*/
#include <sys/types.h>
#include <sys/stat.h>
-#include <sys/queue.h>
#include <sys/param.h>
#include <sys/uio.h>
#include <sys/mman.h>
#define mp_pgno mp_p.p_pgno
union padded {
pgno_t p_pgno; /* page number */
- void * p_pad;
+ void * p_align; /* for IL32P64 */
} mp_p;
#define P_BRANCH 0x01 /* branch page */
#define P_LEAF 0x02 /* leaf page */
static int mdb_touch(MDB_txn *txn, MDB_pageparent *mp);
typedef struct MDB_ppage { /* ordered list of pages */
- SLIST_ENTRY(MDB_ppage) mp_entry;
MDB_page *mp_page;
unsigned int mp_ki; /* cursor index on page */
} 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)
-#define CURSOR_PUSH(c,p) SLIST_INSERT_HEAD(&(c)->mc_stack, p, mp_entry)
+#define CURSOR_TOP(c) (&(c)->mc_stack[(c)->mc_snum-1])
+#define CURSOR_PARENT(c) (&(c)->mc_stack[(c)->mc_snum-2])
struct MDB_xcursor;
struct MDB_cursor {
MDB_txn *mc_txn;
- struct page_stack mc_stack; /* stack of parent pages */
+ MDB_ppage mc_stack[32]; /* stack of parent pages */
+ unsigned int mc_snum; /* number of pushed pages */
MDB_dbi mc_dbi;
short mc_initialized; /* 1 if initialized */
short mc_eof; /* 1 if end is reached */
return MDB_VERSION_STRING;
}
+static const char *errstr[] = {
+ "MDB_KEYEXIST: Key/data pair already exists",
+ "MDB_NOTFOUND: No matching key/data pair found",
+ "MDB_PAGE_NOTFOUND: Requested page not found",
+ "MDB_CORRUPTED: Located page was wrong type",
+ "MDB_PANIC: Update of meta page failed",
+ "MDB_VERSION_MISMATCH: Database environment version mismatch"
+};
+
+char *
+mdb_strerror(int err)
+{
+ if (!err)
+ return ("Successful return: 0");
+
+ if (err >= MDB_KEYEXIST && err <= MDB_VERSION_MISMATCH)
+ return (char *)errstr[err - MDB_KEYEXIST];
+
+ return strerror(err);
+}
+
int
mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b)
{
env = txn->mt_env;
if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
- DPRINTF("attempt to commit read-only transaction");
mdb_txn_abort(txn);
- return EPERM;
+ return MDB_SUCCESS;
}
if (txn != env->me_txn) {
/* Commit up to MDB_COMMIT_PAGES dirty pages to disk until done.
*/
next = 0;
+ i = 1;
do {
n = 0;
done = 1;
size = 0;
- for (i=1; i<=txn->mt_u.dirty_list[0].mid; i++) {
+ for (; i<=txn->mt_u.dirty_list[0].mid; i++) {
dp = txn->mt_u.dirty_list[i].mptr;
if (dp->p.mp_pgno != next) {
if (n) {
mdb_txn_abort(txn);
return n;
}
- env->me_txn = NULL;
+done:
+ env->me_txn = NULL;
/* update the DB tables */
{
int toggle = !env->me_db_toggle;
pthread_mutex_unlock(&env->me_txns->mti_wmutex);
free(txn);
- txn = NULL;
-
-done:
- mdb_txn_abort(txn);
return MDB_SUCCESS;
}
{
MDB_ppage *top;
- top = CURSOR_TOP(cursor);
- CURSOR_POP(cursor);
-
- DPRINTF("popped page %lu off cursor %p", top->mp_page->mp_pgno, (void *) cursor);
+ if (cursor->mc_snum) {
+ top = CURSOR_TOP(cursor);
+ cursor->mc_snum--;
- free(top);
+ DPRINTF("popped page %lu off cursor %p", top->mp_page->mp_pgno, (void *) cursor);
+ }
}
static MDB_ppage *
DPRINTF("pushing page %lu on cursor %p", mp->mp_pgno, (void *) cursor);
- if ((ppage = calloc(1, sizeof(MDB_ppage))) == NULL)
- return NULL;
+ ppage = &cursor->mc_stack[cursor->mc_snum++];
ppage->mp_page = mp;
- CURSOR_PUSH(cursor, ppage);
+ ppage->mp_ki = 0;
return ppage;
}
DPRINTF("read overflow page %lu failed", pgno);
return MDB_PAGE_NOTFOUND;
}
- data->mv_data = omp;
+ data->mv_data = METADATA(omp);
return MDB_SUCCESS;
}
{
int rc;
MDB_node *indx;
- MDB_ppage *parent, *top;
+ MDB_ppage *parent;
MDB_page *mp;
- top = CURSOR_TOP(cursor);
- if ((parent = SLIST_NEXT(top, mp_entry)) == NULL) {
+ if (cursor->mc_snum < 2) {
return MDB_NOTFOUND; /* root has no siblings */
}
+ parent = CURSOR_PARENT(cursor);
