1 /* $Id: mfile.c,v 1.51 2003-03-25 23:47:23 adam Exp $
2 Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003
5 This file is part of the Zebra server.
7 Zebra is free software; you can redistribute it and/or modify it under
8 the terms of the GNU General Public License as published by the Free
9 Software Foundation; either version 2, or (at your option) any later
12 Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
13 WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17 You should have received a copy of the GNU General Public License
18 along with Zebra; see the file LICENSE.zebra. If not, write to the
19 Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
27 * TODO: The size estimates in init may not be accurate due to
28 * only partially written final blocks.
31 #include <sys/types.h>
45 #include <zebra-lock.h>
49 static int scan_areadef(MFile_area ma, const char *ad, const char *base)
52 * If no definition is given, use current directory, unlimited.
54 char dirname[FILENAME_MAX+1];
55 mf_dir **dp = &ma->dirs, *dir = *dp;
62 int i = 0, fact = 1, multi;
65 while (*ad == ' ' || *ad == '\t')
69 if (!yaz_is_abspath(ad) && base)
71 strcpy (dirname, base);
77 if (*ad == ':' && strchr ("+-0123456789", ad[1]))
86 logf (LOG_WARN, "Missing colon after path: %s", ad0);
91 logf (LOG_WARN, "Empty path: %s", ad0);
94 while (*ad == ' ' || *ad == '\t')
104 if (*ad < '0' || *ad > '9')
106 logf (LOG_FATAL, "Missing size after path: %s", ad0);
110 while (*ad >= '0' && *ad <= '9')
111 size = size*10 + (*ad++ - '0');
114 case 'B': case 'b': multi = 1; break;
115 case 'K': case 'k': multi = 1024; break;
116 case 'M': case 'm': multi = 1048576; break;
117 case 'G': case 'g': multi = 1073741824; break;
119 logf (LOG_FATAL, "Missing unit: %s", ad0);
122 logf (LOG_FATAL, "Illegal unit: %c in %s", *ad, ad0);
126 *dp = dir = (mf_dir *) xmalloc(sizeof(mf_dir));
128 strcpy(dir->name, dirname);
129 dir->max_bytes = dir->avail_bytes = fact * size * multi;
135 static int file_position(MFile mf, int pos, int offset)
137 int off = 0, c = mf->cur_file, ps;
139 if ((c > 0 && pos <= mf->files[c-1].top) ||
140 (c < mf->no_files -1 && pos > mf->files[c].top))
143 while (c + 1 < mf->no_files && mf->files[c].top < pos)
145 off += mf->files[c].blocks;
148 assert(c < mf->no_files);
151 off = c ? (mf->files[c-1].top + 1) : 0;
152 if (mf->files[c].fd < 0)
154 if ((mf->files[c].fd = open(mf->files[c].path,
156 (O_BINARY|O_RDWR|O_CREAT) :
157 (O_BINARY|O_RDONLY), 0666)) < 0)
159 if (!mf->wr && errno == ENOENT && off == 0)
161 logf (LOG_WARN|LOG_ERRNO, "Failed to open %s", mf->files[c].path);
165 if (mfile_seek(mf->files[c].fd, (ps = pos - off) * mf->blocksize + offset,
168 logf (LOG_WARN|LOG_ERRNO, "Failed to seek in %s", mf->files[c].path);
175 static int cmp_part_file(const void *p1, const void *p2)
177 return ((part_file *)p1)->number - ((part_file *)p2)->number;
181 * Create a new area, cotaining metafiles in directories.
182 * Find the part-files in each directory, and inventory the existing metafiles.
