28 1998/05/20 10:00:35 adam
43 * Fixed register spec so that colon isn't treated as size separator
44 * unless followed by [0-9+-] in order to allow DOS drive specifications.
46 * Revision 1.27 1998/02/10 11:55:07 adam
49 * Revision 1.26 1997/10/27 14:25:38 adam
52 * Revision 1.25 1997/09/18 08:59:16 adam
53 * Extra generic handle for the character mapping routines.
55 * Revision 1.24 1997/09/17 12:19:06 adam
56 * Zebra version corresponds to YAZ version 1.4.
57 * Changed Zebra server so that it doesn't depend on global common_resource.
59 * Revision 1.23 1997/09/09 13:37:53 adam
60 * Partial port to WIN95/NT.
62 * Revision 1.22 1997/09/04 13:56:39 adam
63 * Added O_BINARY to open calls.
65 * Revision 1.21 1996/10/29 13:56:18 adam
66 * Include of zebrautl.h instead of alexutil.h.
68 * Revision 1.20 1996/05/14 12:10:16 quinn
71 * Revision 1.19 1996/05/01 07:16:30 quinn
74 * Revision 1.18 1996/04/09 06:47:30 adam
75 * Function scan_areadef doesn't use sscanf (%n fails on this Linux).
77 * Revision 1.17 1996/03/20 13:29:11 quinn
80 * Revision 1.16 1995/12/12 15:57:57 adam
81 * Implemented mf_unlink. cf_unlink uses mf_unlink.
83 * Revision 1.15 1995/12/08 16:21:14 adam
84 * Work on commit/update.
86 * Revision 1.14 1995/12/05 13:12:37 quinn
89 * Revision 1.13 1995/11/30 17:00:50 adam
90 * Several bug fixes. Commit system runs now.
92 * Revision 1.12 1995/11/24 17:26:11 quinn
93 * Mostly about making some ISAM stuff in the config file optional.
95 * Revision 1.11 1995/11/13 09:32:43 quinn
98 * Revision 1.10 1995/09/04 12:33:22 adam
99 * Various cleanup. YAZ util used instead.
101 * Revision 1.9 1994/11/04 14:26:39 quinn
104 * Revision 1.8 1994/10/05 16:56:42 quinn
107 * Revision 1.7 1994/09/19 14:12:37 quinn
110 * Revision 1.6 1994/09/14 13:10:15 quinn
111 * Corrected some bugs in the init-phase
113 * Revision 1.5 1994/09/12 08:01:51 quinn
116 * Revision 1.4 1994/09/01 14:51:07 quinn
117 * Allowed mf_write to write beyond eof+1.
119 * Revision 1.3 1994/08/24 09:37:17 quinn
120 * Changed reaction to read return values.
122 * Revision 1.2 1994/08/23 14:50:48 quinn
125 * Revision 1.1 1994/08/23 14:41:33 quinn
126 * First functional version.
132 * TODO: The size estimates in init may not be accurate due to
133 * only partially written final blocks.
136 #include <sys/types.h>
150 #include <zebra-lock.h>
151 #include <zebrautl.h>
154 static int scan_areadef(MFile_area ma, const char *name, const char *ad)
157 * If no definition is given, use current directory, unlimited.
159 char dirname[FILENAME_MAX+1];
160 mf_dir **dp = &ma->dirs, *dir = *dp;
166 const char *ad0 = ad;
167 int i = 0, fact = 1, multi;
170 while (*ad == ' ' || *ad == '\t')
176 if (*ad == ':' && strchr ("+-0123456789", ad[1]))
178 if (i < FILENAME_MAX)
185 logf (LOG_WARN, "Missing colon after path: %s", ad0);
190 logf (LOG_WARN, "Empty path: %s", ad0);
193 while (*ad == ' ' || *ad == '\t')
203 if (*ad < '0' || *ad > '9')
205 logf (LOG_FATAL, "Missing size after path: %s", ad0);
209 while (*ad >= '0' && *ad <= '9')
210 size = size*10 + (*ad++ - '0');
213 case 'B': case 'b': multi = 1; break;
214 case 'K': case 'k': multi = 1024; break;
215 case 'M': case 'm': multi = 1048576; break;
216 case 'G': case 'g': multi = 1073741824; break;
218 logf (LOG_FATAL, "Missing unit: %s", ad0);
221 logf (LOG_FATAL, "Illegal unit: %c in %s", *ad, ad0);
225 *dp = dir = (mf_dir *) xmalloc(sizeof(mf_dir));
227 strcpy(dir->name, dirname);
228 dir->max_bytes = dir->avail_bytes = fact * size * multi;
234 static int file_position(MFile mf, int pos, int offset)
236 int off = 0, c = mf->cur_file, ps;
238 if ((c > 0 && pos <= mf->files[c-1].top) ||
239 (c < mf->no_files -1 && pos > mf->files[c].top))
242 while (c + 1 < mf->no_files && mf->files[c].top < pos)
244 off += mf->files[c].blocks;
247 assert(c < mf->no_files);
250 off = c ? (mf->files[c-1].top + 1) : 0;
251 if (mf->files[c].fd < 0 && (mf->files[c].fd = open(mf->files[c].path,
252 mf->wr ? (O_BINARY|O_RDWR|O_CREAT) : (O_BINARY|O_RDONLY), 0666)) < 0)
254 if (!mf->wr && errno == ENOENT && off == 0)
256 logf (LOG_WARN|LOG_ERRNO, "Failed to open %s", mf->files[c].path);
259 if (lseek(mf->files[c].fd, (ps = pos - off) * mf->blocksize + offset,
262 logf (LOG_WARN|LOG_ERRNO, "Failed to seek in %s", mf->files[c].path);
269 static int cmp_part_file(const void *p1, const void *p2)
271 return ((part_file *)p1)->number - ((part_file *)p2)->number;
275 * Create a new area, cotaining metafiles in directories.
