kata-containers/src/runtime/pkg/govmm/qemu/qmp.go
Fabiano Fidêncio 8bcaed0b4f govmm: Adapt license headers to kata-containers
Both projects follow the same license, Apache-2.0, but the header saying
that comes from govmm is different from the one expected for the tests
present on the kata-containers repo.

Signed-off-by: Fabiano Fidêncio <fabiano.fidencio@intel.com>
2022-01-19 18:02:46 +01:00

1664 lines
51 KiB
Go

// Copyright contributors to the Virtual Machine Manager for Go project
//
// SPDX-License-Identifier: Apache-2.0
//
package qemu
import (
"bufio"
"container/list"
"encoding/json"
"errors"
"fmt"
"io"
"net"
"os"
"strconv"
"syscall"
"time"
"context"
"strings"
)
// QMPLog is a logging interface used by the qemu package to log various
// interesting pieces of information. Rather than introduce a dependency
// on a given logging package, qemu presents this interface that allows
// clients to provide their own logging type which they can use to
// seamlessly integrate qemu's logs into their own logs. A QMPLog
// implementation can be specified in the QMPConfig structure.
type QMPLog interface {
// V returns true if the given argument is less than or equal
// to the implementation's defined verbosity level.
V(int32) bool
// Infof writes informational output to the log. A newline will be
// added to the output if one is not provided.
Infof(string, ...interface{})
// Warningf writes warning output to the log. A newline will be
// added to the output if one is not provided.
Warningf(string, ...interface{})
// Errorf writes error output to the log. A newline will be
// added to the output if one is not provided.
Errorf(string, ...interface{})
}
type qmpNullLogger struct{}
func (l qmpNullLogger) V(level int32) bool {
return false
}
func (l qmpNullLogger) Infof(format string, v ...interface{}) {
}
func (l qmpNullLogger) Warningf(format string, v ...interface{}) {
}
func (l qmpNullLogger) Errorf(format string, v ...interface{}) {
}
// QMPConfig is a configuration structure that can be used to specify a
// logger and a channel to which logs and QMP events are to be sent. If
// neither of these fields are specified, or are set to nil, no logs will be
// written and no QMP events will be reported to the client.
type QMPConfig struct {
// eventCh can be specified by clients who wish to receive QMP
// events.
EventCh chan<- QMPEvent
// logger is used by the qmpStart function and all the go routines
// it spawns to log information.
Logger QMPLog
// specify the capacity of buffer used by receive QMP response.
MaxCapacity int
}
type qmpEventFilter struct {
eventName string
dataKey string
dataValue string
}
// QMPEvent contains a single QMP event, sent on the QMPConfig.EventCh channel.
// nolint: govet
type QMPEvent struct {
// The name of the event, e.g., DEVICE_DELETED
Name string
// The data associated with the event. The contents of this map are
// unprocessed by the qemu package. It is simply the result of
// unmarshalling the QMP json event. Here's an example map
// map[string]interface{}{
// "driver": "virtio-blk-pci",
// "drive": "drive_3437843748734873483",
// }
Data map[string]interface{}
// The event's timestamp converted to a time.Time object.
Timestamp time.Time
}
type qmpResult struct {
response interface{}
err error
}
// nolint: govet
type qmpCommand struct {
ctx context.Context
res chan qmpResult
name string
args map[string]interface{}
filter *qmpEventFilter
resultReceived bool
oob []byte
}
// QMP is a structure that contains the internal state used by startQMPLoop and
// the go routines it spwans. All the contents of this structure are private.
// nolint: govet
type QMP struct {
cmdCh chan qmpCommand
conn io.ReadWriteCloser
cfg QMPConfig
connectedCh chan<- *QMPVersion
disconnectedCh chan struct{}
version *QMPVersion
}
// QMPVersion contains the version number and the capabailities of a QEMU
// instance, as reported in the QMP greeting message.
// nolint: govet
type QMPVersion struct {
Major int
Minor int
Micro int
Capabilities []string
}
// CPUProperties contains the properties of a CPU instance
type CPUProperties struct {
Node int `json:"node-id"`
Socket int `json:"socket-id"`
Die int `json:"die-id"`
Core int `json:"core-id"`
Thread int `json:"thread-id"`
}
// HotpluggableCPU represents a hotpluggable CPU
// nolint: govet
type HotpluggableCPU struct {
Type string `json:"type"`
VcpusCount int `json:"vcpus-count"`
Properties CPUProperties `json:"props"`
QOMPath string `json:"qom-path"`
}
// MemoryDevicesData cotains the data describes a memory device
// nolint: govet
type MemoryDevicesData struct {
Slot int `json:"slot"`
Node int `json:"node"`
Addr uint64 `json:"addr"`
Memdev string `json:"memdev"`
ID string `json:"id"`
Hotpluggable bool `json:"hotpluggable"`
Hotplugged bool `json:"hotplugged"`
Size uint64 `json:"size"`
}
// MemoryDevices represents memory devices of vm
// nolint: govet
type MemoryDevices struct {
Data MemoryDevicesData `json:"data"`
Type string `json:"type"`
}
// CPUInfo represents information about each virtual CPU
// nolint: govet
type CPUInfo struct {
CPU int `json:"CPU"`
Current bool `json:"current"`
Halted bool `json:"halted"`
QomPath string `json:"qom_path"`
Arch string `json:"arch"`
Pc int `json:"pc"`
ThreadID int `json:"thread_id"`
Props CPUProperties `json:"props"`
}
// CPUInfoFast represents information about each virtual CPU
// nolint: govet
type CPUInfoFast struct {
CPUIndex int `json:"cpu-index"`
QomPath string `json:"qom-path"`
Arch string `json:"arch"`
ThreadID int `json:"thread-id"`
Target string `json:"target"`
Props CPUProperties `json:"props"`
}
// MigrationRAM represents migration ram status
type MigrationRAM struct {
Total int64 `json:"total"`
Remaining int64 `json:"remaining"`
Transferred int64 `json:"transferred"`
TotalTime int64 `json:"total-time"`
SetupTime int64 `json:"setup-time"`
ExpectedDowntime int64 `json:"expected-downtime"`
Duplicate int64 `json:"duplicate"`
Normal int64 `json:"normal"`
NormalBytes int64 `json:"normal-bytes"`
DirtySyncCount int64 `json:"dirty-sync-count"`
}
// MigrationDisk represents migration disk status
