#include <stdlib.h>
#include <string.h>
#include <time.h>
+
+#ifdef _MSC_VER
+#include <io.h>
+typedef SSIZE_T ssize_t;
+#else
#include <unistd.h>
+#endif
#if defined(__sun) || defined(ANDROID)
/* Most platforms have posix_memalign, older may only have memalign */
#define MDB_OWNERDEAD EOWNERDEAD /**< #LOCK_MUTEX0() result if dead owner */
#endif
-#ifdef MDB_OWNERDEAD
+#ifdef __GLIBC__
+#define GLIBC_VER ((__GLIBC__ << 16 )| __GLIBC_MINOR__)
+#endif
+/** Some platforms define the EOWNERDEAD error code
+ * even though they don't support Robust Mutexes.
+ * Compile with -DMDB_USE_ROBUST=0, or use some other
+ * mechanism like -DMDB_USE_SYSV_SEM instead of
+ * -DMDB_USE_POSIX_MUTEX. (SysV semaphores are
+ * also Robust, but some systems don't support them
+ * either.)
+ */
+#ifndef MDB_USE_ROBUST
+/* Android currently lacks Robust Mutex support. So does glibc < 2.4. */
+# if defined(MDB_USE_POSIX_MUTEX) && (defined(ANDROID) || \
+ (defined(__GLIBC__) && GLIBC_VER < 0x020004))
+# define MDB_USE_ROBUST 0
+# else
+# define MDB_USE_ROBUST 1
+/* glibc < 2.10 only provided _np API */
+# if defined(__GLIBC__) && GLIBC_VER < 0x02000a
+# define PTHREAD_MUTEX_ROBUST PTHREAD_MUTEX_ROBUST_NP
+# define pthread_mutexattr_setrobust(attr, flag) pthread_mutexattr_setrobust_np(attr, flag)
+# define pthread_mutex_consistent(mutex) pthread_mutex_consistent_np(mutex)
+# endif
+# endif
+#endif /* MDB_USE_ROBUST */
+
+#if defined(MDB_OWNERDEAD) && MDB_USE_ROBUST
#define MDB_ROBUST_SUPPORTED 1
#endif
*
* @note If O_DSYNC is undefined but exists in /usr/include,
* preferably set some compiler flag to get the definition.
- * Otherwise compile with the less efficient -DMDB_DSYNC=O_SYNC.
*/
#ifndef MDB_DSYNC
+# ifdef O_DSYNC
# define MDB_DSYNC O_DSYNC
+# else
+# define MDB_DSYNC O_SYNC
+# endif
#endif
#endif
#define FREE_DBI 0
/** Handle for the default DB. */
#define MAIN_DBI 1
+ /** Number of DBs in metapage (free and main) - also hardcoded elsewhere */
+#define CORE_DBS 2
+
+ /** Number of meta pages - also hardcoded elsewhere */
+#define NUM_METAS 2
/** Meta page content.
* A meta page is the start point for accessing a database snapshot.
uint32_t mm_version;
void *mm_address; /**< address for fixed mapping */
size_t mm_mapsize; /**< size of mmap region */
- MDB_db mm_dbs[2]; /**< first is free space, 2nd is main db */
+ MDB_db mm_dbs[CORE_DBS]; /**< first is free space, 2nd is main db */
/** The size of pages used in this DB */
-#define mm_psize mm_dbs[0].md_pad
+#define mm_psize mm_dbs[FREE_DBI].md_pad
/** Any persistent environment flags. @ref mdb_env */
-#define mm_flags mm_dbs[0].md_flags
+#define mm_flags mm_dbs[FREE_DBI].md_flags
pgno_t mm_last_pg; /**< last used page in file */
volatile txnid_t mm_txnid; /**< txnid that committed this page */
} MDB_meta;
*/
struct MDB_txn {
MDB_txn *mt_parent; /**< parent of a nested txn */
- MDB_txn *mt_child; /**< nested txn under this txn */
+ /** Nested txn under this txn, set together with flag #MDB_TXN_HAS_CHILD */
+ MDB_txn *mt_child;
pgno_t mt_next_pgno; /**< next unallocated page */
/** The ID of this transaction. IDs are integers incrementing from 1.
* Only committed write transactions increment the ID. If a transaction
MDB_cursor **mt_cursors;
/** Array of flags for each DB */
unsigned char *mt_dbflags;
- /** Number of DB records in use. This number only ever increments;
- * we don't decrement it when individual DB handles are closed.
+ /** Number of DB records in use, or 0 when the txn is finished.
+ * This number only ever increments until the txn finishes; we
+ * don't decrement it when individual DB handles are closed.
*/
MDB_dbi mt_numdbs;
#define MDB_TXN_RDONLY MDB_RDONLY /**< read-only transaction */
/* internal txn flags */
#define MDB_TXN_WRITEMAP MDB_WRITEMAP /**< copy of #MDB_env flag in writers */
+#define MDB_TXN_FINISHED 0x01 /**< txn is finished or never began */
#define MDB_TXN_ERROR 0x02 /**< txn is unusable after an error */
#define MDB_TXN_DIRTY 0x04 /**< must write, even if dirty list is empty */
#define MDB_TXN_SPILLS 0x08 /**< txn or a parent has spilled pages */
+#define MDB_TXN_HAS_CHILD 0x10 /**< txn has an #MDB_txn.%mt_child */
+ /** most operations on the txn are currently illegal */
+#define MDB_TXN_BLOCKED (MDB_TXN_FINISHED|MDB_TXN_ERROR|MDB_TXN_HAS_CHILD)
/** @} */
unsigned int mt_flags; /**< @ref mdb_txn */
/** #dirty_list room: Array size - \#dirty pages visible to this txn.