276 * Find the part-files in each directory, and inventory the existing metafiles.
278 MFile_area mf_init(const char *name, const char *spec)
280 MFile_area ma = (MFile_area) xmalloc(sizeof(*ma));
283 part_file *part_f = 0;
287 char metaname[FILENAME_MAX+1], tmpnam[FILENAME_MAX+1];
289 logf (LOG_DEBUG, "mf_init(%s)", name);
290 strcpy(ma->name, name);
293 if (scan_areadef(ma, name, spec) < 0)
295 logf (LOG_WARN, "Failed to access description of '%s'", name);
298 /* look at each directory */
299 for (dirp = ma->dirs; dirp; dirp = dirp->next)
301 if (!(dd = opendir(dirp->name)))
303 logf (LOG_WARN|LOG_ERRNO, "Failed to open directory %s",
307 /* look at each file */
308 while ((dent = readdir(dd)))
310 if (*dent->d_name == '.')
312 if (sscanf(dent->d_name, "%[^-]-%d.mf", metaname, &number) != 2)
314 logf (LOG_DEBUG, "bf: %s is not a part-file.", dent->d_name);
317 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
320 if (!strcmp(meta_f->name, metaname))
322 part_f = &meta_f->files[meta_f->no_files++];
329 meta_f = (meta_file *) xmalloc(sizeof(*meta_f));
330 zebra_mutex_init (&meta_f->mutex);
332 meta_f->next = ma->mfiles;
334 meta_f->cur_file = -1;
336 strcpy(meta_f->name, metaname);
337 part_f = &meta_f->files[0];
338 meta_f->no_files = 1;
340 part_f->number = number;
343 sprintf(tmpnam, "%s/%s", dirp->name, dent->d_name);
344 part_f->path = xstrdup(tmpnam);
346 if ((fd = open(part_f->path, O_BINARY|O_RDONLY)) < 0)
348 logf (LOG_FATAL|LOG_ERRNO, "Failed to access %s",
352 if ((part_f->bytes = lseek(fd, 0, SEEK_END)) < 0)
354 logf (LOG_FATAL|LOG_ERRNO, "Failed to seek in %s",
359 if (dirp->max_bytes >= 0)
360 dirp->avail_bytes -= part_f->bytes;
364 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
366 logf (LOG_DEBUG, "mf_init: %s consists of %d part(s)", meta_f->name,
368 qsort(meta_f->files, meta_f->no_files, sizeof(part_file),
374 void mf_destroy(MFile_area ma)
392 meta_file *m = meta_f;
394 for (i = 0; i<m->no_files; i++)
396 xfree (m->files[i].path);
398 zebra_mutex_destroy (&meta_f->mutex);
399 meta_f = meta_f->next;
405 void mf_reset(MFile_area ma)
415 meta_file *m = meta_f;
418 for (i = 0; i<m->no_files; i++)
420 unlink (m->files[i].path);
421 xfree (m->files[i].path);
423 meta_f = meta_f->next;
431 * If !ma, Use MF_DEFAULT_AREA.
