[BMC] Add BMC CLIs and integrate BMC dump into show techsupport

scripts/bmc_techsupport.py

@@ -0,0 +1,140 @@
+#!/usr/bin/env python3
+
+"""
+bmc_techsupport script.
+    This script is invoked by the generate_dump script for BMC techsupport fetching,
+    but also can be invoked manually to trigger and collect BMC debug log dump.
+
+    The usage of this script is divided into two parts:
+        1. Triggering BMC debug log dump Redfish task
+            * In this case the script triggers a POST request to BMC to start collecting debug log dump.
+            * In this script we will print the new task-id to the console output
+              to collect the debug log dump once the task-id has finished.
+            * This step is non-blocking, task-id is returned immediately.
+            * It is invoked with the parameter '--mode trigger'
+              E.g.: /usr/local/bin/bmc_techsupport.py --mode trigger
+
+        2. Collecting BMC debug log dump
+            * In this step we will wait for the task-id to finish if it has not finished.
+            * Blocking action until we get the file or encounter an ERROR or Timeout.
+            * It is invoked with the parameter '--mode collect --task <task-id> --path <path>'
+              E.g.: /usr/local/bin/bmc_techsupport.py --mode collect --path <path> --task <task-id>
+
+    Basically, in the generate_dump script we will call the first method
+    at the beginning of its process and the second method towards the end of the process.
+"""
+
+
+import argparse
+import os
+import sonic_platform
+import time
+from sonic_py_common.syslogger import SysLogger
+
+
+TIMEOUT_FOR_GET_BMC_DEBUG_LOG_DUMP_IN_SECONDS = 60
+SYSLOG_IDENTIFIER = "bmc_techsupport"
+log = SysLogger(SYSLOG_IDENTIFIER)
+
+
+class BMCDebugDumpExtractor:
+    '''
+        Class to trigger and extract BMC debug log dump
+    '''
+
+    INVALID_TASK_ID = '-1'
+    TRIGGER_MODE = 'trigger'
+    COLLECT_MODE = 'collect'
+
+    def __init__(self):
+        platform = sonic_platform.platform.Platform()
+        chassis = platform.get_chassis()
+        self.bmc = chassis.get_bmc()
+
+    def trigger_debug_dump(self):
+        '''
+        Trigger BMC debug log dump and prints the running task id to the console output
+        '''
+        try:
+            task_id = BMCDebugDumpExtractor.INVALID_TASK_ID
+            log.log_info("Triggering BMC debug log dump Redfish task")
+            (ret, (task_id, err_msg)) = self.bmc.trigger_bmc_debug_log_dump()
+            if ret != 0:
+                raise Exception(err_msg)
+            log.log_info(f'Successfully triggered BMC debug log dump - Task-id: {task_id}')
+        except Exception as e:
+            log.log_error(f'Failed to trigger BMC debug log dump - {str(e)}')
+        finally:
+            # generate_dump script captures the task id from the console output via $(...) syntax
+            print(f'{task_id}')
+
+    def extract_debug_dump_file(self, task_id, filepath):
+        '''
+            Extract BMC debug log dump file for the given task id and save it to the given filepath
+        '''
+        try:
+            if task_id is None or task_id == BMCDebugDumpExtractor.INVALID_TASK_ID:
+                raise Exception('Invalid Task-ID')
+            log_dump_dir = os.path.dirname(filepath)
+            log_dump_filename = os.path.basename(filepath)
+            if not log_dump_dir or not log_dump_filename:
+                raise Exception(f'Invalid given filepath: {filepath}')
+            if not log_dump_filename.endswith('.tar.xz'):
+                raise Exception(f'Invalid given filepath extension, should be .tar.xz: {log_dump_filename}')
+
+            start_time = time.time()
+            log.log_info("Collecting BMC debug log dump")
+            ret, err_msg = self.bmc.get_bmc_debug_log_dump(
+                task_id=task_id,
+                filename=log_dump_filename,
+                path=log_dump_dir,
+                timeout=TIMEOUT_FOR_GET_BMC_DEBUG_LOG_DUMP_IN_SECONDS
+            )
+            end_time = time.time()
+            duration = end_time - start_time
+            if ret != 0:
+                timeout_msg = (
+                    f'BMC debug log dump does not finish within '
+                    f'{TIMEOUT_FOR_GET_BMC_DEBUG_LOG_DUMP_IN_SECONDS} seconds: {err_msg}'
+                )
+                log.log_error(timeout_msg)
+                raise Exception(err_msg)
+            log.log_info(f'Finished successfully collecting BMC debug log dump. Duration: {duration} seconds')
+        except Exception as e:
+            log.log_error(f'Failed to collect BMC debug log dump - {str(e)}')
+
+
+def main(mode, task_id, filepath):
+    try:
+        extractor = BMCDebugDumpExtractor()
+        if extractor.bmc is None:
+            raise Exception('BMC instance is not available')
+    except Exception as e:
+        log.log_error(f'Failed to initialize BMCDebugDumpExtractor: {str(e)}')
+        if mode == BMCDebugDumpExtractor.TRIGGER_MODE:
+            print(f'{BMCDebugDumpExtractor.INVALID_TASK_ID}')
+        return
+    if mode == BMCDebugDumpExtractor.TRIGGER_MODE:
+        extractor.trigger_debug_dump()
+    elif mode == BMCDebugDumpExtractor.COLLECT_MODE:
+        if not task_id or not filepath:
+            log.log_error("Both --task and --path arguments are required for 'collect' mode")
+            return
+        extractor.extract_debug_dump_file(task_id, filepath)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="BMC tech-support generator script.")
+    parser.add_argument(
+        '-m', '--mode',
+        choices=['collect', 'trigger'],
+        required=True,
+        help="Mode of operation: 'collect' for collecting debug dump or 'trigger' for triggering debug dump task."
+    )
+    parser.add_argument('-p', '--path', help="Path to save the BMC debug log dump file.")
+    parser.add_argument('-t', '--task', help="Task-ID to monitor and collect the debug dump from.")
+    args = parser.parse_args()
+    mode = args.mode
+    task_id = args.task
+    filepath = args.path
+    main(mode, task_id, filepath)

