E2E image mode enablement for DPUs

clustersConfig.py

@@ -77,15 +77,16 @@ class ExtraConfigArgs:
     kickstart: Optional[str] = None
     remove_args: Optional[str] = None
 
-    registry: Optional[list[RegistryInfo]] = None
-
     # Custom OVN repo URL
     ovn_repo: Optional[str] = None
     # Custom OVN ref, it should be existing commit hash or branch
     ovn_ref: Optional[str] = None
 
     registries: Optional[list[RegistryInfo]] = None
-    import_pull_secret: bool = False
+    import_pull_secret: bool = True
+    bootc_dir: str = "rhel-image-mode-4-dpu"
+    bootc_build_local: bool = True
+    iso_builder_auth_file: Optional[str] = None
 
     def __post_init__(self) -> None:
         if self.registries is not None:

dpuVendor.py

@@ -39,14 +39,25 @@ def setup(self, dpu: host.Host) -> None:
     def download_p4_tar(self, rh: host.Host) -> None:
         logger.info("Downloading p4.tar.gz")
         rh.run_or_die(f"curl -L {self.P4_URL} -o /tmp/p4.tar.gz")
-        rh.run("rm -rf /opt/p4")
-        rh.run_or_die("tar -U -C /opt/ -xzf /tmp/p4.tar.gz")
-        rh.run("mv /opt/intel-ipu-acc-components-2.0.0.11126 /opt/p4")
+        rh.run("rm -rf /opt/p4/*")
+        rh.run_or_die("tar -U -C /tmp/ -xzf /tmp/p4.tar.gz")
+        rh.run("mv /tmp/intel-ipu-acc-components-2.0.0.11126/* /opt/p4/")
         rh.run("mv /opt/p4/p4-cp /opt/p4/p4-cp-nws")
         rh.run("mv /opt/p4/p4-sde /opt/p4/p4sde")
+        rh.run("rm -rf /tmp/intel-ipu-acc-components-2.0.0.11126")
 
     def configure_p4_hugepages(self, rh: host.Host) -> None:
         logger.info("Configuring hugepages for p4 pod")
+
+        # Check if the service already exists - skip if it does
+        try:
+            result = rh.run("systemctl list-unit-files hugepages-setup.service")
+            if "hugepages-setup.service" in result.out:
+                logger.info("hugepages-setup.service already exists, skipping configuration")
+                return
+        except Exception as e:
+            logger.debug(f"Could not check for existing service: {e}, proceeding with setup")
+
         # The p4 container typically sets this up. If we are running the container as a daemonset in microshift, we need to
         # ensure this resource is available prior to the pod starting to ensure dpdk is successful
         hugepages_service = """[Unit]

extraConfigHostRegistry.py

@@ -1,5 +1,5 @@
 from concurrent.futures import Future
-from typing import Optional
+from typing import Optional, Any
 from auth import import_secret_path
 from clustersConfig import ClustersConfig
 from logger import logger
@@ -12,31 +12,25 @@
 def ExtraConfigHostRegistry(cc: ClustersConfig, cfg: ExtraConfigArgs, futures: dict[str, Future[Optional[host.Result]]]) -> None:
     [f.result() for (_, f) in futures.items()]
 
-    auth_path = "/run/user/0/containers/auth.json"
-
-    # Load existing data
-    if os.path.exists(auth_path):
-        with open(auth_path, "r") as f:
-            try:
-                auth_data = json.load(f)
-            except json.JSONDecodeError:
-                logger.warning(f"{auth_path} is invalid. Resetting.")
-                auth_data = {"auths": {}}
-    else:
-        auth_data = {"auths": {}}
-
-    # Merge pull secret
-    if cfg.import_pull_secret:
-        logger.info(f"Importing secrets_path: {cc.secrets_path}")
-        auth_data["auths"].update(import_secret_path(cc.secrets_path))
-
     # Merge other user-defined registries
     if cfg.registries:
         logger.info(f"Preparing registries: {cfg.registries}")
         for r in cfg.registries:
+            auth_data: dict[str, Any] = {"auths": {}}
+            if cfg.import_pull_secret:
+                logger.info(f"Importing secrets_path: {cc.secrets_path}")
+                auth_data["auths"].update(import_secret_path(cc.secrets_path))
+            if os.path.exists(r.auth_path):
+                with open(r.auth_path, "r") as f:
+                    try:
+                        auth_data = json.load(f)
+                    except json.JSONDecodeError:
+                        logger.warning(f"{r.auth_path} is invalid. Resetting.")
+                        auth_data = {"auths": {}}
+            else:
+                auth_data = {"auths": {}}
             auth_data["auths"].update(r.prep_auth())
-
-    # Write once
-    os.makedirs(os.path.dirname(auth_path), exist_ok=True)
-    with open(auth_path, "w") as f:
-        json.dump(auth_data, f, indent=2)
+            os.makedirs(os.path.dirname(r.auth_path), exist_ok=True)
+            with open(r.auth_path, "w") as f:
+                json.dump(auth_data, f, indent=2)
+            logger.info(f"Successfully wrote {len(auth_data['auths'])} registry auth entries to {r.auth_path}")

