Set AI_ADDRCONFIG when making getaddrinfo(3) calls for outgoing conns #7295
Conversation
| @@ -140,4 +140,168 @@ def test_accept_multithread | |||
| server_threads.each(&:join) | |||
| end | |||
| end | |||
|
|
|||
| def test_ai_addrconfig | |||
There was a problem hiding this comment.
I wrote this test, but it's so complicated and environment dependent that I wonder if it's worth having at all - maybe I should delete this?
There was a problem hiding this comment.
I agree that it is too complicated.
I talked about this test with naruse.
He said this test should be retained but should be always ommitted.
There was a problem hiding this comment.
Thank you for merging this!
I see you committed the skip in abf192e. I agree this makes sense, even without the skip this test wouldn't actually do anything on CI anyway - unshare(2) won't work without root privileges and presumably we don't run the tests as root in Github Actions or rubyci.org. So better to just skip and be clear about it.
|
@akr hey, I was wondering if you could have a look at this? 🙏 thank you! |
When making an outgoing TCP or UDP connection, set AI_ADDRCONFIG in the hints we send to getaddrinfo(3) (if supported). This will prompt the resolver to _NOT_ issue A or AAAA queries if the system does not actually have an IPv4 or IPv6 address (respectively). This makes outgoing connections marginally more efficient on non-dual-stack systems, since we don't have to try connecting to an address which can't possibly work. More importantly, however, this works around a race condition present in some older versions of glibc on aarch64 where it could accidently send the two outgoing DNS queries with the same DNS txnid, and get confused when receiving the responses. This manifests as outgoing connections sometimes taking 5 seconds (the DNS timeout before retry) to be made. Fixes #19144
mame
force-pushed
the
ktsanaktsidis/ai_addrconfig
branch
from
4fec80c
to
673ed41
Compare
| @@ -140,4 +140,168 @@ def test_accept_multithread | |||
| server_threads.each(&:join) | |||
| end | |||
| end | |||
|
|
|||
| def test_ai_addrconfig | |||
There was a problem hiding this comment.
I agree that it is too complicated.
I talked about this test with naruse.
He said this test should be retained but should be always ommitted.
When making an outgoing TCP or UDP connection, set AI_ADDRCONFIG in the hints we send to getaddrinfo(3) (if supported). This will prompt the resolver to NOT issue A or AAAA queries if the system does not actually have an IPv4 or IPv6 address (respectively).
This makes outgoing connections marginally more efficient on non-dual-stack systems, since we don't have to try connecting to an address which can't possibly work.
More importantly, however, this works around a race condition present in some older versions of glibc on aarch64 where it could accidently send the two outgoing DNS queries with the same DNS txnid, and get confused when receiving the responses. This manifests as outgoing connections sometimes taking 5 seconds (the DNS timeout before retry) to be made.
Fixes #19144
https://bugs.ruby-lang.org/issues/19144