Re: prevent bgpd from starting when control socket already used

2018-11-19 Thread Remi Locherer
On Mon, Nov 19, 2018 at 09:45:55AM +0100, Claudio Jeker wrote: > On Sun, Nov 18, 2018 at 11:40:40PM +0100, Remi Locherer wrote: > > Hi, > > > > from the discussion I understand nobody rejects the functionality. > > > > To ease the review here again the diff (with incorporated feedback from > > an

Re: prevent bgpd from starting when control socket already used

2018-11-19 Thread Claudio Jeker
On Sun, Nov 18, 2018 at 11:40:40PM +0100, Remi Locherer wrote: > Hi, > > from the discussion I understand nobody rejects the functionality. > > To ease the review here again the diff (with incorporated feedback from > anton@ (redundant parens)). > > Any comments or OKs? Not sure about this, sin

Re: prevent bgpd from starting when control socket already used

2018-11-18 Thread Remi Locherer
Hi, from the discussion I understand nobody rejects the functionality. To ease the review here again the diff (with incorporated feedback from anton@ (redundant parens)). Any comments or OKs? Remi cvs diff: Diffing . Index: bgpd.c ==

Re: prevent bgpd from starting when control socket already used

2018-11-12 Thread Sebastian Benoit
Stuart Henderson(s...@spacehopper.org) on 2018.11.11 21:55:19 +: > On 2018/11/11 22:45, Job Snijders wrote: > > Shouldnt we already bomb out at the following? > > > > cannot bind to 0.0.0.0:179: Address already in use > > cannot bind to [::]:179: Address already in use > > > > In any regard,

Re: prevent bgpd from starting when control socket already used

2018-11-12 Thread Remi Locherer
On Mon, Nov 12, 2018 at 08:12:37AM +0100, Claudio Jeker wrote: > On Sun, Nov 11, 2018 at 04:40:54PM -0700, Theo de Raadt wrote: > > Makes sense to me, I suppose. > > > > Isn't another approach to swap the opening of the sockets? > > > > Or why does failure to control :179 sockets not stop startup

Re: prevent bgpd from starting when control socket already used

2018-11-11 Thread Claudio Jeker
On Sun, Nov 11, 2018 at 04:40:54PM -0700, Theo de Raadt wrote: > Makes sense to me, I suppose. > > Isn't another approach to swap the opening of the sockets? > > Or why does failure to control :179 sockets not stop startup? Because you can change them during reload and if they are incorrect for

Re: prevent bgpd from starting when control socket already used

2018-11-11 Thread Theo de Raadt
Makes sense to me, I suppose. Isn't another approach to swap the opening of the sockets? Or why does failure to control :179 sockets not stop startup? > I heard from two devs that started a 2nd bgpd by accident (forgot -n for > a config check) which then caused downtime. > > Below diff adds a c

Re: prevent bgpd from starting when control socket already used

2018-11-11 Thread Stuart Henderson
On 2018/11/11 22:45, Job Snijders wrote: > Shouldnt we already bomb out at the following? > > cannot bind to 0.0.0.0:179: Address already in use > cannot bind to [::]:179: Address already in use > > In any regard, I agree with the functionality proposed. No strong opinion > on the diff itself. S

Re: prevent bgpd from starting when control socket already used

2018-11-11 Thread Job Snijders
Shouldnt we already bomb out at the following? cannot bind to 0.0.0.0:179: Address already in use cannot bind to [::]:179: Address already in use In any regard, I agree with the functionality proposed. No strong opinion on the diff itself. Kind regards, Job On Sun, Nov 11, 2018 at 22:35 Remi L

prevent bgpd from starting when control socket already used

2018-11-11 Thread Remi Locherer
Hi, I heard from two devs that started a 2nd bgpd by accident (forgot -n for a config check) which then caused downtime. Below diff adds a check to bgpd similar to the one we have now in ospfd and ospf6d: if another process is listening on the control socket bgpd exits. The situation is a bit di