#include <time.h>
#include <unistd.h>
#include <pthread.h>
-#include <endian.h>
#include "mdb.h"
#define ULONG unsigned long
typedef ULONG pgno_t;
-#include "idl.h"
+#include "midl.h"
#ifndef DEBUG
#define DEBUG 1
} MDB_rxbody;
#ifndef CACHELINE
-#define CACHELINE 64 /* most CPUs. Itanium uses 128 */
+# ifdef __APPLE__
+# define CACHELINE 128 /* 64 is too small to contain a mutex */
+# else
+# define CACHELINE 64 /* most CPUs. Itanium uses 128 */
+# endif
#endif
typedef struct MDB_reader {
} MDB_ppage;
SLIST_HEAD(page_stack, MDB_ppage);
+/* FIXME: tree depth is mostly bounded, we should just
+ * use a fixed array and avoid malloc/pointer chasing
+ */
#define CURSOR_EMPTY(c) SLIST_EMPTY(&(c)->mc_stack)
#define CURSOR_TOP(c) SLIST_FIRST(&(c)->mc_stack)
#define CURSOR_POP(c) SLIST_REMOVE_HEAD(&(c)->mc_stack, mp_entry)
MDB_cursor *cursor, int modify,
MDB_pageparent *mpp);
-static int mdbenv_read_header(MDB_env *env, MDB_meta *meta);
-static int mdbenv_read_meta(MDB_env *env, int *which);
-static int mdbenv_write_meta(MDB_txn *txn);
+static int mdb_env_read_header(MDB_env *env, MDB_meta *meta);
+static int mdb_env_read_meta(MDB_env *env, int *which);
+static int mdb_env_write_meta(MDB_txn *txn);
static MDB_page *mdb_get_page(MDB_txn *txn, pgno_t pgno);
static MDB_node *mdb_search_node(MDB_txn *txn, MDB_dbi dbi, MDB_page *mp,
MDB_val *key, MDB_val *data);
static void mdb_xcursor_init0(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx);
-static void mdb_xcursor_init1(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx, MDB_db *db);
+static void mdb_xcursor_init1(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx, MDB_node *node);
static void mdb_xcursor_fini(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx);
static size_t mdb_leaf_size(MDB_env *env, MDB_val *key,
return *p1 - *p2;
}
+char *
+mdb_version(int *maj, int *min, int *pat)
+{
+ *maj = MDB_VERSION_MAJOR;
+ *min = MDB_VERSION_MINOR;
+ *pat = MDB_VERSION_PATCH;
+ return MDB_VERSION_STRING;
+}
+
int
mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b)
{
if ((dp = mdb_alloc_page(txn, pp->mp_parent, pp->mp_pi, 1)) == NULL)
return ENOMEM;
DPRINTF("touched page %lu -> %lu", mp->mp_pgno, dp->p.mp_pgno);
- mdb_idl_insert(txn->mt_free_pgs, mp->mp_pgno);
+ mdb_midl_insert(txn->mt_free_pgs, mp->mp_pgno);
pgno = dp->p.mp_pgno;
memcpy(&dp->p, mp, txn->mt_env->me_psize);
mp = &dp->p;
}
int
-mdbenv_sync(MDB_env *env)
+mdb_env_sync(MDB_env *env, int force)
{
int rc = 0;
- if (!F_ISSET(env->me_flags, MDB_NOSYNC)) {
+ if (force || !F_ISSET(env->me_flags, MDB_NOSYNC)) {
if (fsync(env->me_fd))
rc = errno;
}
txn->mt_env = env;
- if ((rc = mdbenv_read_meta(env, &toggle)) != MDB_SUCCESS) {
+ if ((rc = mdb_env_read_meta(env, &toggle)) != MDB_SUCCESS) {
mdb_txn_abort(txn);
return rc;
}
free(dp);
}
- if ((n = mdbenv_sync(env)) != 0 ||
- (n = mdbenv_write_meta(txn)) != MDB_SUCCESS ||
- (n = mdbenv_sync(env)) != 0) {
+ if ((n = mdb_env_sync(env, 0)) != 0 ||
+ (n = mdb_env_write_meta(txn)) != MDB_SUCCESS ||
+ (n = mdb_env_sync(env, 0)) != 0) {
mdb_txn_abort(txn);
return n;
}
}
static int
