4 files changed, 437 insertions, 34 deletions

diff --git a/meta/lib/oeqa/core/target/__init__.py b/meta/lib/oeqa/core/target/__init__.py
index 1382aa9b52..177f648fe3 100644
--- a/meta/lib/oeqa/core/target/__init__.py
+++ b/meta/lib/oeqa/core/target/__init__.py
@@ -10,6 +10,7 @@ class OETarget(object):
 
     def __init__(self, logger, *args, **kwargs):
         self.logger = logger
+        self.runner = None
 
     @abstractmethod
     def start(self):
diff --git a/meta/lib/oeqa/core/target/qemu.py b/meta/lib/oeqa/core/target/qemu.py
index 0f29414df5..d93b3ac94a 100644
--- a/meta/lib/oeqa/core/target/qemu.py
+++ b/meta/lib/oeqa/core/target/qemu.py
@@ -8,20 +8,21 @@ import os
 import sys
 import signal
 import time
+import glob
+import subprocess
 from collections import defaultdict
 
 from .ssh import OESSHTarget
 from oeqa.utils.qemurunner import QemuRunner
-from oeqa.utils.dump import TargetDumper
 
 supported_fstypes = ['ext3', 'ext4', 'cpio.gz', 'wic']
 
 class OEQemuTarget(OESSHTarget):
     def __init__(self, logger, server_ip, timeout=300, user='root',
             port=None, machine='', rootfs='', kernel='', kvm=False, slirp=False,
-            dump_dir='', dump_host_cmds='', display='', bootlog='',
+            dump_dir='', display='', bootlog='',
             tmpdir='', dir_image='', boottime=60, serial_ports=2,
-            boot_patterns = defaultdict(str), ovmf=False, **kwargs):
+            boot_patterns = defaultdict(str), ovmf=False, tmpfsdir=None, **kwargs):
 
         super(OEQemuTarget, self).__init__(logger, None, server_ip, timeout,
                 user, port)
@@ -35,17 +36,15 @@ class OEQemuTarget(OESSHTarget):
         self.ovmf = ovmf
         self.use_slirp = slirp
         self.boot_patterns = boot_patterns
+        self.dump_dir = dump_dir
+        self.bootlog = bootlog
 
         self.runner = QemuRunner(machine=machine, rootfs=rootfs, tmpdir=tmpdir,
                                  deploy_dir_image=dir_image, display=display,
                                  logfile=bootlog, boottime=boottime,
-                                 use_kvm=kvm, use_slirp=slirp, dump_dir=dump_dir,
-                                 dump_host_cmds=dump_host_cmds, logger=logger,
+                                 use_kvm=kvm, use_slirp=slirp, dump_dir=dump_dir, logger=logger,
                                  serial_ports=serial_ports, boot_patterns = boot_patterns, 
-                                 use_ovmf=ovmf)
-        dump_target_cmds = kwargs.get("testimage_dump_target")
-        self.target_dumper = TargetDumper(dump_target_cmds, dump_dir, self.runner)
-        self.target_dumper.create_dir("qemu")
+                                 use_ovmf=ovmf, tmpfsdir=tmpfsdir)
 
     def start(self, params=None, extra_bootparams=None, runqemuparams=''):
         if self.use_slirp and not self.server_ip:
@@ -68,7 +67,28 @@ class OEQemuTarget(OESSHTarget):
                 self.server_ip = self.runner.server_ip
         else:
             self.stop()
-            raise RuntimeError("FAILED to start qemu - check the task log and the boot log")
+            # Display the first 20 lines of top and
+            # last 20 lines of the bootlog when the
+            # target is not being booted up.
+            topfile = glob.glob(self.dump_dir + "/*_qemu/host_*_top")
+            msg = "\n\n===== start: snippet =====\n\n"
+            for f in topfile:
+                msg += "file: %s\n\n" % f
+                with open(f) as tf:
+                    for x in range(20):
+                        msg += next(tf)
+            msg += "\n\n===== end: snippet =====\n\n"
+            blcmd = ["tail", "-20", self.bootlog]
+            msg += "===== start: snippet =====\n\n"
+            try:
+                out = subprocess.check_output(blcmd, stderr=subprocess.STDOUT, timeout=1).decode('utf-8')
+                msg += "file: %s\n\n" % self.bootlog
+                msg += out
+            except (subprocess.CalledProcessError, subprocess.TimeoutExpired, FileNotFoundError) as err:
+                msg += "Error running command: %s\n%s\n" % (blcmd, err)
+            msg += "\n\n===== end: snippet =====\n"
+
+            raise RuntimeError("FAILED to start qemu - check the task log and the boot log %s" % (msg))
 
