On Fri, Mar 11, 2011 at 11:38:47AM +0100, Jiri Denemark wrote:

On Tue, Mar 01, 2011 at 18:57:01 +0100, Philipp Hahn wrote:

...

awk splits the line on consecutive spaces, which breaks getting the name of a domain whose name contains spaces. Use sed instead to strip the "Name:" prefix from the line

Signed-off-by: Philipp Hahn <hahn@univention.de> --- tools/libvirt-guests.init.sh | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/tools/libvirt-guests.init.sh b/tools/libvirt-guests.init.sh index 8823d06..ea2189e 100644 --- a/tools/libvirt-guests.init.sh +++ b/tools/libvirt-guests.init.sh @@ -105,7 +105,7 @@ guest_name() { uuid=$2

name=$(run_virsh_c $uri dominfo $uuid 2>/dev/null | \ - awk '/^Name:/{print $2}') + sed -ne 's/^Name: *//p') [ -n "$name" ] || name=$uuid

echo "$name"

Ah yes, domain names with spaces... I guess we should enhance virsh so that it can start domains based on UUID so the we can get rid of the last domain name usage when calling virsh in this script.

virsh already *can* start VMs based on UUID. Any virsh command which needs an inactive domain can accept either name of UUID interchangably. Any virsh command which needs an active domain can accept name, UUID or ID Daniel -- |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|

On 03/01/2011 10:57 AM, Philipp Hahn wrote:

awk splits the line on consecutive spaces, which breaks getting the name of a domain whose name contains spaces. Use sed instead to strip the "Name:" prefix from the line

Signed-off-by: Philipp Hahn <hahn@univention.de> --- tools/libvirt-guests.init.sh | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/tools/libvirt-guests.init.sh b/tools/libvirt-guests.init.sh index 8823d06..ea2189e 100644 --- a/tools/libvirt-guests.init.sh +++ b/tools/libvirt-guests.init.sh @@ -105,7 +105,7 @@ guest_name() { uuid=$2

name=$(run_virsh_c $uri dominfo $uuid 2>/dev/null | \ - awk '/^Name:/{print $2}') + sed -ne 's/^Name: *//p') [ -n "$name" ] || name=$uuid

echo "$name"

ACK - this patch is independently useful from Jirka's patch to not use $name in virsh, because we also use the output of guest_name in informational messages and should not truncate there. I've pushed it. -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org

On 03/09/2011 01:54 AM, Philipp Hahn wrote:

At least protect the $uri variable against further expansion by properly quoting it. While doing that, also quote all other variables to protect against shell meta characters.

Hmm. Valid URIs can contain '?', which _is_ a glob character, so I agree that we need to quote "$uri" everywhere that we have already correctly obtatined it in order to avoid inadvertent globbing. Meanwhile, $URIS is indeed a user-settable variable, via /etc/sysconfig/libvirt-guests, so we should want to protect ourselves from malicious input. (On the other hand, we already blindly source /etc/sysconfig/libvirt-guests without validating that it contains _only_ shell assignments, so a malicious user could already put arbitrary shell code in there. The real way to be safe would be to parse out _just_ the assignments we care about, using sed or read, and eval those assignments rather than sourcing the entire file. I guess we have to draw a line somewhere at how much paranoia we want to express. If we can trust that /etc/sysconfig/libvirt-guests can't be subverted by a malicious user, then we can trust that it won't contain a malicious $URIS and instead chalk up bugs in that file as system administration shortcomings.) However, that points out a shortcoming in your patch - we already do constructs like for uri in $URIS; do which _already_ corrupted any globbing and performed IFS splitting, at which point $uri would not contain any further globbing (except in the weird case of a glob that expanded to an existing file name which happens to also be a valid glob), so quoting just "$uri" as in your patch is a lost cause unless we address the problem from the point of the user input to begin with. Meanwhile, we already know that a valid URI does not contain whitespace (a valid URI can represent whitespace via encoding, where needed), so it's already acceptable to use IFS splitting on $URIS to break it into individual URIs, it's just that we need to suppress globbing in the process. 'set -f' can be used for this, if needed. If we could use bash shell arrays, it would be a matter of creating an array variable ${URI[]} that contains the individual URIs, doing the work of safely splitting things only once. But libvirt-guests has to run under /bin/sh == dash for Debian, which lacks shell arrays. I'm still thinking about how to handle this one... Meanwhile, I'm not a fan of blindly quoting everything; there are documented cases where you can trigger shell bugs by doing too much quoting. For example: foo=`some_command "with quotes"` is portable, but foo="`some_command "with quotes"`" is not. So I prefer to quote variables that come from external source, but not internal variables that are obviously safe (that said, quoting generally doesn't hurt, so even if something is overkill does not meant that I am rejecting the patch).

+++ b/tools/libvirt-guests.init.sh @@ -66,12 +66,10 @@ run_virsh() { shift

if [ "x$uri" = xdefault ]; then - conn= + virsh "$@" </dev/null else - conn="-c $uri" + virsh -c "$uri" "$@" </dev/null fi - - virsh $conn "$@" </dev/null

This part's good.

@@ -89,7 +87,7 @@ list_guests() {

uuids= for id in $(echo "$list" | awk 'NR > 2 {print $1}'); do - uuid=$(run_virsh_c $uri dominfo $id | awk '/^UUID:/{print $2}') + uuid=$(run_virsh_c "$uri" dominfo "$id" | awk '/^UUID:/{print $2}')

Quoting "$id" is overkill; we know it's numeric. But it probably doesn't hurt.

