Merge pull request #1856 from fidencio/2.1.0-branch-bump

# Kata Containers 2.1.0

VERSION

@@ -1 +1 @@
-2.1.0-rc0
+2.1.0

ci/install_yq.sh

@@ -18,7 +18,9 @@ function install_yq() {
 	GOPATH=${GOPATH:-${HOME}/go}
 	local yq_path="${GOPATH}/bin/yq"
 	local yq_pkg="github.com/mikefarah/yq"
-	[ -x  "${GOPATH}/bin/yq" ] && return
+	local yq_version=3.4.1
+
+	[ -x  "${GOPATH}/bin/yq" ] && [ "`${GOPATH}/bin/yq --version`"X == "yq version ${yq_version}"X ] && return
 
 	read -r -a sysInfo <<< "$(uname -sm)"
 
@@ -56,8 +58,6 @@ function install_yq() {
 		die "Please install curl"
 	fi
 
-	local yq_version=3.4.1
-
 	## NOTE: ${var,,} => gives lowercase value of var
 	local yq_url="https://${yq_pkg}/releases/download/${yq_version}/yq_${goos,,}_${goarch}"
 	curl -o "${yq_path}" -LSsf "${yq_url}"

docs/Limitations.md

@@ -28,7 +28,7 @@
 * [Appendices](#appendices)
     * [The constraints challenge](#the-constraints-challenge)
 
----
+***
 
 # Overview
 
@@ -94,7 +94,9 @@ This section lists items that might be possible to fix.
 ### checkpoint and restore
 
 The runtime does not provide `checkpoint` and `restore` commands. There
-are discussions about using VM save and restore to give [`criu`](https://github.com/checkpoint-restore/criu)-like functionality, which might provide a solution.
+are discussions about using VM save and restore to give us a
+`[criu](https://github.com/checkpoint-restore/criu)`-like functionality,
+which might provide a solution.
 
 Note that the OCI standard does not specify `checkpoint` and `restore`
 commands.

docs/how-to/how-to-set-sandbox-config-kata.md

@@ -26,6 +26,7 @@ There are several kinds of Kata configurations and they are listed below.
 | `io.katacontainers.config.runtime.disable_new_netns` | `boolean` | determines if a new netns is created for the hypervisor process |
 | `io.katacontainers.config.runtime.internetworking_model` | string| determines how the VM should be connected to the container network interface. Valid values are `macvtap`, `tcfilter` and `none` |
 | `io.katacontainers.config.runtime.sandbox_cgroup_only`| `boolean` | determines if Kata processes are managed only in sandbox cgroup |
+| `io.katacontainers.config.runtime.enable_pprof` | `boolean` | enables Golang `pprof` for `containerd-shim-kata-v2` process |
 
 ## Agent Options
 | Key | Value Type | Comments |

src/agent/kata-agent.service.in

@@ -15,7 +15,7 @@ Wants=kata-containers.target
 StandardOutput=tty
 Type=simple
 ExecStart=@BINDIR@/@AGENT_NAME@
-LimitNOFILE=infinity
+LimitNOFILE=1048576
 # ExecStop is required for static agent tracing; in all other scenarios
 # the runtime handles shutting down the VM.
 ExecStop=/bin/sync ; /usr/bin/systemctl --force poweroff

src/agent/rustjail/src/container.rs

@@ -48,6 +48,7 @@ use oci::State as OCIState;
 use std::collections::HashMap;
 use std::os::unix::io::FromRawFd;
 use std::str::FromStr;
+use std::sync::Arc;
 
 use slog::{info, o, Logger};
 
@@ -57,6 +58,7 @@ use crate::sync_with_async::{read_async, write_async};
 use async_trait::async_trait;
 use rlimit::{setrlimit, Resource, Rlim};
 use tokio::io::AsyncBufReadExt;
+use tokio::sync::Mutex;
 
 use crate::utils;
 
@@ -106,6 +108,9 @@ pub type Config = CreateOpts;
 type NamespaceType = String;
 
 lazy_static! {
+    // This locker ensures the child exit signal will be received by the right receiver.
+    pub static ref WAIT_PID_LOCKER: Arc<Mutex<bool>> = Arc::new(Mutex::new(false));
+
     static ref NAMESPACES: HashMap<&'static str, CloneFlags> = {
         let mut m = HashMap::new();
         m.insert("user", CloneFlags::CLONE_NEWUSER);
@@ -1467,6 +1472,8 @@ async fn execute_hook(logger: &Logger, h: &Hook, st: &OCIState) -> Result<()> {
         })
         .collect();
 
+    // Avoid the exit signal to be reaped by the global reaper.
+    let _wait_locker = WAIT_PID_LOCKER.lock().await;
     let mut child = tokio::process::Command::new(path)
         .args(args.iter())
         .envs(env.iter())

