Systems and methods for network and I/O device drivers

Abstract

The present invention is directed to systems and methods for transferring storage data and network data using the same interface circuit. Storage data is transferred from an upper layer storage driver to a lower layer storage driver. Network data is transferred from a network driver to the lower layer storage drive. The storage data and the network data are transferred to a communications link interface circuit, wherein the storage data is transferred using a storage protocol and the network data is transferred using a network protocol.

Claims

What is claimed is:

1. A system for sending both network and storage protocols over the same host adapter, said system comprising:

a network driver, including at least one network miniport;

a SCSI miniport driver layer coupled to said network driver to receive network related information from said network miniport;

a SCSI driver layer higher than said SCSI miniport driver layer, said higher SCSI layer passing storage related information to said SCSI miniport driver layer;

a bus adapter coupled to said SCSI miniport driver layer, wherein said SCSI miniport driver layer transfers both network and storage related information to said bus adapter; and

a communication medium link coupled to said bus adapter, wherein said communication link is configured to receive both said network related information and said storage related information transferred from the SCSI miniport driver layer, and said communication medium link is configured to couple a first computer system to a second computer system.

2. The system for sending both network and storage protocols over the same host adapter as defined in claim 1, wherein said network driver includes at least a intermediate driver.

3. The system for sending both network and storage protocols over the same host adapter as defined in claim 1, wherein said communication medium link is fibre channel-compatible.

4. The system for sending both network and storage protocols over the same host adapter as defined in claim 1, wherein said communication medium link is SCSI-compatible.

5. The system for sending both network and storage protocols over the same host adapter as defined in claim 1, wherein said network related information is transferred using an Ethernet protocol.

6. The system for sending both network and storage protocols over the same host adapter as defined in claim 1, wherein said storage related data is transferred using a SCSI protocol.

7. The system for sending both network and storage protocols over the same host adapter as defined in claim 1, wherein said storage related data is transferred using a fibre channel protocol.

8. The system of claim 1, wherein the SCSI miniport driver can process multiple packets send in a single call by the network miniport.

9. The system of claim 1, wherein the SCSI miniport driver builds control blocks for sending unencapsulated data packets along with scatter/gather lists to the bus adapter.

10. A method of transferring data using a storage protocol and a network protocol over the same host adapter, said method comprising the acts of:

transferring storage data from an upper layer storage driver layer to a lower layer storage driver;

transferring network data from a network driver to said lower layer storage driver; and

transferring said storage data and said network data to a communication link interface circuit.

11. The method of transferring data as defined in claim 10, wherein said lower layer storage driver is a miniport driver.

12. The method of transferring data as defined in claim 10, wherein said upper layer storage driver is a port driver.

13. The method of transferring data as defined in claim 10, wherein said network driver includes at least a network miniport driver.

14. The method of transferring data as defined in claim 10, further comprising the act of encapsulating said network data in a storage command.

15. The method of transferring data as defined in claim 10, further comprising the act of passing a network scatter/gather list to said lower layer storage driver.

16. The method of transferring data as defined in claim 10, further comprising the act of determining if said lower layer storage driver supports an interface to said network driver.

17. A system for sending both network and storage data using the same driver, said system comprising:

a network driver; and

a storage driver coupled to said network driver to receive network related information, said storage driver configured to receive storage related information from an operating system, and said storage driver configured to transfer both said network and said storage related data to an interface card.

18. The system for sending both network and storage data as defined in claim 17, wherein said storage driver further comprises at least one entry point for communication with said network driver.

19. The system for sending both network and storage data as defined in claim 17, further comprising a spinlock managed by said storage driver, said spinlock used to control access to at least a first resource.

20. The system for sending both network and storage data as defined in claim 17, wherein said interface card is configured to be coupled to a fibre channel network.

21. The system for sending both network and storage data as defined in claim 17, wherein said interface card is configured to be coupled to a SCSI bus.

