Pkg xen devel - Jul 2019 - Bug#932336: xen-utils-common: vif-nat exits with code 1 even if successful

Package: xen-utils-common
Version: 4.11.1+92-g6c33308a8d-2
Severity: normal

Dear Maintainer,

This happened after upgrading from Debian 9 to Debian 10, and also after
commenting out line 88 of /etc/xen/scripts/vif-nat.

After I worked around another bug (which hasn't been assigned a number yet -
"vif-nat fails to find backend/vif/x/x/domain") in vif-nat on Debian
10, vif-nat still said it was returning code 1 and was thus still preventing
domUs from being created.
As I attempted to add "echo"s to the script to find out where it was
erroring out, to my surprise I found that just adding the echos made the script
work!
I narrowed it down to one strategic place to add an "echo" to let the
script work normally: before the "handle_iptable" on line 182.
I don't know why this makes it work - perhaps an exit code is being carried
through where it shouldn't.

Thanks,
Paul Sutor 


-- System Information:
Debian Release: 10.0
  APT prefers stable
  APT policy: (500, 'stable')
Architecture: amd64 (x86_64)

Kernel: Linux 4.19.0-5-amd64 (SMP w/32 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8),
LANGUAGE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)
LSM: AppArmor: enabled

Versions of packages xen-utils-common depends on:
ii  libc6           2.28-10
ii  libxenstore3.0  4.11.1+92-g6c33308a8d-2
ii  lsb-base        10.2019051400
ii  python          2.7.16-1
ii  ucf             3.0038+nmu1
ii  udev            241-5
ii  xenstore-utils  4.11.1+92-g6c33308a8d-2

xen-utils-common recommends no packages.

Versions of packages xen-utils-common suggests:
pn  xen-doc  <none>

-- Configuration Files:
/etc/default/xendomains changed:
XENDOMAINS_RESTORE=false
XENDOMAINS_AUTO=/etc/xen/auto
XENDOMAINS_STOP_MAXWAIT=300

