From 34209269d66e17d373c065bfd3f0f90e2a655318 Mon Sep 17 00:00:00 2001
From: sujkang <sujkang@microsoft.com>
Date: Fri, 19 Feb 2021 13:57:50 -0800
Subject: [PATCH 1/4] add soft-reboot

---
 scripts/soft-reboot | 229 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 229 insertions(+)
 create mode 100755 scripts/soft-reboot

diff --git a/scripts/soft-reboot b/scripts/soft-reboot
new file mode 100755
index 0000000000..13526b7ede
--- /dev/null
+++ b/scripts/soft-reboot
@@ -0,0 +1,229 @@
+#!/bin/bash
+DEVPATH="/usr/share/sonic/device"
+REBOOT_CAUSE_FILE="/host/reboot-cause/reboot-cause.txt"
+REBOOT_TIME=$(date)
+REBOOT_METHOD="/sbin/kexec -e"
+LOG_SSD_HEALTH="/usr/local/bin/log_ssd_health"
+
+
+EXIT_SUCCESS=0
+EXIT_FAILURE=1
+EXIT_NOT_SUPPORTED=2
+EXIT_FILE_SYSTEM_FULL=3
+EXIT_NEXT_IMAGE_NOT_EXISTS=4
+
+# Reboot immediately if we run the kdump capture kernel
+VMCORE_FILE=/proc/vmcore
+if [ -e $VMCORE_FILE -a -s $VMCORE_FILE ]; then
+        echo "We have a /proc/vmcore, then we just kdump'ed"
+        echo "User issued 'kdump' command [User: kdump, Time: ${REBOOT_TIME}]" > ${REBOOT_CAUSE_FILE}
+        sync
+        PLATFORM=$(grep -oP 'sonic_platform=\K\S+' /proc/cmdline)
+        if [ ! -z "${PLATFORM}" -a -x ${DEVPATH}/${PLATFORM}/${PLAT_REBOOT} ]; then
+            exec ${DEVPATH}/${PLATFORM}/${PLAT_REBOOT}
+        fi
+        # If no platform-specific reboot tool, just run /sbin/reboot
+        /sbin/reboot
+        echo 1 > /proc/sys/kernel/sysrq
+        echo b > /proc/sysrq-trigger
+fi
+
+REBOOT_USER=$(logname)
+PLATFORM=$(sonic-cfggen -H -v DEVICE_METADATA.localhost.platform)
+ASIC_TYPE=$(sonic-cfggen -y /etc/sonic/sonic_version.yml -v asic_type)
+VERBOSE=no
+EXIT_NEXT_IMAGE_NOT_EXISTS=4
+EXIT_SONIC_INSTALLER_VERIFY_REBOOT=21
+SSD_FW_UPDATE="ssd-fw-upgrade"
+REBOOT_SCRIPT_NAME=$(basename $0)
+REBOOT_TYPE="${REBOOT_SCRIPT_NAME}"
+PLATFORM_PLUGIN="${REBOOT_TYPE}_plugin"
+TAG_LATEST=yes
+
+function debug()
+{
+    if [[ x"${VERBOSE}" == x"yes" ]]; then
+        echo `date` $@
+    fi
+    logger "$@"
+}
+
+function tag_images()
+{
+    if test -f /usr/local/bin/ctrmgr_tools.py
+    then
+        if [[ x"${TAG_LATEST}" == x"yes" ]]; then
+            /usr/local/bin/ctrmgr_tools.py tag-all
+        fi
+    fi
+}
+
+function stop_sonic_services()
+{
+    if [[ x"$ASIC_TYPE" != x"mellanox" ]]; then
+        debug "Stopping syncd process..."
+        docker exec -i syncd /usr/bin/syncd_request_shutdown --cold > /dev/null
+        sleep 3