char *me_path; /**< path to the DB files */
char *me_map; /**< the memory map of the data file */
MDB_txninfo *me_txns; /**< the memory map of the lock file or NULL */
- MDB_meta *me_metas[2]; /**< pointers to the two meta pages */
+ MDB_meta *me_metas[NUM_METAS]; /**< pointers to the two meta pages */
void *me_pbuf; /**< scratch area for DUPSORT put() */
MDB_txn *me_txn; /**< current write transaction */
MDB_txn *me_txn0; /**< prealloc'd write transaction */
static int mdb_page_new(MDB_cursor *mc, uint32_t flags, int num, MDB_page **mp);
static int mdb_page_touch(MDB_cursor *mc);
+#define MDB_END_NAMES {"committed", "empty-commit", "abort", "reset", \
+ "reset-tmp", "fail-begin", "fail-beginchild"}
+enum {
+ /* mdb_txn_end operation number, for logging */
+ MDB_END_COMMITTED, MDB_END_EMPTY_COMMIT, MDB_END_ABORT, MDB_END_RESET,
+ MDB_END_RESET_TMP, MDB_END_FAIL_BEGIN, MDB_END_FAIL_BEGINCHILD
+};
+#define MDB_END_OPMASK 0x0F /**< mask for #mdb_txn_end() operation number */
+#define MDB_END_UPDATE 0x10 /**< update env state (DBIs) */
+#define MDB_END_FREE 0x20 /**< free txn unless it is #MDB_env.%me_txn0 */
+#define MDB_END_SLOT MDB_NOTLS /**< release any reader slot if #MDB_NOTLS */
+static void mdb_txn_end(MDB_txn *txn, unsigned mode);
+
static int mdb_page_get(MDB_txn *txn, pgno_t pgno, MDB_page **mp, int *lvl);
static int mdb_page_search_root(MDB_cursor *mc,
MDB_val *key, int modify);
pgno_t newpgno, unsigned int nflags);
static int mdb_env_read_header(MDB_env *env, MDB_meta *meta);
-static int mdb_env_pick_meta(const MDB_env *env);
+static MDB_meta *mdb_env_pick_meta(const MDB_env *env);
static int mdb_env_write_meta(MDB_txn *txn);
#ifdef MDB_USE_POSIX_MUTEX /* Drop unused excl arg */
# define mdb_env_close0(env, excl) mdb_env_close1(env)
"MDB_MAP_RESIZED: Database contents grew beyond environment mapsize",
"MDB_INCOMPATIBLE: Operation and DB incompatible, or DB flags changed",
"MDB_BAD_RSLOT: Invalid reuse of reader locktable slot",
- "MDB_BAD_TXN: Transaction cannot recover - it must be aborted",
+ "MDB_BAD_TXN: Transaction must abort, has a child, or is invalid",
"MDB_BAD_VALSIZE: Unsupported size of key/DB name/data, or wrong DUPFIXED size",
"MDB_BAD_DBI: The specified DBI handle was closed/changed unexpectedly",
};
;
}
buf[0] = 0;
- FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM |
+ FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM |
FORMAT_MESSAGE_IGNORE_INSERTS,
NULL, err, 0, ptr, sizeof(buf), (va_list *)pad);
return ptr;
/** assert(3) variant in cursor context */
#define mdb_cassert(mc, expr) mdb_assert0((mc)->mc_txn->mt_env, expr, #expr)
/** assert(3) variant in transaction context */
-#define mdb_tassert(mc, expr) mdb_assert0((txn)->mt_env, expr, #expr)
+#define mdb_tassert(txn, expr) mdb_assert0((txn)->mt_env, expr, #expr)
/** assert(3) variant in environment context */
#define mdb_eassert(env, expr) mdb_assert0(env, expr, #expr)
mdb_tassert(txn, rc == MDB_NOTFOUND);
}
}
- if (freecount + count + 2 /* metapages */ != txn->mt_next_pgno) {
+ if (freecount + count + NUM_METAS != txn->mt_next_pgno) {
fprintf(stderr, "audit: %lu freecount: %lu count: %lu total: %lu next_pgno: %lu\n",
- txn->mt_txnid, freecount, count+2, freecount+count+2, txn->mt_next_pgno);
+ txn->mt_txnid, freecount, count+NUM_METAS,
+ freecount+count+NUM_METAS, txn->mt_next_pgno);
}
}
#endif
/* Estimate how much space this op will take */
i = m0->mc_db->md_depth;
/* Named DBs also dirty the main DB */
- if (m0->mc_dbi > MAIN_DBI)
+ if (m0->mc_dbi >= CORE_DBS)
i += txn->mt_dbs[MAIN_DBI].md_depth;
/* For puts, roughly factor in the key+data size */
if (key)
}
}
-#if !(MDB_DEBUG)
-#define mdb_txn_reset0(txn, act) mdb_txn_reset0(txn)
-#endif
-static void
-mdb_txn_reset0(MDB_txn *txn, const char *act);
-
#if !(MDB_PIDLOCK) /* Currently the same as defined(_WIN32) */
enum Pidlock_op {
Pidset, Pidcheck
if ((flags &= MDB_TXN_RDONLY) != 0) {
if (!ti) {
- meta = env->me_metas[ mdb_env_pick_meta(env) ];
+ meta = mdb_env_pick_meta(env);
txn->mt_txnid = meta->mm_txnid;
txn->mt_u.reader = NULL;
} else {
txn->mt_u.reader = r;
meta = env->me_metas[txn->mt_txnid & 1];
}
- txn->mt_dbxs = env->me_dbxs; /* mostly static anyway */
+
} else {
/* Not yet touching txn == env->me_txn0, it may be active */
if (ti) {
txn->mt_txnid = ti->mti_txnid;
meta = env->me_metas[txn->mt_txnid & 1];
} else {
- meta = env->me_metas[ mdb_env_pick_meta(env) ];
+ meta = mdb_env_pick_meta(env);
txn->mt_txnid = meta->mm_txnid;
}
txn->mt_txnid++;
}
/* Copy the DB info and flags */
- memcpy(txn->mt_dbs, meta->mm_dbs, 2 * sizeof(MDB_db));
+ memcpy(txn->mt_dbs, meta->mm_dbs, CORE_DBS * sizeof(MDB_db));
/* Moved to here to avoid a data race in read TXNs */
txn->mt_next_pgno = meta->mm_last_pg+1;
/* Setup db info */
txn->mt_numdbs = env->me_numdbs;
- for (i=2; i<txn->mt_numdbs; i++) {
+ for (i=CORE_DBS; i<txn->mt_numdbs; i++) {
x = env->me_dbflags[i];
txn->mt_dbs[i].md_flags = x & PERSISTENT_FLAGS;
txn->mt_dbflags[i] = (x & MDB_VALID) ? DB_VALID|DB_USRVALID|DB_STALE : 0;
txn->mt_dbflags[MAIN_DBI] = DB_VALID|DB_USRVALID;
txn->mt_dbflags[FREE_DBI] = DB_VALID;
- if (env->me_maxpg < txn->mt_next_pgno) {
- mdb_txn_reset0(txn, "renew0-mapfail");
- if (new_notls) {
- txn->mt_u.reader->mr_pid = 0;
- txn->mt_u.reader = NULL;
- }
- return MDB_MAP_RESIZED;
+ if (env->me_flags & MDB_FATAL_ERROR) {
+ DPUTS("environment had fatal error, must shutdown!");
+ rc = MDB_PANIC;
+ } else if (env->me_maxpg < txn->mt_next_pgno) {
+ rc = MDB_MAP_RESIZED;
