relax bind mount regex

the source path can be cached from the first container now

Signed-off-by: Aurélien Bombo <abombo@microsoft.com>

src/runtime-rs/Makefile

@@ -180,7 +180,7 @@ DEFNETQUEUES := 1
 DEFENABLEANNOTATIONS := [\"enable_iommu\", \"virtio_fs_extra_args\", \"kernel_params\", \"kernel_verity_params\", \"default_vcpus\", \"default_memory\"]
 DEFENABLEANNOTATIONS_COCO := [\"enable_iommu\", \"virtio_fs_extra_args\", \"kernel_params\", \"kernel_verity_params\", \"default_vcpus\", \"default_memory\", \"cc_init_data\"]
 DEFDISABLEGUESTSECCOMP := true
-DEFDISABLEGUESTEMPTYDIR := false
+DEFDISABLEGUESTEMPTYDIR := true
 ##VAR DEFAULTEXPFEATURES=[features] Default experimental features enabled
 DEFAULTEXPFEATURES := []
 DEFDISABLESELINUX := false

src/runtime/Makefile

@@ -220,7 +220,7 @@ DEFBRIDGES := 1
 DEFENABLEANNOTATIONS := [\"enable_iommu\", \"virtio_fs_extra_args\", \"kernel_params\", \"kernel_verity_params\"]
 DEFENABLEANNOTATIONS_COCO := [\"enable_iommu\", \"virtio_fs_extra_args\", \"kernel_params\", \"kernel_verity_params\", \"default_vcpus\", \"default_memory\", \"cc_init_data\"]
 DEFDISABLEGUESTSECCOMP := true
-DEFDISABLEGUESTEMPTYDIR := false
+DEFDISABLEGUESTEMPTYDIR := true
 #Default experimental features enabled
 DEFAULTEXPFEATURES := []
 

src/runtime/virtcontainers/fs_share_linux.go

@@ -83,9 +83,13 @@ type FilesystemShare struct {
 	configVolRegex *regexp.Regexp
 	// Regex to match only the timestamped directory inside the k8's volume mount
 	timestampDirRegex *regexp.Regexp
-	// The same volume mount can be shared by multiple containers in the same sandbox (pod)
+	// srcDstMap tracks file-level source to destination mappings for configmap/secret watching
-	srcDstMap            map[string][]string
+	srcDstMap     map[string][]string
-	srcDstMapLock        sync.Mutex
+	srcDstMapLock sync.Mutex
+	// srcGuestMap caches volume source path to guest path, enabling multiple containers
+	// in the same pod to share the same volume mount
+	srcGuestMap          map[string]string
+	srcGuestMapLock      sync.Mutex
 	eventLoopStarted     bool
 	eventLoopStartedLock sync.Mutex
 	watcherDoneChannel   chan bool
@@ -108,6 +112,7 @@ func NewFilesystemShare(s *Sandbox) (*FilesystemShare, error) {
 		sandbox:            s,
 		watcherDoneChannel: make(chan bool),
 		srcDstMap:          make(map[string][]string),
+		srcGuestMap:        make(map[string]string),
 		watcher:            watcher,
 		configVolRegex:     configVolRegex,
 		timestampDirRegex:  timestampDirRegex,
@@ -309,6 +314,13 @@ func (f *FilesystemShare) ShareFile(ctx context.Context, c *Container, m *Mount)
 	// bind mount it in the shared directory.
 	caps := f.sandbox.hypervisor.Capabilities(ctx)
 	if !caps.IsFsSharingSupported() {
+		f.srcGuestMapLock.Lock()
+		if guestPath, ok := f.srcGuestMap[m.Source]; ok {
+			f.srcGuestMapLock.Unlock()
+			return &SharedFile{guestPath: guestPath}, nil
+		}
+		f.srcGuestMapLock.Unlock()
+
 		f.Logger().Debug("filesystem sharing is not supported, files will be copied")
 
 		var ignored bool
@@ -418,6 +430,11 @@ func (f *FilesystemShare) ShareFile(ctx context.Context, c *Container, m *Mount)
 		m.HostPath = mountDest
 	}
 
