Merge pull request #441 from liubin/feature/245-add-debug-console

kata 2.0: add debug console service

src/agent/src/main.rs

@@ -31,16 +31,20 @@ extern crate netlink;
 use crate::netlink::{RtnlHandle, NETLINK_ROUTE};
 use anyhow::{anyhow, Context, Result};
 use nix::fcntl::{self, OFlag};
+use nix::fcntl::{FcntlArg, FdFlag};
+use nix::libc::{STDERR_FILENO, STDIN_FILENO, STDOUT_FILENO};
+use nix::pty;
+use nix::sys::select::{select, FdSet};
 use nix::sys::socket::{self, AddressFamily, SockAddr, SockFlag, SockType};
 use nix::sys::wait::{self, WaitStatus};
-use nix::unistd;
-use nix::unistd::dup;
+use nix::unistd::{self, close, dup, dup2, fork, setsid, ForkResult};
 use prctl::set_child_subreaper;
 use signal_hook::{iterator::Signals, SIGCHLD};
 use std::collections::HashMap;
 use std::env;
-use std::ffi::OsStr;
+use std::ffi::{CStr, CString, OsStr};
 use std::fs::{self, File};
+use std::io::{Read, Write};
 use std::os::unix::ffi::OsStrExt;
 use std::os::unix::fs as unixfs;
 use std::os::unix::io::AsRawFd;
@@ -75,6 +79,8 @@ const NAME: &str = "kata-agent";
 const KERNEL_CMDLINE_FILE: &str = "/proc/cmdline";
 const CONSOLE_PATH: &str = "/dev/console";
 
+const DEFAULT_BUF_SIZE: usize = 8 * 1024;
+
 lazy_static! {
     static ref GLOBAL_DEVICE_WATCHER: Arc<Mutex<HashMap<String, Sender<String>>>> =
         Arc::new(Mutex::new(HashMap::new()));
@@ -213,7 +219,7 @@ fn start_sandbox(logger: &Logger, config: &agentConfig, init_mode: bool) -> Resu
 
         let handle = builder.spawn(move || {
             let shells = shells.lock().unwrap();
-            let result = setup_debug_console(shells.to_vec(), debug_console_vport);
+            let result = setup_debug_console(&thread_logger, shells.to_vec(), debug_console_vport);
             if result.is_err() {
                 // Report error, but don't fail
                 warn!(thread_logger, "failed to setup debug console";
@@ -406,9 +412,9 @@ use crate::config::agentConfig;
 use nix::sys::stat::Mode;
 use std::os::unix::io::{FromRawFd, RawFd};
 use std::path::PathBuf;
-use std::process::{exit, Command, Stdio};
+use std::process::exit;
 
-fn setup_debug_console(shells: Vec<String>, port: u32) -> Result<()> {
+fn setup_debug_console(logger: &Logger, shells: Vec<String>, port: u32) -> Result<()> {
     let mut shell: &str = "";
     for sh in shells.iter() {
         let binary = PathBuf::from(sh);
@@ -422,7 +428,7 @@ fn setup_debug_console(shells: Vec<String>, port: u32) -> Result<()> {
         return Err(anyhow!("no shell found to launch debug console"));
     }
 
-    let f: RawFd = if port > 0 {
+    if port > 0 {
         let listenfd = socket::socket(
             AddressFamily::Vsock,
             SockType::Stream,
@@ -432,29 +438,201 @@ fn setup_debug_console(shells: Vec<String>, port: u32) -> Result<()> {
         let addr = SockAddr::new_vsock(libc::VMADDR_CID_ANY, port);
         socket::bind(listenfd, &addr)?;
         socket::listen(listenfd, 1)?;
-        socket::accept4(listenfd, SockFlag::SOCK_CLOEXEC)?
+        loop {
+            let f: RawFd = socket::accept4(listenfd, SockFlag::SOCK_CLOEXEC)?;
+            match run_debug_console_shell(logger, shell, f) {
+                Ok(_) => {
+                    info!(logger, "run_debug_console_shell session finished");
+                }
+                Err(err) => {
+                    error!(logger, "run_debug_console_shell failed: {:?}", err);
+                }
+            }
+        }
     } else {
         let mut flags = OFlag::empty();
         flags.insert(OFlag::O_RDWR);
         flags.insert(OFlag::O_CLOEXEC);
-        fcntl::open(CONSOLE_PATH, flags, Mode::empty())?
+        loop {
+            let f: RawFd = fcntl::open(CONSOLE_PATH, flags, Mode::empty())?;
+            match run_debug_console_shell(logger, shell, f) {
+                Ok(_) => {
+                    info!(logger, "run_debug_console_shell session finished");
+                }
+                Err(err) => {
+                    error!(logger, "run_debug_console_shell failed: {:?}", err);
+                }
+            }
+        }
+    };
+}
+
+fn io_copy<R: ?Sized, W: ?Sized>(reader: &mut R, writer: &mut W) -> io::Result<u64>
+where
+    R: Read,
+    W: Write,
+{
+    let mut buf = [0; DEFAULT_BUF_SIZE];
+    let buf_len;
+
+    match reader.read(&mut buf) {
+        Ok(0) => return Ok(0),
+        Ok(len) => buf_len = len,
+        Err(err) => return Err(err),
     };
 