+ } else {
+ return MDB_SUCCESS;
}
-
- return MDB_SUCCESS;
+ mdb_txn_end(txn, new_notls /*0 or MDB_END_SLOT*/ | MDB_END_FAIL_BEGIN);
+ return rc;
}
int
{
int rc;
- if (!txn || txn->mt_dbxs) /* A reset txn has mt_dbxs==NULL */
+ if (!txn || !F_ISSET(txn->mt_flags, MDB_TXN_RDONLY|MDB_TXN_FINISHED))
return EINVAL;
- if (txn->mt_env->me_flags & MDB_FATAL_ERROR) {
- DPUTS("environment had fatal error, must shutdown!");
- return MDB_PANIC;
- }
-
rc = mdb_txn_renew0(txn);
if (rc == MDB_SUCCESS) {
DPRINTF(("renew txn %"Z"u%c %p on mdbenv %p, root page %"Z"u",
flags &= MDB_TXN_BEGIN_FLAGS;
flags |= env->me_flags & MDB_WRITEMAP;
- if (env->me_flags & MDB_FATAL_ERROR) {
- DPUTS("environment had fatal error, must shutdown!");
- return MDB_PANIC;
- }
if (env->me_flags & MDB_RDONLY & ~flags) /* write txn in RDONLY env */
return EACCES;
if (parent) {
/* Nested transactions: Max 1 child, write txns only, no writemap */
flags |= parent->mt_flags;
- if (parent->mt_child ||
- (flags & (MDB_RDONLY|MDB_WRITEMAP|MDB_TXN_ERROR)))
- {
+ if (flags & (MDB_RDONLY|MDB_WRITEMAP|MDB_TXN_BLOCKED)) {
return (parent->mt_flags & MDB_TXN_RDONLY) ? EINVAL : MDB_BAD_TXN;
}
/* Child txns save MDB_pgstate and use own copy of cursors */
DPRINTF(("calloc: %s", strerror(errno)));
return ENOMEM;
}
+ txn->mt_dbxs = env->me_dbxs; /* static */
txn->mt_dbs = (MDB_db *) ((char *)txn + tsize);
txn->mt_dbflags = (unsigned char *)txn + size - env->me_maxdbs;
txn->mt_flags = flags;
txn->mt_u.dirty_list[0].mid = 0;
txn->mt_spill_pgs = NULL;
txn->mt_next_pgno = parent->mt_next_pgno;
+ parent->mt_flags |= MDB_TXN_HAS_CHILD;
parent->mt_child = txn;
txn->mt_parent = parent;
txn->mt_numdbs = parent->mt_numdbs;
- txn->mt_dbxs = parent->mt_dbxs;
memcpy(txn->mt_dbs, parent->mt_dbs, txn->mt_numdbs * sizeof(MDB_db));
/* Copy parent's mt_dbflags, but clear DB_NEW */
for (i=0; i<txn->mt_numdbs; i++)
if (!rc)
rc = mdb_cursor_shadow(parent, txn);
if (rc)
- mdb_txn_reset0(txn, "beginchild-fail");
+ mdb_txn_end(txn, MDB_END_FAIL_BEGINCHILD);
} else { /* MDB_RDONLY */
txn->mt_dbiseqs = env->me_dbiseqs;
renew:
MDB_env *env = txn->mt_env;
unsigned char *tdbflags = txn->mt_dbflags;
- for (i = n; --i >= 2;) {
+ for (i = n; --i >= CORE_DBS;) {
if (tdbflags[i] & DB_NEW) {
if (keep) {
env->me_dbflags[i] = txn->mt_dbs[i].md_flags | MDB_VALID;
env->me_numdbs = n;
}
-/** Common code for #mdb_txn_reset() and #mdb_txn_abort().
+/** End a transaction, except successful commit of a nested transaction.
* May be called twice for readonly txns: First reset it, then abort.
- * @param[in] txn the transaction handle to reset
- * @param[in] act why the transaction is being reset
+ * @param[in] txn the transaction handle to end
+ * @param[in] mode why and how to end the transaction
*/
static void
-mdb_txn_reset0(MDB_txn *txn, const char *act)
+mdb_txn_end(MDB_txn *txn, unsigned mode)
{
MDB_env *env = txn->mt_env;
+#if MDB_DEBUG
+ static const char *const names[] = MDB_END_NAMES;
+#endif
- /* Close any DBI handles opened in this txn */
- mdb_dbis_update(txn, 0);
+ /* Export or close DBI handles opened in this txn */
+ mdb_dbis_update(txn, mode & MDB_END_UPDATE);
DPRINTF(("%s txn %"Z"u%c %p on mdbenv %p, root page %"Z"u",
- act, txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w',
+ names[mode & MDB_END_OPMASK],
+ txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w',
(void *) txn, (void *)env, txn->mt_dbs[MAIN_DBI].md_root));
if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
if (txn->mt_u.reader) {
txn->mt_u.reader->mr_txnid = (txnid_t)-1;
- if (!(env->me_flags & MDB_NOTLS))
+ if (!(env->me_flags & MDB_NOTLS)) {
txn->mt_u.reader = NULL; /* txn does not own reader */
+ } else if (mode & MDB_END_SLOT) {
+ txn->mt_u.reader->mr_pid = 0;
+ txn->mt_u.reader = NULL;
+ } /* else txn owns the slot until it does MDB_END_SLOT */
}
- txn->mt_numdbs = 0; /* close nothing if called again */
- txn->mt_dbxs = NULL; /* mark txn as reset */
- } else {
+ txn->mt_numdbs = 0; /* prevent further DBI activity */
+ txn->mt_flags |= MDB_TXN_FINISHED;
+
+ } else if (!F_ISSET(txn->mt_flags, MDB_TXN_FINISHED)) {
pgno_t *pghead = env->me_pghead;
- mdb_cursors_close(txn, 0);
+ if (!(mode & MDB_END_UPDATE)) /* !(already closed cursors) */
+ mdb_cursors_close(txn, 0);
if (!(env->me_flags & MDB_WRITEMAP)) {
mdb_dlist_free(txn);
}
+ txn->mt_numdbs = 0;
+ txn->mt_flags = MDB_TXN_FINISHED;
+
if (!txn->mt_parent) {
mdb_midl_shrink(&txn->mt_free_pgs);
env->me_free_pgs = txn->mt_free_pgs;
env->me_pglast = 0;
env->me_txn = NULL;
+ mode = 0; /* txn == env->me_txn0, do not free() it */
+
/* The writer mutex was locked in mdb_txn_begin. */
if (env->me_txns)
UNLOCK_MUTEX(env->me_wmutex);
} else {
txn->mt_parent->mt_child = NULL;
+ txn->mt_parent->mt_flags &= ~MDB_TXN_HAS_CHILD;
env->me_pgstate = ((MDB_ntxn *)txn)->mnt_pgstate;
mdb_midl_free(txn->mt_free_pgs);
mdb_midl_free(txn->mt_spill_pgs);
mdb_midl_free(pghead);
}
+
+ if (mode & MDB_END_FREE)
+ free(txn);
}
void
if (!(txn->mt_flags & MDB_TXN_RDONLY))
return;
- mdb_txn_reset0(txn, "reset");
+ mdb_txn_end(txn, MDB_END_RESET);
}
void
if (txn->mt_child)
mdb_txn_abort(txn->mt_child);
- mdb_txn_reset0(txn, "abort");
- /* Free reader slot tied to this txn (if MDB_NOTLS && writable FS) */
- if ((txn->mt_flags & MDB_TXN_RDONLY) && txn->mt_u.reader)
- txn->mt_u.reader->mr_pid = 0;
-
- if (txn != txn->mt_env->me_txn0)
- free(txn);
+ mdb_txn_end(txn, MDB_END_ABORT|MDB_END_SLOT|MDB_END_FREE);
}
/** Save the freelist as of this transaction to the freeDB.
