Scribd Logo
Upload

Welcome to Scribd!

  • 56 views

    Virtio 1.0: Paravirtualized I/O For KVM and Beyond

    Uploaded by

    Mehmet Demir
    This document introduces the VIRTIO 1.0 standard for paravirtualized I/O. It describes how VIRTIO uses virtqueues to transfer data between devices and drivers, with features negotiated through feature bits. Device configuration is stored in a configuration space. VIRTIO provides a standardized interface that enables compatibility across hypervisors and operating systems for virtual devices like network and SCSI controllers.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Virtio 1.0: Paravirtualized I/O For KVM and Beyond

    Uploaded by

    Mehmet Demir
    56 views15 pages
    This document introduces the VIRTIO 1.0 standard for paravirtualized I/O. It describes how VIRTIO uses virtqueues to transfer data between devices and drivers, with features negotiated through feature bits. Device configuration is stored in a configuration space. VIRTIO provides a standardized interface that enables compatibility across hypervisors and operating systems for virtual devices like network and SCSI controllers.

    Original Title

    virtio-devconf-2014

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document introduces the VIRTIO 1.0 standard for paravirtualized I/O. It describes how VIRTIO uses virtqueues to transfer data between devices and drivers, with features negotiated through feature bits. Device configuration is stored in a configuration space. VIRTIO provides a standardized interface that enables compatibility across hypervisors and operating systems for virtual devices like network and SCSI controllers.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    56 views15 pages

    Virtio 1.0: Paravirtualized I/O For KVM and Beyond

    Uploaded by

    Mehmet Demir
    This document introduces the VIRTIO 1.0 standard for paravirtualized I/O. It describes how VIRTIO uses virtqueues to transfer data between devices and drivers, with features negotiated through feature bits. Device configuration is stored in a configuration space. VIRTIO provides a standardized interface that enables compatibility across hypervisors and operating systems for virtual devices like network and SCSI controllers.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 15

    VIRTIO 1.

    0
    Paravirtualized I/O for KVM and beyond

    Stefan Hajnoczi <[email protected]>


    th
    8 February 2014
    1 INTERNAL ONLY | PRESENTER NAME
    What we're going to cover

    How VIRTIO 1.0 works

    You want to:


    ● Understand paravirtualized I/O
    ● Design custom devices
    ● Get familiar before tackling the spec/code

    Not covering every VIRTIO 1.0 spec change


    See Rusty's linuxconf.au talk: http://goo.gl/wd9Xfp

    2 INTERNAL ONLY | PRESENTER NAME


    What is virtio?

    “Straightforward, efficient, standard and extensible


    mechanism for virtual devices”
    ● Network card, SCSI controller, etc

    Designed for situations where accessing device is


    expensive, device accessing memory is cheap
    ● Real hardware is the opposite!

    Like USB class-compliant devices, a standard driver


    means compatibility across OSes and hypervisors

    3 INTERNAL ONLY | PRESENTER NAME


    What's happening in virtio land?
    Community (led by Rusty Russell)
    VIRTIO Independent, informal document
    0.9.5
    QEMU, lguest, Linux, FreeBSD, VirtualBox

    OASIS Committee (chaired by Rusty Russell)


    VIRTIO Formal process, formal document
    1.0
    QEMU, lguest, Linux, FreeBSD, VirtualBox,
    Xen, etc

    4 INTERNAL ONLY | PRESENTER NAME


    Virtio architecture

    Three layers defined by virtio:

    net scsi ... Device types

    Feature Config Core device model


    Virtqueue
    bits space

    PCI MMIO CCW Transports

    5 INTERNAL ONLY | PRESENTER NAME


    Device lifecycle and device status field

    Find matching driver


    Reset Ack Driver

    Select
    ... Reset features

    Check device
    Driver accepted features Features
    Failed
    OK OK

    6 INTERNAL ONLY | PRESENTER NAME


    Feature bit negotiation

    The feature bit field enables extensibility


    ● New features can be added to spec in future

    Steps for negotiation:


    1.Device shows all supported feature bits
    2.Driver selects subset of features it supports
    3.Driver sets FEATURES_OK in status field
    4.Device leaves FEATURES_OK set if ok

    7 INTERNAL ONLY | PRESENTER NAME


    Configuration space

    Contains device parameters


    ● Read/write
    ● 32-bit atomic access (careful with bigger accesses)
    ● Version counter for consistent >32-bit reads
    ● No consistent >32-bit writes!
    ● Device notifies driver via interrupt on update

    Consider using a config virtqueue for complex device


    configuration or error handling.

    8 INTERNAL ONLY | PRESENTER NAME


    Virtqueues and the device model

    Devices have virtqueues to transfer data buffers

    Driver adds buffer, device processes and returns it

    Buffers may be:


    ● Scatter-gather lists (multiple memory regions)
    ● Handled out-of-order by device, if appropriate

    Interrupt notifies driver of buffer completion

    9 INTERNAL ONLY | PRESENTER NAME


    Virtqueue programming interface example

    void
    virtqueue_add_sgs(struct virtqueue *vq,
                  struct scatterlist sg[],
                  unsigned int out_sgs,
                  unsigned int in_sgs,
                  void *data, gfp_t gfp);
    void *virtqueue_get_buf(
                  struct virtqueue *vq,
                  unsigned int *len);

    10 INTERNAL ONLY | PRESENTER NAME


    Virtqueue memory layout (aka vring)
    Driver allocates vring and configures device with its address:

    Descriptor table
    Index Addr Len Flags Next
    0
    1
    ...

    Available ring Used ring


    ... ...

    Device Driver
    Driver Device
    11 INTERNAL ONLY | PRESENTER NAME
    Adding buffers to the vring
    Driver puts scatter-gather list into descriptor table, adds head
    index to available ring, and then kicks device.

    Descriptor table
    Index Addr Len Flags Next
    0 0x8000000000000000 4096 NEXT 1
    1 0x8000000000040000 128 WRITE 0
    ...

    Available ring Used ring


    0 ... ...

    Device Driver Driver Device


    12 INTERNAL ONLY | PRESENTER NAME
    Returning completed buffers to the vring
    Device adds head index to used ring and then notifies driver.

    Descriptor table
    Index Addr Len Flags Next
    0 0x8000000000000000 4096 NEXT 1
    1 0x8000000000040000 128 WRITE 0
    ...

    Available ring Used ring


    0 ... 0 ...

    Device Driver Driver Device


    13 INTERNAL ONLY | PRESENTER NAME
    Example device: virtio-scsi
    Virtqueues: Configuration space:
    0.Control
    struct virtio_scsi_config {
    1.Events le32 num_queues;
    2.Requests le32 seg_max;
    le32 max_sectors;
    3.Requests le32 cmd_per_lun;
    4.... (multiqueue) le32 event_info_size;
    le32 sense_size;
    struct virtio_scsi_req_cmd { le32 cdb_size;
    u8 lun[8]; le64 id; le16 max_channel;
    ... le16 max_target;
    char cdb[cdb_size]; le32 max_lun;
    char dataout[]; };
    ...
    };
    14 INTERNAL ONLY | PRESENTER NAME
    More information

    VIRTIO 1.0 draft: http://goo.gl/BQ1Kbu


    Mailing list: [email protected]

    QEMU virtio code: hw/virtio/


    Linux virtio driver code: drivers/virtio/
    Linux vhost device code: drivers/vhost/

    My blog: http://blog.vmsplice.net/
    My email: [email protected]

    16 INTERNAL ONLY | PRESENTER NAME

    You might also like