merge #3876 into opencontainers/runc:main

Chethan Suresh (1):
  Support time namespace

LGTMs: kolyskin cyphar
Closes #3876

Author: cyphar

libcontainer/configs/config.go

@@ -216,6 +216,9 @@ type Config struct {
 	// Do not try to remount a bind mount again after the first attempt failed on source
 	// filesystems that have nodev, noexec, nosuid, noatime, relatime, strictatime, nodiratime set
 	NoMountFallback bool `json:"no_mount_fallback,omitempty"`
+
+	// TimeOffsets specifies the offset for supporting time namespaces.
+	TimeOffsets map[string]specs.LinuxTimeOffset `json:"time_offsets,omitempty"`
 }
 
 type (

libcontainer/configs/namespaces_linux.go

@@ -14,6 +14,7 @@ const (
 	NEWIPC    NamespaceType = "NEWIPC"
 	NEWUSER   NamespaceType = "NEWUSER"
 	NEWCGROUP NamespaceType = "NEWCGROUP"
+	NEWTIME   NamespaceType = "NEWTIME"
 )
 
 var (
@@ -38,6 +39,8 @@ func NsName(ns NamespaceType) string {
 		return "uts"
 	case NEWCGROUP:
 		return "cgroup"
+	case NEWTIME:
+		return "time"
 	}
 	return ""
 }
@@ -72,6 +75,7 @@ func NamespaceTypes() []NamespaceType {
 		NEWPID,
 		NEWNS,
 		NEWCGROUP,
+		NEWTIME,
 	}
 }
 

libcontainer/configs/namespaces_syscall.go

@@ -17,6 +17,7 @@ var namespaceInfo = map[NamespaceType]int{
 	NEWUTS:    unix.CLONE_NEWUTS,
 	NEWPID:    unix.CLONE_NEWPID,
 	NEWCGROUP: unix.CLONE_NEWCGROUP,
+	NEWTIME:   unix.CLONE_NEWTIME,
 }
 
 // CloneFlags parses the container's Namespaces options to set the correct

libcontainer/configs/validate/validator.go

@@ -116,6 +116,12 @@ func namespaces(config *configs.Config) error {
 		}
 	}
 
+	if config.Namespaces.Contains(configs.NEWTIME) {
+		if _, err := os.Stat("/proc/self/timens_offsets"); os.IsNotExist(err) {
+			return errors.New("time namespaces aren't enabled in the kernel")
+		}
+	}
+
 	return nil
 }
 

libcontainer/container_linux.go

@@ -1191,6 +1191,18 @@ func (c *Container) bootstrapData(cloneFlags uintptr, nsMaps map[configs.Namespa
 		})
 	}
 
+	// write boottime and monotonic time ns offsets.
+	if c.config.Namespaces.Contains(configs.NEWTIME) && c.config.TimeOffsets != nil {
+		var offsetSpec bytes.Buffer
+		for clock, offset := range c.config.TimeOffsets {
+			fmt.Fprintf(&offsetSpec, "%s %d %d\n", clock, offset.Secs, offset.Nanosecs)
+		}
+		r.AddData(&Bytemsg{
+			Type:  TimeOffsetsAttr,
+			Value: offsetSpec.Bytes(),
+		})
+	}
+
 	return bytes.NewReader(r.Serialize()), nil
 }
 

libcontainer/message_linux.go

@@ -23,6 +23,7 @@ const (
 	GidmapPathAttr   uint16 = 27289
 	MountSourcesAttr uint16 = 27290
 	IdmapSourcesAttr uint16 = 27291
+	TimeOffsetsAttr  uint16 = 27292
 )
 
 type Int32msg struct {

libcontainer/nsenter/namespace.h

@@ -28,5 +28,8 @@
 #ifndef CLONE_NEWNET
 #	define CLONE_NEWNET 0x40000000 /* New network namespace */
 #endif
+#ifndef CLONE_NEWTIME
+#	define CLONE_NEWTIME 0x00000080	/* New time namespace */
+#endif
 
 #endif /* NSENTER_NAMESPACE_H */

libcontainer/nsenter/nsexec.c

@@ -104,6 +104,10 @@ struct nlconfig_t {
 	/* Idmap sources opened outside the container userns which will be id mapped. */
 	char *idmapsources;
 	size_t idmapsources_len;
+
+	/* Time NS offsets. */
+	char *timensoffset;
+	size_t timensoffset_len;
 };
 
 /*
@@ -122,6 +126,7 @@ struct nlconfig_t {
 #define GIDMAPPATH_ATTR		27289
 #define MOUNT_SOURCES_ATTR	27290
 #define IDMAP_SOURCES_ATTR	27291
+#define TIMENSOFFSET_ATTR	27292
 
 /*
  * Use the raw syscall for versions of glibc which don't include a function for
@@ -351,6 +356,8 @@ static int nsflag(char *name)
 		return CLONE_NEWUSER;
 	else if (!strcmp(name, "uts"))
 		return CLONE_NEWUTS;
+	else if (!strcmp(name, "time"))
+		return CLONE_NEWTIME;
 
 	/* If we don't recognise a name, fallback to 0. */
 	return 0;
@@ -445,6 +452,10 @@ static void nl_parse(int fd, struct nlconfig_t *config)
 			config->idmapsources = current;
 			config->idmapsources_len = payload_len;
 			break;
+		case TIMENSOFFSET_ATTR:
+			config->timensoffset = current;
+			config->timensoffset_len = payload_len;
+			break;
 		default:
 			bail("unknown netlink message type %d", nlattr->nla_type);
 		}
@@ -747,6 +758,17 @@ void receive_idmapsources(int sockfd)
 	receive_fd_sources(sockfd, "_LIBCONTAINER_IDMAP_FDS");
 }
 
+static void update_timens(char *map, size_t map_len)
+{
+	if (map == NULL || map_len == 0)
+		return;
+	write_log(DEBUG, "update /proc/self/timens_offsets to '%s'", map);
+	if (write_file(map, map_len, "/proc/self/timens_offsets") < 0) {
+		if (errno != EPERM)
+			bail("failed to update /proc/self/timens_offsets");
+	}
+}
+
 void nsexec(void)
 {
 	int pipenum;
@@ -1185,6 +1207,11 @@ void nsexec(void)
 					bail("failed to sync with parent: SYNC_MOUNT_IDMAP_ACK: got %u", s);
 			}
 
+			/*
+			 * set boottime and monotonic timens offsets.
+			 */
+			update_timens(config.timensoffset, config.timensoffset_len);
+
 			/*
 			 * TODO: What about non-namespace clone flags that we're dropping here?
 			 *

libcontainer/specconv/spec_linux.go

@@ -49,6 +49,7 @@ func initMaps() {
 			specs.IPCNamespace:     configs.NEWIPC,
 			specs.UTSNamespace:     configs.NEWUTS,
 			specs.CgroupNamespace:  configs.NEWCGROUP,
+			specs.TimeNamespace:    configs.NEWTIME,
 		}
 
 		mountPropagationMapping = map[string]int{
@@ -435,6 +436,9 @@ func CreateLibcontainerConfig(opts *CreateOpts) (*configs.Config, error) {
 				MemBwSchema:   spec.Linux.IntelRdt.MemBwSchema,
 			}
 		}
+
+		// update timens offsets
+		config.TimeOffsets = spec.Linux.TimeOffsets
 	}
 
 	// Set the host UID that should own the container's cgroup.