src/agent/rustjail/src/mount.rs

@@ -68,6 +68,8 @@ lazy_static! {
         m.insert("rprivate", MsFlags::MS_PRIVATE | MsFlags::MS_REC);
         m.insert("slave", MsFlags::MS_SLAVE);
         m.insert("rslave", MsFlags::MS_SLAVE | MsFlags::MS_REC);
+        m.insert("unbindable", MsFlags::MS_UNBINDABLE);
+        m.insert("runbindable", MsFlags::MS_UNBINDABLE | MsFlags::MS_REC);
         m
     };
     static ref OPTIONS: HashMap<&'static str, (bool, MsFlags)> = {
@@ -93,17 +95,6 @@ lazy_static! {
         m.insert("nodiratime", (false, MsFlags::MS_NODIRATIME));
         m.insert("bind", (false, MsFlags::MS_BIND));
         m.insert("rbind", (false, MsFlags::MS_BIND | MsFlags::MS_REC));
-        m.insert("unbindable", (false, MsFlags::MS_UNBINDABLE));
-        m.insert(
-            "runbindable",
-            (false, MsFlags::MS_UNBINDABLE | MsFlags::MS_REC),
-        );
-        m.insert("private", (false, MsFlags::MS_PRIVATE));
-        m.insert("rprivate", (false, MsFlags::MS_PRIVATE | MsFlags::MS_REC));
-        m.insert("shared", (false, MsFlags::MS_SHARED));
-        m.insert("rshared", (false, MsFlags::MS_SHARED | MsFlags::MS_REC));
-        m.insert("slave", (false, MsFlags::MS_SLAVE));
-        m.insert("rslave", (false, MsFlags::MS_SLAVE | MsFlags::MS_REC));
         m.insert("relatime", (false, MsFlags::MS_RELATIME));
         m.insert("norelatime", (true, MsFlags::MS_RELATIME));
         m.insert("strictatime", (false, MsFlags::MS_STRICTATIME));
@@ -192,7 +183,7 @@ pub fn init_rootfs(
 
     let mut bind_mount_dev = false;
     for m in &spec.mounts {
-        let (mut flags, data) = parse_mount(&m);
+        let (mut flags, pgflags, data) = parse_mount(&m);
         if !m.destination.starts_with('/') || m.destination.contains("..") {
             return Err(anyhow!(
                 "the mount destination {} is invalid",
@@ -234,13 +225,15 @@ pub fn init_rootfs(
             // effective.
             // first check that we have non-default options required before attempting a
             // remount
-            if m.r#type == "bind" {
-                for o in &m.options {
-                    if let Some(fl) = PROPAGATION.get(o.as_str()) {
-                        let dest = secure_join(rootfs, &m.destination);
-                        mount(None::<&str>, dest.as_str(), None::<&str>, *fl, None::<&str>)?;
-                    }
-                }
+            if m.r#type == "bind" && !pgflags.is_empty() {
+                let dest = secure_join(rootfs, &m.destination);
+                mount(
+                    None::<&str>,
+                    dest.as_str(),
+                    None::<&str>,
+                    pgflags,
+                    None::<&str>,
+                )?;
             }
         }
     }
@@ -657,26 +650,27 @@ pub fn ms_move_root(rootfs: &str) -> Result<bool> {
     Ok(true)
 }
 
-fn parse_mount(m: &Mount) -> (MsFlags, String) {
+fn parse_mount(m: &Mount) -> (MsFlags, MsFlags, String) {
     let mut flags = MsFlags::empty();
+    let mut pgflags = MsFlags::empty();
     let mut data = Vec::new();
 
     for o in &m.options {
-        match OPTIONS.get(o.as_str()) {
-            Some(v) => {
-                let (clear, fl) = *v;
-                if clear {
-                    flags &= !fl;
-                } else {
-                    flags |= fl;
-                }
+        if let Some(v) = OPTIONS.get(o.as_str()) {
+            let (clear, fl) = *v;
+            if clear {
+                flags &= !fl;
+            } else {
+                flags |= fl;
             }
-
-            None => data.push(o.clone()),
+        } else if let Some(fl) = PROPAGATION.get(o.as_str()) {
+            pgflags |= *fl;
+        } else {
+            data.push(o.clone());
         }
     }
 
-    (flags, data.join(","))
+    (flags, pgflags, data.join(","))
 }
 
 // This function constructs a canonicalized path by combining the `rootfs` and `unsafe_path` elements.
@@ -922,7 +916,7 @@ pub fn finish_rootfs(cfd_log: RawFd, spec: &Spec) -> Result<()> {
 
     for m in spec.mounts.iter() {
         if m.destination == "/dev" {
-            let (flags, _) = parse_mount(m);
+            let (flags, _, _) = parse_mount(m);
             if flags.contains(MsFlags::MS_RDONLY) {
                 mount(
                     Some("/dev"),

src/agent/src/signal.rs

@@ -22,6 +22,9 @@ async fn handle_sigchild(logger: Logger, sandbox: Arc<Mutex<Sandbox>>) -> Result
     info!(logger, "handling signal"; "signal" => "SIGCHLD");
 
     loop {
+        // Avoid reaping the undesirable child's signal, e.g., execute_hook's
+        // The lock should be released immediately.
+        rustjail::container::WAIT_PID_LOCKER.lock().await;
         let result = wait::waitpid(
             Some(Pid::from_raw(-1)),
             Some(WaitPidFlag::WNOHANG | WaitPidFlag::__WALL),

src/runtime/cli/config/configuration-acrn.toml.in

@@ -150,6 +150,10 @@ block_device_driver = "@DEFBLOCKSTORAGEDRIVER_ACRN@"
 
 #debug_console_enabled = true
 
+# Agent connection dialing timeout value in seconds
+# (default: 30)
+#dial_timeout = 30
+
 [netmon]
 # If enabled, the network monitoring process gets started when the
 # sandbox is created. This allows for the detection of some additional

src/runtime/cli/config/configuration-clh.toml.in

@@ -165,6 +165,10 @@ block_device_driver = "virtio-blk"
 
 #debug_console_enabled = true
 
+# Agent connection dialing timeout value in seconds
+# (default: 30)
+#dial_timeout = 30
+
 [netmon]
 # If enabled, the network monitoring process gets started when the
 # sandbox is created. This allows for the detection of some additional

src/runtime/cli/config/configuration-fc.toml.in

@@ -287,6 +287,10 @@ kernel_modules=[]
 