-    let cmd = Command::new(shell)
-        .arg("-i")
-        .stdin(unsafe { Stdio::from_raw_fd(f) })
-        .stdout(unsafe { Stdio::from_raw_fd(f) })
-        .stderr(unsafe { Stdio::from_raw_fd(f) })
-        .spawn();
+    // write and return
+    match writer.write_all(&buf[..buf_len]) {
+        Ok(_) => return Ok(buf_len as u64),
+        Err(err) => return Err(err),
+    }
+}
 
-    let mut cmd = match cmd {
-        Ok(c) => c,
-        Err(_) => return Err(anyhow!("failed to spawn shell")),
-    };
+fn run_debug_console_shell(logger: &Logger, shell: &str, socket_fd: RawFd) -> Result<()> {
+    let pseduo = pty::openpty(None, None)?;
+    let _ = fcntl::fcntl(pseduo.master, FcntlArg::F_SETFD(FdFlag::FD_CLOEXEC));
+    let _ = fcntl::fcntl(pseduo.slave, FcntlArg::F_SETFD(FdFlag::FD_CLOEXEC));
 
-    cmd.wait()?;
+    let slave_fd = pseduo.slave;
 
-    return Ok(());
+    match fork() {
+        Ok(ForkResult::Child) => {
+            // create new session with child as session leader
+            setsid()?;
+
+            // dup stdin, stdout, stderr to let child act as a terminal
+            dup2(slave_fd, STDIN_FILENO)?;
+            dup2(slave_fd, STDOUT_FILENO)?;
+            dup2(slave_fd, STDERR_FILENO)?;
+
+            // set tty
+            unsafe {
+                libc::ioctl(0, libc::TIOCSCTTY);
+            }
+
+            let cmd = CString::new(shell).unwrap();
+            let args: Vec<&CStr> = vec![];
+
+            // run shell
+            if let Err(e) = unistd::execvp(cmd.as_c_str(), args.as_slice()) {
+                match e {
+                    nix::Error::Sys(errno) => {
+                        std::process::exit(errno as i32);
+                    }
+                    _ => std::process::exit(-2),
+                }
+            }
+        }
+
+        Ok(ForkResult::Parent { child: child_pid }) => {
+            info!(logger, "get debug shell pid {:?}", child_pid);
+
+            let (rfd, wfd) = unistd::pipe2(OFlag::O_CLOEXEC)?;
+            let master_fd = pseduo.master;
+            let debug_shell_logger = logger.clone();
+
+            // channel that used to sync between thread and main process
+            let (tx, rx) = mpsc::channel::<i32>();
+
+            // start a thread to do IO copy between socket and pseduo.master
+            thread::spawn(move || {
+                let mut master_reader = unsafe { File::from_raw_fd(master_fd) };
+                let mut master_writer = unsafe { File::from_raw_fd(master_fd) };
+                let mut socket_reader = unsafe { File::from_raw_fd(socket_fd) };
+                let mut socket_writer = unsafe { File::from_raw_fd(socket_fd) };
+
+                loop {
+                    let mut fd_set = FdSet::new();
+                    fd_set.insert(rfd);
+                    fd_set.insert(master_fd);
+                    fd_set.insert(socket_fd);
+
+                    match select(
+                        Some(fd_set.highest().unwrap() + 1),
+                        &mut fd_set,
+                        None,
+                        None,
+                        None,
+                    ) {
+                        Ok(_) => (),
+                        Err(e) => {
+                            if e == nix::Error::from(nix::errno::Errno::EINTR) {
+                                continue;
+                            } else {
+                                error!(debug_shell_logger, "select error {:?}", e);
+                                tx.send(1).unwrap();
+                                break;
+                            }
+                        }
+                    }
+
+                    if fd_set.contains(rfd) {
+                        info!(
+                            debug_shell_logger,
+                            "debug shell process {} exited", child_pid
+                        );
+                        tx.send(1).unwrap();
+                        break;
+                    }
+
+                    if fd_set.contains(master_fd) {
+                        match io_copy(&mut master_reader, &mut socket_writer) {
+                            Ok(0) => {
+                                debug!(debug_shell_logger, "master fd closed");
+                                tx.send(1).unwrap();
+                                break;
+                            }
+                            Ok(_) => {}
+                            Err(ref e) if e.kind() == std::io::ErrorKind::Interrupted => continue,
+                            Err(e) => {
+                                error!(debug_shell_logger, "read master fd error {:?}", e);
+                                tx.send(1).unwrap();
+                                break;
+                            }
+                        }
+                    }
+
+                    if fd_set.contains(socket_fd) {
+                        match io_copy(&mut socket_reader, &mut master_writer) {
+                            Ok(0) => {
+                                debug!(debug_shell_logger, "socket fd closed");
+                                tx.send(1).unwrap();
+                                break;
+                            }
+                            Ok(_) => {}
+                            Err(ref e) if e.kind() == std::io::ErrorKind::Interrupted => continue,
+                            Err(e) => {
+                                error!(debug_shell_logger, "read socket fd error {:?}", e);
+                                tx.send(1).unwrap();
+                                break;
+                            }
+                        }
+                    }
+                }
+            });
+
+            let wait_status = wait::waitpid(child_pid, None);
+            info!(logger, "debug console process exit code: {:?}", wait_status);
+
+            info!(logger, "notify debug monitor thread to exit");
+            // close pipe to exit select loop
+            let _ = close(wfd);
+
+            // wait for thread exit.
+            let _ = rx.recv().unwrap();
+            info!(logger, "debug monitor thread has exited");
+
+            // close files
+            let _ = close(rfd);
+            let _ = close(master_fd);
+            let _ = close(slave_fd);
+        }
+        Err(err) => {
+            return Err(anyhow!("fork error: {:?}", err));
+        }
+    }
+
+    Ok(())
 }
 