scripts/generate_dump

@@ -1950,6 +1950,77 @@ save_log_files() {
     enable_logrotate
 }
 
+###############################################################################
+# Check BMC presence
+# Arguments:
+#  None
+# Returns:
+#  0 if BMC is supported, 1 otherwise
+###############################################################################
+is_bmc_supported() {
+    local platform=$(python3 -c "from sonic_py_common import device_info; print(device_info.get_platform())")
+    # Check if the required file exists
+    if [ ! -f /usr/share/sonic/device/$platform/bmc.json ]; then
+        return 1
+    else
+        return 0
+    fi
+}
+
+###############################################################################
+# Trigger BMC debug log dump task
+# Arguments:
+#  None
+# Returns:
+#  None
+###############################################################################
+trigger_bmc_debug_log_dump() {
+    trap 'handle_error $? $LINENO' ERR
+    if ! is_bmc_supported; then
+        echo "INFO: BMC is not found on this platform. Skipping..."
+        return
+    fi
+    # Trigger BMC redfish API to start BMC debug log dump task
+    local task_id=$(python3 /usr/local/bin/bmc_techsupport.py -m trigger)
+    echo "$task_id"
+}
+
+###############################################################################
+# Save BMC debug log dump files
+# Globals:
+#  MKDIR, CP, TARDIR, TECHSUPPORT_TIME_INFO
+# Arguments:
+#  $1 - BMC debug log dump task ID
+# Returns:
+#  None
+###############################################################################
+collect_bmc_files() {
+    $MKDIR $V -p $TARDIR/bmc
+    trap 'handle_error $? $LINENO' ERR
+    start_t=$(date +%s%3N)
+    if ! is_bmc_supported; then
+        return
+    fi
+
+    local bmc_debug_log_dump_task_id=$1
+    local TARBALL_XZ="/tmp/bmc_debug_log_dump.tar.xz"
+    # Remove existing tarball files if they exist
+    [ -f "$TARBALL_XZ" ] && rm -f "$TARBALL_XZ"
+
+    # Invoke BMC redfish API to extract BMC debug log dump to "/tmp/bmc_debug_log_dump.tar.xz"
+    python3 /usr/local/bin/bmc_techsupport.py -m collect -p "$TARBALL_XZ" -t "$bmc_debug_log_dump_task_id"
+    if [ -f "$TARBALL_XZ" ]; then
+        $CP $V -rf "$TARBALL_XZ" $TARDIR/bmc
+    else
+        echo "ERROR: File $TARBALL_XZ does not exist."
+    fi
+
+    # Cleanup
+    [ -f "$TARBALL_XZ" ] && rm -f "$TARBALL_XZ"
+    end_t=$(date +%s%3N)
+    echo "[ collect_bmc_files ] : $(($end_t-$start_t)) msec" >> $TECHSUPPORT_TIME_INFO
+}
+
 ###############################################################################
 # Save warmboot files
 # Globals:
@@ -2176,6 +2247,12 @@ main() {
     echo $BASE > $TECHSUPPORT_TIME_INFO
     start_t=$(date +%s%3N)
 
