Re: [PATCH] resolvconf support (Re: NM 0.7 Debian Lenny - strangest issue...)



Alexander Sack wrote:
On Mon, Jul 07, 2008 at 03:51:22PM -0400, Dan Williams wrote:
On Mon, 2008-07-07 at 20:26 +0200, Michael Biebl wrote:
2.) If resolvconf is installed, change NM to not manage /etc/resolv.conf itself, but only pass the dns information to resolvconf (which then will write /etc/resolv.conf). The downside is, that resolvconf might have own dns configuration. So the resulting /etc/resolv.conf might differ from what NM expects it to be.

I'm leaning towards 2.), but I'm open to other suggestions and comments.
I'm fine with #2 also; I'll take patches for this upstream.

But I have one question that I pointed out earlier.  If you install
resolvconf, does that _necessarily_ imply that it is turned on?  i.e. I
can install apache, but I don't actually have to run apache even if it's
installed.  Does resolvconf have an on/off switch somewhere, or do you
have to uninstall it to turn it off?


Below a first patch introducing basic resolvconf support. I decided to

Ok, so you basically agree with #2.

not add/delete nameserver entries for individual NMIP4Config's, but
instead just add/delete _one_ config named "NetworkManager", which is
based on the merged NMIP4Config that NM uses in rewrite_resolv_conf.

The code will fall back to normal resolv.conf rewriting if
resolvconf fails. Further I guarded the new code by #ifdef
WITH_RESOLVCONF. However, I didnt add a configure flag.



I wouldn't necessarily add a configure flag and #ifdef WITH_RESOLVCONF.

I'd simply check at runtime, if the resolvconf binary is available ( g_file_test (RESOLVCONF, G_FILE_TEST_IS_EXECUTABLE)) { .. } )

Compiling NM with resolvconf support doesn't necessarily mean, that the machine NM is installed on, has resolvconf installed.

Comments?


--- src/named-manager/nm-named-manager.c	2008-06-26 19:33:13 +0000
+++ src/named-manager/nm-named-manager.c	2008-07-08 10:07:08 +0000
@@ -29,25 +29,30 @@
 #include <glib.h>
#include <glib/gi18n.h> #include "nm-named-manager.h"
 #include "nm-ip4-config.h"
 #include "nm-utils.h"
 #include "NetworkManagerSystem.h"
+#include "NetworkManagerUtils.h"
#ifdef HAVE_SELINUX
 #include <selinux/selinux.h>
 #endif
#ifndef RESOLV_CONF
 #define RESOLV_CONF "/etc/resolv.conf"
 #endif
+#ifndef RESOLV_CONF_PATH
+#define RESOLV_CONF_PATH "/sbin/resolvconf"
+#endif


I'd name that simply RESOLVCONF

 G_DEFINE_TYPE(NMNamedManager, nm_named_manager, G_TYPE_OBJECT)
#define NM_NAMED_MANAGER_GET_PRIVATE(o) (G_TYPE_INSTANCE_GET_PRIVATE ((o), \
                                          NM_TYPE_NAMED_MANAGER, \
                                          NMNamedManagerPrivate))
struct NMNamedManagerPrivate {
@@ -291,23 +296,51 @@ rewrite_resolv_conf (NMNamedManager *mgr
 				     g_strerror (errno));
 		}
 	}
g_free (domain);
 	g_free (searches);
if (*error == NULL) {
-		if (rename (tmp_resolv_conf, RESOLV_CONF) < 0) {
+		gboolean do_rewrite = TRUE;
+
+#ifdef WITH_RESOLVCONF
+		gint exit_code;
+		char *cmd;
+
+		// delete NM resolvconf
+		nm_spawn_process(RESOLV_CONF_PATH " -d NetworkManager");
+		cmd = g_strconcat (RESOLV_CONF_PATH " -a nm1 ",
+					 tmp_resolv_conf,
+					 NULL);

Don't you think, we should use the real interface name here?
(And is that a typo: Networkmanager vs. nm1?)

I haven't actually tested the code yet.

Cheers,
Michael
--
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?

Attachment: signature.asc
Description: OpenPGP digital signature



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]