confd: major behavior change, run container --read-only always

Container support in Infix was released with v24.02, so this change may
unfortunately break a few use-cases out there.  Regrettable as this is,
the default behavior, including how containers are started after boot,
break other use-cases that were considered more important.

As of this commit:

 - all containers in Infix run in read-only mode, use volumes and
   mounts for persistence across reboot/stop/start/upgrade
 - all containers are now "recreated" at boot or related config changes,
   this ensures an OCI image embedded in the Infix image, /lib/oci/, is
   always used as the base for a running container

Fixes #823

Signed-off-by: Joachim Wiberg <troglobit@gmail.com>

board/common/rootfs/usr/sbin/container

@@ -108,9 +108,9 @@ create()
 	logging="--log-driver k8s-file --log-opt path=/run/containers/$name.fifo"
     fi
 
-    args="$args --replace --quiet --cgroup-parent=containers $caps"
+    args="$args --read-only --replace --quiet --cgroup-parent=containers $caps"
     args="$args --restart=$restart --systemd=false --tz=local $privileged"
-    args="$args $ro $vol $mount $hostname $entrypoint $env $port $logging"
+    args="$args $vol $mount $hostname $entrypoint $env $port $logging"
     pidfn=/run/container:${name}.pid
 
     [ -n "$quiet" ] || log "---------------------------------------"
@@ -136,8 +136,7 @@ create()
     # shellcheck disable=SC2048
     log "Calling podman create --name $name --conmon-pidfile=$pidfn $args $image $*"
     if podman create --name "$name" --conmon-pidfile="$pidfn" $args "$image" $*; then
-	[ -n "$quiet" ] || log "Successfully created container $name from $image"
-	rm -f "/run/containers/env/${name}.env"
+	[ -n "$quiet"  ] || log "Successfully created container $name from $image"
 	[ -n "$manual" ] || start "$name"
 	exit 0
     fi
@@ -267,7 +266,6 @@ options:
                            Syntax: [[ip:][hostPort]:]containerPort[/protocol]
   -q, --quiet              Quiet operation, called from confd
   -r, --restart POLICY     One of "no", "always", or "on-failure:NUM"
-      --read-only          Do not create a writable layer
   -s, --simple             Show output in simplified format
   -v, --volume NAME:PATH   Create named volume mounted inside container on PATH
 
@@ -386,9 +384,6 @@ while [ "$1" != "" ]; do
 	    shift
 	    restart=$1
 	    ;;
-	--read-only)
-	    ro="--read-only=true"
-	    ;;
 	-s | --simple)
 	    simple=true
 	    ;;
@@ -627,7 +622,7 @@ case $cmd in
 	;;
     upgrade)
 	# Start script used to initially create container
-	script=/var/lib/containers/active/S01-${1}.sh
+	script=/run/containers/active/S01-${1}.sh
 
 	# Find container image
 	img=$(podman inspect "$1" | jq -r .[].ImageName)

package/execd/execd.conf

@@ -1,2 +1,2 @@
 service log:prio:local1.err,tag:container \
-	[2345] execd /run/containers/queue /var/lib/containers/active -- Container job runner
+	[2345] execd /run/containers/queue /run/containers/active -- Container job runner

package/execd/tmpfiles.conf

@@ -1,8 +1,8 @@
 d  /run/containers/args		0700 - -
 d  /run/containers/files	0700 - -
-d  /var/lib/containers/active	0700 - -
 d  /var/lib/containers/oci	0755 - -
 d  /run/containers/inbox	0700 - -
 d  /run/containers/queue	0700 - -
+d  /run/containers/active	0700 - -
 d  /run/cni			0755 - -
 L+ /var/lib/cni                 -    - - - /run/cni

src/confd/src/infix-containers.c

@@ -18,7 +18,7 @@
 #define  CFG_XPATH    "/infix-containers:containers"
 #define  INBOX_QUEUE  "/run/containers/inbox"
 #define  JOB_QUEUE    "/run/containers/queue"
-#define  ACTIVE_QUEUE "/var/lib/containers/active"
+#define  ACTIVE_QUEUE "/run/containers/active"
 #define  LOGGER       "logger -t container -p local1.notice"
 
