/** The version number for a database's lockfile format. */
#define MDB_LOCK_VERSION 1
- /** @brief The max size of a key we can write, or 0 for dynamic max.
+ /** @brief The max size of a key we can write, or 0 for computed max.
*
- * Define this as 0 to compute the max from the page size. 511
- * is default for backwards compat: liblmdb <= 0.9.10 can break
- * when modifying a DB with keys/dupsort data bigger than its max.
- * #MDB_DEVEL sets the default to 0.
+ * This macro should normally be left alone or set to 0.
+ * Note that a database with big keys or dupsort data cannot be
+ * reliably modified by a liblmdb which uses a smaller max.
+ * The default is 511 for backwards compat, or 0 when #MDB_DEVEL.
+ *
+ * Other values are allowed, for backwards compat. However:
+ * A value bigger than the computed max can break if you do not
+ * know what you are doing, and liblmdb <= 0.9.10 can break when
+ * modifying a DB with keys/dupsort data bigger than its max.
*
* Data items in an #MDB_DUPSORT database are also limited to
* this size, since they're actually keys of a sub-DB. Keys and
static MDB_cmp_func mdb_cmp_memn, mdb_cmp_memnr, mdb_cmp_int, mdb_cmp_cint, mdb_cmp_long;
/** @endcond */
+/** Compare two items pointing at size_t's of unknown alignment. */
+#ifdef MISALIGNED_OK
+# define mdb_cmp_clong mdb_cmp_long
+#else
+# define mdb_cmp_clong mdb_cmp_cint
+#endif
+
#ifdef _WIN32
static SECURITY_DESCRIPTOR mdb_null_sd;
static SECURITY_ATTRIBUTES mdb_all_sa;
int
mdb_dcmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b)
{
- return txn->mt_dbxs[dbi].md_dcmp(a, b);
+ MDB_cmp_func *dcmp = txn->mt_dbxs[dbi].md_dcmp;
+#if UINT_MAX < SIZE_MAX
+ if (dcmp == mdb_cmp_int && a->mv_size == sizeof(size_t))
+ dcmp = mdb_cmp_clong;
+#endif
+ return dcmp(a, b);
}
/** Allocate memory for a page.
MDB_env *env = txn->mt_env;
MDB_txninfo *ti = env->me_txns;
MDB_meta *meta;
- unsigned int i, nr;
+ unsigned int i, nr, flags = txn->mt_flags;
uint16_t x;
int rc, new_notls = 0;
- if (txn->mt_flags & MDB_TXN_RDONLY) {
- txn->mt_flags = MDB_TXN_RDONLY;
- /* Setup db info */
- txn->mt_numdbs = env->me_numdbs;
- txn->mt_dbxs = env->me_dbxs; /* mostly static anyway */
+ if ((flags &= MDB_TXN_RDONLY) != 0) {
if (!ti) {
meta = env->me_metas[ mdb_env_pick_meta(env) ];
txn->mt_txnid = meta->mm_txnid;
txn->mt_u.reader = r;
meta = env->me_metas[txn->mt_txnid & 1];
}
+ txn->mt_dbxs = env->me_dbxs; /* mostly static anyway */
} else {
if (ti) {
LOCK_MUTEX_W(env);
meta = env->me_metas[ mdb_env_pick_meta(env) ];
txn->mt_txnid = meta->mm_txnid;
}
- /* Setup db info */
- txn->mt_numdbs = env->me_numdbs;
txn->mt_txnid++;
#if MDB_DEBUG
if (txn->mt_txnid == mdb_debug_start)
mdb_debug = 1;
#endif
- txn->mt_flags = 0;
txn->mt_child = NULL;
txn->mt_loose_pgs = NULL;
txn->mt_loose_count = 0;
/* Moved to here to avoid a data race in read TXNs */
txn->mt_next_pgno = meta->mm_last_pg+1;
+ txn->mt_flags = flags;
+
+ /* Setup db info */
+ txn->mt_numdbs = env->me_numdbs;
for (i=2; i<txn->mt_numdbs; i++) {
x = env->me_dbflags[i];
txn->mt_dbs[i].md_flags = x & PERSISTENT_FLAGS;
}
if (!txn->mt_parent) {
- if (mdb_midl_shrink(&txn->mt_free_pgs))
- env->me_free_pgs = txn->mt_free_pgs;
+ mdb_midl_shrink(&txn->mt_free_pgs);
+ env->me_free_pgs = txn->mt_free_pgs;
/* me_pgstate: */
env->me_pghead = NULL;
env->me_pglast = 0;
/* Write up to MDB_COMMIT_PAGES dirty pages at a time. */
if (pos!=next_pos || n==MDB_COMMIT_PAGES || wsize+size>MAX_WRITE) {
if (n) {
+retry_write:
/* Write previous page(s) */