mdb_txn_commit(MDB_txn *txn)
{
int rc;
- unsigned int i;
+ unsigned int i, end_mode;
MDB_env *env;
if (txn == NULL)
return EINVAL;
+ /* mdb_txn_end() mode for a commit which writes nothing */
+ end_mode = MDB_END_EMPTY_COMMIT|MDB_END_UPDATE|MDB_END_SLOT|MDB_END_FREE;
+
if (txn->mt_child) {
rc = mdb_txn_commit(txn->mt_child);
if (rc)
env = txn->mt_env;
if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
- mdb_dbis_update(txn, 1);
- txn->mt_numdbs = 2; /* so txn_abort() doesn't close any new handles */
- mdb_txn_abort(txn);
- return MDB_SUCCESS;
+ goto done;
}
- if (F_ISSET(txn->mt_flags, MDB_TXN_ERROR)) {
- DPUTS("error flag is set, can't commit");
+ if (txn->mt_flags & (MDB_TXN_FINISHED|MDB_TXN_ERROR)) {
+ DPUTS("txn has failed/finished, can't commit");
if (txn->mt_parent)
txn->mt_parent->mt_flags |= MDB_TXN_ERROR;
rc = MDB_BAD_TXN;
/* Update parent's DB table. */
memcpy(parent->mt_dbs, txn->mt_dbs, txn->mt_numdbs * sizeof(MDB_db));
parent->mt_numdbs = txn->mt_numdbs;
- parent->mt_dbflags[0] = txn->mt_dbflags[0];
- parent->mt_dbflags[1] = txn->mt_dbflags[1];
- for (i=2; i<txn->mt_numdbs; i++) {
+ parent->mt_dbflags[FREE_DBI] = txn->mt_dbflags[FREE_DBI];
+ parent->mt_dbflags[MAIN_DBI] = txn->mt_dbflags[MAIN_DBI];
+ for (i=CORE_DBS; i<txn->mt_numdbs; i++) {
/* preserve parent's DB_NEW status */
x = parent->mt_dbflags[i] & DB_NEW;
parent->mt_dbflags[i] = txn->mt_dbflags[i] | x;
txn->mt_txnid, (void*)txn, (void*)env, txn->mt_dbs[MAIN_DBI].md_root));
/* Update DB root pointers */
- if (txn->mt_numdbs > 2) {
+ if (txn->mt_numdbs > CORE_DBS) {
MDB_cursor mc;
MDB_dbi i;
MDB_val data;
data.mv_size = sizeof(MDB_db);
mdb_cursor_init(&mc, txn, MAIN_DBI, NULL);
- for (i = 2; i < txn->mt_numdbs; i++) {
+ for (i = CORE_DBS; i < txn->mt_numdbs; i++) {
if (txn->mt_dbflags[i] & DB_DIRTY) {
if (TXN_DBI_CHANGED(txn, i)) {
rc = MDB_BAD_DBI;
mdb_midl_free(env->me_pghead);
env->me_pghead = NULL;
mdb_midl_shrink(&txn->mt_free_pgs);
- env->me_free_pgs = txn->mt_free_pgs;
#if (MDB_DEBUG) > 2
mdb_audit(txn);
(rc = mdb_env_sync(env, 0)) ||
(rc = mdb_env_write_meta(txn)))
goto fail;
-
- /* Free P_LOOSE pages left behind in dirty_list */
- if (!(env->me_flags & MDB_WRITEMAP))
- mdb_dlist_free(txn);
+ end_mode = MDB_END_COMMITTED|MDB_END_UPDATE;
done:
- env->me_pglast = 0;
- env->me_txn = NULL;
- mdb_dbis_update(txn, 1);
-
- if (env->me_txns)
- UNLOCK_MUTEX(env->me_wmutex);
- if (txn != env->me_txn0)
- free(txn);
-
+ mdb_txn_end(txn, end_mode);
return MDB_SUCCESS;
fail:
* Read both meta pages so we can use the latest one.
*/
- for (i=off=0; i<2; i++, off = meta->mm_psize) {
+ for (i=off=0; i<NUM_METAS; i++, off += meta->mm_psize) {
#ifdef _WIN32
DWORD len;
OVERLAPPED ov;
meta->mm_version = MDB_DATA_VERSION;
meta->mm_mapsize = env->me_mapsize;
meta->mm_psize = env->me_psize;
- meta->mm_last_pg = 1;
+ meta->mm_last_pg = NUM_METAS-1;
meta->mm_flags = env->me_flags & 0xffff;
- meta->mm_flags |= MDB_INTEGERKEY;
- meta->mm_dbs[0].md_root = P_INVALID;
- meta->mm_dbs[1].md_root = P_INVALID;
+ meta->mm_flags |= MDB_INTEGERKEY; /* this is mm_dbs[FREE_DBI].md_flags */
+ meta->mm_dbs[FREE_DBI].md_root = P_INVALID;
+ meta->mm_dbs[MAIN_DBI].md_root = P_INVALID;
}
/** Write the environment parameters of a freshly created DB environment.
