summaryrefslogtreecommitdiff
path: root/lib/tdb/docs/mutex.txt
diff options
context:
space:
mode:
Diffstat (limited to 'lib/tdb/docs/mutex.txt')
-rw-r--r--lib/tdb/docs/mutex.txt4
1 files changed, 2 insertions, 2 deletions
diff --git a/lib/tdb/docs/mutex.txt b/lib/tdb/docs/mutex.txt
index a5a75420c1d..0003c7f64be 100644
--- a/lib/tdb/docs/mutex.txt
+++ b/lib/tdb/docs/mutex.txt
@@ -15,7 +15,7 @@ with a fcntl lock.
The external locking API of tdb also allows one to lock the complete database, and
ctdb uses this facility to freeze databases during a recovery. While the
so-called allrecord lock is held, all linked lists and all individual records
-are frozen alltogether. Tdb achieves this by locking the complete file range
+are frozen altogether. Tdb achieves this by locking the complete file range
with a single fcntl lock. Individual 1-byte locks for the linked lists
conflict with this. Access to records is prevented by the one large fnctl byte
range lock.
@@ -73,7 +73,7 @@ The setpshared functions indicate to the kernel that the mutex is about to be
shared between processes in a common shared memory area.
The process shared posix mutexes have the potential to replace fcntl locking
-to coordinate mmap access for tdbs. However, they are missing the criticial
+to coordinate mmap access for tdbs. However, they are missing the critical
auto-cleanup property that fcntl provides when a process dies. A process that
dies hard while holding a shared mutex has no chance to clean up the protected
data structures and unlock the shared mutex. Thus with a pure process shared