     def stop(self):
         self.runner.stop()
diff --git a/meta/lib/oeqa/core/target/serial.py b/meta/lib/oeqa/core/target/serial.py
new file mode 100644
index 0000000000..7c2cd8b248
--- /dev/null
+++ b/meta/lib/oeqa/core/target/serial.py
@@ -0,0 +1,315 @@
+#
+# SPDX-License-Identifier: MIT
+#
+
+import base64
+import logging
+import os
+from threading import Lock
+from . import OETarget
+
+class OESerialTarget(OETarget):
+
+    def __init__(self, logger, target_ip, server_ip, server_port=0,
+                 timeout=300, serialcontrol_cmd=None, serialcontrol_extra_args=None,
+                 serialcontrol_ps1=None, serialcontrol_connect_timeout=None, 
+                 machine=None, **kwargs):
+        if not logger:
+            logger = logging.getLogger('target')
+            logger.setLevel(logging.INFO)
+            filePath = os.path.join(os.getcwd(), 'remoteTarget.log')
+            fileHandler = logging.FileHandler(filePath, 'w', 'utf-8')
+            formatter = logging.Formatter(
+                        '%(asctime)s.%(msecs)03d %(levelname)s: %(message)s',
+                        '%H:%M:%S')
+            fileHandler.setFormatter(formatter)
+            logger.addHandler(fileHandler)
+
+        super(OESerialTarget, self).__init__(logger)
+
+        if serialcontrol_ps1:
+            self.target_ps1 = serialcontrol_ps1
+        elif machine:
+            # fallback to a default value which assumes root@machine
+            self.target_ps1 = f'root@{machine}:.*# '
+        else:
+            raise ValueError("Unable to determine shell command prompt (PS1) format.")
+
+        if not serialcontrol_cmd:
+            raise ValueError("Unable to determine serial control command.")
+
+        if serialcontrol_extra_args:
+            self.connection_script = f'{serialcontrol_cmd} {serialcontrol_extra_args}'
+        else:
+            self.connection_script = serialcontrol_cmd
+
+        if serialcontrol_connect_timeout:
+            self.connect_timeout = serialcontrol_connect_timeout
+        else:
+            self.connect_timeout = 10 # default to 10s connection timeout
+
+        self.default_command_timeout = timeout
+        self.ip = target_ip
+        self.server_ip = server_ip
+        self.server_port = server_port 
+        self.conn = None
+        self.mutex = Lock()
+
+    def start(self, **kwargs):
+        pass
+
+    def stop(self, **kwargs):
+        pass
+
+    def get_connection(self):
+        if self.conn is None:
+            self.conn = SerialConnection(self.connection_script,
+                                         self.target_ps1,
+                                         self.connect_timeout,
+                                         self.default_command_timeout)
+
+        return self.conn
+
+    def run(self, cmd, timeout=None):
+        """
+            Runs command on target over the provided serial connection.
+            The first call will open the connection, and subsequent
+            calls will re-use the same connection to send new commands.
+
+            command:    Command to run on target.
+            timeout:    <value>:    Kill command after <val> seconds.
+                        None:       Kill command default value seconds.
+                        0:          No timeout, runs until return.
+        """
+        # Lock needed to avoid multiple threads running commands concurrently
+        # A serial connection can only be used by one caller at a time
+        with self.mutex:
+            conn = self.get_connection()
+
+            self.logger.debug(f"[Running]$ {cmd}")
+            # Run the command, then echo $? to get the command's return code
+            try:
+                output = conn.run_command(cmd, timeout)
+                status = conn.run_command("echo $?")
+                self.logger.debug(f"   [stdout]: {output}")
+                self.logger.debug(f"   [ret code]: {status}\n\n")
+            except SerialTimeoutException as e:
+                self.logger.debug(e)
+                output = ""
+                status = 255
+
+            # Return to $HOME after each command to simulate a stateless SSH connection
+            conn.run_command('cd "$HOME"')
+
+        return (int(status), output)
+
+    def copyTo(self, localSrc, remoteDst):
+        """
+            Copies files by converting them to base 32, then transferring
+            the ASCII text to the target, and decoding it in place on the
