Abstract: A communication apparatus includes a plurality of switch ports, each switch port including one or more port buffers for buffering data that traverses the switch port. A switch fabric is coupled to transfer the data between the switch ports. A switch control unit is configured to reassign at least one port buffer of a given switch port to buffer a part of the data that does not enter or exit the apparatus via the given switch port, and to cause the switch fabric to forward the part of the data to a destination switch port via the at least one reassigned port buffer.

Abstract: An input/output (I/O) device includes a host interface for connection to a host device having a memory and a network interface, which is configured to receive, over a network, data packets associated with I/O operations directed to specified virtual addresses in the memory. Packet processing hardware is configured to translate the virtual addresses into physical addresses and to perform the I/O operations using the physical addresses, and upon an occurrence of a page fault in translating one of the virtual addresses, to transmit a response packet over the network to a source of the data packets so as to cause the source to refrain from transmitting further data packets while the page fault is serviced.

Abstract: An interface adapter for a packet network includes a first plurality of execution engines, coupled to a host interface so as to read from a memory work items corresponding to messages to be sent over the network, and to generate gather entries defining packets to be transmitted over the network responsive to the work items. A scheduling processor assigns the work items to the execution engines for generation of the gather entries. Switching circuitry couples the execution engines to a plurality of gather engines, which generate the packets responsive to the gather entries.

Abstract: A method for preventing deadlock in communication between a host software application and a network interface card (NIC), comprises writing a doorbell associated with at least one descriptor having a descriptor context to a buffer in the NIC, dropping at least one doorbell from the buffer if the buffer is full, thereby allowing a write of a new doorbell to the buffer, and recovering each dropped doorbell for further execution of descriptors associated with this doorbell. The descriptor execution is in order of posting by the application to the NIC. A system implementing the method comprises a doorbell drop mechanism and a doorbell recovery mechanism.

Abstract: A method for data storage includes mapping a queue pair (QP) of a channel adapter to a specified Fibre Channel (FC) exchange for communication with a storage device. Upon receiving at the channel adapter from a host computer a storage command directed to the storage device, the storage command is executed by transmitting data packets over a switched network from the channel adapter to the storage device using the specified exchange and performing a remote direct memory access (RDMA) operation on the channel adapter using the mapped QP.

Abstract: A network interface adapter includes a network interface and a client interface, for coupling to a client device so as to receive from the client device work requests to send messages over the network using a plurality of transport service instances. Message processing circuitry, coupled between the network interface and the client interface, includes an execution unit, which generates the messages in response to the work requests and passes the messages to the network interface to be sent over the network. A memory stores records of the messages that have been generated by the execution unit in respective lists according to the transport service instances with which the messages are associated. A completion unit receives the records from the memory and, responsive thereto, reports to the client device upon completion of the messages.

Abstract: A method for controlling access by processes running on a host device to a communication network includes assigning to each of the processes a respective doorbell address on a network interface adapter that couples the host device to the network and allocating instances of a communication service on the network, to be provided via the adapter, to the processes. Upon receiving a request submitted by a given one of the processes to its respective doorbell address to access one of the allocated service instances, the adapter conveys the data over the network using the specified instance of the service, subject to verifying, based on the doorbell address to which the request was submitted, that the specified instance was allocated to the given process.

Abstract: A method for controlling access by processes running on a host device to a communication network includes assigning to each of the processes a respective doorbell address on a network interface adapter that couples the host device to the network and allocating instances of a communication service on the network, to be provided via the adapter, to the processes. Upon receiving a request submitted by a given one of the processes to its respective doorbell address to access one of the allocated service instances, the adapter conveys the data over the network using the specified instance of the service, subject to verifying, based on the doorbell address to which the request was submitted, that the specified instance was allocated to the given process.

Abstract: A system and method of digital communication wherein a host on an InfiniBand network transmits Fibre Channel packets encapsulated within InfiniBand packets to a gateway which forwards the Fibre Channel packets to Fibre Channel device via a Fibre Channel network, and wherein Fibre Channel packets addressed to a host on an InfiniBand network are transmitted by a Fibre Channel device to a gateway, the gateway encapsulating the Fibre Channel packets within InfiniBand packets and transmitting the InfiniBand packets to an InfiniBand host, where the Fibre Channel packet is extracted.

Abstract: A network interface device includes a fabric interface, adapted to exchange messages over a switch fabric with a plurality of host processors, the messages containing data, and a network interface, including one or more ports adapted to be coupled to a network external to the switch fabric. Message processing circuitry is coupled between the fabric interface and the network interface, so as to enable at least first and second host processors among the plurality of the host processors to use a single one of the ports substantially simultaneously so as to transmit and receive frames containing the data over the network.