type MigrationDisk struct {
Total int64 `json:"total"`
Remaining int64 `json:"remaining"`
Transferred int64 `json:"transferred"`
}
// MigrationXbzrleCache represents migration XbzrleCache status
type MigrationXbzrleCache struct {
CacheSize int64 `json:"cache-size"`
Bytes int64 `json:"bytes"`
Pages int64 `json:"pages"`
CacheMiss int64 `json:"cache-miss"`
CacheMissRate int64 `json:"cache-miss-rate"`
Overflow int64 `json:"overflow"`
}
// MigrationStatus represents migration status of a vm
type MigrationStatus struct {
Status string `json:"status"`
Capabilities []map[string]interface{} `json:"capabilities,omitempty"`
RAM MigrationRAM `json:"ram,omitempty"`
Disk MigrationDisk `json:"disk,omitempty"`
XbzrleCache MigrationXbzrleCache `json:"xbzrle-cache,omitempty"`
}
// SchemaInfo represents all QMP wire ABI
type SchemaInfo struct {
MetaType string `json:"meta-type"`
Name string `json:"name"`
}
// StatusInfo represents guest running status
// nolint: govet
type StatusInfo struct {
Running bool `json:"running"`
SingleStep bool `json:"singlestep"`
Status string `json:"status"`
}
func (q *QMP) readLoop(fromVMCh chan<- []byte) {
scanner := bufio.NewScanner(q.conn)
if q.cfg.MaxCapacity > 0 {
buffer := make([]byte, q.cfg.MaxCapacity)
scanner.Buffer(buffer, q.cfg.MaxCapacity)
}
for scanner.Scan() {
line := scanner.Bytes()
// Since []byte channel type transfer slice info(include slice underlying array pointer, len, cap)
// between channel sender and receiver. scanner.Bytes() returned slice's underlying array
// may point to data that will be overwritten by a subsequent call to Scan(reference from:
// https://golang.org/pkg/bufio/#Scanner.Bytes), which may make receiver read mixed data,
// so we need to copy line to new allocated space and then send to channel receiver
sendLine := make([]byte, len(line))
copy(sendLine, line)
fromVMCh <- sendLine
}
q.cfg.Logger.Infof("scanner return error: %v", scanner.Err())
close(fromVMCh)
}
func (q *QMP) processQMPEvent(cmdQueue *list.List, name interface{}, data interface{},
timestamp interface{}) {
strname, ok := name.(string)
if !ok {
return
}
var eventData map[string]interface{}
if data != nil {
eventData, _ = data.(map[string]interface{})
}
cmdEl := cmdQueue.Front()
if cmdEl != nil {
cmd := cmdEl.Value.(*qmpCommand)
filter := cmd.filter
if filter != nil {
if filter.eventName == strname {
match := filter.dataKey == ""
if !match && eventData != nil {
match = eventData[filter.dataKey] == filter.dataValue
}
if match {
if cmd.resultReceived {
q.finaliseCommand(cmdEl, cmdQueue, true)
} else {
cmd.filter = nil
}
}
}
}
}
if q.cfg.EventCh != nil {
ev := QMPEvent{
Name: strname,
Data: eventData,
}
if timestamp != nil {
timestamp, ok := timestamp.(map[string]interface{})
if ok {
seconds, _ := timestamp["seconds"].(float64)
microseconds, _ := timestamp["microseconds"].(float64)
ev.Timestamp = time.Unix(int64(seconds), int64(microseconds))
}
}
q.cfg.EventCh <- ev
}
}
func (q *QMP) finaliseCommandWithResponse(cmdEl *list.Element, cmdQueue *list.List, succeeded bool, response interface{}) {
cmd := cmdEl.Value.(*qmpCommand)
cmdQueue.Remove(cmdEl)
select {
case <-cmd.ctx.Done():
default:
if succeeded {
cmd.res <- qmpResult{response: response}
} else {
cmd.res <- qmpResult{err: fmt.Errorf("QMP command failed: %v", response)}
}
}
if cmdQueue.Len() > 0 {
q.writeNextQMPCommand(cmdQueue)
}
}
func (q *QMP) finaliseCommand(cmdEl *list.Element, cmdQueue *list.List, succeeded bool) {
q.finaliseCommandWithResponse(cmdEl, cmdQueue, succeeded, nil)
}
func (q *QMP) errorDesc(errorData interface{}) (string, error) {
// convert error to json
data, err := json.Marshal(errorData)
if err != nil {
return "", fmt.Errorf("unable to extract error information: %v", err)
}
// see: https://github.com/qemu/qemu/blob/stable-2.12/qapi/qmp-dispatch.c#L125
var qmpErr map[string]string
// convert json to qmpError
if err = json.Unmarshal(data, &qmpErr); err != nil {
return "", fmt.Errorf("unable to convert json to qmpError: %v", err)
}
return qmpErr["desc"], nil
}
func (q *QMP) processQMPInput(line []byte, cmdQueue *list.List) {
var vmData map[string]interface{}
err := json.Unmarshal(line, &vmData)
if err != nil {
q.cfg.Logger.Warningf("Unable to decode response [%s] from VM: %v",
string(line), err)
return
}
if evname, found := vmData["event"]; found {
q.processQMPEvent(cmdQueue, evname, vmData["data"], vmData["timestamp"])
return
}
response, succeeded := vmData["return"]
errData, failed := vmData["error"]
if !succeeded && !failed {
return
}
cmdEl := cmdQueue.Front()
if cmdEl == nil {
q.cfg.Logger.Warningf("Unexpected command response received [%s] from VM",
string(line))
return
}
cmd := cmdEl.Value.(*qmpCommand)
if failed || cmd.filter == nil {
if errData != nil {
desc, err := q.errorDesc(errData)
if err != nil {
q.cfg.Logger.Infof("Get error description failed: %v", err)
} else {
response = desc
}
}
q.finaliseCommandWithResponse(cmdEl, cmdQueue, succeeded, response)
} else {
cmd.resultReceived = true
}
}
func currentCommandDoneCh(cmdQueue *list.List) <-chan struct{} {
cmdEl := cmdQueue.Front()
if cmdEl == nil {
return nil
}
cmd := cmdEl.Value.(*qmpCommand)
return cmd.ctx.Done()
}
func (q *QMP) writeNextQMPCommand(cmdQueue *list.List) {
cmdEl := cmdQueue.Front()
cmd := cmdEl.Value.(*qmpCommand)
cmdData := make(map[string]interface{})
cmdData["execute"] = cmd.name
if cmd.args != nil {
cmdData["arguments"] = cmd.args
}
encodedCmd, err := json.Marshal(&cmdData)
if err != nil {
cmd.res <- qmpResult{
err: fmt.Errorf("unable to marhsall command %s: %v",
cmd.name, err),
}
cmdQueue.Remove(cmdEl)
}
encodedCmd = append(encodedCmd, '\n')
if unixConn, ok := q.conn.(*net.UnixConn); ok && len(cmd.oob) > 0 {
_, _, err = unixConn.WriteMsgUnix(encodedCmd, cmd.oob, nil)
} else {
_, err = q.conn.Write(encodedCmd)
}
if err != nil {
cmd.res <- qmpResult{
err: fmt.Errorf("unable to write command to qmp socket %v", err),
}
cmdQueue.Remove(cmdEl)
}
}
func failOutstandingCommands(cmdQueue *list.List) {
for e := cmdQueue.Front(); e != nil; e = e.Next() {
cmd := e.Value.(*qmpCommand)
select {
case cmd.res <- qmpResult{