+            target.
+
+            On a 115k baud serial connection, this method transfers at
+            roughly 30kbps.
+        """
+        with open(localSrc, 'rb') as file:
+            data = file.read()
+
+        b32 = base64.b32encode(data).decode('utf-8')
+
+        # To avoid shell line limits, send a chunk at a time
+        SPLIT_LEN = 512 
+        lines = [b32[i:i+SPLIT_LEN] for i in range(0, len(b32), SPLIT_LEN)]
+
+        with self.mutex:
+            conn = self.get_connection()
+
+            filename = os.path.basename(localSrc)
+            TEMP = f'/tmp/{filename}.b32'
+
+            # Create or empty out the temp file
+            conn.run_command(f'echo -n "" > {TEMP}')
+
+            for line in lines:
+                conn.run_command(f'echo -n {line} >> {TEMP}')
+
+            # Check to see whether the remoteDst is a directory
+            is_directory = conn.run_command(f'[[ -d {remoteDst} ]]; echo $?')
+            if int(is_directory) == 0:
+                # append the localSrc filename to the end of remoteDst
+                remoteDst = os.path.join(remoteDst, filename)
+
+            conn.run_command(f'base32 -d {TEMP} > {remoteDst}')
+            conn.run_command(f'rm {TEMP}')
+
+        return 0, 'Success'
+
+    def copyFrom(self, remoteSrc, localDst):
+        """
+            Copies files by converting them to base 32 on the target, then
+            transferring the ASCII text to the host. That text is then
+            decoded here and written out to the destination.
+
+            On a 115k baud serial connection, this method transfers at
+            roughly 30kbps.
+        """
+        with self.mutex:
+            b32 = self.get_connection().run_command(f'base32 {remoteSrc}')
+
+            data = base64.b32decode(b32.replace('\r\n', ''))
+
+            # If the local path is a directory, get the filename from
+            # the remoteSrc path and append it to localDst
+            if os.path.isdir(localDst):
+                filename = os.path.basename(remoteSrc)
+                localDst = os.path.join(localDst, filename)
+
+            with open(localDst, 'wb') as file:
+                file.write(data)
+
+        return 0, 'Success'
+
+    def copyDirTo(self, localSrc, remoteDst):
+        """
+            Copy recursively localSrc directory to remoteDst in target.
+        """
+
+        for root, dirs, files in os.walk(localSrc):
+            # Create directories in the target as needed
+            for d in dirs:
+                tmpDir = os.path.join(root, d).replace(localSrc, "")
+                newDir = os.path.join(remoteDst, tmpDir.lstrip("/"))
+                cmd = "mkdir -p %s" % newDir
+                self.run(cmd)
+
+            # Copy files into the target
+            for f in files:
+                tmpFile = os.path.join(root, f).replace(localSrc, "")
+                dstFile = os.path.join(remoteDst, tmpFile.lstrip("/"))
+                srcFile = os.path.join(root, f)
+                self.copyTo(srcFile, dstFile)
+
+    def deleteFiles(self, remotePath, files):
+        """
+            Deletes files in target's remotePath.
+        """
+
+        cmd = "rm"
+        if not isinstance(files, list):
+            files = [files]
+
+        for f in files:
+            cmd = "%s %s" % (cmd, os.path.join(remotePath, f))
+
+        self.run(cmd)
+
+    def deleteDir(self, remotePath):
+        """
+            Deletes target's remotePath directory.
+        """
+
+        cmd = "rmdir %s" % remotePath
+        self.run(cmd)
+
+    def deleteDirStructure(self, localPath, remotePath):
+        """
+        Delete recursively localPath structure directory in target's remotePath.
+
+        This function is useful to delete a package that is installed in the
+        device under test (DUT) and the host running the test has such package
+        extracted in tmp directory.
+
+        Example:
+            pwd: /home/user/tmp
+            tree:   .
+                    └── work
+                        ├── dir1
+                        │   └── file1
+                        └── dir2
+
+            localpath = "/home/user/tmp" and remotepath = "/home/user"
+
+            With the above variables this function will try to delete the
+            directory in the DUT in this order:
+                /home/user/work/dir1/file1
+                /home/user/work/dir1        (if dir is empty)