184 MFile_area mf_init(const char *name, const char *spec, const char *base)
186 MFile_area ma = (MFile_area) xmalloc(sizeof(*ma));
189 part_file *part_f = 0;
193 char metaname[FILENAME_MAX+1], tmpnam[FILENAME_MAX+1];
195 logf (LOG_DEBUG, "mf_init(%s)", name);
196 strcpy(ma->name, name);
199 if (scan_areadef(ma, spec, base) < 0)
201 logf (LOG_WARN, "Failed to access description of '%s'", name);
204 /* look at each directory */
205 for (dirp = ma->dirs; dirp; dirp = dirp->next)
207 if (!(dd = opendir(dirp->name)))
209 logf (LOG_WARN|LOG_ERRNO, "Failed to open directory %s",
213 /* look at each file */
214 while ((dent = readdir(dd)))
216 int len = strlen(dent->d_name);
217 const char *cp = strrchr (dent->d_name, '-');
218 if (strchr (".-", *dent->d_name))
220 if (len < 5 || !cp || strcmp (dent->d_name + len - 3, ".mf"))
223 memcpy (metaname, dent->d_name, cp - dent->d_name);
224 metaname[ cp - dent->d_name] = '\0';
226 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
229 if (!strcmp(meta_f->name, metaname))
231 part_f = &meta_f->files[meta_f->no_files++];
238 meta_f = (meta_file *) xmalloc(sizeof(*meta_f));
239 zebra_mutex_init (&meta_f->mutex);
241 meta_f->next = ma->mfiles;
243 meta_f->cur_file = -1;
244 meta_f->unlink_flag = 0;
246 strcpy(meta_f->name, metaname);
247 part_f = &meta_f->files[0];
248 meta_f->no_files = 1;
250 part_f->number = number;
253 sprintf(tmpnam, "%s/%s", dirp->name, dent->d_name);
254 part_f->path = xstrdup(tmpnam);
256 if ((fd = open(part_f->path, O_BINARY|O_RDONLY)) < 0)
258 logf (LOG_FATAL|LOG_ERRNO, "Failed to access %s",
262 if ((part_f->bytes = mfile_seek(fd, 0, SEEK_END)) < 0)
264 logf (LOG_FATAL|LOG_ERRNO, "Failed to seek in %s",
272 if (dirp->max_bytes >= 0)
273 dirp->avail_bytes -= part_f->bytes;
277 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
279 logf (LOG_DEBUG, "mf_init: %s consists of %d part(s)", meta_f->name,
281 qsort(meta_f->files, meta_f->no_files, sizeof(part_file),
287 void mf_destroy(MFile_area ma)
305 meta_file *m = meta_f;
307 for (i = 0; i<m->no_files; i++)
309 xfree (m->files[i].path);
311 zebra_mutex_destroy (&meta_f->mutex);
312 meta_f = meta_f->next;
318 void mf_reset(MFile_area ma)
328 meta_file *m = meta_f;
331 for (i = 0; i<m->no_files; i++)
333 unlink (m->files[i].path);
334 xfree (m->files[i].path);
336 meta_f = meta_f->next;
344 * If !ma, Use MF_DEFAULT_AREA.
346 MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
350 char tmp[FILENAME_MAX+1];
353 logf(LOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
354 wflag ? "RW" : "RDONLY");
356 for (mnew = ma->mfiles; mnew; mnew = mnew->next)
357 if (!strcmp(name, mnew->name))
366 mnew = (meta_file *) xmalloc(sizeof(*mnew));
367 strcpy(mnew->name, name);
368 /* allocate one, empty file */
369 zebra_mutex_init (&mnew->mutex);
371 mnew->files[0].bytes = 0;
372 mnew->files[0].blocks = 0;
373 mnew->files[0].top = -1;
374 mnew->files[0].number = 0;
375 mnew->files[0].fd = -1;
376 mnew->unlink_flag = 0;
377 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
378 for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
379 mnew->min_bytes_creat; dp = dp->next);
382 logf (LOG_FATAL, "Insufficient space for new mfile.");
385 mnew->files[0].dir = dp;
386 sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
387 mnew->files[0].path = xstrdup(tmp);
389 mnew->next = ma->mfiles;
394 for (i = 0; i < mnew->no_files; i++)
396 if (mnew->files[i].bytes % block_size)
397 mnew->files[i].bytes += block_size - mnew->files[i].bytes %
399 mnew->files[i].blocks = (int) (mnew->files[i].bytes / block_size);
403 mnew->blocksize = block_size;
404 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
409 for (i = 0; i < mnew->no_files; i++)
411 mnew->files[i].blocks = (int)(mnew->files[i].bytes / mnew->blocksize);
412 if (i == mnew->no_files - 1)
413 mnew->files[i].top = -1;
416 i ? (mnew->files[i-1].top + mnew->files[i].blocks)
417 : (mnew->files[i].blocks - 1);
425 int mf_close(MFile mf)
429 logf (LOG_DEBUG, "mf_close(%s)", mf->name);
431 for (i = 0; i < mf->no_files; i++)
433 if (mf->files[i].fd >= 0)
436 fsync(mf->files[i].fd);
438 close(mf->files[i].fd);
439 mf->files[i].fd = -1;
442 unlink(mf->files[i].path);
449 * Read one block from a metafile. Interface mirrors bfile.