 #[cfg(test)]
@@ -472,8 +650,9 @@ mod tests {
         let shells_ref = SHELLS.clone();
         let mut shells = shells_ref.lock().unwrap();
         shells.clear();
+        let logger = slog_scope::logger();
 
-        let result = setup_debug_console(shells.to_vec(), 0);
+        let result = setup_debug_console(&logger, shells.to_vec(), 0);
 
         assert!(result.is_err());
         assert_eq!(
@@ -498,8 +677,9 @@ mod tests {
             .to_string();
 
         shells.push(shell);
+        let logger = slog_scope::logger();
 
-        let result = setup_debug_console(shells.to_vec(), 0);
+        let result = setup_debug_console(&logger, shells.to_vec(), 0);
 
         assert!(result.is_err());
         assert_eq!(

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

@@ -127,6 +127,13 @@ block_device_driver = "@DEFBLOCKSTORAGEDRIVER_ACRN@"
 #trace_mode = "dynamic"
 #trace_type = "isolated"
 
+# Enable debug console.
+
+# If enabled, user can connect guest OS running inside hypervisor
+# through "kata-runtime exec <sandbox-id>" command
+
+#debug_console_enabled = true
+
 [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

@@ -125,6 +125,12 @@ block_device_driver = "virtio-blk"
 #trace_mode = "dynamic"
 #trace_type = "isolated"
 
+# Enable debug console.
+
+# If enabled, user can connect guest OS running inside hypervisor
+# through "kata-runtime exec <sandbox-id>" command
+
+#debug_console_enabled = true
 
 [netmon]
 # If enabled, the network monitoring process gets started when the

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

@@ -256,6 +256,13 @@ block_device_driver = "@DEFBLOCKSTORAGEDRIVER_FC@"
 #
 kernel_modules=[]
 
+# Enable debug console.
+
+# If enabled, user can connect guest OS running inside hypervisor
+# through "kata-runtime exec <sandbox-id>" command
+
+#debug_console_enabled = true
+
 [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-virtiofs.toml.in

@@ -352,6 +352,12 @@ vhost_user_store_path = "@DEFVHOSTUSERSTOREPATH@"
 #
 kernel_modules=[]
 
+# Enable debug console.
+
+# If enabled, user can connect guest OS running inside hypervisor
+# through "kata-runtime exec <sandbox-id>" command
+
+#debug_console_enabled = true
 
 [netmon]
 # If enabled, the network monitoring process gets started when the

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

@@ -375,6 +375,12 @@ vhost_user_store_path = "@DEFVHOSTUSERSTOREPATH@"
 #
 kernel_modules=[]
 
+# Enable debug console.
+
+# If enabled, user can connect guest OS running inside hypervisor
+# through "kata-runtime exec <sandbox-id>" command
+
+#debug_console_enabled = true
 
