Comment 0 for bug 1849658

Revision history for this message
Dan Streetman (ddstreet) wrote :

[impact]

for DNS UDP replies larger than 512 bytes, fallback to TCP is used. For example 'host toomany.ddstreet.org'.

Due to a bug in resolved in refcounting DNS stream types, the refcount underflows for type 0 streams (which resolved uses to talk to upstream nameservers), resulting in resolved being unable to fallback to TCP to handle truncated UDP replies.

[test case]

ubuntu@sf247344-upstream:~$ dig +noanswer +noedns toomany.ddstreet.org
;; Truncated, retrying in TCP mode.

; <<>> DiG 9.11.3-1ubuntu1.9-Ubuntu <<>> +noanswer +noedns toomany.ddstreet.org
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 2683
;; flags: qr rd ra; QUERY: 1, ANSWER: 40, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;toomany.ddstreet.org. IN A

;; Query time: 0 msec
;; SERVER: 127.0.0.53#53(127.0.0.53)
;; WHEN: Thu Oct 24 11:40:29 UTC 2019
;; MSG SIZE rcvd: 678

ubuntu@sf247344-upstream:~$ sudo resolvectl flush-caches
ubuntu@sf247344-upstream:~$ dig +noanswer +noedns toomany.ddstreet.org

; <<>> DiG 9.11.3-1ubuntu1.9-Ubuntu <<>> +noanswer +noedns toomany.ddstreet.org
;; global options: +cmd
;; connection timed out; no servers could be reached

[regression potential]

very low, as this only properly sets the stream type in the DnsStream object; any regression would be a failure to be able to use TCP for DNS requests or replies.

[other info]

https://github.com/systemd/systemd/pull/13838