+                /home/user/work/dir2        (if dir is empty)
+                /home/user/work             (if dir is empty)
+        """
+
+        for root, dirs, files in os.walk(localPath, topdown=False):
+            # Delete files first
+            tmpDir = os.path.join(root).replace(localPath, "")
+            remoteDir = os.path.join(remotePath, tmpDir.lstrip("/"))
+            self.deleteFiles(remoteDir, files)
+
+            # Remove dirs if empty
+            for d in dirs:
+                tmpDir = os.path.join(root, d).replace(localPath, "")
+                remoteDir = os.path.join(remotePath, tmpDir.lstrip("/"))
+                self.deleteDir(remoteDir)
+
+class SerialTimeoutException(Exception):
+    def __init__(self, msg):
+        self.msg = msg
+    def __str__(self):
+        return self.msg
+
+class SerialConnection:
+
+    def __init__(self, script, target_prompt, connect_timeout, default_command_timeout):
+        import pexpect # limiting scope to avoid build dependency
+        self.prompt = target_prompt
+        self.connect_timeout = connect_timeout
+        self.default_command_timeout = default_command_timeout
+        self.conn = pexpect.spawn('/bin/bash', ['-c', script], encoding='utf8')
+        self._seek_to_clean_shell()
+        # Disable echo to avoid the need to parse the outgoing command
+        self.run_command('stty -echo')
+
+    def _seek_to_clean_shell(self):
+        """
+            Attempts to find a clean shell, meaning it is clear and
+            ready to accept a new command. This is necessary to ensure
+            the correct output is captured from each command.
+        """
+        import pexpect # limiting scope to avoid build dependency
+        # Look for a clean shell
+        # Wait a short amount of time for the connection to finish
+        pexpect_code = self.conn.expect([self.prompt, pexpect.TIMEOUT],
+                                        timeout=self.connect_timeout)
+
+        # if a timeout occurred, send an empty line and wait for a clean shell
+        if pexpect_code == 1:
+            # send a newline to clear and present the shell
+            self.conn.sendline("")
+            pexpect_code = self.conn.expect(self.prompt)
+
+    def run_command(self, cmd, timeout=None):
+        """
+            Runs command on target over the provided serial connection.
+            Returns any output on the shell while the command was run.
+
+            command:    Command to run on target.
+            timeout:    <value>:    Kill command after <val> seconds.
+                        None:       Kill command default value seconds.
+                        0:          No timeout, runs until return.
+        """
+        import pexpect # limiting scope to avoid build dependency
+        # Convert from the OETarget defaults to pexpect timeout values
+        if timeout is None:
+            timeout = self.default_command_timeout
+        elif timeout == 0:
+            timeout = None # passing None to pexpect is infinite timeout
+
+        self.conn.sendline(cmd)
+        pexpect_code = self.conn.expect([self.prompt, pexpect.TIMEOUT], timeout=timeout)
+
+        # check for timeout
+        if pexpect_code == 1:
+            self.conn.send('\003') # send Ctrl+C
+            self._seek_to_clean_shell()
+            raise SerialTimeoutException(f'Timeout executing: {cmd} after {timeout}s')
+
+        return self.conn.before.removesuffix('\r\n')
+
diff --git a/meta/lib/oeqa/core/target/ssh.py b/meta/lib/oeqa/core/target/ssh.py
index 461448dbc5..8b5c450a05 100644
--- a/meta/lib/oeqa/core/target/ssh.py
+++ b/meta/lib/oeqa/core/target/ssh.py
@@ -34,12 +34,17 @@ class OESSHTarget(OETarget):
         self.timeout = timeout
         self.user = user
         ssh_options = [
+                '-o', 'ServerAliveCountMax=2',
+                '-o', 'ServerAliveInterval=30',
                 '-o', 'UserKnownHostsFile=/dev/null',
                 '-o', 'StrictHostKeyChecking=no',
                 '-o', 'LogLevel=ERROR'
                 ]
+        scp_options = [
+                '-r'
+        ]
         self.ssh = ['ssh', '-l', self.user ] + ssh_options
-        self.scp = ['scp'] + ssh_options
+        self.scp = ['scp'] + ssh_options + scp_options
         if port:
             self.ssh = self.ssh + [ '-p', port ]
             self.scp = self.scp + [ '-P', port ]
@@ -50,14 +55,14 @@ class OESSHTarget(OETarget):
     def stop(self, **kwargs):
         pass
 
