The purpose of virtio and this specification is that virtual environments and guests should have a straightforward, efficient, standard and extensible mechanism for virtual devices, rather than boutique per-environment or per-OS mechanisms.
Technical Committee members should send comments on this specification to the Technical Committee’s email list. Others should send comments to the Technical Committee by using the “Send A Comment” button on the Technical Committee’s web page at https://www.oasis-open.org/committees/virtio/.
For information on whether any patents have been disclosed that may be essential to implementing this specification, and any offers of patent licensing terms, please refer to the Intellectual Property Rights section of the Technical Committee web page (https://www.oasis-open.org/committees/virtio/ipr.php).
Virtual I/O Device (VIRTIO) Version 1.0. Edited by Rusty Russell, Michael S. Tsirkin, Cornelia Huck, and Pawel Moll. 03 March 2016. OASIS Committee Specification 04. http://docs.oasis-open.org/virtio/virtio/v1.0/cs04/virtio-v1.0-cs04.html. Latest version: http://docs.oasis-open.org/virtio/virtio/v1.0/virtio-v1.0.html.
Copyright © OASIS Open 2015. All Rights Reserved.
All capitalized terms in the following text have the meanings assigned to them in the OASIS Intellectual Property Rights Policy (the "OASIS IPR Policy"). The full Policy may be found at the OASIS website.
This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published, and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this section are included on all such copies and derivative works. However, this document itself may not be modified in any way, including by removing the copyright notice or references to OASIS, except as needed for the purpose of developing any document or deliverable produced by an OASIS Technical Committee (in which case the rules applicable to copyrights, as set forth in the OASIS IPR Policy, must be followed) or as required to translate it into languages other than English.
The limited permissions granted above are perpetual and will not be revoked by OASIS or its successors or assigns.
This document and the information contained herein is provided on an "AS IS" basis and OASIS DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY OWNERSHIP RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
OASIS requests that any OASIS Party or any other party that believes it has patent claims that would necessarily be infringed by implementations of this OASIS Committee Specification or OASIS Standard, to notify OASIS TC Administrator and provide an indication of its willingness to grant patent licenses to such patent claims in a manner consistent with the IPR Mode of the OASIS Technical Committee that produced this specification.
OASIS invites any party to contact the OASIS TC Administrator if it is aware of a claim of ownership of any patent claims that would necessarily be infringed by implementations of this specification by a patent holder that is not willing to provide a license to such patent claims in a manner consistent with the IPR Mode of the OASIS Technical Committee that produced this specification. OASIS may include such claims on its website, but disclaims any obligation to do so.
OASIS takes no position regarding the validity or scope of any intellectual property or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; neither does it represent that it has made any effort to identify any such rights. Information on OASIS’ procedures with respect to rights in any document or deliverable produced by an OASIS Technical Committee can be found on the OASIS website. Copies of claims of rights made available for publication and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this OASIS Committee Specification or OASIS Standard, can be obtained from the OASIS TC Administrator. OASIS makes no representation that any information or list of intellectual property rights will at any time be complete, or that any claims in such list are, in fact, Essential Claims.
The name "OASIS" is a trademark of OASIS, the owner and developer of this
specification, and should be used only to refer to the organization and its official
outputs. OASIS welcomes reference to, and implementation and use of, specifications,
while reserving the right to enforce its marks against misleading uses. Please see
https://www.oasis-open.org/policies-guidelines/trademark for above guidance.
The purpose of virtio and this specification is that virtual environments and guests should have a straightforward, efficient, standard and extensible mechanism for virtual devices, rather than boutique per-environment or per-OS mechanisms.
Bradner S., “Key words for use in RFCs to Indicate Requirement Levels”, BCP 14,
RFC 2119, March 1997.
z/Architecture Principles of Operation, IBM Publication SA22-7832,
|[S390 Common I/O]||
ESA/390 Common I/O-Device and Self-Description, IBM Publication SA22-7204,
Conventional PCI Specifications,
PCI Express Specifications
IEEE Standard for Local and Metropolitan Area Networks: Overview and
SCSI Architectural Model,
SCSI Multimedia Commands,
|[Virtio PCI Draft]||
Virtio PCI Draft Specification
The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in [RFC2119].
Earlier drafts of this specification (i.e. revisions before 1.0, see e.g. [Virtio PCI Draft]) defined a similar, but different interface between the driver and the device. Since these are widely deployed, this specification accommodates OPTIONAL features to simplify transition from these earlier draft interfaces.
Specifically devices and drivers MAY support:
Legacy devices and legacy drivers are not compliant with this specification.
To simplify transition from these earlier draft interfaces, a device MAY implement:
Similarly, a driver MAY implement:
Devices or drivers with no legacy compatibility are referred to as non-transitional devices and drivers, respectively.
For devices and drivers already implementing the legacy interface, some changes will have to be made to support this specification.
In this case, it might be beneficial for the reader to focus on sections tagged "Legacy Interface" in the section title. These highlight the changes made since the earlier drafts.
Many device and driver in-memory structure layouts are documented using the C struct syntax. All structures are assumed to be without additional padding. To stress this, cases where common C compilers are known to insert extra padding within structures are tagged using the GNU C __attribute__((packed)) syntax.
For the integer data types used in the structure definitions, the following conventions are used:
During device initialization by a driver, the driver follows the sequence of steps specified in 3.1.
The device status field provides a simple low-level indication of the completed steps of this sequence. It’s most useful to imagine it hooked up to traffic lights on the console indicating the status of each device. The following bits are defined (listed below in the order in which they would be typically set):
The driver MUST update device status, setting bits to indicate the completed steps of the driver initialization sequence specified in 3.1. The driver MUST NOT clear a device status bit. If the driver sets the FAILED bit, the driver MUST later reset the device before attempting to re-initialize.
The driver SHOULD NOT rely on completion of operations of a device if DEVICE_NEEDS_RESET is set. Note: For example, the driver can’t assume requests in flight will be completed if DEVICE_NEEDS_RESET is set, nor can it assume that they have not been completed. A good implementation will try to recover by issuing a reset.
The device MUST initialize device status to 0 upon reset.
The device MUST NOT consume buffers or notify the driver before DRIVER_OK.
The device SHOULD set DEVICE_NEEDS_RESET when it enters an error state that a reset is needed. If DRIVER_OK is set, after it sets DEVICE_NEEDS_RESET, the device MUST send a device configuration change notification to the driver.
Each virtio device offers all the features it understands. During device initialization, the driver reads this and tells the device the subset that it accepts. The only way to renegotiate is to reset the device.
This allows for forwards and backwards compatibility: if the device is enhanced with a new feature bit, older drivers will not write that feature bit back to the device. Similarly, if a driver is enhanced with a feature that the device doesn’t support, it see the new feature is not offered.
Feature bits are allocated as follows:
In particular, new fields in the device configuration space are indicated by offering a new feature bit.
The driver MUST NOT accept a feature which the device did not offer, and MUST NOT accept a feature which requires another feature which was not accepted.
The driver SHOULD go into backwards compatibility mode if the device does not offer a feature it understands, otherwise MUST set the FAILED device status bit and cease initialization.
The device MUST NOT offer a feature which requires another feature which was not offered. The device SHOULD accept any valid subset of features the driver accepts, otherwise it MUST fail to set the FEATURES_OK device status bit when the driver writes it.
Transitional Drivers MUST detect Legacy Devices by detecting that the feature bit VIRTIO_F_VERSION_1 is not offered. Transitional devices MUST detect Legacy drivers by detecting that VIRTIO_F_VERSION_1 has not been acknowledged by the driver.
In this case device is used through the legacy interface.
Legacy interface support is OPTIONAL. Thus, both transitional and non-transitional devices and drivers are compliant with this specification.
Requirements pertaining to transitional devices and drivers is contained in sections named ’Legacy Interface’ like this one.
When device is used through the legacy interface, transitional devices and transitional drivers MUST operate according to the requirements documented within these legacy interface sections. Specification text within these sections generally does not apply to non-transitional devices.
Device configuration space is generally used for rarely-changing or initialization-time parameters. Where configuration fields are optional, their existence is indicated by feature bits: Future versions of this specification will likely extend the device configuration space by adding extra fields at the tail. Note: The device configuration space uses the little-endian format for multi-byte fields.
Each transport also provides a generation count for the device configuration space, which will change whenever there is a possibility that two accesses to the device configuration space can see different versions of that space.
Drivers MUST NOT assume reads from fields greater than 32 bits wide are atomic, nor are reads from multiple fields: drivers SHOULD read device configuration space fields like so:
For optional configuration space fields, the driver MUST check that the corresponding feature is offered before accessing that part of the configuration space. Note: See section 3.1 for details on feature negotiation.
Drivers MUST NOT limit structure size and device configuration space size. Instead, drivers SHOULD only check that device configuration space is large enough to contain the fields necessary for device operation. Note: For example, if the specification states that device configuration space ’includes a single 8-bit field’ drivers should understand this to mean that the device configuration space might also include an arbitrary amount of tail padding, and accept any device configuration space size equal to or greater than the specified 8-bit size.
The device MUST allow reading of any device-specific configuration field before FEATURES_OK is set by the driver. This includes fields which are conditional on feature bits, as long as those feature bits are offered by the device.
Note that for legacy interfaces, device configuration space is generally the guest’s native endian, rather than PCI’s little-endian. The correct endian-ness is documented for each device.
Legacy devices did not have a configuration generation field, thus are susceptible to race conditions if configuration is updated. This affects the block capacity (see 5.2.4) and network mac (see 5.1.4) fields; when using the legacy interface, drivers SHOULD read these fields multiple times until two reads generate a consistent result.
The mechanism for bulk data transport on virtio devices is pretentiously called a virtqueue. Each device can have zero or more virtqueues3. Each queue has a 16-bit queue size parameter, which sets the number of entries and implies the total size of the queue.
Each virtqueue consists of three parts:
where each part is physically-contiguous in guest memory, and has different alignment requirements.
The memory aligment and size requirements, in bytes, of each part of the virtqueue are summarized in the following table:
|Descriptor Table||16||16∗(Queue Size)|
|Available Ring||2||6 + 2∗(Queue Size)|
|Used Ring||4||6 + 8∗(Queue Size)|
The Alignment column gives the minimum alignment for each part of the virtqueue.
The Size column gives the total number of bytes for each part of the virtqueue.
Queue Size corresponds to the maximum number of buffers in the virtqueue4. Queue Size value is always a power of 2. The maximum Queue Size value is 32768. This value is specified in a bus-specific way.
When the driver wants to send a buffer to the device, it fills in a slot in the descriptor table (or chains several together), and writes the descriptor index into the available ring. It then notifies the device. When the device has finished a buffer, it writes the descriptor index into the used ring, and sends an interrupt.
The driver MUST ensure that the physical address of the first byte of each virtqueue part is a multiple of the specified alignment value in the above table.
For Legacy Interfaces, several additional restrictions are placed on the virtqueue layout:
Each virtqueue occupies two or more physically-contiguous pages (usually defined as 4096 bytes, but depending on the transport; henceforth referred to as Queue Align) and consists of three parts:
|Descriptor Table||Available Ring (…padding…)||Used Ring|
The bus-specific Queue Size field controls the total number of bytes for the virtqueue. When using the legacy interface, the transitional driver MUST retrieve the Queue Size field from the device and MUST allocate the total number of bytes for the virtqueue according to the following formula (Queue Align given in qalign and Queue Size given in qsz):
This wastes some space with padding. When using the legacy interface, both transitional devices and drivers MUST use the following virtqueue layout structure to locate elements of the virtqueue:
Note that when using the legacy interface, transitional devices and drivers MUST use the native endian of the guest as the endian of fields and in the virtqueue. This is opposed to little-endian for non-legacy interface as specified by this standard. It is assumed that the host is already aware of the guest endian.
The framing of messages with descriptors is independent of the contents of the buffers. For example, a network transmit buffer consists of a 12 byte header followed by the network packet. This could be most simply placed in the descriptor table as a 12 byte output descriptor followed by a 1514 byte output descriptor, but it could also consist of a single 1526 byte output descriptor in the case where the header and packet are adjacent, or even three or more descriptors (possibly with loss of efficiency in that case).
Note that, some device implementations have large-but-reasonable restrictions on total descriptor size (such as based on IOV_MAX in the host OS). This has not been a problem in practice: little sympathy will be given to drivers which create unreasonably-sized descriptors such as by dividing a network packet into 1500 single-byte descriptors!
The device MUST NOT make assumptions about the particular arrangement of descriptors. The device MAY have a reasonable limit of descriptors it will allow in a chain.
The driver MUST place any device-writable descriptor elements after any device-readable descriptor elements.
The driver SHOULD NOT use an excessive number of descriptors to describe a buffer.
Regrettably, initial driver implementations used simple layouts, and devices came to rely on it, despite this specification wording. In addition, the specification for virtio_blk SCSI commands required intuiting field lengths from frame boundaries (see 184.108.40.206 Legacy Interface: Device Operation)
Thus when using the legacy interface, the VIRTIO_F_ANY_LAYOUT feature indicates to both the device and the driver that no assumptions were made about framing. Requirements for transitional drivers when this is not negotiated are included in each device section.
The descriptor table refers to the buffers the driver is using for the device. addr is a physical address, and the buffers can be chained via next. Each descriptor describes a buffer which is read-only for the device (“device-readable”) or write-only for the device (“device-writable”), but a chain of descriptors can contain both device-readable and device-writable buffers.