#ifdef MDB_USE_PWRITEV
wres = pwritev(env->me_fd, iov, n, wpos);
if (n == 1) {
wres = pwrite(env->me_fd, iov[0].iov_base, wsize, wpos);
} else {
+retry_seek:
if (lseek(env->me_fd, wpos, SEEK_SET) == -1) {
rc = ErrCode();
+ if (rc == EINTR)
+ goto retry_seek;
DPRINTF(("lseek: %s", strerror(rc)));
return rc;
}
if (wres != wsize) {
if (wres < 0) {
rc = ErrCode();
+ if (rc == EINTR)
+ goto retry_write;
DPRINTF(("Write error: %s", strerror(rc)));
} else {
rc = EIO; /* TODO: Use which error code? */
goto fail;
}
data.mv_data = &txn->mt_dbs[i];
- rc = mdb_cursor_put(&mc, &txn->mt_dbxs[i].md_name, &data, 0);
+ rc = mdb_cursor_put(&mc, &txn->mt_dbxs[i].md_name, &data,
+ F_SUBDATA);
if (rc)
goto fail;
}
mdb_midl_free(env->me_pghead);
env->me_pghead = NULL;
- if (mdb_midl_shrink(&txn->mt_free_pgs))
- env->me_free_pgs = txn->mt_free_pgs;
+ mdb_midl_shrink(&txn->mt_free_pgs);
+ env->me_free_pgs = txn->mt_free_pgs;
#if (MDB_DEBUG) > 2
mdb_audit(txn);
int len;
#define DO_PWRITE(rc, fd, ptr, size, len, pos) do { \
len = pwrite(fd, ptr, size, pos); \
- rc = (len >= 0); } while(0)
+ if (len == -1 && ErrCode() == EINTR) continue; \
+ rc = (len >= 0); break; } while(1)
#endif
DPUTS("writing new meta page");
mdb_env_init_meta0(env, meta);
p = calloc(2, psize);
+ if (!p)
+ return ENOMEM;
+
p->mp_pgno = 0;
p->mp_flags = P_META;
*(MDB_meta *)METADATA(p) = *meta;
rc = -1;
}
#else
+retry_write:
rc = pwrite(mfd, ptr, len, off);
#endif
if (rc != len) {
rc = rc < 0 ? ErrCode() : EIO;
+#ifndef _WIN32
+ if (rc == EINTR)
+ goto retry_write;
+#endif
DPUTS("write failed, disk error?");
/* On a failure, the pagecache still contains the new data.
* Write some old data back, to prevent it from being used.
*(size_t *)a->mv_data > *(size_t *)b->mv_data;
}
-/** Compare two items pointing at aligned unsigned int's */
+/** Compare two items pointing at aligned unsigned int's.
+ *
+ * This is also set as #MDB_INTEGERDUP|#MDB_DUPFIXED's #MDB_dbx.%md_dcmp,
+ * but #mdb_cmp_clong() is called instead if the data type is size_t.
+ */
static int
mdb_cmp_int(const MDB_val *a, const MDB_val *b)
{
#endif
}
-/** Compare two items pointing at size_t's of unknown alignment. */
-#ifdef MISALIGNED_OK
-# define mdb_cmp_clong mdb_cmp_long
-#else
-# define mdb_cmp_clong mdb_cmp_cint
-#endif
-
/** Compare two items lexically */
static int
mdb_cmp_memn(const MDB_val *a, const MDB_val *b)
&mc->mc_dbx->md_name, &exact);
if (!exact)
return MDB_NOTFOUND;
+ if ((leaf->mn_flags & (F_DUPDATA|F_SUBDATA)) != F_SUBDATA)
+ return MDB_INCOMPATIBLE; /* not a named DB */
rc = mdb_node_read(mc->mc_txn, leaf, &data);
if (rc)
return rc;
if (leaf == NULL) {
DPUTS("===> inexact leaf not found, goto sibling");
- if ((rc = mdb_cursor_sibling(mc, 1)) != MDB_SUCCESS)
+ if ((rc = mdb_cursor_sibling(mc, 1)) != MDB_SUCCESS) {
+ mc->mc_flags |= C_EOF;
return rc; /* no entries matched */
+ }
mp = mc->mc_pg[mc->mc_top];
mdb_cassert(mc, IS_LEAF(mp));
leaf = NODEPTR(mp, 0);
return rc;
}
} else if (op == MDB_GET_BOTH || op == MDB_GET_BOTH_RANGE) {
- MDB_val d2;
- if ((rc = mdb_node_read(mc->mc_txn, leaf, &d2)) != MDB_SUCCESS)
+ MDB_val olddata;
+ MDB_cmp_func *dcmp;
+ if ((rc = mdb_node_read(mc->mc_txn, leaf, &olddata)) != MDB_SUCCESS)
return rc;
- rc = mc->mc_dbx->md_dcmp(data, &d2);
+ dcmp = mc->mc_dbx->md_dcmp;
+#if UINT_MAX < SIZE_MAX
+ if (dcmp == mdb_cmp_int && olddata.mv_size == sizeof(size_t))
+ dcmp = mdb_cmp_clong;
+#endif
+ rc = dcmp(data, &olddata);
if (rc) {
if (op == MDB_GET_BOTH || rc > 0)
return MDB_NOTFOUND;
rc = 0;
- *data = d2;
