BUG#14727815 - CRASH IN PTHREAD_RWLOCK_WRLOCK/SRW_UNLOCK
IN QUERY CACHE CODE DESCRIPTION: MySQL Server crashes sporadically when Query Caching is on and the server has high contention among clients. ANALYSIS : Scenario 1: In Query_cache::move_by_type() when handling RESULT or its related blocks, Write Lock is acquired on its parent Query block. However the next and prev pointers are cached in local variables before lock acquisition. In an extremely high contention scenario there exists a possibility that Query_cache::append_result_data() is operating on the same query block and as a consequence might append a new Result block to the end of Result blocks Linked List of the Query. This would manipulate the next, prev pointers of the Block being processed in move_by_type(), however the local pointers still point to previous nodes there by causing Data Corruption leading to crash. Scenario 2: In Windows SDK "BOOL" is typedefed as "int" and BOOLEAN is typedefed as "usigned char". The function pointer definition "srw_bool_func" mistakenly uses BOOL instead of BOOLEAN thereby virtually making the function my_TryAcquireSRWLockExclusive() always succeed because only the LSB of EAX has the actual result of the call, however due to type mismatch all bytes of EAX are used for evaluation. Again during high contention scenarios in Query_cache::free_old_query() calls try_lock_writing() on a Query, this call always succeeds and the query is freed, even though it is used by some other thread, in this case Query_cache::send_result_to_client() was using it and the code causes a crash because it accessed free or reallocated memory. FIX : Scenario 1: The next, prev pointers are now accessed only after Lock acquisition in Query_cache::move_by_type(). Scenario 2: In the definition of "srw_bool_func" BOOL has been replaced with "BOOLEAN"
Showing
Please register or sign in to comment