The actual contents of the memory offered to the device depends on the device type. Most common is to begin the data with a header (containing little-endian fields) for the device to read, and postfix it with a status tailer for the device to write.
The number of descriptors in the table is defined by the queue size for this virtqueue: this is the maximum possible descriptor chain length. Note: The legacy [Virtio PCI Draft] referred to this structure as vring_desc, and the constants as VRING_DESC_F_NEXT, etc, but the layout and values were identical.
A device MUST NOT write to a device-readable buffer, and a device SHOULD NOT read a device-writable buffer (it MAY do so for debugging or diagnostic purposes).
Drivers MUST NOT add a descriptor chain over than 232 bytes long in total; this implies that loops in the descriptor chain are forbidden!
Some devices benefit by concurrently dispatching a large number of large requests. The VIRTIO_F_INDIRECT_DESC feature allows this (see A virtio_queue.h). To increase ring capacity the driver can store a table of indirect descriptors anywhere in memory, and insert a descriptor in main virtqueue (with flags&VIRTQ_DESC_F_INDIRECT on) that refers to memory buffer containing this indirect descriptor table; addr and len refer to the indirect table address and length in bytes, respectively.
The indirect table layout structure looks like this (len is the length of the descriptor that refers to this table, which is a variable, so this code won’t compile):
The first indirect descriptor is located at start of the indirect descriptor table (index 0), additional indirect descriptors are chained by next. An indirect descriptor without a valid next (with flags&VIRTQ_DESC_F_NEXT off) signals the end of the descriptor. A single indirect descriptor table can include both device-readable and device-writable descriptors.
A driver MUST NOT create a descriptor chain longer than the Queue Size of the device.
A driver MUST NOT set both VIRTQ_DESC_F_INDIRECT and VIRTQ_DESC_F_NEXT in flags.
The device MUST handle the case of zero or more normal chained descriptors followed by a single descriptor with flags&VIRTQ_DESC_F_INDIRECT. Note: While unusual (most implementations either create a chain solely using non-indirect descriptors, or use a single indirect element), such a layout is valid.
The driver uses the available ring to offer buffers to the device: each ring entry refers to the head of a descriptor chain. It is only written by the driver and read by the device.
idx field indicates where the driver would put the next descriptor entry in the ring (modulo the queue size). This starts at 0, and increases. Note: The legacy [Virtio PCI Draft] referred to this structure as vring_avail, and the constant as VRING_AVAIL_F_NO_INTERRUPT, but the layout and value were identical.
If the VIRTIO_F_EVENT_IDX feature bit is not negotiated, the flags field in the available ring offers a crude mechanism for the driver to inform the device that it doesn’t want interrupts when buffers are used. Otherwise used_event is a more performant alternative where the driver specifies how far the device can progress before interrupting.
Neither of these interrupt suppression methods are reliable, as they are not synchronized with the device, but they serve as useful optimizations.
If the VIRTIO_F_EVENT_IDX feature bit is not negotiated:
Otherwise, if the VIRTIO_F_EVENT_IDX feature bit is negotiated:
The driver MUST handle spurious interrupts from the device.
If the VIRTIO_F_EVENT_IDX feature bit is not negotiated:
Otherwise, if the VIRTIO_F_EVENT_IDX feature bit is negotiated:
The used ring is where the device returns buffers once it is done with them: it is only written to by the device, and read by the driver.
Each entry in the ring is a pair: id indicates the head entry of the descriptor chain describing the buffer (this matches an entry placed in the available ring by the guest earlier), and len the total of bytes written into the buffer. Note: len is particularly useful for drivers using untrusted buffers: if a driver does not know exactly how much has been written by the device, the driver would have to zero the buffer in advance to ensure no data leakage occurs.
For example, a network driver may hand a received buffer directly to an unprivileged userspace application. If the network device has not overwritten the bytes which were in that buffer, this could leak the contents of freed memory from other processes to the application.
idx field indicates where the driver would put the next descriptor entry in the ring (modulo the queue size). This starts at 0, and increases. Note: The legacy [Virtio PCI Draft] referred to these structures as vring_used and vring_used_elem, and the constant as VRING_USED_F_NO_NOTIFY, but the layout and value were identical.
Historically, many drivers ignored the len value, as a result, many devices set len incorrectly. Thus, when using the legacy interface, it is generally a good idea to ignore the len value in used ring entries if possible. Specific known issues are listed per device type.
The device MUST set len prior to updating the used idx.
The device MUST write at least len bytes to descriptor, beginning at the first device-writable buffer, prior to updating the used idx.
The device MAY write more than len bytes to descriptor. Note: There are potential error cases where a device might not know what parts of the buffers have been written. This is why len is permitted to be an underestimate: that’s preferable to the driver believing that uninitialized memory has been overwritten when it has not.
The driver MUST NOT make assumptions about data in device-writable buffers beyond the first len bytes, and SHOULD ignore this data.
The device can suppress notifications in a manner analogous to the way drivers can suppress interrupts as detailed in section 2.4.7. The device manipulates flags or avail_event in the used ring the same way the driver manipulates flags or used_event in the available ring.
The driver MUST initialize flags in the used ring to 0 when allocating the used ring.
If the VIRTIO_F_EVENT_IDX feature bit is not negotiated:
Otherwise, if the VIRTIO_F_EVENT_IDX feature bit is negotiated:
If the VIRTIO_F_EVENT_IDX feature bit is not negotiated:
Otherwise, if the VIRTIO_F_EVENT_IDX feature bit is negotiated:
The device MUST handle spurious notifications from the driver.
The Linux Kernel Source code contains the definitions above and helper routines in a more usable form, in include/uapi/linux/virtio_ring.h. This was explicitly licensed by IBM and Red Hat under the (3-clause) BSD license so that it can be freely used by all other projects, and is reproduced (with slight variation) in A virtio_queue.h.
The driver MUST follow this sequence to initialize a device:
If any of these steps go irrecoverably wrong, the driver SHOULD set the FAILED status bit to indicate that it has given up on the device (it can reset the device later to restart if desired). The driver MUST NOT continue initialization in that case.
The driver MUST NOT notify the device before setting DRIVER_OK.
Legacy devices did not support the FEATURES_OK status bit, and thus did not have a graceful way for the device to indicate unsupported feature combinations. They also did not provide a clear mechanism to end feature negotiation, which meant that devices finalized features on first-use, and no features could be introduced which radically changed the initial operation of the device.
Legacy driver implementations often used the device before setting the DRIVER_OK bit, and sometimes even before writing the feature bits to the device.
The result was the steps 5 and 6 were omitted, and steps 4, 7 and 8 were conflated.
Therefore, when using the legacy interface:
There are two parts to device operation: supplying new buffers to the device, and processing used buffers from the device. Note: As an example, the simplest virtio network device has two virtqueues: the transmit virtqueue and the receive virtqueue. The driver adds outgoing (device-readable) packets to the transmit virtqueue, and then frees them after they are used. Similarly, incoming (device-writable) buffers are added to the receive virtqueue, and processed after they are used.
The driver offers buffers to one of the device’s virtqueues as follows:
Note that the above code does not take precautions against the available ring buffer wrapping around: this is not possible since the ring buffer is the same size as the descriptor table, so step (1) will prevent such a condition.
In addition, the maximum queue size is 32768 (the highest power of 2 which fits in 16 bits), so the 16-bit idx value can always distinguish between a full and empty buffer.
What follows is the requirements of each stage in more detail.
A buffer consists of zero or more device-readable physically-contiguous elements followed by zero or more physically-contiguous device-writable elements (each has at least one element). This algorithm maps it into the descriptor table to form a descriptor chain:
for each buffer element, b:
In practice, d.next is usually used to chain free descriptors, and a separate count kept to check there are enough free descriptors before beginning the mappings.
The descriptor chain head is the first d in the algorithm above, ie. the index of the descriptor table entry referring to the first part of the buffer. A naive driver implementation MAY do the following (with the appropriate conversion to-and-from little-endian assumed):
However, in general the driver MAY add many descriptor chains before it updates idx (at which point they become visible to the device), so it is common to keep a counter of how many the driver has added:
idx always increments, and wraps naturally at 65536:
Once available idx is updated by the driver, this exposes the descriptor and its contents. The device MAY access the descriptor chains the driver created and the memory they refer to immediately.
The actual method of device notification is bus-specific, but generally it can be expensive. So the device MAY suppress such notifications if it doesn’t need them, as detailed in section 2.4.9.
The driver has to be careful to expose the new idx value before checking if notifications are suppressed.
Once the device has used buffers referred to by a descriptor (read from or written to them, or parts of both, depending on the nature of the virtqueue and the device), it interrupts the driver as detailed in section 2.4.7. Note: For optimal performance, a driver MAY disable interrupts while processing the used ring, but beware the problem of missing interrupts between emptying the ring and reenabling interrupts. This is usually handled by re-checking for more used buffers after interrups are re-enabled:
For devices where the device-specific configuration information can be changed, an interrupt is delivered when a device-specific configuration change occurs.
In addition, this interrupt is triggered by the device setting DEVICE_NEEDS_RESET (see 2.1.2).
Once the driver has set the DRIVER_OK status bit, all the configured virtqueue of the device are considered live. None of the virtqueues of a device are live once the device has been reset.
A driver MUST NOT alter descriptor table entries which have been exposed in the available ring (and not marked consumed by the device in the used ring) of a live virtqueue.
A driver MUST NOT decrement the available idx on a live virtqueue (ie. there is no way to “unexpose” buffers).
Thus a driver MUST ensure a virtqueue isn’t live (by device reset) before removing exposed buffers.
Virtio devices are commonly implemented as PCI devices.
A Virtio device can be implemented as any kind of PCI device: a Conventional PCI device or a PCI Express device. To assure designs meet the latest level requirements, see the PCI-SIG home page at http://www.pcisig.com for any approved changes.
A Virtio device using Virtio Over PCI Bus MUST expose to guest an interface that meets the specification requirements of the appropriate PCI specification: [PCI] and [PCIe] respectively.
Any PCI device with PCI Vendor ID 0x1AF4, and PCI Device ID 0x1000 through 0x107F inclusive is a virtio device. The actual value within this range indicates which virtio device is supported by the device. The PCI Device ID is calculated by adding 0x1040 to the Virtio Device ID, as indicated in section 5. Additionally, devices MAY utilize a Transitional PCI Device ID range, 0x1000 to 0x103F depending on the device type.
Devices MUST have the PCI Vendor ID 0x1AF4. Devices MUST either have the PCI Device ID calculated by adding 0x1040 to the Virtio Device ID, as indicated in section 5 or have the Transitional PCI Device ID depending on the device type, as follows:
|Transitional PCI Device ID||Virtio Device|
|0x1002||memory ballooning (traditional)|
For example, the network card device with the Virtio Device ID 1 has the PCI Device ID 0x1041 or the Transitional PCI Device ID 0x1000.
The PCI Subsystem Vendor ID and the PCI Subsystem Device ID MAY reflect the PCI Vendor and Device ID of the environment (for informational purposes by the driver).
Non-transitional devices SHOULD have a PCI Device ID in the range 0x1040 to 0x107f. Non-transitional devices SHOULD have a PCI Revision ID of 1 or higher. Non-transitional devices SHOULD have a PCI Subsystem Device ID of 0x40 or higher.
This is to reduce the chance of a legacy driver attempting to drive the device.
Drivers MUST match devices with the PCI Vendor ID 0x1AF4 and the PCI Device ID in the range 0x1040 to 0x107f, calculated by adding 0x1040 to the Virtio Device ID, as indicated in section 5. Drivers for device types listed in section 4.1.2 MUST match devices with the PCI Vendor ID 0x1AF4 and the Transitional PCI Device ID indicated in section 4.1.2.
Drivers MUST match any PCI Revision ID value. Drivers MAY match any PCI Subsystem Vendor ID and any PCI Subsystem Device ID value.
Transitional devices MUST have a PCI Revision ID of 0. Transitional devices MUST have the PCI Subsystem Device ID matching the Virtio Device ID, as indicated in section 5. Transitional devices MUST have the Transitional PCI Device ID in the range 0x1000 to 0x103f.
This is to match legacy drivers.
The device is configured via I/O and/or memory regions (though see 220.127.116.11 for access via the PCI configuration space), as specified by Virtio Structure PCI Capabilities.
Fields of different sizes are present in the device configuration regions. All 64-bit, 32-bit and 16-bit fields are little-endian. 64-bit fields are to be treated as two 32-bit fields, with low 32 bit part followed by the high 32 bit part.
For device configuration access, the driver MUST use 8-bit wide accesses for 8-bit wide fields, 16-bit wide and aligned accesses for 16-bit wide fields and 32-bit wide and aligned accesses for 32-bit and 64-bit wide fields. For 64-bit fields, the driver MAY access each of the high and low 32-bit parts of the field independently.
For 64-bit device configuration fields, the device MUST allow driver independent access to high and low 32-bit parts of the field.
The virtio device configuration layout includes several structures:
Each structure can be mapped by a Base Address register (BAR) belonging to the function, or accessed via the special VIRTIO_PCI_CAP_PCI_CFG field in the PCI configuration space.
The location of each structure is specified using a vendor-specific PCI capability located on the capability list in PCI configuration space of the device. This virtio structure capability uses little-endian format; all fields are read-only for the driver unless stated otherwise:
This structure can be followed by extra data, depending on cfg_type, as documented below.
The fields are interpreted as follows:
Any other value is reserved for future use.
Each structure is detailed individually below.