psize = env->me_psize;
- p = calloc(2, psize);
+ p = calloc(NUM_METAS, psize);
if (!p)
return ENOMEM;
q->mp_flags = P_META;
*(MDB_meta *)METADATA(q) = *meta;
- DO_PWRITE(rc, env->me_fd, p, psize * 2, len, 0);
+ DO_PWRITE(rc, env->me_fd, p, psize * NUM_METAS, len, 0);
if (!rc)
rc = ErrCode();
- else if ((unsigned) len == psize * 2)
+ else if ((unsigned) len == psize * NUM_METAS)
rc = MDB_SUCCESS;
else
rc = ENOSPC;
if (flags & MDB_WRITEMAP) {
mp->mm_mapsize = mapsize;
- mp->mm_dbs[0] = txn->mt_dbs[0];
- mp->mm_dbs[1] = txn->mt_dbs[1];
+ mp->mm_dbs[FREE_DBI] = txn->mt_dbs[FREE_DBI];
+ mp->mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI];
mp->mm_last_pg = txn->mt_next_pgno - 1;
#if (__GNUC__ * 100 + __GNUC_MINOR__ >= 404) && /* TODO: portability */ \
!(defined(__i386__) || defined(__x86_64__))
if (!(flags & (MDB_NOMETASYNC|MDB_NOSYNC))) {
unsigned meta_size = env->me_psize;
rc = (env->me_flags & MDB_MAPASYNC) ? MS_ASYNC : MS_SYNC;
- ptr = env->me_map;
- if (toggle) {
+ ptr = (char *)mp - PAGEHDRSZ;
#ifndef _WIN32 /* POSIX msync() requires ptr = start of OS page */
- if (meta_size < env->me_os_psize)
- meta_size += meta_size;
- else
+ r2 = (ptr - env->me_map) & (env->me_os_psize - 1);
+ ptr -= r2;
+ meta_size += r2;
#endif
- ptr += meta_size;
- }
if (MDB_MSYNC(ptr, meta_size, rc)) {
rc = ErrCode();
goto fail;
}
goto done;
}
- metab.mm_txnid = env->me_metas[toggle]->mm_txnid;
- metab.mm_last_pg = env->me_metas[toggle]->mm_last_pg;
+ metab.mm_txnid = mp->mm_txnid;
+ metab.mm_last_pg = mp->mm_last_pg;
meta.mm_mapsize = mapsize;
- meta.mm_dbs[0] = txn->mt_dbs[0];
- meta.mm_dbs[1] = txn->mt_dbs[1];
+ meta.mm_dbs[FREE_DBI] = txn->mt_dbs[FREE_DBI];
+ meta.mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI];
meta.mm_last_pg = txn->mt_next_pgno - 1;
meta.mm_txnid = txn->mt_txnid;
off = offsetof(MDB_meta, mm_mapsize);
ptr = (char *)&meta + off;
len = sizeof(MDB_meta) - off;
- if (toggle)
- off += env->me_psize;
- off += PAGEHDRSZ;
+ off += (char *)mp - env->me_map;
/* Write to the SYNC fd */
mfd = (flags & (MDB_NOSYNC|MDB_NOMETASYNC)) ? env->me_fd : env->me_mfd;
/** Check both meta pages to see which one is newer.
* @param[in] env the environment handle
- * @return meta toggle (0 or 1).
+ * @return newest #MDB_meta.
*/
-static int
+static MDB_meta *
mdb_env_pick_meta(const MDB_env *env)
{
- return (env->me_metas[0]->mm_txnid < env->me_metas[1]->mm_txnid);
+ MDB_meta *const *metas = env->me_metas;
+ return metas[ metas[0]->mm_txnid < metas[1]->mm_txnid ];
}
int ESECT
return ENOMEM;
e->me_maxreaders = DEFAULT_READERS;
- e->me_maxdbs = e->me_numdbs = 2;
+ e->me_maxdbs = e->me_numdbs = CORE_DBS;
e->me_fd = INVALID_HANDLE_VALUE;
e->me_lfd = INVALID_HANDLE_VALUE;
e->me_mfd = INVALID_HANDLE_VALUE;
void *old;
if (env->me_txn)
return EINVAL;
- meta = env->me_metas[mdb_env_pick_meta(env)];
+ meta = mdb_env_pick_meta(env);
if (!size)
size = meta->mm_mapsize;
{
{
if (env->me_map)
return EINVAL;
- env->me_maxdbs = dbs + 2; /* Named databases + main and free DB */
+ env->me_maxdbs = dbs + CORE_DBS;
return MDB_SUCCESS;
}
#if MDB_DEBUG
{
- int toggle = mdb_env_pick_meta(env);
- MDB_db *db = &env->me_metas[toggle]->mm_dbs[MAIN_DBI];
+ MDB_meta *meta = mdb_env_pick_meta(env);
+ MDB_db *db = &meta->mm_dbs[MAIN_DBI];
DPRINTF(("opened database version %u, pagesize %u",
- env->me_metas[0]->mm_version, env->me_psize));
- DPRINTF(("using meta page %d", toggle));
+ meta->mm_version, env->me_psize));
+ DPRINTF(("using meta page %d", (int) (meta->mm_txnid & 1)));
DPRINTF(("depth: %u", db->md_depth));
DPRINTF(("entries: %"Z"u", db->md_entries));
DPRINTF(("branch pages: %"Z"u", db->md_branch_pages));
static int ESECT
mdb_env_share_locks(MDB_env *env, int *excl)
{
- int rc = 0, toggle = mdb_env_pick_meta(env);
+ int rc = 0;
+ MDB_meta *meta = mdb_env_pick_meta(env);
- env->me_txns->mti_txnid = env->me_metas[toggle]->mm_txnid;
+ env->me_txns->mti_txnid = meta->mm_txnid;
#ifdef _WIN32
{
off_t size, rsize;
#ifdef _WIN32
- env->me_lfd = CreateFile(lpath, GENERIC_READ|GENERIC_WRITE,
+ env->me_lfd = CreateFileA(lpath, GENERIC_READ|GENERIC_WRITE,
FILE_SHARE_READ|FILE_SHARE_WRITE, NULL, OPEN_ALWAYS,
FILE_ATTRIBUTE_NORMAL, NULL);
#else
mdb_hash_enc(&val, encbuf);
sprintf(env->me_txns->mti_rmname, "Global\\MDBr%s", encbuf);
sprintf(env->me_txns->mti_wmname, "Global\\MDBw%s", encbuf);
- env->me_rmutex = CreateMutex(&mdb_all_sa, FALSE, env->me_txns->mti_rmname);
+ env->me_rmutex = CreateMutexA(&mdb_all_sa, FALSE, env->me_txns->mti_rmname);
if (!env->me_rmutex) goto fail_errno;
- env->me_wmutex = CreateMutex(&mdb_all_sa, FALSE, env->me_txns->mti_wmname);
+ env->me_wmutex = CreateMutexA(&mdb_all_sa, FALSE, env->me_txns->mti_wmname);
if (!env->me_wmutex) goto fail_errno;
#elif defined(MDB_USE_POSIX_SEM)
struct stat stbuf;
goto fail;
}
#ifdef _WIN32
- env->me_rmutex = OpenMutex(SYNCHRONIZE, FALSE, env->me_txns->mti_rmname);
+ env->me_rmutex = OpenMutexA(SYNCHRONIZE, FALSE, env->me_txns->mti_rmname);
if (!env->me_rmutex) goto fail_errno;
- env->me_wmutex = OpenMutex(SYNCHRONIZE, FALSE, env->me_txns->mti_wmname);
+ env->me_wmutex = OpenMutexA(SYNCHRONIZE, FALSE, env->me_txns->mti_wmname);
if (!env->me_wmutex) goto fail_errno;
#elif defined(MDB_USE_POSIX_SEM)
env->me_rmutex = sem_open(env->me_txns->mti_rmname, 0);
len = OPEN_ALWAYS;
}
mode = FILE_ATTRIBUTE_NORMAL;
- env->me_fd = CreateFile(dpath, oflags, FILE_SHARE_READ|FILE_SHARE_WRITE,
+ env->me_fd = CreateFileA(dpath, oflags, FILE_SHARE_READ|FILE_SHARE_WRITE,
NULL, len, mode, NULL);
#else
if (F_ISSET(flags, MDB_RDONLY))
*/
#ifdef _WIN32
len = OPEN_EXISTING;
- env->me_mfd = CreateFile(dpath, oflags,
+ env->me_mfd = CreateFileA(dpath, oflags,
FILE_SHARE_READ|FILE_SHARE_WRITE, NULL, len,
