For an overview of the changes, see this introduction article for MySQL 5.5 with InnoDB 1.1. The following is a condensed version of the change log.
Fix for Bug #52580: Crash in ha_innobase::open on executing INSERT with concurrent ALTER TABLE.
Change in MySQL Bug #51557 releases the mutex LOCK_open before ha_innobase::open(), causing racing condition for index translation table creation. Fix it by adding dict_sys mutex for the operation.
Add support for multiple buffer pools.
Fix Bug #26590: MySQL does not allow more than 1023 open transactions. Create additional rollback segments on startup. Reduce the upper limit of total rollback segments from 256 to 128. This is because we can't use the sign bit. It has not caused problems in the past because we only created one segment. InnoDB has always had the capability to use the additional rollback segments, therefore this patch is backward compatible. The only requirement to maintain backward compatibility has been to ensure that the additional segments are created after the double write buffer. This is to avoid breaking assumptions in the existing code.
Implement Performance Schema in InnoDB. Objects in four different modules in InnoDB have been performance instrumented, these modules are: mutexes, rwlocks, file I/O, and threads. We mostly preserved the existing APIs, but APIs would point to instrumented function wrappers if performance schema is defined. There are 4 different defines that control the instrumentation of each module. The feature is off by default, and will be compiled in with a special build option, and require a configure option to turn it on when the server boots.
Implement the buf_pool_watch for DeleteBuffering in the page hash table. This serves two purposes. It allows multiple watches to be set at the same time (by multiple purge threads) and it removes a race condition when the read of a block completes about the time the buffer pool watch is being set.
Introduce a new mutex to protect flush_list. Redesign mtr_commit() in a way that log_sys mutex is not held while all mtr_memos are popped and is released just after the modified blocks are inserted into the flush_list. This should reduce contention on log_sys mutex.
Implement the global variable
innodb_change_buffering, with the
none: buffer nothing
inserts: buffer inserts (like InnoDB so
deletes: buffer delete-marks
changes: buffer inserts and delete-marks
purges: buffer delete-marks and deletes
all: buffer all operations (insert,
The default is
all. All values except
inserts will make
InnoDB write new-format records to the change buffer, even for
Provide support for native AIO on Linux.