err: errors.New("exitting QMP loop, command cancelled"),
}:
case <-cmd.ctx.Done():
}
}
}
func (q *QMP) cancelCurrentCommand(cmdQueue *list.List) {
cmdEl := cmdQueue.Front()
cmd := cmdEl.Value.(*qmpCommand)
if cmd.resultReceived {
q.finaliseCommand(cmdEl, cmdQueue, false)
} else {
cmd.filter = nil
}
}
func (q *QMP) parseVersion(version []byte) *QMPVersion {
var qmp map[string]interface{}
err := json.Unmarshal(version, &qmp)
if err != nil {
q.cfg.Logger.Errorf("Invalid QMP greeting: %s", string(version))
return nil
}
versionMap := qmp
for _, k := range []string{"QMP", "version", "qemu"} {
versionMap, _ = versionMap[k].(map[string]interface{})
if versionMap == nil {
return nil
}
}
micro, _ := versionMap["micro"].(float64)
minor, _ := versionMap["minor"].(float64)
major, _ := versionMap["major"].(float64)
capabilities, _ := qmp["QMP"].(map[string]interface{})["capabilities"].([]interface{})
stringcaps := make([]string, 0, len(capabilities))
for _, c := range capabilities {
if cap, ok := c.(string); ok {
stringcaps = append(stringcaps, cap)
}
}
return &QMPVersion{Major: int(major),
Minor: int(minor),
Micro: int(micro),
Capabilities: stringcaps,
}
}
// The qemu package allows multiple QMP commands to be submitted concurrently
// from different Go routines. Unfortunately, QMP doesn't really support parallel
// commands as there is no way reliable way to associate a command response
// with a request. For this reason we need to submit our commands to
// QMP serially. The qemu package performs this serialisation using a
// queue (cmdQueue owned by mainLoop). We use a queue rather than a simple
// mutex so we can support cancelling of commands (see below) and ordered
// execution of commands, i.e., if command B is issued before command C,
// it should be executed before command C even if both commands are initially
// blocked waiting for command A to finish. This would be hard to achieve with
// a simple mutex.
//
// Cancelling is a little tricky. Commands such as ExecuteQMPCapabilities
// can be cancelled by cancelling or timing out their contexts. When a
// command is cancelled the calling function, e.g., ExecuteQMPCapabilities,
// will return but we may not be able to remove the command's entry from
// the command queue or issue the next command. There are two scenarios
// here.
//
// 1. The command has been processed by QMP, i.e., we have received a
// return or an error, but is still blocking as it is waiting for
// an event. For example, the ExecuteDeviceDel blocks until a DEVICE_DELETED
// event is received. When such a command is cancelled we can remove it
// from the queue and start issuing the next command. When the DEVICE_DELETED
// event eventually arrives it will just be ignored.
//
// 2. The command has not been processed by QMP. In this case the command
// needs to remain on the cmdQueue until the response to this command is
// received from QMP. During this time no new commands can be issued. When the
// response is received, it is discarded (as no one is interested in the result
// any more), the entry is removed from the cmdQueue and we can proceed to
// execute the next command.
func (q *QMP) mainLoop() {
cmdQueue := list.New().Init()
fromVMCh := make(chan []byte)
go q.readLoop(fromVMCh)
defer func() {
if q.cfg.EventCh != nil {
close(q.cfg.EventCh)
}
/* #nosec */
_ = q.conn.Close()
<-fromVMCh
failOutstandingCommands(cmdQueue)
close(q.disconnectedCh)
}()
var cmdDoneCh <-chan struct{}
var version *QMPVersion
ready := false
for {
select {
case cmd, ok := <-q.cmdCh:
if !ok {
return
}
_ = cmdQueue.PushBack(&cmd)
// We only want to execute the new cmd if QMP is
// ready and there are no other commands pending.
// If there are commands pending our new command
// will get run when the pending commands complete.
if ready && cmdQueue.Len() == 1 {
q.writeNextQMPCommand(cmdQueue)
cmdDoneCh = currentCommandDoneCh(cmdQueue)
}
case line, ok := <-fromVMCh:
if !ok {
return
}
if !ready {
// Not ready yet. Check if line is the QMP version.
// Sometimes QMP events are thrown before the QMP version,
// hence it's not a guarantee that the first data read from
// the channel is the QMP version.
version = q.parseVersion(line)
if version != nil {
q.connectedCh <- version
ready = true
}
// Do not process QMP input to avoid deadlocks.
break
}
q.processQMPInput(line, cmdQueue)
cmdDoneCh = currentCommandDoneCh(cmdQueue)
case <-cmdDoneCh:
q.cancelCurrentCommand(cmdQueue)
cmdDoneCh = currentCommandDoneCh(cmdQueue)
}
}
}
func startQMPLoop(conn io.ReadWriteCloser, cfg QMPConfig,
connectedCh chan<- *QMPVersion, disconnectedCh chan struct{}) *QMP {
q := &QMP{
cmdCh: make(chan qmpCommand),
conn: conn,
cfg: cfg,
connectedCh: connectedCh,
disconnectedCh: disconnectedCh,
}
go q.mainLoop()
return q
}
func (q *QMP) executeCommandWithResponse(ctx context.Context, name string, args map[string]interface{},
oob []byte, filter *qmpEventFilter) (interface{}, error) {
var err error
var response interface{}
resCh := make(chan qmpResult)
select {
case <-q.disconnectedCh:
err = errors.New("exitting QMP loop, command cancelled")
case q.cmdCh <- qmpCommand{
ctx: ctx,
res: resCh,
name: name,
args: args,
filter: filter,
oob: oob,
}:
}
if err != nil {
return response, err
}
select {
case res := <-resCh:
err = res.err
response = res.response
case <-ctx.Done():
err = ctx.Err()
}
return response, err
}
func (q *QMP) executeCommand(ctx context.Context, name string, args map[string]interface{},
filter *qmpEventFilter) error {
_, err := q.executeCommandWithResponse(ctx, name, args, nil, filter)
return err
}
// QMPStart connects to a unix domain socket maintained by a QMP instance. It
// waits to receive the QMP welcome message via the socket and spawns some go
// routines to manage the socket. The function returns a *QMP which can be
// used by callers to send commands to the QEMU instance or to close the
// socket and all the go routines that have been spawned to monitor it. A
// *QMPVersion is also returned. This structure contains the version and
// capabilities information returned by the QEMU instance in its welcome
// message.
