Detecting and recovering from timeouts in scalable mesh networks in processor-based devices

An integrated circuit (IC), comprising:

• an interconnect comprising a scalable mesh network;
• a plurality of agents communicatively coupled to the interconnect via a respective plurality of bridge devices, wherein:
• the plurality of agents comprises a source agent and a target agent;
• the plurality of bridge devices comprises a source bridge device and a target bridge device;
• the source bridge device is configured to provide an interface between a communications protocol of the interconnect and a first interface protocol of the source agent;
• and
• the target bridge device is configured to provide an interface between a communications protocol of the interconnect and a second interface protocol of the target agent;
• and
• the target bridge device configured to:
• receive, from the source agent, a first transaction directed to the target agent;
• responsive to receiving the first transaction, initiate a timeout counter;
• and
• responsive to expiration of the timeout counter:
• determine whether no response to the first transaction was received from the target agent;
• and
• responsive to determining that no response to the first transaction was received, transmit, to the source bridge device, an indication that no response to the first transaction was received,
• wherein:
• the second interface protocol comprises a custom interface protocol;
• the target bridge device comprises:
• a plurality of monitors corresponding to a respective plurality of message classes;
• and
• a plurality of timeout counters corresponding to the respective plurality of message classes;
• the target bridge device is further configured to identify, using a monitor of the plurality of monitors, a message class of the first transaction;
• the timeout counter comprises a timeout counter corresponding to the message class among the plurality of timeout counters;
• and
• the indication comprises a quiesce packet corresponding to the message class, the quiesce packet comprising an identifier of the target agent.

A method for detecting and recovering from timeouts in scalable mesh networks, comprising:

• receiving, by a target bridge device from a source agent via an interconnect comprising a scalable mesh network, a first transaction directed to a target agent;
• responsive to receiving the first transaction, initiating, by the target bridge device, a timeout counter;
• and
• responsive to expiration of the timeout counter:
• determining, by the target bridge device, that no response to the first transaction was received from the target agent;
• and
• responsive to determining that no response to the first transaction was received, transmitting, by the target bridge device to a source bridge device, an indication that no response to the first transaction was received, wherein:
• a second interface protocol comprises an Advanced Microcontroller Bus Architecture (AMBA) Advanced extensible Interface (AXI) interface protocol;
• the method further comprises, further responsive to determining that no response to the first transaction was received, generating, by the target bridge device, a first error response;
• and
• the indication comprises the first error response,
• wherein the method further comprises, further responsive to determining that no response to the first transaction was received:
• placing the target bridge device in a timeout mode;
• receiving, by the target bridge device from the source agent, a second transaction directed to the target agent;
• and
• responsive to receiving the second transaction in the timeout mode:
• generating, by the target bridge device, a second error response;
• and
• transmitting, by the target bridge device, the second error response to the source bridge device, and wherein the method further comprises,
• receiving, by the target agent, an indication to correct a nonresponsive condition;
• receiving, by the target bridge device, an indication to return to a normal operation mode;
• and
• responsive to receiving the indication to return to the normal operation mode, placing the target bridge device in the normal operation mode.

A method for detecting and recovering from timeouts in scalable mesh networks, comprising:

• receiving, by a target bridge device from a source agent via an interconnect comprising a scalable mesh network, a first transaction directed to a target agent;
• responsive to receiving the first transaction, initiating, by the target bridge device, a timeout counter;
• and
• responsive to expiration of the timeout counter:
• determining, by the target bridge device, that no response to the first transaction was received from the target agent;
• and
• responsive to determining that no response to the first transaction was received, transmitting, by the target bridge device to a source bridge device, an indication that no response to the first transaction was received,
• wherein:
• a second interface protocol comprises a custom interface protocol;
• the target bridge device comprises:
• a plurality of monitors corresponding to a respective plurality of message classes;
• and
• a plurality of timeout counters corresponding to the respective plurality of message classes;
• the method further comprises identifying, by a monitor of the plurality of monitors, a message class of the first transaction;
• the timeout counter comprises a timeout counter corresponding to the message class among the plurality of timeout counters;
• and
• the indication comprises a quiesce packet corresponding to the message class, the quiesce packet comprising an identifier of the target agent.