Attachment 653599 Details for Bug 880682 – patch against pmie/pmlogger rc.d scripts

[patch] patch against pmie/pmlogger rc.d scripts

bz880682.patch (text/plain), 5.06 KB, created by Frank Ch. Eigler on 2012-11-28 16:19:56 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Frank Ch. Eigler

Created: 2012-11-28 16:19:56 UTC

Size: 5.06 KB

patch

obsolete

>amended subset of commit 26129ab3fc91a77aefa3c2807c2b49dec4631036
>Author: Nathan Scott <nathans@redhat.com>
>Date:   Wed Nov 28 14:30:31 2012 +1100
>
>    Fix race conditions in pmie and pmlogger startup scripts
>    
>    Recent changes to tempfile handling has regressed the pmie and
>    pmlogger startup scripts.  Errors of the form:
>    /etc/rc.d/init.d/pmlogger: line 100: /var/tmp/pcp.5vfQsSHKo/pmcheck: No such file or directory
>    are now produced.
>    
>    Because sections of these two scripts are run in parallel with
>    the original script, we are open to race conditions where the
>    main script exits and removes the temporary directory before the
>    co-process has an opportunity to create its temporary file.  We
>    can resolve this using separate temporary dirs and trap handling
>    which thus no longer race.
>    
>    Worse still, QA failed to detect the problem.  At least one test
>    that should have found the problem trivially (575) failed to, as
>    a result of aggressively discarding stderr and stdout instead of
>    using filtering.  This has been rectified and common pmie filter
>    routines abstracted (from tests 115, 504 and 575) to simplify the
>    task for all current and future tests.
>
>diff --git pmie.orig pmie
>index bcb0ba7..698c45a 100644
>--- pmie.orig
>+++ pmie
>@@ -1,5 +1,6 @@
> #!/bin/sh
> #
>+# Copyright (c) 2012 Red Hat.
> # Copyright (c) 2000 Silicon Graphics, Inc.  All Rights Reserved.
> # 
> # This program is free software; you can redistribute it and/or modify it
>@@ -109,37 +110,48 @@ _reboot_setup()
>     [ ! -d "$LOGDIR" ] && mkdir -p "$LOGDIR" && chown pcp:pcp "$LOGDIR"
> }
> 
>+# Note: _start_pmie is running in the background, in parallel with
>+# the rest of the script.  It might complete well after the caller
>+# so tmpfile handling is especially problematic.  Goal is to speed
>+# bootup by starting potentially slow (remote monitoring) processes
>+# in the background.
>+#
> _start_pmie()
> {
>+    bgstatus=0
>+    bgtmp=`mktemp -d /var/tmp/pcp.XXXXXXXXX` || exit 1
>+    trap "rm -rf $bgtmp; exit \$bgstatus" 0 1 2 3 15
>+
>     wait_option=''
>     [ ! -z "$PMCD_WAIT_TIMEOUT" ] && wait_option="-t $PMCD_WAIT_TIMEOUT" 
> 
>     if pmcd_wait $wait_option
>     then
>-	pmie_check >$tmp/pmie 2>&1
>-	if [ -s $tmp/pmie ]
>+	pmie_check >$bgtmp/pmie 2>&1
>+	bgstatus=$?
>+	if [ -s $bgtmp/pmie ]
> 	then
> 	    pmpost "pmie_check start failed in $prog, mailing output to root"
> 	    if [ ! -z "$MAIL" ]
> 	    then
>-		$MAIL -s "pmie_check start failed in $prog" root <$tmp/pmie >/dev/null 2>&1
>+		$MAIL -s "pmie_check start failed in $prog" root <$bgtmp/pmie >/dev/null 2>&1
> 	    else
> 		echo "$prog: pmie_check start failed ..."
>-		cat $tmp/pmie
>+		cat $bgtmp/pmie
> 	    fi
> 	fi
>-	rm -f $tmp/pmie
>     else
>-	status=$?
>-	pmpost "pmcd_wait failed in $prog: exit status: $status" 
>+	bgstatus=$?
>+	pmpost "pmcd_wait failed in $prog: exit status: $bgstatus" 
> 	if [ ! -z "$MAIL" ]
> 	then
>-	    echo "pmcd_wait: exit status: $status" | $MAIL -s "pmcd_wait failed in $prog" root
>+	    echo "pmcd_wait: exit status: $bgstatus" | $MAIL -s "pmcd_wait failed in $prog" root
> 	else
> 	    echo "$prog: pmcd_wait failed ..."
>-	    echo "pmcd_wait: exit status: $status"
>+	    echo "pmcd_wait: exit status: $bgstatus"
> 	fi
>     fi
>+    exit $bgstatus	# co-process is now complete
> }
> 
> _shutdown()
>diff --git pmlogger.orig pmlogger
>index 6f9949f..84f3382 100644
>--- pmlogger
>+++ pmlogger
>@@ -1,5 +1,6 @@
> #!/bin/sh
> #
>+# Copyright (c) 2012 Red Hat.
> # Copyright (c) 2000-2008 Silicon Graphics, Inc.  All Rights Reserved.
> # 
> # This program is free software; you can redistribute it and/or modify it
>@@ -12,10 +13,6 @@
> # or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
> # for more details.
> # 
>-# You should have received a copy of the GNU General Public License along
>-# with this program; if not, write to the Free Software Foundation, Inc.,
>-# 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
>-# 
> # Start or Stop the Performance Co-Pilot pmlogger processes.
> #
> # The following is for chkconfig on RedHat based systems
>@@ -95,21 +92,32 @@ in
> 	;;
> esac
> 
>+# Note: _start_pmcheck() runs in the background, in parallel with
>+# the rest of the script.  It might complete well after the caller
>+# so tmpfile handling is especially problematic.  Goal is to speed
>+# bootup by starting potentially slow (remote monitoring) pmlogger
>+# processes in the background.
>+#
> _start_pmcheck()
> {
>-    pmlogger_check $VFLAG >$tmp/pmcheck 2>&1
>-    if [ -s $tmp/pmcheck ]
>+    bgstatus=0
>+    bgtmp=`mktemp -d /var/tmp/pcp.XXXXXXXXX` || exit 1
>+    trap "rm -rf $bgtmp; exit \$bgstatus" 0 1 2 3 15
>+
>+    pmlogger_check $VFLAG >$bgtmp/pmcheck 2>&1
>+    bgstatus=$?
>+    if [ -s $bgtmp/pmcheck ]
>     then
> 	pmpost "pmlogger_check failed in $prog, mailing output to root"
> 	if [ ! -z "$MAIL" ]
> 	then
>-	    $MAIL -s "pmlogger_check failed in $prog" root <$tmp/pmcheck
>+	    $MAIL -s "pmlogger_check failed in $prog" root <$bgtmp/pmcheck
> 	else
> 	    echo "$prog: pmlogger_check failed ..."
>-	    cat $tmp/pmcheck
>+	    cat $bgtmp/pmcheck
> 	fi
>     fi
>-    rm -f $tmp/pmcheck
>+    exit $bgstatus	# co-process is now complete
> }
> 
> _start_pmlogger()

Actions: View | Diff

Attachments on bug 880682: 653599