//
// socket contains the path to the domain socket. cfg contains some options
// that can be specified by the caller, namely where the qemu package should
// send logs and QMP events. disconnectedCh is a channel that must be supplied
// by the caller. It is closed when an error occurs openning or writing to
// or reading from the unix domain socket. This implies that the QEMU instance
// that opened the socket has closed.
//
// If this function returns without error, callers should call QMP.Shutdown
// when they wish to stop monitoring the QMP instance. This is not strictly
// necessary if the QEMU instance exits and the disconnectedCh is closed, but
// doing so will not cause any problems.
//
// Commands can be sent to the QEMU instance via the QMP.Execute methods.
// These commands are executed serially, even if the QMP.Execute methods
// are called from different go routines. The QMP.Execute methods will
// block until they have received a success or failure message from QMP,
// i.e., {"return": {}} or {"error":{}}, and in some cases certain events
// are received.
//
// QEMU currently requires that the "qmp_capabilties" command is sent before any
// other command. Therefore you must call qmp.ExecuteQMPCapabilities() before
// you execute any other command.
func QMPStart(ctx context.Context, socket string, cfg QMPConfig, disconnectedCh chan struct{}) (*QMP, *QMPVersion, error) {
if cfg.Logger == nil {
cfg.Logger = qmpNullLogger{}
}
dialer := net.Dialer{Cancel: ctx.Done()}
conn, err := dialer.Dial("unix", socket)
if err != nil {
cfg.Logger.Warningf("Unable to connect to unix socket (%s): %v", socket, err)
close(disconnectedCh)
return nil, nil, err
}
connectedCh := make(chan *QMPVersion)
q := startQMPLoop(conn, cfg, connectedCh, disconnectedCh)
select {
case <-ctx.Done():
q.Shutdown()
<-disconnectedCh
return nil, nil, fmt.Errorf("canceled by caller")
case <-disconnectedCh:
return nil, nil, fmt.Errorf("lost connection to VM")
case q.version = <-connectedCh:
if q.version == nil {
return nil, nil, fmt.Errorf("failed to find QMP version information")
}
}
if q.version.Major < 5 {
return nil, nil, fmt.Errorf("govmm requires qemu version 5.0 or later, this is qemu (%d.%d)", q.version.Major, q.version.Minor)
}
return q, q.version, nil
}
// Shutdown closes the domain socket used to monitor a QEMU instance and
// terminates all the go routines spawned by QMPStart to manage that instance.
// QMP.Shutdown does not shut down the running instance. Calling QMP.Shutdown
// will result in the disconnectedCh channel being closed, indicating that we
// have lost connection to the QMP instance. In this case it does not indicate
// that the instance has quit.
//
// QMP.Shutdown should not be called concurrently with other QMP methods. It
// should not be called twice on the same QMP instance.
//
// Calling QMP.Shutdown after the disconnectedCh channel is closed is permitted but
// will not have any effect.
func (q *QMP) Shutdown() {
close(q.cmdCh)
}
// ExecuteQMPCapabilities executes the qmp_capabilities command on the instance.
func (q *QMP) ExecuteQMPCapabilities(ctx context.Context) error {
return q.executeCommand(ctx, "qmp_capabilities", nil, nil)
}
// ExecuteStop sends the stop command to the instance.
func (q *QMP) ExecuteStop(ctx context.Context) error {
return q.executeCommand(ctx, "stop", nil, nil)
}
// ExecuteCont sends the cont command to the instance.
func (q *QMP) ExecuteCont(ctx context.Context) error {
return q.executeCommand(ctx, "cont", nil, nil)
}
// ExecuteSystemPowerdown sends the system_powerdown command to the instance.
// This function will block until the SHUTDOWN event is received.
func (q *QMP) ExecuteSystemPowerdown(ctx context.Context) error {
filter := &qmpEventFilter{
eventName: "POWERDOWN",
}
return q.executeCommand(ctx, "system_powerdown", nil, filter)
}
// ExecuteQuit sends the quit command to the instance, terminating
// the QMP instance immediately.
func (q *QMP) ExecuteQuit(ctx context.Context) error {
return q.executeCommand(ctx, "quit", nil, nil)
}
func (q *QMP) blockdevAddBaseArgs(driver, device, blockdevID string, ro bool) (map[string]interface{}, map[string]interface{}) {
var args map[string]interface{}
blockdevArgs := map[string]interface{}{
"driver": "raw",
"read-only": ro,
"file": map[string]interface{}{
"driver": driver,
"filename": device,
},
}
blockdevArgs["node-name"] = blockdevID
args = blockdevArgs
return args, blockdevArgs
}
// ExecuteBlockdevAdd sends a blockdev-add to the QEMU instance. device is the
// path of the device to add, e.g., /dev/rdb0, and blockdevID is an identifier
// used to name the device. As this identifier will be passed directly to QMP,
// it must obey QMP's naming rules, e,g., it must start with a letter.
func (q *QMP) ExecuteBlockdevAdd(ctx context.Context, device, blockdevID string, ro bool) error {
args, _ := q.blockdevAddBaseArgs("host_device", device, blockdevID, ro)
return q.executeCommand(ctx, "blockdev-add", args, nil)
}
// ExecuteBlockdevAddWithCache has two more parameters direct and noFlush
// than ExecuteBlockdevAdd.
// They are cache-related options for block devices that are described in
// https://github.com/qemu/qemu/blob/master/qapi/block-core.json.
// direct denotes whether use of O_DIRECT (bypass the host page cache)
// is enabled. noFlush denotes whether flush requests for the device are
// ignored.
func (q *QMP) ExecuteBlockdevAddWithCache(ctx context.Context, device, blockdevID string, direct, noFlush, ro bool) error {
args, blockdevArgs := q.blockdevAddBaseArgs("host_device", device, blockdevID, ro)
blockdevArgs["cache"] = map[string]interface{}{
"direct": direct,
"no-flush": noFlush,
}
return q.executeCommand(ctx, "blockdev-add", args, nil)
}
// ExecuteBlockdevAddWithDriverCache has three one parameter driver
// than ExecuteBlockdevAddWithCache.
// Parameter driver can set the driver of block device.
func (q *QMP) ExecuteBlockdevAddWithDriverCache(ctx context.Context, driver, device, blockdevID string, direct, noFlush, ro bool) error {
args, blockdevArgs := q.blockdevAddBaseArgs(driver, device, blockdevID, ro)
blockdevArgs["cache"] = map[string]interface{}{
"direct": direct,
"no-flush": noFlush,
}
return q.executeCommand(ctx, "blockdev-add", args, nil)
}
// ExecuteDeviceAdd adds the guest portion of a device to a QEMU instance
// using the device_add command. blockdevID should match the blockdevID passed
// to a previous call to ExecuteBlockdevAdd. devID is the id of the device to
// add. Both strings must be valid QMP identifiers. driver is the name of the
// driver,e.g., virtio-blk-pci, and bus is the name of the bus. bus is optional.
