X-Git-Url: http://git.ozlabs.org/?p=ccan;a=blobdiff_plain;f=ccan%2Ftdb%2Flock.c;h=280e3679e3762e8ede315b300b61abf8d5e13ef6;hp=19b92d9ae09c63152b87d10062b1363a8c063516;hb=d665ff828977d76f8ec6ed53e6aab2e6fa124675;hpb=a879e3e2859e17496f29e14b8f393583b2c93f19

diff --git a/ccan/tdb/lock.c b/ccan/tdb/lock.c
index 19b92d9a..280e3679 100644
--- a/ccan/tdb/lock.c
+++ b/ccan/tdb/lock.c
@@ -119,6 +119,12 @@ static int fcntl_unlock(struct tdb_context *tdb, int rw, off_t off, off_t len)
 	return fcntl(tdb->fd, F_SETLKW, &fl);
 }
 
+/* list -1 is the alloc list, otherwise a hash chain. */
+static tdb_off_t lock_offset(int list)
+{
+	return FREELIST_TOP + 4*list;
+}
+
 /* a byte range locking function - return 0 on success
    this functions locks/unlocks 1 byte at the specified offset.
 
@@ -146,6 +152,14 @@ int tdb_brlock(struct tdb_context *tdb,
 		return -1;
 	}
 
+	/* Sanity check */
+	if (tdb->transaction && offset >= lock_offset(-1) && len != 0) {
+		tdb->ecode = TDB_ERR_RDONLY;
+		TDB_LOG((tdb, TDB_DEBUG_TRACE, "tdb_brlock attempted in transaction at offset %d rw_type=%d flags=%d len=%d\n",
+			 offset, rw_type, flags, (int)len));
+		return -1;
+	}
+
 	do {
 		ret = fcntl_lock(tdb, rw_type, offset, len,
 				 flags & TDB_LOCK_WAIT);
@@ -197,13 +211,29 @@ int tdb_brunlock(struct tdb_context *tdb,
   deadlock detection and claim a deadlock when progress can be
   made. For those OSes we may loop for a while.  
 */
-int tdb_brlock_upgrade(struct tdb_context *tdb, tdb_off_t offset, size_t len)
+int tdb_allrecord_upgrade(struct tdb_context *tdb)
 {
 	int count = 1000;
+
+	if (tdb->allrecord_lock.count != 1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "tdb_allrecord_upgrade failed: count %u too high\n",
+			 tdb->allrecord_lock.count));
+		return -1;
+	}
+
+	if (tdb->allrecord_lock.off != 1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR,
+			 "tdb_allrecord_upgrade failed: already upgraded?\n"));
+		return -1;
+	}
+
 	while (count--) {
 		struct timeval tv;
-		if (tdb_brlock(tdb, F_WRLCK, offset, len,
+		if (tdb_brlock(tdb, F_WRLCK, FREELIST_TOP, 0,
 			       TDB_LOCK_WAIT|TDB_LOCK_PROBE) == 0) {
+			tdb->allrecord_lock.ltype = F_WRLCK;
+			tdb->allrecord_lock.off = 0;
 			return 0;
 		}
 		if (errno != EDEADLK) {
@@ -214,16 +244,10 @@ int tdb_brlock_upgrade(struct tdb_context *tdb, tdb_off_t offset, size_t len)
 		tv.tv_usec = 1;
 		select(0, NULL, NULL, NULL, &tv);
 	}
-	TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_brlock_upgrade failed at offset %d\n", offset));
+	TDB_LOG((tdb, TDB_DEBUG_TRACE,"tdb_allrecord_upgrade failed\n"));
 	return -1;
 }
 
-/* list -1 is the alloc list, otherwise a hash chain. */
-static tdb_off_t lock_offset(int list)
-{
-	return FREELIST_TOP + 4*list;
-}
-
 static struct tdb_lock_type *find_nestlock(struct tdb_context *tdb,
 					   tdb_off_t offset)
 {
@@ -273,16 +297,14 @@ int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype,
 
 	/* Since fcntl locks don't nest, we do a lock for the first one,
 	   and simply bump the count for future ones */
-	if (tdb->methods->brlock(tdb, ltype, offset, 1, flags)) {
+	if (tdb_brlock(tdb, ltype, offset, 1, flags)) {
 		return -1;
 	}
 
-	tdb->num_locks++;
-
 	tdb->lockrecs[tdb->num_lockrecs].off = offset;
 	tdb->lockrecs[tdb->num_lockrecs].count = 1;
 	tdb->lockrecs[tdb->num_lockrecs].ltype = ltype;
-	tdb->num_lockrecs += 1;
+	tdb->num_lockrecs++;
 
 	return 0;
 }
@@ -366,19 +388,14 @@ int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype,
 	if (mark_lock) {
 		ret = 0;
 	} else {
-		ret = tdb->methods->brunlock(tdb, ltype, offset, 1);
+		ret = tdb_brunlock(tdb, ltype, offset, 1);
 	}
-	tdb->num_locks--;
 