 
@@ -56,9 +56,6 @@ static int add(const char *name, struct lyd_node *cif)
 			fprintf(fp, " --hostname %s", buf);
 	}
 
-	if (lydx_is_enabled(cif, "read-only"))
-		fprintf(fp, " --read-only");
-
 	if (lydx_is_enabled(cif, "privileged"))
 		fprintf(fp, " --privileged");
 
@@ -392,15 +389,7 @@ static void cleanup(sr_session_ctx_t *session, struct confd *confd)
 /*
  * Containers depend on a lot of other system resources being properly
  * set up, e.g., networking, which is run by dagger.  So we need to wait
- * for all that before we can launch new, or modified, containers.  The
- * latter is the tricky part.
- *
- * By default, containers get a writable layer which is preserved across
- * restarts/reboots of container or host -- provided we don't recreate
- * them on a reboot.  Hence the cmp magic below: we check if the command
- * to create a container is the same as what is already activated, if it
- * is already activated we know 'podman create' has done its thing and
- * we can safely start the container.
+ * for all that before we can launch new, or modified, containers.
  */
 void infix_containers_post_hook(sr_session_ctx_t *session, struct confd *confd)
 {
@@ -416,33 +405,26 @@ void infix_containers_post_hook(sr_session_ctx_t *session, struct confd *confd)
 	}
 
 	while ((d = readdir(dir))) {
-		char curr[strlen(ACTIVE_QUEUE) + strlen(d->d_name) + 2];
 		char next[strlen(INBOX_QUEUE) + strlen(d->d_name) + 2];
+		char name[strlen(d->d_name) + 1];
+		char *ptr;
 
 		if (d->d_name[0] == '.')
 			continue;
 
-		snprintf(curr, sizeof(curr), "%s/%s", ACTIVE_QUEUE, d->d_name);
 		snprintf(next, sizeof(next), "%s/%s", INBOX_QUEUE, d->d_name);
-		if (!systemf("cmp %s %s >/dev/null 2>&1", curr, next)) {
-			char name[strlen(d->d_name) + 1];
-			char *ptr;
 
-			strlcpy(name, d->d_name, sizeof(name));
-			ptr = strstr(name, ".sh");
-			if (ptr) {
-				char *nm = NULL;
+		strlcpy(name, d->d_name, sizeof(name));
+		ptr = strstr(name, ".sh");
+		if (ptr) {
+			char *nm = NULL;
 
-				*ptr = 0;
-				if (!strncmp(name, "S01-", 4))
-					nm = &name[4];
+			*ptr = 0;
+			if (!strncmp(name, "S01-", 4))
+				nm = &name[4];
 
-				/* New job is already active, no changes, skipping ... */
-				if (nm && !is_manual(session, nm))
-					systemf("initctl -bnq cond set container:%s", nm);
-			}
-			remove(next);
-			continue;
+			if (nm && !is_manual(session, nm))
+				systemf("initctl -bnq cond set container:%s", nm);
 		}
 
 		if (movefile(next, JOB_QUEUE))

src/confd/yang/infix-containers.yang

@@ -26,6 +26,11 @@ module infix-containers {
     prefix infix-sys;
   }
 
+  revision 2024-11-12  {
+    description "Deprecate read-only, it is now always true.";
+    reference "internal";
+  }
+
   revision 2024-10-14  {
     description "Two major changes:
                   - Allow changing name of host interfaces inside container
@@ -256,6 +261,7 @@ module infix-containers {
       }
 
       leaf read-only {
+        status deprecated; // This is now the default, setting kept only to not break configs
 	description "Create a read-only container.  Use volumes for writable directories.";
 	type boolean;
       }