The device MAY offer more than one structure of any type - this makes it possible for the device to expose multiple interfaces to drivers. The order of the capabilities in the capability list specifies the order of preference suggested by the device. Note: For example, on some hypervisors, notifications using IO accesses are faster than memory accesses. In this case, the device would expose two capabilities with cfg_type set to VIRTIO_PCI_CAP_NOTIFY_CFG: the first one addressing an I/O BAR, the second one addressing a memory BAR. In this example, the driver would use the I/O BAR if I/O resources are available, and fall back on memory BAR when I/O resources are unavailable.
Any other value is reserved for future use.
length MAY include padding, or fields unused by the driver, or future extensions. Note: For example, a future device might present a large structure size of several MBytes. As current devices never utilize structures larger than 4KBytes in size, driver MAY limit the mapped structure size to e.g. 4KBytes (thus ignoring parts of structure after the first 4KBytes) to allow forward compatibility with such devices without loss of functionality and without wasting resources.
The driver MUST ignore any vendor-specific capability structure which has a reserved cfg_type value.
The driver SHOULD use the first instance of each virtio structure type they can support.
The driver MUST accept a cap_len value which is larger than specified here.
The driver MUST ignore any vendor-specific capability structure which has a reserved bar value.
The drivers SHOULD only map part of configuration structure large enough for device operation. The drivers MUST handle an unexpectedly large length, but MAY check that length is large enough for device operation.
The driver MUST NOT write into any field of the capability structure, with the exception of those with cap_type VIRTIO_PCI_CAP_PCI_CFG as detailed in 18.104.22.168.2.
The device MUST include any extra data (from the beginning of the cap_vndr field through end of the extra data fields if any) in cap_len. The device MAY append extra data or padding to any structure beyond that.
If the device presents multiple structures of the same type, it SHOULD order them from optimal (first) to least-optimal (last).
The common configuration structure is found at the bar and offset within the VIRTIO_PCI_CAP_COMMON_CFG capability; its layout is below.
The device MUST present at least one common configuration capability.
The device MUST present the feature bits it is offering in device_feature, starting at bit device_feature_select ∗ 32 for any device_feature_select written by the driver. Note: This means that it will present 0 for any device_feature_select other than 0 or 1, since no feature defined here exceeds 63.
The device MUST present any valid feature bits the driver has written in driver_feature, starting at bit driver_feature_select ∗ 32 for any driver_feature_select written by the driver. Valid feature bits are those which are subset of the corresponding device_feature bits. The device MAY present invalid bits written by the driver. Note: This means that a device can ignore writes for feature bits it never offers, and simply present 0 on reads. Or it can just mirror what the driver wrote (but it will still have to check them when the driver sets FEATURES_OK). Note: A driver shouldn’t write invalid bits anyway, as per 3.1.1, but this attempts to handle it.
The device MUST present a changed config_generation after the driver has read a device-specific configuration value which has changed since any part of the device-specific configuration was last read. Note: As config_generation is an 8-bit value, simply incrementing it on every configuration change could violate this requirement due to wrap. Better would be to set an internal flag when it has changed, and if that flag is set when the driver reads from the device-specific configuration, increment config_generation and clear the flag.
The device MUST reset when 0 is written to device_status, and present a 0 in device_status once that is done.
The device MUST present a 0 in queue_enable on reset.
The device MUST present a 0 in queue_size if the virtqueue corresponding to the current queue_select is unavailable.
The driver MUST NOT write a value which is not a power of 2 to queue_size.
The driver MUST configure the other virtqueue fields before enabling the virtqueue with queue_enable.
After writing 0 to device_status, the driver MUST wait for a read of device_status to return 0 before reinitializing the device.
The driver MUST NOT write a 0 to queue_enable.
The notification location is found using the VIRTIO_PCI_CAP_NOTIFY_CFG capability. This capability is immediately followed by an additional field, like so:
notify_off_multiplier is combined with the queue_notify_off to derive the Queue Notify address within a BAR for a virtqueue:
The cap.offset and notify_off_multiplier are taken from the notification capability structure above, and the queue_notify_off is taken from the common configuration structure. Note: For example, if notifier_off_multiplier is 0, the device uses the same Queue Notify address for all queues.
The cap.offset MUST be 2-byte aligned.
The device MUST either present notify_off_multiplier as an even power of 2, or present notify_off_multiplier as 0.
The value cap.length presented by the device MUST be at least 2 and MUST be large enough to support queue notification offsets for all supported queues in all possible configurations.
For all queues, the value cap.length presented by the device MUST satisfy:
The VIRTIO_PCI_CAP_ISR_CFG capability refers to at least a single byte, which contains the 8-bit ISR status field to be used for INT#x interrupt handling.
The offset for the ISR status has no alignment requirements.
The ISR bits allow the device to distinguish between device-specific configuration change interrupts and normal virtqueue interrupts:
|Bits||0||1||2 to 31|
|Purpose||Queue Interrupt||Device Configuration Interrupt||Reserved|
To avoid an extra access, simply reading this register resets it to 0 and causes the device to de-assert the interrupt.
In this way, driver read of ISR status causes the device to de-assert an interrupt.
See sections 22.214.171.124 and 126.96.36.199 for how this is used.
The device MUST set the Device Configuration Interrupt bit in ISR status before sending a device configuration change notification to the driver.
If MSI-X capability is disabled, the device MUST set the Queue Interrupt bit in ISR status before sending a virtqueue notification to the driver.
If MSI-X capability is disabled, the device MUST set the Interrupt Status bit in the PCI Status register in the PCI Configuration Header of the device to the logical OR of all bits in ISR status of the device. The device then asserts/deasserts INT#x interrupts unless masked according to standard PCI rules [PCI].
The device MUST reset ISR status to 0 on driver read.
The device MUST present at least one VIRTIO_PCI_CAP_DEVICE_CFG capability for any device type which has a device-specific configuration.
The VIRTIO_PCI_CAP_PCI_CFG capability creates an alternative (and likely suboptimal) access method to the common configuration, notification, ISR and device-specific configuration regions.
The capability is immediately followed by an additional field like so:
The fields cap.bar, cap.length, cap.offset and pci_cfg_data are read-write (RW) for the driver.
To access a device region, the driver writes into the capability structure (ie. within the PCI configuration space) as follows:
At that point, pci_cfg_data will provide a window of size cap.length into the given cap.bar at offset cap.offset.
Upon detecting driver write access to pci_cfg_data, the device MUST execute a write access at offset cap.offset at BAR selected by cap.bar using the first cap.length bytes from pci_cfg_data.
Upon detecting driver read access to pci_cfg_data, the device MUST execute a read access of length cap.length at offset cap.offset at BAR selected by cap.bar and store the first cap.length bytes in pci_cfg_data.
The driver MUST NOT read or write pci_cfg_data unless cap.bar, cap.length and cap.offset address cap.length bytes within a BAR range specified by some other Virtio Structure PCI Capability of type other than VIRTIO_PCI_CAP_PCI_CFG.
Transitional devices MUST present part of configuration registers in a legacy configuration structure in BAR0 in the first I/O region of the PCI device, as documented below. When using the legacy interface, transitional drivers MUST use the legacy configuration structure in BAR0 in the first I/O region of the PCI device, as documented below.
When using the legacy interface the driver MAY access the device-specific configuration region using any width accesses, and a transitional device MUST present driver with the same results as when accessed using the “natural” access method (i.e. 32-bit accesses for 32-bit fields, etc).
Note that this is possible because while the virtio common configuration structure is PCI (i.e. little) endian, when using the legacy interface the device-specific configuration region is encoded in the native endian of the guest (where such distinction is applicable).
When used through the legacy interface, the virtio common configuration structure looks as follows:
|Read / Write||R||R+W||R+W||R||R+W||R+W||R+W||R|
|Purpose||Device Features bits 0:31||Driver Features bits 0:31||Queue Address||queue_size||queue_select||Queue Notify||Device Status|| ISR
If MSI-X is enabled for the device, two additional fields immediately follow this header:
Note: When MSI-X capability is enabled, device-specific configuration starts at byte offset 24 in virtio common configuration structure structure. When MSI-X capability is not enabled, device-specific configuration starts at byte offset 20 in virtio header. ie. once you enable MSI-X on the device, the other fields move. If you turn it off again, they move back!
Any device-specific configuration space immediately follows these general headers:
|Bits||Device Specific|| |
|Read / Write||Device Specific|
When accessing the device-specific configuration space using the legacy interface, transitional drivers MUST access the device-specific configuration space at an offset immediately following the general headers.
When using the legacy interface, transitional devices MUST present the device-specific configuration space if any at an offset immediately following the general headers.
Note that only Feature Bits 0 to 31 are accessible through the Legacy Interface. When used through the Legacy Interface, Transitional Devices MUST assume that Feature Bits 32 to 63 are not acknowledged by Driver.
As legacy devices had no config_generation field, see 2.3.4 Legacy Interface: Device Configuration Space for workarounds.
All known legacy drivers check either the PCI Revision or the Device and Vendor IDs, and thus won’t attempt to drive a non-transitional device.
A buggy legacy driver might mistakenly attempt to drive a non-transitional device. If support for such drivers is required (as opposed to fixing the bug), the following would be the recommended way to detect and handle them. Note: Such buggy drivers are not currently known to be used in production.
This documents PCI-specific steps executed during Device Initialization.
Legacy devices did not have the Virtio PCI Capability in their capability list.
Transitional devices MUST expose the Legacy Interface in I/O space in BAR0.
Transitional drivers MUST look for the Virtio PCI Capabilities on the capability list. If these are not present, driver MUST assume a legacy device, and use it through the legacy interface.
Non-transitional drivers MUST look for the Virtio PCI Capabilities on the capability list. If these are not present, driver MUST assume a legacy device, and fail gracefully.
Writing a valid MSI-X Table entry number, 0 to 0x7FF, to config_msix_vector/queue_msix_vector maps interrupts triggered by the configuration change/selected queue events respectively to the corresponding MSI-X vector. To disable interrupts for an event type, the driver unmaps this event by writing a special NO_VECTOR value:
Note that mapping an event to vector might require device to allocate internal device resources, and thus could fail.
Device MUST support mapping any event type to any valid vector 0 to MSI-X Table Size. Device MUST support unmapping any event type.
The device MUST return vector mapped to a given event, (NO_VECTOR if unmapped) on read of config_msix_vector/queue_msix_vector. The device MUST have all queue and configuration change events are unmapped upon reset.
Devices SHOULD NOT cause mapping an event to vector to fail unless it is impossible for the device to satisfy the mapping request. Devices MUST report mapping failures by returning the NO_VECTOR value when the relevant config_msix_vector/queue_msix_vector field is read.
Driver MAY intepret the Table Size as a hint from the device for the suggested number of MSI-X vectors to use.
Driver MUST NOT attempt to map an event to a vector outside the MSI-X Table supported by the device, as reported by Table Size in the MSI-X Capability.
After mapping an event to vector, the driver MUST verify success by reading the Vector field value: on success, the previously written value is returned, and on failure, NO_VECTOR is returned. If a mapping failure is detected, the driver MAY retry mapping with fewer vectors, disable MSI-X or report device failure.
The driver typically does this as follows, for each virtqueue a device has:
The driver notifies the device by writing the 16-bit virtqueue index of this virtqueue to the Queue Notify address. See 188.8.131.52 for how to calculate this address.
If an interrupt is necessary for a virtqueue, the device would typically act as follows:
Some virtio PCI devices can change the device configuration state, as reflected in the device-specific configuration region of the device. In this case:
A single interrupt MAY indicate both that one or more virtqueue has been used and that the configuration space has changed.
The driver interrupt handler would typically:
Virtual environments without PCI support (a common situation in embedded devices models) might use simple memory mapped device (“virtio-mmio”) instead of the PCI device.
The memory mapped virtio device behaviour is based on the PCI device specification. Therefore most operations including device initialization, queues configuration and buffer transfers are nearly identical. Existing differences are described in the following sections.
Unlike PCI, MMIO provides no generic device discovery mechanism. For each device, the guest OS will need to know the location of the registers and interrupt(s) used. The suggested binding for systems using flattened device trees is shown in this example:
MMIO virtio devices provide a set of memory mapped control registers followed by a device-specific configuration space, described in the table 4.1.
All register values are organized as Little Endian.
Virtio Subsystem Device ID
Virtio Subsystem Vendor ID
Flags representing features the device supports
Device (host) features word selection.
Flags representing device features understood
and activated by the driver
Activated (guest) features word selection
Virtual queue index
Maximum virtual queue size
Virtual queue size
Virtual queue ready bit
Virtual queue’s Descriptor Table 64 bit long
Virtual queue’s Available Ring 64 bit long
Virtual queue’s Used Ring 64 bit long physical
Configuration atomicity value
The device MUST return 0x74726976 in MagicValue.
The device MUST return value 0x2 in Version.
The device MUST present each event by setting the corresponding bit in InterruptStatus from the moment it takes place, until the driver acknowledges the interrupt by writing a corresponding bit mask to the InterruptACK register. Bits which do not represent events which took place MUST be zero.
Upon reset, the device MUST clear all bits in InterruptStatus and ready bits in the QueueReady register for all queues in the device.
The device MUST change value returned in ConfigGeneration if there is any risk of a driver seeing an inconsistent configuration state.
The device MUST NOT access virtual queue contents when QueueReady is zero (0x0).
The driver MUST NOT access memory locations not described in the table 4.1 (or, in case of the configuration space, described in the device specification), MUST NOT write to the read-only registers (direction R) and MUST NOT read from the write-only registers (direction W).
