/* $Id: mfile.c,v 1.52 2003/04/05 12:32:34 adam Exp $
Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003
Index Data Aps
This file is part of the Zebra server.
Zebra is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 2, or (at your option) any later
version.
Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with Zebra; see the file LICENSE.zebra. If not, write to the
Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
02111-1307, USA.
*/
/*
* TODO: The size estimates in init may not be accurate due to
* only partially written final blocks.
*/
#include <sys/types.h>
#include <fcntl.h>
#ifdef WIN32
#include <io.h>
#else
#include <unistd.h>
#endif
#include <direntz.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include <assert.h>
#include <errno.h>
#include <zebra-lock.h>
#include <zebrautl.h>
#include <mfile.h>
static int scan_areadef(MFile_area ma, const char *ad, const char *base)
{
/*
* If no definition is given, use current directory, unlimited.
*/
char dirname[FILENAME_MAX+1];
mf_dir **dp = &ma->dirs, *dir = *dp;
if (!ad)
ad = ".:-1b";
for (;;)
{
const char *ad0 = ad;
int i = 0, fact = 1, multi;
mfile_off_t size = 0;
while (*ad == ' ' || *ad == '\t')
ad++;
if (!*ad)
break;
if (!yaz_is_abspath(ad) && base)
{
strcpy (dirname, base);
i = strlen(dirname);
dirname[i++] = '/';
}
while (*ad)
{
if (*ad == ':' && strchr ("+-0123456789", ad[1]))
break;
if (i < FILENAME_MAX)
dirname[i++] = *ad;
ad++;
}
dirname[i] = '\0';
if (*ad++ != ':')
{
logf (LOG_WARN, "Missing colon after path: %s", ad0);
return -1;
}
if (i == 0)
{
logf (LOG_WARN, "Empty path: %s", ad0);
return -1;
}
while (*ad == ' ' || *ad == '\t')
ad++;
if (*ad == '-')
{
fact = -1;
ad++;
}
else if (*ad == '+')
ad++;
size = 0;
if (*ad < '0' || *ad > '9')
{
logf (LOG_FATAL, "Missing size after path: %s", ad0);
return -1;
}
size = 0;
while (*ad >= '0' && *ad <= '9')
size = size*10 + (*ad++ - '0');
switch (*ad)
{
case 'B': case 'b': multi = 1; break;
case 'K': case 'k': multi = 1024; break;
case 'M': case 'm': multi = 1048576; break;
case 'G': case 'g': multi = 1073741824; break;
case '\0':
logf (LOG_FATAL, "Missing unit: %s", ad0);
return -1;
default:
logf (LOG_FATAL, "Illegal unit: %c in %s", *ad, ad0);
return -1;
}
ad++;
*dp = dir = (mf_dir *) xmalloc(sizeof(mf_dir));
dir->next = 0;
strcpy(dir->name, dirname);
dir->max_bytes = dir->avail_bytes = fact * size * multi;
dp = &dir->next;
}
return 0;
}
static int file_position(MFile mf, int pos, int offset)
{
int off = 0, c = mf->cur_file, ps;
if ((c > 0 && pos <= mf->files[c-1].top) ||
(c < mf->no_files -1 && pos > mf->files[c].top))
{
c = 0;
while (c + 1 < mf->no_files && mf->files[c].top < pos)
{
off += mf->files[c].blocks;
c++;
}
assert(c < mf->no_files);
}
else
off = c ? (mf->files[c-1].top + 1) : 0;
if (mf->files[c].fd < 0)
{
if ((mf->files[c].fd = open(mf->files[c].path,
mf->wr ?
(O_BINARY|O_RDWR|O_CREAT) :
(O_BINARY|O_RDONLY), 0666)) < 0)
{
if (!mf->wr && errno == ENOENT && off == 0)
return -2;
logf (LOG_WARN|LOG_ERRNO, "Failed to open %s", mf->files[c].path);
return -1;
}
}
ps = pos - off;
if (mfile_seek(mf->files[c].fd, ps * (mfile_off_t) mf->blocksize + offset,
SEEK_SET) < 0)
{
logf (LOG_WARN|LOG_ERRNO, "Failed to seek in %s", mf->files[c].path);
logf(LOG_WARN, "pos=%d off=%d blocksize=%d offset=%d",
pos, off, mf->blocksize, offset);
return -1;
}
mf->cur_file = c;
return ps;
}
static int cmp_part_file(const void *p1, const void *p2)
{
return ((part_file *)p1)->number - ((part_file *)p2)->number;
}
/*
* Create a new area, cotaining metafiles in directories.
