1 /* $Id: mfile.c,v 1.52.2.4 2006-12-05 21:14:38 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 yaz_log(YLOG_WARN, "Missing colon after path: %s", ad0);
91 yaz_log(YLOG_WARN, "Empty path: %s", ad0);
94 while (*ad == ' ' || *ad == '\t')
104 if (*ad < '0' || *ad > '9')
106 yaz_log(YLOG_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 yaz_log(YLOG_FATAL, "Missing unit: %s", ad0);
122 yaz_log(YLOG_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 yaz_log(YLOG_WARN|YLOG_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 yaz_log(YLOG_WARN|YLOG_ERRNO, "Failed to seek in %s", mf->files[c].path);
170 yaz_log(YLOG_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 yaz_log(YLOG_DEBUG, "mf_init(%s)", name);
199 strcpy(ma->name, name);
202 if (scan_areadef(ma, spec, base) < 0)
204 yaz_log(YLOG_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 yaz_log(YLOG_WARN|YLOG_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 yaz_log(YLOG_FATAL|YLOG_ERRNO, "Failed to access %s",
264 if ((part_f->bytes = mfile_seek(fd, 0, SEEK_END)) < 0)
266 yaz_log(YLOG_FATAL|YLOG_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 yaz_log(YLOG_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)
306 void mf_reset(MFile_area ma, int unlink_flag)
316 meta_file *m = meta_f;
318 meta_f = meta_f->next;
321 for (i = 0; i<m->no_files; i++)
324 unlink (m->files[i].path);
325 xfree (m->files[i].path);
327 zebra_mutex_destroy (&m->mutex);
336 MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
340 char tmp[FILENAME_MAX+1];
343 yaz_log(YLOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
344 wflag ? "RW" : "RDONLY");
346 for (mnew = ma->mfiles; mnew; mnew = mnew->next)
347 if (!strcmp(name, mnew->name))
356 mnew = (meta_file *) xmalloc(sizeof(*mnew));
357 strcpy(mnew->name, name);
358 /* allocate one, empty file */
359 zebra_mutex_init (&mnew->mutex);
361 mnew->files[0].bytes = 0;
362 mnew->files[0].blocks = 0;
363 mnew->files[0].top = -1;
364 mnew->files[0].number = 0;
365 mnew->files[0].fd = -1;
366 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
367 for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
368 mnew->min_bytes_creat; dp = dp->next);
371 yaz_log(YLOG_FATAL, "Insufficient space for new mfile.");
374 mnew->files[0].dir = dp;
375 sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
376 mnew->files[0].path = xstrdup(tmp);
378 mnew->next = ma->mfiles;
383 for (i = 0; i < mnew->no_files; i++)
385 if (mnew->files[i].bytes % block_size)
386 mnew->files[i].bytes += block_size - mnew->files[i].bytes %
388 mnew->files[i].blocks = (int) (mnew->files[i].bytes / block_size);
392 mnew->blocksize = block_size;
393 mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
398 for (i = 0; i < mnew->no_files; i++)
400 mnew->files[i].blocks = (int)(mnew->files[i].bytes / mnew->blocksize);
401 if (i == mnew->no_files - 1)
402 mnew->files[i].top = -1;
405 i ? (mnew->files[i-1].top + mnew->files[i].blocks)
406 : (mnew->files[i].blocks - 1);
414 int mf_close(MFile mf)
418 yaz_log(YLOG_DEBUG, "mf_close(%s)", mf->name);
420 for (i = 0; i < mf->no_files; i++)
422 if (mf->files[i].fd >= 0)
425 fsync(mf->files[i].fd);
427 close(mf->files[i].fd);
428 mf->files[i].fd = -1;
436 * Read one block from a metafile. Interface mirrors bfile.
438 int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
442 zebra_mutex_lock (&mf->mutex);
443 if ((rd = file_position(mf, no, offset)) < 0)
447 zebra_mutex_unlock (&mf->mutex);
452 yaz_log(YLOG_FATAL, "mf_read %s internal error", mf->name);
456 toread = nbytes ? nbytes : mf->blocksize;
457 if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
459 yaz_log(YLOG_FATAL|YLOG_ERRNO, "mf_read: Read failed (%s)",
460 mf->files[mf->cur_file].path);
463 zebra_mutex_unlock (&mf->mutex);
473 int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
475 int ps, nblocks, towrite;
477 char tmp[FILENAME_MAX+1];
478 unsigned char dummych = '\xff';
480 zebra_mutex_lock (&mf->mutex);
481 if ((ps = file_position(mf, no, offset)) < 0)
483 yaz_log(YLOG_FATAL, "mf_write %s internal error (1)", mf->name);
486 /* file needs to grow */
487 while (ps >= mf->files[mf->cur_file].blocks)
489 mfile_off_t needed = (ps - mf->files[mf->cur_file].blocks + 1) *
491 /* file overflow - allocate new file */
492 if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
493 needed > mf->files[mf->cur_file].dir->avail_bytes)
496 if ((nblocks = (int) (mf->files[mf->cur_file].dir->avail_bytes /
499 yaz_log(YLOG_DEBUG, "Capping off file %s at pos %d",
500 mf->files[mf->cur_file].path, nblocks);
501 if ((ps = file_position(mf,
502 (mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
503 mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
505 yaz_log(YLOG_FATAL, "mf_write %s internal error (2)",
509 yaz_log(YLOG_DEBUG, "ps = %d", ps);
510 if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
512 yaz_log(YLOG_ERRNO|YLOG_FATAL, "mf_write %s internal error (3)",
516 mf->files[mf->cur_file].blocks += nblocks;
517 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
518 mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
522 yaz_log(YLOG_DEBUG, "Creating new file.");
523 for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
524 dp->avail_bytes < needed; dp = dp->next);
527 yaz_log(YLOG_FATAL, "Cannot allocate more space for %s",
531 mf->files[mf->cur_file].top = (mf->cur_file ?
532 mf->files[mf->cur_file-1].top : -1) +
533 mf->files[mf->cur_file].blocks;
534 mf->files[++(mf->cur_file)].top = -1;
535 mf->files[mf->cur_file].dir = dp;
536 mf->files[mf->cur_file].number =
537 mf->files[mf->cur_file-1].number + 1;
538 mf->files[mf->cur_file].blocks = 0;
539 mf->files[mf->cur_file].bytes = 0;
540 mf->files[mf->cur_file].fd = -1;
541 sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
542 mf->files[mf->cur_file].number);
543 mf->files[mf->cur_file].path = xstrdup(tmp);
545 /* open new file and position at beginning */
546 if ((ps = file_position(mf, no, offset)) < 0)
548 yaz_log(YLOG_FATAL, "mf_write %s internal error (4)",
555 nblocks = ps - mf->files[mf->cur_file].blocks + 1;
556 mf->files[mf->cur_file].blocks += nblocks;
557 mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
558 if (mf->files[mf->cur_file].dir->max_bytes >= 0)
559 mf->files[mf->cur_file].dir->avail_bytes -=
560 nblocks * mf->blocksize;
563 towrite = nbytes ? nbytes : mf->blocksize;
564 if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
566 yaz_log(YLOG_FATAL|YLOG_ERRNO, "Write failed for file %s part %d",
567 mf->name, mf->cur_file);
570 zebra_mutex_unlock (&mf->mutex);