[201811] Check platform reboot cause to see if any reset happened during fast/warm-reboot

files/build_templates/docker_image_ctl.j2

@@ -36,6 +36,23 @@ function preStartAction()
         echo -n > /tmp/dump.rdb
         docker cp /tmp/dump.rdb database:/var/lib/redis/
     fi
+{%- elif docker_container_name == "swss" %}
+    if [[ "$BOOT_TYPE" == "fast" ]] && [[ -d /host/fast-reboot ]]; then
+        if [[ -f /host/reboot-cause/previous-reboot-cause.json ]]; then
+            REG_BOOT_TYPE="fast*"
+	    CAUSE_NO_AVAIL="\"N/A\""
+            REBOOT_CAUSE="$(cat /host/reboot-cause/previous-reboot-cause.json | jq '.cause')"
+            EXTRA_CAUSE="$(cat /host/reboot-cause/previous-reboot-cause.json | jq '.comment')"
+
+            # Clear the FAST_REBOOT|system db setting if EXTRA_REBOOT_CAUSE is not "N/A" before starting swss
+            if [[ $REBOOT_CAUSE =~ $REG_BOOT_TYPE ]]; then
+                if [[ "${EXTRA_CAUSE}" != "${CAUSE_NO_AVAIL}" ]]; then
+                    # Delete the FAST_REBOOT|system db setting
+                    $SONIC_DB_CLI STATE_DB DEL "FAST_REBOOT|system" &>/dev/null
+                 fi
+             fi
+         fi
+    fi
 {%- else %}
     : # nothing
 {%- endif %}

files/image_config/process-reboot-cause/process-reboot-cause

@@ -138,6 +138,35 @@ def find_hardware_reboot_cause():
     return hardware_reboot_cause
 
 
+def get_reboot_cause_dict(previous_reboot_cause, comment, gen_time):
+    """Store the key infomation of device reboot into a dictionary by parsing the string in
+    previous_reboot_cause.
+    If user issused a command to reboot device, then user, command and time will be
+    stored into a dictionary.
+    If device was rebooted due to the kernel panic, then the string `Kernel Panic`
+    and time will be stored into a dictionary.
+    """
+    reboot_cause_dict = {}
+    reboot_cause_dict['gen_time'] = gen_time
+    reboot_cause_dict['cause'] = previous_reboot_cause
+    reboot_cause_dict['user'] = "N/A"
+    reboot_cause_dict['time'] = "N/A"
+    reboot_cause_dict['comment'] = comment if comment is not None else "N/A"
+    if re.search(r'User issued', previous_reboot_cause):
+        # Match with "User issued '{}' command [User: {}, Time: {}]"
+        match = re.search(r'User issued \'(.*)\' command \[User: (.*), Time: (.*)\]', previous_reboot_cause)
+        if match is not None:
+            reboot_cause_dict['cause'] = match.group(1)
+            reboot_cause_dict['user'] = match.group(2)
+            reboot_cause_dict['time'] = match.group(3)
+    elif re.search(r'Kernel Panic', previous_reboot_cause):
+        match = re.search(r'Kernel Panic \[Time: (.*)\]', previous_reboot_cause)
+        if match is not None:
+            reboot_cause_dict['cause'] = "Kernel Panic"
+            reboot_cause_dict['time'] = match.group(1)
+
+    return reboot_cause_dict
+
 def main():
     log_info("Starting up...")
 
@@ -167,6 +196,8 @@ def main():
     #    reboot info. We will use it as the previous cause.
     software_reboot_cause = find_software_reboot_cause()
 
+    additional_reboot_info = None
+
     # The main decision logic of the reboot cause:
     # If there is a reboot cause indicated by /proc/cmdline, it should be warmreboot/fastreboot
     #   the software_reboot_cause which is the content of /hosts/reboot-cause/reboot-cause.txt 
@@ -176,11 +207,34 @@ def main():
     # Else the software_reboot_cause will be treated as the reboot cause
     if proc_cmdline_reboot_cause is not None:
         previous_reboot_cause = software_reboot_cause
+        if not hardware_reboot_cause.startswith(REBOOT_CAUSE_NON_HARDWARE):
+            # Add the hardware_reboot_cause into additional_reboot_info
+            additional_reboot_info = hardware_reboot_cause
     elif hardware_reboot_cause is not None:
         previous_reboot_cause = hardware_reboot_cause
+        # Check if any software reboot was issued before this hardware reboot happened
+        if software_reboot_cause is not REBOOT_CAUSE_UNKNOWN:
+            additional_reboot_info = software_reboot_cause
     else:
         previous_reboot_cause = software_reboot_cause
 
+    # Current time
+    reboot_cause_gen_time = str(datetime.datetime.now().strftime('%Y_%m_%d_%H_%M_%S'))
+
+    # Save the previous cause info into its history file as json format
+    reboot_cause_dict = get_reboot_cause_dict(previous_reboot_cause, additional_reboot_info, reboot_cause_gen_time)
+
+    # Create reboot-cause-#time#.json under history directory
+    REBOOT_CAUSE_HISTORY_FILE_JSON = os.path.join(REBOOT_CAUSE_HISTORY_DIR, "reboot-cause-{}.json".format(reboot_cause_gen_time))
+
+    # Create REBOOT_CAUSE_HISTORY_DIR if it doesn't exist
+    if not os.path.exists(REBOOT_CAUSE_HISTORY_DIR):
+        os.makedirs(REBOOT_CAUSE_HISTORY_DIR)
+
+    # Write the previous reboot cause to REBOOT_CAUSE_HISTORY_FILE_JSON as a JSON format
+    with open(REBOOT_CAUSE_HISTORY_FILE_JSON, "w") as reboot_cause_history_file:
+        json.dump(reboot_cause_dict, reboot_cause_history_file)
+
     # Write the previous reboot cause to PREVIOUS_REBOOT_CAUSE_FILE
     with open(PREVIOUS_REBOOT_CAUSE_FILE, "w") as prev_cause_file:
         prev_cause_file.write(previous_reboot_cause)