extraConfigIsoBuilder.py

@@ -38,6 +38,9 @@ def ExtraConfigIsoBuilder(
     activation_key: str = cast(str, cfg.activation_key)
     image_mode_url: str = cfg.image_mode_url
     iso_builder_url: str = cfg.iso_builder_url
+    bootc_build_local: bool = cfg.bootc_build_local
+    bootc_dir: str = cfg.bootc_dir
+    iso_builder_auth_file: Optional[str] = cfg.iso_builder_auth_file
 
     # Optional bootc iso build params
     input_iso: Optional[str] = cfg.input_iso
@@ -52,6 +55,10 @@ def ExtraConfigIsoBuilder(
     if dpu_flavor == "ipu":
         extra_args = " ip=192.168.0.2:::255.255.255.0::enp0s1f0:off netroot=iscsi:192.168.0.1::::iqn.e2000:acc acpi=force"
         kernel_args = (kernel_args or "") + extra_args
+        remove_args = "rd.live.check"
+        grub_replacements = [
+            "timeout=60|timeout=5",
+        ]
 
     # Build the ISO
     BootcIsoBuilder(
@@ -67,4 +74,8 @@ def ExtraConfigIsoBuilder(
         kernel_args=kernel_args,
         remove_args=remove_args,
         dpu_flavor=dpu_flavor,
+        auth_file_path=iso_builder_auth_file,  # Use auth file from iso_builder config
+        bootc_build_local=bootc_build_local,
+        grub_replacements=grub_replacements,
+        bootc_dir=bootc_dir,
     ).build()

extraConfigMicroshift.py

@@ -88,47 +88,97 @@ def ExtraConfigMicroshift(cc: ClustersConfig, cfg: ExtraConfigArgs, futures: dic
 
     # Configure firewalld for microshift
     logger.info("Configuring firewall for microshift")
-    acc.run("systemctl disable firewalld")
-    acc.run("systemctl stop firewalld")
+    fw_active = acc.run("systemctl is-active firewalld")
+    fw_enabled = acc.run("systemctl is-enabled firewalld")
+
+    if fw_active.success() and "active" in fw_active.out:
+        logger.info("Stopping firewalld service")
+        acc.run("systemctl stop firewalld")
+
+    if fw_enabled.success() and "enabled" in fw_enabled.out:
+        logger.info("Disabling firewalld service")
+        acc.run("systemctl disable firewalld")
 
     # Adjust the timeout for microshift service to ensure it starts successfully
-    acc.run_or_die("mkdir -p /etc/systemd/system/microshift.service.d/")
-    acc.write("/etc/systemd/system/microshift.service.d/override.conf", "[Service]\nTimeoutStartSec=15m")
+    acc.run("mkdir -p /etc/systemd/system/microshift.service.d/")
+    override_content = "[Service]\nTimeoutStartSec=15m"
+    existing_override = acc.run("cat /etc/systemd/system/microshift.service.d/override.conf 2>/dev/null")
+    service_override_changed = False
+    if not existing_override.success() or override_content not in existing_override.out:
+        logger.info("Writing microshift service timeout override")
+        acc.write("/etc/systemd/system/microshift.service.d/override.conf", override_content)
+        acc.run("systemctl daemon-reload")
+        service_override_changed = True
+
+    # Check if microshift is already installed
+    ms_installed = acc.run("rpm -q microshift")
+    multus_installed = acc.run("rpm -q microshift-multus")
+    microshift_already_installed = ms_installed.success() and multus_installed.success()
+
+    # Only add early access repo if microshift is not already installed
+    if not microshift_already_installed:
+        repo_exists = acc.run("test -f /etc/yum.repos.d/microshift-canidate.repo")
+        if not repo_exists.success():
+            logger.info("Writing microshift candidate repository")
+            acc.write("/etc/yum.repos.d/microshift-canidate.repo", early_access_microshift())
+    else:
+        logger.info("Microshift already installed, skipping early access repository setup")
 
