Kdump improvements (sonic-net#1284)

- Added support for reporting kdump reboot cause to detect and report
a kernel crash
- Added additional commands to sonic-kdump-config wrapper script to
dump Kdump configuration and status information in JSON format
- Automatically save Kdump configuration in startup configuration file.
This is an added convenience to the user as Kdump configuration changes
require a system reboot and user is expected to save the configuration so
that they are activated post reboot.
- Additional bug fixes and improvements to allow hostcfgd to process Kdump
configuration changes in ConfigDB. sonic-kdump-config script will write to
ConfigDB.
- Moved kdump CLI commands to a dedicated file

Signed-off-by: Rajendra Dendukuri <rajendra.dendukuri@broadcom.com>

config/kdump.py

@@ -0,0 +1,44 @@
+import os
+import click
+import utilities_common.cli as clicommon
+from swsssdk import ConfigDBConnector
+
+@click.group(cls=clicommon.AbbreviationGroup, name="kdump")
+def kdump():
+ """ Configure kdump """
+ if os.geteuid() != 0:
+ exit("Root privileges are required for this operation")
+
+@kdump.command()
+def disable():
+ """Disable kdump operation"""
+ config_db = ConfigDBConnector()
+ if config_db is not None:
+ config_db.connect()
+ config_db.mod_entry("KDUMP", "config", {"enabled": "false"})
+
+@kdump.command()
+def enable():
+ """Enable kdump operation"""
+ config_db = ConfigDBConnector()
+ if config_db is not None:
+ config_db.connect()
+ config_db.mod_entry("KDUMP", "config", {"enabled": "true"})
+
+@kdump.command()
+@click.argument('kdump_memory', metavar='<kdump_memory>', required=True)
+def memory(kdump_memory):
+ """Set memory allocated for kdump capture kernel"""
+ config_db = ConfigDBConnector()
+ if config_db is not None:
+ config_db.connect()
+ config_db.mod_entry("KDUMP", "config", {"memory": kdump_memory})
+
+@kdump.command('num-dumps')
+@click.argument('kdump_num_dumps', metavar='<kdump_num_dumps>', required=True, type=int)
+def num_dumps(kdump_num_dumps):
+ """Set max number of dump files for kdump"""
+ config_db = ConfigDBConnector()
+ if config_db is not None:
+ config_db.connect()
+ config_db.mod_entry("KDUMP", "config", {"num_dumps": kdump_num_dumps})

config/main.py

@@ -28,6 +28,7 @@
 from . import chassis_modules
 from . import console
 from . import feature
+from . import kdump
 from . import kube
 from . import mlnx
 from . import muxcable
@@ -878,6 +879,7 @@ def config(ctx):
 config.add_command(chassis_modules.chassis_modules)
 config.add_command(console.console)
 config.add_command(feature.feature)
+config.add_command(kdump.kdump)
 config.add_command(kube.kubernetes)
 config.add_command(muxcable.muxcable)
 config.add_command(nat.nat)
@@ -1901,50 +1903,6 @@ def shutdown():
  """Shut down BGP session(s)"""
  pass
 
-@config.group(cls=clicommon.AbbreviationGroup)
-def kdump():
- """ Configure kdump """
- if os.geteuid() != 0:
- exit("Root privileges are required for this operation")
-
-@kdump.command()
-def disable():
- """Disable kdump operation"""
- config_db = ConfigDBConnector()
- if config_db is not None:
- config_db.connect()
- config_db.mod_entry("KDUMP", "config", {"enabled": "false"})
- clicommon.run_command("sonic-kdump-config --disable")
-
-@kdump.command()
-def enable():
- """Enable kdump operation"""
- config_db = ConfigDBConnector()
- if config_db is not None:
- config_db.connect()
- config_db.mod_entry("KDUMP", "config", {"enabled": "true"})
- clicommon.run_command("sonic-kdump-config --enable")
-
-@kdump.command()
-@click.argument('kdump_memory', metavar='<kdump_memory>', required=True)
-def memory(kdump_memory):
- """Set memory allocated for kdump capture kernel"""
- config_db = ConfigDBConnector()
- if config_db is not None:
- config_db.connect()
- config_db.mod_entry("KDUMP", "config", {"memory": kdump_memory})
- clicommon.run_command("sonic-kdump-config --memory %s" % kdump_memory)
-
-@kdump.command('num-dumps')
-@click.argument('kdump_num_dumps', metavar='<kdump_num_dumps>', required=True, type=int)
-def num_dumps(kdump_num_dumps):
- """Set max number of dump files for kdump"""
- config_db = ConfigDBConnector()
- if config_db is not None:
- config_db.connect()
- config_db.mod_entry("KDUMP", "config", {"num_dumps": kdump_num_dumps})
- clicommon.run_command("sonic-kdump-config --num_dumps %d" % kdump_num_dumps)
-
 # 'all' subcommand
 @shutdown.command()
 @click.option('-v', '--verbose', is_flag=True, help="Enable verbose output")

scripts/reboot

@@ -1,19 +1,28 @@
 #!/bin/bash
+DEVPATH="/usr/share/sonic/device"
+PLAT_REBOOT="platform_reboot"
+REBOOT_CAUSE_FILE="/host/reboot-cause/reboot-cause.txt"
+REBOOT_TIME=$(date)
 
 # 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"
- /sbin/reboot
+ 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 '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=$(logname)
-REBOOT_TIME=$(date)
 PLATFORM=$(sonic-cfggen -H -v DEVICE_METADATA.localhost.platform)
 ASIC_TYPE=$(sonic-cfggen -y /etc/sonic/sonic_version.yml -v asic_type)
-DEVPATH="/usr/share/sonic/device"
-PLAT_REBOOT="platform_reboot"
-REBOOT_CAUSE_FILE="/host/reboot-cause/reboot-cause.txt"
 VERBOSE=no
 EXIT_NEXT_IMAGE_NOT_EXISTS=4
 EXIT_SONIC_INSTALLER_VERIFY_REBOOT=21