The driver MUST only use 32 bit wide and aligned reads and writes to access the control registers described in table 4.1. For the device-specific configuration space, the driver MUST use 8 bit wide accesses for 8 bit wide fields, 16 bit wide and aligned accesses for 16 bit wide fields and 32 bit wide and aligned accesses for 32 and 64 bit wide fields.
The driver MUST ignore a device with MagicValue which is not 0x74726976, although it MAY report an error.
The driver MUST ignore a device with Version which is not 0x2, although it MAY report an error.
The driver MUST ignore a device with DeviceID 0x0, but MUST NOT report any error.
Before reading from DeviceFeatures, the driver MUST write a value to DeviceFeaturesSel.
Before writing to the DriverFeatures register, the driver MUST write a value to the DriverFeaturesSel register.
The driver MUST write a value to QueueNum which is less than or equal to the value presented by the device in QueueNumMax.
When QueueReady is not zero, the driver MUST NOT access QueueNum, QueueDescLow, QueueDescHigh, QueueAvailLow, QueueAvailHigh, QueueUsedLow, QueueUsedHigh.
To stop using the queue the driver MUST write zero (0x0) to this QueueReady and MUST read the value back to ensure synchronization.
The driver MUST ignore undefined bits in InterruptStatus.
The driver MUST write a value with a bit mask describing events it handled into InterruptACK when it finishes handling an interrupt and MUST NOT set any of the undefined bits in the value.
Further initialization MUST follow the procedure described in 3.1 Device Initialization.
The driver will typically initialize the virtual queue in the following way:
The driver notifies the device about new buffers being available in a queue by writing the index of the updated queue to QueueNotify.
The memory mapped virtio device is using a single, dedicated interrupt signal, which is asserted when at least one of the bits described in the description of InterruptStatus is set. This is how the device notifies the driver about a new used buffer being available in the queue or about a change in the device configuration.
The legacy MMIO transport used page-based addressing, resulting in a slightly different control register layout, the device initialization and the virtual queue configuration procedure.
Table 4.2 presents control registers layout, omitting descriptions of registers which did not change their function nor behaviour:
Device version number
Virtio Subsystem Device ID
Virtio Subsystem Vendor ID
Flags representing features the device supports
Device (host) features word selection.
Flags representing device features understood
and activated by the driver
Activated (guest) features word selection
Guest page size
Virtual queue index
Maximum virtual queue size
Virtual queue size
Used Ring alignment in the virtual queue
Guest physical page number of the virtual queue
The virtual queue page size is defined by writing to GuestPageSize, as written by the guest. The driver does this before the virtual queues are configured.
The virtual queue layout follows p. 2.4.2 Legacy Interfaces: A Note on Virtqueue Layout, with the alignment defined in QueueAlign.
The virtual queue is configured as follows:
Notification mechanisms did not change.
S/390 based virtual machines support neither PCI nor MMIO, so a different transport is needed there.
virtio-ccw uses the standard channel I/O based mechanism used for the majority of devices on S/390. A virtual channel device with a special control unit type acts as proxy to the virtio device (similar to the way virtio-pci uses a PCI device) and configuration and operation of the virtio device is accomplished (mostly) via channel commands. This means virtio devices are discoverable via standard operating system algorithms, and adding virtio support is mainly a question of supporting a new control unit type.
As the S/390 is a big endian machine, the data structures transmitted via channel commands are big-endian: this is made clear by use of the types be16, be32 and be64.
As a proxy device, virtio-ccw uses a channel-attached I/O control unit with a special control unit type (0x3832) and a control unit model corresponding to the attached virtio device’s subsystem device ID, accessed via a virtual I/O subchannel and a virtual channel path of type 0x32. This proxy device is discoverable via normal channel subsystem device discovery (usually a STORE SUBCHANNEL loop) and answers to the basic channel commands:
For a virtio-ccw proxy device, SENSE ID will return the following information:
|1-2||control unit type||0x3832|
|3||control unit model||
|4-5||device type||zeroes (unset)|
|6||device model||zeroes (unset)|
|7-255||extended SenseId data||zeroes (unset)|
In addition to the basic channel commands, virtio-ccw defines a set of channel commands related to configuration and operation of virtio:
The virtio-ccw device acts like a normal channel device, as specified in [S390 PoP] and [S390 Common I/O]. In particular:
A driver for virtio-ccw devices MUST check for a control unit type of 0x3832 and MUST ignore the device type and model.
A driver SHOULD attempt to provide the correct length in a channel command even if it suppresses length checks for that command.
virtio-ccw uses several channel commands to set up a device.
CCW_CMD_SET_VIRTIO_REV is issued by the driver to set the revision of the virtio-ccw transport it intends to drive the device with. It uses the following communication structure:
revision contains the desired revision id, length the length of the data portion and data revision-dependent additional desired options.
The following values are supported:
|0||0|| ||legacy interface; transitional devices only|
|1||0|| ||Virtio 1.0|
|2||0|| ||CCW_CMD_READ_STATUS support|
|3-n||reserved for later revisions|
Note that a change in the virtio standard does not necessarily correspond to a change in the virtio-ccw revision.
A device MUST answer with command reject to any virtio-ccw specific channel command that is not contained in the revision selected by the driver.
A device MUST answer with command reject to any attempt to select a different revision after a revision has been successfully selected by the driver.
A device MUST treat the revision as unset from the time the associated subchannel has been enabled until a revision has been successfully set by the driver. This implies that revisions are not persistent across disabling and enabling of the associated subchannel.
A driver MUST NOT issue any other virtio-ccw specific channel commands prior to setting the revision.
After a revision has been successfully selected by the driver, it MUST NOT attempt to select a different revision.
A legacy driver will not issue the CCW_CMD_SET_VIRTIO_REV prior to issuing other virtio-ccw specific channel commands. A non-transitional device therefore MUST answer any such attempts with a command reject. A transitional device MUST assume in this case that the driver is a legacy driver and continue as if the driver selected revision 0. This implies that the device MUST reject any command not valid for revision 0, including a subsequent CCW_CMD_SET_VIRTIO_REV.
CCW_CMD_READ_VQ_CONF is issued by the driver to obtain information about a queue. It uses the following structure for communicating:
The requested number of buffers for queue index is returned in max_num.
Afterwards, CCW_CMD_SET_VQ is issued by the driver to inform the device about the location used for its queue. The transmitted structure is
desc, avail and used contain the guest addresses for the descriptor table, available ring and used ring for queue index, respectively. The actual virtqueue size (number of allocated buffers) is transmitted in num.
queue contains the guest address for queue index, num the number of buffers and align the alignment. The queue layout follows 2.4.2 Legacy Interfaces: A Note on Virtqueue Layout.
The driver changes the status of a device via the CCW_CMD_WRITE_STATUS command, which transmits an 8 bit status value.
As described in 2.2.2, a device sometimes fails to set the status field: For example, it might fail to accept the FEATURES_OK status bit during device initialization.
With revision 2, CCW_CMD_READ_STATUS is defined: It reads an 8 bit status value from the device and acts as a reverse operation to CCW_CMD_WRITE_STATUS.
If at least revision 2 has been negotiated, the driver SHOULD use the CCW_CMD_READ_STATUS command to retrieve the status field after a configuration change has been detected.
If not at least revision 2 has been negotiated, the driver MUST NOT attempt to issue the CCW_CMD_READ_STATUS command.
If at least revision 2 has been negotiated, the device MUST return the current status field if the CCW_CMD_READ_STATUS command is issued.
Feature bits are arranged in an array of 32 bit values, making for a total of 8192 feature bits. Feature bits are in little-endian byte order.
The CCW commands dealing with features use the following communication block:
features are the 32 bits of features currently accessed, while index describes which of the feature bit values is to be accessed. No padding is added at the end of the structure, it is exactly 5 bytes in length.
The guest obtains the device’s device feature set via the CCW_CMD_READ_FEAT command. The device stores the features at index to features.
For communicating its supported features to the device, the driver uses the CCW_CMD_WRITE_FEAT command, denoting a features/index combination.
The device’s configuration space is located in host memory.
To obtain information from the configuration space, the driver uses CCW_CMD_READ_CONF, specifying the guest memory for the device to write to.
For changing configuration information, the driver uses CCW_CMD_WRITE_CONF, specifying the guest memory for the device to read from.
In both cases, the complete configuration space is transmitted. This allows the driver to compare the new configuration space with the old version, and keep a generation count internally whenever it changes.
In order to set up the indicator bits for host->guest notification, the driver uses different channel commands depending on whether it wishes to use traditional I/O interrupts tied to a subchannel or adapter I/O interrupts for virtqueue notifications. For any given device, the two mechanisms are mutually exclusive.
For the configuration change indicators, only a mechanism using traditional I/O interrupts is provided, regardless of whether traditional or adapter I/O interrupts are used for virtqueue notifications.
To communicate the location of the indicator bits for host->guest notification, the driver uses the CCW_CMD_SET_IND command, pointing to a location containing the guest address of the indicators in a 64 bit value.
If the driver has already set up two-staged queue indicators via the CCW_CMD_SET_IND_ADAPTER command, the device MUST post a unit check with command reject to any subsequent CCW_CMD_SET_IND command.
To communicate the location of the indicator bits used in the configuration change host->guest notification, the driver issues the CCW_CMD_SET_CONF_IND command, pointing to a location containing the guest address of the indicators in a 64 bit value.
To communicate the location of the summary and queue indicator bits, the driver uses the CCW_CMD_SET_IND_ADAPTER command with the following payload:
summary_indicator contains the guest address of the 8 bit summary indicator. indicator contains the guest address of an area wherein the indicators for the devices are contained, starting at bit_nr, one bit per virtqueue of the device. Bit numbers start at the left, i.e. the most significant bit in the first byte is assigned the bit number 0. isc contains the I/O interruption subclass to be used for the adapter I/O interrupt. It MAY be different from the isc used by the proxy virtio-ccw device’s subchannel. No padding is added at the end of the structure, it is exactly 25 bytes in length.
There are two modes of operation regarding host->guest notification, classic I/O interrupts and adapter I/O interrupts. The mode to be used is determined by the driver by using CCW_CMD_SET_IND respectively CCW_CMD_SET_IND_ADAPTER to set up queue indicators.
For configuration changes, the driver always uses classic I/O interrupts.
For notifying the driver of virtqueue buffers, the device sets the corresponding bit in the guest-provided indicators. If an interrupt is not already pending for the subchannel, the device generates an unsolicited I/O interrupt.
If the device wants to notify the driver about configuration changes, it sets bit 0 in the configuration indicators and generates an unsolicited I/O interrupt, if needed. This also applies if adapter I/O interrupts are used for queue notifications.
For notifying the driver of virtqueue buffers, the device sets the bit in the guest-provided indicator area at the corresponding offset. The guest-provided summary indicator is set to 0x01. An adapter I/O interrupt for the corresponding interruption subclass is generated.
The recommended way to process an adapter I/O interrupt by the driver is as follows:
For notifying the device of virtqueue buffers, the driver unfortunately can’t use a channel command (the asynchronous characteristics of channel I/O interact badly with the host block I/O backend). Instead, it uses a diagnose 0x500 call with subcode 3 specifying the queue, as follows:
|GPR||Input Value||Output Value|
|2||Subchannel ID||Host Cookie|
The device MAY return a 64-bit host cookie in GPR2 to speed up the notification execution.
In order to reset a device, a driver sends the CCW_CMD_VDEV_RESET command.
The following device IDs are used to identify different types of virtio devices. Some device IDs are reserved for devices which are not currently defined in this standard.
Discovering what devices are available and their type is bus-dependent.
|Device ID||Virtio Device|
|5||memory ballooning (traditional)|
|21||Signal Distribution Module|
|24||vhost-user device backend|
Some of the devices above are unspecified by this document, because they are seen as immature or especially niche. Be warned that some are only specified by the sole existing implementation; they could become part of a future specification, be abandoned entirely, or live on outside this standard. We shall speak of them no further.
The virtio network device is a virtual ethernet card, and is the most complex of the devices supported so far by virtio. It has enhanced rapidly and demonstrates clearly how support for new features are added to an existing device. Empty buffers are placed in one virtqueue for receiving packets, and outgoing packets are enqueued into another for transmission in that order. A third command queue is used to control advanced filtering features.
N=1 if VIRTIO_NET_F_MQ is not negotiated, otherwise N is set by max_virtqueue_pairs.
controlq only exists if VIRTIO_NET_F_CTRL_VQ set.
Some networking feature bits require other networking feature bits (see 2.2.1):
This was supposed to indicate segmentation offload support, but upon further investigation it became clear that multiple bits were needed.
Three driver-read-only configuration fields are currently defined. The mac address field always exists (though is only valid if VIRTIO_NET_F_MAC is set), and status only exists if VIRTIO_NET_F_STATUS is set. Two read-only bits (for the driver) are currently defined for the status field: VIRTIO_NET_S_LINK_UP and VIRTIO_NET_S_ANNOUNCE.
The following driver-read-only field, max_virtqueue_pairs only exists if VIRTIO_NET_F_MQ is set. This field specifies the maximum number of each of transmit and receive virtqueues (receiveq1…receiveqN and transmitq1…transmitqN respectively) that can be configured once VIRTIO_NET_F_MQ is negotiated.
The following driver-read-only field, mtu only exists if VIRTIO_NET_F_MTU is set. This field specifies the maximum MTU for the driver to use.
The device MUST set max_virtqueue_pairs to between 1 and 0x8000 inclusive, if it offers VIRTIO_NET_F_MQ.