-    # Check on the status of the cluster
-    acc.write("/etc/yum.repos.d/microshift-canidate.repo", early_access_microshift())
     time.sleep(1)
     logger.info("Checking if time is set properly to avoid OCSR errors")
     logger.info(acc.run("systemctl status chronyd --no-pager -l"))
     lh_date = host.LocalHost().run("date").out.strip()
-    acc_date = host.LocalHost().run("date").out.strip()
+    acc_date = acc.run("date").out.strip()
     logger.info(f"LocalHost date: {lh_date}")
     logger.info(f"ACC date: {acc_date}")
     logger.info("Manually synchronizing time")
     host.sync_time(lh, acc)
     lh_date = host.LocalHost().run("date").out.strip()
-    acc_date = host.LocalHost().run("date").out.strip()
+    acc_date = acc.run("date").out.strip()
     logger.info(f"LocalHost date: {lh_date}")
     logger.info(f"ACC date: {acc_date}")
 
-    logger.info("Installing microshift")
-    acc.run_or_die("dnf install -y microshift microshift-multus", retry=60)
+    # Install microshift packages (idempotent)
+    config_changed = service_override_changed
+
+    if not microshift_already_installed:
+        logger.info("Installing microshift")
+        acc.run_or_die("dnf install -y microshift microshift-multus", retry=60)
+        config_changed = True
+    else:
+        logger.info("Microshift packages already installed")
+
+    # Configure crio runtime (idempotent)
     ret = acc.run(r"grep '\[crio.runtime.runtimes.crun\]' /etc/crio/crio.conf")
     if not ret.success():
+        logger.info("Adding crun configuration to crio.conf")
         crun_conf_lines = ['[crio.runtime.runtimes.crun]', 'runtime_path = "/usr/bin/crun"', 'runtime_type = "oci"', 'runtime_root = "/run/crun"']
         for line in crun_conf_lines:
             acc.run(f'echo \'{line}\' >> /etc/crio/crio.conf')
-    acc.run("systemctl restart crio.service")
-    logger.info("Starting microshift")
-    acc.run("systemctl restart microshift")
-    acc.run("systemctl enable microshift")
+        acc.run("systemctl restart crio.service")
+        config_changed = True
+    else:
+        logger.info("crun configuration already present in crio.conf")
+
+    # Start and enable microshift (idempotent)
+    logger.info("Managing microshift service")
+    ms_enabled = acc.run("systemctl is-enabled microshift")
+    if not ms_enabled.success() or "enabled" not in ms_enabled.out:
+        logger.info("Enabling microshift service")
+        acc.run("systemctl enable microshift")
+        config_changed = True
+
+    ms_active = acc.run("systemctl is-active microshift")
+    if not ms_active.success() or "active" not in ms_active.out:
+        logger.info("Starting microshift service")
+        acc.run("systemctl restart microshift")
+    elif config_changed:
+        logger.info("Configuration changed, restarting microshift service")
+        acc.run("systemctl restart microshift")
 
     contents = read_prep_microshift_kubeconfig(acc)
     kubeconfig = write_microshift_kubeconfig(contents, host.LocalHost())
 
-    acc.run("systemctl stop firewalld")
-    acc.run("systemctl disable firewalld")
-
     def cb() -> None:
         acc.run("ip r del default via 192.168.0.1")
 

ipu.py

@@ -93,9 +93,9 @@ def _wait_for_acc_with_retry(self, acc: host.Host) -> None:
         acc.ssh_connect("root", "redhat")
         logger.info(acc.run("uname -a"))
         logger.info("Connected to ACC")
-        if acc.exists("/cda-install"):
-            logger.error_and_exit("Found /cda-install file from a previous installation (install failed)")
-        acc.write("/cda-install", f"{time.time()}")
+        if acc.exists("/var/cda-install"):
+            logger.error_and_exit("Found /var/cda-install file from a previous installation (install failed)")
+        acc.write("/var/cda-install", f"{time.time()}")
 
     def start(self, iso_or_image_path: str) -> bool:
         assert self.config.bmc is not None
@@ -183,7 +183,6 @@ def helper(cmd: str) -> host.Result:
         # after an IMC reboot (which occurs during the RHEL installation)
         assert self.config.dpu_host is not None
         logger.info(f"Reloading idpf on host side {self.config.dpu_host}")
-
         ipu_host = host.RemoteHost(self.config.dpu_host)
         ipu_host.ssh_connect("core")
         ipu_host.run("sudo rmmod idpf")

iso-surgeon/README.md

@@ -39,6 +39,7 @@ You can optionally provide:
 * `--input_iso`: use an existing RHEL boot ISO
 * `--kickstart`: custom kickstart file
 * `--kernel_args`: GRUB arguments to inject
+* `--grub_replace`: GRUB replacements in format 'old_text:new_text' (can be used multiple times, splits on "|")
 * `--output_iso`: destination path for the new ISO
 * `--rhel_version`: the version of RHEL to generate if no input iso is given