/etc/xen/scripts/vif-nat changed:
dir=$(dirname "$0")
. "$dir/vif-common.sh"
if [ -f /etc/dhcpd.conf ]
then
	dhcp=${dhcp:-yes}
else
	dhcp=${dhcp:-no}
fi
if [ "$dhcp" != 'no' ]
then
  dhcpd_conf_file=$(find_dhcpd_conf_file)
  dhcpd_init_file=$(find_dhcpd_init_file)
  dhcpd_arg_file=$(find_dhcpd_arg_file)
  if [ -z "$dhcpd_conf_file" ] || [ -z "$dhcpd_init_file" ]
|| [ -z "$dhcpd_arg_file" ]
  then
    echo 'Failed to find dhcpd configuration or init or args file.'
>&2
    exit 1
  fi
fi
domid=$(xenstore_read "$XENBUS_PATH/frontend-id")
vifid=$(xenstore_read "$XENBUS_PATH/handle")
vifid=$(( $vifid + 1 ))
ip_from_dom()
{
  local domid1=$(( $domid / 256 ))
  local domid2=$(( $domid % 256 ))
  echo "10.$domid1.$domid2.$vifid/16"
}
routing_ip()
{
  echo $(echo $1 | awk -F. '{print
$1"."$2"."$3"."$4 + 127}')
}
dotted_quad()
{
 echo\
 $(( ($1 & 0xFF000000) >> 24))\
.$(( ($1 & 0x00FF0000) >> 16))\
.$(( ($1 & 0x0000FF00) >> 8 ))\
.$((  $1 & 0x000000FF       ))
}
if [ "$ip" = "" ]
then
  ip=$(ip_from_dom)
fi
router_ip=$(routing_ip "$ip")
vif_ip=`echo ${ip} | awk -F/ '{print $1}'`
if [ "$vifid" != "1" ]
then
  hostname="$hostname-$vifid"
fi
dhcparg_remove_entry()
{
  local tmpfile=$(mktemp)
  sed -e "s/${dev} //" "$dhcpd_arg_file"
>"$tmpfile"
  if diff "$tmpfile" "$dhcpd_arg_file" >/dev/null
  then
    rm "$tmpfile"
  else
    mv "$tmpfile" "$dhcpd_arg_file"
  fi
}
dhcparg_add_entry()
{
  dhcparg_remove_entry
  local tmpfile=$(mktemp)
  # handle Red Hat, SUSE, and Debian styles, with or without quotes
  sed -e
's/^DHCPDARGS="*\([^"]*\)"*/DHCPDARGS="\1'"${dev}
"'"/' \
     "$dhcpd_arg_file" >"$tmpfile" && mv
"$tmpfile" "$dhcpd_arg_file"
  sed -e
's/^DHCPD_INTERFACE="*\([^"]*\)"*/DHCPD_INTERFACE="\1'"${dev}
"'"/' \
     "$dhcpd_arg_file" >"$tmpfile" && mv
"$tmpfile" "$dhcpd_arg_file"
  sed -e
's/^INTERFACES="*\([^"]*\)"*/INTERFACES="\1'"${dev}
"'"/' \
     "$dhcpd_arg_file" >"$tmpfile" && mv
"$tmpfile" "$dhcpd_arg_file"
  rm -f "$tmpfile"
}
dhcp_remove_entry()
{
  local tmpfile=$(mktemp)
  grep -v "host $hostname" "$dhcpd_conf_file"
>"$tmpfile"
  if diff "$tmpfile" "$dhcpd_conf_file" >/dev/null
  then
    rm "$tmpfile"
  else
    mv "$tmpfile" "$dhcpd_conf_file"
  fi
  dhcparg_remove_entry
}
dhcp_up()
{
  claim_lock "vif-nat-dhcp"
  dhcp_remove_entry
  mac=$(xenstore_read "$XENBUS_PATH/mac")
  echo >>"$dhcpd_conf_file" \
"host $hostname { hardware ethernet $mac; fixed-address $vif_ip; option
routers $router_ip; option host-name \"$hostname\"; }"
  dhcparg_add_entry
  release_lock "vif-nat-dhcp"
  "$dhcpd_init_file" restart || true
}
dhcp_down()
{
  claim_lock "vif-nat-dhcp"
  dhcp_remove_entry
  release_lock "vif-nat-dhcp"
  "$dhcpd_init_file" restart || true # We need to ignore failure
because
                                     # ISC dhcpd 3 borks if there is nothing
                                     # for it to do, which is the case if
                                     # the outgoing interface is not
                                     # configured to offer leases and there
                                     # are no vifs.
}
case "$command" in
    online)
        if ip route | grep -q "dev ${dev}"
        then
          log debug "${dev} already up"
          exit 0
        fi
        do_or_die ip link set dev "${dev}" up arp on
        do_or_die ip addr add "$router_ip" dev "${dev}"
        do_or_die ip route add "$vif_ip" dev "${dev}" src
"$router_ip"
        echo 1 >/proc/sys/net/ipv4/conf/${dev}/proxy_arp
        [ "$dhcp" != 'no' ] && dhcp_up
        ;;
    offline)
        [ "$dhcp" != 'no' ] && dhcp_down
        do_without_error ifconfig "${dev}" down
        ;;
esac
echo hi
handle_iptable
call_hooks vif post
log debug "Successful vif-nat $command for ${dev}."
if [ "$command" = "online" ]
then
  success
fi

/etc/xen/xend-config.sxp changed:
(dom0-min-mem 640)
(enable-dom0-ballooning yes)
(total_available_memory 0) 
(dom0-cpus 0)
(vncpasswd '')

/etc/xen/xl.conf changed:
vif.default.script="vif-nat"


-- no debconf information

I stumbled over exactly the same issue today, tried to narrow it down by
adding debug messages to the script, and suddenly it worked.

I think the bad guy is

	[ "$dhcp" != 'no' ] && dhcp_up

which returns 1 in case no dhcp is used. Maybe using "if" here is
better...

Also handle_iptable (in vif-common.sh) only does a "return" which
again
returns 1. Maybe using "return 0" here is better...

The question (for me) is why this leads to an exit. Even if "set -e"
is
used in some scripts, why does it exit at handle_iptable() and not
already after the false [ "$dhcp" != 'no' ]?

Stephan