Atomic Accesses

6.3Locked access

When the ARLOCK[1:0] or AWLOCK[1:0] signals for a transaction show that it is a locked transfer then the interconnect must ensure that only that master is allowed access to the slave region until an unlocked transfer from the same master completes. The arbiter within the interconnect is used to enforce this restriction.

When a master starts a locked sequence of either read or write transactions it must ensure that it has no other outstanding transactions waiting to complete.

Any transaction with ARLOCK[1:0] or AWLOCK[1:0] set to indicate a locked sequence forces the interconnect to lock the following transaction. Therefore, a locked sequence must always complete with a final transaction that does not have ARLOCK[1:0] or AWLOCK[1:0] set to indicate a locked access. This final transaction is included in the locked sequence and effectively removes the lock.

When completing a locked sequence a master must ensure that all previous locked transactions are complete before issuing the final unlocking transaction. It must then ensure that the final unlocking transaction has fully completed before any further transactions are commenced.

The master must ensure that all transactions within a locked sequence have the same

ARID or AWID value.

Note

Locked accesses require that the interconnect prevents any other transactions occurring while the locked sequence is in progress and can therefore have an impact on the interconnect performance. It is recommended that locked accesses are only used to support legacy devices.

The following restrictions are recommended but not mandatory:

•keep all locked transaction sequences within the same 4KB address region

•limit locked transaction sequences to two transactions.

Atomic Accesses