METHOD FOR PREVENTING NODE CONTROLLER DEADLOCK AND NODE CONTROLLER

摘要

A method for preventing node controller deadlock, comprising: the node controller (NC) of a node receiving a request message sent by any node and writing the request message into a processing queue, the request message is used for requesting a system address (301); the NC monitoring whether a cache data block containing the system address is cached in other nodes: if so, then invalidating the cache data block containing the system address that is cached on other nodes, and returning directly a feedback message to a home agent (HA) when the NC receives a first monitoring message transmitted by the HA to avoid the first monitoring message being blocked by the request message when writing into the processing queue; the first monitoring message is used for monitoring whether the above system address is cached in other nodes; the feedback message is used for indicating that the above system address cached in other nodes is invalid, making the HA transmitting the system address cached thereon to a caching agent (CA) (302); and the NC transmitting the request message written into the processing queue to the HA (303). The technical solution can prevent node controller deadlock, and avoid the crash of a non uniform memory access (NUMA) system.