mode | FILE_FLAG_WRITE_THROUGH, NULL);
#else
txn->mt_dbflags = (unsigned char *)(txn->mt_dbiseqs + env->me_maxdbs);
txn->mt_env = env;
txn->mt_dbxs = env->me_dbxs;
+ txn->mt_flags = MDB_TXN_FINISHED;
env->me_txn0 = txn;
} else {
rc = ENOMEM;
/* Doing this here since me_dbxs may not exist during mdb_env_close */
if (env->me_dbxs) {
- for (i = env->me_maxdbs; --i > MAIN_DBI; )
+ for (i = env->me_maxdbs; --i >= CORE_DBS; )
free(env->me_dbxs[i].md_name.mv_data);
free(env->me_dbxs);
}
/* Make sure the txn is still viable, then find the root from
* the txn's db table and set it as the root of the cursor's stack.
*/
- if (F_ISSET(mc->mc_txn->mt_flags, MDB_TXN_ERROR)) {
- DPUTS("transaction has failed, must abort");
+ if (mc->mc_txn->mt_flags & MDB_TXN_BLOCKED) {
+ DPUTS("transaction may not be used now");
return MDB_BAD_TXN;
} else {
/* Make sure we're using an up-to-date root */
if (!key || !data || !TXN_DBI_EXIST(txn, dbi, DB_USRVALID))
return EINVAL;
- if (txn->mt_flags & MDB_TXN_ERROR)
+ if (txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
mdb_cursor_init(&mc, txn, dbi, &mx);
if (mc == NULL)
return EINVAL;
- if (mc->mc_txn->mt_flags & MDB_TXN_ERROR)
+ if (mc->mc_txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
switch (op) {
{
int rc = MDB_SUCCESS;
- if (mc->mc_dbi > MAIN_DBI && !(*mc->mc_dbflag & DB_DIRTY)) {
+ if (mc->mc_dbi >= CORE_DBS && !(*mc->mc_dbflag & DB_DIRTY)) {
MDB_cursor mc2;
MDB_xcursor mcx;
if (TXN_DBI_CHANGED(mc->mc_txn, mc->mc_dbi))
nospill = flags & MDB_NOSPILL;
flags &= ~MDB_NOSPILL;
- if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_ERROR))
+ if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED))
return (mc->mc_txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN;
if (key->mv_size-1 >= ENV_MAXKEY(env))
* update branch key if there is a parent page
*/
if (mc->mc_top && !mc->mc_ki[mc->mc_top]) {
- unsigned short top = mc->mc_top;
+ unsigned short dtop = 1;
mc->mc_top--;
/* slot 0 is always an empty key, find real slot */
- while (mc->mc_top && !mc->mc_ki[mc->mc_top])
+ while (mc->mc_top && !mc->mc_ki[mc->mc_top]) {
mc->mc_top--;
+ dtop++;
+ }
if (mc->mc_ki[mc->mc_top])
rc2 = mdb_update_key(mc, key);
else
rc2 = MDB_SUCCESS;
- mc->mc_top = top;
+ mc->mc_top += dtop;
if (rc2)
return rc2;
}
MDB_page *mp;
int rc;
- if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_ERROR))
+ if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED))
return (mc->mc_txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN;
if (!(mc->mc_flags & C_INITIALIZED))
if (!ret || !TXN_DBI_EXIST(txn, dbi, DB_VALID))
return EINVAL;
- if (txn->mt_flags & MDB_TXN_ERROR)
+ if (txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
- /* Allow read access to the freelist */
- if (!dbi && !F_ISSET(txn->mt_flags, MDB_TXN_RDONLY))
+ if (dbi == FREE_DBI && !F_ISSET(txn->mt_flags, MDB_TXN_RDONLY))
return EINVAL;
if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT)
if ((mc->mc_flags & C_UNTRACK) || txn->mt_cursors)
return EINVAL;
- if (txn->mt_flags & MDB_TXN_ERROR)
+ if (txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
mdb_cursor_init(mc, txn, mc->mc_dbi, mc->mc_xcursor);
if (mc->mc_xcursor == NULL)
return MDB_INCOMPATIBLE;
- if (mc->mc_txn->mt_flags & MDB_TXN_ERROR)
+ if (mc->mc_txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
if (!(mc->mc_flags & C_INITIALIZED))
/* Adjust other cursors pointing to mp */
MDB_cursor *m2, *m3;
MDB_dbi dbi = csrc->mc_dbi;
- MDB_page *mp;
-
- mp = cdst->mc_pg[csrc->mc_top];
- for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) {
- if (csrc->mc_flags & C_SUB)
- m3 = &m2->mc_xcursor->mx_cursor;
- else
- m3 = m2;
- if (m3 == cdst) continue;
- if (m3->mc_pg[csrc->mc_top] == mp && m3->mc_ki[csrc->mc_top] >=
- cdst->mc_ki[csrc->mc_top]) {
- m3->mc_ki[csrc->mc_top]++;
+ MDB_page *mpd, *mps;
+
+ mps = csrc->mc_pg[csrc->mc_top];
+ /* If we're adding on the left, bump others up */
+ if (!cdst->mc_ki[csrc->mc_top]) {
+ mpd = cdst->mc_pg[csrc->mc_top];
+ for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) {
+ if (csrc->mc_flags & C_SUB)
+ m3 = &m2->mc_xcursor->mx_cursor;
+ else
+ m3 = m2;
+ if (m3 != cdst &&
+ m3->mc_pg[csrc->mc_top] == mpd &&
+ m3->mc_ki[csrc->mc_top] >= cdst->mc_ki[csrc->mc_top]) {
+ m3->mc_ki[csrc->mc_top]++;
+ }
+ if (m3 !=csrc &&
+ m3->mc_pg[csrc->mc_top] == mps &&
+ m3->mc_ki[csrc->mc_top] == csrc->mc_ki[csrc->mc_top]) {
+ m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top];
+ m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top];
+ m3->mc_ki[csrc->mc_top-1]++;
+ }
}
- }
-
- mp = csrc->mc_pg[csrc->mc_top];
- for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) {
- if (csrc->mc_flags & C_SUB)
- m3 = &m2->mc_xcursor->mx_cursor;
- else
- m3 = m2;
- if (m3 == csrc) continue;
- if (m3->mc_pg[csrc->mc_top] == mp && m3->mc_ki[csrc->mc_top] ==
- csrc->mc_ki[csrc->mc_top]) {
- m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top];
- m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top];
+ } else
+ /* Adding on the right, bump others down */
+ {
+ for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) {
+ if (csrc->mc_flags & C_SUB)
+ m3 = &m2->mc_xcursor->mx_cursor;
+ else
+ m3 = m2;
+ if (m3 == csrc) continue;
+ if (m3->mc_pg[csrc->mc_top] == mps) {
+ if (!m3->mc_ki[csrc->mc_top]) {
+ m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top];
+ m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top];
+ m3->mc_ki[csrc->mc_top-1]--;
+ } else {
+ m3->mc_ki[csrc->mc_top]--;
+ }
+ }
}
}
}
if ((rc = mdb_page_touch(cdst)))
return rc;
+ /* get dst page again now that we've touched it. */
+ pdst = cdst->mc_pg[cdst->mc_top];
+
/* Move all nodes from src to dst.