+    fi
+}
+
+function clear_warm_boot()
+{
+    # If reboot is requested, make sure the outstanding warm-boot is cleared
+    # So the system will come up from a cold boot.
+    WARM_DIR="/host/warmboot"
+    REDIS_FILE=dump.rdb
+    TIMESTAMP=`date +%Y%m%d-%H%M%S`
+    if [[ -f ${WARM_DIR}/${REDIS_FILE} ]]; then
+        mv -f ${WARM_DIR}/${REDIS_FILE} ${WARM_DIR}/${REDIS_FILE}.${TIMESTAMP} || /bin/true
+    fi
+    /sbin/kexec -u || /bin/true
+}
+
+SCRIPT=$0
+
+function show_help_and_exit()
+{
+    echo "Usage ${SCRIPT} [options]"
+    echo "    Request rebooting the device. Invoke platform-specific tool when available."
+    echo "    This script will shutdown syncd before rebooting."
+    echo " "
+    echo "    Available options:"
+    echo "        -h, -? : getting this help"
+
+    exit "${EXIT_SUCCESS}"
+}
+
+function setup_reboot_variables()
+{
+    # Kernel and initrd image
+    NEXT_SONIC_IMAGE=$(sonic-installer list | grep "Next: " | cut -d ' ' -f 2)
+    IMAGE_PATH="/host/image-${NEXT_SONIC_IMAGE#SONiC-OS-}"
+    if grep -q aboot_platform= /host/machine.conf; then
+        if is_secureboot; then
+            KERNEL_IMAGE=""
+            BOOT_OPTIONS="SONIC_BOOT_TYPE=${BOOT_TYPE_ARG} secure_boot_enable=1"
+        else
+            KERNEL_IMAGE="$(ls $IMAGE_PATH/boot/vmlinuz-*)"
+            BOOT_OPTIONS="$(cat "$IMAGE_PATH/kernel-cmdline" | tr '\n' ' ') SONIC_BOOT_TYPE=${BOOT_TYPE_ARG}"
+        fi
+    elif grep -q onie_platform= /host/machine.conf; then
+        KERNEL_OPTIONS=$(cat /host/grub/grub.cfg | sed "/$NEXT_SONIC_IMAGE'/,/}/"'!'"g" | grep linux)
+        KERNEL_IMAGE="/host$(echo $KERNEL_OPTIONS | cut -d ' ' -f 2)"
+        BOOT_OPTIONS="$(echo $KERNEL_OPTIONS | sed -e 's/\s*linux\s*/BOOT_IMAGE=/') SONIC_BOOT_TYPE=${BOOT_TYPE_ARG}"
+    else
+        error "Unknown bootloader. ${REBOOT_TYPE} is not supported."
+        exit "${EXIT_NOT_SUPPORTED}"
+    fi
+    INITRD=$(echo $KERNEL_IMAGE | sed 's/vmlinuz/initrd.img/g')
+}
+
+function load_aboot_secureboot_kernel() {
+    local next_image="$IMAGE_PATH/sonic.swi"
+    echo "Loading next image from $next_image"
+    unzip -qp "$next_image" boot0 | \
+        swipath=$next_image kexec=true loadonly=true ENV_EXTRA_CMDLINE="$BOOT_OPTIONS" bash -
+}
+
+function load_kernel() {
+    # Load kernel into the memory
+    /sbin/kexec -l "$KERNEL_IMAGE" --initrd="$INITRD" --append="$BOOT_OPTIONS"
+}
+
+
+function reboot_pre_check()
+{
+    # Make sure that the file system is normal: read-write able
