Files
runc/libcontainer/configs/validate/validator.go
T
Xiaochen Shen 27560ace2f libcontainer: intelrdt: add support for Intel RDT/MBA in runc
Memory Bandwidth Allocation (MBA) is a resource allocation sub-feature
of Intel Resource Director Technology (RDT) which is supported on some
Intel Xeon platforms. Intel RDT/MBA provides indirect and approximate
throttle over memory bandwidth for the software. A user controls the
resource by indicating the percentage of maximum memory bandwidth.

Hardware details of Intel RDT/MBA can be found in section 17.18 of
Intel Software Developer Manual:
https://software.intel.com/en-us/articles/intel-sdm

In Linux 4.12 kernel and newer, Intel RDT/MBA is enabled by kernel
config CONFIG_INTEL_RDT. If hardware support, CPU flags `rdt_a` and
`mba` will be set in /proc/cpuinfo.

Intel RDT "resource control" filesystem hierarchy:
mount -t resctrl resctrl /sys/fs/resctrl
tree /sys/fs/resctrl
/sys/fs/resctrl/
|-- info
|   |-- L3
|   |   |-- cbm_mask
|   |   |-- min_cbm_bits
|   |   |-- num_closids
|   |-- MB
|       |-- bandwidth_gran
|       |-- delay_linear
|       |-- min_bandwidth
|       |-- num_closids
|-- ...
|-- schemata
|-- tasks
|-- <container_id>
    |-- ...
    |-- schemata
    |-- tasks

For MBA support for `runc`, we will reuse the infrastructure and code
base of Intel RDT/CAT which implemented in #1279. We could also make
use of `tasks` and `schemata` configuration for memory bandwidth
resource constraints.

The file `tasks` has a list of tasks that belongs to this group (e.g.,
<container_id>" group). Tasks can be added to a group by writing the
task ID to the "tasks" file (which will automatically remove them from
the previous group to which they belonged). New tasks created by
fork(2) and clone(2) are added to the same group as their parent.

The file `schemata` has a list of all the resources available to this
group. Each resource (L3 cache, memory bandwidth) has its own line and
format.

Memory bandwidth schema:
It has allocation values for memory bandwidth on each socket, which
contains L3 cache id and memory bandwidth percentage.
    Format: "MB:<cache_id0>=bandwidth0;<cache_id1>=bandwidth1;..."

The minimum bandwidth percentage value for each CPU model is predefined
and can be looked up through "info/MB/min_bandwidth". The bandwidth
granularity that is allocated is also dependent on the CPU model and
can be looked up at "info/MB/bandwidth_gran". The available bandwidth
control steps are: min_bw + N * bw_gran. Intermediate values are
rounded to the next control step available on the hardware.

For more information about Intel RDT kernel interface:
https://www.kernel.org/doc/Documentation/x86/intel_rdt_ui.txt

An example for runc:
Consider a two-socket machine with two L3 caches where the minimum
memory bandwidth of 10% with a memory bandwidth granularity of 10%.
Tasks inside the container may use a maximum memory bandwidth of 20%
on socket 0 and 70% on socket 1.

"linux": {
    "intelRdt": {
        "memBwSchema": "MB:0=20;1=70"
    }
}

Signed-off-by: Xiaochen Shen <xiaochen.shen@intel.com>
2018-10-16 14:29:29 +08:00