*/
j = nkeys = NUMKEYS(pdst);
/* Adjust other cursors pointing to mp */
MDB_cursor *m2, *m3;
MDB_dbi dbi = csrc->mc_dbi;
+ unsigned int top = csrc->mc_top;
for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) {
if (csrc->mc_flags & C_SUB)
m3 = m2;
if (m3 == csrc) continue;
if (m3->mc_snum < csrc->mc_snum) continue;
- if (m3->mc_pg[csrc->mc_top] == psrc) {
- m3->mc_pg[csrc->mc_top] = pdst;
- m3->mc_ki[csrc->mc_top] += nkeys;
+ if (m3->mc_pg[top] == psrc) {
+ m3->mc_pg[top] = pdst;
+ m3->mc_ki[top] += nkeys;
+ m3->mc_ki[top-1] = cdst->mc_ki[top-1];
}
}
}
m3 = &m2->mc_xcursor->mx_cursor;
else
m3 = m2;
- if (m3 == mc || m3->mc_snum < mc->mc_snum) continue;
+ if (m3 == mc) continue;
if (m3->mc_pg[0] == mp) {
- for (i=0; i<m3->mc_snum; i++) {
+ for (i=0; i<mc->mc_db->md_depth; i++) {
m3->mc_pg[i] = m3->mc_pg[i+1];
m3->mc_ki[i] = m3->mc_ki[i+1];
}
*/
if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= thresh && NUMKEYS(mn.mc_pg[mn.mc_top]) > minkeys) {
rc = mdb_node_move(&mn, mc);
- if (mc->mc_ki[mc->mc_top-1]) {
+ if (!mc->mc_ki[mc->mc_top]) {
+ /* if we inserted on left, bump position up */
oldki++;
}
} else {
if (!key || !TXN_DBI_EXIST(txn, dbi, DB_USRVALID))
return EINVAL;
- if (txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_ERROR))
+ if (txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED))
return (txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN;
if (!F_ISSET(txn->mt_dbs[dbi].md_flags, MDB_DUPSORT)) {
rp->mp_pad = mp->mp_pad;
DPRINTF(("new right sibling: page %"Z"u", rp->mp_pgno));
- if (mc->mc_snum < 2) {
+ /* Usually when splitting the root page, the cursor
+ * height is 1. But when called from mdb_update_key,
+ * the cursor height may be greater because it walks
+ * up the stack while finding the branch slot to update.
+ */
+ if (mc->mc_top < 1) {
if ((rc = mdb_page_new(mc, P_BRANCH, 1, &pp)))
goto done;
/* shift current top to make room for new parent */
- mc->mc_pg[1] = mc->mc_pg[0];
- mc->mc_ki[1] = mc->mc_ki[0];
+ for (i=mc->mc_snum; i>0; i--) {
+ mc->mc_pg[i] = mc->mc_pg[i-1];
+ mc->mc_ki[i] = mc->mc_ki[i-1];
+ }
mc->mc_pg[0] = pp;
mc->mc_ki[0] = 0;
mc->mc_db->md_root = pp->mp_pgno;
mc->mc_db->md_depth--;
goto done;
}
- mc->mc_snum = 2;
- mc->mc_top = 1;
+ mc->mc_snum++;
+ mc->mc_top++;
ptop = 0;
} else {
ptop = mc->mc_top-1;
rp->mp_upper -= ksize - sizeof(indx_t);
mc->mc_ki[mc->mc_top] = x;
mc->mc_pg[mc->mc_top] = rp;
+ mc->mc_ki[ptop]++;
}
} else {
int psize, nsize, k;
if (flags & ~(MDB_NOOVERWRITE|MDB_NODUPDATA|MDB_RESERVE|MDB_APPEND|MDB_APPENDDUP))
return EINVAL;
+ if (txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED))
+ return (txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN;
+
mdb_cursor_init(&mc, txn, dbi, &mx);
return mdb_cursor_put(&mc, key, data, flags);
}
my.mc_wlen[1] = 0;
my.mc_olen[0] = 0;
my.mc_olen[1] = 0;
- my.mc_next_pgno = 2;
+ my.mc_next_pgno = NUM_METAS;
my.mc_status = 0;
my.mc_new = 1;
my.mc_toggle = 0;
return rc;
mp = (MDB_page *)my.mc_wbuf[0];
- memset(mp, 0, 2*env->me_psize);
+ memset(mp, 0, NUM_METAS * env->me_psize);
mp->mp_pgno = 0;
mp->mp_flags = P_META;
mm = (MDB_meta *)METADATA(mp);
mdb_cursor_init(&mc, txn, FREE_DBI, NULL);
while ((rc = mdb_cursor_get(&mc, &key, &data, MDB_NEXT)) == 0)
freecount += *(MDB_ID *)data.mv_data;
- freecount += txn->mt_dbs[0].md_branch_pages +
- txn->mt_dbs[0].md_leaf_pages +
- txn->mt_dbs[0].md_overflow_pages;
+ freecount += txn->mt_dbs[FREE_DBI].md_branch_pages +
+ txn->mt_dbs[FREE_DBI].md_leaf_pages +
+ txn->mt_dbs[FREE_DBI].md_overflow_pages;
/* Set metapage 1 */
mm->mm_last_pg = txn->mt_next_pgno - freecount - 1;
- mm->mm_dbs[1] = txn->mt_dbs[1];
- if (mm->mm_last_pg > 1) {
- mm->mm_dbs[1].md_root = mm->mm_last_pg;
+ mm->mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI];
+ if (mm->mm_last_pg > NUM_METAS-1) {
+ mm->mm_dbs[MAIN_DBI].md_root = mm->mm_last_pg;
mm->mm_txnid = 1;
} else {