 #debug_console_enabled = true
 
+# Agent connection dialing timeout value in seconds
+# (default: 30)
+#dial_timeout = 30
+
 [netmon]
 # If enabled, the network monitoring process gets started when the
 # sandbox is created. This allows for the detection of some additional

src/runtime/cli/config/configuration-qemu.toml.in

@@ -437,6 +437,10 @@ kernel_modules=[]
 
 #debug_console_enabled = true
 
+# Agent connection dialing timeout value in seconds
+# (default: 30)
+#dial_timeout = 30
+
 [netmon]
 # If enabled, the network monitoring process gets started when the
 # sandbox is created. This allows for the detection of some additional

src/runtime/cli/kata-check.go

@@ -389,13 +389,6 @@ EXAMPLES:
 		if verbose {
 			kataLog.Logger.SetLevel(logrus.InfoLevel)
 		}
-		ctx, err := cliContextToContext(context)
-		if err != nil {
-			return err
-		}
-
-		span, _ := katautils.Trace(ctx, "check")
-		defer span.End()
 
 		if !context.Bool("no-network-checks") && os.Getenv(noNetworkEnvVar) == "" {
 			cmd := RelCmdCheck
@@ -407,8 +400,7 @@ EXAMPLES:
 			if os.Geteuid() == 0 {
 				kataLog.Warn("Not running network checks as super user")
 			} else {
-
-				err = HandleReleaseVersions(cmd, version, context.Bool("include-all-releases"))
+				err := HandleReleaseVersions(cmd, version, context.Bool("include-all-releases"))
 				if err != nil {
 					return err
 				}
@@ -424,7 +416,7 @@ EXAMPLES:
 			return errors.New("check: cannot determine runtime config")
 		}
 
-		err = setCPUtype(runtimeConfig.HypervisorType)
+		err := setCPUtype(runtimeConfig.HypervisorType)
 		if err != nil {
 			return err
 		}
@@ -437,7 +429,6 @@ EXAMPLES:
 		}
 
 		err = hostIsVMContainerCapable(details)
-
 		if err != nil {
 			return err
 		}

src/runtime/cli/kata-env.go

@@ -13,7 +13,6 @@ import (
 	"strings"
 
 	"github.com/BurntSushi/toml"
-	"github.com/kata-containers/kata-containers/src/runtime/pkg/katautils"
 	"github.com/kata-containers/kata-containers/src/runtime/pkg/utils"
 	vc "github.com/kata-containers/kata-containers/src/runtime/virtcontainers"
 	exp "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/experimental"
@@ -448,14 +447,6 @@ var kataEnvCLICommand = cli.Command{
 		},
 	},
 	Action: func(context *cli.Context) error {
-		ctx, err := cliContextToContext(context)
-		if err != nil {
-			return err
-		}
-
-		span, _ := katautils.Trace(ctx, "kata-env")
-		defer span.End()
-
 		return handleSettings(defaultOutputFile, context)
 	},
 }

src/runtime/cli/kata-exec.go

@@ -14,7 +14,6 @@ import (
 	"net/http"
 	"net/url"
 	"os"
-	"path/filepath"
 	"strings"
 
 	"sync"
@@ -26,7 +25,6 @@ import (
 	clientUtils "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/agent/protocols/client"
 	"github.com/pkg/errors"
 	"github.com/urfave/cli"
-	"go.opentelemetry.io/otel/label"
 )
 
 const (
@@ -38,10 +36,8 @@ const (
 
 	subCommandName = "exec"
 	// command-line parameters name
-	paramRuntimeNamespace                    = "runtime-namespace"
 	paramDebugConsolePort                    = "kata-debug-port"
 	defaultKernelParamDebugConsoleVPortValue = 1026
-	defaultRuntimeNamespace                  = "k8s.io"
 )
 
 var (
@@ -57,34 +53,16 @@ var kataExecCLICommand = cli.Command{
 	Name:  subCommandName,
 	Usage: "Enter into guest by debug console",
 	Flags: []cli.Flag{
-		cli.StringFlag{
-			Name:  paramRuntimeNamespace,
-			Usage: "Namespace that containerd or CRI-O are using for containers. (Default: k8s.io, only works for containerd)",
-		},
 		cli.Uint64Flag{
 			Name:  paramDebugConsolePort,
 			Usage: "Port that debug console is listening on. (Default: 1026)",
 		},
 	},
 	Action: func(context *cli.Context) error {
-		ctx, err := cliContextToContext(context)
-		if err != nil {
-			return err
-		}
-		span, _ := katautils.Trace(ctx, subCommandName)
-		defer span.End()
-
-		namespace := context.String(paramRuntimeNamespace)
-		if namespace == "" {
-			namespace = defaultRuntimeNamespace
-		}
-		span.SetAttributes(label.Key("namespace").String(namespace))
-
 		port := context.Uint64(paramDebugConsolePort)
 		if port == 0 {
 			port = defaultKernelParamDebugConsoleVPortValue
 		}
-		span.SetAttributes(label.Key("port").Uint64(port))
 
 		sandboxID := context.Args().Get(0)
 
@@ -92,9 +70,8 @@ var kataExecCLICommand = cli.Command{
 			return err
 		}
 
-		span.SetAttributes(label.Key("sandbox").String(sandboxID))
+		conn, err := getConn(sandboxID, port)
 