The device MUST set mtu to between 68 and 65535 inclusive, if it offers VIRTIO_NET_F_MTU.
The device SHOULD set mtu to at least 1280, if it offers VIRTIO_NET_F_MTU.
The device MUST NOT modify mtu once it has been set.
The device MUST NOT pass received packets that exceed mtu (plus low level ethernet header length) size with gso_type NONE or ECN after VIRTIO_NET_F_MTU has been successfully negotiated.
The device MUST forward transmitted packets of up to mtu (plus low level ethernet header length) size with gso_type NONE or ECN, and do so without fragmentation, after VIRTIO_NET_F_MTU has been successfully negotiated.
A driver SHOULD negotiate VIRTIO_NET_F_MAC if the device offers it. If the driver negotiates the VIRTIO_NET_F_MAC feature, the driver MUST set the physical address of the NIC to mac. Otherwise, it SHOULD use a locally-administered MAC address (see IEEE 802, “9.2 48-bit universal LAN MAC addresses”).
If the driver does not negotiate the VIRTIO_NET_F_STATUS feature, it SHOULD assume the link is active, otherwise it SHOULD read the link status from the bottom bit of status.
A driver SHOULD negotiate VIRTIO_NET_F_MTU if the device offers it.
If the driver negotiates VIRTIO_NET_F_MTU, it MUST supply enough receive buffers to receive at least one receive packet of size mtu (plus low level ethernet header length) with gso_type NONE or ECN.
If the driver negotiates VIRTIO_NET_F_MTU, it MUST NOT transmit packets of size exceeding the value of mtu (plus low level ethernet header length) with gso_type NONE or ECN.
When using the legacy interface, transitional devices and drivers MUST format status and max_virtqueue_pairs in struct virtio_net_config according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
When using the legacy interface, mac is driver-writable which provided a way for drivers to update the MAC without negotiating VIRTIO_NET_F_CTRL_MAC_ADDR.
A driver would perform a typical initialization routine like so:
A truly minimal driver would only accept VIRTIO_NET_F_MAC and ignore everything else.
Packets are transmitted by placing them in the transmitq1…transmitqN, and buffers for incoming packets are placed in the receiveq1…receiveqN. In each case, the packet itself is preceded by a header:
The controlq is used to control device features such as filtering.
When using the legacy interface, transitional devices and drivers MUST format the fields in struct virtio_net_hdr according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
The legacy driver only presented num_buffers in the struct virtio_net_hdr when VIRTIO_NET_F_MRG_RXBUF was negotiated; without that feature the structure was 2 bytes shorter.
When using the legacy interface, the driver SHOULD ignore the len value in used ring entries for the transmit queues and the controlq queue. Note: Historically, some devices put the total descriptor length there, even though no data was actually written.
Transmitting a single packet is simple, but varies depending on the different features the driver negotiated.
If VIRTIO_NET_F_CSUM is not negotiated, the driver MUST set flags to zero and SHOULD supply a fully checksummed packet to the device.
If VIRTIO_NET_F_HOST_TSO4 is negotiated, the driver MAY set gso_type to VIRTIO_NET_HDR_GSO_TCPV4 to request TCPv4 segmentation, otherwise the driver MUST NOT set gso_type to VIRTIO_NET_HDR_GSO_TCPV4.
If VIRTIO_NET_F_HOST_TSO6 is negotiated, the driver MAY set gso_type to VIRTIO_NET_HDR_GSO_TCPV6 to request TCPv6 segmentation, otherwise the driver MUST NOT set gso_type to VIRTIO_NET_HDR_GSO_TCPV6.
If VIRTIO_NET_F_HOST_UFO is negotiated, the driver MAY set gso_type to VIRTIO_NET_HDR_GSO_UDP to request UDP segmentation, otherwise the driver MUST NOT set gso_type to VIRTIO_NET_HDR_GSO_UDP.
The driver SHOULD NOT send to the device TCP packets requiring segmentation offload which have the Explicit Congestion Notification bit set, unless the VIRTIO_NET_F_HOST_ECN feature is negotiated, in which case the driver MUST set the VIRTIO_NET_HDR_GSO_ECN bit in gso_type.
If the VIRTIO_NET_F_CSUM feature has been negotiated, the driver MAY set the VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags, if so:
If none of the VIRTIO_NET_F_HOST_TSO4, TSO6 or UFO options have been negotiated, the driver MUST set gso_type to VIRTIO_NET_HDR_GSO_NONE.
If gso_type differs from VIRTIO_NET_HDR_GSO_NONE, then the driver MUST also set the VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags and MUST set gso_size to indicate the desired MSS.
If one of the VIRTIO_NET_F_HOST_TSO4, TSO6 or UFO options have been negotiated, the driver SHOULD set hdr_len to a value not less than the length of the headers, including the transport header.
The driver MUST NOT set the VIRTIO_NET_HDR_F_DATA_VALID bit in flags.
If VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags is not set, the device MUST NOT use the csum_start and csum_offset.
If one of the VIRTIO_NET_F_HOST_TSO4, TSO6 or UFO options have been negotiated, the device MAY use hdr_len only as a hint about the transport header size. The device MUST NOT rely on hdr_len to be correct. Note: This is due to various bugs in implementations.
If VIRTIO_NET_HDR_F_NEEDS_CSUM is not set, the device MUST NOT rely on the packet checksum being correct.
The normal behavior in this interrupt handler is to retrieve and new descriptors from the used ring and free the corresponding headers and packets.
It is generally a good idea to keep the receive virtqueue as fully populated as possible: if it runs out, network performance will suffer.
If the VIRTIO_NET_F_GUEST_TSO4, VIRTIO_NET_F_GUEST_TSO6 or VIRTIO_NET_F_GUEST_UFO features are used, the maximum incoming packet will be to 65550 bytes long (the maximum size of a TCP or UDP packet, plus the 14 byte ethernet header), otherwise 1514 bytes. The 12-byte struct virtio_net_hdr is prepended to this, making for 65562 or 1526 bytes.
If VIRTIO_NET_F_MQ is negotiated, each of receiveq1…receiveqN that will be used SHOULD be populated with receive buffers.
The device MUST use only a single descriptor if VIRTIO_NET_F_MRG_RXBUF was not negotiated. Note: This means that num_buffers will always be 1 if VIRTIO_NET_F_MRG_RXBUF is not negotiated.
When a packet is copied into a buffer in the receiveq, the optimal path is to disable further interrupts for the receiveq (see 3.2.2 Receiving Used Buffers From The Device) and process packets until no more are found, then re-enable them.
Processing incoming packets involves:
Additionally, VIRTIO_NET_F_GUEST_CSUM, TSO4, TSO6, UDP and ECN features enable receive checksum, large receive offload and ECN support which are the input equivalents of the transmit checksum, transmit segmentation offloading and ECN features, as described in 184.108.40.206:
If VIRTIO_NET_F_MRG_RXBUF has been negotiated, the device MUST set num_buffers to indicate the number of buffers the packet (including the header) is spread over.
If a receive packet is spread over multiple buffers, the device MUST use all buffers but the last (i.e. the first numbuffers− 1 buffers) completely up to the full length of each buffer supplied by the driver.
The device MUST use all buffers used by a single receive packet together, by atomically incrementing idx in the used ring by the num_buffers value.
If VIRTIO_NET_F_GUEST_CSUM is not negotiated, the device MUST set flags to zero and SHOULD supply a fully checksummed packet to the driver.
If VIRTIO_NET_F_GUEST_TSO4 is not negotiated, the device MUST NOT set gso_type to VIRTIO_NET_HDR_GSO_TCPV4.
If VIRTIO_NET_F_GUEST_UDP is not negotiated, the device MUST NOT set gso_type to VIRTIO_NET_HDR_GSO_UDP.
If VIRTIO_NET_F_GUEST_TSO6 is not negotiated, the device MUST NOT set gso_type to VIRTIO_NET_HDR_GSO_TCPV6.
The device SHOULD NOT send to the driver TCP packets requiring segmentation offload which have the Explicit Congestion Notification bit set, unless the VIRTIO_NET_F_GUEST_ECN feature is negotiated, in which case the device MUST set the VIRTIO_NET_HDR_GSO_ECN bit in gso_type.
If the VIRTIO_NET_F_GUEST_CSUM feature has been negotiated, the device MAY set the VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags, if so:
If none of the VIRTIO_NET_F_GUEST_TSO4, TSO6 or UFO options have been negotiated, the device MUST set gso_type to VIRTIO_NET_HDR_GSO_NONE.
If gso_type differs from VIRTIO_NET_HDR_GSO_NONE, then the device MUST also set the VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags MUST set gso_size to indicate the desired MSS.
If one of the VIRTIO_NET_F_GUEST_TSO4, TSO6 or UFO options have been negotiated, the device SHOULD set hdr_len to a value not less than the length of the headers, including the transport header.
If the VIRTIO_NET_F_GUEST_CSUM feature has been negotiated, the device MAY set the VIRTIO_NET_HDR_F_DATA_VALID bit in flags, if so, the device MUST validate the packet checksum (in case of multiple encapsulated protocols, one level of checksums is validated).
If VIRTIO_NET_HDR_F_NEEDS_CSUM bit in flags is not set, the driver MUST NOT use the csum_start and csum_offset.
If one of the VIRTIO_NET_F_GUEST_TSO4, TSO6 or UFO options have been negotiated, the driver MAY use hdr_len only as a hint about the transport header size. The driver MUST NOT rely on hdr_len to be correct. Note: This is due to various bugs in implementations.
If neither VIRTIO_NET_HDR_F_NEEDS_CSUM nor VIRTIO_NET_HDR_F_DATA_VALID is set, the driver MUST NOT rely on the packet checksum being correct.
The driver uses the control virtqueue (if VIRTIO_NET_F_CTRL_VQ is negotiated) to send commands to manipulate various features of the device which would not easily map into the configuration space.
All commands are of the following form:
The class, command and command-specific-data are set by the driver, and the device sets the ack byte. There is little it can do except issue a diagnostic if ack is not VIRTIO_NET_OK.
If the VIRTIO_NET_F_CTRL_RX_EXTRA feature has been negotiated, the device MUST support the following VIRTIO_NET_CTRL_RX class commands:
If the VIRTIO_NET_F_CTRL_RX_EXTRA feature has not been negotiated, the driver MUST NOT issue commands VIRTIO_NET_CTRL_RX_ALLUNI, VIRTIO_NET_CTRL_RX_NOMULTI, VIRTIO_NET_CTRL_RX_NOUNI or VIRTIO_NET_CTRL_RX_NOBCAST.
The device can filter incoming packets by any number of destination MAC addresses9. This table is set using the class VIRTIO_NET_CTRL_MAC and the command VIRTIO_NET_CTRL_MAC_TABLE_SET. The command-specific-data is two variable length tables of 6-byte MAC addresses (as described in struct virtio_net_ctrl_mac). The first table contains unicast addresses, and the second contains multicast addresses.
The VIRTIO_NET_CTRL_MAC_ADDR_SET command is used to set the default MAC address which rx filtering accepts (and if VIRTIO_NET_F_MAC_ADDR has been negotiated, this will be reflected in mac in config space).
The command-specific-data for VIRTIO_NET_CTRL_MAC_ADDR_SET is the 6-byte MAC address.
The device MUST update the MAC filtering table before it consumes the VIRTIO_NET_CTRL_MAC_TABLE_SET command.
The device MUST update mac in config space before it consumes the VIRTIO_NET_CTRL_MAC_ADDR_SET command, if VIRTIO_NET_F_MAC_ADDR has been negotiated.
The device SHOULD drop incoming packets which have a destination MAC which matches neither the mac (or that set with VIRTIO_NET_CTRL_MAC_ADDR_SET) nor the MAC filtering table.
If VIRTIO_NET_F_CTRL_RX has been negotiated, the driver SHOULD issue VIRTIO_NET_CTRL_MAC_ADDR_SET to set the default mac if it is different from mac.
The driver MUST follow the VIRTIO_NET_CTRL_MAC_TABLE_SET command by a le32 number, followed by that number of non-multicast MAC addresses, followed by another le32 number, followed by that number of multicast addresses. Either number MAY be 0.
Legacy drivers that didn’t negotiate VIRTIO_NET_F_CTRL_MAC_ADDR changed mac in config space when NIC is accepting incoming packets. These drivers always wrote the mac value from first to last byte, therefore after detecting such drivers, a transitional device MAY defer MAC update, or MAY defer processing incoming packets until driver writes the last byte of mac in the config space.
Both the VIRTIO_NET_CTRL_VLAN_ADD and VIRTIO_NET_CTRL_VLAN_DEL command take a little-endian 16-bit VLAN id as the command-specific-data.
The driver checks VIRTIO_NET_S_ANNOUNCE bit in the device configuration status field when it notices the changes of device configuration. The command VIRTIO_NET_CTRL_ANNOUNCE_ACK is used to indicate that driver has received the notification and device clears the VIRTIO_NET_S_ANNOUNCE bit in status.
Processing this notification involves:
Multiqueue is disabled by default. The driver enables multiqueue by executing the VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command, specifying the number of the transmit and receive queues to be used up to max_virtqueue_pairs; subsequently, transmitq1…transmitqn and receiveq1…receiveqn where n=virtqueue_pairs MAY be used.
When multiqueue is enabled, the device MUST use automatic receive steering based on packet flow. Programming of the receive steering classificator is implicit. After the driver transmitted a packet of a flow on transmitqX, the device SHOULD cause incoming packets for that flow to be steered to receiveqX. For uni-directional protocols, or where no packets have been transmitted yet, the device MAY steer a packet to a random queue out of the specified receiveq1…receiveqn.