22. A device driver for communicating both network and storage information over the same interface circuit, said device driver comprising:

a first entry point used to receive storage data intended for transfer to at least one communication link; and

a second entry point used to receive network data intended for transfer to said at least one communication link; and

an interface used to communicate said storage data using a storage protocol and said network data using a network protocol to a communication link interface circuit.

23. The device driver as defined in claim 22, further comprising at least a standard Windows NT class layer.

24. The device driver as defined in claim 22, further comprising at least a class layer, a port layer, and a miniport layer, wherein said second entry point is associated with said miniport layer.

25. A networked system, including at least two computer systems coupled by a communication link which transfers data between said computer systems using both network and storage protocols, said system comprising:

at least a first processor located in at least a first of said at least two computer systems;

a first operating system configured to be executed by said first processor;

a first network driver configured to be executed by said first computer system, said first network driver having an interface used to receive network related information from said operating system; and

a first storage driver, configured to be executed by said first computer system, said first storage driver having at least a first interface used to receive said network related information from said network driver, and said first storage driver having at least a second interface used to receive storage related information from said operating system;

a first bus adapter in communication with said first storage driver, said first bus adapter used to receive both said network related information and said storage related information; and

a communication medium in communication with said first adapter, said communication medium used to transfer both said network and said storage related data, using respectively, a network protocol and a storage protocol, to at least a second of said at least two computer systems.

26. The networked system as defined by claim 25, further comprising:

a second host adapter coupled to at least a second processor in a second of said at least two computers, said second host adapter coupled to said communication medium; and

a second storage driver, configured to be executed by said second computer system, said second storage driver having at least a first interface used to receive said network related information and storage related information from said second host adapter;

a second network driver configured to be executed by said second computer system, said second network driver having at least a first interface for receiving network related information from said storage driver; and

a second operating system configured to be executed by at least a first processor of said second computer system, said second operating system having at least one interface for receiving information from said storage driver and said network driver.

27. The networked system as defined by claim 25, wherein said communication medium is a fibre channel.

28. The networked system as defined by claim 25, wherein said communication medium is SCSI-compatible.

29. The networked system as defined by claim 25, wherein said first operating system is Windows NT.

30. The networked system as defined by claim 25, wherein said first computer system further comprises at least a second processor configured to execute said first operating system.

31. A networked system, including at least two computer systems coupled by at least two communication links, said networked system comprising:

a first computer system;

a second computer system;

a first communication link coupling said first computer system and said second computer system, said first communication link configured to transfer data using an I/O protocol;

a second communication link coupling said first computer system and said second computer system, said second communication link configured to transfer data using a network protocol;

a failure detection routine configured to detect the failure of said first communication link and said second communication link, wherein upon detecting the failure of either of said first and said second communication links, said failure detection routine causes both I/O and network protocols to be used to transfer, respectively, storage and network data, on the non-failed communication link.

32. The networked system as defined by claim 31, wherein said I/O protocol is a SCSI protocol.

33. A method of providing redundant communication between at least two computer systems, said method comprising the acts of:

transferring data over a first communication link using an I/O protocol;

transferring data over a second communication link using a network protocol;

detecting when either said first communication link or said second communication link have experienced at least a first mode of failure;

transferring data using said network protocol over said first communication link in response to detecting said failure in said second communication link; and

transferring data using said I/O protocol over said second communication link in response to detecting said failure in said first communication link.

34. A device driver for communicating both network and storage information using the same interface circuit, said device driver comprising:

a first interface used to receive storage data intended for transfer to at least one communication link; and

a second interface used to receive network data intended for transfer to at least one communication link; and

a third interface used to communicate said storage data using said storage protocol and said network data, where said network data is encapsulated using said storage protocol, to a communication link interface circuit.

35. A device driver for communicating both network and I/O information using the same interface circuit using respectively network and I/O protocols, said device driver comprising:

a means for receiving I/O data intended for transfer to at least one communication link; and

a means for receiving network data intended for transfer to at least one communication link; and

a means for communicating said I/O data using an I/O protocol and said network data using said network protocol to a means for interfacing to a communication medium.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to methods and systems for handling network and I/O communications, and in particular, to network and I/O device drivers.