-mdbenv_read_header(MDB_env *env, MDB_meta *meta)
+mdb_env_read_header(MDB_env *env, MDB_meta *meta)
{
char page[PAGESIZE];
MDB_page *p;
}
static int
-mdbenv_init_meta(MDB_env *env, MDB_meta *meta)
+mdb_env_init_meta(MDB_env *env, MDB_meta *meta)
{
MDB_page *p, *q;
MDB_meta *m;
}
static int
-mdbenv_write_meta(MDB_txn *txn)
+mdb_env_write_meta(MDB_txn *txn)
{
MDB_env *env;
MDB_meta meta;
}
static int
-mdbenv_read_meta(MDB_env *env, int *which)
+mdb_env_read_meta(MDB_env *env, int *which)
{
int toggle = 0;
}
int
-mdbenv_create(MDB_env **env)
+mdb_env_create(MDB_env **env)
{
MDB_env *e;
}
int
-mdbenv_set_mapsize(MDB_env *env, size_t size)
+mdb_env_set_mapsize(MDB_env *env, size_t size)
{
if (env->me_map)
return EINVAL;
}
int
-mdbenv_set_maxdbs(MDB_env *env, int dbs)
+mdb_env_set_maxdbs(MDB_env *env, int dbs)
{
env->me_maxdbs = dbs;
return MDB_SUCCESS;
}
int
-mdbenv_set_maxreaders(MDB_env *env, int readers)
+mdb_env_set_maxreaders(MDB_env *env, int readers)
{
env->me_maxreaders = readers;
return MDB_SUCCESS;
}
int
-mdbenv_get_maxreaders(MDB_env *env, int *readers)
+mdb_env_get_maxreaders(MDB_env *env, int *readers)
{
if (!env || !readers)
return EINVAL;
}
static int
-mdbenv_open2(MDB_env *env, unsigned int flags)
+mdb_env_open2(MDB_env *env, unsigned int flags)
{
int i, newenv = 0;
MDB_meta meta;
memset(&meta, 0, sizeof(meta));
- if ((i = mdbenv_read_header(env, &meta)) != 0) {
+ if ((i = mdb_env_read_header(env, &meta)) != 0) {
if (i != ENOENT)
return i;
DPRINTF("new mdbenv");
meta.mm_mapsize = env->me_mapsize;
if (flags & MDB_FIXEDMAP)
meta.mm_address = env->me_map;
- i = mdbenv_init_meta(env, &meta);
+ i = mdb_env_init_meta(env, &meta);
if (i != MDB_SUCCESS) {
munmap(env->me_map, env->me_mapsize);
return i;
env->me_metas[0] = METADATA(p);
env->me_metas[1] = (MDB_meta *)((char *)env->me_metas[0] + meta.mm_psize);
- if ((i = mdbenv_read_meta(env, NULL)) != 0)
+ if ((i = mdb_env_read_meta(env, NULL)) != 0)
return i;
DPRINTF("opened database version %u, pagesize %u",
}
static void
-mdbenv_reader_dest(void *ptr)
+mdb_env_reader_dest(void *ptr)
{
MDB_reader *reader = ptr;
/* downgrade the exclusive lock on the region back to shared */
static void
-mdbenv_share_locks(MDB_env *env)
+mdb_env_share_locks(MDB_env *env)
{
struct flock lock_info;
}
static int
-mdbenv_setup_locks(MDB_env *env, char *lpath, int mode, int *excl)
+mdb_env_setup_locks(MDB_env *env, char *lpath, int mode, int *excl)
{
int rc;
off_t size, rsize;
#define LOCKNAME "/lock.mdb"
#define DATANAME "/data.mdb"
int
-mdbenv_open(MDB_env *env, const char *path, unsigned int flags, mode_t mode)
+mdb_env_open(MDB_env *env, const char *path, unsigned int flags, mode_t mode)
{
int oflags, rc, len, excl;
char *lpath, *dpath;
sprintf(lpath, "%s" LOCKNAME, path);
sprintf(dpath, "%s" DATANAME, path);
- rc = mdbenv_setup_locks(env, lpath, mode, &excl);
+ rc = mdb_env_setup_locks(env, lpath, mode, &excl);
if (rc)
goto leave;
if ((env->me_fd = open(dpath, oflags, mode)) == -1)
return errno;
- if ((rc = mdbenv_open2(env, flags)) != MDB_SUCCESS) {
+ if ((rc = mdb_env_open2(env, flags)) != MDB_SUCCESS) {
