From d874c17af14b52e60238d75596d654723e0dd238 Mon Sep 17 00:00:00 2001
From: Bin Qian <bin.qian@windriver.com>
Date: Thu, 9 Jan 2025 14:18:07 +0000
Subject: [PATCH] rsync repo only for same sw-version

software-controller and sw-patch-controller both arbitrarily perform
repo rsync from the active controller. This operation deletes the newly
created to-release repo during a DX major release deploy, as the
to-release is first deployed to the standby controller.

This change applies the same change to both services, to verify that the
active controller is actually running the same sw-version before
performing rsync.

This change works together with [1]

TCs:
    passed: on a DX system, upgrade from 22.12 to 24.09 then patch to
            24.09 with success

Story: 2010676
Task: 51516

[1] https://review.opendev.org/c/starlingx/update/+/938409

Change-Id: I952d1633387512ccdc40dcbc8c062cbe8efedcd6
Signed-off-by: Bin Qian <bin.qian@windriver.com>
---
 .../service-files/software-controller-init.sh | 29 ++++++++++++++++---
 sw-patch/bin/sw-patch-controller-init.sh      | 21 +++++++++++---
 2 files changed, 42 insertions(+), 8 deletions(-)

diff --git a/software/service-files/software-controller-init.sh b/software/service-files/software-controller-init.sh
index 59bd5a8d..7868c337 100644
--- a/software/service-files/software-controller-init.sh
+++ b/software/service-files/software-controller-init.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 #
-# Copyright (c) 2023 Wind River Systems, Inc.
+# Copyright (c) 2023-2025 Wind River Systems, Inc.
 #
 # SPDX-License-Identifier: Apache-2.0
 #
@@ -52,10 +52,17 @@ function do_setup {
         grep -xq "gpg-verify=false" $FEED_OSTREE_CONFIG || sed -i '$a gpg-verify=false' $FEED_OSTREE_CONFIG
         grep -xq "gpg-verify=false" $SYSROOT_OSTREE_CONFIG || sed -i '$a gpg-verify=false' $SYSROOT_OSTREE_CONFIG
 
+        LOG "apt-ostree repo init --feed $REPO_DIR --release $RELEASE --origin $REPO_ID ..."
         apt-ostree repo init \
             --feed $REPO_DIR \
             --release $RELEASE \
             --origin $REPO_ID
+
+        if [ $? -eq 0 ]; then
+            LOG "... done"
+        else
+            LOG "... failed"
+        fi
     fi
 
     if [ ! -d $PATCHING_DIR ]; then
@@ -76,13 +83,27 @@ function do_setup {
         return 0
     fi
 
+    # TODO(bqian) review this rsync below, it could break the atomic data sync mechanism
     # Sync the software dir
     LOG_TO_FILE "rsync -acv --delete rsync://controller/software/ ${PATCHING_DIR}/"
     rsync -acv --delete rsync://controller/software/ ${PATCHING_DIR}/ >> $logfile 2>&1
 
-    # Sync the repo dir
-    LOG_TO_FILE "rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/"
-    rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/ >> $logfile 2>&1
+    # sync the repo from peer controller if both are running the same sw_version
+    tmp_dir=$(mktemp -d)
+    rsync -acv rsync://controller/platform/platform.conf ${tmp_dir}
+    my_tag="^sw_version=${SW_VERSION}$"
+    grep ${my_tag} ${tmp_dir}/platform.conf
+    rc=$?
+    rm ${tmp_dir}/platform.conf
+    rmdir ${tmp_dir}
+
+    if [ ${rc} -eq 0 ]; then
+        # Sync the repo dir
+        LOG_TO_FILE "rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/"
+        rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/ >> $logfile 2>&1
+    else
+        LOG "Skip rsync. Peer is not running the same software version"
+    fi
 }
 
 case "$1" in
diff --git a/sw-patch/bin/sw-patch-controller-init.sh b/sw-patch/bin/sw-patch-controller-init.sh
index 3650833b..1f58ef3f 100644
--- a/sw-patch/bin/sw-patch-controller-init.sh
+++ b/sw-patch/bin/sw-patch-controller-init.sh
@@ -1,6 +1,6 @@
 #!/bin/bash
 #
-# Copyright (c) 2014 Wind River Systems, Inc.
+# Copyright (c) 2014-2025 Wind River Systems, Inc.
 #
 # SPDX-License-Identifier: Apache-2.0
 #
@@ -66,9 +66,22 @@ function do_setup {
     LOG_TO_FILE "rsync -acv --delete rsync://controller/patching/ ${PATCHING_DIR}/"
     rsync -acv --delete rsync://controller/patching/ ${PATCHING_DIR}/ >> $logfile 2>&1
 
-    # Sync the patching dir
-    LOG_TO_FILE "rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/"
-    rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/ >> $logfile 2>&1
+    # sync the repo from peer controller if both are running the same sw_version
+    tmp_dir=$(mktemp -d)
+    rsync -acv rsync://controller/platform/platform.conf ${tmp_dir}
+    my_tag="^sw_version=${SW_VERSION}$"
+    grep ${my_tag} ${tmp_dir}/platform.conf
+    rc=$?
+    rm ${tmp_dir}/platform.conf
+    rmdir ${tmp_dir}
+
+    if [ ${rc} -eq 0 ]; then
+        # Sync the repo dir
+        LOG_TO_FILE "rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/"
+        rsync -acv --delete rsync://controller/repo/ ${REPO_ROOT}/ >> $logfile 2>&1
+    else
+        LOG "Skip rsync. Peer is not running the same software version"
+    fi
 }
 
 case "$1" in