X-Git-Url: https://git.ozlabs.org/?p=ccan;a=blobdiff_plain;f=ccan%2Ftdb%2Flock.c;h=a59ba365213e473630617ee69ff1f238585d6782;hp=66cfcbdeea6eab9cd943f7b18a9dd8f66e4c9f8e;hb=88541fe7eb0542150669312675ddc00e9684de66;hpb=57c618f35d2aa244d444670c63a21e7a677c63c2

diff --git a/ccan/tdb/lock.c b/ccan/tdb/lock.c
index 66cfcbde..a59ba365 100644
--- a/ccan/tdb/lock.c
+++ b/ccan/tdb/lock.c
@@ -197,13 +197,29 @@ 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) {
@@ -214,7 +230,7 @@ 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;
 }
 
@@ -224,12 +240,24 @@ 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)
+{
+	unsigned int i;
+
+	for (i=0; i<tdb->num_lockrecs; i++) {
+		if (tdb->lockrecs[i].off == offset) {
+			return &tdb->lockrecs[i];
+		}
+	}
+	return NULL;
+}
+
 /* lock an offset in the database. */
 int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype,
 		  enum tdb_lock_flags flags)
 {
 	struct tdb_lock_type *new_lck;
-	int i;
 
 	if (offset >= lock_offset(tdb->header.hash_size)) {
 		tdb->ecode = TDB_ERR_LOCK;
@@ -240,24 +268,14 @@ int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype,
 	if (tdb->flags & TDB_NOLOCK)
 		return 0;
 
-	for (i=0; i<tdb->num_lockrecs; i++) {
-		if (tdb->lockrecs[i].off == offset) {
-			if (tdb->lockrecs[i].count == 0) {
-				/*
-				 * Can't happen, see tdb_unlock(). It should
-				 * be an assert.
-				 */
-				TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_lock: "
-					 "lck->count == 0 for offset %u",
-					 offset));
-			}
-			/*
-			 * Just increment the in-memory struct, posix locks
-			 * don't stack.
-			 */
-			tdb->lockrecs[i].count++;
-			return 0;
-		}
+	new_lck = find_nestlock(tdb, offset);
+	if (new_lck) {
+		/*
+		 * Just increment the in-memory struct, posix locks
+		 * don't stack.
+		 */
+		new_lck->count++;
+		return 0;
 	}
 
 	new_lck = (struct tdb_lock_type *)realloc(
@@ -275,12 +293,10 @@ int tdb_nest_lock(struct tdb_context *tdb, uint32_t offset, int ltype,
 		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;
 }
@@ -332,8 +348,7 @@ int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype,
 		    bool mark_lock)
 {
 	int ret = -1;
-	int i;
-	struct tdb_lock_type *lck = NULL;
+	struct tdb_lock_type *lck;
 
 	if (tdb->flags & TDB_NOLOCK)
 		return 0;
@@ -344,13 +359,7 @@ int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype,
 		return ret;
 	}
 
-	for (i=0; i<tdb->num_lockrecs; i++) {
-		if (tdb->lockrecs[i].off == offset) {
-			lck = &tdb->lockrecs[i];
-			break;
-		}
-	}
-
+	lck = find_nestlock(tdb, offset);
 	if ((lck == NULL) || (lck->count == 0)) {
 		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlock: count is 0\n"));
 		return -1;
@@ -373,17 +382,12 @@ int tdb_nest_unlock(struct tdb_context *tdb, uint32_t offset, int ltype,
 	} else {
 		ret = tdb->methods->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
@@ -420,18 +424,7 @@ int tdb_unlock(struct tdb_context *tdb, int list, int ltype)
  */
 int tdb_transaction_lock(struct tdb_context *tdb, int ltype)
 {
-	if (tdb->transaction_lock_count > 0) {
-		tdb->transaction_lock_count++;
-		return 0;
-	}
-
-	if (tdb->methods->brlock(tdb, ltype, TRANSACTION_LOCK, 1, TDB_LOCK_WAIT) == -1) {
-		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_transaction_lock: failed to get transaction lock\n"));
-		tdb->ecode = TDB_ERR_LOCK;
-		return -1;
-	}
-	tdb->transaction_lock_count++;
-	return 0;
+	return tdb_nest_lock(tdb, TRANSACTION_LOCK, ltype, TDB_LOCK_WAIT);
 }
 