 [netmon]
 # If enabled, the network monitoring process gets started when the

src/runtime/cli/kata-exec.go

@@ -0,0 +1,218 @@
+// Copyright (c) 2017-2019 Intel Corporation
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+package main
+
+import (
+	"fmt"
+	"io"
+	"io/ioutil"
+	"net"
+	"net/http"
+	"net/url"
+	"os"
+	"strings"
+
+	"sync"
+	"time"
+
+	"github.com/containerd/console"
+	"github.com/kata-containers/kata-containers/src/runtime/pkg/katautils"
+	clientUtils "github.com/kata-containers/kata-containers/src/runtime/virtcontainers/pkg/agent/protocols/client"
+	"github.com/pkg/errors"
+	"github.com/urfave/cli"
+)
+
+const (
+
+	// The buffer size used to specify the buffer for IO streams copy
+	bufSize = 1024 * 2
+
+	defaultTimeout = 3 * time.Second
+
+	subCommandName = "exec"
+	// command-line parameters name
+	paramKataMonitorAddr                     = "kata-monitor-addr"
+	paramDebugConsolePort                    = "kata-debug-port"
+	defaultKernelParamDebugConsoleVPortValue = 1026
+	defaultParamKataMonitorAddr              = "http://localhost:8090"
+)
+
+var (
+	bufPool = sync.Pool{
+		New: func() interface{} {
+			buffer := make([]byte, bufSize)
+			return &buffer
+		},
+	}
+)
+
+var kataExecCLICommand = cli.Command{
+	Name:  subCommandName,
+	Usage: "Enter into guest by debug console",
+	Flags: []cli.Flag{
+		cli.StringFlag{
+			Name:  paramKataMonitorAddr,
+			Usage: "Kata monitor listen address.",
+		},
+		cli.Uint64Flag{
+			Name:  paramDebugConsolePort,
+			Usage: "Port that debug console is listening on.",
+		},
+	},
+	Action: func(context *cli.Context) error {
+		ctx, err := cliContextToContext(context)
+		if err != nil {
+			return err
+		}
+		span, _ := katautils.Trace(ctx, subCommandName)
+		defer span.Finish()
+
+		endPoint := context.String(paramKataMonitorAddr)
+		if endPoint == "" {
+			endPoint = defaultParamKataMonitorAddr
+		}
+
+		port := context.Uint64(paramDebugConsolePort)
+		if port == 0 {
+			port = defaultKernelParamDebugConsoleVPortValue
+		}
+
+		sandboxID := context.Args().Get(0)
+		if sandboxID == "" {
+			return fmt.Errorf("SandboxID not found")
+		}
+
+		conn, err := getConn(endPoint, sandboxID, port)
+		if err != nil {
+			return err
+		}
+		defer conn.Close()
+
+		con := console.Current()
+		defer con.Reset()
+
+		if err := con.SetRaw(); err != nil {
+			return err
+		}
+
+		iostream := &iostream{
+			conn:   conn,
+			exitch: make(chan struct{}),
+			closed: false,
+		}
+
+		ioCopy(iostream, con)
+
+		<-iostream.exitch
+		return nil
+	},
+}
+
+func ioCopy(stream *iostream, con console.Console) {
+	var wg sync.WaitGroup
+
+	// stdin
+	go func() {
+		p := bufPool.Get().(*[]byte)
+		defer bufPool.Put(p)
+		io.CopyBuffer(stream, con, *p)
+	}()
+
+	// stdout
+	wg.Add(1)
+	go func() {
+		p := bufPool.Get().(*[]byte)
+		defer bufPool.Put(p)
+		io.CopyBuffer(os.Stdout, stream, *p)
+		wg.Done()
+	}()
+
+	wg.Wait()
+	close(stream.exitch)
+}
+
+type iostream struct {
+	conn   net.Conn
+	exitch chan struct{}
+	closed bool
+}
+
+func (s *iostream) Write(data []byte) (n int, err error) {
+	if s.closed {
+		return 0, errors.New("stream closed")
+	}
+	return s.conn.Write(data)
+}
+
+func (s *iostream) Close() error {
+	if s.closed {
+		return errors.New("stream closed")
+	}
+
+	err := s.conn.Close()
+	if err == nil {
+		s.closed = true
+	}
+
+	return err
+}
+
+func (s *iostream) Read(data []byte) (n int, err error) {
+	if s.closed {
+		return 0, errors.New("stream closed")
+	}
+
+	return s.conn.Read(data)
+}
+
+func getConn(endPoint, sandboxID string, port uint64) (net.Conn, error) {
+	shimURL := fmt.Sprintf("%s/agent-url?sandbox=%s", endPoint, sandboxID)
+	resp, err := http.Get(shimURL)
+	if err != nil {
+		return nil, err
+	}
+
+	if resp.StatusCode != http.StatusOK {
+		return nil, fmt.Errorf("Failed to get %s: %d", shimURL, resp.StatusCode)
+	}
+
+	defer resp.Body.Close()
+	data, err := ioutil.ReadAll(resp.Body)
+	if err != nil {
+		return nil, err
+	}
+
+	sock := strings.TrimSuffix(string(data), "\n")
+	addr, err := url.Parse(sock)
+	if err != nil {
+		return nil, err
+	}
+
+	// validate more
+	switch addr.Scheme {
+	case clientUtils.VSockSocketScheme:
+		// vsock://31513974:1024
+		cidAndPort := strings.Split(addr.Host, ":")
+		if len(cidAndPort) != 2 {
+			return nil, fmt.Errorf("Invalid vsock scheme: %s", sock)
+		}
+		shimAddr := fmt.Sprintf("%s:%s:%d", clientUtils.VSockSocketScheme, cidAndPort[0], port)
+		return clientUtils.VsockDialer(shimAddr, defaultTimeout)
+
+	case clientUtils.HybridVSockScheme:
+		// addr: hvsock:///run/vc/firecracker/340b412c97bf1375cdda56bfa8f18c8a/root/kata.hvsock:1024
+		hvsocket := strings.Split(addr.Path, ":")
+		if len(hvsocket) != 2 {
+			return nil, fmt.Errorf("Invalid hybrid vsock scheme: %s", sock)
+		}
+
+		// hvsock:///run/vc/firecracker/340b412c97bf1375cdda56bfa8f18c8a/root/kata.hvsock
+		shimAddr := fmt.Sprintf("%s:%s:%d", clientUtils.HybridVSockScheme, hvsocket[0], port)
+		return clientUtils.HybridVSockDialer(shimAddr, defaultTimeout)
+	}
+
+	return nil, fmt.Errorf("schema %s not found", addr.Scheme)
+}