+ *data = olddata;
}
} else {
/* Was a single item before, must convert now */
if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) {
+ MDB_cmp_func *dcmp;
/* Just overwrite the current item */
if (flags == MDB_CURRENT)
goto current;
-
+ dcmp = mc->mc_dbx->md_dcmp;
#if UINT_MAX < SIZE_MAX
- if (mc->mc_dbx->md_dcmp == mdb_cmp_int && olddata.mv_size == sizeof(size_t))
- mc->mc_dbx->md_dcmp = mdb_cmp_clong;
+ if (dcmp == mdb_cmp_int && olddata.mv_size == sizeof(size_t))
+ dcmp = mdb_cmp_clong;
#endif
/* does data match? */
- if (!mc->mc_dbx->md_dcmp(data, &olddata)) {
+ if (!dcmp(data, &olddata)) {
if (flags & MDB_NODUPDATA)
return MDB_KEYEXIST;
/* overwrite it */
goto new_sub;
}
current:
+ /* LMDB passes F_SUBDATA in 'flags' to write a DB record */
+ if ((leaf->mn_flags ^ flags) & F_SUBDATA)
+ return MDB_INCOMPATIBLE;
/* overflow page overwrites need special handling */
if (F_ISSET(leaf->mn_flags, F_BIGDATA)) {
MDB_page *omp;
goto fail;
}
}
+ /* LMDB passes F_SUBDATA in 'flags' to delete a DB record */
+ else if ((leaf->mn_flags ^ flags) & F_SUBDATA) {
+ rc = MDB_INCOMPATIBLE;
+ goto fail;
+ }
/* add overflow pages to free list */
if (F_ISSET(leaf->mn_flags, F_BIGDATA)) {
{
MDB_node *node;
int rc;
- unsigned int ptop, minkeys;
+ unsigned int ptop, minkeys, thresh;
MDB_cursor mn;
indx_t oldki;
- minkeys = 1 + (IS_BRANCH(mc->mc_pg[mc->mc_top]));
+ if (IS_BRANCH(mc->mc_pg[mc->mc_top])) {
+ minkeys = 2;
+ thresh = 1;
+ } else {
+ minkeys = 1;
+ thresh = FILL_THRESHOLD;
+ }
DPRINTF(("rebalancing %s page %"Z"u (has %u keys, %.1f%% full)",
IS_LEAF(mc->mc_pg[mc->mc_top]) ? "leaf" : "branch",
mdb_dbg_pgno(mc->mc_pg[mc->mc_top]), NUMKEYS(mc->mc_pg[mc->mc_top]),
(float)PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) / 10));
- if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= FILL_THRESHOLD &&
+ if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= thresh &&
NUMKEYS(mc->mc_pg[mc->mc_top]) >= minkeys) {
DPRINTF(("no need to rebalance page %"Z"u, above fill threshold",
mdb_dbg_pgno(mc->mc_pg[mc->mc_top])));
* move one key from it. Otherwise we should try to merge them.
* (A branch page must never have less than 2 keys.)
*/
- minkeys = 1 + (IS_BRANCH(mn.mc_pg[mn.mc_top]));
- if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= FILL_THRESHOLD && NUMKEYS(mn.mc_pg[mn.mc_top]) > minkeys) {
+ 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[ptop]) {
oldki++;
MDB_cursor *m2, *m3;
MDB_dbi dbi = mc->mc_dbi;
+ /* DB is totally empty now, just bail out.
+ * Other cursors adjustments were already done
+ * by mdb_rebalance and aren't needed here.
+ */
+ if (!mc->mc_snum)
+ return rc;
+
mp = mc->mc_pg[mc->mc_top];
nkeys = NUMKEYS(mp);
psize = 0;
if (newindx <= split_indx || newindx >= nkeys) {
i = 0; j = 1;
- k = newindx >= nkeys ? nkeys : split_indx+2;
+ k = newindx >= nkeys ? nkeys : split_indx+1+IS_LEAF(mp);
} else {
i = nkeys; j = -1;
k = split_indx-1;
if (rc == MDB_SUCCESS) {
/* make sure this is actually a DB */
MDB_node *node = NODEPTR(mc.mc_pg[mc.mc_top], mc.mc_ki[mc.mc_top]);
- if (!(node->mn_flags & F_SUBDATA))
+ if ((node->mn_flags & (F_DUPDATA|F_SUBDATA)) != F_SUBDATA)
return MDB_INCOMPATIBLE;
} else if (rc == MDB_NOTFOUND && (flags & MDB_CREATE)) {
/* Create if requested */
/* Can't delete the main DB */
if (del && dbi > MAIN_DBI) {
- rc = mdb_del0(txn, MAIN_DBI, &mc->mc_dbx->md_name, NULL, 0);
+ rc = mdb_del0(txn, MAIN_DBI, &mc->mc_dbx->md_name, NULL, F_SUBDATA);
if (!rc) {
txn->mt_dbflags[dbi] = DB_STALE;
mdb_dbi_close(txn->mt_env, dbi);