close(env->me_fd);
env->me_fd = -1;
} else {
env->me_path = strdup(path);
DPRINTF("opened dbenv %p", (void *) env);
- pthread_key_create(&env->me_txkey, mdbenv_reader_dest);
+ pthread_key_create(&env->me_txkey, mdb_env_reader_dest);
if (excl)
- mdbenv_share_locks(env);
+ mdb_env_share_locks(env);
env->me_dbxs = calloc(env->me_maxdbs, sizeof(MDB_dbx));
env->me_dbs[0] = calloc(env->me_maxdbs, sizeof(MDB_db));
env->me_dbs[1] = calloc(env->me_maxdbs, sizeof(MDB_db));
}
void
-mdbenv_close(MDB_env *env)
+mdb_env_close(MDB_env *env)
{
if (env == NULL)
return;
free(env->me_dbxs);
free(env->me_path);
+ pthread_key_delete(env->me_txkey);
+
if (env->me_map) {
munmap(env->me_map, env->me_mapsize);
}
close(env->me_fd);
if (env->me_txns) {
+ pid_t pid = getpid();
size_t size = (env->me_maxreaders-1) * sizeof(MDB_reader) + sizeof(MDB_txninfo);
+ int i;
+ for (i=0; i<env->me_txns->mti_numreaders; i++)
+ if (env->me_txns->mti_readers[i].mr_pid == pid)
+ env->me_txns->mti_readers[i].mr_pid = 0;
munmap(env->me_txns, size);
}
close(env->me_lfd);
MDB_xcursor mx;
mdb_xcursor_init0(txn, dbi, &mx);
- mdb_xcursor_init1(txn, dbi, &mx, NODEDATA(leaf));
+ mdb_xcursor_init1(txn, dbi, &mx, leaf);
rc = mdb_search_page(&mx.mx_txn, mx.mx_cursor.mc_dbi, NULL, NULL, 0, &mpp);
if (rc != MDB_SUCCESS)
return rc;
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
- mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, NODEDATA(leaf));
+ mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, leaf);
rc = mdb_cursor_first(&cursor->mc_xcursor->mx_cursor, data, NULL);
if (rc != MDB_SUCCESS)
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
if (op == MDB_PREV || op == MDB_PREV_DUP) {
- rc = mdb_cursor_next(&cursor->mc_xcursor->mx_cursor, data, NULL, MDB_PREV);
+ rc = mdb_cursor_prev(&cursor->mc_xcursor->mx_cursor, data, NULL, MDB_PREV);
if (op != MDB_PREV || rc == MDB_SUCCESS)
return rc;
}
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
- mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, NODEDATA(leaf));
+ mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, leaf);
rc = mdb_cursor_last(&cursor->mc_xcursor->mx_cursor, data, NULL);
if (rc != MDB_SUCCESS)
return rc;
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
- mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, NODEDATA(leaf));
+ mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, leaf);
if (op == MDB_SET || op == MDB_SET_RANGE) {
rc = mdb_cursor_first(&cursor->mc_xcursor->mx_cursor, data, NULL);
} else {
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
- mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, NODEDATA(leaf));
+ mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, leaf);
rc = mdb_cursor_first(&cursor->mc_xcursor->mx_cursor, data, NULL);
if (rc)
return rc;
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
- mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, NODEDATA(leaf));
+ mdb_xcursor_init1(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor, leaf);
rc = mdb_cursor_last(&cursor->mc_xcursor->mx_cursor, data, NULL);
if (rc)
return rc;
}
static void