Multiqueue is disabled by setting virtqueue_pairs to 1 (this is the default) and waiting for the device to use the command buffer.
The driver MUST NOT request a virtqueue_pairs of 0 or greater than max_virtqueue_pairs in the device configuration space.
The driver MUST queue packets only on any transmitq1 before the VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command.
The driver MUST NOT queue packets on transmit queues greater than virtqueue_pairs once it has placed the VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command in the available ring.
The device MUST NOT queue packets on receive queues greater than virtqueue_pairs once it has placed the VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET command in the used ring.
The class VIRTIO_NET_CTRL_GUEST_OFFLOADS has one command: VIRTIO_NET_CTRL_GUEST_OFFLOADS_SET applies the new offloads configuration.
le64 value passed as command data is a bitmask, bits set define offloads to be enabled, bits cleared - offloads to be disabled.
There is a corresponding device feature for each offload. Upon feature negotiation corresponding offload gets enabled to preserve backward compartibility.
When using legacy interfaces, transitional drivers which have not negotiated VIRTIO_F_ANY_LAYOUT MUST use a single descriptor for the struct virtio_net_hdr on both transmit and receive, with the network data in the following descriptors.
Additionally, when using the control virtqueue (see 220.127.116.11) , transitional drivers which have not negotiated VIRTIO_F_ANY_LAYOUT MUST:
The virtio block device is a simple virtual block device (ie. disk). Read and write requests (and other exotic requests) are placed in the queue, and serviced (probably out of order) by the device except where noted.
The capacity of the device (expressed in 512-byte sectors) is always present. The availability of the others all depend on various feature bits as indicated above.
When using the legacy interface, transitional devices and drivers MUST format the fields in struct virtio_blk_config according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
Drivers SHOULD NOT negotiate VIRTIO_BLK_F_FLUSH if they are incapable of sending VIRTIO_BLK_T_FLUSH commands.
If neither VIRTIO_BLK_F_CONFIG_WCE nor VIRTIO_BLK_F_FLUSH are negotiated, the driver MAY deduce the presence of a writethrough cache. If VIRTIO_BLK_F_CONFIG_WCE was not negotiated but VIRTIO_BLK_F_FLUSH was, the driver SHOULD assume presence of a writeback cache.
The driver MUST NOT read writeback before setting the FEATURES_OK status bit.
Devices SHOULD always offer VIRTIO_BLK_F_FLUSH, and MUST offer it if they offer VIRTIO_BLK_F_CONFIG_WCE.
If VIRTIO_BLK_F_CONFIG_WCE is negotiated but VIRTIO_BLK_F_FLUSH is not, the device MUST initialize writeback to 0.
Because legacy devices do not have FEATURES_OK, transitional devices MUST implement slightly different behavior around feature negotiation when used through the legacy interface. In particular, when using the legacy interface:
The driver queues requests to the virtqueue, and they are used by the device (not necessarily in order). Each request is of form:
The type of the request is either a read (VIRTIO_BLK_T_IN), a write (VIRTIO_BLK_T_OUT), or a flush (VIRTIO_BLK_T_FLUSH).
The sector number indicates the offset (multiplied by 512) where the read or write is to occur. This field is unused and set to 0 for scsi packet commands and for flush commands.
The final status byte is written by the device: either VIRTIO_BLK_S_OK for success, VIRTIO_BLK_S_IOERR for device or driver error or VIRTIO_BLK_S_UNSUPP for a request unsupported by device:
A driver MUST NOT submit a request which would cause a read or write beyond capacity.
A driver SHOULD accept the VIRTIO_BLK_F_RO feature if offered.
A driver MUST set sector to 0 for a VIRTIO_BLK_T_FLUSH request. A driver SHOULD NOT include any data in a VIRTIO_BLK_T_FLUSH request.
If the VIRTIO_BLK_F_CONFIG_WCE feature is negotiated, the driver MAY switch to writethrough or writeback mode by writing respectively 0 and 1 to the writeback field. After writing a 0 to writeback, the driver MUST NOT assume that any volatile writes have been committed to persistent device backend storage.
A device MUST set the status byte to VIRTIO_BLK_S_IOERR for a write request if the VIRTIO_BLK_F_RO feature if offered, and MUST NOT write any data.
A write is considered volatile when it is submitted; the contents of sectors covered by a volatile write are undefined in persistent device backend storage until the write becomes stable. A write becomes stable once it is completed and one or more of the following conditions is true:
If the device is backed by persistent storage, the device MUST ensure that stable writes are committed to it, before reporting completion of the write (cases 1 and 2) or the flush (case 3). Failure to do so can cause data loss in case of a crash.
If the driver changes writeback between the submission of the write and its completion, the write could be either volatile or stable when its completion is reported; in other words, the exact behavior is undefined.
If VIRTIO_BLK_F_FLUSH was not offered by the device11, the device MAY also commit writes to persistent device backend storage before reporting their completion. Unlike case 1, however, this is not an absolute requirement of the specification. Note: An implementation that does not offer VIRTIO_BLK_F_FLUSH and does not commit completed writes will not be resilient to data loss in case of crashes. Not offering VIRTIO_BLK_F_FLUSH is an absolute requirement for implementations that do not wish to be safe against such data losses.
When using the legacy interface, transitional devices and drivers MUST format the fields in struct virtio_blk_req according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
When using the legacy interface, transitional drivers SHOULD ignore the len value in used ring entries. Note: Historically, some devices put the total descriptor length, or the total length of device-writable buffers there, even when only the status byte was actually written.
The reserved field was previously called ioprio. ioprio is a hint about the relative priorities of requests to the device: higher numbers indicate more important requests.
The command VIRTIO_BLK_T_FLUSH_OUT was a synonym for VIRTIO_BLK_T_FLUSH; a driver MUST treat it as a VIRTIO_BLK_T_FLUSH command.
If the device has VIRTIO_BLK_F_BARRIER feature the high bit (VIRTIO_BLK_T_BARRIER) indicates that this request acts as a barrier and that all preceding requests SHOULD be complete before this one, and all following requests SHOULD NOT be started until this is complete. Note: A barrier does not flush caches in the underlying backend device in host, and thus does not serve as data consistency guarantee. Only a VIRTIO_BLK_T_FLUSH request does that.
Some older legacy devices did not commit completed writes to persistent device backend storage when VIRTIO_BLK_F_FLUSH was offered but not negotiated. In order to work around this, the driver MAY set the writeback to 0 (if available) or it MAY send an explicit flush request after every completed write.
If the device has VIRTIO_BLK_F_SCSI feature, it can also support scsi packet command requests, each of these requests is of form:
A request type can also be a scsi packet command (VIRTIO_BLK_T_SCSI_CMD or VIRTIO_BLK_T_SCSI_CMD_OUT). The two types are equivalent, the device does not distinguish between them:
The cmd field is only present for scsi packet command requests, and indicates the command to perform. This field MUST reside in a single, separate device-readable buffer; command length can be derived from the length of this buffer.
Note that these first three (four for scsi packet commands) fields are always device-readable: data is either device-readable or device-writable, depending on the request. The size of the read or write can be derived from the total size of the request buffers.
sense is only present for scsi packet command requests, and indicates the buffer for scsi sense data.
data_len is only present for scsi packet command requests, this field is deprecated, and SHOULD be ignored by the driver. Historically, devices copied data length there.
sense_len is only present for scsi packet command requests and indicates the number of bytes actually written to the sense buffer.
residual field is only present for scsi packet command requests and indicates the residual size, calculated as data length - number of bytes actually transferred.
When using legacy interfaces, transitional drivers which have not negotiated VIRTIO_F_ANY_LAYOUT:
The virtio console device is a simple device for data input and output. A device MAY have one or more ports. Each port has a pair of input and output virtqueues. Moreover, a device has a pair of control IO virtqueues. The control virtqueues are used to communicate information between the device and the driver about ports being opened and closed on either side of the connection, indication from the device about whether a particular port is a console port, adding new ports, port hot-plug/unplug, etc., and indication from the driver about whether a port or a device was successfully added, port open/close, etc. For data IO, one or more empty buffers are placed in the receive queue for incoming data and outgoing characters are placed in the transmit queue.
The port 0 receive and transmit queues always exist: other queues only exist if VIRTIO_CONSOLE_F_MULTIPORT is set.
The size of the console is supplied in the configuration space if the VIRTIO_CONSOLE_F_SIZE feature is set. Furthermore, if the VIRTIO_CONSOLE_F_MULTIPORT feature is set, the maximum number of ports supported by the device can be fetched.
If VIRTIO_CONSOLE_F_EMERG_WRITE is set then the driver can use emergency write to output a single character without initializing virtio queues, or even acknowledging the feature.
When using the legacy interface, transitional devices and drivers MUST format the fields in struct virtio_console_config according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
The device MUST allow a write to emerg_wr, even on an unconfigured device.
The device SHOULD transmit the lower byte written to emerg_wr to an appropriate log or output method.
The driver MUST NOT put a device-readable in a receiveq. The driver MUST NOT put a device-writable buffer in a transmitq.
If the driver negotiated the VIRTIO_CONSOLE_F_MULTIPORT, the two control queues are used to manipulate the different console ports: the control receiveq for messages from the device to the driver, and the control sendq for driver-to-device messages. The layout of the control messages is:
The values for event are:
The device MUST NOT specify a port in VIRTIO_CONSOLE_DEVICE_REMOVE which has not been created with a previous VIRTIO_CONSOLE_DEVICE_ADD.
Upon receipt of a VIRTIO_CONSOLE_CONSOLE_PORT message, the driver SHOULD treat the port in a manner suitable for text console access and MUST respond with a VIRTIO_CONSOLE_PORT_OPEN message, which MUST have value set to 1.
When using the legacy interface, transitional devices and drivers MUST format the fields in struct virtio_console_control according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
When using the legacy interface, the driver SHOULD ignore the len value in used ring entries for the transmit queues and the control transmitq. Note: Historically, some devices put the total descriptor length there, even though no data was actually written.
When using legacy interfaces, transitional drivers which have not negotiated VIRTIO_F_ANY_LAYOUT MUST use only a single descriptor for all buffers in the control receiveq and control transmitq.
The virtio entropy device supplies high-quality randomness for guest use.
None currently defined
None currently defined.
When the driver requires random bytes, it places the descriptor of one or more buffers in the queue. It will be completely filled by random data by the device.
The driver MUST NOT place driver-readable buffers into the queue.
The driver MUST examine the length written by the device to determine how many random bytes were received.
The device MUST place one or more random bytes into the buffer, but it MAY use less than the entire buffer length.
This is the traditional balloon device. The device number 13 is reserved for a new memory balloon interface, with different semantics, which is expected in a future version of the standard.
The traditional virtio memory balloon device is a primitive device for managing guest memory: the device asks for a certain amount of memory, and the driver supplies it (or withdraws it, if the device has more than it asks for). This allows the guest to adapt to changes in allowance of underlying physical memory. If the feature is negotiated, the device can also be used to communicate guest memory statistics to the host.
Virtqueue 2 only exists if VIRTIO_BALLON_F_STATS_VQ set.
The driver SHOULD accept the VIRTIO_BALLOON_F_MUST_TELL_HOST feature if offered by the device.
If the device offers the VIRTIO_BALLOON_F_MUST_TELL_HOST feature bit, and if the driver did not accept this feature bit, the device MAY signal failure by failing to set FEATURES_OK device status bit when the driver writes it.
Both fields of this configuration are always available.
The device initialization process is outlined below:
The device is driven either by the receipt of a configuration change interrupt, or by changing guest memory needs, such as performing memory compaction or responding to out of memory conditions.
The driver SHOULD supply pages to the balloon when num_pages is greater than the actual number of pages in the balloon.
The driver MAY use pages from the balloon when num_pages is less than the actual number of pages in the balloon.
The driver MAY supply pages to the balloon when num_pages is greater than or equal to the actual number of pages in the balloon.
If VIRTIO_BALLOON_F_DEFLATE_ON_OOM has not been negotiated, the driver MUST NOT use pages from the balloon when num_pages is less than or equal to the actual number of pages in the balloon.
If VIRTIO_BALLOON_F_DEFLATE_ON_OOM has been negotiated, the driver MAY use pages from the balloon when num_pages is less than or equal to the actual number of pages in the balloon if this is required for system stability (e.g. if memory is required by applications running within the guest).
The driver MUST use the deflateq to inform the device of pages that it wants to use from the balloon.
If the VIRTIO_BALLOON_F_MUST_TELL_HOST feature is negotiated, the driver MUST NOT use pages from the balloon until the device has acknowledged the deflate request.
Otherwise, if the VIRTIO_BALLOON_F_MUST_TELL_HOST feature is not negotiated, the driver MAY begin to re-use pages previously given to the balloon before the device has acknowledged the deflate request.
In any case, the driver MUST NOT use pages from the balloon after adding the pages to the balloon, but before the device has acknowledged the inflate request.
The driver MUST NOT request deflation of pages in the balloon before the device has acknowledged the inflate request.
The driver MUST update actual after changing the number of pages in the balloon.
The driver MAY update actual once after multiple inflate and deflate operations.
The device MAY modify the contents of a page in the balloon after detecting its physical number in an inflate request and before acknowledging the inflate request by using the inflateq descriptor.