451 int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
455 zebra_mutex_lock (&mf->mutex);
456 if ((rd = file_position(mf, no, offset)) < 0)
460 zebra_mutex_unlock (&mf->mutex);
465 yaz_log (LOG_FATAL, "mf_read %s internal error", mf->name);
469 toread = nbytes ? nbytes : mf->blocksize;
470 if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
472 logf (LOG_FATAL|LOG_ERRNO, "mf_read: Read failed (%s)",
473 mf->files[mf->cur_file].path);
476 zebra_mutex_unlock (&mf->mutex);
486 int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
488 int ps, nblocks, towrite;
490 char tmp[FILENAME_MAX+1];
491 unsigned char dummych = '\xff';
493 zebra_mutex_lock (&mf->mutex);
494 if ((ps = file_position(mf, no, offset)) < 0)
496 yaz_log (LOG_FATAL, "mf_write %s internal error (1)", mf->name);
499 /* file needs to grow */
500 while (ps >= mf->files[mf->cur_file].blocks)
502 mfile_off_t needed = (ps - mf->files[mf->cur_file].blocks + 1) *
504 /* file overflow - allocate new file */
505 if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
506 needed > mf->files[mf->cur_file].dir->avail_bytes)
509 if ((nblocks = (int) (mf->files[mf->cur_file].dir->avail_bytes /
512 logf (LOG_DEBUG, "Capping off file %s at pos %d",
513 mf->files[mf->cur_file].path, nblocks);
514 if ((ps = file_position(mf,
515 (mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
516 mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
518 yaz_log (LOG_FATAL, "mf_write %s internal error (2)",
522 logf (LOG_DEBUG, "ps = %d", ps);
523 if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
525 logf (LOG_ERRNO|LOG_FATAL, "mf_write %s internal error (3)",
529 mf->files[mf->cur_file].blocks += nblocks;
530 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
531 mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
535 logf (LOG_DEBUG, "Creating new file.");
536 for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
537 dp->avail_bytes < needed; dp = dp->next);
540 logf (LOG_FATAL, "Cannot allocate more space for %s",
544 mf->files[mf->cur_file].top = (mf->cur_file ?
545 mf->files[mf->cur_file-1].top : -1) +
546 mf->files[mf->cur_file].blocks;
547 mf->files[++(mf->cur_file)].top = -1;
548 mf->files[mf->cur_file].dir = dp;
549 mf->files[mf->cur_file].number =
550 mf->files[mf->cur_file-1].number + 1;
551 mf->files[mf->cur_file].blocks = 0;
552 mf->files[mf->cur_file].bytes = 0;
553 mf->files[mf->cur_file].fd = -1;
554 sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
555 mf->files[mf->cur_file].number);
556 mf->files[mf->cur_file].path = xstrdup(tmp);
558 /* open new file and position at beginning */
559 if ((ps = file_position(mf, no, offset)) < 0)
561 yaz_log (LOG_FATAL, "mf_write %s internal error (4)",
568 nblocks = ps - mf->files[mf->cur_file].blocks + 1;
569 mf->files[mf->cur_file].blocks += nblocks;
570 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
571 if (mf->files[mf->cur_file].dir->max_bytes >= 0)
572 mf->files[mf->cur_file].dir->avail_bytes -=
573 nblocks * mf->blocksize;
576 towrite = nbytes ? nbytes : mf->blocksize;
577 if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
579 logf (LOG_FATAL|LOG_ERRNO, "Write failed for file %s part %d",
580 mf->name, mf->cur_file);
583 zebra_mutex_unlock (&mf->mutex);
588 * Destroy a metafile, unlinking component files. File must be open.
590 int mf_unlink(MFile mf)
597 for (i = 0; i<mf->no_files; i++)
598 unlink(mf->files[i].path);
604 * Unlink the file by name, rather than MFile-handle. File should be closed.
606 int mf_unlink_name(MFile_area ma, const char *name)