-    def _run(self, command, timeout=None, ignore_status=True):
+    def _run(self, command, timeout=None, ignore_status=True, raw=False):
         """
             Runs command in target using SSHProcess.
         """
         self.logger.debug("[Running]$ %s" % " ".join(command))
 
         starttime = time.time()
-        status, output = SSHCall(command, self.logger, timeout)
+        status, output = SSHCall(command, self.logger, timeout, raw)
         self.logger.debug("[Command returned '%d' after %.2f seconds]"
                  "" % (status, time.time() - starttime))
 
@@ -67,7 +72,7 @@ class OESSHTarget(OETarget):
 
         return (status, output)
 
-    def run(self, command, timeout=None):
+    def run(self, command, timeout=None, ignore_status=True, raw=False):
         """
             Runs command in target.
 
@@ -86,10 +91,12 @@ class OESSHTarget(OETarget):
         else:
             processTimeout = self.timeout
 
-        status, output = self._run(sshCmd, processTimeout, True)
-        self.logger.debug('Command: %s\nOutput:  %s\n' % (command, output))
-        if (status == 255) and (('No route to host') in output):
-            self.target_dumper.dump_target()
+        status, output = self._run(sshCmd, processTimeout, ignore_status, raw)
+        if len(output) > (64 * 1024):
+            self.logger.debug('Command: %s\nStatus: %d Output length:  %s\n' % (command, status, len(output)))
+        else:
+            self.logger.debug('Command: %s\nStatus: %d Output:  %s\n' % (command, status, output))
+
         return (status, output)
 
     def copyTo(self, localSrc, remoteDst):
@@ -202,32 +209,51 @@ class OESSHTarget(OETarget):
                 remoteDir = os.path.join(remotePath, tmpDir.lstrip("/"))
                 self.deleteDir(remoteDir)
 
-def SSHCall(command, logger, timeout=None, **opts):
+def SSHCall(command, logger, timeout=None, raw=False, **opts):
 