433 MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
437 char tmp[FILENAME_MAX+1];
440 logf(LOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
441 wflag ? "RW" : "RDONLY");
443 for (mnew = ma->mfiles; mnew; mnew = mnew->next)
444 if (!strcmp(name, mnew->name))
453 mnew = (meta_file *) xmalloc(sizeof(*mnew));
454 strcpy(mnew->name, name);
455 /* allocate one, empty file */
456 zebra_mutex_init (&mnew->mutex);
458 mnew->files[0].bytes = 0;
459 mnew->files[0].blocks = 0;
460 mnew->files[0].top = -1;
461 mnew->files[0].number = 0;
462 mnew->files[0].fd = -1;
463 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
464 for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
465 mnew->min_bytes_creat; dp = dp->next);
468 logf (LOG_FATAL, "Insufficient space for new mfile.");
471 mnew->files[0].dir = dp;
472 sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
473 mnew->files[0].path = xstrdup(tmp);
475 mnew->next = ma->mfiles;
480 for (i = 0; i < mnew->no_files; i++)
482 if (mnew->files[i].bytes % block_size)
483 mnew->files[i].bytes += block_size - mnew->files[i].bytes %
485 mnew->files[i].blocks = mnew->files[i].bytes / block_size;
489 mnew->blocksize = block_size;
490 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
495 for (i = 0; i < mnew->no_files; i++)
497 mnew->files[i].blocks = mnew->files[i].bytes / mnew->blocksize;
498 if (i == mnew->no_files - 1)
499 mnew->files[i].top = -1;
502 i ? (mnew->files[i-1].top + mnew->files[i].blocks)
503 : (mnew->files[i].blocks - 1);
511 int mf_close(MFile mf)
515 logf (LOG_DEBUG, "mf_close(%s)", mf->name);
517 for (i = 0; i < mf->no_files; i++)
518 if (mf->files[i].fd >= 0)
520 close(mf->files[i].fd);
521 mf->files[i].fd = -1;
528 * Read one block from a metafile. Interface mirrors bfile.
530 int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
534 zebra_mutex_lock (&mf->mutex);
535 if ((rd = file_position(mf, no, offset)) < 0)
539 zebra_mutex_unlock (&mf->mutex);
545 toread = nbytes ? nbytes : mf->blocksize;
546 if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
548 logf (LOG_FATAL|LOG_ERRNO, "mf_read: Read failed (%s)",
549 mf->files[mf->cur_file].path);
552 zebra_mutex_unlock (&mf->mutex);
562 int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
564 int ps, nblocks, towrite;
566 char tmp[FILENAME_MAX+1];
567 unsigned char dummych = '\xff';
569 zebra_mutex_lock (&mf->mutex);
570 if ((ps = file_position(mf, no, offset)) < 0)
572 /* file needs to grow */
573 while (ps >= mf->files[mf->cur_file].blocks)
575 /* file overflow - allocate new file */
576 if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
577 (ps - mf->files[mf->cur_file].blocks + 1) * mf->blocksize >
578 mf->files[mf->cur_file].dir->avail_bytes)
581 if ((nblocks = mf->files[mf->cur_file].dir->avail_bytes /
584 logf (LOG_DEBUG, "Capping off file %s at pos %d",
585 mf->files[mf->cur_file].path, nblocks);
586 if ((ps = file_position(mf,
587 (mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
588 mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
590 logf (LOG_DEBUG, "ps = %d", ps);
591 if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
593 logf (LOG_ERRNO|LOG_FATAL, "write dummy");
596 mf->files[mf->cur_file].blocks += nblocks;
597 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
598 mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
602 logf (LOG_DEBUG, "Creating new file.");
603 for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
604 dp->avail_bytes < mf->min_bytes_creat; dp = dp->next);
607 logf (LOG_FATAL, "Cannot allocate more space for %s",
611 mf->files[mf->cur_file].top = (mf->cur_file ?
612 mf->files[mf->cur_file-1].top : -1) +
613 mf->files[mf->cur_file].blocks;
614 mf->files[++(mf->cur_file)].top = -1;
615 mf->files[mf->cur_file].dir = dp;
616 mf->files[mf->cur_file].number =
617 mf->files[mf->cur_file-1].number + 1;
618 mf->files[mf->cur_file].blocks =
619 mf->files[mf->cur_file].bytes = 0;
620 mf->files[mf->cur_file].fd = -1;
621 sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
622 mf->files[mf->cur_file].number);
623 mf->files[mf->cur_file].path = xstrdup(tmp);
625 /* open new file and position at beginning */
626 if ((ps = file_position(mf, no, offset)) < 0)
631 nblocks = ps - mf->files[mf->cur_file].blocks + 1;
632 mf->files[mf->cur_file].blocks += nblocks;
633 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
634 if (mf->files[mf->cur_file].dir->max_bytes >= 0)
635 mf->files[mf->cur_file].dir->avail_bytes -=
636 nblocks * mf->blocksize;
639 towrite = nbytes ? nbytes : mf->blocksize;
640 if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
642 logf (LOG_FATAL|LOG_ERRNO, "Write failed for file %s part %d",
643 mf->name, mf->cur_file);
646 zebra_mutex_unlock (&mf->mutex);
651 * Destroy a metafile, unlinking component files. File must be open.
653 int mf_unlink(MFile mf)
657 for (i = 0; i < mf->no_files; i++)
658 unlink (mf->files[i].path);
663 * Unlink the file by name, rather than MFile-handle. File should be closed.
665 int mf_unlink_name(MFile_area ma, const char *name)