Files
kata-containers/virtcontainers/pkg/oci/utils.go
Sebastien Boeuf 9c6ed93f80 hook: Move OCI hooks handling to the CLI
The CLI being the implementation of the OCI specification, and the
hooks being OCI specific, it makes sense to move the handling of any
OCI hooks to the CLI level. This changes allows the Kata API to
become OCI agnostic.

Fixes #599

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2018-08-24 15:07:27 -07:00

694 lines
18 KiB
Go

// Copyright (c) 2017 Intel Corporation
//
// SPDX-License-Identifier: Apache-2.0
//
package oci
import (
"context"
"encoding/json"
"errors"
"fmt"
"io/ioutil"
"path/filepath"
"strconv"
"strings"
"syscall"
criContainerdAnnotations "github.com/containerd/cri-containerd/pkg/annotations"
crioAnnotations "github.com/kubernetes-incubator/cri-o/pkg/annotations"
spec "github.com/opencontainers/runtime-spec/specs-go"
"github.com/sirupsen/logrus"
vc "github.com/kata-containers/runtime/virtcontainers"
"github.com/kata-containers/runtime/virtcontainers/device/config"
vcAnnotations "github.com/kata-containers/runtime/virtcontainers/pkg/annotations"
dockershimAnnotations "github.com/kata-containers/runtime/virtcontainers/pkg/annotations/dockershim"
"github.com/kata-containers/runtime/virtcontainers/utils"
)
type annotationContainerType struct {
annotation string
containerType vc.ContainerType
}
var (
// ErrNoLinux is an error for missing Linux sections in the OCI configuration file.
ErrNoLinux = errors.New("missing Linux section")
// CRIContainerTypeKeyList lists all the CRI keys that could define
// the container type from annotations in the config.json.
CRIContainerTypeKeyList = []string{criContainerdAnnotations.ContainerType, crioAnnotations.ContainerType, dockershimAnnotations.ContainerTypeLabelKey}
// CRISandboxNameKeyList lists all the CRI keys that could define
// the sandbox ID (sandbox ID) from annotations in the config.json.
CRISandboxNameKeyList = []string{criContainerdAnnotations.SandboxID, crioAnnotations.SandboxID, dockershimAnnotations.SandboxIDLabelKey}
// CRIContainerTypeList lists all the maps from CRI ContainerTypes annotations
// to a virtcontainers ContainerType.
CRIContainerTypeList = []annotationContainerType{
{crioAnnotations.ContainerTypeSandbox, vc.PodSandbox},
{crioAnnotations.ContainerTypeContainer, vc.PodContainer},
{criContainerdAnnotations.ContainerTypeSandbox, vc.PodSandbox},
{criContainerdAnnotations.ContainerTypeContainer, vc.PodContainer},
{dockershimAnnotations.ContainerTypeLabelSandbox, vc.PodSandbox},
{dockershimAnnotations.ContainerTypeLabelContainer, vc.PodContainer},
}
)
const (
// StateCreated represents a container that has been created and is
// ready to be run.
StateCreated = "created"
// StateRunning represents a container that's currently running.
StateRunning = "running"
// StateStopped represents a container that has been stopped.
StateStopped = "stopped"
// StatePaused represents a container that has been paused.
StatePaused = "paused"
)
// CompatOCIProcess is a structure inheriting from spec.Process defined
// in runtime-spec/specs-go package. The goal is to be compatible with
// both v1.0.0-rc4 and v1.0.0-rc5 since the latter introduced a change
// about the type of the Capabilities field.
// Refer to: https://github.com/opencontainers/runtime-spec/commit/37391fb
type CompatOCIProcess struct {
spec.Process
Capabilities interface{} `json:"capabilities,omitempty" platform:"linux"`
}
// CompatOCISpec is a structure inheriting from spec.Spec defined
// in runtime-spec/specs-go package. It relies on the CompatOCIProcess
// structure declared above, in order to be compatible with both
// v1.0.0-rc4 and v1.0.0-rc5.
// Refer to: https://github.com/opencontainers/runtime-spec/commit/37391fb
type CompatOCISpec struct {
spec.Spec
Process *CompatOCIProcess `json:"process,omitempty"`
}
// FactoryConfig is a structure to set the VM factory configuration.
type FactoryConfig struct {
// Template enables VM templating support in VM factory.