src/runtime/cli/kata-monitor/main.go

@@ -36,6 +36,7 @@ func main() {
 	m := http.NewServeMux()
 	m.Handle("/metrics", http.HandlerFunc(km.ProcessMetricsRequest))
 	m.Handle("/sandboxes", http.HandlerFunc(km.ListSandboxes))
+	m.Handle("/agent-url", http.HandlerFunc(km.GetAgentURL))
 
 	// for debug shim process
 	m.Handle("/debug/vars", http.HandlerFunc(km.ExpvarHandler))

src/runtime/cli/main.go

@@ -125,6 +125,7 @@ var runtimeCommands = []cli.Command{
 	// Kata Containers specific extensions
 	kataCheckCLICommand,
 	kataEnvCLICommand,
+	kataExecCLICommand,
 	factoryCLICommand,
 }
 

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

@@ -8,6 +8,7 @@ package containerdshim
 import (
 	"context"
 	"expvar"
+	"fmt"
 	"io"
 	"net/http"
 	"net/http/pprof"
@@ -34,6 +35,18 @@ var (
 	shimMgtLog               = shimLog.WithField("subsystem", "shim-management")
 )
 
+// agentURL returns URL for agent
+func (s *service) agentURL(w http.ResponseWriter, r *http.Request) {
+	url, err := s.sandbox.GetAgentURL()
+	if err != nil {
+		w.WriteHeader(http.StatusInternalServerError)
+		w.Write([]byte(err.Error()))
+		return
+	}
+
+	fmt.Fprint(w, url)
+}
+
 // serveMetrics handle /metrics requests
 func (s *service) serveMetrics(w http.ResponseWriter, r *http.Request) {
 
@@ -139,6 +152,7 @@ func (s *service) startManagementServer(ctx context.Context, ociSpec *specs.Spec
 	// bind hanlder
 	m := http.NewServeMux()
 	m.Handle("/metrics", http.HandlerFunc(s.serveMetrics))
+	m.Handle("/agent-url", http.HandlerFunc(s.agentURL))
 	s.mountPprofHandle(m, ociSpec)
 
 	// register shim metrics

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

@@ -10,7 +10,6 @@ import (
 	"compress/gzip"
 	"io"
 	"io/ioutil"
-	"net"
 	"net/http"
 	"path/filepath"
 	"sort"
@@ -236,33 +235,7 @@ func (km *KataMonitor) aggregateSandboxMetrics(encoder expfmt.Encoder) error {
 
 // getSandboxMetrics will get sandbox's metrics from shim
 func (km *KataMonitor) getSandboxMetrics(sandboxID, namespace string) ([]*dto.MetricFamily, error) {
-	socket, err := km.getMonitorAddress(sandboxID, namespace)
-	if err != nil {
-		return nil, err
-	}
-
-	transport := &http.Transport{
-		DisableKeepAlives: true,
-		Dial: func(proto, addr string) (conn net.Conn, err error) {
-			return net.Dial("unix", "\x00"+socket)
-		},
-	}
-
-	client := http.Client{
-		Timeout:   3 * time.Second,
-		Transport: transport,
-	}
-
-	resp, err := client.Get("http://shim/metrics")
-	if err != nil {
-		return nil, err
-	}
-
-	defer func() {
-		resp.Body.Close()
-	}()
-
-	body, err := ioutil.ReadAll(resp.Body)
+	body, err := km.doGet(sandboxID, namespace, defaultTimeout, "metrics")
 	if err != nil {
 		return nil, err
 	}