2. Description of the Related Art

Computer systems typically utilize an operating system that provides an interface between application programs and the hardware. For example, operating systems are used to schedule tasks, allocate storage, and handle the interface to peripheral hardware, such as hard drives, optical disk drives, tape drives, network devices and the like. The operating system may be split into a kernel and various system programs. The system programs use facilities provided by the kernel to perform higher-level housekeeping task, often acting as servers in a client-server relationship.

As previously discussed, an operating system interfaces application programs to hardware. For example, an application program may transmit a request to access peripheral hardware. This request is received by the operating system, which in turn translates the request to a format useable by the device. The operating system then transmits the translated request to the appropriate device and handles communications with the device.

The software used to control a peripheral device is typically called a device driver. A driver is usually a routine or set of routines that implements device-specific aspects of generic I/O operations. The device driver may be responsible for accessing the hardware registers of the device, starting and completing I/O operations, performing error processing, and often includes an interrupt handler to service interrupts generated by the device. Device drivers are typically kernel mode drivers and often form part of the lowest level of the operating system kernel, with which they are linked when the kernel is built. Some operating systems, such as Windows NT, have loadable device drivers that can be installed from files after the operating system is running.

In Windows NT, an I/O manager handles communication between a driver and an application program or kernel component. The I/O manager responds to an I/O request by issuing an I/O request packet to the appropriate device driver. The driver translates the request into an appropriate form for the targeted device and causes the device to begin the requested I/O operation. Once the device completes the I/O operation, the device generates an interrupt which is serviced by the device driver. The I/O manager then completes the I/O request by performing a variety of operations, including, for example, recording the outcome of the I/O operation.

Some operating systems, such as Windows NT, or UNIX, use a layered approach in implementing device drivers. These operating systems allow several driver layers to exist between an application program and a device. For example, one or more intermediate drivers may be layered on top of a physical device driver. These intermediate driver layers allow hardware-specific operations to be separated from more general management issues. In Windows NT parlance, one such intermediate driver type is referred to as a "class driver," while the hardware driver is called a "port driver."

One example of a device driver is a SCSI drive device driver. The SCSI device driver typically consists of several layers. For example, the SCSI device driver may include a class driver, a SCSI port driver, and a SCSI miniport driver. When an application program issues a read request for a file located on the disk, the operating system invokes the appropriate SCSI class driver and passes the read request to the class driver. The class driver translates the received I/O request packets into I/O request packets with system defined SCSI request blocks (SRBs) containing SCSI command descriptor blocks. The translated I/O request packets are then sent to the next lower driver, which may be, for example, a SCSI port driver. The SCSI port driver translates the SCSI request blocks from the class driver and passes the SCSI request blocks and the command descriptor blocks to the SCSI miniport. The SCSI miniport driver is dynamically linked with the SCSI port driver and provides hardware-specific support for a particular SCSI host bus adapter (HBA). The SCSI drive performs the read operation and generates an interrupt to the SCSI miniport that then services the interrupt.

Another example of a device driver is a network driver. The network driver handles communication with networking hardware, such as a NIC (Network Interface Card). Many operating systems, such as Windows NT, use driver layering to disengage network protocol management from actual data transfers. Thus, a network driver may include a Windows NT LAN (Large Area Network) driver followed by an NDIS or NIC miniport driver. The network driver layers may be interconnected using a Windows NT NDIS (Network Device Interface Specification) interface. In conventional systems, the NDIS miniport driver interfaces to the NIC. Thus, in conventional systems, the NDIS miniport performs the hardware-specific operations needed to manage the NIC.

