execdriver

Documentation

Index

• Variables
• func GetAllCapabilities() []string
• func InitContainer(c *Command) *configs.Config
• func SetupCgroups(container *configs.Config, c *Command) error
• func TweakCapabilities(basics, adds, drops []string) ([]string, error)
• type Capabilities
• type CapabilityMapping
  • func GetCapability(key string) *CapabilityMapping
  • func (c *CapabilityMapping) String() string
• type Command
• type Context
• type Driver
• type DriverCallback
• type ExitStatus
• type Hooks
• type Info
• type Ipc
• type Mount
• type Network
• type Pid
• type Pipes
  • func NewPipes(stdin io.ReadCloser, stdout, stderr io.Writer, useStdin bool) *Pipes
• type ProcessConfig
• type ResourceStats
  • func Stats(containerDir string, containerMemoryLimit int64, machineMemory int64) (*ResourceStats, error)
• type Resources
• type StdConsole
  • func NewStdConsole(processConfig *ProcessConfig, pipes *Pipes) (*StdConsole, error)
  • func (s *StdConsole) AttachPipes(command *exec.Cmd, pipes *Pipes) error
  • func (s *StdConsole) Close() error
  • func (s *StdConsole) Resize(h, w int) error
• type Terminal
• type UTS

Variables

var (
	ErrNotRunning              = errors.New("Container is not running")
	ErrWaitTimeoutReached      = errors.New("Wait timeout reached")
	ErrDriverAlreadyRegistered = errors.New("A driver already registered this docker init function")
	ErrDriverNotFound          = errors.New("The requested docker init has not been found")
)

Define error messages

Functions

func GetAllCapabilities

func GetAllCapabilities() []string

GetAllCapabilities returns all of the capabilities

func InitContainer

func InitContainer(c *Command) *configs.Config

InitContainer is the initialization of a container config. It returns the initial configs for a container. It's mostly defined by the default template.

func SetupCgroups

func SetupCgroups(container *configs.Config, c *Command) error

SetupCgroups setups cgroup resources for a container.

func TweakCapabilities

func TweakCapabilities(basics, adds, drops []string) ([]string, error)

TweakCapabilities can tweak capabilities by adding or dropping capabilities based on the basics capabilities.

Types

type Capabilities

type Capabilities []*CapabilityMapping

Capabilities contains all CapabilityMapping

type CapabilityMapping

type CapabilityMapping struct {
	Key   string         `json:"key,omitempty"`
	Value capability.Cap `json:"value,omitempty"`
}

CapabilityMapping maps linux capability name to its value of capability.Cap type Capabilities is one of the security systems in Linux Security Module (LSM) framework provided by the kernel. For more details on capabilities, see http://man7.org/linux/man-pages/man7/capabilities.7.html

func GetCapability

func GetCapability(key string) *CapabilityMapping

GetCapability returns CapabilityMapping which contains specific key

func (*CapabilityMapping) String

func (c *CapabilityMapping) String() string

String returns <key> of CapabilityMapping

type Command

type Command struct {
	ID                 string            `json:"id"`
	Rootfs             string            `json:"rootfs"` // root fs of the container
	ReadonlyRootfs     bool              `json:"readonly_rootfs"`
	InitPath           string            `json:"initpath"` // dockerinit
	WorkingDir         string            `json:"working_dir"`
	ConfigPath         string            `json:"config_path"` // this should be able to be removed when the lxc template is moved into the driver
	Network            *Network          `json:"network"`
	Ipc                *Ipc              `json:"ipc"`
	Pid                *Pid              `json:"pid"`
	UTS                *UTS              `json:"uts"`
	Resources          *Resources        `json:"resources"`
	Mounts             []Mount           `json:"mounts"`
	AllowedDevices     []*configs.Device `json:"allowed_devices"`
	AutoCreatedDevices []*configs.Device `json:"autocreated_devices"`
	CapAdd             []string          `json:"cap_add"`
	CapDrop            []string          `json:"cap_drop"`
	GroupAdd           []string          `json:"group_add"`
	ContainerPid       int               `json:"container_pid"`  // the pid for the process inside a container
	ProcessConfig      ProcessConfig     `json:"process_config"` // Describes the init process of the container.
	ProcessLabel       string            `json:"process_label"`
	MountLabel         string            `json:"mount_label"`
	LxcConfig          []string          `json:"lxc_config"`
	AppArmorProfile    string            `json:"apparmor_profile"`
	CgroupParent       string            `json:"cgroup_parent"` // The parent cgroup for this command.
	FirstStart         bool              `json:"first_start"`
	LayerPaths         []string          `json:"layer_paths"` // Windows needs to know the layer paths and folder for a command
	LayerFolder        string            `json:"layer_folder"`
}