     def run():
         nonlocal output
         nonlocal process
+        output_raw = bytearray()
         starttime = time.time()
+        progress = time.time()
         process = subprocess.Popen(command, **options)
+        has_timeout = False
+        appendline = None
         if timeout:
             endtime = starttime + timeout
             eof = False
-            while time.time() < endtime and not eof:
-                logger.debug('time: %s, endtime: %s' % (time.time(), endtime))
+            os.set_blocking(process.stdout.fileno(), False)
+            while not has_timeout and not eof:
                 try:
                     if select.select([process.stdout], [], [], 5)[0] != []:
-                        reader = codecs.getreader('utf-8')(process.stdout, 'ignore')
-                        data = reader.read(1024, 4096)
+                        # wait a bit for more data, tries to avoid reading single characters
+                        time.sleep(0.2)
+                        data = process.stdout.read()
                         if not data:
-                            process.stdout.close()
                             eof = True
                         else:
-                            output += data
-                            logger.debug('Partial data from SSH call: %s' % data)
+                            output_raw.extend(data)
+                            # ignore errors to capture as much as possible
+                            #logger.debug('Partial data from SSH call:\n%s' % data.decode('utf-8', errors='ignore'))
                             endtime = time.time() + timeout
                 except InterruptedError:
+                    logger.debug('InterruptedError')
+                    continue
+                except BlockingIOError:
+                    logger.debug('BlockingIOError')
                     continue
 
+                if time.time() >= endtime:
+                    logger.debug('SSHCall has timeout! Time: %s, endtime: %s' % (time.time(), endtime))
+                    has_timeout = True
+
+                if time.time() >= (progress + 60):
+                    logger.debug('Waiting for process output at time: %s with datasize: %s' % (time.time(), len(output_raw)))
+                    progress = time.time()
+
+            process.stdout.close()
+
             # process hasn't returned yet
             if not eof:
                 process.terminate()
@@ -235,20 +261,58 @@ def SSHCall(command, logger, timeout=None, **opts):
                 try:
                     process.kill()
                 except OSError:
+                    logger.debug('OSError when killing process')
                     pass
                 endtime = time.time() - starttime
-                lastline = ("\nProcess killed - no output for %d seconds. Total"
+                appendline = ("\nProcess killed - no output for %d seconds. Total"
                             " running time: %d seconds." % (timeout, endtime))
-                logger.debug('Received data from SSH call %s ' % lastline)
-                output += lastline
-
+                logger.debug('Received data from SSH call:\n%s ' % appendline)
+                process.wait()
+
+            if raw:
+                output = bytes(output_raw)
+                if appendline:
+                    output += bytes(appendline, "utf-8")
+            else:
+                output = output_raw.decode('utf-8', errors='ignore')
+                if appendline:
+                    output += appendline
         else:
-            output = process.communicate()[0].decode('utf-8', errors='ignore')
-            logger.debug('Data from SSH call: %s' % output.rstrip())
+            output = output_raw = process.communicate()[0]
+            if not raw:
+                output = output_raw.decode('utf-8', errors='ignore')
+
+        if len(output) < (64 * 1024):
+            if output.rstrip():
+                logger.debug('Data from SSH call:\n%s' % output.rstrip())
+            else:
+                logger.debug('No output from SSH call')
+
+        # timout or not, make sure process exits and is not hanging
+        if process.returncode == None:
+            try:
+                process.wait(timeout=5)
+            except TimeoutExpired:
+                try:
+                    process.kill()
+                except OSError:
+                    logger.debug('OSError')
+                    pass
+                process.wait()
+
+        if has_timeout:
+            # Version of openssh before 8.6_p1 returns error code 0 when killed
+            # by a signal, when the timeout occurs we will receive a 0 error
+            # code because the process is been terminated and it's wrong because
+            # that value means success, but the process timed out.
+            # Afterwards, from version 8.6_p1 onwards, the returned code is 255.
+            # Fix this behaviour by checking the return code
+            if process.returncode == 0:
+                process.returncode = 255
 
     options = {
         "stdout": subprocess.PIPE,
-        "stderr": subprocess.STDOUT,
+        "stderr": subprocess.STDOUT if not raw else None,
         "stdin": None,
         "shell": False,
         "bufsize": -1,
@@ -271,6 +335,9 @@ def SSHCall(command, logger, timeout=None, **opts):
         # whilst running and ensure we don't leave a process behind.
         if process.poll() is None:
             process.kill()
+        if process.returncode == None:
+            process.wait()
         logger.debug('Something went wrong, killing SSH process')
         raise
-    return (process.wait(), output.rstrip())
+
+    return (process.returncode, output if raw else output.rstrip())