LU-13595 scripts: Add a debug option to lustre_rmmod

The option is "-d" or "--debug-kernel", it uses "lctl debug_kernel" to
get debug message before unload libcfs.
For each module the script add a debug mark before unloading.

Example:
	$ lustre_rmmod -d > /tmp/lustre_rmmod.log
	$ lustre_rmmod -d lustre lnet > /tmp/lustre_rmmod2.log

Test-Parameters: trivial
Signed-off-by: Etienne AUJAMES <eaujames@ddn.com>
Change-Id: I6e44a24f2e786c08faf1db27de94e0f88ca65dc7
Reviewed-on: https://review.whamcloud.com/38941
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: Olaf Faaland-LLNL <faaland1@llnl.gov>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>

diff --git a/lustre/scripts/lustre_rmmod b/lustre/scripts/lustre_rmmod
index 53c8ae2..9da25ff 100755 (executable)
--- a/lustre/scripts/lustre_rmmod
+++ b/lustre/scripts/lustre_rmmod
@@ -5,9 +5,33 @@
 # returned.
 ###############################################################################
 
 # returned.
 ###############################################################################
 

+SCRIPT_NAME="$(basename "$0")"
+LCTL=${LCTL:-lctl}
+DEBUG='false'
+
+# Print help message
+print_usage() {
+       echo "$SCRIPT_NAME -h|--help"
+       echo "$SCRIPT_NAME [-d|--debug-kernel] [modulesname...]"
+       echo
+       echo -e "\t-h, --help\t\tDisplay this help message"
+       echo -e "\t-d, --debug-kernel\tDisplay lustre kernel debug messages"
+       echo -e "\tmodulesname\t\tList of lustre modules to unload. By default"
+       echo -e "\t\t\t\tldiskfs and libcfs (and their dependencies) are"
+       echo -e "\t\t\t\tselected."
+}
+
+# Print kernel debug message for lustre modules
+print_debug() {
+       $LCTL mark "$SCRIPT_NAME : Stop debug"
+       $LCTL debug_kernel
+       DEBUG='false'
+}
+

 # Unload all modules dependent on $1 (exclude removal of $1)
 unload_dep_modules_exclusive() {
        local MODULE=$1
 # Unload all modules dependent on $1 (exclude removal of $1)
 unload_dep_modules_exclusive() {
        local MODULE=$1

+

        local DEPS="$(lsmod | awk '($1 == "'$MODULE'") { print $4 }')"
        for SUBMOD in $(echo $DEPS | tr ',' ' '); do
                unload_dep_modules_inclusive $SUBMOD || return 1
        local DEPS="$(lsmod | awk '($1 == "'$MODULE'") { print $4 }')"
        for SUBMOD in $(echo $DEPS | tr ',' ' '); do
                unload_dep_modules_inclusive $SUBMOD || return 1


@@ -22,17 +46,49 @@ unload_dep_modules_inclusive() {
        # if $MODULE not loaded, return 0
        lsmod | egrep -q "^\<$MODULE\>" || return 0
        unload_dep_modules_exclusive $MODULE || return 1
        # if $MODULE not loaded, return 0
        lsmod | egrep -q "^\<$MODULE\>" || return 0
        unload_dep_modules_exclusive $MODULE || return 1

+
+       if $DEBUG; then
+               if [ "$MODULE" = 'libcfs' ]; then
+                       print_debug
+               fi
+               $LCTL mark "$SCRIPT_NAME : Unload $MODULE"
+       fi
+

        rmmod $MODULE || return 1
        return 0
 }
 
        rmmod $MODULE || return 1
        return 0
 }
 

-modules="$@"
+declare -a modules
+while [ $# -gt 0 ]; do
+       case "$1" in
+               -h|--help)
+                       print_usage >&2
+                       exit 0
+                       ;;
+               -d|--debug-kernel)
+                       if lsmod | egrep -q '^libcfs'; then
+                               DEBUG='true'
+                       else
+                               echo "Debug unavailable: libcfs is not loaded" >&2
+                       fi
+                       ;;
+               -*)
+                       echo "Error invalid option" >&2
+                       print_usage >&2
+                       exit 2
+                       ;;
+               *)
+                       modules+=("$1")
+                       ;;
+       esac
+       shift
+done

 
 # To maintain backwards compatibility, ldiskfs and libcfs must be
 # unloaded if no parameters are given, or if only the ldiskfs parameter
 # is given. It's ugly, but is needed to emulate the prior functionality
 
 # To maintain backwards compatibility, ldiskfs and libcfs must be
 # unloaded if no parameters are given, or if only the ldiskfs parameter
 # is given. It's ugly, but is needed to emulate the prior functionality

-if [ -z "$modules" ] || [ "$modules" = "ldiskfs" ]; then
-       modules="ptlrpc lnet_selftest ldiskfs libcfs"
+if [ "${#modules[@]}" -eq 0 ] || [ "${modules[*]}" = "ldiskfs" ]; then
+       modules=('ptlrpc' 'lnet_selftest' 'ldiskfs' 'libcfs')

 fi
 
 if [ -f /sys/kernel/debug/kmemleak ] ; then
 fi
 
 if [ -f /sys/kernel/debug/kmemleak ] ; then


@@ -46,10 +102,23 @@ if [ -f /sys/kernel/debug/kmemleak ] ; then
        echo clear > /sys/kernel/debug/kmemleak
 fi
 
        echo clear > /sys/kernel/debug/kmemleak
 fi
 

-for mod in $modules; do
+# Manage debug
+if $DEBUG; then
+       echo "Lustre debug parameters:" >&2
+       $LCTL get_param debug >&2
+       $LCTL get_param debug_mb >&2
+
+       $LCTL mark "$SCRIPT_NAME : Start debug"
+fi
+
+for mod in ${modules[*]}; do

        unload_dep_modules_inclusive $mod || exit 1
 done
 
        unload_dep_modules_inclusive $mod || exit 1
 done
 

+if $DEBUG; then
+       print_debug
+fi
+

 if [ -f /sys/kernel/debug/kmemleak ] ; then
        echo scan > /sys/kernel/debug/kmemleak
        cat /sys/kernel/debug/kmemleak > /tmp/kmemleak-after-unload.txt
 if [ -f /sys/kernel/debug/kmemleak ] ; then
        echo scan > /sys/kernel/debug/kmemleak
        cat /sys/kernel/debug/kmemleak > /tmp/kmemleak-after-unload.txt