Many conventional computer systems include at least two interface cards, an I/O HBA and a network interface card, to respectively handle I/O and network protocols. As described above, these cards are managed respectively by a SCSI miniport and an NDIS miniport. Each card is typically connected to separate communication media. For example, the I/O HBA might be connected to a high speed Fibre Channel communication link, which in turn might be connected to one or more server systems or other computer systems. The Fibre Channel link would then handle storage related communications. The NDIS miniport might be connected via a NIC to an Ethernet link, which in turn may be connected to other computer systems. The Ethernet link would then handle the network related communications, such as, in the case of a clustered server system, "heartbeat" information. The network heartbeat traffic is typically much less than the storage related traffic, and therefore the lower bandwidth, lower cost, Ethernet link is considered well suited to handle such lower bandwidth communication. Thus, for example, in a typical clustered server system having host systems and storage subsystems, each host system would have at least one I/O HBA and one Ethernet interface card. In addition, each storage subsystem would have at least one I/O HBA. The host systems communicate heartbeat information to each other over the Ethernet link. Both the host systems and the storage subsystems communicate storage information over the Fibre Channel.

However, the conventional approach described above disadvantageously requires at least two slots in each host system computer, one for the storage or I/O HBA and one for the NIC, and two communication links, in order to handle both storage related communication and network related communication. This situation is further exacerbated in systems incorporating redundant channels for both the storage and communication links. In conventional systems, such redundancy requires four host bus adapters, which in turn require four card slots in each host computer system. However, many standard computers have a very limited number of slots available. Hence, in order to accommodate four HBAs, either a non-standard, larger chassis is required, or an additional chassis having additional slot positions must be connected to the host computer system. Either alternative results in an expensive, large system.

SUMMARY OF THE INVENTION

One embodiment of the present invention advantageously enables both network protocol and an I/O protocol to be transferred to a communication link using the same interface circuit. For example, in one embodiment two or more computer systems are interconnected using a communication link, such as a Fibre Channel link. Both an I/O protocol, such as a SCSI protocol, and a network protocol, such as an Internet protocol (IP), are used to transmit data over the same Fibre Channel link using the same host bus adapter. Thus, in comparison with conventional systems, one embodiment of the present requires half as many communication links and half as many host bus adapters to carry data using both SCSI and IP protocols.

The reduction in the number of communication links and the number of host bus adapters is even greater in networked systems having redundant links. For example, in one embodiment of the present invention, two or more computer systems are interconnected using two communication links. During normal operation, one link is used to carry data using an I/O protocol, such as SCSI, while the other link is used to carry data using a network protocol, such as IP. In the event of failure of either one of the two links, one embodiment of the present invention detects the failure and uses the remaining link to carry data using both the I/O and network protocols. Thus, one embodiment of the present invention provides redundant links for both network and I/O protocol data using two links and two host bus adapters. This contrasts with conventional systems, which typically use two redundant communication links and two host bus adapters for I/O protocol data, and two additional redundant links and two additional host bus adapters for network protocol data.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates one embodiment of the present invention;

FIG. 2 illustrates one embodiment of the present invention with redundant communication links;

FIG. 3 illustrates a conventional network driver architecture;

FIG. 4 illustrates a conventional I/O driver architecture; and

FIG. 5 illustrates one embodiment of the driver architecture of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

The present invention provides methods and systems for running network and storage protocols over the same communication link. Thus, by way of example, one embodiment of the present invention advantageously reduces the number of interface cards, interface circuits, and communication links, required to network clustered server systems together. The reduction in the number of interface cards, interface circuits, and communication links provides for reduced costs and smaller systems as compared with conventional systems.

FIG. 1 provides an overview of the hardware of an exemplary system incorporating one embodiment of the present invention. The illustrated system is a clustered server system. A clustered server system typically includes two or more independent severs that can access a common body of data storage and provide services to a common set of clients. However, the present invention is not limited to clustered server systems. For example, the present invention can be incorporated into other systems that utilize both network and storage protocols.