+    filename="/host/test-`date +%Y%m%d-%H%M%S`"
+    ERR=0
+    touch ${filename} || ERR=$?
+    if [[ ${ERR} -ne 0 ]]; then
+        # Continue rebooting in this case, but log the error
+        VERBOSE=yes debug "Filesystem might be read-only or full ..."
+    fi
+    rm ${filename}
+
+    # Verify the next image by sonic-installer
+    local message=$(sonic-installer verify-next-image 2>&1)
+    if [ $? -ne 0 ]; then
+        VERBOSE=yes debug "Failed to verify next image: ${message}"
+        exit ${EXIT_SONIC_INSTALLER_VERIFY_REBOOT}
+    fi
+}
+
+function parse_options()
+{
+    while getopts "h?v" opt; do
+        case ${opt} in
+            h|\? )
+                show_help_and_exit
+                ;;
+            v )
+                VERBOSE=yes
+                ;;
+            t )
+                TAG_LATEST=no
+                ;;
+        esac
+    done
+}
+
+parse_options $@
+
+# Exit if not superuser
+if [[ "$EUID" -ne 0 ]]; then
+    echo "This command must be run as root" >&2
+    exit "${EXIT_FAILURE}"
+fi
+
+debug "User requested rebooting device ..."
+
+setup_reboot_variables
+reboot_pre_check
+
+# Tag remotely deployed images as local
+tag_images
+
+# Stop SONiC services gracefully.
+stop_sonic_services
+
+clear_warm_boot
+
+if is_secureboot && grep -q aboot_machine= /host/machine.conf; then
+    load_aboot_secureboot_kernel
+else
+    load_kernel
+fi
+
+if [ -x ${LOG_SSD_HEALTH} ]; then
+    debug "Collecting logs to check ssd health before ${REBOOT_TYPE}..."
+    ${LOG_SSD_HEALTH}
+fi
+
+# Update the reboot cause file to reflect that user issued 'reboot' command
+# Upon next boot, the contents of this file will be used to determine the
+# cause of the previous reboot
+echo "User issued '${REBOOT_SCRIPT_NAME}' command [User: ${REBOOT_USER}, Time: ${REBOOT_TIME}]" > ${REBOOT_CAUSE_FILE}
+sync
+/sbin/fstrim -av
+sleep 3
+
+# sync the current system time to CMOS
+if [ -x /sbin/hwclock ]; then
+    /sbin/hwclock -w || /bin/true
+fi
+
+# Run platform specific reboot plugin
+if [ -x ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN} ]; then
+    debug "Running ${PLATFORM} specific plugin..."
+    ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN}
+fi
+
+# Reboot: explicitly call Linux native reboot under sbin
+debug "Rebooting with ${REBOOT_METHOD} to ${NEXT_SONIC_IMAGE} ..."
+exec ${REBOOT_METHOD}
+
+# Should never reach here
+error "${REBOOT_TYPE} failed!"
+exit "${EXIT_FAILURE}"
+

