1 /* $Id: mfile.c,v 1.52.2.2 2006-10-04 09:07:19 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 this program; if not, write to the Free Software
19 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
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);
166 if (mfile_seek(mf->files[c].fd, ps * (mfile_off_t) mf->blocksize + offset,
169 logf (LOG_WARN|LOG_ERRNO, "Failed to seek in %s", mf->files[c].path);
170 logf(LOG_WARN, "pos=%d off=%d blocksize=%d offset=%d",
171 pos, off, mf->blocksize, offset);
178 static int cmp_part_file(const void *p1, const void *p2)
180 return ((part_file *)p1)->number - ((part_file *)p2)->number;
184 * Create a new area, cotaining metafiles in directories.
185 * Find the part-files in each directory, and inventory the existing metafiles.
187 MFile_area mf_init(const char *name, const char *spec, const char *base)
189 MFile_area ma = (MFile_area) xmalloc(sizeof(*ma));
192 part_file *part_f = 0;
196 char metaname[FILENAME_MAX+1], tmpnam[FILENAME_MAX+1];
198 logf (LOG_DEBUG, "mf_init(%s)", name);
199 strcpy(ma->name, name);
202 if (scan_areadef(ma, spec, base) < 0)
204 logf (LOG_WARN, "Failed to access description of '%s'", name);
207 /* look at each directory */
208 for (dirp = ma->dirs; dirp; dirp = dirp->next)
210 if (!(dd = opendir(dirp->name)))
212 logf (LOG_WARN|LOG_ERRNO, "Failed to open directory %s",
216 /* look at each file */
217 while ((dent = readdir(dd)))
219 int len = strlen(dent->d_name);
220 const char *cp = strrchr (dent->d_name, '-');
221 if (strchr (".-", *dent->d_name))
223 if (len < 5 || !cp || strcmp (dent->d_name + len - 3, ".mf"))
226 memcpy (metaname, dent->d_name, cp - dent->d_name);
227 metaname[ cp - dent->d_name] = '\0';
229 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
232 if (!strcmp(meta_f->name, metaname))
234 part_f = &meta_f->files[meta_f->no_files++];
241 meta_f = (meta_file *) xmalloc(sizeof(*meta_f));
242 zebra_mutex_init (&meta_f->mutex);
244 meta_f->next = ma->mfiles;
246 meta_f->cur_file = -1;
248 strcpy(meta_f->name, metaname);
249 part_f = &meta_f->files[0];
250 meta_f->no_files = 1;
252 part_f->number = number;
255 sprintf(tmpnam, "%s/%s", dirp->name, dent->d_name);
256 part_f->path = xstrdup(tmpnam);
258 if ((fd = open(part_f->path, O_BINARY|O_RDONLY)) < 0)
260 logf (LOG_FATAL|LOG_ERRNO, "Failed to access %s",
264 if ((part_f->bytes = mfile_seek(fd, 0, SEEK_END)) < 0)
266 logf (LOG_FATAL|LOG_ERRNO, "Failed to seek in %s",
274 if (dirp->max_bytes >= 0)
275 dirp->avail_bytes -= part_f->bytes;
279 for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
281 logf (LOG_DEBUG, "mf_init: %s consists of %d part(s)", meta_f->name,
283 qsort(meta_f->files, meta_f->no_files, sizeof(part_file),
289 void mf_destroy(MFile_area ma)
307 void mf_reset(MFile_area ma, int unlink_flag)
317 meta_file *m = meta_f;
319 meta_f = meta_f->next;
322 for (i = 0; i<m->no_files; i++)
325 unlink (m->files[i].path);
326 xfree (m->files[i].path);
328 zebra_mutex_destroy (&m->mutex);
337 MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
341 char tmp[FILENAME_MAX+1];
344 logf(LOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
345 wflag ? "RW" : "RDONLY");
347 for (mnew = ma->mfiles; mnew; mnew = mnew->next)
348 if (!strcmp(name, mnew->name))
357 mnew = (meta_file *) xmalloc(sizeof(*mnew));
358 strcpy(mnew->name, name);
359 /* allocate one, empty file */
360 zebra_mutex_init (&mnew->mutex);
362 mnew->files[0].bytes = 0;
363 mnew->files[0].blocks = 0;
364 mnew->files[0].top = -1;
365 mnew->files[0].number = 0;
366 mnew->files[0].fd = -1;
367 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
368 for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
369 mnew->min_bytes_creat; dp = dp->next);
372 logf (LOG_FATAL, "Insufficient space for new mfile.");
375 mnew->files[0].dir = dp;
376 sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
377 mnew->files[0].path = xstrdup(tmp);
379 mnew->next = ma->mfiles;
384 for (i = 0; i < mnew->no_files; i++)
386 if (mnew->files[i].bytes % block_size)
387 mnew->files[i].bytes += block_size - mnew->files[i].bytes %
389 mnew->files[i].blocks = (int) (mnew->files[i].bytes / block_size);
393 mnew->blocksize = block_size;
394 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
399 for (i = 0; i < mnew->no_files; i++)
401 mnew->files[i].blocks = (int)(mnew->files[i].bytes / mnew->blocksize);
402 if (i == mnew->no_files - 1)
403 mnew->files[i].top = -1;
406 i ? (mnew->files[i-1].top + mnew->files[i].blocks)
407 : (mnew->files[i].blocks - 1);
415 int mf_close(MFile mf)
419 logf (LOG_DEBUG, "mf_close(%s)", mf->name);
421 for (i = 0; i < mf->no_files; i++)
423 if (mf->files[i].fd >= 0)
426 fsync(mf->files[i].fd);
428 close(mf->files[i].fd);
429 mf->files[i].fd = -1;
437 * Read one block from a metafile. Interface mirrors bfile.