Command wraps an os/exec.Cmd to add more metadata

TODO Windows: Factor out unused fields such as LxcConfig, AppArmorProfile, and CgroupParent.

type Context

type Context map[string]string

Context is a generic key value pair that allows arbatrary data to be sent

type Driver

type Driver interface {
	// Run executes the process, blocks until the process exits and returns
	// the exit code. It's the last stage on Docker side for running a container.
	Run(c *Command, pipes *Pipes, hooks Hooks) (ExitStatus, error)

	// Exec executes the process in an existing container, blocks until the
	// process exits and returns the exit code.
	Exec(c *Command, processConfig *ProcessConfig, pipes *Pipes, hooks Hooks) (int, error)

	// Kill sends signals to process in container.
	Kill(c *Command, sig int) error

	// Pause pauses a container.
	Pause(c *Command) error

	// Unpause unpauses a container.
	Unpause(c *Command) error

	// Name returns the name of the driver.
	Name() string

	// Info returns the configuration stored in the driver struct,
	// "temporary" hack (until we move state from core to plugins).
	Info(id string) Info

	// GetPidsForContainer returns a list of pid for the processes running in a container.
	GetPidsForContainer(id string) ([]int, error)

	// Terminate kills a container by sending signal SIGKILL.
	Terminate(c *Command) error

	// Clean removes all traces of container exec.
	Clean(id string) error

	// Stats returns resource stats for a running container
	Stats(id string) (*ResourceStats, error)

	// SupportsHooks refers to the driver capability to exploit pre/post hook functionality
	SupportsHooks() bool
}

Driver is an interface for drivers to implement including all basic functions a driver should have

type DriverCallback

type DriverCallback func(processConfig *ProcessConfig, pid int, chOOM <-chan struct{}) error

DriverCallback defines a callback function which is used in "Run" and "Exec". This allows work to be done in the parent process when the child is passing through PreStart, Start and PostStop events. Callbacks are provided a processConfig pointer and the pid of the child. The channel will be used to notify the OOM events.

type ExitStatus

type ExitStatus struct {
	// The exit code with which the container exited.
	ExitCode int

	// Whether the container encountered an OOM.
	OOMKilled bool
}

ExitStatus provides exit reasons for a container.

type Hooks

type Hooks struct {
	// PreStart is called before container's CMD/ENTRYPOINT is executed
	PreStart []DriverCallback
	// Start is called after the container's process is full started
	Start DriverCallback
	// PostStop is called after the container process exits
	PostStop []DriverCallback
}

Hooks is a struct containing function pointers to callbacks used by any execdriver implementation exploiting hooks capabilities

type Info

type Info interface {
	IsRunning() bool
}

Info is driver specific information based on processes registered with the driver

type Ipc

type Ipc struct {
	ContainerID string `json:"container_id"` // id of the container to join ipc.
	HostIpc     bool   `json:"host_ipc"`
}

Ipc settings of the container It is for IPC namespace setting. Usually different containers have their own IPC namespace, however this specifies to use an existing IPC namespace. You can join the host's or a container's IPC namespace.

type Mount

type Mount struct {
	Source      string `json:"source"`
	Destination string `json:"destination"`
	Writable    bool   `json:"writable"`
	Private     bool   `json:"private"`
	Slave       bool   `json:"slave"`
}

Mount contains information for a mount operation.