Template bool
}
// RuntimeConfig aggregates all runtime specific settings
type RuntimeConfig struct {
VMConfig vc.Resources
HypervisorType vc.HypervisorType
HypervisorConfig vc.HypervisorConfig
FactoryConfig FactoryConfig
AgentType vc.AgentType
AgentConfig interface{}
ProxyType vc.ProxyType
ProxyConfig vc.ProxyConfig
ShimType vc.ShimType
ShimConfig interface{}
Console string
//Determines how the VM should be connected to the
//the container network interface
InterNetworkModel vc.NetInterworkingModel
}
// AddKernelParam allows the addition of new kernel parameters to an existing
// hypervisor configuration stored inside the current runtime configuration.
func (config *RuntimeConfig) AddKernelParam(p vc.Param) error {
return config.HypervisorConfig.AddKernelParam(p)
}
var ociLog = logrus.WithFields(logrus.Fields{
"source": "virtcontainers",
"subsystem": "oci",
})
// SetLogger sets the logger for oci package.
func SetLogger(ctx context.Context, logger *logrus.Entry) {
fields := ociLog.Data
ociLog = logger.WithFields(fields)
}
func cmdEnvs(spec CompatOCISpec, envs []vc.EnvVar) []vc.EnvVar {
for _, env := range spec.Process.Env {
kv := strings.Split(env, "=")
if len(kv) < 2 {
continue
}
envs = append(envs,
vc.EnvVar{
Var: kv[0],
Value: kv[1],
})
}
return envs
}
func newMount(m spec.Mount) vc.Mount {
return vc.Mount{
Source: m.Source,
Destination: m.Destination,
Type: m.Type,
Options: m.Options,
}
}
func containerMounts(spec CompatOCISpec) []vc.Mount {
ociMounts := spec.Spec.Mounts
if ociMounts == nil {
return []vc.Mount{}
}
var mnts []vc.Mount
for _, m := range ociMounts {
mnts = append(mnts, newMount(m))
}
return mnts
}
func contains(s []string, e string) bool {
for _, a := range s {
if a == e {
return true
}
}
return false
}
func newLinuxDeviceInfo(d spec.LinuxDevice) (*config.DeviceInfo, error) {
allowedDeviceTypes := []string{"c", "b", "u", "p"}
if !contains(allowedDeviceTypes, d.Type) {
return nil, fmt.Errorf("Unexpected Device Type %s for device %s", d.Type, d.Path)
}
if d.Path == "" {
return nil, fmt.Errorf("Path cannot be empty for device")
}
deviceInfo := config.DeviceInfo{
ContainerPath: d.Path,
DevType: d.Type,
Major: d.Major,
Minor: d.Minor,
}
if d.UID != nil {
deviceInfo.UID = *d.UID
}
if d.GID != nil {
deviceInfo.GID = *d.GID
}
if d.FileMode != nil {
deviceInfo.FileMode = *d.FileMode
}
return &deviceInfo, nil
}
func containerDeviceInfos(spec CompatOCISpec) ([]config.DeviceInfo, error) {
ociLinuxDevices := spec.Spec.Linux.Devices
if ociLinuxDevices == nil {
return []config.DeviceInfo{}, nil
}
var devices []config.DeviceInfo
for _, d := range ociLinuxDevices {
linuxDeviceInfo, err := newLinuxDeviceInfo(d)
if err != nil {
return []config.DeviceInfo{}, err
}
devices = append(devices, *linuxDeviceInfo)
}
return devices, nil
}
func containerCapabilities(s CompatOCISpec) (vc.LinuxCapabilities, error) {
capabilities := s.Process.Capabilities
var c vc.LinuxCapabilities
// In spec v1.0.0-rc4, capabilities was a list of strings. This was changed
// to an object with v1.0.0-rc5.