439 int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
443 zebra_mutex_lock (&mf->mutex);
444 if ((rd = file_position(mf, no, offset)) < 0)
448 zebra_mutex_unlock (&mf->mutex);
453 yaz_log (LOG_FATAL, "mf_read %s internal error", mf->name);
457 toread = nbytes ? nbytes : mf->blocksize;
458 if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
460 logf (LOG_FATAL|LOG_ERRNO, "mf_read: Read failed (%s)",
461 mf->files[mf->cur_file].path);
464 zebra_mutex_unlock (&mf->mutex);
474 int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
476 int ps, nblocks, towrite;
478 char tmp[FILENAME_MAX+1];
479 unsigned char dummych = '\xff';
481 zebra_mutex_lock (&mf->mutex);
482 if ((ps = file_position(mf, no, offset)) < 0)
484 yaz_log (LOG_FATAL, "mf_write %s internal error (1)", mf->name);
487 /* file needs to grow */
488 while (ps >= mf->files[mf->cur_file].blocks)
490 mfile_off_t needed = (ps - mf->files[mf->cur_file].blocks + 1) *
492 /* file overflow - allocate new file */
493 if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
494 needed > mf->files[mf->cur_file].dir->avail_bytes)
497 if ((nblocks = (int) (mf->files[mf->cur_file].dir->avail_bytes /
500 logf (LOG_DEBUG, "Capping off file %s at pos %d",
501 mf->files[mf->cur_file].path, nblocks);
502 if ((ps = file_position(mf,
503 (mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
504 mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
506 yaz_log (LOG_FATAL, "mf_write %s internal error (2)",
510 logf (LOG_DEBUG, "ps = %d", ps);
511 if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
513 logf (LOG_ERRNO|LOG_FATAL, "mf_write %s internal error (3)",
517 mf->files[mf->cur_file].blocks += nblocks;
518 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
519 mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
523 logf (LOG_DEBUG, "Creating new file.");
524 for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
525 dp->avail_bytes < needed; dp = dp->next);
528 logf (LOG_FATAL, "Cannot allocate more space for %s",
532 mf->files[mf->cur_file].top = (mf->cur_file ?
533 mf->files[mf->cur_file-1].top : -1) +
534 mf->files[mf->cur_file].blocks;
535 mf->files[++(mf->cur_file)].top = -1;
536 mf->files[mf->cur_file].dir = dp;
537 mf->files[mf->cur_file].number =
538 mf->files[mf->cur_file-1].number + 1;
539 mf->files[mf->cur_file].blocks = 0;
540 mf->files[mf->cur_file].bytes = 0;
541 mf->files[mf->cur_file].fd = -1;
542 sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
543 mf->files[mf->cur_file].number);
544 mf->files[mf->cur_file].path = xstrdup(tmp);
546 /* open new file and position at beginning */
547 if ((ps = file_position(mf, no, offset)) < 0)
549 yaz_log (LOG_FATAL, "mf_write %s internal error (4)",
556 nblocks = ps - mf->files[mf->cur_file].blocks + 1;
557 mf->files[mf->cur_file].blocks += nblocks;
558 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
559 if (mf->files[mf->cur_file].dir->max_bytes >= 0)
560 mf->files[mf->cur_file].dir->avail_bytes -=
561 nblocks * mf->blocksize;
564 towrite = nbytes ? nbytes : mf->blocksize;
565 if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
567 logf (LOG_FATAL|LOG_ERRNO, "Write failed for file %s part %d",
568 mf->name, mf->cur_file);
571 zebra_mutex_unlock (&mf->mutex);