// shared denotes if the drive can be shared allowing it to be passed more than once.
// disableModern indicates if virtio version 1.0 should be replaced by the
// former version 0.9, as there is a KVM bug that occurs when using virtio
// 1.0 in nested environments.
func (q *QMP) ExecuteDeviceAdd(ctx context.Context, blockdevID, devID, driver, bus, romfile string, shared, disableModern bool) error {
args := map[string]interface{}{
"id": devID,
"driver": driver,
"drive": blockdevID,
}
var transport VirtioTransport
if transport.isVirtioCCW(nil) {
args["devno"] = bus
} else if bus != "" {
args["bus"] = bus
}
if shared {
args["share-rw"] = "on"
}
if transport.isVirtioPCI(nil) {
args["romfile"] = romfile
if disableModern {
args["disable-modern"] = disableModern
}
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteSCSIDeviceAdd adds the guest portion of a block device to a QEMU instance
// using a SCSI driver with the device_add command. blockdevID should match the
// blockdevID passed to a previous call to ExecuteBlockdevAdd. devID is the id of
// the device to add. Both strings must be valid QMP identifiers. driver is the name of the
// scsi driver,e.g., scsi-hd, and bus is the name of a SCSI controller bus.
// scsiID is the SCSI id, lun is logical unit number. scsiID and lun are optional, a negative value
// for scsiID and lun is ignored. shared denotes if the drive can be shared allowing it
// to be passed more than once.
// disableModern indicates if virtio version 1.0 should be replaced by the
// former version 0.9, as there is a KVM bug that occurs when using virtio
// 1.0 in nested environments.
func (q *QMP) ExecuteSCSIDeviceAdd(ctx context.Context, blockdevID, devID, driver, bus, romfile string, scsiID, lun int, shared, disableModern bool) error {
// TBD: Add drivers for scsi passthrough like scsi-generic and scsi-block
drivers := []string{"scsi-hd", "scsi-cd", "scsi-disk"}
isSCSIDriver := false
for _, d := range drivers {
if driver == d {
isSCSIDriver = true
break
}
}
if !isSCSIDriver {
return fmt.Errorf("invalid SCSI driver provided %s", driver)
}
args := map[string]interface{}{
"id": devID,
"driver": driver,
"drive": blockdevID,
"bus": bus,
}
if scsiID >= 0 {
args["scsi-id"] = scsiID
}
if lun >= 0 {
args["lun"] = lun
}
if shared {
args["share-rw"] = "on"
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteBlockdevDel deletes a block device by sending blockdev-del
// command. blockdevID is the id of the block device to be deleted.
// Typically, this will match the id passed to ExecuteBlockdevAdd. It
// must be a valid QMP id.
func (q *QMP) ExecuteBlockdevDel(ctx context.Context, blockdevID string) error {
args := map[string]interface{}{}
args["node-name"] = blockdevID
return q.executeCommand(ctx, "blockdev-del", args, nil)
}
// ExecuteChardevDel deletes a char device by sending a chardev-remove command.
// chardevID is the id of the char device to be deleted. Typically, this will
// match the id passed to ExecuteCharDevUnixSocketAdd. It must be a valid QMP id.
func (q *QMP) ExecuteChardevDel(ctx context.Context, chardevID string) error {
args := map[string]interface{}{
"id": chardevID,
}
return q.executeCommand(ctx, "chardev-remove", args, nil)
}
// ExecuteNetdevAdd adds a Net device to a QEMU instance
// using the netdev_add command. netdevID is the id of the device to add.
// Must be valid QMP identifier.
func (q *QMP) ExecuteNetdevAdd(ctx context.Context, netdevType, netdevID, ifname, downscript, script string, queues int) error {
args := map[string]interface{}{
"type": netdevType,
"id": netdevID,
"ifname": ifname,
"downscript": downscript,
"script": script,
}
if queues > 1 {
args["queues"] = queues
}
return q.executeCommand(ctx, "netdev_add", args, nil)
}
// ExecuteNetdevChardevAdd adds a Net device to a QEMU instance
// using the netdev_add command. netdevID is the id of the device to add.
// Must be valid QMP identifier.
func (q *QMP) ExecuteNetdevChardevAdd(ctx context.Context, netdevType, netdevID, chardev string, queues int) error {
args := map[string]interface{}{
"type": netdevType,
"id": netdevID,
"chardev": chardev,
}
if queues > 1 {
args["queues"] = queues
}
return q.executeCommand(ctx, "netdev_add", args, nil)
}
// ExecuteNetdevAddByFds adds a Net device to a QEMU instance
// using the netdev_add command by fds and vhostfds. netdevID is the id of the device to add.
// Must be valid QMP identifier.
func (q *QMP) ExecuteNetdevAddByFds(ctx context.Context, netdevType, netdevID string, fdNames, vhostFdNames []string) error {
fdNameStr := strings.Join(fdNames, ":")
args := map[string]interface{}{
"type": netdevType,
"id": netdevID,
"fds": fdNameStr,
}
if len(vhostFdNames) > 0 {
vhostFdNameStr := strings.Join(vhostFdNames, ":")
args["vhost"] = true
args["vhostfds"] = vhostFdNameStr
}
return q.executeCommand(ctx, "netdev_add", args, nil)
}
// ExecuteNetdevDel deletes a Net device from a QEMU instance
// using the netdev_del command. netdevID is the id of the device to delete.
func (q *QMP) ExecuteNetdevDel(ctx context.Context, netdevID string) error {
args := map[string]interface{}{
"id": netdevID,
}
return q.executeCommand(ctx, "netdev_del", args, nil)
}
// ExecuteNetPCIDeviceAdd adds a Net PCI device to a QEMU instance
// using the device_add command. devID is the id of the device to add.
// Must be valid QMP identifier. netdevID is the id of nic added by previous netdev_add.
// queues is the number of queues of a nic.
// disableModern indicates if virtio version 1.0 should be replaced by the
// former version 0.9, as there is a KVM bug that occurs when using virtio
// 1.0 in nested environments.
func (q *QMP) ExecuteNetPCIDeviceAdd(ctx context.Context, netdevID, devID, macAddr, addr, bus, romfile string, queues int, disableModern bool) error {
args := map[string]interface{}{
"id": devID,
"driver": VirtioNetPCI,
"romfile": romfile,
}
if bus != "" {
args["bus"] = bus
}
if addr != "" {
args["addr"] = addr
}
if macAddr != "" {
args["mac"] = macAddr
}
if netdevID != "" {
args["netdev"] = netdevID
}
if disableModern {
args["disable-modern"] = disableModern
}
if queues > 0 {
// (2N+2 vectors, N for tx queues, N for rx queues, 1 for config, and one for possible control vq)
// -device virtio-net-pci,mq=on,vectors=2N+2...