// Check for the interface type to support both the versions.
switch caps := capabilities.(type) {
case map[string]interface{}:
for key, value := range caps {
switch val := value.(type) {
case []interface{}:
var list []string
for _, str := range val {
list = append(list, str.(string))
}
switch key {
case "bounding":
c.Bounding = list
case "effective":
c.Effective = list
case "inheritable":
c.Inheritable = list
case "ambient":
c.Ambient = list
case "permitted":
c.Permitted = list
}
default:
return c, fmt.Errorf("Unexpected format for capabilities: %v", caps)
}
}
case []interface{}:
var list []string
for _, str := range caps {
list = append(list, str.(string))
}
c = vc.LinuxCapabilities{
Bounding: list,
Effective: list,
Inheritable: list,
Ambient: list,
Permitted: list,
}
case nil:
ociLog.Debug("Empty capabilities have been passed")
return c, nil
default:
return c, fmt.Errorf("Unexpected format for capabilities: %v", caps)
}
return c, nil
}
// ContainerCapabilities return a LinuxCapabilities for virtcontainer
func ContainerCapabilities(s CompatOCISpec) (vc.LinuxCapabilities, error) {
if s.Process == nil {
return vc.LinuxCapabilities{}, fmt.Errorf("ContainerCapabilities, Process is nil")
}
return containerCapabilities(s)
}
func networkConfig(ocispec CompatOCISpec, config RuntimeConfig) (vc.NetworkConfig, error) {
linux := ocispec.Linux
if linux == nil {
return vc.NetworkConfig{}, ErrNoLinux
}
var netConf vc.NetworkConfig
for _, n := range linux.Namespaces {
if n.Type != spec.NetworkNamespace {
continue
}
if n.Path != "" {
netConf.NetNSPath = n.Path
}
}
netConf.InterworkingModel = config.InterNetworkModel
return netConf, nil
}
// getConfigPath returns the full config path from the bundle
// path provided.
func getConfigPath(bundlePath string) string {
return filepath.Join(bundlePath, "config.json")
}
// ParseConfigJSON unmarshals the config.json file.
func ParseConfigJSON(bundlePath string) (CompatOCISpec, error) {
configPath := getConfigPath(bundlePath)
ociLog.Debugf("converting %s", configPath)
configByte, err := ioutil.ReadFile(configPath)
if err != nil {
return CompatOCISpec{}, err
}
var ocispec CompatOCISpec
if err := json.Unmarshal(configByte, &ocispec); err != nil {
return CompatOCISpec{}, err
}
caps, err := ContainerCapabilities(ocispec)
if err != nil {
return CompatOCISpec{}, err
}
ocispec.Process.Capabilities = caps
return ocispec, nil
}
// GetContainerType determines which type of container matches the annotations
// table provided.
func GetContainerType(annotations map[string]string) (vc.ContainerType, error) {
if containerType, ok := annotations[vcAnnotations.ContainerTypeKey]; ok {
return vc.ContainerType(containerType), nil
}
ociLog.Errorf("Annotations[%s] not found, cannot determine the container type",
vcAnnotations.ContainerTypeKey)
return vc.UnknownContainerType, fmt.Errorf("Could not find container type")
}
// ContainerType returns the type of container and if the container type was
// found from CRI servers annotations.
func (spec *CompatOCISpec) ContainerType() (vc.ContainerType, error) {
for _, key := range CRIContainerTypeKeyList {
containerTypeVal, ok := spec.Annotations[key]
if !ok {
continue
}
for _, t := range CRIContainerTypeList {
if t.annotation == containerTypeVal {
return t.containerType, nil
}
}
return vc.UnknownContainerType, fmt.Errorf("Unknown container type %s", containerTypeVal)
}
return vc.PodSandbox, nil
}
// SandboxID determines the sandbox ID related to an OCI configuration. This function
// is expected to be called only when the container type is "PodContainer".
func (spec *CompatOCISpec) SandboxID() (string, error) {
for _, key := range CRISandboxNameKeyList {
sandboxID, ok := spec.Annotations[key]
if ok {
return sandboxID, nil
}
}
return "", fmt.Errorf("Could not find sandbox ID")
}
func vmConfig(ocispec CompatOCISpec, config RuntimeConfig) (vc.Resources, error) {
resources := config.VMConfig
if ocispec.Linux == nil || ocispec.Linux.Resources == nil {
return resources, nil
}
if ocispec.Linux.Resources.Memory != nil &&
ocispec.Linux.Resources.Memory.Limit != nil {
memBytes := *ocispec.Linux.Resources.Memory.Limit
if memBytes <= 0 {
return vc.Resources{}, fmt.Errorf("Invalid OCI memory limit %d", memBytes)
}
// Use some math magic to round up to the nearest Mb.
