tdb.c

自己精简过的PPPD代码。在嵌入中应用可以更好的发挥。比原先的小了很多

 /* 
   Unix SMB/CIFS implementation.

   trivial database library

   Copyright (C) Andrew Tridgell              1999-2004
   Copyright (C) Paul `Rusty' Russell		   2000
   Copyright (C) Jeremy Allison			   2000-2003
   
     ** NOTE! The following LGPL license applies to the tdb
     ** library. This does NOT imply that all of Samba is released
     ** under the LGPL
   
   This library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2 of the License, or (at your option) any later version.

   This library 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
   Lesser General Public License for more details.
   
   You should have received a copy of the GNU Lesser General Public
   License along with this library; if not, write to the Free Software
   Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
*/


/* NOTE: If you use tdbs under valgrind, and in particular if you run
 * tdbtorture, you may get spurious "uninitialized value" warnings.  I
 * think this is because valgrind doesn't understand that the mmap'd
 * area may be written to by other processes.  Memory can, from the
 * point of view of the grinded process, spontaneously become
 * initialized.
 *
 * I can think of a few solutions.  [mbp 20030311]
 *
 * 1 - Write suppressions for Valgrind so that it doesn't complain
 * about this.  Probably the most reasonable but people need to
 * remember to use them.
 *
 * 2 - Use IO not mmap when running under valgrind.  Not so nice.
 *
 * 3 - Use the special valgrind macros to mark memory as valid at the
 * right time.  Probably too hard -- the process just doesn't know.
 */ 

#include <stdlib.h>
#include <stdio.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <fcntl.h>
#include <errno.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <signal.h>
#include "tdb.h"
#include "spinlock.h"

#define TDB_MAGIC_FOOD "TDB file\n"
#define TDB_VERSION (0x26011967 + 6)
#define TDB_MAGIC (0x26011999U)
#define TDB_FREE_MAGIC (~TDB_MAGIC)
#define TDB_DEAD_MAGIC (0xFEE1DEAD)
#define TDB_ALIGNMENT 4
#define MIN_REC_SIZE (2*sizeof(struct list_struct) + TDB_ALIGNMENT)
#define DEFAULT_HASH_SIZE 131
#define TDB_PAGE_SIZE 0x2000
#define FREELIST_TOP (sizeof(struct tdb_header))
#define TDB_ALIGN(x,a) (((x) + (a)-1) & ~((a)-1))
#define TDB_BYTEREV(x) (((((x)&0xff)<<24)|((x)&0xFF00)<<8)|(((x)>>8)&0xFF00)|((x)>>24))
#define TDB_DEAD(r) ((r)->magic == TDB_DEAD_MAGIC)
#define TDB_BAD_MAGIC(r) ((r)->magic != TDB_MAGIC && !TDB_DEAD(r))
#define TDB_HASH_TOP(hash) (FREELIST_TOP + (BUCKET(hash)+1)*sizeof(tdb_off))
#define TDB_DATA_START(hash_size) (TDB_HASH_TOP(hash_size-1) + TDB_SPINLOCK_SIZE(hash_size))


/* NB assumes there is a local variable called "tdb" that is the
 * current context, also takes doubly-parenthesized print-style
 * argument. */
#define TDB_LOG(x) (tdb->log_fn?((tdb->log_fn x),0) : 0)

/* lock offsets */
#define GLOBAL_LOCK 0
#define ACTIVE_LOCK 4

#ifndef MAP_FILE
#define MAP_FILE 0
#endif

#ifndef MAP_FAILED
#define MAP_FAILED ((void *)-1)
#endif

/* free memory if the pointer is valid and zero the pointer */
#ifndef SAFE_FREE
#define SAFE_FREE(x) do { if ((x) != NULL) {free((x)); (x)=NULL;} } while(0)
#endif

#define BUCKET(hash) ((hash) % tdb->header.hash_size)
TDB_DATA tdb_null;

/* all contexts, to ensure no double-opens (fcntl locks don't nest!) */
static TDB_CONTEXT *tdbs = NULL;

static int tdb_munmap(TDB_CONTEXT *tdb)
{
	if (tdb->flags & TDB_INTERNAL)
		return 0;

#ifdef HAVE_MMAP
	if (tdb->map_ptr) {
		int ret = munmap(tdb->map_ptr, tdb->map_size);
		if (ret != 0)
			return ret;
	}
#endif
	tdb->map_ptr = NULL;
	return 0;
}

static void tdb_mmap(TDB_CONTEXT *tdb)
{
	if (tdb->flags & TDB_INTERNAL)
		return;

