Re: [PATCH] birdc: Do not execute cmd on non interactive help request
On Thu, Dec 28, 2023 at 12:37:26PM +0100, Inrin via Bird-users wrote: > Hi! > > A pal of mine noticed that `echo ? | birdc` executes the command > after showing it's help text is shown. I later confirmed that `birdcl` > in > comparison to `birdc` does not show this behaviour. > > Since this is a non intuitive side effect on birdc's side, here a > proposed patch to align birdc and birdcl's behaviour. > > Please advise if fixing it on another part might be preferred. > I chose birdc.c, to minimize the changes needed. Thanks, merged. Sorry for the delay, i missed / forgot about the patch. -- Elen sila lumenn' omentielvo Ondrej 'Santiago' Zajicek (email: santi...@crfreenet.org) "To err is human -- to blame it on a computer is even more so."
Re: Strict Bind usability
On Sun, Jan 14, 2024 at 10:50:18PM +0100, Sebastian Hahn wrote: > Hi list, > > I've been using strict bind for bgp, but there might be a lack in my > understanding or I have a proposal for a new feature. > Strict bind seems to be a bit more restrictive than I'd like (and than I > think it is necessary): > > I would like to use dynamic bgp (using a range for neighbour) with > auto-configured LL IPv6 adresses. This doesn't work, of course, if I need > to statically configure the source address, which I am forced to with the > strict bind option. I am wondering, is this restriction really needed? Hi You are right, this is something that could be easily changed. Will look at it. > In addition, it would be great if a strict bind option could be added for the > babel protocol. Babel (and OSPF, RIP) already uses one socket per interface. Such socket is not bound to specific IP address using bind(), but is bound to the specific interface using setsockopt(SO_BINDTODEVICE), at least in Linux. We cannot use bind() with local IP address for multicast sockets, as they would reject multicast packets (as they have different dst address as one the socket is bound). Anyway, it should not be necessary. AFAIK, setups that require strict bind in BGP should work out of the box in Babel. Or what specific setup does not work for you? > Also let me say Thanks for bird, it's a pleasure to use! Thanks! -- Elen sila lumenn' omentielvo Ondrej 'Santiago' Zajicek (email: santi...@crfreenet.org) "To err is human -- to blame it on a computer is even more so."
Re: Strict Bind usability
Hi Ondrej, > On 6. Feb 2024, at 19:33, Ondrej Zajicek wrote: >> >> I would like to use dynamic bgp (using a range for neighbour) with >> auto-configured LL IPv6 adresses. This doesn't work, of course, if I need >> to statically configure the source address, which I am forced to with the >> strict bind option. I am wondering, is this restriction really needed? > > Hi > > You are right, this is something that could be easily changed. Will look at > it. That would be awesome :) >> In addition, it would be great if a strict bind option could be added for >> the babel protocol. > > Babel (and OSPF, RIP) already uses one socket per interface. Such socket > is not bound to specific IP address using bind(), but is bound to the > specific interface using setsockopt(SO_BINDTODEVICE), at least in Linux. > > We cannot use bind() with local IP address for multicast sockets, as they > would reject multicast packets (as they have different dst address as one > the socket is bound). > > Anyway, it should not be necessary. AFAIK, setups that require strict > bind in BGP should work out of the box in Babel. Or what specific setup > does not work for you? Oh, I think I was confused by the output of netstat. Using ss instead, I can see that bird correctly binds to just the devices specified in the configuration. I had previously assumed it was binding to all interfaces. Very sorry for this basic mistake. Thanks! Sebastian