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

@@ -80,6 +80,28 @@ func (km *KataMonitor) initSandboxCache() error {
 	return nil
 }
 
+// GetAgentURL returns agent URL
+func (km *KataMonitor) GetAgentURL(w http.ResponseWriter, r *http.Request) {
+	sandboxID, err := getSandboxIdFromReq(r)
+	if err != nil {
+		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")
+	if err != nil {
+		commonServeError(w, http.StatusBadRequest, err)
+		return
+	}
+
+	fmt.Fprintln(w, string(data))
+}
+
 // ListSandboxes list all sandboxes running in Kata
 func (km *KataMonitor) ListSandboxes(w http.ResponseWriter, r *http.Request) {
 	sandboxes := km.getSandboxList()

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

@@ -12,14 +12,6 @@ import (
 	"net/http"
 )
 
-func getSandboxIdFromReq(r *http.Request) (string, error) {
-	sandbox := r.URL.Query().Get("sandbox")
-	if sandbox != "" {
-		return sandbox, nil
-	}
-	return "", fmt.Errorf("sandbox not found in %+v", r.URL.Query())
-}
-
 func serveError(w http.ResponseWriter, status int, txt string) {
 	w.Header().Set("Content-Type", "text/plain; charset=utf-8")
 	w.Header().Set("X-Go-Pprof", "1")

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

@@ -0,0 +1,81 @@
+// Copyright (c) 2020 Ant Financial
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+
+package katamonitor
+
+import (
+	"fmt"
+	"io/ioutil"
+	"net"
+	"net/http"
+	"time"
+)
+
+const (
+	defaultTimeout = 3 * time.Second
+)
+
+func commonServeError(w http.ResponseWriter, status int, err error) {
+	w.Header().Set("Content-Type", "text/plain; charset=utf-8")
+	w.WriteHeader(status)
+	if err != nil {
+		fmt.Fprintln(w, err.Error())
+	}
+}
+
+func getSandboxIdFromReq(r *http.Request) (string, error) {
+	sandbox := r.URL.Query().Get("sandbox")
+	if sandbox != "" {
+		return sandbox, nil
+	}
+	return "", fmt.Errorf("sandbox not found in %+v", r.URL.Query())
+}
+
+func (km *KataMonitor) buildShimClient(sandboxID, namespace string, timeout time.Duration) (*http.Client, error) {
+	socket, err := km.getMonitorAddress(sandboxID, namespace)
+	if err != nil {
+		return nil, err
+	}
+
+	transport := &http.Transport{
+		DisableKeepAlives: true,
+		Dial: func(proto, addr string) (conn net.Conn, err error) {
+			return net.Dial("unix", "\x00"+socket)
+		},
+	}
+
+	client := &http.Client{
+		Transport: transport,
+	}
+
+	if timeout > 0 {
+		client.Timeout = timeout
+	}
+
+	return client, nil
+}
+
+func (km *KataMonitor) doGet(sandboxID, namespace string, timeoutInSeconds time.Duration, urlPath string) ([]byte, error) {
+	client, err := km.buildShimClient(sandboxID, namespace, timeoutInSeconds)
+	if err != nil {
+		return nil, err
+	}
+
+	resp, err := client.Get(fmt.Sprintf("http://shim/%s", urlPath))
+	if err != nil {
+		return nil, err
+	}
+
+	defer func() {
+		resp.Body.Close()
+	}()
+
+	body, err := ioutil.ReadAll(resp.Body)
+	if err != nil {
+		return nil, err
+	}
+
+	return body, nil
+}

src/runtime/pkg/katautils/config.go

@@ -132,11 +132,12 @@ type runtime struct {
 }
 
 type agent struct {
-	Debug         bool     `toml:"enable_debug"`
-	Tracing       bool     `toml:"enable_tracing"`
-	TraceMode     string   `toml:"trace_mode"`
-	TraceType     string   `toml:"trace_type"`
-	KernelModules []string `toml:"kernel_modules"`
+	Debug               bool     `toml:"enable_debug"`
+	Tracing             bool     `toml:"enable_tracing"`
+	TraceMode           string   `toml:"trace_mode"`
+	TraceType           string   `toml:"trace_type"`
+	KernelModules       []string `toml:"kernel_modules"`
+	DebugConsoleEnabled bool     `toml:"debug_console_enabled"`
 }
 
 type netmon struct {
@@ -441,6 +442,10 @@ func (h hypervisor) getIOMMUPlatform() bool {
 	return h.IOMMUPlatform
 }
 