DPRINTF("parent page is page %lu, index %u",
parent->mp_page->mp_pgno, parent->mp_ki);
assert(key);
assert(key->mv_size > 0);
- while (CURSOR_TOP(cursor) != NULL)
- cursor_pop_page(cursor);
+ cursor->mc_snum = 0;
rc = mdb_search_page(cursor->mc_txn, cursor->mc_dbi, key, cursor, 0, &mpp);
if (rc != MDB_SUCCESS)
cursor->mc_eof = 0;
if (data) {
- if ((rc = mdb_read_data(cursor->mc_txn, leaf, data)) != MDB_SUCCESS)
+ MDB_val d2;
+ if ((rc = mdb_read_data(cursor->mc_txn, leaf, &d2)) != MDB_SUCCESS)
return rc;
if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
if (rc != MDB_SUCCESS)
return rc;
}
+ } else {
+ *data = d2;
}
+
}
rc = mdb_set_key(leaf, key);
MDB_pageparent mpp;
MDB_node *leaf;
- while (CURSOR_TOP(cursor) != NULL)
- cursor_pop_page(cursor);
+ cursor->mc_snum = 0;
rc = mdb_search_page(cursor->mc_txn, cursor->mc_dbi, NULL, cursor, 0, &mpp);
if (rc != MDB_SUCCESS)
MDB_node *leaf;
MDB_val lkey;
- while (CURSOR_TOP(cursor) != NULL)
- cursor_pop_page(cursor);
+ cursor->mc_snum = 0;
lkey.mv_size = MAXKEYSIZE+1;
lkey.mv_data = NULL;
mx->mx_dbxs[dbn+1].md_dirty = 0;
mx->mx_txn.mt_numdbs = dbn+2;
- SLIST_INIT(&mx->mx_cursor.mc_stack);
+ mx->mx_cursor.mc_snum = 0;
mx->mx_cursor.mc_txn = &mx->mx_txn;
mx->mx_cursor.mc_dbi = dbn+1;
}
size += sizeof(MDB_xcursor);
if ((cursor = calloc(1, size)) != NULL) {
- SLIST_INIT(&cursor->mc_stack);
cursor->mc_dbi = dbi;
cursor->mc_txn = txn;
if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) {
mdb_cursor_close(MDB_cursor *cursor)
{
if (cursor != NULL) {
- while(!CURSOR_EMPTY(cursor))
- cursor_pop_page(cursor);
- if (cursor->mc_txn->mt_dbs[cursor->mc_dbi].md_flags & MDB_DUPSORT) {
- mdb_xcursor_fini(cursor->mc_txn, cursor->mc_dbi, cursor->mc_xcursor);
- while(!CURSOR_EMPTY(&cursor->mc_xcursor->mx_cursor))
- cursor_pop_page(&cursor->mc_xcursor->mx_cursor);
- }
-
free(cursor);
}
}
unsigned int i;
cursor_pop_page(&mx.mx_cursor);
- top = CURSOR_TOP(&mx.mx_cursor);
- if (top != NULL) {
- parent = SLIST_NEXT(top, mp_entry);
- while (parent != NULL) {
+ if (mx.mx_cursor.mc_snum) {
+ top = CURSOR_TOP(&mx.mx_cursor);
+ while (mx.mx_cursor.mc_snum > 1) {
+ parent = CURSOR_PARENT(&mx.mx_cursor);
for (i=0; i<NUMKEYS(top->mp_page); i++) {
ni = NODEPTR(top->mp_page, i);
mdb_midl_insert(txn->mt_free_pgs, ni->mn_pgno);
}
- if (parent) {
- parent->mp_ki++;
- if (parent->mp_ki >= NUMKEYS(parent->mp_page)) {
- cursor_pop_page(&mx.mx_cursor);
- top = CURSOR_TOP(&mx.mx_cursor);
- parent = SLIST_NEXT(top, mp_entry);
- } else {
- ni = NODEPTR(parent->mp_page, parent->mp_ki);
- top->mp_page = mdb_get_page(&mx.mx_txn, ni->mn_pgno);
- }
+ parent->mp_ki++;
+ if (parent->mp_ki >= NUMKEYS(parent->mp_page)) {
+ cursor_pop_page(&mx.mx_cursor);
+ top = parent;
+ } else {
+ ni = NODEPTR(parent->mp_page, parent->mp_ki);
+ top->mp_page = mdb_get_page(&mx.mx_txn, ni->mn_pgno);
}
}
}
mpp.mp_page = &dp->p;
txn->mt_dbs[dbi].md_root = mpp.mp_page->mp_pgno;
txn->mt_dbs[dbi].md_depth++;
+ txn->mt_dbxs[dbi].md_dirty = 1;
ki = 0;
}
else
return mdb_put0(txn, dbi, key, data, flags);
}
+int
+mdb_env_set_flags(MDB_env *env, unsigned int flag, int onoff)
+{
+#define CHANGEABLE (MDB_NOSYNC)
+ if ((flag & CHANGEABLE) != flag)
+ return EINVAL;
+ if (onoff)
+ env->me_flags |= flag;
+ else
+ env->me_flags &= ~flag;
+ return MDB_SUCCESS;
+}
+
int
mdb_env_get_flags(MDB_env *env, unsigned int *arg)
{
txn->mt_dbxs[txn->mt_numdbs].md_dirty = dirty;
memcpy(&txn->mt_dbs[txn->mt_numdbs], data.mv_data, sizeof(MDB_db));
*dbi = txn->mt_numdbs;
+ txn->mt_env->me_dbs[0][txn->mt_numdbs] = txn->mt_dbs[txn->mt_numdbs];
+ txn->mt_env->me_dbs[1][txn->mt_numdbs] = txn->mt_dbs[txn->mt_numdbs];
txn->mt_numdbs++;
}