// This has the side effect that we can never have <1Mb assigned.
resources.Memory = uint((memBytes + (1024*1024 - 1)) / (1024 * 1024))
}
return resources, nil
}
func addAssetAnnotations(ocispec CompatOCISpec, config *vc.SandboxConfig) {
assetAnnotations := []string{
vcAnnotations.KernelPath,
vcAnnotations.ImagePath,
vcAnnotations.InitrdPath,
vcAnnotations.KernelHash,
vcAnnotations.ImageHash,
vcAnnotations.AssetHashType,
}
for _, a := range assetAnnotations {
value, ok := ocispec.Annotations[a]
if !ok {
continue
}
config.Annotations[a] = value
}
}
// SandboxConfig converts an OCI compatible runtime configuration file
// to a virtcontainers sandbox configuration structure.
func SandboxConfig(ocispec CompatOCISpec, runtime RuntimeConfig, bundlePath, cid, console string, detach bool) (vc.SandboxConfig, error) {
containerConfig, err := ContainerConfig(ocispec, bundlePath, cid, console, detach)
if err != nil {
return vc.SandboxConfig{}, err
}
shmSize, err := getShmSize(containerConfig)
if err != nil {
return vc.SandboxConfig{}, err
}
networkConfig, err := networkConfig(ocispec, runtime)
if err != nil {
return vc.SandboxConfig{}, err
}
resources, err := vmConfig(ocispec, runtime)
if err != nil {
return vc.SandboxConfig{}, err
}
ociSpecJSON, err := json.Marshal(ocispec)
if err != nil {
return vc.SandboxConfig{}, err
}
sandboxConfig := vc.SandboxConfig{
ID: cid,
Hostname: ocispec.Hostname,
VMConfig: resources,
HypervisorType: runtime.HypervisorType,
HypervisorConfig: runtime.HypervisorConfig,
AgentType: runtime.AgentType,
AgentConfig: runtime.AgentConfig,
ProxyType: runtime.ProxyType,
ProxyConfig: runtime.ProxyConfig,
ShimType: runtime.ShimType,
ShimConfig: runtime.ShimConfig,
NetworkModel: vc.DefaultNetworkModel,
NetworkConfig: networkConfig,
Containers: []vc.ContainerConfig{containerConfig},
Annotations: map[string]string{
vcAnnotations.ConfigJSONKey: string(ociSpecJSON),
vcAnnotations.BundlePathKey: bundlePath,
},
ShmSize: shmSize,
}
addAssetAnnotations(ocispec, &sandboxConfig)
return sandboxConfig, nil
}
// ContainerConfig converts an OCI compatible runtime configuration
// file to a virtcontainers container configuration structure.
func ContainerConfig(ocispec CompatOCISpec, bundlePath, cid, console string, detach bool) (vc.ContainerConfig, error) {
ociSpecJSON, err := json.Marshal(ocispec)
if err != nil {
return vc.ContainerConfig{}, err
}
rootfs := ocispec.Root.Path
if !filepath.IsAbs(rootfs) {
rootfs = filepath.Join(bundlePath, ocispec.Root.Path)
}
ociLog.Debugf("container rootfs: %s", rootfs)
cmd := vc.Cmd{
Args: ocispec.Process.Args,
Envs: cmdEnvs(ocispec, []vc.EnvVar{}),
WorkDir: ocispec.Process.Cwd,
User: strconv.FormatUint(uint64(ocispec.Process.User.UID), 10),
PrimaryGroup: strconv.FormatUint(uint64(ocispec.Process.User.GID), 10),
Interactive: ocispec.Process.Terminal,
Console: console,
Detach: detach,
NoNewPrivileges: ocispec.Process.NoNewPrivileges,
}
cmd.SupplementaryGroups = []string{}
for _, gid := range ocispec.Process.User.AdditionalGids {
cmd.SupplementaryGroups = append(cmd.SupplementaryGroups, strconv.FormatUint(uint64(gid), 10))
}
deviceInfos, err := containerDeviceInfos(ocispec)
if err != nil {
return vc.ContainerConfig{}, err
}
if ocispec.Process != nil {
caps, ok := ocispec.Process.Capabilities.(vc.LinuxCapabilities)
if !ok {