+	// Cache the guestPath for this volume source so other containers can share it
+	f.srcGuestMapLock.Lock()
+	defer f.srcGuestMapLock.Unlock()
+	f.srcGuestMap[m.Source] = guestPath
+
 	return &SharedFile{
 		guestPath: guestPath,
 	}, nil
@@ -442,6 +459,10 @@ func (f *FilesystemShare) UnshareFile(ctx context.Context, c *Container, m *Moun
 		}
 	}
 
+	f.srcGuestMapLock.Lock()
+	delete(f.srcGuestMap, m.Source)
+	f.srcGuestMapLock.Unlock()
+
 	return nil
 }
 

src/tools/genpolicy/genpolicy-settings.json

@@ -152,17 +152,6 @@
         }
     },
     "volumes": {
-        "emptyDir": {
-            "mount_type": "local",
-            "mount_source": "^$(cpath)/$(sandbox-id)/rootfs/local/",
-            "mount_point": "^$(cpath)/$(sandbox-id)/rootfs/local/",
-            "driver": "local",
-            "source": "local",
-            "fstype": "local",
-            "options": [
-                "mode=0777"
-            ]
-        },
         "emptyDir_memory": {
             "mount_type": "bind",
             "mount_source": "^/run/kata-containers/sandbox/ephemeral/",

src/tools/genpolicy/rules.rego

@@ -1160,7 +1160,7 @@ mount_source_allows(p_mount, i_mount, bundle_id, sandbox_id) if {
     regex3 := replace(regex2, "$(cpath)", policy_data.common.cpath)
     print("mount_source_allows 2: regex3 =", regex3)
 
-    regex4 := replace(regex3, "$(sandbox-id)", sandbox_id)
+    regex4 := replace(regex3, "$(bundle-id)", "[a-z0-9]{64}")
     print("mount_source_allows 2: regex4 =", regex4)
     regex.match(regex4, i_mount.source)
 

src/tools/genpolicy/src/mount_and_storage.rs

@@ -105,7 +105,6 @@ pub fn get_mount_and_storage(
     storages: &mut Vec<agent::Storage>,
     yaml_volume: &volume::Volume,
     yaml_mount: &pod::VolumeMount,
-    pod_security_context: &Option<pod::PodSecurityContext>,
 ) {
     debug!(
         "get_mount_and_storage: adding mount and storage for: {:?}",
@@ -113,27 +112,18 @@ pub fn get_mount_and_storage(
     );
 
     if let Some(emptyDir) = &yaml_volume.emptyDir {
-        let settings_volumes = &settings.volumes;
+        let is_tmpfs = emptyDir.medium.as_ref().is_some_and(|m| m == "Memory");
-        let mut volume: Option<&settings::EmptyDirVolume> = None;
+        if is_tmpfs {
-
+            get_memory_empty_dir_mount_and_storage(settings, p_mounts, storages, yaml_mount);
-        if let Some(medium) = &emptyDir.medium {
+        } else {
-            if medium == "Memory" {
+            let access = if yaml_mount.readOnly == Some(true) {
-                volume = Some(&settings_volumes.emptyDir_memory);
+                debug!("setting read only access for emptyDir mount");
-            }
+                "ro"
+            } else {
+                "rw"
+            };
+            get_shared_bind_mount(yaml_mount, p_mounts, "rprivate", access);
         }
-
-        if volume.is_none() {
-            volume = Some(&settings_volumes.emptyDir);
-        }
-
-        get_empty_dir_mount_and_storage(
-            settings,
-            p_mounts,
-            storages,
-            yaml_mount,
-            volume.unwrap(),
-            pod_security_context,
-        );
     } else if yaml_volume.persistentVolumeClaim.is_some() || yaml_volume.azureFile.is_some() {
         get_shared_bind_mount(yaml_mount, p_mounts, "rprivate", "rw");
     } else if yaml_volume.hostPath.is_some() {
@@ -149,50 +139,25 @@ pub fn get_mount_and_storage(
     }
 }
 
-fn get_empty_dir_mount_and_storage(
+fn get_memory_empty_dir_mount_and_storage(
     settings: &settings::Settings,
     p_mounts: &mut Vec<policy::KataMount>,
     storages: &mut Vec<agent::Storage>,
     yaml_mount: &pod::VolumeMount,
-    settings_empty_dir: &settings::EmptyDirVolume,
-    pod_security_context: &Option<pod::PodSecurityContext>,
 ) {
-    debug!("Settings emptyDir: {:?}", settings_empty_dir);
+    let settings_empty_dir = &settings.volumes.emptyDir_memory;
+    debug!("Settings emptyDir_memory: {:?}", settings_empty_dir);
 