-mdb_xcursor_init1(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx, MDB_db *db)
+mdb_xcursor_init1(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx, MDB_node *node)
{
+ MDB_db *db = NODEDATA(node);
+ MDB_dbi dbn;
mx->mx_dbs[0] = txn->mt_dbs[0];
mx->mx_dbs[1] = txn->mt_dbs[1];
if (dbi > 1) {
mx->mx_dbs[2] = txn->mt_dbs[dbi];
- mx->mx_dbs[3] = *db;
+ dbn = 3;
} else {
- mx->mx_dbs[2] = *db;
+ dbn = 2;
}
+ mx->mx_dbs[dbn] = *db;
+ mx->mx_dbxs[dbn].md_name.mv_data = NODEKEY(node);
+ mx->mx_dbxs[dbn].md_name.mv_size = node->mn_ksize;
+ mx->mx_txn.mt_next_pgno = txn->mt_next_pgno;
+ mx->mx_txn.mt_oldest = txn->mt_oldest;
+ mx->mx_txn.mt_u = txn->mt_u;
}
static void
mdb_xcursor_fini(MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx)
{
+ txn->mt_next_pgno = mx->mx_txn.mt_next_pgno;
+ txn->mt_oldest = mx->mx_txn.mt_oldest;
+ txn->mt_u = mx->mx_txn.mt_u;
txn->mt_dbs[0] = mx->mx_dbs[0];
txn->mt_dbs[1] = mx->mx_dbs[1];
txn->mt_dbxs[0].md_dirty = mx->mx_dbxs[0].md_dirty;
ovpages = OVPAGES(NODEDSZ(leaf), txn->mt_env->me_psize);
for (i=0; i<ovpages; i++) {
DPRINTF("freed ov page %lu", pg);
- mdb_idl_insert(txn->mt_free_pgs, pg);
+ mdb_midl_insert(txn->mt_free_pgs, pg);
pg++;
}
}
MDB_pageparent mp2;
mdb_xcursor_init0(txn, dbi, &mx);
- mdb_xcursor_init1(txn, dbi, &mx, NODEDATA(leaf));
+ mdb_xcursor_init1(txn, dbi, &mx, leaf);
if (flags == MDB_DEL_DUP) {
rc = mdb_del(&mx.mx_txn, mx.mx_cursor.mc_dbi, data, NULL, 0);
mdb_xcursor_fini(txn, dbi, &mx);
while (parent != NULL) {
for (i=0; i<NUMKEYS(top->mp_page); i++) {
ni = NODEPTR(top->mp_page, i);
- mdb_idl_insert(txn->mt_free_pgs, ni->mn_pgno);
+ mdb_midl_insert(txn->mt_free_pgs, ni->mn_pgno);
}
if (parent) {
parent->mp_ki++;
}
}
}
- mdb_idl_insert(txn->mt_free_pgs, mx.mx_txn.mt_dbs[mx.mx_cursor.mc_dbi].md_root);
+ mdb_midl_insert(txn->mt_free_pgs, mx.mx_txn.mt_dbs[mx.mx_cursor.mc_dbi].md_root);
}
}
}
rdata = data;
}
- if (SIZELEFT(mpp.mp_page) < mdb_leaf_size(txn->mt_env, key, data)) {
- rc = mdb_split(txn, dbi, &mpp.mp_page, &ki, key, data, P_INVALID);
+ if (SIZELEFT(mpp.mp_page) < mdb_leaf_size(txn->mt_env, key, rdata)) {
+ rc = mdb_split(txn, dbi, &mpp.mp_page, &ki, key, rdata, P_INVALID);
} else {
/* There is room already in this leaf page. */
- rc = mdb_add_node(txn, dbi, mpp.mp_page, ki, key, data, 0, 0);
+ rc = mdb_add_node(txn, dbi, mpp.mp_page, ki, key, rdata, 0, 0);
}
if (rc != MDB_SUCCESS)
leaf = NODEPTR(mpp.mp_page, ki);
put_sub:
mdb_xcursor_init0(txn, dbi, &mx);
- mdb_xcursor_init1(txn, dbi, &mx, NODEDATA(leaf));
+ mdb_xcursor_init1(txn, dbi, &mx, leaf);
xdata.mv_size = 0;
xdata.mv_data = "";
if (flags == MDB_NODUPDATA)
}
int
-mdbenv_get_flags(MDB_env *env, unsigned int *arg)
+mdb_env_get_flags(MDB_env *env, unsigned int *arg)
{
if (!env || !arg)
return EINVAL;
}
int
-mdbenv_get_path(MDB_env *env, const char **arg)
+mdb_env_get_path(MDB_env *env, const char **arg)
{
if (!env || !arg)
return EINVAL;
return MDB_SUCCESS;
}
int
-mdbenv_stat(MDB_env *env, MDB_stat *arg)
+mdb_env_stat(MDB_env *env, MDB_stat *arg)
{
if (env == NULL || arg == NULL)
return EINVAL;
projects / openldap / blobdiff