* Find the part-files in each directory, and inventory the existing metafiles.
*/
MFile_area mf_init(const char *name, const char *spec, const char *base)
{
MFile_area ma = (MFile_area) xmalloc(sizeof(*ma));
mf_dir *dirp;
meta_file *meta_f;
part_file *part_f = 0;
DIR *dd;
struct dirent *dent;
int fd, number;
char metaname[FILENAME_MAX+1], tmpnam[FILENAME_MAX+1];
logf (LOG_DEBUG, "mf_init(%s)", name);
strcpy(ma->name, name);
ma->mfiles = 0;
ma->dirs = 0;
if (scan_areadef(ma, spec, base) < 0)
{
logf (LOG_WARN, "Failed to access description of '%s'", name);
return 0;
}
/* look at each directory */
for (dirp = ma->dirs; dirp; dirp = dirp->next)
{
if (!(dd = opendir(dirp->name)))
{
logf (LOG_WARN|LOG_ERRNO, "Failed to open directory %s",
dirp->name);
return 0;
}
/* look at each file */
while ((dent = readdir(dd)))
{
int len = strlen(dent->d_name);
const char *cp = strrchr (dent->d_name, '-');
if (strchr (".-", *dent->d_name))
continue;
if (len < 5 || !cp || strcmp (dent->d_name + len - 3, ".mf"))
continue;
number = atoi(cp+1);
memcpy (metaname, dent->d_name, cp - dent->d_name);
metaname[ cp - dent->d_name] = '\0';
for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
{
/* known metafile */
if (!strcmp(meta_f->name, metaname))
{
part_f = &meta_f->files[meta_f->no_files++];
break;
}
}
/* new metafile */
if (!meta_f)
{
meta_f = (meta_file *) xmalloc(sizeof(*meta_f));
zebra_mutex_init (&meta_f->mutex);
meta_f->ma = ma;
meta_f->next = ma->mfiles;
meta_f->open = 0;
meta_f->cur_file = -1;
meta_f->unlink_flag = 0;
ma->mfiles = meta_f;
strcpy(meta_f->name, metaname);
part_f = &meta_f->files[0];
meta_f->no_files = 1;
}
part_f->number = number;
part_f->dir = dirp;
part_f->fd = -1;
sprintf(tmpnam, "%s/%s", dirp->name, dent->d_name);
part_f->path = xstrdup(tmpnam);
/* get size */
if ((fd = open(part_f->path, O_BINARY|O_RDONLY)) < 0)
{
logf (LOG_FATAL|LOG_ERRNO, "Failed to access %s",
dent->d_name);
return 0;
}
if ((part_f->bytes = mfile_seek(fd, 0, SEEK_END)) < 0)
{
logf (LOG_FATAL|LOG_ERRNO, "Failed to seek in %s",
dent->d_name);
return 0;
}
#ifndef WIN32
fsync(fd);
#endif
close(fd);
if (dirp->max_bytes >= 0)
dirp->avail_bytes -= part_f->bytes;
}
closedir(dd);
}
for (meta_f = ma->mfiles; meta_f; meta_f = meta_f->next)
{
logf (LOG_DEBUG, "mf_init: %s consists of %d part(s)", meta_f->name,
meta_f->no_files);
qsort(meta_f->files, meta_f->no_files, sizeof(part_file),
cmp_part_file);
}
return ma;
}
void mf_destroy(MFile_area ma)
{
mf_dir *dp;
meta_file *meta_f;
if (!ma)
return;
dp = ma->dirs;
while (dp)
{
mf_dir *d = dp;
dp = dp->next;
xfree (d);
}
meta_f = ma->mfiles;
while (meta_f)
{
int i;
meta_file *m = meta_f;
for (i = 0; i<m->no_files; i++)
{
xfree (m->files[i].path);
}
zebra_mutex_destroy (&meta_f->mutex);
meta_f = meta_f->next;
xfree (m);
}
xfree (ma);
}
void mf_reset(MFile_area ma)
{
meta_file *meta_f;
if (!ma)
return;
meta_f = ma->mfiles;
while (meta_f)
{
int i;
meta_file *m = meta_f;
assert (!m->open);
for (i = 0; i<m->no_files; i++)
{
unlink (m->files[i].path);
xfree (m->files[i].path);
}
meta_f = meta_f->next;
xfree (m);
}
ma->mfiles = 0;
}
/*
* Open a metafile.