-    if yaml_mount.subPathExpr.is_none() {
+    storages.push(agent::Storage {
-        let mut options = settings_empty_dir.options.clone();
+        driver: settings_empty_dir.driver.clone(),
-        if let Some(gid) = pod_security_context.as_ref().and_then(|sc| sc.fsGroup) {
+        driver_options: Vec::new(),
-            // This matches the runtime behavior of only setting the fsgid if the mountpoint GID is not 0.
+        source: settings_empty_dir.source.clone(),
-            // https://github.com/kata-containers/kata-containers/blob/b69da5f3ba8385c5833b31db41a846a203812675/src/runtime/virtcontainers/kata_agent.go#L1602-L1607
+        fstype: settings_empty_dir.fstype.clone(),
-            if gid != 0 {
+        options: settings_empty_dir.options.clone(),
-                options.push(format!("fsgid={gid}"));
+        mount_point: format!("{}{}$", &settings_empty_dir.mount_point, &yaml_mount.name),
-            }
+        fs_group: protobuf::MessageField::none(),
-        }
+        special_fields: ::protobuf::SpecialFields::new(),
-        storages.push(agent::Storage {
+    });
-            driver: settings_empty_dir.driver.clone(),
-            driver_options: Vec::new(),
-            source: settings_empty_dir.source.clone(),
-            fstype: settings_empty_dir.fstype.clone(),
-            options,
-            mount_point: format!("{}{}$", &settings_empty_dir.mount_point, &yaml_mount.name),
-            fs_group: protobuf::MessageField::none(),
-            special_fields: ::protobuf::SpecialFields::new(),
-        });
-    }
-
-    let source = if yaml_mount.subPathExpr.is_some() {
-        let file_name = Path::new(&yaml_mount.mountPath).file_name().unwrap();
-        let name = OsString::from(file_name).into_string().unwrap();
-        format!("{}{name}$", &settings.volumes.configMap.mount_source)
-    } else {
-        format!("{}{}$", &settings_empty_dir.mount_source, &yaml_mount.name)
-    };
-
-    let mount_type = if yaml_mount.subPathExpr.is_some() {
-        "bind"
-    } else {
-        &settings_empty_dir.mount_type
-    };
 
     let access = match yaml_mount.readOnly {
         Some(true) => {
@@ -204,8 +169,8 @@ fn get_empty_dir_mount_and_storage(
 
     p_mounts.push(policy::KataMount {
         destination: yaml_mount.mountPath.to_string(),
-        type_: mount_type.to_string(),
+        type_: settings_empty_dir.mount_type.clone(),
-        source,
+        source: format!("{}{}$", &settings_empty_dir.mount_source, &yaml_mount.name),
         options: vec![
             "rbind".to_string(),
             "rprivate".to_string(),
@@ -318,13 +283,7 @@ fn get_shared_bind_mount(
     propagation: &str,
     access: &str,
 ) {
-    // The Kata Shim filepath.Base() to extract the last element of this path, in
+    let source = "$(sfprefix)[a-zA-Z0-9_.-]+$".to_string();
-    // https://github.com/kata-containers/kata-containers/blob/5e46f814dd79ab6b34588a83825260413839735a/src/runtime/virtcontainers/fs_share_linux.go#L305
-    // In Rust, Path::file_name() has a similar behavior.
-    let path = Path::new(&yaml_mount.mountPath);
-    let mount_path = path.file_name().unwrap().to_str().unwrap();
-
-    let source = format!("$(sfprefix){mount_path}$");
 
     let dest = yaml_mount.mountPath.clone();
     let type_ = "bind".to_string();

src/tools/genpolicy/src/settings.rs

@@ -31,7 +31,6 @@ pub struct Settings {
 /// Volume settings loaded from genpolicy-settings.json.
 #[derive(Clone, Debug, Serialize, Deserialize)]
 pub struct Volumes {
-    pub emptyDir: EmptyDirVolume,
     pub emptyDir_memory: EmptyDirVolume,
     pub configMap: ConfigMapVolume,
     pub image_volume: ImageVolume,

src/tools/genpolicy/src/yaml.rs

@@ -304,7 +304,6 @@ pub fn get_container_mounts_and_storages(
                             storages,
                             volume,
                             volume_mount,
-                            &podSpec.securityContext,
                         );
                     }
                 }

tests/gha-run-k8s-common.sh

@@ -9,6 +9,8 @@ source "${tests_dir}/common.bash"
 kubernetes_dir="${tests_dir}/integration/kubernetes"
 helm_chart_dir="${repo_root_dir}/tools/packaging/kata-deploy/helm-chart/kata-deploy"
 