// enable mq in guest by 'ethtool -L eth0 combined $queue_num'
// Clearlinux automatically sets up the queues properly
// The agent implementation should do this to ensure that it is
// always set
args["mq"] = "on"
args["vectors"] = 2*queues + 2
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteNetCCWDeviceAdd adds a Net CCW device to a QEMU instance
// using the device_add command. devID is the id of the device to add.
// Must be valid QMP identifier. netdevID is the id of nic added by previous netdev_add.
// queues is the number of queues of a nic.
func (q *QMP) ExecuteNetCCWDeviceAdd(ctx context.Context, netdevID, devID, macAddr, bus string, queues int) error {
args := map[string]interface{}{
"id": devID,
"driver": VirtioNetCCW,
"netdev": netdevID,
"mac": macAddr,
"devno": bus,
}
if queues > 0 {
args["mq"] = "on"
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteDeviceDel deletes guest portion of a QEMU device by sending a
// device_del command. devId is the identifier of the device to delete.
// Typically it would match the devID parameter passed to an earlier call
// to ExecuteDeviceAdd. It must be a valid QMP identidier.
//
// This method blocks until a DEVICE_DELETED event is received for devID.
func (q *QMP) ExecuteDeviceDel(ctx context.Context, devID string) error {
args := map[string]interface{}{
"id": devID,
}
filter := &qmpEventFilter{
eventName: "DEVICE_DELETED",
dataKey: "device",
dataValue: devID,
}
return q.executeCommand(ctx, "device_del", args, filter)
}
// ExecutePCIDeviceAdd is the PCI version of ExecuteDeviceAdd. This function can be used
// to hot plug PCI devices on PCI(E) bridges, unlike ExecuteDeviceAdd this function receive the
// device address on its parent bus. bus is optional. queues specifies the number of queues of
// a block device. shared denotes if the drive can be shared allowing it to be passed more than once.
// disableModern indicates if virtio version 1.0 should be replaced by the
// former version 0.9, as there is a KVM bug that occurs when using virtio
// 1.0 in nested environments.
func (q *QMP) ExecutePCIDeviceAdd(ctx context.Context, blockdevID, devID, driver, addr, bus, romfile string, queues int, shared, disableModern bool) error {
args := map[string]interface{}{
"id": devID,
"driver": driver,
"drive": blockdevID,
"addr": addr,
}
if bus != "" {
args["bus"] = bus
}
if shared {
args["share-rw"] = "on"
}
if queues > 0 {
args["num-queues"] = strconv.Itoa(queues)
}
var transport VirtioTransport
if transport.isVirtioPCI(nil) {
args["romfile"] = romfile
if disableModern {
args["disable-modern"] = disableModern
}
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecutePCIVhostUserDevAdd adds a vhost-user device to a QEMU instance using the device_add command.
// This function can be used to hot plug vhost-user devices on PCI(E) bridges.
// It receives the bus and the device address on its parent bus. bus is optional.
// devID is the id of the device to add.Must be valid QMP identifier. chardevID
// is the QMP identifier of character device using a unix socket as backend.
// driver is the name of vhost-user driver, like vhost-user-blk-pci.
func (q *QMP) ExecutePCIVhostUserDevAdd(ctx context.Context, driver, devID, chardevID, addr, bus string) error {
args := map[string]interface{}{
"driver": driver,
"id": devID,
"chardev": chardevID,
"addr": addr,
}
if bus != "" {
args["bus"] = bus
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteVFIODeviceAdd adds a VFIO device to a QEMU instance using the device_add command.
// devID is the id of the device to add. Must be valid QMP identifier.
// bdf is the PCI bus-device-function of the pci device.
// bus is optional. When hot plugging a PCIe device, the bus can be the ID of the pcie-root-port.
func (q *QMP) ExecuteVFIODeviceAdd(ctx context.Context, devID, bdf, bus, romfile string) error {
var driver string
var transport VirtioTransport
if transport.isVirtioCCW(nil) {
driver = string(VfioCCW)
} else {
driver = string(VfioPCI)
}
args := map[string]interface{}{
"id": devID,
"driver": driver,
"host": bdf,
"romfile": romfile,
}
if bus != "" {
args["bus"] = bus
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecutePCIVFIODeviceAdd adds a VFIO device to a QEMU instance using the device_add command.
// This function can be used to hot plug VFIO devices on PCI(E) bridges, unlike
// ExecuteVFIODeviceAdd this function receives the bus and the device address on its parent bus.
// bus is optional. devID is the id of the device to add.Must be valid QMP identifier. bdf is the
// PCI bus-device-function of the pci device.
func (q *QMP) ExecutePCIVFIODeviceAdd(ctx context.Context, devID, bdf, addr, bus, romfile string) error {
args := map[string]interface{}{
"id": devID,
"driver": VfioPCI,
"host": bdf,
"addr": addr,
"romfile": romfile,
}
if bus != "" {
args["bus"] = bus
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecutePCIVFIOMediatedDeviceAdd adds a VFIO mediated device to a QEMU instance using the device_add command.
// This function can be used to hot plug VFIO mediated devices on PCI(E) bridges or root bus, unlike
// ExecuteVFIODeviceAdd this function receives the bus and the device address on its parent bus.
// devID is the id of the device to add. Must be valid QMP identifier. sysfsdev is the VFIO mediated device.
// Both bus and addr are optional. If they are both set to be empty, the system will pick up an empty slot on root bus.
func (q *QMP) ExecutePCIVFIOMediatedDeviceAdd(ctx context.Context, devID, sysfsdev, addr, bus, romfile string) error {
args := map[string]interface{}{
"id": devID,
"driver": VfioPCI,
"sysfsdev": sysfsdev,
"romfile": romfile,
}
if bus != "" {
args["bus"] = bus
}
if addr != "" {
args["addr"] = addr
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteAPVFIOMediatedDeviceAdd adds a VFIO mediated AP device to a QEMU instance using the device_add command.
func (q *QMP) ExecuteAPVFIOMediatedDeviceAdd(ctx context.Context, sysfsdev string) error {
args := map[string]interface{}{
"driver": VfioAP,
"sysfsdev": sysfsdev,
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// isSocketIDSupported returns if the cpu driver supports the socket id option
func isSocketIDSupported(driver string) bool {
if driver == "host-s390x-cpu" || driver == "host-powerpc64-cpu" {
return false
}
return true
}
// isThreadIDSupported returns if the cpu driver supports the thread id option
func isThreadIDSupported(driver string) bool {
if driver == "host-s390x-cpu" || driver == "host-powerpc64-cpu" {
return false
}
return true
}
// isDieIDSupported returns if the cpu driver and the qemu version support the die id option
func (q *QMP) isDieIDSupported(driver string) bool {
return driver == "host-x86_64-cpu"
}
// ExecuteCPUDeviceAdd adds a CPU to a QEMU instance using the device_add command.