-		conn, err := getConn(namespace, sandboxID, port)
 		if err != nil {
 			return err
 		}
@@ -177,9 +154,8 @@ func (s *iostream) Read(data []byte) (n int, err error) {
 	return s.conn.Read(data)
 }
 
-func getConn(namespace, sandboxID string, port uint64) (net.Conn, error) {
-	socketAddr := filepath.Join(string(filepath.Separator), "containerd-shim", namespace, sandboxID, "shim-monitor.sock")
-	client, err := kataMonitor.BuildUnixSocketClient(socketAddr, defaultTimeout)
+func getConn(sandboxID string, port uint64) (net.Conn, error) {
+	client, err := kataMonitor.BuildShimClient(sandboxID, defaultTimeout)
 	if err != nil {
 		return nil, err
 	}
@@ -190,7 +166,7 @@ func getConn(namespace, sandboxID string, port uint64) (net.Conn, error) {
 	}
 
 	if resp.StatusCode != http.StatusOK {
-		return nil, fmt.Errorf("Failed to get %s: %d", socketAddr, resp.StatusCode)
+		return nil, fmt.Errorf("Failure from %s shim-monitor: %d", sandboxID, resp.StatusCode)
 	}
 
 	defer resp.Body.Close()

src/runtime/cli/kata-metrics.go

@@ -0,0 +1,38 @@
+// Copyright (c) 2021 Apple Inc.
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+package main
+
+import (
+	"fmt"
+
+	kataMonitor "github.com/kata-containers/kata-containers/src/runtime/pkg/kata-monitor"
+	"github.com/kata-containers/kata-containers/src/runtime/pkg/katautils"
+	"github.com/urfave/cli"
+)
+
+var kataMetricsCLICommand = cli.Command{
+	Name:      "metrics",
+	Usage:     "gather metrics associated with infrastructure used to run a sandbox",
+	UsageText: "metrics <sandbox id>",
+	Action: func(context *cli.Context) error {
+
+		sandboxID := context.Args().Get(0)
+
+		if err := katautils.VerifyContainerID(sandboxID); err != nil {
+			return err
+		}
+
+		// Get the metrics!
+		metrics, err := kataMonitor.GetSandboxMetrics(sandboxID)
+		if err != nil {
+			return err
+		}
+
+		fmt.Printf("%s\n", metrics)
+
+		return nil
+	},
+}

src/runtime/cli/main.go

@@ -27,9 +27,6 @@ import (
 	specs "github.com/opencontainers/runtime-spec/specs-go"
 	"github.com/sirupsen/logrus"
 	"github.com/urfave/cli"
-	"go.opentelemetry.io/otel"
-	"go.opentelemetry.io/otel/label"
-	otelTrace "go.opentelemetry.io/otel/trace"
 )
 
 // specConfig is the name of the file holding the containers configuration
@@ -125,6 +122,7 @@ var runtimeCommands = []cli.Command{
 	kataCheckCLICommand,
 	kataEnvCLICommand,
 	kataExecCLICommand,
+	kataMetricsCLICommand,
 	factoryCLICommand,
 }
 
@@ -132,10 +130,6 @@ var runtimeCommands = []cli.Command{
 // parsing occurs.
 var runtimeBeforeSubcommands = beforeSubcommands
 
-// runtimeAfterSubcommands is the function to run after the command-line
-// has been parsed.
-var runtimeAfterSubcommands = afterSubcommands
-
 // runtimeCommandNotFound is the function to handle an invalid sub-command.
 var runtimeCommandNotFound = commandNotFound
 
@@ -168,10 +162,6 @@ func init() {
 
 // setupSignalHandler sets up signal handling, starting a go routine to deal
 // with signals as they arrive.
-//
-// Note that the specified context is NOT used to create a trace span (since the
-// first (root) span must be created in beforeSubcommands()): it is simply
-// used to pass to the crash handling functions to finalise tracing.
 func setupSignalHandler(ctx context.Context) {
 	signals.SetLogger(kataLog)
 
@@ -181,10 +171,6 @@ func setupSignalHandler(ctx context.Context) {
 		signal.Notify(sigCh, sig)
 	}
 
-	dieCb := func() {
-		katautils.StopTracing(ctx)
-	}
-
 	go func() {
 		for {
 			sig := <-sigCh
@@ -198,7 +184,6 @@ func setupSignalHandler(ctx context.Context) {
 
 			if signals.FatalSignal(nativeSignal) {
 				kataLog.WithField("signal", sig).Error("received fatal signal")
-				signals.Die(dieCb)
 			} else if debug && signals.NonFatalSignal(nativeSignal) {
 				kataLog.WithField("signal", sig).Debug("handling signal")
 				signals.Backtrace()
@@ -210,18 +195,6 @@ func setupSignalHandler(ctx context.Context) {
 // setExternalLoggers registers the specified logger with the external
 // packages which accept a logger to handle their own logging.
 func setExternalLoggers(ctx context.Context, logger *logrus.Entry) {
-	var span otelTrace.Span
-
-	// Only create a new span if a root span already exists. This is
-	// required to ensure that this function will not disrupt the root
-	// span logic by creating a span before the proper root span has been
-	// created.
-
-	if otelTrace.SpanFromContext(ctx) != nil {
-		span, ctx = katautils.Trace(ctx, "setExternalLoggers")
-		defer span.End()
-	}
-
 	// Set virtcontainers logger.
 	vci.SetLogger(ctx, logger)
 
@@ -244,7 +217,6 @@ func beforeSubcommands(c *cli.Context) error {
 	var configFile string
 	var runtimeConfig oci.RuntimeConfig
 	var err error
-	var traceFlushFunc func()
 
 	katautils.SetConfigOptions(name, defaultRuntimeConfiguration, defaultSysConfRuntimeConfiguration)
 
