zookeeper-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "H Y (JIRA)" <j...@apache.org>
Subject [jira] [Created] (ZOOKEEPER-2878) some issues in c code of lock of recipes
Date Mon, 21 Aug 2017 06:34:01 GMT
H Y created ZOOKEEPER-2878:

             Summary: some issues in c code of lock of recipes 
                 Key: ZOOKEEPER-2878
                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2878
             Project: ZooKeeper
          Issue Type: Bug
          Components: recipes
            Reporter: H Y
            Priority: Minor

There are three issues in the c code of lock.
1, It not multi-thread safe, because  pmutex is the local mutex of zkr_lock_mutex_t, if there
are more than one thread is calling zkr_lock_lock, it may watch children node fail(retry_zoowexists
may return NOT ZOK, 'unable to watch my predecessor' will output. I suggest that changing
the pmutex to global.
2,child_floor function is not correct, it should compare the sequence of the node.(zoo_lock.c,
line145 should be 'if (strcmp((sorted_data[i] + 19), (element + 19)) < 0)'
3, Logic mistaking in zkr_lock_operation of zoo_lock.c at line 256. mutex->id should be
allocated by getName function. So, I think that we should delete from line 249 to 257.

This message was sent by Atlassian JIRA

View raw message