* If !ma, Use MF_DEFAULT_AREA.
*/
MFile mf_open(MFile_area ma, const char *name, int block_size, int wflag)
{
meta_file *mnew;
int i;
char tmp[FILENAME_MAX+1];
mf_dir *dp;
logf(LOG_DEBUG, "mf_open(%s bs=%d, %s)", name, block_size,
wflag ? "RW" : "RDONLY");
assert (ma);
for (mnew = ma->mfiles; mnew; mnew = mnew->next)
if (!strcmp(name, mnew->name))
{
if (mnew->open)
abort();
else
break;
}
if (!mnew)
{
mnew = (meta_file *) xmalloc(sizeof(*mnew));
strcpy(mnew->name, name);
/* allocate one, empty file */
zebra_mutex_init (&mnew->mutex);
mnew->no_files = 1;
mnew->files[0].bytes = 0;
mnew->files[0].blocks = 0;
mnew->files[0].top = -1;
mnew->files[0].number = 0;
mnew->files[0].fd = -1;
mnew->unlink_flag = 0;
mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
for (dp = ma->dirs; dp && dp->max_bytes >= 0 && dp->avail_bytes <
mnew->min_bytes_creat; dp = dp->next);
if (!dp)
{
logf (LOG_FATAL, "Insufficient space for new mfile.");
return 0;
}
mnew->files[0].dir = dp;
sprintf(tmp, "%s/%s-%d.mf", dp->name, mnew->name, 0);
mnew->files[0].path = xstrdup(tmp);
mnew->ma = ma;
mnew->next = ma->mfiles;
ma->mfiles = mnew;
}
else
{
for (i = 0; i < mnew->no_files; i++)
{
if (mnew->files[i].bytes % block_size)
mnew->files[i].bytes += block_size - mnew->files[i].bytes %
block_size;
mnew->files[i].blocks = (int) (mnew->files[i].bytes / block_size);
}
assert(!mnew->open);
}
mnew->blocksize = block_size;
mnew->min_bytes_creat = MF_MIN_BLOCKS_CREAT * block_size;
mnew->wr=wflag;
mnew->cur_file = 0;
mnew->open = 1;
for (i = 0; i < mnew->no_files; i++)
{
mnew->files[i].blocks = (int)(mnew->files[i].bytes / mnew->blocksize);
if (i == mnew->no_files - 1)
mnew->files[i].top = -1;
else
mnew->files[i].top =
i ? (mnew->files[i-1].top + mnew->files[i].blocks)
: (mnew->files[i].blocks - 1);
}
return mnew;
}
/*
* Close a metafile.
*/
int mf_close(MFile mf)
{
int i;
logf (LOG_DEBUG, "mf_close(%s)", mf->name);
assert(mf->open);
for (i = 0; i < mf->no_files; i++)
{
if (mf->files[i].fd >= 0)
{
#ifndef WIN32
fsync(mf->files[i].fd);
#endif
close(mf->files[i].fd);
mf->files[i].fd = -1;
}
if (mf->unlink_flag)
unlink(mf->files[i].path);
}
mf->open = 0;
return 0;
}
/*
* Read one block from a metafile. Interface mirrors bfile.
*/
int mf_read(MFile mf, int no, int offset, int nbytes, void *buf)
{
int rd, toread;
zebra_mutex_lock (&mf->mutex);
if ((rd = file_position(mf, no, offset)) < 0)
{
if (rd == -2)
{
zebra_mutex_unlock (&mf->mutex);
return 0;
}
else
{
yaz_log (LOG_FATAL, "mf_read %s internal error", mf->name);
exit(1);
}
}
toread = nbytes ? nbytes : mf->blocksize;
if ((rd = read(mf->files[mf->cur_file].fd, buf, toread)) < 0)
{
logf (LOG_FATAL|LOG_ERRNO, "mf_read: Read failed (%s)",
mf->files[mf->cur_file].path);
exit(1);
}
zebra_mutex_unlock (&mf->mutex);
if (rd < toread)
return 0;
else
return 1;
}
/*
* Write.