- mm->mm_dbs[1].md_root = P_INVALID;
+ mm->mm_dbs[MAIN_DBI].md_root = P_INVALID;
}
}
- my.mc_wlen[0] = env->me_psize * 2;
+ my.mc_wlen[0] = env->me_psize * NUM_METAS;
my.mc_txn = txn;
pthread_mutex_lock(&my.mc_mutex);
while(my.mc_new)
pthread_cond_wait(&my.mc_cond, &my.mc_mutex);
pthread_mutex_unlock(&my.mc_mutex);
- rc = mdb_env_cwalk(&my, &txn->mt_dbs[1].md_root, 0);
+ rc = mdb_env_cwalk(&my, &txn->mt_dbs[MAIN_DBI].md_root, 0);
if (rc == MDB_SUCCESS && my.mc_wlen[my.mc_toggle])
rc = mdb_env_cthr_toggle(&my, 1);
mdb_env_cthr_toggle(&my, -1);
if (env->me_txns) {
/* We must start the actual read txn after blocking writers */
- mdb_txn_reset0(txn, "reset-stage1");
+ mdb_txn_end(txn, MDB_END_RESET_TMP);
/* Temporarily block writers until we snapshot the meta pages */
wmutex = env->me_wmutex;
}
}
- wsize = env->me_psize * 2;
+ wsize = env->me_psize * NUM_METAS;
ptr = env->me_map;
w2 = wsize;
while (w2 > 0) {
* already in the OS cache.
*/
#ifdef _WIN32
- newfd = CreateFile(lpath, GENERIC_WRITE, 0, NULL, CREATE_NEW,
+ newfd = CreateFileA(lpath, GENERIC_WRITE, 0, NULL, CREATE_NEW,
FILE_FLAG_NO_BUFFERING|FILE_FLAG_WRITE_THROUGH, NULL);
#else
newfd = open(lpath, O_WRONLY|O_CREAT|O_EXCL, 0666);
int ESECT
mdb_env_stat(MDB_env *env, MDB_stat *arg)
{
- int toggle;
+ MDB_meta *meta;
if (env == NULL || arg == NULL)
return EINVAL;
- toggle = mdb_env_pick_meta(env);
+ meta = mdb_env_pick_meta(env);
- return mdb_stat0(env, &env->me_metas[toggle]->mm_dbs[MAIN_DBI], arg);
+ return mdb_stat0(env, &meta->mm_dbs[MAIN_DBI], arg);
}
int ESECT
mdb_env_info(MDB_env *env, MDB_envinfo *arg)
{
- int toggle;
+ MDB_meta *meta;
if (env == NULL || arg == NULL)
return EINVAL;
- toggle = mdb_env_pick_meta(env);
- arg->me_mapaddr = env->me_metas[toggle]->mm_address;
+ meta = mdb_env_pick_meta(env);
+ arg->me_mapaddr = meta->mm_address;
+ arg->me_last_pgno = meta->mm_last_pg;
+ arg->me_last_txnid = meta->mm_txnid;
+
arg->me_mapsize = env->me_mapsize;
arg->me_maxreaders = env->me_maxreaders;
arg->me_numreaders = env->me_txns ? env->me_txns->mti_numreaders : 0;
-
- arg->me_last_pgno = env->me_metas[toggle]->mm_last_pg;
- arg->me_last_txnid = env->me_metas[toggle]->mm_txnid;
return MDB_SUCCESS;
}
if (flags & ~VALID_FLAGS)
return EINVAL;
- if (txn->mt_flags & MDB_TXN_ERROR)
+ if (txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
/* main DB? */
/* Is the DB already open? */
len = strlen(name);
- for (i=2; i<txn->mt_numdbs; i++) {
+ for (i=CORE_DBS; i<txn->mt_numdbs; i++) {
if (!txn->mt_dbxs[i].md_name.mv_size) {
/* Remember this free slot */
if (!unused) unused = i;
if (!arg || !TXN_DBI_EXIST(txn, dbi, DB_VALID))
return EINVAL;
- if (txn->mt_flags & MDB_TXN_ERROR)
+ if (txn->mt_flags & MDB_TXN_BLOCKED)
return MDB_BAD_TXN;
if (txn->mt_dbflags[dbi] & DB_STALE) {
void mdb_dbi_close(MDB_env *env, MDB_dbi dbi)
{
char *ptr;
- if (dbi <= MAIN_DBI || dbi >= env->me_maxdbs)
+ if (dbi < CORE_DBS || dbi >= env->me_maxdbs)
return;
ptr = env->me_dbxs[dbi].md_name.mv_data;
/* If there was no name, this was already closed */
goto leave;
/* Can't delete the main DB */
- if (del && dbi > MAIN_DBI) {
+ if (del && dbi >= CORE_DBS) {
rc = mdb_del0(txn, MAIN_DBI, &mc->mc_dbx->md_name, NULL, F_SUBDATA);
if (!rc) {
txn->mt_dbflags[dbi] = DB_STALE;
static int ESECT
mdb_mutex_failed(MDB_env *env, mdb_mutexref_t mutex, int rc)
{
- int toggle, rlocked, rc2;
+ int rlocked, rc2;
+ MDB_meta *meta;
if (rc == MDB_OWNERDEAD) {
/* We own the mutex. Clean up after dead previous owner. */
/* Keep mti_txnid updated, otherwise next writer can
* overwrite data which latest meta page refers to.
*/
- toggle = mdb_env_pick_meta(env);
- env->me_txns->mti_txnid = env->me_metas[toggle]->mm_txnid;
+ meta = mdb_env_pick_meta(env);
+ env->me_txns->mti_txnid = meta->mm_txnid;
/* env is hosed if the dead thread was ours */
if (env->me_txn) {
env->me_flags |= MDB_FATAL_ERROR;
projects / openldap / blobdiff