As illustrated in FIG. 1, the exemplary system 100 includes two storage subsystems 106, 108 and two host systems 102, 104, though the clustered server system 100 can have more or fewer host systems and storage subsystems. Each storage subsystem 106, 108 and host system 102, 104 may include one or more processors, such as, by way of example, a Pentium.RTM. II processor or an Alpha.RTM. processor. In one embodiment, each host system 102, 104 is running an operating system, such as Microsoft Windows.RTM. NT versions 4 or 5, or Unix. For the purposes of illustrating exemplary embodiments of the present invention, it will be assumed that the systems are running a version of Microsoft Windows.RTM. NT. Each storage subsystem 106, 108 and host system 102, 104 typically includes random access memory, as well as hard drives and optical drives, such as a CD-ROM drive or a DVD drive (not shown). In addition, each host system typically includes an interface card, such as a host bus adapter (not shown) from one of the QLogic ISP21xx or ISP22xx families. In one embodiment, the host adapter has a processor, such as CISC (complex instruction set computer) or a RISC (reduced instruction set computer) processor, and memory, such as RAM or EEPROM. In one embodiment, at least a portion of the memory is used to hold adapter code downloaded from the host computer system. In one embodiment, each host adapter is located on a separate circuit card. In another embodiment, more than one host adapter is located on a circuit card. Each host bus adapter card is typically connected to a local computer bus slot, such as a PCI bus slot.

In contrast to conventional systems, the system 100 illustrated in FIG. 1 runs both storage and network communications using the same host bus adapter interface circuit on a common communication link 110. Thus, in contrast to conventional systems, which use two interface cards and two corresponding communication links per host system, the embodiment illustrated in FIG. 1 only needs one interface card or circuit and one communication link per host system. Furthermore, as discussed below, in one embodiment, use of the present invention is transparent to the operating system. Additionally, in one embodiment of the present invention, Windows NT standard class drivers may be used without modification.

In one embodiment of the present invention, the cluster storage protocol is a SCSI protocol, such as SCSI II, while the network protocol is the Internet Protocol (IP). In one embodiment, the communication link 110 connecting the various cluster systems or "nodes" is a Fibre Channel Loop (FCL). Fibre Channel is a high-speed data transfer interface technology that advantageously maps common transport protocols, such as SCSI and IP. Thus, using Fibre Channel technology, it is possible to merge high-speed I/O, such as SCSI, and networking functionality in a single connectivity technology. However, alternative embodiments can use other bus technologies, such as a SCSI bus, to run both I/O and networking protocols on a common link. Thus, in one embodiment, the network and storage packets are transferred between computer systems using standard network and I/O protocols, such as the IP and SCSI protocols. This embodiment may be used when the HBA, such as one based on the QLogic ISP2200, supports both IP and SCSI protocols. In another embodiment, if the HBA supports the SCSI protocol, but not the IP protocol, the network packets are encapsulated in SCSI packets or commands. In the "encapsulation" embodiment, the HBA may support the SCSI target mode, as well as the more typical initiator mode, thereby allowing the HBA to receive SCSI packets encapsulating IP packets.

The Fibre Channel Loop 110 is connected to the host bus adapter of the host systems 102, 104 and the storage subsystems 106, 108. Thus, storage and network packets can be routed between the host systems 102, 104 and the storage subsystems 106, 108. The data transferred using the IP protocol may include "heartbeat" related information. "Heartbeats" are transferred across the clustered nodes to ensure that everything is properly synchronized, and so that each node can ensure that the other nodes are functioning.

The Fibre Channel Loop 110 is connected to the host bus adapters residing in each node 102-108. Thus, data is transferred from and to the Fibre Channel Loop 110 via respective host bus adapters. The operating systems running each node communicate with their respective host bus adapters using device drivers. In one embodiment, the device driver is responsible for accessing the host bus adapter's internal registers, starting and completing I/O operations, performing error processing, and servicing interrupts generated by the device.

In one embodiment, a fail-over path is provided to ensure high reliability and to avoid a shutdown of the clustered sever system in the event of a communication link failure. By way of example, FIG. 2 illustrates a clustered server system 200 which has two Fibre Channel Loops 210, 212. In normal operation, Loop 1210 handles the I/O traffic, while Loop 2212 handles the network traffic. Thus, both communication loops 210, 212 are efficiently utilized in normal operation, with the bandwidth of Loop 1 allocated to I/O traffic, and the bandwidth of Loop 2 allocated to network traffic. However, in the event of a Fibre Channel Loop failure, such as the failure of Loop 1210, one embodiment of the present invention detects the failure. Upon detecting the failure, the remaining operating Fibre Channel Loop, such as Loop 2212, is used as a fail-over path to carry both network traffic and I/O traffic. Thus, in contrast to conventional systems, which would utilize four communication links to provide redundancy for both network and I/O traffic, the illustrated embodiment 200 of the present invention utilizes two communication links 210, 212. Furthermore, in contrast to conventional systems, which would utilize at least two HBAs and two NICs to provide such redundancy, the illustrated embodiment 200 uses two HBAs per host system 202, 204 to provide the desired fail-over path.