 /*
@@ -439,24 +432,14 @@ int tdb_transaction_lock(struct tdb_context *tdb, int ltype)
  */
 int tdb_transaction_unlock(struct tdb_context *tdb, int ltype)
 {
-	int ret;
-	if (tdb->transaction_lock_count > 1) {
-		tdb->transaction_lock_count--;
-		return 0;
-	}
-	ret = tdb->methods->brunlock(tdb, ltype, TRANSACTION_LOCK, 1);
-	if (ret == 0) {
-		tdb->transaction_lock_count = 0;
-	}
-	return ret;
+	return tdb_nest_unlock(tdb, TRANSACTION_LOCK, ltype, false);
 }
 
 
-
-
-/* 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) {
@@ -474,16 +457,20 @@ static int _tdb_lockall(struct tdb_context *tdb, int ltype,
 		tdb->ecode = TDB_ERR_LOCK;
 		return -1;
 	}
-	
-	if (tdb->num_locks != 0) {
+
+	if (tdb_have_extra_locks(tdb)) {
 		/* can't combine global and chain locks */
 		tdb->ecode = TDB_ERR_LOCK;
 		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->methods->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)));
 		}
@@ -491,7 +478,10 @@ 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;
 }
@@ -499,7 +489,7 @@ 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) {
@@ -507,7 +497,14 @@ 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;
 	}
@@ -518,8 +515,7 @@ static int _tdb_unlockall(struct tdb_context *tdb, int ltype, bool mark_lock)
 	}
 
 	if (!mark_lock &&
-	    tdb->methods->brunlock(tdb, ltype,
-				   FREELIST_TOP, 4*tdb->header.hash_size)) {
+	    tdb->methods->brunlock(tdb, ltype, FREELIST_TOP, 0)) {
 		TDB_LOG((tdb, TDB_DEBUG_ERROR, "tdb_unlockall failed (%s)\n", strerror(errno)));
 		return -1;
 	}
@@ -534,27 +530,27 @@ 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;
 }
@@ -563,20 +559,20 @@ 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;
 }
@@ -585,7 +581,7 @@ 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
@@ -691,3 +687,54 @@ int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off)
 			count++;
 	return (count == 1 ? tdb->methods->brunlock(tdb, F_RDLCK, off, 1) : 0);
 }
+
+bool tdb_have_extra_locks(struct tdb_context *tdb)
+{
+	unsigned int extra = tdb->num_lockrecs;
+
+	/* A transaction holds the lock for all records. */
+	if (!tdb->transaction && tdb->allrecord_lock.count) {
+		return true;
+	}
+
+	/* We always hold the active lock if CLEAR_IF_FIRST. */
+	if (find_nestlock(tdb, ACTIVE_LOCK)) {
+		extra--;
+	}
+
+	/* In a transaction, we expect to hold the transaction lock */
+	if (tdb->transaction && find_nestlock(tdb, TRANSACTION_LOCK)) {
+		extra--;
+	}
+
+	return extra;
+}
+
+/* The transaction code uses this to remove all locks.  Note that this
+   may include OPEN_LOCK. */
+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, 0);
+		tdb->allrecord_lock.count = 0;
+	}
+
+	for (i=0;i<tdb->num_lockrecs;i++) {
+		struct tdb_lock_type *lck = &tdb->lockrecs[i];
+
+		/* Don't release transaction or active locks! */
+		if (tdb->transaction && lck->off == TRANSACTION_LOCK) {
+			tdb->lockrecs[extra++] = *lck;
+		} else if (lck->off == ACTIVE_LOCK) {
+			tdb->lockrecs[extra++] = *lck;
+		} else {
+			tdb_brunlock(tdb, lck->ltype, lck->off, 1);
+		}
+	}
+	tdb->num_lockrecs = extra;
+	if (tdb->num_lockrecs == 0) {
+		SAFE_FREE(tdb->lockrecs);
+	}
+}