On 11/21/2011 06:28 AM, Stefan Berger wrote:
Introduce a shell variable 'IBT' to invoke the ip(6)tables
command.
Tested with libvirt-tck.
---
src/nwfilter/nwfilter_ebiptables_driver.c | 313 ++++++++++++++----------------
1 file changed, 155 insertions(+), 158 deletions(-)
Index: libvirt-acl/src/nwfilter/nwfilter_ebiptables_driver.c
===================================================================
--- libvirt-acl.orig/src/nwfilter/nwfilter_ebiptables_driver.c
+++ libvirt-acl/src/nwfilter/nwfilter_ebiptables_driver.c
@@ -146,6 +146,10 @@ static const char ebiptables_script_set_
#define NWFILTER_SET_EBTABLES_SHELLVAR(BUFPTR) \
virBufferAsprintf(BUFPTR, "EBT=%s\n", ebtables_cmd_path);
These variable assignments will fail if the absolute path to
ebtables_cmd_path and friends contain any whitespace or shell
metacharacters, but then again, the old code pre-conversion would have
failed in similar manners, so we already had an implicit assumption that
the user isn't calling ./configure with an insane PATH containing a
directory with whitespace at the front as the location for their
preferred commands. I see nothing wrong with continuing that
assumption, although someday we may want to make configure fail if it
detects a tool name that cannot be safely passed unquoted through shell
(at least, for the three tools mentioned in this .c file where we are
using the shell), as a safety valve in the interim of us ever making the
much bigger change of avoiding shell in the first place.
ACK as-is.
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org