In one embodiment, the fail-over technique operates as follows. Typically, a Windows NT Cluster Administrator is used to manage and configure network paths and storage devices. In conventional systems, if multiple network paths, with corresponding multiple communication links, have been configured, one path is designated as the primary network path. If the Cluster Administrator determines, via a loss of signal indication or the like, that it is unable to communicate over the primary path, the Windows NT Cluster Administrator can switch to a secondary network path and a corresponding secondary communication link. In contrast to conventional systems, one embodiment of the present invention permits the I/O bus, such as the SCSI bus or the Fibre Channel loop, to be configured as a secondary or primary network path in addition to acting as a I/O or storage path. Thus, upon detection of a network path failure, the Windows NT Cluster Administrator can fail-over to the I/O bus.

To better illustrate the software architecture of one embodiment of the present invention, a description of conventional NIC and SCSI drivers will now be discussed. FIG. 3 illustrates a conventional Windows NT NDIS (Network Device Interface Specification) driver architecture 300. An NDIS interface 304 forms a wrapper about an NDIS NIC miniport driver 302, thus providing an interface between the NDIS NIC miniport driver 302, one or more NDIS intermediate drivers 306, LAN protocol drivers 308, and native-media-aware protocol drivers 310. Additionally, the NDIS interface 304 provides common, pre-defined functions to the interface between the NDIS NIC miniport driver 302 and the NIC 312. The NDIS NIC miniport 302 uses these pre-defined functions, which in turn call other components, to manage the NIC 312 hardware.

FIG. 4 illustrates a conventional Windows SCSI driver architecture 400. At the upper level is a SCSI class driver 404. The SCSI class driver 404 translates I/O requests received from application programs and the like via I/O system services. The translated I/O requests are in the form of packets with system defined SCSI request blocks containing SCSI command descriptor blocks. The translated I/O request packets are then sent to the next lower driver, for example, a SCSI port driver 406. The SCSI port driver 406 translates the SCSI request blocks from the class driver 404 and passes the SCSI request blocks and the command descriptor blocks to the SCSI miniport driver 402. The SCSI miniport driver 402 is dynamically linked with the SCSI port driver 406 and provides hardware-specific support for a particular SCSI host bus adapter 408. The SCSI driver 400 may include other optional driver layers as well, such as encryption drivers or filter drivers. Thus, as illustrated in FIGS. 3 and 4, in conventional systems the NIC driver 300 and SCSI driver 400 independently interface to their respective interface cards 312, 408.

The software architecture and operation of the embodiment 100 illustrated in FIG. 1 will now be discussed in detail. FIG. 5 illustrates one embodiment of a driver architecture 500 of the present invention. For example, the illustrated embodiment 500 could be used in one of the host systems illustrated in FIGS. 1 or 2. The upper layers of the NIC driver and the SCSI driver are similar to those found in the conventional driver architectures 300, 400 illustrated in FIGS. 3 and 4. However, in contrast to the NIC miniport 302 illustrated in FIG. 3, which interfaces to its own network interface card 312, the NDIS miniport 502 illustrated in FIG. 5 redirects communication to an enhanced SCSI miniport driver 508 via an interface 504. The enhanced SCSI miniport driver 508, in turn, routes network traffic from the NDIS miniport 502 onto the common Fibre Channel Loop 110 illustrated in FIG. 1. Therefore, in one embodiment, the network miniport driver 502 and the SCSI miniport driver 508 allow host-to-host network traffic to be overlaid onto a Fibre Channel bus, or, in an alternative embodiment, onto a SCSI bus. This permits the host-to-host traffic to operate in parallel with normal disk storage related traffic on the same bus. Thus, the enhanced driver 502, 508, in combination with the HBA, emulate the network interface card (NIC).