+    # Trigger BMC debug log dump task - Must be the first task to run
+    bmc_debug_log_dump_task_id=$(trigger_bmc_debug_log_dump)
+    if [ "$bmc_debug_log_dump_task_id" == "-1" ]; then
+        echo "INFO: Fail to trigger BMC debug log dump. Skipping..."
+    fi
+
     # Capture /proc state early
     save_proc /proc/buddyinfo /proc/cmdline /proc/consoles \
         /proc/cpuinfo /proc/devices /proc/diskstats /proc/dma \
@@ -2389,6 +2466,11 @@ main() {
     save_log_files &
     save_crash_files &
     save_warmboot_files &
+
+    if [ "$bmc_debug_log_dump_task_id" != "-1" ]; then
+        collect_bmc_files $bmc_debug_log_dump_task_id &
+    fi
+
     wait
 
     save_to_tar

setup.py

@@ -189,6 +189,7 @@
         'scripts/memory_threshold_check.py',
         'scripts/memory_threshold_check_handler.py',
         'scripts/techsupport_cleanup.py',
+        'scripts/bmc_techsupport.py',
         'scripts/storm_control.py',
         'scripts/verify_image_sign.sh',
         'scripts/verify_image_sign_common.sh',

show/platform.py

@@ -70,6 +70,110 @@ def summary(json):
             click.echo("Switch Type: {}".format(switch_type))
 
 
+# 'bmc' subcommand ("show platform bmc")
+@platform.group()
+def bmc():
+    """Show BMC information"""
+    pass
+
+
+# 'summary' subcommand ("show platform bmc summary")
+@bmc.command(name='summary')
+@click.option('--json', is_flag=True, help="Output in JSON format")
+def bmc_summary(json):
+    """Show BMC summary information"""
+    try:
+        import sonic_platform
+        chassis = sonic_platform.platform.Platform().get_chassis()
+        bmc = chassis.get_bmc()
+
+        if bmc is None:
+            click.echo("BMC is not available on this platform")
+            return
+
+        eeprom_info = bmc.get_eeprom()
+        if not eeprom_info:
+            click.echo("Failed to retrieve BMC EEPROM information")
+            return
+
+        # Extract the required fields
+        manufacturer = eeprom_info.get('Manufacturer', 'N/A')
+        model = eeprom_info.get('Model', 'N/A')
+        part_number = eeprom_info.get('PartNumber', 'N/A')
+        power_state = eeprom_info.get('PowerState', 'N/A')
+        serial_number = eeprom_info.get('SerialNumber', 'N/A')
+        bmc_version = bmc.get_version()
+
+        if json:
+            bmc_summary = {
+                'Manufacturer': manufacturer,
+                'Model': model,
+                'PartNumber': part_number,
+                'SerialNumber': serial_number,
+                'PowerState': power_state,
+                'FirmwareVersion': bmc_version
+            }
+            click.echo(clicommon.json_dump(bmc_summary))
+        else:
+            click.echo(f"Manufacturer: {manufacturer}")
+            click.echo(f"Model: {model}")
+            click.echo(f"PartNumber: {part_number}")
+            click.echo(f"SerialNumber: {serial_number}")
+            click.echo(f"PowerState: {power_state}")
+            click.echo(f"FirmwareVersion: {bmc_version}")
+
+    except Exception as e:
+        click.echo(f"Error retrieving BMC information: {str(e)}")
+
+
+# 'eeprom' subcommand ("show platform bmc eeprom")
+@bmc.command()
+@click.option('--json', is_flag=True, help="Output in JSON format")
+def eeprom(json):
+    """Show BMC EEPROM information"""
+    try:
+        import sonic_platform
+        chassis = sonic_platform.platform.Platform().get_chassis()
+        bmc = chassis.get_bmc()
+
+        if bmc is None:
+            click.echo("BMC is not available on this platform")
+            return
+
+        # Get BMC EEPROM information
+        eeprom_info = bmc.get_eeprom()
+
+        if not eeprom_info:
+            click.echo("Failed to retrieve BMC EEPROM information")
+            return
+
+        # Extract the required fields
+        manufacturer = eeprom_info.get('Manufacturer', 'N/A')
+        model = eeprom_info.get('Model', 'N/A')
+        part_number = eeprom_info.get('PartNumber', 'N/A')
+        power_state = eeprom_info.get('PowerState', 'N/A')
+        serial_number = eeprom_info.get('SerialNumber', 'N/A')
+
+        if json:
+            bmc_eeprom = {
+                'Manufacturer': manufacturer,
+                'Model': model,
+                'PartNumber': part_number,
+                'PowerState': power_state,
+                'SerialNumber': serial_number
+            }
+            click.echo(clicommon.json_dump(bmc_eeprom))
+        else:
+            click.echo(f"Manufacturer: {manufacturer}")
+            click.echo(f"Model: {model}")
+            click.echo(f"PartNumber: {part_number}")
+            click.echo(f"PowerState: {power_state}")
+            click.echo(f"SerialNumber: {serial_number}")
+
+    except Exception as e:
+        click.echo(f"Error retrieving BMC EEPROM information: {str(e)}")
+
+
 # 'syseeprom' subcommand ("show platform syseeprom")
 @platform.command()
 @click.option('--verbose', is_flag=True, help="Enable verbose output")