- name=$(run_virsh_c $uri dominfo $uuid 2>/dev/null | \ + name=$(run_virsh_c "$uri" dominfo "$uuid" 2>/dev/null | \

Likewise, if we know $uuid is valid, then it contains neither $IFS nor globbing characters. Should we instead be tightening up the list_guests function to guarantee that we never encounter garbage data?

@@ -146,25 +144,25 @@ start() { while read uri list; do configured=false for confuri in $URIS; do - if [ $confuri = $uri ]; then + if [ "$confuri" = "$uri" ]; then

To be really robust (for example, $confuri could be '(', which is known to trip up some buggy test(1) implementations), this should be: [ "x$confuri" = "x$uri" ]

configured=true break fi done - if ! $configured; then + if ! "$configured"; then

This one's overkill - we explicitly set $configured to either 'true' or 'false'; no external data possible.

virsh_pid=$! while true; do sleep 1 - kill -0 $virsh_pid >/dev/null 2>&1 || break - progress=$(run_virsh_c $uri domjobinfo $guest 2>/dev/null | \ + kill -0 "$virsh_pid" >/dev/null 2>&1 || break + progress=$(run_virsh_c "$uri" domjobinfo "$guest" 2>/dev/null | \

Overkill - we know $virsh_pid is numeric.

timeout=$SHUTDOWN_TIMEOUT - while [ $timeout -gt 0 ]; do + while [ "$timeout" -gt 0 ]; do

Reasonable, since SHUTDOWN_TIMEOUT might come from external sources.

sleep 1 timeout=$((timeout - 1)) - guest_is_on $uri $guest || return - $guest_running || break - printf '\r%s%-12d ' "$label" $timeout + guest_is_on "$uri" "$guest" || return + "$guest_running" || break + printf '\r%s%-12d ' "$label" "$timeout"

Overkill, since once you are inside the while loop, you are guaranteed that $timeout is now numeric.

done

- if guest_is_on $uri $guest; then - if $guest_running; then + if guest_is_on "$uri" "$guest"; then + if "$guest_running"; then

Overkill, since $guest_running is under our control.

- list=$(list_guests $uri) + list=$(list_guests "$uri") if [ $? -eq 0 ]; then empty=true for uuid in $list; do - $empty || printf ", " - printf %s "$(guest_name $uri $uuid)" + "$empty" || printf ", "

Overkill, since $empty is under our control.

while read uri list; do - if $suspending; then + if "$suspending"; then

Likewise.

@@ -330,7 +328,7 @@ case "$1" in usage 0 ;; start|stop|gueststatus) - $1 + "$1"

Overkill, since we know that $1 is one of three safe values. I think I'll apply your patch as-is, in spite of the overkill, then worry about how to safely split $URIS as a separate patch. -- Eric Blake eblake@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org

Hello Eric, On Friday Mar 11th 2011 21:48:03 Eric Blake wrote:

On 03/09/2011 01:54 AM, Philipp Hahn wrote:

...

At least protect the $uri variable against further expansion by properly quoting it. While doing that, also quote all other variables to protect against shell meta characters.

Meanwhile, $URIS is indeed a user-settable variable, via /etc/sysconfig/libvirt-guests,

I would call it root-configurable instead of user-settable, because that file is normally owned by root and has mod 0644. A sane root hopefully does not put a 'rm -rf /' in there ;-) It would be something else if that file was sourced by a program running with more privileges than the user, who is allowed to edit that file. But yes, even root should expect decent bahavior and should not be surprised by the shell expanding string containing wildcards.

Meanwhile, I'm not a fan of blindly quoting everything; there are documented cases where you can trigger shell bugs by doing too much quoting. For example:

foo=`some_command "with quotes"`

is portable, but

foo="`some_command "with quotes"`"

is not. So I prefer to quote variables that come from external source, but not internal variables that are obviously safe (that said, quoting generally doesn't hurt, so even if something is overkill does not meant that I am rejecting the patch).

I've seen to many scripts not doing any quoting at all or all wrong, which fail in interesting to horrible ways when given arguments containing shell meta characters. I personally consider them a much greater risk than some ancient shell not being able to parse properly quoted strings.

...

@@ -89,7 +87,7 @@ list_guests() {

uuids= for id in $(echo "$list" | awk 'NR > 2 {print $1}'); do - uuid=$(run_virsh_c $uri dominfo $id | awk '/^UUID:/{print $2}') + uuid=$(run_virsh_c "$uri" dominfo "$id" | awk '/^UUID:/{print $2}')

Quoting "$id" is overkill; we know it's numeric. But it probably doesn't hurt.

My rule of thumb is this: if the command dominfo is expecting one argument, I make sure it gets exactly on argument by properly quoting it. I don't care that I know that it will normally not contain blanks or wildcard, so even when someone other later changes the variable in some stupid way, it will still pass one required argument to the function. I prefer virsh to report an error like "'foo bar' is invalid" instead of "'foo' not found" or 'unexpected additional parameter "bar"'.

This one's overkill - we explicitly set $configured to either 'true' or 'false'; no external data possible.

I'm waiting for the day when someone dictates, that the UNIXroot-directory is no longer called '/' but '/ /'. That will give some lovely firefork ;-)

Overkill, since $guest_running is under our control.

...

+ "$1"

Overkill, since we know that $1 is one of three safe values.

I think I'll apply your patch as-is, in spite of the overkill, then worry about how to safely split $URIS as a separate patch.

Only #1 was the really important fix, the others were done because I was already looking for whitespace problems. To thanks for applying this one two. Sincerely Philipp -- Philipp Hahn Open Source Software Engineer hahn@univention.de Univention GmbH Linux for Your Business fon: +49 421 22 232- 0 Mary-Somerville-Str.1 D-28359 Bremen fax: +49 421 22 232-99 http://www.univention.de/