fix(s2n-quic-transport): only allow GSO of MTU-sized packets #1613
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
camshaft
force-pushed
the
camshaft/gso-fixes
branch
from
7db775a
to
305edf3
Compare
| @@ -34,7 +34,7 @@ impl Snapshot { | |||
| assert!(rss < 30_000, "{rss}"); | |||
| } | |||
| "post-close" => { | |||
| assert_eq!(rss, 0, "{rss}"); | |||
| assert!(rss < 128, "{rss}"); | |||
There was a problem hiding this comment.
why did this change?
There was a problem hiding this comment.
I think it has to do with one of our dependencies. It's failing for me in other PRs as well so it doesn't seem related here:
https://github.com/aws/s2n-quic/actions/runs/4028993317/jobs/6926443934
| }); | ||
|
|
||
| if transmit_result.is_ok() { | ||
| if queue.has_capacity() { |
There was a problem hiding this comment.
The only error I see that was coming out of this previously was ConnectionOnTransmitError::NoDatagram, just wanted to check that its ok to continue iteration in that case
There was a problem hiding this comment.
Yeah the idea is to not rely on if the connection decided to transmit something or not. Instead we want to know what the actual capacity of the queue is and make the decision that way.
Co-authored-by: Wesley Rosenblum <[email protected]>
WesleyRosenblum
approved these changes
Jan 27, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description of changes:
I noticed an issue when determining if a transmission can be included as a GSO segment or not. The current logic is to allow any packet sizes to be considered for GSO, as long we aren't MTU probing. This isn't really ideal, since a previous transmission to the same tuple could have been a small ACK packet. This means we're artificially clamping MTU to the small ACK size, rather than allowing the connection to transmit a full MTU.
This change, instead, only allows GSO on transmissions if the segment size is
>=than the path's MTU for the current transmission mode. This results in more connections transmitting correctly, rather than bailing since we may end up deciding to not write a packet if the segment size is too small (initial packets, etc).Call-outs:
In experimentation, I also noticed the socket buffer queue sizes were somewhat small and will many concurrent connections I was filling it up a lot. This PR includes a default that is double the previous, as well as an unstable configuration of the value via environment variables.
I'm going to be revamping the IO implementation in the next coming weeks, and this environment variable will likely be gone after I'm done.
Testing:
I wrote a handshake fuzzer for this test. I'm planning on submitting a PR and put it under
/toolslater.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.