#ifdef HAVE_MMAP
	if (!(tdb->flags & TDB_NOMMAP)) {
		tdb->map_ptr = mmap(NULL, tdb->map_size, 
				    PROT_READ|(tdb->read_only? 0:PROT_WRITE), 
				    MAP_SHARED|MAP_FILE, tdb->fd, 0);

		/*
		 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
		 */

		if (tdb->map_ptr == MAP_FAILED) {
			tdb->map_ptr = NULL;
			TDB_LOG((tdb, 2, "tdb_mmap failed for size %d (%s)\n", 
				 tdb->map_size, strerror(errno)));
		}
	} else {
		tdb->map_ptr = NULL;
	}
#else
	tdb->map_ptr = NULL;
#endif
}

/* Endian conversion: we only ever deal with 4 byte quantities */
static void *convert(void *buf, u32 size)
{
	u32 i, *p = buf;
	for (i = 0; i < size / 4; i++)
		p[i] = TDB_BYTEREV(p[i]);
	return buf;
}
#define DOCONV() (tdb->flags & TDB_CONVERT)
#define CONVERT(x) (DOCONV() ? convert(&x, sizeof(x)) : &x)

/* the body of the database is made of one list_struct for the free space
   plus a separate data list for each hash value */
struct list_struct {
	tdb_off next; /* offset of the next record in the list */
	tdb_len rec_len; /* total byte length of record */
	tdb_len key_len; /* byte length of key */
	tdb_len data_len; /* byte length of data */
	u32 full_hash; /* the full 32 bit hash of the key */
	u32 magic;   /* try to catch errors */
	/* the following union is implied:
		union {
			char record[rec_len];
			struct {
				char key[key_len];
				char data[data_len];
			}
			u32 totalsize; (tailer)
		}
	*/
};

/***************************************************************
 Allow a caller to set a "alarm" flag that tdb can check to abort
 a blocking lock on SIGALRM.
***************************************************************/

static sig_atomic_t *palarm_fired;

void tdb_set_lock_alarm(sig_atomic_t *palarm)
{
	palarm_fired = palarm;
}

/* a byte range locking function - return 0 on success
   this functions locks/unlocks 1 byte at the specified offset.

   On error, errno is also set so that errors are passed back properly
   through tdb_open(). */
static int tdb_brlock(TDB_CONTEXT *tdb, tdb_off offset, 
		      int rw_type, int lck_type, int probe)
{
	struct flock fl;
	int ret;

	if (tdb->flags & TDB_NOLOCK)
		return 0;
	if ((rw_type == F_WRLCK) && (tdb->read_only)) {
		errno = EACCES;
		return -1;
	}

	fl.l_type = rw_type;
	fl.l_whence = SEEK_SET;
	fl.l_start = offset;
	fl.l_len = 1;
	fl.l_pid = 0;

	do {
		ret = fcntl(tdb->fd,lck_type,&fl);
		if (ret == -1 && errno == EINTR && palarm_fired && *palarm_fired)
			break;
	} while (ret == -1 && errno == EINTR);

	if (ret == -1) {
		if (!probe && lck_type != F_SETLK) {
			/* Ensure error code is set for log fun to examine. */
			if (errno == EINTR && palarm_fired && *palarm_fired)
				tdb->ecode = TDB_ERR_LOCK_TIMEOUT;
			else
				tdb->ecode = TDB_ERR_LOCK;
			TDB_LOG((tdb, 5,"tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d\n", 
				 tdb->fd, offset, rw_type, lck_type));
		}
		/* Was it an alarm timeout ? */
		if (errno == EINTR && palarm_fired && *palarm_fired) {
			TDB_LOG((tdb, 5, "tdb_brlock timed out (fd=%d) at offset %d rw_type=%d lck_type=%d\n", 
				 tdb->fd, offset, rw_type, lck_type));
			return TDB_ERRCODE(TDB_ERR_LOCK_TIMEOUT, -1);
		}
		/* Otherwise - generic lock error. errno set by fcntl.
		 * EAGAIN is an expected return from non-blocking
		 * locks. */
		if (errno != EAGAIN) {
			TDB_LOG((tdb, 5, "tdb_brlock failed (fd=%d) at offset %d rw_type=%d lck_type=%d: %s\n", 
				 tdb->fd, offset, rw_type, lck_type, 
				 strerror(errno)));
		}
		return TDB_ERRCODE(TDB_ERR_LOCK, -1);
	}
	return 0;
}