// driver is the CPU model, cpuID must be a unique ID to identify the CPU, socketID is the socket number within
// node/board the CPU belongs to, coreID is the core number within socket the CPU belongs to, threadID is the
// thread number within core the CPU belongs to. Note that socketID and threadID are not a requirement for
// architecures like ppc64le.
func (q *QMP) ExecuteCPUDeviceAdd(ctx context.Context, driver, cpuID, socketID, dieID, coreID, threadID, romfile string) error {
args := map[string]interface{}{
"driver": driver,
"id": cpuID,
"core-id": coreID,
}
if socketID != "" && isSocketIDSupported(driver) {
args["socket-id"] = socketID
}
if threadID != "" && isThreadIDSupported(driver) {
args["thread-id"] = threadID
}
if q.isDieIDSupported(driver) {
if dieID != "" {
args["die-id"] = dieID
}
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteQueryHotpluggableCPUs returns a slice with the list of hotpluggable CPUs
func (q *QMP) ExecuteQueryHotpluggableCPUs(ctx context.Context) ([]HotpluggableCPU, error) {
response, err := q.executeCommandWithResponse(ctx, "query-hotpluggable-cpus", nil, nil, nil)
if err != nil {
return nil, err
}
// convert response to json
data, err := json.Marshal(response)
if err != nil {
return nil, fmt.Errorf("unable to extract CPU information: %v", err)
}
var cpus []HotpluggableCPU
// convert json to []HotpluggableCPU
if err = json.Unmarshal(data, &cpus); err != nil {
return nil, fmt.Errorf("unable to convert json to hotpluggable CPU: %v", err)
}
return cpus, nil
}
// ExecSetMigrationCaps sets migration capabilities
func (q *QMP) ExecSetMigrationCaps(ctx context.Context, caps []map[string]interface{}) error {
args := map[string]interface{}{
"capabilities": caps,
}
return q.executeCommand(ctx, "migrate-set-capabilities", args, nil)
}
// ExecSetMigrateArguments sets the command line used for migration
func (q *QMP) ExecSetMigrateArguments(ctx context.Context, url string) error {
args := map[string]interface{}{
"uri": url,
}
return q.executeCommand(ctx, "migrate", args, nil)
}
// ExecQueryMemoryDevices returns a slice with the list of memory devices
func (q *QMP) ExecQueryMemoryDevices(ctx context.Context) ([]MemoryDevices, error) {
response, err := q.executeCommandWithResponse(ctx, "query-memory-devices", nil, nil, nil)
if err != nil {
return nil, err
}
// convert response to json
data, err := json.Marshal(response)
if err != nil {
return nil, fmt.Errorf("unable to extract memory devices information: %v", err)
}
var memoryDevices []MemoryDevices
// convert json to []MemoryDevices
if err = json.Unmarshal(data, &memoryDevices); err != nil {
return nil, fmt.Errorf("unable to convert json to memory devices: %v", err)
}
return memoryDevices, nil
}
// ExecQueryCpus returns a slice with the list of `CpuInfo`
// Since qemu 2.12, we have `query-cpus-fast` as a better choice in production
// we can still choose `ExecQueryCpus` for compatibility though not recommended.
func (q *QMP) ExecQueryCpus(ctx context.Context) ([]CPUInfo, error) {
response, err := q.executeCommandWithResponse(ctx, "query-cpus", nil, nil, nil)
if err != nil {
return nil, err
}
// convert response to json
data, err := json.Marshal(response)
if err != nil {
return nil, fmt.Errorf("unable to extract memory devices information: %v", err)
}
var cpuInfo []CPUInfo
// convert json to []CPUInfo
if err = json.Unmarshal(data, &cpuInfo); err != nil {
return nil, fmt.Errorf("unable to convert json to CPUInfo: %v", err)
}
return cpuInfo, nil
}
// ExecQueryCpusFast returns a slice with the list of `CpuInfoFast`
// This is introduced since 2.12, it does not incur a performance penalty and
// should be used in production instead of query-cpus.
func (q *QMP) ExecQueryCpusFast(ctx context.Context) ([]CPUInfoFast, error) {
response, err := q.executeCommandWithResponse(ctx, "query-cpus-fast", nil, nil, nil)
if err != nil {
return nil, err
}
// convert response to json
data, err := json.Marshal(response)
if err != nil {
return nil, fmt.Errorf("unable to extract memory devices information: %v", err)
}
var cpuInfoFast []CPUInfoFast
// convert json to []CPUInfoFast
if err = json.Unmarshal(data, &cpuInfoFast); err != nil {
return nil, fmt.Errorf("unable to convert json to CPUInfoFast: %v", err)
}
return cpuInfoFast, nil
}
// ExecMemdevAdd adds size of MiB memory device to the guest
func (q *QMP) ExecMemdevAdd(ctx context.Context, qomtype, id, mempath string, size int, share bool, driver, driverID, addr, bus string) error {
args := map[string]interface{}{
"qom-type": qomtype,
"id": id,
"size": uint64(size) << 20,
}
if mempath != "" {
args["mem-path"] = mempath
}
if share {
args["share"] = true
}
err := q.executeCommand(ctx, "object-add", args, nil)
if err != nil {
return err
}
defer func() {
if err != nil {
q.cfg.Logger.Errorf("Unable to add memory device %s: %v", id, err)
err = q.executeCommand(ctx, "object-del", map[string]interface{}{"id": id}, nil)
if err != nil {
q.cfg.Logger.Warningf("Unable to clean up memory object %s: %v", id, err)
}
}
}()
args = map[string]interface{}{
"driver": driver,
"id": driverID,
"memdev": id,
}
if bus != "" {
args["bus"] = bus
}
if addr != "" {
args["addr"] = addr
}
err = q.executeCommand(ctx, "device_add", args, nil)
return err
}
// ExecHotplugMemory adds size of MiB memory to the guest
func (q *QMP) ExecHotplugMemory(ctx context.Context, qomtype, id, mempath string, size int, share bool) error {
return q.ExecMemdevAdd(ctx, qomtype, id, mempath, size, share, "pc-dimm", "dimm"+id, "", "")
}
// ExecuteNVDIMMDeviceAdd adds a block device to a QEMU instance using
// a NVDIMM driver with the device_add command.
// id is the id of the device to add. It must be a valid QMP identifier.