type Network

type Network struct {
	Mtu            int    `json:"mtu"`
	ContainerID    string `json:"container_id"` // id of the container to join network.
	NamespacePath  string `json:"namespace_path"`
	HostNetworking bool   `json:"host_networking"`
}

Network settings of the container

type Pid

type Pid struct {
	HostPid bool `json:"host_pid"`
}

Pid settings of the container It is for PID namespace setting. Usually different containers have their own PID namespace, however this specifies to use an existing PID namespace. Joining the host's PID namespace is currently the only supported option.

type Pipes

type Pipes struct {
	Stdin          io.ReadCloser
	Stdout, Stderr io.Writer
}

Pipes is a wrapper around a container's output for stdin, stdout, stderr

func NewPipes

func NewPipes(stdin io.ReadCloser, stdout, stderr io.Writer, useStdin bool) *Pipes

NewPipes returns a wrapper around a container's output

type ProcessConfig

type ProcessConfig struct {
	exec.Cmd `json:"-"`

	Privileged  bool     `json:"privileged"`
	User        string   `json:"user"`
	Tty         bool     `json:"tty"`
	Entrypoint  string   `json:"entrypoint"`
	Arguments   []string `json:"arguments"`
	Terminal    Terminal `json:"-"` // standard or tty terminal
	Console     string   `json:"-"` // dev/console path
	ConsoleSize [2]int   `json:"-"` // h,w of initial console size
}

ProcessConfig describes a process that will be run inside a container.

type ResourceStats

type ResourceStats struct {
	*libcontainer.Stats
	Read        time.Time `json:"read"`
	MemoryLimit int64     `json:"memory_limit"`
	SystemUsage uint64    `json:"system_usage"`
}

ResourceStats contains information about resource usage by a container.

func Stats

func Stats(containerDir string, containerMemoryLimit int64, machineMemory int64) (*ResourceStats, error)

Stats collects all the resource usage information from a container.

type Resources

type Resources struct {
	Memory            int64            `json:"memory"`
	MemorySwap        int64            `json:"memory_swap"`
	MemoryReservation int64            `json:"memory_reservation"`
	KernelMemory      int64            `json:"kernel_memory"`
	CPUShares         int64            `json:"cpu_shares"`
	CpusetCpus        string           `json:"cpuset_cpus"`
	CpusetMems        string           `json:"cpuset_mems"`
	CPUPeriod         int64            `json:"cpu_period"`
	CPUQuota          int64            `json:"cpu_quota"`
	BlkioWeight       int64            `json:"blkio_weight"`
	Rlimits           []*ulimit.Rlimit `json:"rlimits"`
	OomKillDisable    bool             `json:"oom_kill_disable"`
	MemorySwappiness  int64            `json:"memory_swappiness"`
}

Resources contains all resource configs for a driver. Currently these are all for cgroup configs. TODO Windows: Factor out ulimit.Rlimit

type StdConsole

type StdConsole struct {
}

StdConsole defines standard console operations for execdriver

func NewStdConsole

func NewStdConsole(processConfig *ProcessConfig, pipes *Pipes) (*StdConsole, error)

NewStdConsole returns a new StdConsole struct

func (*StdConsole) AttachPipes

func (s *StdConsole) AttachPipes(command *exec.Cmd, pipes *Pipes) error

AttachPipes attaches given pipes to exec.Cmd

func (*StdConsole) Close

func (s *StdConsole) Close() error

Close implements Close method of Terminal interface

func (*StdConsole) Resize

func (s *StdConsole) Resize(h, w int) error

Resize implements Resize method of Terminal interface

type Terminal

type Terminal interface {
	io.Closer
	Resize(height, width int) error
}

Terminal represents a pseudo TTY, it is for when using a container interactively.

type UTS

type UTS struct {
	HostUTS bool `json:"host_uts"`
}

UTS settings of the container It is for UTS namespace setting. Usually different containers have their own UTS namespace, however this specifies to use an existing UTS namespace. Joining the host's UTS namespace is currently the only supported option.