Abstract: A method for static rate flow control includes receiving a sequence of data packets for transmission over a network, including at least first and second packets having a common destination address on the network, the first and second packets having respective first and second lengths, and transmitting the first packet to the destination address. Responsive to transmitting the first packet, an entry is placed in a flow control table, and a timeout period is set for the entry responsive to the first length. The second packet is transmitted only after the timeout period has expired.

Abstract: A network interface adapter includes a network interface, coupled to send and receive data packets over a network and a host interface, for coupling to a host processor and to a system memory associated therewith. The system memory contains context information with respect to a plurality of transport service instances used to send and receive the data packets over the network, each of the data packets belonging to a respective one of the service instances. Packet processing circuitry, coupled between the network interface and the host interface, processes the data packets using the context information of the respective service instances. A cache memory associated with the packet processing circuitry is coupled to load from the system memory and store the context information of the respective transport service instances for the data packets being processed by the packet processing circuitry.

Abstract: A method for allocating a processing resource among multiple inputs includes defining a sequence of multiplexing iterations, each such iteration including a first plurality of windows, each such window containing a second plurality of time slots. A respective weight is assigned to each of the inputs, and each of the inputs is allotted one of the time slots in each of a respective number of the windows in each of the iterations, the respective number being determined by the respective weight. Each of the inputs is then provided with access to the processing resource during the time slots allotted thereto.

Abstract: A network interface adapter includes an outgoing packet generator, adapted to generate an outgoing request packet for delivery to a remote responder responsive to a request submitted by a host processor and a network output port, coupled to transmit the outgoing request packet over a network to the remote responder. A network input port receives an incoming response packet from the remote responder, in response to the outgoing request packet sent thereto, as well as an incoming request packet sent by a remote requester. An incoming packet processor receives and processes both the incoming response packet and the incoming request packet, and causes the outgoing packet generator, responsive to the incoming request packet, to generate, in addition to the outgoing request packet, an outgoing response packet for transmission to the remote requester.

Abstract: A method for communication over a network includes receiving from a host processor a descriptor defining a message including message data to be sent over the network, and responsive to the descriptor, generating a sequence of packets each containing a respective portion of the message data. An indication is entered in a selected packet among the packets in the sequence, other than the final packet, requesting that a recipient of the packets acknowledge the selected packet. Following an interruption in the sequence of the packets subsequent to the selected packet, sending of the packets in the sequence resumes beginning after the selected packet.

Abstract: A method and apparatus for conveying data over a packet-switching network (26). Data are received from a peripheral device (25) for transmission via the network to a memory (22) associated with a central processing unit (CPU) (21), followed by an interrupt signal from the peripheral device associated with the data. One or more data packets containing the data are sent over the network to a host network interface (32) serving the memory and the CPU, followed by an interrupt packet sent over the network to the host network interface. Responsive to the interrupt packet, an interrupt input of the CPU is asserted only after the one or more data packets have arrived at the host network interface.

Abstract: A method for preventing deadlock in communication between a host software application and a network interface card (NIC), comprises writing a doorbell associated with at least one descriptor having a descriptor context to a buffer in the NIC, dropping at least one doorbell from the buffer if the buffer is full, thereby allowing a write of a new doorbell to the buffer, and recovering each dropped doorbell for further execution of descriptors associated with this doorbell. The descriptor execution is in order of posting by the application to the NIC. A system implementing the method comprises a doorbell drop mechanism and a doorbell recovery mechanism.

Abstract: A method for link-level flow control includes establishing a plurality of logical links over a physical link between a transmitting entity and a receiving entity in a network. Respective maximum limits of transmission credits are assigned to the logical links, the credits corresponding to space available to the links in a dynamically allocable portion of a receive buffer at the receiving entity, such that a sum of the maximum limits for all of the logical links corresponds to an amount of space substantially larger than a total volume of the space in the dynamically allocable portion of the receive buffer.

Abstract: A method of direct memory access (DMA) includes receiving a first notification at a DMA engine that a first list of descriptors has been prepared, each of the descriptors in the list including an instruction for execution by the DMA engine and a link to a succeeding one of the descriptors, except for a final descriptor in the list, which has a null link. The DMA engine reads and executes the descriptors in the first list. When the DMA engine receives a second notification that a second list of the descriptors has been prepared, it rereads at least a part of the final descriptor in the first list to determine a changed value of the link, indicating a first descriptor in the second list. It then reads and executes the descriptors in the second list responsive to the changed value of the link.

Abstract: A method for allocating a processing resource among multiple inputs includes defining a sequence of multiplexing iterations, each such iteration including a first plurality of windows, each such window containing a second plurality of time slots. A respective weight is assigned to each of the inputs, and each of the inputs is allotted one of the time slots in each of a respective number of the windows in each of the iterations, the respective number being determined by the respective weight. Each of the inputs is then provided with access to the processing resource during the time slots allotted thereto.