+AZ_REGION="${AZ_REGION:-eastus}"
+AZ_NODEPOOL_TAGS="${AZ_NODEPOOL_TAGS:-}"
 GENPOLICY_PULL_METHOD="${GENPOLICY_PULL_METHOD:-oci-distribution}"
 GH_PR_NUMBER="${GH_PR_NUMBER:-}"
 HELM_DEFAULT_INSTALLATION="${HELM_DEFAULT_INSTALLATION:-false}"
@@ -111,7 +113,7 @@ function create_cluster() {
 		"GENPOLICY_PULL_METHOD=${GENPOLICY_PULL_METHOD:0:1}")
 
 	az group create \
-		-l eastus \
+		-l "${AZ_REGION}" \
 		-n "${rg}"
 
 	# Required by e.g. AKS App Routing for KBS installation.
@@ -129,7 +131,8 @@ function create_cluster() {
 		--node-count 1 \
 		--generate-ssh-keys \
 		--tags "${tags[@]}" \
-		$([[ "${KATA_HOST_OS}" = "cbl-mariner" ]] && echo "--os-sku AzureLinux --workload-runtime KataVmIsolation")
+		$([[ "${KATA_HOST_OS}" = "cbl-mariner" ]] && echo "--os-sku AzureLinux --workload-runtime KataVmIsolation") \
+		$([ -n "${AZ_NODEPOOL_TAGS}" ] && echo "--nodepool-tags "${AZ_NODEPOOL_TAGS}"")
 }
 
 function install_bats() {

tests/integration/kubernetes/tests_common.sh

@@ -37,6 +37,8 @@ K8S_TEST_DIR="${kubernetes_dir:-"${BATS_TEST_DIRNAME}"}"
 
 AUTO_GENERATE_POLICY="${AUTO_GENERATE_POLICY:-}"
 GENPOLICY_PULL_METHOD="${GENPOLICY_PULL_METHOD:-}"
+GENPOLICY_BINARY="${GENPOLICY_BINARY:-"/opt/kata/bin/genpolicy"}"
+GENPOLICY_SETTINGS_DIR="${GENPOLICY_SETTINGS_DIR:-"/opt/kata/share/defaults/kata-containers"}"
 KATA_HYPERVISOR="${KATA_HYPERVISOR:-}"
 KATA_HOST_OS="${KATA_HOST_OS:-}"
 
@@ -191,12 +193,11 @@ adapt_common_policy_settings() {
 # and change these settings to use Kata CI cluster's default namespace.
 create_common_genpolicy_settings() {
 	declare -r genpolicy_settings_dir="$1"
-	declare -r default_genpolicy_settings_dir="/opt/kata/share/defaults/kata-containers"
 
 	auto_generate_policy_enabled || return 0
 
-	cp "${default_genpolicy_settings_dir}/genpolicy-settings.json" "${genpolicy_settings_dir}"
+	cp "${GENPOLICY_SETTINGS_DIR}/genpolicy-settings.json" "${genpolicy_settings_dir}"
-	cp "${default_genpolicy_settings_dir}/rules.rego" "${genpolicy_settings_dir}"
+	cp "${GENPOLICY_SETTINGS_DIR}/rules.rego" "${genpolicy_settings_dir}"
 
 	adapt_common_policy_settings "${genpolicy_settings_dir}"
 }
@@ -247,7 +248,7 @@ auto_generate_policy_no_added_flags() {
 	declare -r additional_flags="${4:-""}"
 
 	auto_generate_policy_enabled || return 0
-	local genpolicy_command="RUST_LOG=info /opt/kata/bin/genpolicy -u -y ${yaml_file}"
+	local genpolicy_command="RUST_LOG=info ${GENPOLICY_BINARY} -u -y ${yaml_file}"
 	genpolicy_command+=" -p ${settings_dir}/rules.rego"
 	genpolicy_command+=" -j ${settings_dir}/genpolicy-settings.json"