@@ -270,7 +242,6 @@ func beforeSubcommands(c *cli.Context) error {
 	// Issue: https://github.com/kata-containers/runtime/issues/2428
 
 	ignoreConfigLogs := false
-	var traceRootSpan string
 
 	subCmdIsCheckCmd := (c.NArg() >= 1 && ((c.Args()[0] == "kata-check") || (c.Args()[0] == "check")))
 	if subCmdIsCheckCmd {
@@ -302,16 +273,13 @@ func beforeSubcommands(c *cli.Context) error {
 		cmdName := c.Args().First()
 		if c.App.Command(cmdName) != nil {
 			kataLog = kataLog.WithField("command", cmdName)
-
-			// Name for the root span (used for tracing) now the
-			// sub-command name is known.
-			traceRootSpan = name + " " + cmdName
 		}
 
-		// Since a context is required, pass a new (throw-away) one - we
-		// cannot use the main context as tracing hasn't been enabled yet
-		// (meaning any spans created at this point will be silently ignored).
-		setExternalLoggers(context.Background(), kataLog)
+		ctx, err := cliContextToContext(c)
+		if err != nil {
+			return err
+		}
+		setExternalLoggers(ctx, kataLog)
 
 		if c.NArg() == 1 && (c.Args()[0] == "kata-env" || c.Args()[0] == "env") {
 			// simply report the logging setup
@@ -325,20 +293,6 @@ func beforeSubcommands(c *cli.Context) error {
 	}
 	if !subCmdIsCheckCmd {
 		debug = runtimeConfig.Debug
-
-		if traceRootSpan != "" {
-			// Create the tracer.
-			//
-			// Note: no spans are created until the command-line has been parsed.
-			// This delays collection of trace data slightly but benefits the user by
-			// ensuring the first span is the name of the sub-command being
-			// invoked from the command-line.
-			traceFlushFunc, err = setupTracing(c, traceRootSpan, &runtimeConfig)
-			if err != nil {
-				return err
-			}
-			defer traceFlushFunc()
-		}
 	}
 
 	args := strings.Join(c.Args(), " ")
@@ -377,36 +331,6 @@ func handleShowConfig(context *cli.Context) {
 	}
 }
 
-func setupTracing(context *cli.Context, rootSpanName string, config *oci.RuntimeConfig) (func(), error) {
-	flush, err := katautils.CreateTracer(name, config)
-	if err != nil {
-		return nil, err
-	}
-
-	ctx, err := cliContextToContext(context)
-	if err != nil {
-		return nil, err
-	}
-
-	// Create the root span now that the sub-command name is
-	// known.
-	//
-	// Note that this "Before" function is called (and returns)
-	// before the subcommand handler is called. As such, we cannot
-	// "Finish()" the span here - that is handled in the .After
-	// function.
-	tracer := otel.Tracer("kata")
-	newCtx, span := tracer.Start(ctx, rootSpanName)
-
-	span.SetAttributes(label.Key("subsystem").String("runtime"))
-
-	// Add tracer to metadata and update the context
-	context.App.Metadata["tracer"] = tracer
-	context.App.Metadata["context"] = newCtx
-
-	return flush, nil
-}
-
 // add supported experimental features in context
 func addExpFeatures(clictx *cli.Context, runtimeConfig oci.RuntimeConfig) error {
 	ctx, err := cliContextToContext(clictx)
@@ -420,22 +344,11 @@ func addExpFeatures(clictx *cli.Context, runtimeConfig oci.RuntimeConfig) error
 	}
 
 	ctx = exp.ContextWithExp(ctx, exps)
-	// Add tracer to metadata and update the context
+	// Add experimental features to metadata and update the context
 	clictx.App.Metadata["context"] = ctx
 	return nil
 }
 
-func afterSubcommands(c *cli.Context) error {
-	ctx, err := cliContextToContext(c)
-	if err != nil {
-		return err
-	}
-
-	katautils.StopTracing(ctx)
-
-	return nil
-}
-
 // function called when an invalid command is specified which causes the
 // runtime to error.
 func commandNotFound(c *cli.Context, command string) {
@@ -502,7 +415,6 @@ func createRuntimeApp(ctx context.Context, args []string) error {
 	app.Flags = runtimeFlags
 	app.Commands = runtimeCommands
 	app.Before = runtimeBeforeSubcommands
-	app.After = runtimeAfterSubcommands
 	app.EnableBashCompletion = true
 
 	// allow sub-commands to access context
@@ -578,12 +490,5 @@ func cliContextToContext(c *cli.Context) (context.Context, error) {
 func main() {
 	// create a new empty context
 	ctx := context.Background()
-
-	dieCb := func() {
-		katautils.StopTracing(ctx)
-	}
-
-	defer signals.HandlePanic(dieCb)
-
 	createRuntime(ctx)
 }

src/runtime/cli/version.go

@@ -6,7 +6,6 @@
 package main
 
 import (
-	"github.com/kata-containers/kata-containers/src/runtime/pkg/katautils"
 	"github.com/urfave/cli"
 )
 
@@ -14,14 +13,6 @@ var versionCLICommand = cli.Command{
 	Name:  "version",
 	Usage: "display version details",
 	Action: func(context *cli.Context) error {
-		ctx, err := cliContextToContext(context)
-		if err != nil {
-			return err
-		}
-
-		span, _ := katautils.Trace(ctx, "version")
-		defer span.End()
-
 		cli.VersionPrinter(context)
 		return nil
 	},

src/runtime/containerd-shim-v2/shim_management.go

@@ -16,7 +16,6 @@ import (
 	"strconv"
 	"strings"
 