From 8007a333d122892601026a159de88e637dbdac91 Mon Sep 17 00:00:00 2001
From: sujkang <sujkang@microsoft.com>
Date: Fri, 19 Feb 2021 18:39:57 -0800
Subject: [PATCH 2/4] soft-reboot

---
 scripts/soft-reboot | 40 ++++++++++++++--------------------------
 1 file changed, 14 insertions(+), 26 deletions(-)

diff --git a/scripts/soft-reboot b/scripts/soft-reboot
index 13526b7ede..db33374a8a 100755
--- a/scripts/soft-reboot
+++ b/scripts/soft-reboot
@@ -38,6 +38,7 @@ SSD_FW_UPDATE="ssd-fw-upgrade"
 REBOOT_SCRIPT_NAME=$(basename $0)
 REBOOT_TYPE="${REBOOT_SCRIPT_NAME}"
 PLATFORM_PLUGIN="${REBOOT_TYPE}_plugin"
+BOOT_TYPE_ARG="soft"
 TAG_LATEST=yes
 
 function debug()
@@ -100,13 +101,8 @@ function setup_reboot_variables()
     NEXT_SONIC_IMAGE=$(sonic-installer list | grep "Next: " | cut -d ' ' -f 2)
     IMAGE_PATH="/host/image-${NEXT_SONIC_IMAGE#SONiC-OS-}"
     if grep -q aboot_platform= /host/machine.conf; then
-        if is_secureboot; then
-            KERNEL_IMAGE=""
-            BOOT_OPTIONS="SONIC_BOOT_TYPE=${BOOT_TYPE_ARG} secure_boot_enable=1"
-        else
-            KERNEL_IMAGE="$(ls $IMAGE_PATH/boot/vmlinuz-*)"
-            BOOT_OPTIONS="$(cat "$IMAGE_PATH/kernel-cmdline" | tr '\n' ' ') SONIC_BOOT_TYPE=${BOOT_TYPE_ARG}"
-        fi
+        KERNEL_IMAGE="$(ls $IMAGE_PATH/boot/vmlinuz-*)"
+        BOOT_OPTIONS="$(cat "$IMAGE_PATH/kernel-cmdline" | tr '\n' ' ') SONIC_BOOT_TYPE=${BOOT_TYPE_ARG}"
     elif grep -q onie_platform= /host/machine.conf; then
         KERNEL_OPTIONS=$(cat /host/grub/grub.cfg | sed "/$NEXT_SONIC_IMAGE'/,/}/"'!'"g" | grep linux)
         KERNEL_IMAGE="/host$(echo $KERNEL_OPTIONS | cut -d ' ' -f 2)"
@@ -118,19 +114,11 @@ function setup_reboot_variables()
     INITRD=$(echo $KERNEL_IMAGE | sed 's/vmlinuz/initrd.img/g')
 }
 
-function load_aboot_secureboot_kernel() {
-    local next_image="$IMAGE_PATH/sonic.swi"
-    echo "Loading next image from $next_image"
-    unzip -qp "$next_image" boot0 | \
-        swipath=$next_image kexec=true loadonly=true ENV_EXTRA_CMDLINE="$BOOT_OPTIONS" bash -
-}
-
 function load_kernel() {
     # Load kernel into the memory
     /sbin/kexec -l "$KERNEL_IMAGE" --initrd="$INITRD" --append="$BOOT_OPTIONS"
 }
 
-
 function reboot_pre_check()
 {
     # Make sure that the file system is normal: read-write able
@@ -176,6 +164,11 @@ if [[ "$EUID" -ne 0 ]]; then
     exit "${EXIT_FAILURE}"
 fi
 
+if [ -x ${LOG_SSD_HEALTH} ]; then
+    debug "Collecting logs to check ssd health before ${REBOOT_TYPE}..."
+    ${LOG_SSD_HEALTH}
+fi
+
 debug "User requested rebooting device ..."
 
 setup_reboot_variables
@@ -189,24 +182,18 @@ stop_sonic_services
 
 clear_warm_boot
 
-if is_secureboot && grep -q aboot_machine= /host/machine.conf; then
-    load_aboot_secureboot_kernel
-else
-    load_kernel
-fi
+echo "$KERNEL_IMAGE : $INITRD : $BOOT_OPTIONS"
 
-if [ -x ${LOG_SSD_HEALTH} ]; then
-    debug "Collecting logs to check ssd health before ${REBOOT_TYPE}..."
-    ${LOG_SSD_HEALTH}
-fi
+load_kernel
 
 # Update the reboot cause file to reflect that user issued 'reboot' command
 # Upon next boot, the contents of this file will be used to determine the
 # cause of the previous reboot
 echo "User issued '${REBOOT_SCRIPT_NAME}' command [User: ${REBOOT_USER}, Time: ${REBOOT_TIME}]" > ${REBOOT_CAUSE_FILE}
+
 sync
-/sbin/fstrim -av
 sleep 3
+sync
 
 # sync the current system time to CMOS
 if [ -x /sbin/hwclock ]; then
@@ -219,8 +206,9 @@ if [ -x ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN} ]; then
     ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN}
 fi
 
-# Reboot: explicitly call Linux native reboot under sbin
+# Reboot: explicitly call Linux "kexec -u"
 debug "Rebooting with ${REBOOT_METHOD} to ${NEXT_SONIC_IMAGE} ..."
+echo `cat ${REBOOT_CAUSE_FILE}`
 exec ${REBOOT_METHOD}
 
 # Should never reach here

From 3947eb15ae938c03391f68409a3877481964f970 Mon Sep 17 00:00:00 2001
From: sujkang <sujkang@microsoft.com>
Date: Fri, 19 Feb 2021 18:45:31 -0800
Subject: [PATCH 3/4] message correction