return vc.ContainerConfig{}, fmt.Errorf("Unexpected format for capabilities: %v", ocispec.Process.Capabilities)
}
cmd.Capabilities = caps
}
var resources vc.ContainerResources
if ocispec.Linux.Resources.CPU != nil {
if ocispec.Linux.Resources.CPU.Quota != nil &&
ocispec.Linux.Resources.CPU.Period != nil {
resources.VCPUs = uint32(utils.ConstraintsToVCPUs(*ocispec.Linux.Resources.CPU.Quota, *ocispec.Linux.Resources.CPU.Period))
}
}
containerConfig := vc.ContainerConfig{
ID: cid,
RootFs: rootfs,
ReadonlyRootfs: ocispec.Spec.Root.Readonly,
Cmd: cmd,
Annotations: map[string]string{
vcAnnotations.ConfigJSONKey: string(ociSpecJSON),
vcAnnotations.BundlePathKey: bundlePath,
},
Mounts: containerMounts(ocispec),
DeviceInfos: deviceInfos,
Resources: resources,
}
cType, err := ocispec.ContainerType()
if err != nil {
return vc.ContainerConfig{}, err
}
containerConfig.Annotations[vcAnnotations.ContainerTypeKey] = string(cType)
return containerConfig, nil
}
func getShmSize(c vc.ContainerConfig) (uint64, error) {
var shmSize uint64
for _, m := range c.Mounts {
if m.Destination != "/dev/shm" {
continue
}
shmSize = vc.DefaultShmSize
if m.Type == "bind" && m.Source != "/dev/shm" {
var s syscall.Statfs_t
if err := syscall.Statfs(m.Source, &s); err != nil {
return 0, err
}
shmSize = uint64(s.Bsize) * s.Blocks
}
break
}
ociLog.Infof("shm-size detected: %d", shmSize)
return shmSize, nil
}
// StatusToOCIState translates a virtcontainers container status into an OCI state.
func StatusToOCIState(status vc.ContainerStatus) spec.State {
return spec.State{
Version: spec.Version,
ID: status.ID,
Status: StateToOCIState(status.State),
Pid: status.PID,
Bundle: status.Annotations[vcAnnotations.BundlePathKey],
Annotations: status.Annotations,
}
}
// StateToOCIState translates a virtcontainers container state into an OCI one.
func StateToOCIState(state vc.State) string {
switch state.State {
case vc.StateReady:
return StateCreated
case vc.StateRunning:
return StateRunning
case vc.StateStopped:
return StateStopped
case vc.StatePaused:
return StatePaused
default:
return ""
}
}
// EnvVars converts an OCI process environment variables slice
// into a virtcontainers EnvVar slice.
func EnvVars(envs []string) ([]vc.EnvVar, error) {
var envVars []vc.EnvVar
envDelimiter := "="
expectedEnvLen := 2
for _, env := range envs {
envSlice := strings.SplitN(env, envDelimiter, expectedEnvLen)
if len(envSlice) < expectedEnvLen {
return []vc.EnvVar{}, fmt.Errorf("Wrong string format: %s, expecting only %v parameters separated with %q",
env, expectedEnvLen, envDelimiter)
}
if envSlice[0] == "" {
return []vc.EnvVar{}, fmt.Errorf("Environment variable cannot be empty")
}
envSlice[1] = strings.Trim(envSlice[1], "' ")
envVar := vc.EnvVar{
Var: envSlice[0],
Value: envSlice[1],
}
envVars = append(envVars, envVar)
}
return envVars, nil
}
// GetOCIConfig returns an OCI spec configuration from the annotation
// stored into the container status.
func GetOCIConfig(status vc.ContainerStatus) (CompatOCISpec, error) {
ociConfigStr, ok := status.Annotations[vcAnnotations.ConfigJSONKey]
if !ok {
return CompatOCISpec{}, fmt.Errorf("Annotation[%s] not found", vcAnnotations.ConfigJSONKey)
}
var ociSpec CompatOCISpec
if err := json.Unmarshal([]byte(ociConfigStr), &ociSpec); err != nil {
return CompatOCISpec{}, err
}
return ociSpec, nil
}