+func (a agent) debugConsoleEnabled() bool {
+	return a.DebugConsoleEnabled
+}
+
 func (a agent) debug() bool {
 	return a.Debug
 }
@@ -866,23 +871,15 @@ func updateRuntimeConfigHypervisor(configPath string, tomlConf tomlConfig, confi
 }
 
 func updateRuntimeConfigAgent(configPath string, tomlConf tomlConfig, config *oci.RuntimeConfig, builtIn bool) error {
-	if builtIn {
-		config.AgentConfig = vc.KataAgentConfig{
-			LongLiveConn:  true,
-			Debug:         config.AgentConfig.Debug,
-			KernelModules: config.AgentConfig.KernelModules,
-		}
-
-		return nil
-	}
-
 	for _, agent := range tomlConf.Agent {
 		config.AgentConfig = vc.KataAgentConfig{
-			Debug:         agent.debug(),
-			Trace:         agent.trace(),
-			TraceMode:     agent.traceMode(),
-			TraceType:     agent.traceType(),
-			KernelModules: agent.kernelModules(),
+			LongLiveConn:       true,
+			Debug:              agent.debug(),
+			Trace:              agent.trace(),
+			TraceMode:          agent.traceMode(),
+			TraceType:          agent.traceType(),
+			KernelModules:      agent.kernelModules(),
+			EnableDebugConsole: agent.debugConsoleEnabled(),
 		}
 	}
 
@@ -1026,12 +1023,10 @@ func initConfig() (config oci.RuntimeConfig, err error) {
 		return oci.RuntimeConfig{}, err
 	}
 
-	defaultAgentConfig := vc.KataAgentConfig{}
-
 	config = oci.RuntimeConfig{
 		HypervisorType:   defaultHypervisor,
 		HypervisorConfig: GetDefaultHypervisorConfig(),
-		AgentConfig:      defaultAgentConfig,
+		AgentConfig:      vc.KataAgentConfig{},
 	}
 
 	return config, nil

src/runtime/pkg/katautils/config_test.go

@@ -167,7 +167,9 @@ func createAllRuntimeConfigFiles(dir, hypervisor string) (config testRuntimeConf
 		VirtioFSCache:         defaultVirtioFSCacheMode,
 	}
 
-	agentConfig := vc.KataAgentConfig{}
+	agentConfig := vc.KataAgentConfig{
+		LongLiveConn: true,
+	}
 
 	netmonConfig := vc.NetmonConfig{
 		Path:   netmonPath,
@@ -519,7 +521,8 @@ func TestMinimalRuntimeConfig(t *testing.T) {
 	# Runtime configuration file
 
 	[agent.kata]
-
+	debug_console_enabled=true
+	kernel_modules=["a", "b", "c"]
 	[netmon]
 	path = "` + netmonPath + `"
 `
@@ -576,7 +579,11 @@ func TestMinimalRuntimeConfig(t *testing.T) {
 		VirtioFSCache:         defaultVirtioFSCacheMode,
 	}
 
-	expectedAgentConfig := vc.KataAgentConfig{}
+	expectedAgentConfig := vc.KataAgentConfig{
+		LongLiveConn:       true,
+		EnableDebugConsole: true,
+		KernelModules:      []string{"a", "b", "c"},
+	}
 
 	expectedNetmonConfig := vc.NetmonConfig{
 		Path:   netmonPath,

src/runtime/virtcontainers/interfaces.go

@@ -75,6 +75,7 @@ type VCSandbox interface {
 
 	UpdateRuntimeMetrics() error
 	GetAgentMetrics() (string, error)
+	GetAgentURL() (string, error)
 }
 
 // VCContainer is the Container interface

src/runtime/virtcontainers/kata_agent.go

@@ -52,6 +52,11 @@ const (
 
 	// path to vfio devices
 	vfioPath = "/dev/vfio/"
+
+	// enable debug console
+	kernelParamDebugConsole           = "agent.debug_console"
+	kernelParamDebugConsoleVPort      = "agent.debug_console_vport"
+	kernelParamDebugConsoleVPortValue = "1026"
 )
 
 var (
@@ -195,13 +200,14 @@ func ephemeralPath() string {
 // KataAgentConfig is a structure storing information needed
 // to reach the Kata Containers agent.
 type KataAgentConfig struct {
-	LongLiveConn      bool
-	Debug             bool
-	Trace             bool
-	ContainerPipeSize uint32
-	TraceMode         string
-	TraceType         string
-	KernelModules     []string
+	LongLiveConn       bool
+	Debug              bool
+	Trace              bool
+	EnableDebugConsole bool
+	ContainerPipeSize  uint32
+	TraceMode          string
+	TraceType          string
+	KernelModules      []string
 }
 