-	"github.com/containerd/containerd/namespaces"
 	cdshim "github.com/containerd/containerd/runtime/v2/shim"
 	vc "github.com/kata-containers/kata-containers/src/runtime/virtcontainers"
 	vcAnnotations "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/annotations"
@@ -129,11 +128,7 @@ func decodeAgentMetrics(body string) []*dto.MetricFamily {
 
 func (s *service) startManagementServer(ctx context.Context, ociSpec *specs.Spec) {
 	// metrics socket will under sandbox's bundle path
-	metricsAddress, err := socketAddress(ctx, s.id)
-	if err != nil {
-		shimMgtLog.WithError(err).Error("failed to create socket address")
-		return
-	}
+	metricsAddress := SocketAddress(s.id)
 
 	listener, err := cdshim.NewSocket(metricsAddress)
 	if err != nil {
@@ -166,7 +161,7 @@ func (s *service) startManagementServer(ctx context.Context, ociSpec *specs.Spec
 	svr.Serve(listener)
 }
 
-// mountServeDebug provides a debug endpoint
+// mountPprofHandle provides a debug endpoint
 func (s *service) mountPprofHandle(m *http.ServeMux, ociSpec *specs.Spec) {
 
 	// return if not enabled
@@ -188,10 +183,8 @@ func (s *service) mountPprofHandle(m *http.ServeMux, ociSpec *specs.Spec) {
 	m.Handle("/debug/pprof/trace", http.HandlerFunc(pprof.Trace))
 }
 
-func socketAddress(ctx context.Context, id string) (string, error) {
-	ns, err := namespaces.NamespaceRequired(ctx)
-	if err != nil {
-		return "", err
-	}
-	return filepath.Join(string(filepath.Separator), "containerd-shim", ns, id, "shim-monitor.sock"), nil
+// SocketAddress returns the address of the abstract domain socket for communicating with the
+// shim management endpoint
+func SocketAddress(id string) string {
+	return filepath.Join(string(filepath.Separator), "run", "vc", id, "shim-monitor")
 }

src/runtime/containerd-shim-v2/wait.go

@@ -142,7 +142,7 @@ func watchOOMEvents(ctx context.Context, s *service) {
 
 	for {
 		select {
-		case <-ctx.Done():
+		case <-s.ctx.Done():
 			return
 		default:
 			containerID, err := s.sandbox.GetOOMEvent(ctx)

src/runtime/pkg/kata-monitor/metrics.go

@@ -176,7 +176,7 @@ func (km *KataMonitor) aggregateSandboxMetrics(encoder expfmt.Encoder) error {
 	for sandboxID, namespace := range sandboxes {
 		wg.Add(1)
 		go func(sandboxID, namespace string, results chan<- []*dto.MetricFamily) {
-			sandboxMetrics, err := km.getSandboxMetrics(sandboxID, namespace)
+			sandboxMetrics, err := getParsedMetrics(sandboxID)
 			if err != nil {
 				monitorLog.WithError(err).WithField("sandbox_id", sandboxID).Errorf("failed to get metrics for sandbox")
 			}
@@ -229,13 +229,12 @@ func (km *KataMonitor) aggregateSandboxMetrics(encoder expfmt.Encoder) error {
 			return err
 		}
 	}
-
 	return nil
+
 }
 
-// getSandboxMetrics will get sandbox's metrics from shim
-func (km *KataMonitor) getSandboxMetrics(sandboxID, namespace string) ([]*dto.MetricFamily, error) {
-	body, err := km.doGet(sandboxID, namespace, defaultTimeout, "metrics")
+func getParsedMetrics(sandboxID string) ([]*dto.MetricFamily, error) {
+	body, err := doGet(sandboxID, defaultTimeout, "metrics")
 	if err != nil {
 		return nil, err
 	}
@@ -243,6 +242,16 @@ func (km *KataMonitor) getSandboxMetrics(sandboxID, namespace string) ([]*dto.Me
 	return parsePrometheusMetrics(sandboxID, body)
 }
 
+// GetSandboxMetrics will get sandbox's metrics from shim
+func GetSandboxMetrics(sandboxID string) (string, error) {
+	body, err := doGet(sandboxID, defaultTimeout, "metrics")
+	if err != nil {
+		return "", err
+	}
+
+	return string(body), nil
+}
+
 // parsePrometheusMetrics will decode metrics from Prometheus text format
 // and return array of *dto.MetricFamily with an ASC order
 func parsePrometheusMetrics(sandboxID string, body []byte) ([]*dto.MetricFamily, error) {

src/runtime/pkg/kata-monitor/monitor.go

@@ -87,13 +87,8 @@ func (km *KataMonitor) GetAgentURL(w http.ResponseWriter, r *http.Request) {
 		commonServeError(w, http.StatusBadRequest, err)
 		return
 	}
-	namespace, err := km.getSandboxNamespace(sandboxID)
-	if err != nil {
-		commonServeError(w, http.StatusBadRequest, err)
-		return
-	}
 
