netdev - Re: [RFC] iproute2: gracefully exit from rtnl

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 1 Sep 2009 15:59:55 -0700
From:	Stephen Hemminger <shemminger@...tta.com>
To:	David Ward <david.ward@...mit.edu>
Cc:	netdev@...r.kernel.org, David Ward <david.ward@...mit.edu>
Subject: Re: [RFC] iproute2: gracefully exit from rtnl_listen()

On Tue,  1 Sep 2009 17:41:26 -0400
David Ward <david.ward@...mit.edu> wrote:

> rtnl_listen() (in iproute2's libnetlink) allows a userspace application to
> monitor an RTNETLINK multicast group, so that the application can react to
> changes in the kernel's routing table, neighbor cache, etc. as they occur.
> (This is in contrast with using RTNETLINK to solicit the kernel for a full
> copy of the routing table or neighbor cache at a specific point in time.)
> 
> However rtnl_listen() creates an infinite loop, which does not break in the
> absence of errors, so there does not seem to be any way to gracefully exit
> from it. Existing applications that call rtnl_listen(), such as rtmon, break
> from this loop by terminating the entire application. There should be some
> mechanism for stopping rtnl_listen() and continuing program execution.
> 
> If you assume that rtnl_listen() will only need to be stopped because the
> application has received a signal (such as SIGINT), then the application's
> signal handler can call rtnl_close() to close the RTNETLINK socket.
> Afterwards, with this patch, rtnl_listen() can detect that the socket was
> closed during the interrupt and exit.
> 
> I am offering this as a starting point for consideration, but I feel that
> this is at best a poor solution. What is a better way to fix rtnl_listen()
> (and similar libnetlink functions for consistency)?
> ---
>  lib/libnetlink.c |   15 ++++++++++++---
>  1 files changed, 12 insertions(+), 3 deletions(-)

Good idea, but why not just change the while(1) loop?

diff --git a/lib/libnetlink.c b/lib/libnetlink.c
index b68e2fd..cecfa34 100644
--- a/lib/libnetlink.c
+++ b/lib/libnetlink.c
@@ -188,7 +188,7 @@ int rtnl_dump_filter(struct rtnl_handle *rth,
 	char buf[16384];
 
 	iov.iov_base = buf;
-	while (1) {
+	while (rtnl->fd >= 0) {
 		int status;
 		struct nlmsghdr *h;
 
@@ -200,12 +200,12 @@ int rtnl_dump_filter(struct rtnl_handle *rth,
 				continue;
 			fprintf(stderr, "netlink receive error %s (%d)\n",
 				strerror(errno), errno);
-			return -1;
+			break;
 		}
 
 		if (status == 0) {
 			fprintf(stderr, "EOF on netlink\n");
-			return -1;
+			break;
 		}
 
 		h = (struct nlmsghdr*)buf;
@@ -251,6 +251,8 @@ skip_it:
 			exit(1);
 		}
 	}
+
+	return -1;
 }
 
 int rtnl_talk(struct rtnl_handle *rtnl, struct nlmsghdr *n, pid_t peer,


--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html