/* lock a list in the database. list -1 is the alloc list */
static int tdb_lock(TDB_CONTEXT *tdb, int list, int ltype)
{
	if (list < -1 || list >= (int)tdb->header.hash_size) {
		TDB_LOG((tdb, 0,"tdb_lock: invalid list %d for ltype=%d\n", 
			   list, ltype));
		return -1;
	}
	if (tdb->flags & TDB_NOLOCK)
		return 0;

	/* Since fcntl locks don't nest, we do a lock for the first one,
	   and simply bump the count for future ones */
	if (tdb->locked[list+1].count == 0) {
		if (!tdb->read_only && tdb->header.rwlocks) {
			if (tdb_spinlock(tdb, list, ltype)) {
				TDB_LOG((tdb, 0, "tdb_lock spinlock failed on list %d ltype=%d\n", 
					   list, ltype));
				return -1;
			}
		} else if (tdb_brlock(tdb,FREELIST_TOP+4*list,ltype,F_SETLKW, 0)) {
			TDB_LOG((tdb, 0,"tdb_lock failed on list %d ltype=%d (%s)\n", 
					   list, ltype, strerror(errno)));
			return -1;
		}
		tdb->locked[list+1].ltype = ltype;
	}
	tdb->locked[list+1].count++;
	return 0;
}

/* unlock the database: returns void because it's too late for errors. */
	/* changed to return int it may be interesting to know there
	   has been an error  --simo */
static int tdb_unlock(TDB_CONTEXT *tdb, int list, int ltype)
{
	int ret = -1;

	if (tdb->flags & TDB_NOLOCK)
		return 0;

	/* Sanity checks */
	if (list < -1 || list >= (int)tdb->header.hash_size) {
		TDB_LOG((tdb, 0, "tdb_unlock: list %d invalid (%d)\n", list, tdb->header.hash_size));
		return ret;
	}

	if (tdb->locked[list+1].count==0) {
		TDB_LOG((tdb, 0, "tdb_unlock: count is 0\n"));
		return ret;
	}

	if (tdb->locked[list+1].count == 1) {
		/* Down to last nested lock: unlock underneath */
		if (!tdb->read_only && tdb->header.rwlocks) {
			ret = tdb_spinunlock(tdb, list, ltype);
		} else {
			ret = tdb_brlock(tdb, FREELIST_TOP+4*list, F_UNLCK, F_SETLKW, 0);
		}
	} else {
		ret = 0;
	}
	tdb->locked[list+1].count--;

	if (ret)
		TDB_LOG((tdb, 0,"tdb_unlock: An error occurred unlocking!\n")); 
	return ret;
}

/* check for an out of bounds access - if it is out of bounds then
   see if the database has been expanded by someone else and expand
   if necessary 
   note that "len" is the minimum length needed for the db
*/
static int tdb_oob(TDB_CONTEXT *tdb, tdb_off len, int probe)
{
	struct stat st;
	if (len <= tdb->map_size)
		return 0;
	if (tdb->flags & TDB_INTERNAL) {
		if (!probe) {
			/* Ensure ecode is set for log fn. */
			tdb->ecode = TDB_ERR_IO;
			TDB_LOG((tdb, 0,"tdb_oob len %d beyond internal malloc size %d\n",
				 (int)len, (int)tdb->map_size));
		}
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}

	if (fstat(tdb->fd, &st) == -1)
		return TDB_ERRCODE(TDB_ERR_IO, -1);

	if (st.st_size < (size_t)len) {
		if (!probe) {
			/* Ensure ecode is set for log fn. */
			tdb->ecode = TDB_ERR_IO;
			TDB_LOG((tdb, 0,"tdb_oob len %d beyond eof at %d\n",
				 (int)len, (int)st.st_size));
		}
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}

	/* Unmap, update size, remap */
	if (tdb_munmap(tdb) == -1)
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	tdb->map_size = st.st_size;
	tdb_mmap(tdb);
	return 0;
}

/* write a lump of data at a specified offset */
static int tdb_write(TDB_CONTEXT *tdb, tdb_off off, void *buf, tdb_len len)
{
	if (tdb_oob(tdb, off + len, 0) != 0)
		return -1;

	if (tdb->map_ptr)
		memcpy(off + (char *)tdb->map_ptr, buf, len);
#ifdef HAVE_PWRITE
	else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
#else
	else if (lseek(tdb->fd, off, SEEK_SET) != off
		 || write(tdb->fd, buf, len) != (ssize_t)len) {
#endif
		/* Ensure ecode is set for log fn. */
		tdb->ecode = TDB_ERR_IO;
		TDB_LOG((tdb, 0,"tdb_write failed at %d len=%d (%s)\n",
			   off, len, strerror(errno)));
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}
	return 0;
}