-	data, err := km.doGet(sandboxID, namespace, defaultTimeout, "agent-url")
+	data, err := doGet(sandboxID, defaultTimeout, "agent-url")
 	if err != nil {
 		commonServeError(w, http.StatusBadRequest, err)
 		return

src/runtime/pkg/kata-monitor/shim_client.go

@@ -11,6 +11,8 @@ import (
 	"net"
 	"net/http"
 	"time"
+
+	shim "github.com/kata-containers/kata-containers/src/runtime/containerd-shim-v2"
 )
 
 const (
@@ -33,16 +35,13 @@ func getSandboxIDFromReq(r *http.Request) (string, error) {
 	return "", fmt.Errorf("sandbox not found in %+v", r.URL.Query())
 }
 
-func (km *KataMonitor) buildShimClient(sandboxID, namespace string, timeout time.Duration) (*http.Client, error) {
-	socketAddr, err := km.getMonitorAddress(sandboxID, namespace)
-	if err != nil {
-		return nil, err
-	}
-	return BuildUnixSocketClient(socketAddr, timeout)
+// BuildShimClient builds and returns an http client for communicating with the provided sandbox
+func BuildShimClient(sandboxID string, timeout time.Duration) (*http.Client, error) {
+	return buildUnixSocketClient(shim.SocketAddress(sandboxID), timeout)
 }
 
-// BuildUnixSocketClient build http client for Unix socket
-func BuildUnixSocketClient(socketAddr string, timeout time.Duration) (*http.Client, error) {
+// buildUnixSocketClient build http client for Unix socket
+func buildUnixSocketClient(socketAddr string, timeout time.Duration) (*http.Client, error) {
 	transport := &http.Transport{
 		DisableKeepAlives: true,
 		Dial: func(proto, addr string) (conn net.Conn, err error) {
@@ -61,8 +60,8 @@ func BuildUnixSocketClient(socketAddr string, timeout time.Duration) (*http.Clie
 	return client, nil
 }
 
-func (km *KataMonitor) doGet(sandboxID, namespace string, timeoutInSeconds time.Duration, urlPath string) ([]byte, error) {
-	client, err := km.buildShimClient(sandboxID, namespace, timeoutInSeconds)
+func doGet(sandboxID string, timeoutInSeconds time.Duration, urlPath string) ([]byte, error) {
+	client, err := BuildShimClient(sandboxID, timeoutInSeconds)
 	if err != nil {
 		return nil, err
 	}

src/runtime/pkg/katautils/config.go

@@ -154,6 +154,7 @@ type agent struct {
 	Debug               bool     `toml:"enable_debug"`
 	Tracing             bool     `toml:"enable_tracing"`
 	DebugConsoleEnabled bool     `toml:"debug_console_enabled"`
+	DialTimeout         uint32   `toml:"dial_timeout"`
 }
 
 type netmon struct {
@@ -471,6 +472,10 @@ func (a agent) debugConsoleEnabled() bool {
 	return a.DebugConsoleEnabled
 }
 
+func (a agent) dialTimout() uint32 {
+	return a.DialTimeout
+}
+
 func (a agent) debug() bool {
 	return a.Debug
 }
@@ -920,6 +925,7 @@ func updateRuntimeConfigAgent(configPath string, tomlConf tomlConfig, config *oc
 			TraceType:          agent.traceType(),
 			KernelModules:      agent.kernelModules(),
 			EnableDebugConsole: agent.debugConsoleEnabled(),
+			DialTimeout:        agent.dialTimout(),
 		}
 	}
 

src/runtime/virtcontainers/kata_agent.go

@@ -217,6 +217,7 @@ type KataAgentConfig struct {
 	ContainerPipeSize  uint32
 	TraceMode          string
 	TraceType          string
+	DialTimeout        uint32
 	KernelModules      []string
 }
 
@@ -236,6 +237,7 @@ type kataAgent struct {
 	keepConn       bool
 	dynamicTracing bool
 	dead           bool
+	dialTimout     uint32
 	kmodules       []string
 
 	vmSocket interface{}
@@ -344,6 +346,7 @@ func (k *kataAgent) init(ctx context.Context, sandbox *Sandbox, config KataAgent
 	disableVMShutdown = k.handleTraceSettings(config)
 	k.keepConn = config.LongLiveConn
 	k.kmodules = config.KernelModules
+	k.dialTimout = config.DialTimeout
 
 	return disableVMShutdown, nil
 }
@@ -1794,7 +1797,7 @@ func (k *kataAgent) connect(ctx context.Context) error {
 	}
 
 	k.Logger().WithField("url", k.state.URL).Info("New client")
-	client, err := kataclient.NewAgentClient(k.ctx, k.state.URL)
+	client, err := kataclient.NewAgentClient(k.ctx, k.state.URL, k.dialTimout)
 	if err != nil {
 		k.dead = true
 		return err

src/runtime/virtcontainers/pkg/agent/protocols/client/client.go

@@ -62,15 +62,21 @@ type dialer func(string, time.Duration) (net.Conn, error)
 //     model, and mediates communication between AF_UNIX sockets (on the host end)
 //     and AF_VSOCK sockets (on the guest end).
 //   - mock://<path>. just for test use.
-func NewAgentClient(ctx context.Context, sock string) (*AgentClient, error) {
+func NewAgentClient(ctx context.Context, sock string, timeout uint32) (*AgentClient, error) {
 	grpcAddr, parsedAddr, err := parse(sock)
 	if err != nil {
 		return nil, err
 	}
 
+	dialTimeout := defaultDialTimeout
+	if timeout > 0 {
+		dialTimeout = time.Duration(timeout) * time.Second
+		agentClientLog.WithField("timeout", timeout).Debug("custom dialing timeout has been set")
+	}
+
 	var conn net.Conn
 	var d = agentDialer(parsedAddr)
-	conn, err = d(grpcAddr, defaultDialTimeout)
+	conn, err = d(grpcAddr, dialTimeout)
 	if err != nil {
 		return nil, err
 	}
@@ -92,7 +98,7 @@ func NewAgentClient(ctx context.Context, sock string) (*AgentClient, error) {
 				grpc.WithStreamInterceptor(otgrpc.OpenTracingStreamClientInterceptor(tracer)))
 		}
 