234 lines
6.8 KiB
Go

package validate
import (
"fmt"
"os"
"path/filepath"
"strings"
"github.com/opencontainers/runc/libcontainer/configs"
"github.com/opencontainers/runc/libcontainer/intelrdt"
selinux "github.com/opencontainers/selinux/go-selinux"
)
type Validator interface {
Validate(*configs.Config) error
}
func New() Validator {
return &ConfigValidator{}
}
type ConfigValidator struct {
}
func (v *ConfigValidator) Validate(config *configs.Config) error {
if err := v.rootfs(config); err != nil {
return err
}
if err := v.network(config); err != nil {
return err
}
if err := v.hostname(config); err != nil {
return err
}
if err := v.security(config); err != nil {
return err
}
if err := v.usernamespace(config); err != nil {
return err
}
if err := v.sysctl(config); err != nil {
return err
}
if err := v.intelrdt(config); err != nil {
return err
}
if config.RootlessEUID {
if err := v.rootlessEUID(config); err != nil {
return err
}
}
return nil
}
// rootfs validates if the rootfs is an absolute path and is not a symlink
// to the container's root filesystem.
func (v *ConfigValidator) rootfs(config *configs.Config) error {
if _, err := os.Stat(config.Rootfs); err != nil {
if os.IsNotExist(err) {
return fmt.Errorf("rootfs (%s) does not exist", config.Rootfs)
}
return err
}
cleaned, err := filepath.Abs(config.Rootfs)
if err != nil {
return err
}
if cleaned, err = filepath.EvalSymlinks(cleaned); err != nil {
return err
}
if filepath.Clean(config.Rootfs) != cleaned {
return fmt.Errorf("%s is not an absolute path or is a symlink", config.Rootfs)
}
return nil
}
func (v *ConfigValidator) network(config *configs.Config) error {
if !config.Namespaces.Contains(configs.NEWNET) {
if len(config.Networks) > 0 || len(config.Routes) > 0 {
return fmt.Errorf("unable to apply network settings without a private NET namespace")
}
}
return nil
}
func (v *ConfigValidator) hostname(config *configs.Config) error {
if config.Hostname != "" && !config.Namespaces.Contains(configs.NEWUTS) {
return fmt.Errorf("unable to set hostname without a private UTS namespace")
}
return nil
}
func (v *ConfigValidator) security(config *configs.Config) error {
// restrict sys without mount namespace
if (len(config.MaskPaths) > 0 || len(config.ReadonlyPaths) > 0) &&
!config.Namespaces.Contains(configs.NEWNS) {
return fmt.Errorf("unable to restrict sys entries without a private MNT namespace")
}
if config.ProcessLabel != "" && !selinux.GetEnabled() {
return fmt.Errorf("selinux label is specified in config, but selinux is disabled or not supported")
}
return nil
}
func (v *ConfigValidator) usernamespace(config *configs.Config) error {
if config.Namespaces.Contains(configs.NEWUSER) {
if _, err := os.Stat("/proc/self/ns/user"); os.IsNotExist(err) {
return fmt.Errorf("USER namespaces aren't enabled in the kernel")
}
} else {
if config.UidMappings != nil || config.GidMappings != nil {
return fmt.Errorf("User namespace mappings specified, but USER namespace isn't enabled in the config")
}
}
return nil
}
// sysctl validates that the specified sysctl keys are valid or not.
// /proc/sys isn't completely namespaced and depending on which namespaces
// are specified, a subset of sysctls are permitted.
func (v *ConfigValidator) sysctl(config *configs.Config) error {
validSysctlMap := map[string]bool{
"kernel.msgmax": true,
"kernel.msgmnb": true,
"kernel.msgmni": true,
"kernel.sem": true,
"kernel.shmall": true,
"kernel.shmmax": true,
"kernel.shmmni": true,
"kernel.shm_rmid_forced": true,
}
for s := range config.Sysctl {
if validSysctlMap[s] || strings.HasPrefix(s, "fs.mqueue.") {
if config.Namespaces.Contains(configs.NEWIPC) {
continue
} else {
return fmt.Errorf("sysctl %q is not allowed in the hosts ipc namespace", s)
}
}
if strings.HasPrefix(s, "net.") {
if config.Namespaces.Contains(configs.NEWNET) {
if path := config.Namespaces.PathOf(configs.NEWNET); path != "" {
if err := checkHostNs(s, path); err != nil {
return err
}
}
continue
} else {
return fmt.Errorf("sysctl %q is not allowed in the hosts network namespace", s)
}
}
if config.Namespaces.Contains(configs.NEWUTS) {
switch s {
case "kernel.domainname":
// This is namespaced and there's no explicit OCI field for it.
continue
case "kernel.hostname":
// This is namespaced but there's a conflicting (dedicated) OCI field for it.
return fmt.Errorf("sysctl %q is not allowed as it conflicts with the OCI %q field", s, "hostname")
}
}
return fmt.Errorf("sysctl %q is not in a separate kernel namespace", s)
}
return nil
}
func (v *ConfigValidator) intelrdt(config *configs.Config) error {
if config.IntelRdt != nil {
if !intelrdt.IsCatEnabled() && !intelrdt.IsMbaEnabled() {
return fmt.Errorf("intelRdt is specified in config, but Intel RDT is not supported or enabled")
}
if !intelrdt.IsCatEnabled() && config.IntelRdt.L3CacheSchema != "" {
return fmt.Errorf("intelRdt.l3CacheSchema is specified in config, but Intel RDT/CAT is not enabled")
}
if !intelrdt.IsMbaEnabled() && config.IntelRdt.MemBwSchema != "" {
return fmt.Errorf("intelRdt.memBwSchema is specified in config, but Intel RDT/MBA is not enabled")
}
if intelrdt.IsCatEnabled() && config.IntelRdt.L3CacheSchema == "" {
return fmt.Errorf("Intel RDT/CAT is enabled and intelRdt is specified in config, but intelRdt.l3CacheSchema is empty")
}
if intelrdt.IsMbaEnabled() && config.IntelRdt.MemBwSchema == "" {
return fmt.Errorf("Intel RDT/MBA is enabled and intelRdt is specified in config, but intelRdt.memBwSchema is empty")
}
}
return nil
}
func isSymbolicLink(path string) (bool, error) {
fi, err := os.Lstat(path)
if err != nil {
return false, err
}
return fi.Mode()&os.ModeSymlink == os.ModeSymlink, nil
}
// checkHostNs checks whether network sysctl is used in host namespace.
func checkHostNs(sysctlConfig string, path string) error {
var currentProcessNetns = "/proc/self/ns/net"
// readlink on the current processes network namespace
destOfCurrentProcess, err := os.Readlink(currentProcessNetns)
if err != nil {
return fmt.Errorf("read soft link %q error", currentProcessNetns)
}
// First check if the provided path is a symbolic link
symLink, err := isSymbolicLink(path)
if err != nil {
return fmt.Errorf("could not check that %q is a symlink: %v", path, err)
}
if symLink == false {
// The provided namespace is not a symbolic link,
// it is not the host namespace.
return nil
}
// readlink on the path provided in the struct
destOfContainer, err := os.Readlink(path)
if err != nil {
return fmt.Errorf("read soft link %q error", path)
}
if destOfContainer == destOfCurrentProcess {
return fmt.Errorf("sysctl %q is not allowed in the hosts network namespace", sysctlConfig)
}
return nil
}