 // KataAgentState is the structure describing the data stored from this
@@ -294,6 +300,11 @@ func KataAgentKernelParams(config KataAgentConfig) []Param {
 		params = append(params, Param{Key: vcAnnotations.ContainerPipeSizeKernelParam, Value: containerPipeSize})
 	}
 
+	if config.EnableDebugConsole {
+		params = append(params, Param{Key: kernelParamDebugConsole, Value: ""})
+		params = append(params, Param{Key: kernelParamDebugConsoleVPort, Value: kernelParamDebugConsoleVPortValue})
+	}
+
 	return params
 }
 
@@ -1208,16 +1219,6 @@ func (k *kataAgent) buildContainerRootfs(sandbox *Sandbox, c *Container, rootPat
 	return nil, nil
 }
 
-func (k *kataAgent) hasAgentDebugConsole(sandbox *Sandbox) bool {
-	for _, p := range sandbox.config.HypervisorConfig.KernelParams {
-		if p.Key == "agent.debug_console" {
-			k.Logger().Info("agent has debug console")
-			return true
-		}
-	}
-	return false
-}
-
 func (k *kataAgent) createContainer(sandbox *Sandbox, c *Container) (p *Process, err error) {
 	span, _ := k.trace("createContainer")
 	defer span.Finish()

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

@@ -178,7 +178,7 @@ func parse(sock string) (string, *url.URL, error) {
 func agentDialer(addr *url.URL) dialer {
 	switch addr.Scheme {
 	case VSockSocketScheme:
-		return vsockDialer
+		return VsockDialer
 	case HybridVSockScheme:
 		return HybridVSockDialer
 	case MockHybridVSockScheme:
@@ -278,7 +278,7 @@ func commonDialer(timeout time.Duration, dialFunc func() (net.Conn, error), time
 	return conn, nil
 }
 
-func vsockDialer(sock string, timeout time.Duration) (net.Conn, error) {
+func VsockDialer(sock string, timeout time.Duration) (net.Conn, error) {
 	cid, port, err := parseGrpcVsockAddr(sock)
 	if err != nil {
 		return nil, err

src/runtime/virtcontainers/pkg/annotations/annotations.go

@@ -245,7 +245,7 @@ const (
 	// The following example can be used to load two kernel modules with parameters
 	///
 	//   annotations:
-	//     io.kata-containers.config.agent.kernel_modules: "e1000e InterruptThrottleRate=3000,3000,3000 EEE=1; i915 enable_ppgtt=0"
+	//     io.katacontainers.config.agent.kernel_modules: "e1000e InterruptThrottleRate=3000,3000,3000 EEE=1; i915 enable_ppgtt=0"
 	//
 	// The first word is considered as the module name and the rest as its parameters.
 	//

src/runtime/virtcontainers/pkg/vcmock/sandbox.go

@@ -247,3 +247,10 @@ func (s *Sandbox) Stats() (vc.SandboxStats, error) {
 	}
 	return vc.SandboxStats{}, nil
 }
+
+func (s *Sandbox) GetAgentURL() (string, error) {
+	if s.GetAgentURLFunc != nil {
+		return s.GetAgentURLFunc()
+	}
+	return "", nil
+}

src/runtime/virtcontainers/pkg/vcmock/types.go

@@ -67,6 +67,7 @@ type Sandbox struct {
 	UpdateRuntimeMetricsFunc func() error
 	GetAgentMetricsFunc      func() (string, error)
 	StatsFunc                func() (vc.SandboxStats, error)
+	GetAgentURLFunc          func() (string, error)
 }
 
 // Container is a fake Container type used for testing

src/runtime/virtcontainers/sandbox.go

@@ -2271,3 +2271,7 @@ func (s *Sandbox) GetPatchedOCISpec() *specs.Spec {
 func (s *Sandbox) GetOOMEvent() (string, error) {
 	return s.agent.getOOMEvent()
 }
+
+func (s *Sandbox) GetAgentURL() (string, error) {
+	return s.agent.getAgentURL()
+}

src/runtime/virtcontainers/vm_test.go

@@ -101,7 +101,15 @@ func TestVMConfigGrpc(t *testing.T) {
 	config := VMConfig{
 		HypervisorType:   QemuHypervisor,
 		HypervisorConfig: newQemuConfig(),
-		AgentConfig:      KataAgentConfig{true, false, false, 0, "", "", []string{}},
+		AgentConfig: KataAgentConfig{
+			LongLiveConn:       true,
+			Debug:              false,
+			Trace:              false,
+			EnableDebugConsole: false,
+			ContainerPipeSize:  0,
+			TraceMode:          "",
+			TraceType:          "",
+			KernelModules:      []string{}},
 	}
 
 	p, err := config.ToGrpc()