Thus, in one embodiment, the NDIS miniport driver 502 emulates an Ethernet connection between multiple systems sharing a common Fibre Channel or SCSI bus. This Ethernet emulation is overlaid by the enhanced SCSI miniport onto the same data channels used to access disk devices shared between the clustered systems. As previously described with reference to FIG. 1, this approach particularly benefits clustered systems that are running short on card slots, such as PCI slots. In addition, as previously described with reference to FIG. 2, in one embodiment, the present invention can provide additional fail-over paths to backup NIC adapters used for local host-to-host communications. In one embodiment, the network driver is adapter-type independent and supports multiple Ethernet emulations through multiple enhanced miniport drivers simultaneously.

In one embodiment, several features are incorporated into the driver architecture 500 to enable the enhanced SCSI miniport 508 to handle both the NDIS miniport communications and the standard SCSI miniport communications. For example, the SCSI miniport 508 is enhanced with additional entry points to accommodate the connection to the NDIS miniport 502. These additional entry points are termed "backdoor" entry points. Similarly, the NDIS miniport 502 addresses communications to the additional SCSI miniport entry points, rather than to the network interface card.

In addition, because the SCSI miniport 508 is not aware of physical addresses of the NIC command data stored in the host system memory, the SCSI miniport 508 cannot build the scatter/gather list for the network related command data and data buffers. As is understood by one of ordinary skill in the art, scatter/gather lists are used to track logically contiguous data that have non-contiguous physical addresses. Thus, in one embodiment, one or more upper layer drivers generate the scatter/gather list for the network data. For example, a SCSI filter driver, the SCSI port driver 406, and/or the NDIS driver 502 may complete the scatter/gather list. The scatter/gather list is then passed to the SCSI miniport 508 via a SCSI Request Block (SRB) extension, as discussed below.

In addition, the SRB extension is used to pass pointer and context parameter information for a Command Callback routine. The Command Callback routine is called by the miniport driver 508 as each command is completed. The Command Callback routine places a pointer to the completed command into a queue of completed command pointers, schedules a DPC (Deferred Procedure Call) service routine, and then returns.

An exemplary backdoor (BD) SRB extension shown below contains link pointers, a completion routine pointer and context, a scatter/gather count, and a list of scatter/gather elements. The exemplary SRB extension may be used when network packets are to be encapsulated in SCSI packets before the packets are transferred over the communications link.

    typedef struct _BD_SRB_EXTENSION
    {
        SRB_EXTENSION              stdExt;
        PSCSI_REQUEST_BLOCK        nextSrb;
        PSCSI_REQUEST_BLOCK        previousSrb;
        PCOMPLETION_CALLBACK       completionRoutine;
        PVOID                      completionContext;
        PVOID                      driverContext;
        LONG                       sgEntryCnt;
        BD_SG_ELEMENT              sgList[MAX_SG_ENTRIES];
    } BD_SRB_EXTENSION, *PBD_SRB_EXTENSION;

    typedef struct   _BD_INQUIRY_DATA
    {
        ULONG                  options;                         // Export
     miniport supported options
        PSTARTIO_BACKDOOR      startIoRoutine;                  // Export
     backdoor StartIo entry
        PVOID                  startIoContext;                  // Export
     miniport pDevExt pointer
        PRESET_BACKDOOR        resetRoutine;                    // Export
     backdoor Reset entry
        PVOID                  resetContext;                    // Export
     miniport pDevExt pointer
        USHORT                 numberBuses;                     // Export
     adapter bus count
        USHORT                 idsPerBus;                       // Export
     number of IDs per bus
        USHORT                 adapterBusId[MAX_BUS_COUNT];     // Export
     adapter bus IDs
        USHORT                 maxDataSegments;                 // Export max
     data segment count
    } BD_INQUIRY_DATA, *PBD_INQUIRY_DATA;

