On Wed, Jan 16 2019, Jeremie Courreges-Anglas <j...@wxcvbn.org> wrote: > On Tue, Jan 15 2019, Brad Smith <b...@comstyle.com> wrote: >> Hi, >> >> Looking for some testing of dnsmasq 2.80.
This was committed since it contains interesting changes. >> If I'm not mistaken >> the issue with IP_SENDSRCADDR was resolved. > > I just used dnsmasq -d, and nope, the issue that plagued dnsmasq is > still here. The kernel problem fixed earlier in 2018 was an edge case > not directly related to dnsmasq. > > I had a diff to unbreak dnsmasq using IP_SENDSRCADDR that I have already > shown, not sure if I got any feedback about it. Updated version below. > > The root cause is that sendmsg(2) fails with EINVAL because of > assumptions about the cmsg API that our kernel rejects. The surrounding > dnsmasq code ignores this error because of an edge case on Linux. If anyone uses dnsmasq and can double check that dnsmasq-2.80 still works properly without specifying -z/--bind-interfaces, that would be great. -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE