[cisco|express-boot]: Add support for cisco express boot in sonic-utilities (#3056)

[cisco|express-boot]: Add support for cisco express boot in sonic-utilities

Author: jhli-cisco

scripts/express-reboot

@@ -0,0 +1 @@
+fast-reboot

scripts/fast-reboot

@@ -171,7 +171,7 @@ function init_warm_reboot_states()
     # If the current running instance was booted up with warm reboot. Then
     # the current DB contents will likely mark warm reboot is done.
     # Clear these states so that the next boot up image won't get confused.
-    if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
+    if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
         sonic-db-cli STATE_DB eval "
             for _, key in ipairs(redis.call('keys', 'WARM_RESTART_TABLE|*')) do
                 redis.call('hdel', key, 'state')
@@ -197,8 +197,13 @@ function request_pre_shutdown()
         debug "Requesting platform reboot pre-check ..."
     	${DEVPATH}/${PLATFORM}/${PLATFORM_REBOOT_PRE_CHECK} ${REBOOT_TYPE}
     fi
-    debug "Requesting pre-shutdown ..."
-    STATE=$(timeout 5s docker exec syncd /usr/bin/syncd_request_shutdown --pre &> /dev/null; if [[ $? == 124 ]]; then echo "timed out"; fi)
+    if [[ "$REBOOT_TYPE" = "express-reboot" ]]; then
+        debug "Requesting express boot pre-shutdown ..."
+        STATE=$(timeout 5s docker exec syncd /usr/bin/syncd_request_shutdown --pxe &> /dev/null; if [[ $? == 124 ]]; then echo "timed out"; fi)
+    else
+        debug "Requesting pre-shutdown ..."
+        STATE=$(timeout 5s docker exec syncd /usr/bin/syncd_request_shutdown --pre &> /dev/null; if [[ $? == 124 ]]; then echo "timed out"; fi)
+    fi
     if [[ x"${STATE}" == x"timed out" ]]; then
         error "Failed to request pre-shutdown"
     fi
@@ -245,7 +250,7 @@ function backup_database()
 {
     debug "Backing up database ..."
 
-    if [[ "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
+    if [[ "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
         # Advanced reboot: dump state to host disk
         sonic-db-cli ASIC_DB FLUSHDB > /dev/null
         sonic-db-cli COUNTERS_DB FLUSHDB > /dev/null
@@ -475,7 +480,7 @@ function check_docker_exec()
 
 function check_db_integrity()
 {
-    if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
+    if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
         CHECK_DB_INTEGRITY=0
         /usr/local/bin/check_db_integrity.py || CHECK_DB_INTEGRITY=$?
         if [[ CHECK_DB_INTEGRITY -ne 0 ]]; then
@@ -520,7 +525,7 @@ function reboot_pre_check()
     # Make sure ASIC configuration has not changed between images
     ASIC_CONFIG_CHECK_SCRIPT="/usr/local/bin/asic_config_check"
     ASIC_CONFIG_CHECK_SUCCESS=0
-    if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" ]]; then
+    if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" ]]; then
         ASIC_CONFIG_CHECK_EXIT_CODE=0
         ${ASIC_CONFIG_CHECK_SCRIPT} || ASIC_CONFIG_CHECK_EXIT_CODE=$?
 
@@ -604,6 +609,9 @@ function check_conflict_boot_in_fw_update() {
         "warm-reboot")
             FW_AU_TASK_FILE_EXP="${FIRMWARE_AU_STATUS_DIR}/warm_fw_au_task"
             ;;
+        "express-reboot")
+            FW_AU_TASK_FILE_EXP="${FIRMWARE_AU_STATUS_DIR}/express_fw_au_task"
+            ;;
     esac
     FW_AU_TASK_FILE=$(compgen -G ${FW_AU_TASK_FILE_REGEX}) || true
     if [[ -n "${FW_AU_TASK_FILE}" ]] && [[ ! -f "${FW_AU_TASK_FILE_EXP}" ]]; then
@@ -631,12 +639,15 @@ if [[ x"${DETACH}" == x"yes" && x"${ALREADY_DETACHED}" == x"" ]]; then
    exit $?
 fi
 