    // IOCTL_BD_INQUIRY options
    #define BDI_FILTER_SUPPORT     0x00000001  // Filter driver supported
    #define BDI_LAN_SUPPORT        0x00000002  // LAN driver supported
    #define BDI_PORT_DATABASE      0x00000004  // Port database supported

    typedef struct   _BD_ENABLE_DATA
    {
        ULONG                      options;             // Import driver
     supported options
        PNOTIFY_CALLBACK           notifyRoutine;       // Import notification
     routine
        PVOID                      notifyContext;       // Import notification
     parameter
        PPORT_DATA_CALLBACK        portDataUpdateRoutine; // Import port
     database routine
        PVOID                      portDataUpdateContext; // Import port
     database parameter
        PGET_BUFFER_CALLBACK       getBufferRoutine;    // Import get buffer
     routine
        PVOID                      getBufferContext;    // Import get buffer
     parameter
        USHORT                     receiveBufferCount;  // Import buffer count
    } BD_ENABLE_DATA, *PBD_ENABLE_DATA;

    // IOCTL_BD_ENABLE options
    #define BDE_NOTIFY_ROUTINE         0x00000001  // Notify routine imported
    #define BDE_DATABASE_ROUTINE       0x00000002  // Port database routine
     imported
    #define BDE_BUFFER_ROUTINE         0x00000004  // Get buffer routine
     imported

    typedef struct _BD_INQUIRY_DATA
    {
        ULONG                  Options;               // Export miniport
     supported options
        PSTARTIO_BACKDOOR      StartIoRoutine;        // Export backdoor
     StartIo entry
        PVOID                  StartIoContext;        // Export miniport DevExt
     pointer
        PRESET_BACKDOOR        ResetRoutine;          // Export backdoor Reset
     entry
        PVOID                  ResetContext;          // Export miniport DevExt
     pointer
        USHORT                 ReceiveBufferQueueSize; // Export receive buffer
     queue size
        PVOID                  ReceiveBufferQueue;    // Export receive buffer
     queue pointer
        PBUFFERS_BACKDOOR      AddBuffersRoutine;     // Export backdoor Add
     Buffers entry
        PVOID                  AddBuffersContext;     // Export miniport DevExt
     pointer
        UCHAR                  acNodeName[8];         // Export adapter node
     name
    } BD_INQUIRY_DATA, *PBD_INQUIRY_DATA;

    typedef struct _BD_ENABLE_DATA
    {
        ULONG                            Options;                   // Import
     driver supported options
        USHORT                           ReceiveBufferCount;        // Import
     buffer count
        ULONG                            ReceiveBufferSize;         // Import
     buffer size
        ULONG                            MaximumTransferSize;       // Import
     maximum transfer size
        PRETURN_PACKETS_CALLBACK         ReturnReceivePacketsRoutine; // Import
     return packets routine
        PVOID                            ReturnReceivePacketsContext; // Import
     return packets parameter
        PVOID                            ReceiveBufferCBs;          // Import
     receive buffer control blocks
        ULONG                            IpAddress;                 // Import
     IP address
    } BD_ENABLE_DATA, *PBD_ENABLE_DATA;

• Inventors
  Shah, Shishir C.; Thompson, Thomas W.; Wagner, David K.;
• Assignee
  QLogic Corporation (Aliso Viejo, CA)
• Published
  Oct-22-2002
• Current US Classes:
  709/250
  719/321
  719/326
• Application #
  280506
• International Classes
  G06F 015/16; G06F 009/00
• Field of Search
  709/250 709/251 709/253 709/230 709/321 709/326 709/327 710/10 710/11 710/105
• Examiner
  Maung; Zarni
• Agent
  Knobbe, Martens Olson & Bear, LLP
• US Patent References:
  5491812
  5613096
  5802391
  5905509
  5991813
  6044415
  6118776
  6148346
  6173346
  6202147
  6209023
  6233619
  6233624
  6314525
  6317842
  6330626