/* read a lump of data at a specified offset, maybe convert */
static int tdb_read(TDB_CONTEXT *tdb,tdb_off off,void *buf,tdb_len len,int cv)
{
	if (tdb_oob(tdb, off + len, 0) != 0)
		return -1;

	if (tdb->map_ptr)
		memcpy(buf, off + (char *)tdb->map_ptr, len);
#ifdef HAVE_PREAD
	else if (pread(tdb->fd, buf, len, off) != (ssize_t)len) {
#else
	else if (lseek(tdb->fd, off, SEEK_SET) != off
		 || read(tdb->fd, buf, len) != (ssize_t)len) {
#endif
		/* Ensure ecode is set for log fn. */
		tdb->ecode = TDB_ERR_IO;
		TDB_LOG((tdb, 0,"tdb_read failed at %d len=%d (%s)\n",
			   off, len, strerror(errno)));
		return TDB_ERRCODE(TDB_ERR_IO, -1);
	}
	if (cv)
		convert(buf, len);
	return 0;
}

/* read a lump of data, allocating the space for it */
static char *tdb_alloc_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_len len)
{
	char *buf;

	if (!(buf = malloc(len))) {
		/* Ensure ecode is set for log fn. */
		tdb->ecode = TDB_ERR_OOM;
		TDB_LOG((tdb, 0,"tdb_alloc_read malloc failed len=%d (%s)\n",
			   len, strerror(errno)));
		return TDB_ERRCODE(TDB_ERR_OOM, buf);
	}
	if (tdb_read(tdb, offset, buf, len, 0) == -1) {
		SAFE_FREE(buf);
		return NULL;
	}
	return buf;
}

/* read/write a tdb_off */
static int ofs_read(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
{
	return tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
}
static int ofs_write(TDB_CONTEXT *tdb, tdb_off offset, tdb_off *d)
{
	tdb_off off = *d;
	return tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
}

/* read/write a record */
static int rec_read(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
{
	if (tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
		return -1;
	if (TDB_BAD_MAGIC(rec)) {
		/* Ensure ecode is set for log fn. */
		tdb->ecode = TDB_ERR_CORRUPT;
		TDB_LOG((tdb, 0,"rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
		return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
	}
	return tdb_oob(tdb, rec->next+sizeof(*rec), 0);
}
static int rec_write(TDB_CONTEXT *tdb, tdb_off offset, struct list_struct *rec)
{
	struct list_struct r = *rec;
	return tdb_write(tdb, offset, CONVERT(r), sizeof(r));
}

/* read a freelist record and check for simple errors */
static int rec_free_read(TDB_CONTEXT *tdb, tdb_off off, struct list_struct *rec)
{
	if (tdb_read(tdb, off, rec, sizeof(*rec),DOCONV()) == -1)
		return -1;

	if (rec->magic == TDB_MAGIC) {
		/* this happens when a app is showdown while deleting a record - we should
		   not completely fail when this happens */
		TDB_LOG((tdb, 0,"rec_free_read non-free magic 0x%x at offset=%d - fixing\n", 
			 rec->magic, off));
		rec->magic = TDB_FREE_MAGIC;
		if (tdb_write(tdb, off, rec, sizeof(*rec)) == -1)
			return -1;
	}

	if (rec->magic != TDB_FREE_MAGIC) {
		/* Ensure ecode is set for log fn. */
		tdb->ecode = TDB_ERR_CORRUPT;
		TDB_LOG((tdb, 0,"rec_free_read bad magic 0x%x at offset=%d\n", 
			   rec->magic, off));
		return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
	}
	if (tdb_oob(tdb, rec->next+sizeof(*rec), 0) != 0)
		return -1;
	return 0;
}

/* update a record tailer (must hold allocation lock) */
static int update_tailer(TDB_CONTEXT *tdb, tdb_off offset,
			 const struct list_struct *rec)
{
	tdb_off totalsize;

	/* Offset of tailer from record header */
	totalsize = sizeof(*rec) + rec->rec_len;
	return ofs_write(tdb, offset + totalsize - sizeof(tdb_off),
			 &totalsize);
}

static tdb_off tdb_dump_record(TDB_CONTEXT *tdb, tdb_off offset)
{