Granted Patent B2

US 12,210,473 · App. 17/980,685 · Granted Jan 28, 2025

Inventors: Nam Sung Kim (Champaign, IL); Mohammad Alian (Urbana, IL)

Assignee: The Board of Trustees of the University of Illinois

G06F13/4027G06F12/10G06F13/1673G06F2212/1016

Abstract

Claims (65)

1. A computing device comprising:

a host processor to:

execute a host driver to create a host-side interface, the host-side interface emulating a first virtual Ethernet interface; and

assign the host-side interface a first medium access control (MAC) address and a first Internet Protocol (IP) address;

a substrate on which is disposed memory components; and

a memory channel network (MCN) processor disposed on the substrate and coupled between the memory components and the host processor, the MCN processor to:

execute an MCN driver to create an MCN-side interface, the MCN-side interface emulating a second virtual Ethernet interface; and

assign the MCN-side interface a second MAC address and a second IP address, which identify the MCN processor as an MCN network node to the host processor; and

wherein the MCN driver, when executed by the MCN processor, is to:

create a network device object;

set up the network device object as an Ethernet device; and

register the Ethernet device with a kernel of the host processor, to make the MCN-side interface visible to a host operating system executing on the host processor.

2. The computing device of claim 1 , wherein, upon receipt of a network packet having a third IP address directed to a second MCN-side interface, the host-side interface is to forward the network packet to the second MCN-side interface comprising a second MCN network node.

3. The computing device of claim 1 , wherein the MCN processor is further to configure the MCN-side interface so that outgoing network packets from the MCN network node are forwarded to the host-side interface for routing.

4. The computing device of claim 1 , wherein the MCN processor comprises:

a core to execute the MCN driver; and

a local buffer coupled to the core, the local buffer to store MCN messages that are sent to and received from the host-side interface, each MCN message comprising a packet length and packet data.

5. The computing device of claim 4 , wherein, in response to a network packet received at the MCN-side interface from a network stack, the MCN-side interface is to:

read a transmit start pointer and a transmit end pointer from the local buffer for the network packet;

in response to sufficient space being available in a transmission buffer of the local buffer, write the packet length followed by the packet data into the transmission buffer, starting at a buffer address to which points the transmit end pointer;

update a value for the transmit end pointer in the local buffer; and

set a transmit polling field of the local buffer to a non-zero value to indicate that a new packet is enqueued in the transmission buffer.

6. The computing device of claim 4 , wherein the local buffer comprises a transmit polling field, and the host-side interface comprises a polling agent to:

periodically read the transmit polling field in a plurality of local buffers across a plurality of MCN nodes, to determine whether there is a pending network packet; and

upon detection of a pending network packet in one of the plurality of local buffers:

read a transmit start pointer and a transmit end pointer from the local buffer for the pending network packet;

read a cache line at a location of the local buffer to which points the transmit start pointer;

retrieve the packet length and a packet destination MAC address from the cache line; and

transmit the pending network packet to a packet forwarding engine of the host-side interface.

7. The computing device of claim 6 , wherein the host-side interface is further to:

if the transmit start pointer moved by a number of bytes read from a transmission buffer of the local buffer is not equal to a value of the transmit end pointer, read an additional at least one more cache line from the plurality of local buffers; and

if the transmit start pointer moved by a number of bytes read from the transmission buffer of the plurality of local buffers is equal to a value of the transmit end pointer, reset a value of a transmit polling field of the local buffer and exit.

8. The computing device of claim 4 , wherein the host-side interface comprises a packet forwarding engine to, in response to receipt of a network packet:

determine that a destination MAC address of the network packet matches the first MAC address;

allocate a socket buffer;

copy data of the network packet from a receiving buffer of the local buffer to the socket buffer; and

transmit the data from the socket buffer to a network stack for processing.

9. The computing device of claim 4 , wherein the host-side interface comprises a packet forwarding engine to, in response to receipt of a network packet:

determine that a destination MAC address of the network packet matches a reserved address for broadcast;

allocate a socket buffer;

copy data of the network packet from a receiving buffer of the local buffer to the socket buffer;

transmit the data from the socket buffer to a network stack for processing; and

transmit the data from the socket buffer to a plurality of MCN network nodes.

10. The computing device of claim 4 , wherein the host-side interface comprises a packet forwarding engine to, in response to receipt of a network packet:

determine that a destination MAC address of the network packet matches a second address of a destination MCN node of a plurality of MCN network nodes; and

transmit the network packet to the destination MCN node.

11. The computing device of claim 1 , wherein the substrate is an MCN dual in-line memory module (DIMM).

12. The computing device of claim 1 , wherein the MCN processor is further to operate according to a memory-based physical layer (PHY) protocol.

13. A method comprising:

executing, by a host processor, a host driver to create a host-side interface, the host-side interface emulating a first virtual Ethernet interface; and

assigning, by the host processor, the host-side interface a first medium access control (MAC) address and a first Internet Protocol (IP) address;

executing, by an MCN processor co-located on a substrate with memory components, an MCN driver to create an MCN-side interface, the MCN-side interface emulating a second virtual Ethernet interface, wherein the MCN processor is coupled between the memory components and the host processor, wherein executing the MCN driver further comprises:

creating a network device object;

setting up the network device object as an Ethernet device; and

registering the Ethernet device with a kernel of the host processor, to make the MCN-side interface visible to a host operating system executing on the host processor; and

assigning, by the MCN processor, the MCN-side interface a second MAC address and a second IP address, which identify the MCN processor as an MCN network node to the host processor.

14. The method of claim 13 , wherein, in response to receiving a network packet having a third IP address directed to a second MCN-side interface, the method further comprises forwarding, by the host-side interface, the network packet to the second MCN-side interface, wherein the second MCN-side interface comprises a second MCN network node.

15. The method of claim 13 , further comprising configuring, by the MCN processor, the MCN-side interface so that outgoing network packets from the MCN network node are forwarded to the host-side interface for routing.

16. The method of claim 13 , further comprising storing, by the MCN processor, at a local buffer, MCN messages that are sent to and received from the host-side interface, each MCN message comprising a packet length and packet data.

17. The method of claim 16 , wherein, in response to receiving a network packet at the MCN-side interface from a network stack, the method further comprises:

reading, by the MCN-side interface, a transmit start pointer and a transmit end pointer from the local buffer for the network packet;

wherein, in response to sufficient space being available in a transmission buffer of the local buffer, the method further comprises writing the packet length followed by the packet data into the transmission buffer, starting at a buffer address to which points the transmit end pointer;

updating a value for the transmit end pointer in the local buffer; and

setting a transmit polling field of the local buffer to a non-zero value to indicate that a new packet is enqueued in the transmission buffer.

18. The method of claim 13 , wherein the method further comprises operating the MCN processor according to a memory-based physical layer (PHY) protocol.

Assignments (1)

Continuity (3)

References Cited (98)