 	/*
 	 * Shrink the array by overwriting the element just unlocked with the
 	 * last array element.
 	 */
-
-	if (tdb->num_lockrecs > 1) {
-		*lck = tdb->lockrecs[tdb->num_lockrecs-1];
-	}
-	tdb->num_lockrecs -= 1;
+	*lck = tdb->lockrecs[--tdb->num_lockrecs];
 
 	/*
 	 * We don't bother with realloc when the array shrinks, but if we have
@@ -427,11 +444,10 @@ int tdb_transaction_unlock(struct tdb_context *tdb, int ltype)
 }
 
 
-
-
-/* lock/unlock entire database */
-static int _tdb_lockall(struct tdb_context *tdb, int ltype,
-			enum tdb_lock_flags flags)
+/* lock/unlock entire database.  It can only be upgradable if you have some
+ * other way of guaranteeing exclusivity (ie. transaction write lock). */
+int tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
+		       enum tdb_lock_flags flags, bool upgradable)
 {
 	/* There are no locks on read-only dbs */
 	if (tdb->read_only || tdb->traverse_read) {
@@ -456,9 +472,13 @@ static int _tdb_lockall(struct tdb_context *tdb, int ltype,
 		return -1;
 	}
 
-	if (tdb->methods->brlock(tdb, ltype,
-				 FREELIST_TOP, 4*tdb->header.hash_size,
-				 flags)) {
+	if (upgradable && ltype != F_RDLCK) {
+		/* tdb error: you can't upgrade a write lock! */
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	if (tdb_brlock(tdb, ltype, FREELIST_TOP, 0, flags)) {
 		if (flags & TDB_LOCK_WAIT) {
 			TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lockall failed (%s)\n", strerror(errno)));
 		}
@@ -466,7 +486,10 @@ static int _tdb_lockall(struct tdb_context *tdb, int ltype,
 	}
 
 	tdb->allrecord_lock.count = 1;
-	tdb->allrecord_lock.ltype = ltype;
+	/* If it's upgradable, it's actually exclusive so we can treat
+	 * it as a write lock. */
+	tdb->allrecord_lock.ltype = upgradable ? F_WRLCK : ltype;
+	tdb->allrecord_lock.off = upgradable;
 
 	return 0;
 }
@@ -474,7 +497,7 @@ static int _tdb_lockall(struct tdb_context *tdb, int ltype,
 
 
 /* unlock entire db */
-static int _tdb_unlockall(struct tdb_context *tdb, int ltype, bool mark_lock)
+int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype, bool mark_lock)
 {
 	/* There are no locks on read-only dbs */
 	if (tdb->read_only || tdb->traverse_read) {
@@ -482,7 +505,14 @@ static int _tdb_unlockall(struct tdb_context *tdb, int ltype, bool mark_lock)
 		return -1;
 	}
 
-	if (tdb->allrecord_lock.ltype != ltype || tdb->allrecord_lock.count == 0) {
+	if (tdb->allrecord_lock.count == 0) {
+		tdb->ecode = TDB_ERR_LOCK;
+		return -1;
+	}
+
+	/* Upgradable locks are marked as write locks. */
+	if (tdb->allrecord_lock.ltype != ltype
+	    && (!tdb->allrecord_lock.off || ltype != F_RDLCK)) {
 		tdb->ecode = TDB_ERR_LOCK;
 		return -1;
 	}
@@ -492,9 +522,7 @@ static int _tdb_unlockall(struct tdb_context *tdb, int ltype, bool mark_lock)
 		return 0;
 	}
 
-	if (!mark_lock &&
-	    tdb->methods->brunlock(tdb, ltype,
-				   FREELIST_TOP, 4*tdb->header.hash_size)) {
+	if (!mark_lock && tdb_brunlock(tdb, ltype, FREELIST_TOP, 0)) {
 		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlockall failed (%s)\n", strerror(errno)));
 		return -1;
 	}
@@ -509,27 +537,27 @@ static int _tdb_unlockall(struct tdb_context *tdb, int ltype, bool mark_lock)
 int tdb_lockall(struct tdb_context *tdb)
 {
 	tdb_trace(tdb, "tdb_lockall");
-	return _tdb_lockall(tdb, F_WRLCK, TDB_LOCK_WAIT);
+	return tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_WAIT, false);
 }
 
 /* lock entire database with write lock - mark only */
 int tdb_lockall_mark(struct tdb_context *tdb)
 {
 	tdb_trace(tdb, "tdb_lockall_mark");
-	return _tdb_lockall(tdb, F_WRLCK, TDB_LOCK_MARK_ONLY);
+	return tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_MARK_ONLY, false);
 }
 
 /* unlock entire database with write lock - unmark only */
 int tdb_lockall_unmark(struct tdb_context *tdb)
 {
 	tdb_trace(tdb, "tdb_lockall_unmark");
-	return _tdb_unlockall(tdb, F_WRLCK, true);
+	return tdb_allrecord_unlock(tdb, F_WRLCK, true);
 }
 