+sonic_asic_type=$(sonic-cfggen -y /etc/sonic/sonic_version.yml -v asic_type)
+if [[ "$REBOOT_TYPE" = "express-reboot" ]] && [[ "$sonic_asic_type" != "cisco-8000" ]]; then
+    echo "eXpress Boot is not supported"
+    exit "${EXIT_FAILURE}"
+fi
 
 check_conflict_boot_in_fw_update
 
 
-sonic_asic_type=$(sonic-cfggen -y /etc/sonic/sonic_version.yml -v asic_type)
-
 # Check reboot type supported
 BOOT_TYPE_ARG="cold"
 case "$REBOOT_TYPE" in
@@ -662,6 +673,12 @@ case "$REBOOT_TYPE" in
         trap clear_boot EXIT HUP INT QUIT TERM KILL ABRT ALRM
         config warm_restart enable system
         ;;
+    "express-reboot")
+        check_warm_restart_in_progress
+        BOOT_TYPE_ARG="express"
+        trap clear_boot EXIT HUP INT QUIT TERM KILL ABRT ALRM
+        config warm_restart enable system
+        ;;
     *)
         error "Not supported reboot type: $REBOOT_TYPE"
         exit "${EXIT_NOT_SUPPORTED}"
@@ -710,7 +727,6 @@ if [[ "$sonic_asic_type" == "mellanox" ]]; then
     fi
 fi
 
-
 if is_secureboot && grep -q aboot_machine= /host/machine.conf; then
     load_aboot_secureboot_kernel
 else
@@ -731,7 +747,7 @@ init_warm_reboot_states
 setup_control_plane_assistant
 
 TEAMD_INCREASE_RETRY_COUNT=0
-if [[ "${REBOOT_TYPE}" = "warm-reboot" || "${REBOOT_TYPE}" = "fastfast-reboot" ]]; then
+if [[ "${REBOOT_TYPE}" = "warm-reboot" || "${REBOOT_TYPE}" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" ]]; then
     TEAMD_RETRY_COUNT_PROBE_RC=0
     /usr/local/bin/teamd_increase_retry_count.py --probe-only || TEAMD_RETRY_COUNT_PROBE_RC=$?
     if [[ ${TEAMD_RETRY_COUNT_PROBE_RC} -ne 0 ]]; then
@@ -746,7 +762,7 @@ if [[ "${REBOOT_TYPE}" = "warm-reboot" || "${REBOOT_TYPE}" = "fastfast-reboot" ]
     fi
 fi
 
-if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
+if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" || "$REBOOT_TYPE" = "fast-reboot" ]]; then
     # Freeze orchagent for warm restart
     # Ask orchagent_restart_check to try freeze 5 times with interval of 2 seconds,
     # it is possible that the orchagent is in transient state and no opportunity to freeze
@@ -774,7 +790,7 @@ if [[ "$REBOOT_TYPE" = "fast-reboot" ]]; then
     fi
 fi
 
-if [[ ( "${REBOOT_TYPE}" = "warm-reboot" || "${REBOOT_TYPE}" = "fastfast-reboot" ) && "${TEAMD_INCREASE_RETRY_COUNT}" -eq 1 ]]; then
+if [[ ( "${REBOOT_TYPE}" = "warm-reboot" || "${REBOOT_TYPE}" = "fastfast-reboot" || "${REBOOT_TYPE}" = "express-reboot" ) && "${TEAMD_INCREASE_RETRY_COUNT}" -eq 1 ]]; then
     /usr/local/bin/teamd_increase_retry_count.py
 fi
 
@@ -817,6 +833,10 @@ for timer in ${TIMERS}; do
     debug "Stopped ${timer} ..."
 done
 
+if [[ "${REBOOT_TYPE}" == "express-reboot" ]]; then
+    SHUTDOWN_ORDER_FILE="/etc/sonic/warm-reboot_order"
+fi
+
 if [[ -f ${SHUTDOWN_ORDER_FILE} ]]; then
     SERVICES_TO_STOP="$(cat ${SHUTDOWN_ORDER_FILE})"
 else
@@ -853,7 +873,7 @@ for service in ${SERVICES_TO_STOP}; do
     debug "Stopped ${service}"
 
     if [[ "${service}" = "swss" ]]; then
-        if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" ]]; then
+        if [[ "$REBOOT_TYPE" = "warm-reboot" || "$REBOOT_TYPE" = "fastfast-reboot" || "$REBOOT_TYPE" = "express-reboot" ]]; then
             # Pre-shutdown syncd
             initialize_pre_shutdown
 

setup.py

@@ -128,6 +128,7 @@
         'scripts/dump_nat_entries.py',
         'scripts/debug_voq_chassis_packet_drops.sh',
         'scripts/ecnconfig',
+        'scripts/express-reboot',
         'scripts/fabricstat',
         'scripts/fanshow',
         'scripts/fast-reboot',