*/
int mf_write(MFile mf, int no, int offset, int nbytes, const void *buf)
{
int ps, nblocks, towrite;
mf_dir *dp;
char tmp[FILENAME_MAX+1];
unsigned char dummych = '\xff';
zebra_mutex_lock (&mf->mutex);
if ((ps = file_position(mf, no, offset)) < 0)
{
yaz_log (LOG_FATAL, "mf_write %s internal error (1)", mf->name);
exit(1);
}
/* file needs to grow */
while (ps >= mf->files[mf->cur_file].blocks)
{
mfile_off_t needed = (ps - mf->files[mf->cur_file].blocks + 1) *
mf->blocksize;
/* file overflow - allocate new file */
if (mf->files[mf->cur_file].dir->max_bytes >= 0 &&
needed > mf->files[mf->cur_file].dir->avail_bytes)
{
/* cap off file? */
if ((nblocks = (int) (mf->files[mf->cur_file].dir->avail_bytes /
mf->blocksize)) > 0)
{
logf (LOG_DEBUG, "Capping off file %s at pos %d",
mf->files[mf->cur_file].path, nblocks);
if ((ps = file_position(mf,
(mf->cur_file ? mf->files[mf->cur_file-1].top : 0) +
mf->files[mf->cur_file].blocks + nblocks - 1, 0)) < 0)
{
yaz_log (LOG_FATAL, "mf_write %s internal error (2)",
mf->name);
exit(1);
}
logf (LOG_DEBUG, "ps = %d", ps);
if (write(mf->files[mf->cur_file].fd, &dummych, 1) < 1)
{
logf (LOG_ERRNO|LOG_FATAL, "mf_write %s internal error (3)",
mf->name);
exit(1);
}
mf->files[mf->cur_file].blocks += nblocks;
mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
mf->files[mf->cur_file].dir->avail_bytes -= nblocks *
mf->blocksize;
}
/* get other bit */
logf (LOG_DEBUG, "Creating new file.");
for (dp = mf->ma->dirs; dp && dp->max_bytes >= 0 &&
dp->avail_bytes < needed; dp = dp->next);
if (!dp)
{
logf (LOG_FATAL, "Cannot allocate more space for %s",
mf->name);
exit(1);
}
mf->files[mf->cur_file].top = (mf->cur_file ?
mf->files[mf->cur_file-1].top : -1) +
mf->files[mf->cur_file].blocks;
mf->files[++(mf->cur_file)].top = -1;
mf->files[mf->cur_file].dir = dp;
mf->files[mf->cur_file].number =
mf->files[mf->cur_file-1].number + 1;
mf->files[mf->cur_file].blocks = 0;
mf->files[mf->cur_file].bytes = 0;
mf->files[mf->cur_file].fd = -1;
sprintf(tmp, "%s/%s-%d.mf", dp->name, mf->name,
mf->files[mf->cur_file].number);
mf->files[mf->cur_file].path = xstrdup(tmp);
mf->no_files++;
/* open new file and position at beginning */
if ((ps = file_position(mf, no, offset)) < 0)
{
yaz_log (LOG_FATAL, "mf_write %s internal error (4)",
mf->name);
exit(1);
}
}
else
{
nblocks = ps - mf->files[mf->cur_file].blocks + 1;
mf->files[mf->cur_file].blocks += nblocks;
mf->files[mf->cur_file].bytes += nblocks * mf->blocksize;
if (mf->files[mf->cur_file].dir->max_bytes >= 0)
mf->files[mf->cur_file].dir->avail_bytes -=
nblocks * mf->blocksize;
}
}
towrite = nbytes ? nbytes : mf->blocksize;
if (write(mf->files[mf->cur_file].fd, buf, towrite) < towrite)
{
logf (LOG_FATAL|LOG_ERRNO, "Write failed for file %s part %d",
mf->name, mf->cur_file);
exit(1);
}
zebra_mutex_unlock (&mf->mutex);
return 0;
}
/*
* Destroy a metafile, unlinking component files. File must be open.
*/
int mf_unlink(MFile mf)
{
if (mf->open)
mf->unlink_flag = 1;
else
{
int i;
for (i = 0; i<mf->no_files; i++)
unlink(mf->files[i].path);
}
return 0;
}
/*
* Unlink the file by name, rather than MFile-handle. File should be closed.
*/
int mf_unlink_name(MFile_area ma, const char *name)
{
abort();
return 0;
}
syntax highlighted by Code2HTML, v. 0.9.1