If the VIRTIO_BALLOON_F_MUST_TELL_HOST feature is negotiated, the device MAY modify the contents of a page in the balloon after detecting its physical number in an inflate request and before detecting its physical number in a deflate request and acknowledging the deflate request.
When using the legacy interface, the driver MUST write out all 4 bytes each time it updates the actual value in the configuration space, using a single atomic operation.
When using the legacy interface, the device SHOULD NOT use the actual value written by the driver in the configuration space, until the last, most-significant byte of the value has been written. Note: Historically, devices used the actual value, even though when using Virtio Over PCI Bus the device-specific configuration space was not guaranteed to be atomic. Using intermediate values during update by driver is best avoided, except for debugging.
Historically, drivers using Virtio Over PCI Bus wrote the actual value by using multiple single-byte writes in order, from the least-significant to the most-significant value.
The stats virtqueue is atypical because communication is driven by the device (not the driver). The channel becomes active at driver initialization time when the driver adds an empty buffer and notifies the device. A request for memory statistics proceeds as follows:
Within the buffer, statistics are an array of 6-byte entries. Each statistic consists of a 16 bit tag and a 64 bit value. All statistics are optional and the driver chooses which ones to supply. To guarantee backwards compatibility, devices omit unsupported statistics.
The driver MUST make at most one buffer available to the device in the statsq, at all times.
After initializing the device, the driver MUST make an output buffer available in the statsq.
Upon detecting that device has used a buffer in the statsq, the driver MUST make an output buffer available in the statsq.
Before making an output buffer available in the statsq, the driver MUST initialize it, including one struct virtio_balloon_stat entry for each statistic that it supports.
Driver MUST use an output buffer size which is a multiple of 6 bytes for all buffers submitted to the statsq.
Driver MAY supply struct virtio_balloon_stat entries in the output buffer submitted to the statsq in any order, without regard to tag values.
Driver MAY supply a subset of all statistics in the output buffer submitted to the statsq.
Driver MUST supply the same subset of statistics in all buffers submitted to the statsq.
Within an output buffer submitted to the statsq, the device MUST ignore entries with tag values that it does not recognize.
Within an output buffer submitted to the statsq, the device MUST accept struct virtio_balloon_stat entries in any order without regard to tag values.
When using the legacy interface, the device SHOULD ignore all values in the first buffer in the statsq supplied by the driver after device initialization. Note: Historically, drivers supplied an uninitialized buffer in the first buffer.
The virtio SCSI host device groups together one or more virtual logical units (such as disks), and allows communicating to them using the SCSI protocol. An instance of the device represents a SCSI host to which many targets and LUNs are attached.
The virtio SCSI device services two kinds of requests:
The device is also able to send out notifications about added and removed logical units. Together, these capabilities provide a SCSI transport protocol that uses virtqueues as the transfer medium. In the transport protocol, the virtio driver acts as the initiator, while the virtio SCSI host provides one or more targets that receive and process the requests.
This section relies on definitions from SAM.
All fields of this configuration are always available.
The driver MUST NOT write to device configuration fields other than sense_size and cdb_size.
The driver MUST NOT send more than cmd_per_lun linked commands to one LUN, and MUST NOT send more than the virtqueue size number of linked commands to one LUN.
On reset, the device MUST set sense_size to 96 and cdb_size to 32.
When using the legacy interface, transitional devices and drivers MUST format the fields in struct virtio_scsi_config according to the native endian of the guest rather than (necessarily when not using the legacy interface) little-endian.
On initialization the driver SHOULD first discover the device’s virtqueues.
If the driver uses the eventq, the driver SHOULD place at least one buffer in the eventq.
The driver MAY immediately issue requests15 or task management functions16.
Device operation consists of operating request queues, the control queue and the event queue.
The driver queues requests to an arbitrary request queue, and they are used by the device on that same queue. It is the responsibility of the driver to ensure strict request ordering for commands placed on different queues, because they will be consumed with no order constraints.
Requests have the following format:
lun addresses the REPORT LUNS well-known logical unit, or a target and logical unit in the virtio-scsi device’s SCSI domain. When used to address the REPORT LUNS logical unit, lun is 0xC1, 0x01 and six zero bytes. The virtio-scsi device SHOULD implement the REPORT LUNS well-known logical unit.
When used to address a target and logical unit, the only supported format for lun is: first byte set to 1, second byte set to target, third and fourth byte representing a single level LUN structure, followed by four zero bytes. With this representation, a virtio-scsi device can serve up to 256 targets and 16384 LUNs per target. The device MAY also support having a well-known logical units in the third and fourth byte.
id is the command identifier (“tag”).
task_attr defines the task attribute as in the table above, but all task attributes MAY be mapped to SIMPLE by the device. Some commands are defined by SCSI standards as "implicit head of queue"; for such commands, all task attributes MAY also be mapped to HEAD OF QUEUE. Drivers and applications SHOULD NOT send a command with the ORDERED task attribute if the command has an implicit HEAD OF QUEUE attribute, because whether the ORDERED task attribute is honored is vendor-specific.
crn may also be provided by clients, but is generally expected to be 0. The maximum CRN value defined by the protocol is 255, since CRN is stored in an 8-bit integer.
The CDB is included in cdb and its size, cdb_size, is taken from the configuration space.
All of these fields are defined in SAM and are always device-readable.
pi_bytesout determines the size of the pi_out field in bytes. If it is nonzero, the pi_out field contains outgoing protection information for write operations. pi_bytesin determines the size of the pi_in field in the device-writable section, in bytes. All three fields are only present if VIRTIO_SCSI_F_T10_PI has been negotiated.
The remainder of the device-readable part is the data output buffer, dataout.
sense and subsequent fields are always device-writable. sense_len indicates the number of bytes actually written to the sense buffer.
residual indicates the residual size, calculated as “data_length - number_of_transferred_bytes”, for read or write operations. For bidirectional commands, the number_of_transferred_bytes includes both read and written bytes. A residual that is less than the size of datain means that dataout was processed entirely. A residual that exceeds the size of datain means that dataout was processed partially and datain was not processed at all.
If the pi_bytesin is nonzero, the pi_in field contains incoming protection information for read operations. pi_in is only present if VIRTIO_SCSI_F_T10_PI has been negotiated17.
The remainder of the device-writable part is the data input buffer, datain.
The device MUST write the response byte as one of the following:
All commands must be completed before the virtio-scsi device is reset or unplugged. The device MAY choose to abort them, or if it does not do so MUST pick the VIRTIO_SCSI_S_FAILURE response.
Upon receiving a VIRTIO_SCSI_S_TARGET_FAILURE response, the driver SHOULD NOT retry the request on other paths.
The controlq is used for other SCSI transport operations. Requests have the following format:
The type identifies the remaining fields.
The following commands are defined:
The type is VIRTIO_SCSI_T_TMF; subtype defines which task management function. All fields except response are filled by the driver.
Other fields which are irrelevant for the requested TMF are ignored but they are still present. lun is in the same format specified for request queues; the single level LUN is ignored when the task management function addresses a whole I_T nexus. When relevant, the value of id is matched against the id values passed on the requestq.
The outcome of the task management function is written by the device in response. The command-specific response values map 1-to-1 with those defined in SAM.
Task management function can affect the response value for commands that are in the request queue and have not been completed yet. For example, the device MUST complete all active commands on a logical unit or target (possibly with a VIRTIO_SCSI_S_RESET response code) upon receiving a "logical unit reset" or "I_T nexus reset" TMF. Similarly, the device MUST complete the selected commands (possibly with a VIRTIO_SCSI_S_ABORTED response code) upon receiving an "abort task" or "abort task set" TMF. Such effects MUST take place before the TMF itself is successfully completed, and the device MUST use memory barriers appropriately in order to ensure that the driver sees these writes in the correct order.
By sending this command, the driver asks the device which events the given LUN can report, as described in paragraphs 6.6 and A.6 of SCSI MMC. The driver writes the events it is interested in into event_requested; the device responds by writing the events that it supports into event_actual.
The type is VIRTIO_SCSI_T_AN_QUERY. lun and event_requested are written by the driver. event_actual and response fields are written by the device.
No command-specific values are defined for the response byte.
By sending this command, the driver asks the specified LUN to report events for its physical interface, again as described in SCSI MMC. The driver writes the events it is interested in into event_requested; the device responds by writing the events that it supports into event_actual.
Event types are the same as for the asynchronous notification query message.
The type is VIRTIO_SCSI_T_AN_SUBSCRIBE. lun and event_requested are written by the driver. event_actual and response are written by the device.
No command-specific values are defined for the response byte.
The eventq is populated by the driver for the device to report information on logical units that are attached to it. In general, the device will not queue events to cope with an empty eventq, and will end up dropping events if it finds no buffer ready. However, when reporting events for many LUNs (e.g. when a whole target disappears), the device can throttle events to avoid dropping them. For this reason, placing 10-15 buffers on the event queue is sufficient.
Buffers returned by the device on the eventq will be referred to as “events” in the rest of this section. Events have the following format:
The devices sets bit 31 in event to report lost events due to missing buffers.
The meaning of reason depends on the contents of event. The following events are defined:
This event is fired in the following cases:
By sending this event, the device signals that a logical unit on a target has been reset, including the case of a new device appearing or disappearing on the bus. The device fills in all fields. event is set to VIRTIO_SCSI_T_TRANSPORT_RESET. lun addresses a logical unit in the SCSI host.
The reason value is one of the three #define values appearing above:
The “removed” and “rescan” events can happen when VIRTIO_SCSI_F_HOTPLUG feature was negotiated; when sent for LUN 0, they MAY apply to the entire target so the driver can ask the initiator to rescan the target to detect this.
Events will also be reported via sense codes (this obviously does not apply to newly appeared buses or targets, since the application has never discovered them):
The preferred way to detect transport reset is always to use events, because sense codes are only seen by the driver when it sends a SCSI command to the logical unit or target. However, in case events are dropped, the initiator will still be able to synchronize with the actual state of the controller if the driver asks the initiator to rescan of the SCSI bus. During the rescan, the initiator will be able to observe the above sense codes, and it will process them as if it the driver had received the equivalent event.
By sending this event, the device signals that an asynchronous event was fired from a physical interface.
All fields are written by the device. event is set to VIRTIO_SCSI_T_ASYNC_NOTIFY. lun addresses a logical unit in the SCSI host. reason is a subset of the events that the driver has subscribed to via the “Asynchronous notification subscription” command.
By sending this event, the device signals a change in the configuration parameters of a logical unit, for example the capacity or cache mode. event is set to VIRTIO_SCSI_T_PARAM_CHANGE. lun addresses a logical unit in the SCSI host.
The same event SHOULD also be reported as a unit attention condition. reason contains the additional sense code and additional sense code qualifier, respectively in bits 0…7 and 8…15. Note: For example, a change in capacity will be reported as asc 0x2a, ascq 0x09 (CAPACITY DATA HAS CHANGED).
For MMC devices (inquiry type 5) there would be some overlap between this event and the asynchronous notification event, so for simplicity the host never reports this event for MMC devices.
If event has bit 31 set, the driver SHOULD poll the logical units for unit attention conditions, and/or do whatever form of bus scan is appropriate for the guest operating system and SHOULD poll for asynchronous events manually using SCSI commands.
When receiving a VIRTIO_SCSI_T_TRANSPORT_RESET message with reason set to VIRTIO_SCSI_EVT_RESET_REMOVED or VIRTIO_SCSI_EVT_RESET_RESCAN for LUN 0, the driver SHOULD ask the initiator to rescan the target, in order to detect the case when an entire target has appeared or disappeared.
The device MUST NOT send VIRTIO_SCSI_T_TRANSPORT_RESET messages with reason set to VIRTIO_SCSI_EVT_RESET_REMOVED or VIRTIO_SCSI_EVT_RESET_RESCAN unless VIRTIO_SCSI_F_HOTPLUG was negotiated.
The device MUST NOT report VIRTIO_SCSI_T_PARAM_CHANGE for MMC devices.
When using legacy interfaces, transitional drivers which have not negotiated VIRTIO_F_ANY_LAYOUT MUST use a single descriptor for the lun, id, task_attr, prio, crn and cdb fields, and MUST only use a single descriptor for the sense_len, residual, status_qualifier, status, response and sense fields.
The vhost-user device backend facilitates vhost-user device emulation through vhost-user protocol exchanges and access to shared memory. Software-defined networking, storage, and other I/O appliances can provide services through this device.
This section relies on definitions from the Vhost-user Protocol. Knowledge of the vhost-user protocol is a prerequisite for understanding this device.
The Vhost-user Protocol was originally designed for processes on a single system communicating over UNIX domain sockets. The virtio vhost-user device backend allows the vhost-user slave to communicate with the vhost-user master over the device instead of a UNIX domain socket. This allows the slave and master to run on two separate systems such as a virtual machine and a hypervisor.
The vhost-user slave program exchanges vhost-user protocol messages with the vhost-user master through this device. How the device implementation communicates with the vhost-user master is beyond the scope of this specification. One possible device implementation uses a UNIX domain socket to relay messages to a vhost-user master process running on the same host.
Existing vhost-user slave programs that communicate over UNIX domain sockets can support the virtio vhost-user device backend without invasive changes because the pre-existing vhost-user wire protocol is used.
No feature bits are defined at this time.
All fields of this configuration are always available.
The driver sets VIRTIO_VHOST_USER_STATUS_SLAVE_UP to indicate readiness for the vhost-user master to connect. The vhost-user master cannot connect unless the driver has set this bit first.
When the driver clears VIRTIO_VHOST_USER_SLAVE_UP while the vhost-user master is connected, the vhost-user master is disconnected.