---
 scripts/soft-reboot | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/scripts/soft-reboot b/scripts/soft-reboot
index db33374a8a..8383c60a86 100755
--- a/scripts/soft-reboot
+++ b/scripts/soft-reboot
@@ -206,7 +206,7 @@ if [ -x ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN} ]; then
     ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN}
 fi
 
-# Reboot: explicitly call Linux "kexec -u"
+# Reboot: explicitly call Linux "kexec -e"
 debug "Rebooting with ${REBOOT_METHOD} to ${NEXT_SONIC_IMAGE} ..."
 echo `cat ${REBOOT_CAUSE_FILE}`
 exec ${REBOOT_METHOD}

From 270070130872052f239f32148e2584c292dbbd3e Mon Sep 17 00:00:00 2001
From: sujkang <sujkang@microsoft.com>
Date: Wed, 24 Feb 2021 16:47:03 -0800
Subject: [PATCH 4/4] review comments

---
 scripts/soft-reboot | 26 ++++++++++++++++++--------
 1 file changed, 18 insertions(+), 8 deletions(-)

diff --git a/scripts/soft-reboot b/scripts/soft-reboot
index 8383c60a86..52ccdd690b 100755
--- a/scripts/soft-reboot
+++ b/scripts/soft-reboot
@@ -4,7 +4,7 @@ REBOOT_CAUSE_FILE="/host/reboot-cause/reboot-cause.txt"
 REBOOT_TIME=$(date)
 REBOOT_METHOD="/sbin/kexec -e"
 LOG_SSD_HEALTH="/usr/local/bin/log_ssd_health"
-
+WATCHDOG_UTIL="/usr/local/bin/watchdogutil"
 
 EXIT_SUCCESS=0
 EXIT_FAILURE=1
@@ -68,10 +68,12 @@ function stop_sonic_services()
     fi
 }
 
-function clear_warm_boot()
+function clear_lingering_reboot_config()
 {
-    # If reboot is requested, make sure the outstanding warm-boot is cleared
-    # So the system will come up from a cold boot.
+    # Clear any outstanding warm-reboot config
+    result=`timeout 10s config warm_restart disable; if [[ $? == 124 ]]; then echo timeout; else echo "code ($?)"; fi` || /bin/true
+    debug "Cancel warm-reboot: ${result}"
+
     WARM_DIR="/host/warmboot"
     REDIS_FILE=dump.rdb
     TIMESTAMP=`date +%Y%m%d-%H%M%S`
@@ -180,9 +182,7 @@ tag_images
 # Stop SONiC services gracefully.
 stop_sonic_services
 
-clear_warm_boot
-
-echo "$KERNEL_IMAGE : $INITRD : $BOOT_OPTIONS"
+clear_lingering_reboot_config
 
 load_kernel
 
@@ -200,6 +200,17 @@ if [ -x /sbin/hwclock ]; then
     /sbin/hwclock -w || /bin/true
 fi
 
+if [ -x ${DEVPATH}/${PLATFORM}/${SSD_FW_UPDATE} ]; then
+    debug "updating ssd fw for${REBOOT_TYPE}"
+    ${DEVPATH}/${PLATFORM}/${SSD_FW_UPDATE} ${REBOOT_TYPE}
+fi
+
+# Enable Watchdog Timer
+if [ -x ${WATCHDOG_UTIL} ]; then
+    debug "Enabling Watchdog before ${REBOOT_TYPE}"
+    ${WATCHDOG_UTIL} arm
+fi
+
 # Run platform specific reboot plugin
 if [ -x ${DEVPATH}/${PLATFORM}/${PLATFORM_PLUGIN} ]; then
     debug "Running ${PLATFORM} specific plugin..."
@@ -208,7 +219,6 @@ fi
 
 # Reboot: explicitly call Linux "kexec -e"
 debug "Rebooting with ${REBOOT_METHOD} to ${NEXT_SONIC_IMAGE} ..."
-echo `cat ${REBOOT_CAUSE_FILE}`
 exec ${REBOOT_METHOD}
 
 # Should never reach here