// mempath is the path of the device to add, e.g., /dev/rdb0. size is
// the data size of the device. pmem is to guarantee the persistence of QEMU writes
// to the vNVDIMM backend.
func (q *QMP) ExecuteNVDIMMDeviceAdd(ctx context.Context, id, mempath string, size int64, pmem *bool) error {
args := map[string]interface{}{
"qom-type": "memory-backend-file",
"id": "nvdimmbackmem" + id,
"mem-path": mempath,
"size": size,
"share": true,
}
if pmem != nil {
args["pmem"] = *pmem
}
err := q.executeCommand(ctx, "object-add", args, nil)
if err != nil {
return err
}
args = map[string]interface{}{
"driver": "nvdimm",
"id": "nvdimm" + id,
"memdev": "nvdimmbackmem" + id,
}
if err = q.executeCommand(ctx, "device_add", args, nil); err != nil {
q.cfg.Logger.Errorf("Unable to hotplug NVDIMM device: %v", err)
err2 := q.executeCommand(ctx, "object-del", map[string]interface{}{"id": "nvdimmbackmem" + id}, nil)
if err2 != nil {
q.cfg.Logger.Warningf("Unable to clean up memory object: %v", err2)
}
}
return err
}
// ExecuteBalloon sets the size of the balloon, hence updates the memory
// allocated for the VM.
func (q *QMP) ExecuteBalloon(ctx context.Context, bytes uint64) error {
args := map[string]interface{}{
"value": bytes,
}
return q.executeCommand(ctx, "balloon", args, nil)
}
// ExecutePCIVSockAdd adds a vhost-vsock-pci bus
// disableModern indicates if virtio version 1.0 should be replaced by the
// former version 0.9, as there is a KVM bug that occurs when using virtio
// 1.0 in nested environments.
func (q *QMP) ExecutePCIVSockAdd(ctx context.Context, id, guestCID, vhostfd, addr, bus, romfile string, disableModern bool) error {
args := map[string]interface{}{
"driver": VHostVSockPCI,
"id": id,
"guest-cid": guestCID,
"vhostfd": vhostfd,
"addr": addr,
"romfile": romfile,
}
if bus != "" {
args["bus"] = bus
}
if disableModern {
args["disable-modern"] = disableModern
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteGetFD sends a file descriptor via SCM rights and assigns it a name
func (q *QMP) ExecuteGetFD(ctx context.Context, fdname string, fd *os.File) error {
oob := syscall.UnixRights(int(fd.Fd()))
args := map[string]interface{}{
"fdname": fdname,
}
_, err := q.executeCommandWithResponse(ctx, "getfd", args, oob, nil)
return err
}
// ExecuteCharDevUnixSocketAdd adds a character device using as backend a unix socket,
// id is an identifier for the device, path specifies the local path of the unix socket,
// wait is to block waiting for a client to connect, server specifies that the socket is a listening socket.
func (q *QMP) ExecuteCharDevUnixSocketAdd(ctx context.Context, id, path string, wait, server bool) error {
data := map[string]interface{}{
"server": server,
"addr": map[string]interface{}{
"type": "unix",
"data": map[string]interface{}{
"path": path,
},
},
}
// wait is only valid for server mode
if server {
data["wait"] = wait
}
args := map[string]interface{}{
"id": id,
"backend": map[string]interface{}{
"type": "socket",
"data": data,
},
}
return q.executeCommand(ctx, "chardev-add", args, nil)
}
// ExecuteVirtSerialPortAdd adds a virtserialport.
// id is an identifier for the virtserialport, name is a name for the virtserialport and
// it will be visible in the VM, chardev is the character device id previously added.
func (q *QMP) ExecuteVirtSerialPortAdd(ctx context.Context, id, name, chardev string) error {
args := map[string]interface{}{
"driver": VirtioSerialPort,
"id": id,
"name": name,
"chardev": chardev,
}
return q.executeCommand(ctx, "device_add", args, nil)
}
// ExecuteQueryMigration queries migration progress.
func (q *QMP) ExecuteQueryMigration(ctx context.Context) (MigrationStatus, error) {
response, err := q.executeCommandWithResponse(ctx, "query-migrate", nil, nil, nil)
if err != nil {
return MigrationStatus{}, err
}
data, err := json.Marshal(response)
if err != nil {
return MigrationStatus{}, fmt.Errorf("unable to extract migrate status information: %v", err)
}
var status MigrationStatus
if err = json.Unmarshal(data, &status); err != nil {
return MigrationStatus{}, fmt.Errorf("unable to convert migrate status information: %v", err)
}
return status, nil
}
// ExecuteMigrationIncoming start migration from incoming uri.
func (q *QMP) ExecuteMigrationIncoming(ctx context.Context, uri string) error {
args := map[string]interface{}{
"uri": uri,
}
return q.executeCommand(ctx, "migrate-incoming", args, nil)
}
// ExecQueryQmpSchema query all QMP wire ABI and returns a slice
func (q *QMP) ExecQueryQmpSchema(ctx context.Context) ([]SchemaInfo, error) {
response, err := q.executeCommandWithResponse(ctx, "query-qmp-schema", nil, nil, nil)
if err != nil {
return nil, err
}
// convert response to json
data, err := json.Marshal(response)
if err != nil {
return nil, fmt.Errorf("unable to extract memory devices information: %v", err)
}
var schemaInfo []SchemaInfo
if err = json.Unmarshal(data, &schemaInfo); err != nil {
return nil, fmt.Errorf("unable to convert json to schemaInfo: %v", err)
}
return schemaInfo, nil
}
// ExecuteQueryStatus queries guest status
func (q *QMP) ExecuteQueryStatus(ctx context.Context) (StatusInfo, error) {
response, err := q.executeCommandWithResponse(ctx, "query-status", nil, nil, nil)
if err != nil {
return StatusInfo{}, err
}
data, err := json.Marshal(response)
if err != nil {
return StatusInfo{}, fmt.Errorf("unable to extract migrate status information: %v", err)
}
var status StatusInfo
if err = json.Unmarshal(data, &status); err != nil {
return StatusInfo{}, fmt.Errorf("unable to convert migrate status information: %v", err)
}
return status, nil
}
// ExecQomSet qom-set path property value
func (q *QMP) ExecQomSet(ctx context.Context, path, property string, value uint64) error {
args := map[string]interface{}{
"path": path,
"property": property,
"value": value,
}
return q.executeCommand(ctx, "qom-set", args, nil)
}
// ExecQomGet qom-get path property
func (q *QMP) ExecQomGet(ctx context.Context, path, property string) (interface{}, error) {
args := map[string]interface{}{
"path": path,
"property": property,
}
response, err := q.executeCommandWithResponse(ctx, "qom-get", args, nil, nil)
if err != nil {
return "", err
}
return response, nil
}
// ExecuteDumpGuestMemory dump guest memory to host
func (q *QMP) ExecuteDumpGuestMemory(ctx context.Context, protocol string, paging bool, format string) error {
args := map[string]interface{}{
"protocol": protocol,
"paging": paging,
"format": format,
}
return q.executeCommand(ctx, "dump-guest-memory", args, nil)
}