 /* lock entire database with write lock - nonblocking varient */
 int tdb_lockall_nonblock(struct tdb_context *tdb)
 {
-	int ret = _tdb_lockall(tdb, F_WRLCK, TDB_LOCK_NOWAIT);
+	int ret = tdb_allrecord_lock(tdb, F_WRLCK, TDB_LOCK_NOWAIT, false);
 	tdb_trace_ret(tdb, "tdb_lockall_nonblock", ret);
 	return ret;
 }
@@ -538,20 +566,20 @@ int tdb_lockall_nonblock(struct tdb_context *tdb)
 int tdb_unlockall(struct tdb_context *tdb)
 {
 	tdb_trace(tdb, "tdb_unlockall");
-	return _tdb_unlockall(tdb, F_WRLCK, false);
+	return tdb_allrecord_unlock(tdb, F_WRLCK, false);
 }
 
 /* lock entire database with read lock */
 int tdb_lockall_read(struct tdb_context *tdb)
 {
 	tdb_trace(tdb, "tdb_lockall_read");
-	return _tdb_lockall(tdb, F_RDLCK, TDB_LOCK_WAIT);
+	return tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_WAIT, false);
 }
 
 /* lock entire database with read lock - nonblock varient */
 int tdb_lockall_read_nonblock(struct tdb_context *tdb)
 {
-	int ret = _tdb_lockall(tdb, F_RDLCK, TDB_LOCK_NOWAIT);
+	int ret = tdb_allrecord_lock(tdb, F_RDLCK, TDB_LOCK_NOWAIT, false);
 	tdb_trace_ret(tdb, "tdb_lockall_read_nonblock", ret);
 	return ret;
 }
@@ -560,7 +588,7 @@ int tdb_lockall_read_nonblock(struct tdb_context *tdb)
 int tdb_unlockall_read(struct tdb_context *tdb)
 {
 	tdb_trace(tdb, "tdb_unlockall_read");
-	return _tdb_unlockall(tdb, F_RDLCK, false);
+	return tdb_allrecord_unlock(tdb, F_RDLCK, false);
 }
 
 /* lock/unlock one hash chain. This is meant to be used to reduce
@@ -627,7 +655,7 @@ int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off)
 	if (tdb->allrecord_lock.count) {
 		return 0;
 	}
-	return off ? tdb->methods->brlock(tdb, F_RDLCK, off, 1, TDB_LOCK_WAIT) : 0;
+	return off ? tdb_brlock(tdb, F_RDLCK, off, 1, TDB_LOCK_WAIT) : 0;
 }
 
 /*
@@ -641,12 +669,21 @@ int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off)
 	for (i = &tdb->travlocks; i; i = i->next)
 		if (i->off == off)
 			return -1;
-	return tdb->methods->brlock(tdb, F_WRLCK, off, 1, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE);
+	if (tdb->allrecord_lock.count) {
+		if (tdb->allrecord_lock.ltype == F_WRLCK) {
+			return 0;
+		}
+		return -1;
+	}
+	return tdb_brlock(tdb, F_WRLCK, off, 1, TDB_LOCK_NOWAIT|TDB_LOCK_PROBE);
 }
 
 int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off)
 {
-	return tdb->methods->brunlock(tdb, F_WRLCK, off, 1);
+	if (tdb->allrecord_lock.count) {
+		return 0;
+	}
+	return tdb_brunlock(tdb, F_WRLCK, off, 1);
 }
 
 /* fcntl locks don't stack: avoid unlocking someone else's */
@@ -664,14 +701,15 @@ int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off)
 	for (i = &tdb->travlocks; i; i = i->next)
 		if (i->off == off)
 			count++;
-	return (count == 1 ? tdb->methods->brunlock(tdb, F_RDLCK, off, 1) : 0);
+	return (count == 1 ? tdb_brunlock(tdb, F_RDLCK, off, 1) : 0);
 }
 
 bool tdb_have_extra_locks(struct tdb_context *tdb)
 {
 	unsigned int extra = tdb->num_lockrecs;
 
-	if (tdb->allrecord_lock.count) {
+	/* A transaction holds the lock for all records. */
+	if (!tdb->transaction && tdb->allrecord_lock.count) {
 		return true;
 	}
 
@@ -695,8 +733,7 @@ void tdb_release_extra_locks(struct tdb_context *tdb)
 	unsigned int i, extra = 0;
 
 	if (tdb->allrecord_lock.count != 0) {
-		tdb_brunlock(tdb, tdb->allrecord_lock.ltype,
-			     FREELIST_TOP, 4*tdb->header.hash_size);
+		tdb_brunlock(tdb, tdb->allrecord_lock.ltype, FREELIST_TOP, 0);
 		tdb->allrecord_lock.count = 0;
 	}
 
@@ -712,7 +749,6 @@ void tdb_release_extra_locks(struct tdb_context *tdb)
 			tdb_brunlock(tdb, lck->ltype, lck->off, 1);
 		}
 	}
-	tdb->num_locks = extra;
 	tdb->num_lockrecs = extra;
 	if (tdb->num_lockrecs == 0) {
 		SAFE_FREE(tdb->lockrecs);