When the vhost-user master disconnects, both VIRTIO_VHOST_USER_STATUS_SLAVE_UP and VIRTIO_VHOST_USER_STATUS_MASTER_UP are cleared by the device. Communication can be restarted by the driver setting VIRTIO_VHOST_USER_STATUS_SLAVE_UP again.
A configuration change notification is sent when the device changes this field unless a write to the field by the driver caused the change.
The driver MUST NOT write to device configuration fields other than status.
The driver MUST NOT set undefined bits in the status configuration field.
The driver SHOULD check the max_vhost_queues configuration field to determine how many queues the vhost-user slave will be able to support.
The driver SHOULD fetch the uuid configuration field to allow vhost-user slave programs to identify a specific device among many.
The driver SHOULD place at least one buffer in rxq before setting the VIRTIO_VHOST_USER_SLAVE_UP bit in the status configuration field.
The driver MUST handle rxq virtqueue notifications that occur before the configuration change notification. It is possible that a vhost-user protocol message from the vhost-user master arrives before the driver has seen the configuration change notification for the VIRTIO_VHOST_USER_STATUS_MASTER_UP status change.
Device operation consists of operating request queues and response queues.
The driver receives vhost-user protocol messages from the vhost-user master on rxq. The driver sends responses to the vhost-user master on txq.
The driver sends slave-initiated requests on txq. The driver receives responses from the vhost-user master on rxq.
All virtqueues offer in-order guaranteed delivery semantics for vhost-user protocol messages.
Each buffer is a vhost-user protocol message as defined by the Vhost-user Protocol. In order to enable cross-endian communication, all message fields are little-endian instead of the native byte order normally used by the protocol.
The appropriate size of rxq buffers is at least as large as the largest message defined by the Vhost-user Protocol standard version that the driver supports. If the vhost-user master sends a message that is too large for an rxq buffer then DEVICE_NEEDS_RESET is set and the driver must reset the device.
File descriptor passing is handled differently by the vhost-user device backend. When a message is received that carries one or more file descriptors according to the vhost-user protocol, additional device resources become available to the driver.
The vhost-user device backend contains additional device resources beyond configuration space and virtqueues. The nature of these resources is transport-specific and therefore only virtio transports that provide these resources support the vhost-user device backend.
The following additional resources exist:
Doorbells are laid out as follows:
Notifications are laid out as follows:
Shared memory is laid out as follows:
The size of vhost memory region 0 is SIZE0, the size of vhost memory region 1 is SIZE1, and so on.
The following vhost-user protocol messages convey access to additional device resources:
Additional resources are configured on the virtio PCI transport by the following struct virtio_pci_cap.cfg_type values:
The doorbell location is found using the VIRTIO_PCI_CAP_DOORBELL_CFG capability. This capability is immediately followed by an additional field, like so:
The doorbell address within a BAR is calculated as follows:
The cap.offset and doorbell_off_multiplier are taken from the notification capability structure above, and the doorbell_idx is the doorbell number.
The cap.offset MUST be 2-byte aligned.
The device MUST either present doorbell_off_multiplier as an even power of 2, or present doorbell_off_multiplier as 0.
The value cap.length presented by the device MUST be at least 2 and MUST be large enough to support doorbell offsets for all supported doorbells in all possible configurations.
The value cap.length presented by the device MUST satisfy:
The number of doorbells is num_doorbells and is dependent on the device.
The notification structure allows MSI-X vectors to be configured for notification interrupts. If MSI-X is not available, bit 2 of the ISR status indicates that a notification occurred.
The notification structure is found using the VIRTIO_PCI_CAP_DOORBELL_CFG capability.
The driver indicates which notification is of interest by writing the notification_select field. The driver then writes the MSI-X vector or VIRTIO_MSI_NO_VECTOR to notification_msix_vector to change the MSI-X vector for that notification.
The shared memory location is found using the VIRTIO_PCI_CAP_SHARED_MEMORY_CFG capability.
The device MUST locate shared memory in a Memory Space BAR.
The device SHOULD locate shared memory in a Prefetchable BAR.
The cap.offset MUST be 4096-byte aligned.
The value cap.length presented by the device MUST be non-zero and 4096-byte aligned.
A driver MUST accept VIRTIO_F_VERSION_1 if it is offered. A driver MAY fail to operate further if VIRTIO_F_VERSION_1 is not offered.
A driver SHOULD accept VIRTIO_F_IOMMU_PLATFORM if it is offered, and it MUST then either disable the IOMMU or configure the IOMMU to translate bus addresses passed to the device into physical addresses in memory. If VIRTIO_F_IOMMU_PLATFORM is not offered, then a driver MUST pass only physical addresses to the device.
A device MUST offer VIRTIO_F_VERSION_1. A device MAY fail to operate further if VIRTIO_F_VERSION_1 is not accepted.
A device SHOULD offer VIRTIO_F_IOMMU_PLATFORM if it is behind an IOMMU that translates bus addresses from the device into physical addresses in memory. A device MAY fail to operate further if VIRTIO_F_IOMMU_PLATFORM is not accepted.
Transitional devices MAY offer the following:
Transitional devices MUST offer, and if offered by the device transitional drivers MUST accept the following:
A driver MUST conform to the following normative statements:
A PCI driver MUST conform to the following normative statements:
An MMIO driver MUST conform to the following normative statements:
A Channel I/O driver MUST conform to the following normative statements:
A network driver MUST conform to the following normative statements:
A block driver MUST conform to the following normative statements:
A console driver MUST conform to the following normative statements:
An entropy driver MUST conform to the following normative statements:
A traditional memory balloon driver MUST conform to the following normative statements:
An SCSI host driver MUST conform to the following normative statements:
A device MUST conform to the following normative statements:
A PCI device MUST conform to the following normative statements:
An MMIO device MUST conform to the following normative statements:
A Channel I/O device MUST conform to the following normative statements:
A network device MUST conform to the following normative statements:
A block device MUST conform to the following normative statements:
A console device MUST conform to the following normative statements:
An entropy device MUST conform to the following normative statements:
A traditional memory balloon device MUST conform to the following normative statements:
An SCSI host device MUST conform to the following normative statements:
A conformant implementation MUST be either transitional or non-transitional, see 1.3.1.
A non-transitional implementation conforms to this specification if it satisfies all of the MUST or REQUIRED level requirements defined above.
An implementation MAY choose to implement OPTIONAL support for the legacy interface, including support for legacy drivers or devices, by additionally conforming to all of the MUST or REQUIRED level requirements for the legacy interface for the transitional devices and drivers.
The requirements for the legacy interface for transitional implementations are located in sections named “Legacy Interface” listed below:
It is possible that a very simple device will operate entirely through its device configuration space, but most will need at least one virtqueue in which it will place requests. A device with both input and output (eg. console and network devices described here) need two queues: one which the driver fills with buffers to receive input, and one which the driver places buffers to transmit output.
Device configuration space should only be used for initialization-time parameters. It is a limited resource with no synchronization between field written by the driver, so for most uses it is better to use a virtqueue to update configuration information (the network device does this for filtering, otherwise the table in the config space could potentially be very large).
Remember that configuration fields over 32 bits wide might not be atomically writable by the driver. Therefore, no writeable field which triggers an action ought to be wider than 32 bits.
Device numbers can be reserved by the OASIS committee: email email@example.com to secure a unique one.
Meanwhile for experimental drivers, use 65535 and work backwards.
Using the optional MSI-X capability devices can speed up interrupt processing by removing the need to read ISR Status register by guest driver (which might be an expensive operation), reducing interrupt sharing between devices and queues within the device, and handling interrupts from multiple CPUs. However, some systems impose a limit (which might be as low as 256) on the total number of MSI-X vectors that can be allocated to all devices. Devices and/or drivers should take this into account, limiting the number of vectors used unless the device is expected to cause a high volume of interrupts. Devices can control the number of vectors used by limiting the MSI-X Table Size or not presenting MSI-X capability in PCI configuration space. Drivers can control this by mapping events to as small number of vectors as possible, or disabling MSI-X capability altogether.
Any change to device configuration space, or new virtqueues, or behavioural changes, should be indicated by negotiation of a new feature bit. This establishes clarity18 and avoids future expansion problems.
Clusters of functionality which are always implemented together can use a single bit, but if one feature makes sense without the others they should not be gratuitously grouped together to conserve feature bits.
The following non-members have provided valuable feedback on this specification and are gratefully acknowledged:
|540||11 Oct 2015||Greg Kurz||
virtqueues: fix trivial typo
|541||11 Oct 2015||Paolo Bonzini||
virtio-blk: fix typo in legacy framing requirements section
|545||18 Oct 2015||Paolo Bonzini||
virtio-blk: restore VIRTIO_BLK_F_FLUSH and VIRTIO_BLK_F_CONFIG_WCE
VIRTIO_BLK_F_CONFIG_WCE is important in order to achieve good performance (up to 2x, though more realistically +30-40%) in latency-bound workloads. However, it was removed by mistake together with VIRTIO_BLK_F_FLUSH.
In addition, even removing VIRTIO_BLK_F_FLUSH was probably not a great idea, because it simplifies simple drivers (e.g. firmware) that are okay with a writethrough cache but still need data to persist after power loss. What really should have been removed is just the possibility that devices not propose VIRTIO_BLK_F_FLUSH, but even that only deserves a "SHOULD" in the new world of conformance statements.
Restore these, with the following changes:
* clarify and use conformance statements in order to define writeback and writethrough caching according to what is commonly done by high-end storage.
* clarify (with conformance statements) the influence of the VIRTIO_BLK_F_FLUSH feature on caching and how to proceed if only one of VIRTIO_BLK_F_FLUSH and VIRTIO_BLK_F_CONFIG_WCE is negotiated.
* strengthen the requirement for persisting writes to MUST after a VIRTIO_BLK_T_FLUSH request (and in other cases too involving the new features).
The suggested behavior upon feature negotiation is okay for the Linux implementation of virtio1, even after the implementation is modified to support the two new features.
This fixes VIRTIO-144.
|546||18 Oct 2015||Michael S. Tsirkin||
pci: clarify configuration access capability rules
The point of the configuration access capability is to enable access to other capabilities. The intent never was to allow writes to a random place within device BARs. Limiting drivers simplifies devices - and devices can always add another capability if drivers ever want to access some other range.
This resolves VIRTIO-145.
|547||18 Oct 2015||Michael S. Tsirkin||
add advice on transition from legacy interfaces
Reading legacy chapters gives a hint about what changed, let’s help readers discover this useful shortcut.
This resolves VIRTIO-146.
|554||16 Feb 2016||Thomas Huth||
virtio-net: fix inconsistent legacy header size
Current text says: The legacy driver only presented num_buffers in the struct virtio_net_hdr when VIRTIO_NET_F_MRG_RXBUF was not negotiated;
Should be: "…was negotiated …" instead of "…was not negotiated …"
To be consistent with the following: without that feature the structure was 2 bytes shorter.
|555||16 Feb 2016||Michael S. Tsirkin||
virtio header: tweak change motivation
The changes are not just to remove Linux assumptions, we have also renamed ring->queue. Tweak the header description accordingly.
|558||16 Feb 2016||Michael S. Tsirkin||
rename virtio_ring.h to virtio_queue.h
Since vring* and VRING* have been replaced with virtq* and VIRTQ* respectively, rename the header virtio_ring.h to virtio_queue.h.
|559||16 Feb 2016||Michael S. Tsirkin||
init: sort status bits
Status bit order is inconsistent: they are neither in increasing order nor in the order they are likely to be used.
The second approach seems more useful since there aren’t that many bits, so the numerical order does not help much.
A typical order of use would be:
Sort the bits accordingly.
1This lack of page-sharing implies that the implementation of the device (e.g. the hypervisor or host) needs full access to the guest memory. Communication with untrusted parties (i.e. inter-guest communication) requires copying.
2The Linux implementation further separates the virtio transport code from the specific virtio drivers: these drivers are shared between different transports.
3For example, the simplest network device has one virtqueue for transmit and one for receive.
4For example, if Queue Size is 4 then at most 4 buffers can be queued at any given time.
5For example, the simplest network device has two virtqueues.
6The 4096 is based on the x86 page size, but it’s also large enough to ensure that the separate parts of the virtqueue are on separate cache lines.
7Due to various bugs in implementations, this field is not useful as a guarantee of the transport header size.
8This case is not handled by some older hardware, so is called out specifically in the protocol.
9Since there are no guarantees, it can use a hash filter or silently switch to allmulti or promiscuous mode if it is given too many addresses.
10Consistent with 18.104.22.168, a writethrough cache can be defined broadly as a cache that commits writes to persistent device backend storage before reporting their completion. For example, a battery-backed writeback cache actually counts as writethrough according to this definition.
12Because this is high importance and low bandwidth, the current Linux implementation polls for the buffer to be used, rather than waiting for an interrupt, simplifying the implementation significantly. However, for generic serial ports with the O_NONBLOCK flag set, the polling limitation is relaxed and the consumed buffers are freed upon the next write or poll call or when a port is closed or hot-unplugged.
13This is historical, and independent of the guest page size.
14In this case, deflation advice is merely a courtesy.
15For example, INQUIRY or REPORT LUNS.
16For example, I_T RESET.
17There is no separate residual size for pi_bytesout and pi_bytesin. It can be computed from the residual field, the size of the data integrity information per sector, and the sizes of pi_out, pi_in, dataout and datain.
18Even if it does mean documenting design or implementation mistakes!