-		ctx, cancel := context.WithTimeout(ctx, defaultDialTimeout)
+		ctx, cancel := context.WithTimeout(ctx, dialTimeout)
 		defer cancel()
 		conn, err := grpc.DialContext(ctx, grpcAddr, dialOpts...)
 		if err != nil {

src/runtime/virtcontainers/qemu.go

@@ -768,8 +768,8 @@ func (q *qemu) setupVirtioMem() error {
 	if err != nil {
 		return err
 	}
-	// 1024 is size for nvdimm
-	sizeMB := int(maxMem) - int(q.config.MemorySize)
+	// backend memory size must be multiple of 2Mib
+	sizeMB := (int(maxMem) - int(q.config.MemorySize)) >> 2 << 2
 
 	share, target, memoryBack, err := q.getMemArgs()
 	if err != nil {
@@ -2043,8 +2043,9 @@ func (q *qemu) resizeMemory(ctx context.Context, reqMemMB uint32, memoryBlockSiz
 	var addMemDevice memoryDevice
 	if q.config.VirtioMem && currentMemory != reqMemMB {
 		q.Logger().WithField("hotplug", "memory").Debugf("resize memory from %dMB to %dMB", currentMemory, reqMemMB)
-		sizeByte := (reqMemMB - q.config.MemorySize) * 1024 * 1024
-		err := q.qmpMonitorCh.qmp.ExecQomSet(q.qmpMonitorCh.ctx, "virtiomem0", "requested-size", uint64(sizeByte))
+		sizeByte := uint64(reqMemMB - q.config.MemorySize)
+		sizeByte = sizeByte * 1024 * 1024
+		err := q.qmpMonitorCh.qmp.ExecQomSet(q.qmpMonitorCh.ctx, "virtiomem0", "requested-size", sizeByte)
 		if err != nil {
 			return 0, memoryDevice{}, err
 		}

src/runtime/virtcontainers/sandbox.go

@@ -686,13 +686,13 @@ func (s *Sandbox) Delete(ctx context.Context) error {
 
 	for _, c := range s.containers {
 		if err := c.delete(ctx); err != nil {
-			return err
+			s.Logger().WithError(err).WithField("cid", c.id).Debug("failed to delete container")
 		}
 	}
 
 	if !rootless.IsRootless() {
 		if err := s.cgroupsDelete(); err != nil {
-			return err
+			s.Logger().WithError(err).Error("failed to cleanup cgroups")
 		}
 	}
 

tools/osbuilder/image-builder/image_builder.sh

@@ -63,11 +63,8 @@ readonly -a systemd_files=(
 # Set a default value
 AGENT_INIT=${AGENT_INIT:-no}
 
-# Align image to (size in MB) according to different architecture.
-case "$(uname -m)" in
-	aarch64) readonly mem_boundary_mb=16 ;;
-	*) readonly mem_boundary_mb=128 ;;
-esac
+# Align image to 128M
+readonly mem_boundary_mb=128
 
 # shellcheck source=../scripts/lib.sh
 source "${lib_file}"

tools/packaging/ccloudvm/README.md

@@ -2,7 +2,8 @@
 
 * [How to use Kata workloads for `ccloudvm`](#how-to-use-kata-workloads-for-ccloudvm)
     * [Create Docker\* and Kata Containers virtualized environment](#create-docker-and-kata-containers-virtualized-environment)
----
+
+***
 
 The [ccloudvm](https://github.com/intel/ccloudvm/) tool is a command
 to create development and demo environments. The tool sets up these development

tools/packaging/kata-deploy/kata-cleanup/base/kata-cleanup.yaml

@@ -18,7 +18,7 @@ spec:
           katacontainers.io/kata-runtime: cleanup
       containers:
       - name: kube-kata-cleanup
-        image: katadocker/kata-deploy:2.1.0-rc0
+        image: katadocker/kata-deploy:2.1.0
         imagePullPolicy: Always
         command: [ "bash", "-c", "/opt/kata-artifacts/scripts/kata-deploy.sh reset" ]
         env:

tools/packaging/kata-deploy/kata-cleanup/overlays/k3s/kustomization.yaml

@@ -0,0 +1,5 @@
+bases:
+- ../../base
+
+patchesStrategicMerge:
+- mount_k3s_conf.yaml

tools/packaging/kata-deploy/kata-cleanup/overlays/k3s/mount_k3s_conf.yaml

@@ -0,0 +1,17 @@
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: kubelet-kata-cleanup
+  namespace: kube-system
+spec:
+  template:
+    spec:
+      containers:
+      - name: kube-kata-cleanup
+        volumeMounts:
+        - name: containerd-conf
+          mountPath: /etc/containerd/
+      volumes:
+        - name: containerd-conf
+          hostPath:
+            path: /var/lib/rancher/k3s/agent/etc/containerd/

tools/packaging/kata-deploy/kata-deploy/base/kata-deploy.yaml

@@ -16,7 +16,7 @@ spec:
       serviceAccountName: kata-label-node
       containers:
       - name: kube-kata
-        image: katadocker/kata-deploy:2.1.0-rc0
+        image: katadocker/kata-deploy:2.1.0
         imagePullPolicy: Always
         lifecycle:
           preStop: