quinn-udp: retry GSO batch after disabling offload#2724
Open
kixelated wants to merge 1 commit into
Open
Conversation
kixelated
requested review from
Ralith,
djc,
gretchenfrage,
mxinden and
thomaseizinger
as code owners
July 9, 2026 21:06
kixelated
changed the title
kixelated
force-pushed
the
codex/gso-disable-retry
branch
from
July 9, 2026 21:12
608e04a to
565fcd6
Compare
Ralith
requested changes
Jul 10, 2026
Collaborator
|
Very much in favor. Thanks for working on it. We (Firefox) ran into the same issue. We currently re-send within Firefox. Instead re-sending in quinn-udp would be much cleanre. |
kixelated
force-pushed
the
codex/gso-disable-retry
branch
2 times, most recently
from
July 12, 2026 14:46
296845f to
adc90b9
Compare
thomaseizinger
requested changes
Jul 13, 2026
thomaseizinger
left a comment
thomaseizinger
left a comment
Collaborator
There was a problem hiding this comment.
Concept ACK, a few nits
| state.max_gso_segments.store(1, Ordering::Relaxed); | ||
|
|
||
| // Immediately retry this batch as individual datagrams rather than waiting | ||
| // for the QUIC layer to time out and retransmit. These sends run with GSO |
Collaborator
There was a problem hiding this comment.
Nit: Refer to "upper" layer here.
| #[allow(unused_variables)] // only used on Linux | ||
| state: &UdpSocketState, | ||
| io: SockRef<'_>, | ||
| io: &SockRef<'_>, |
Collaborator
There was a problem hiding this comment.
We can avoid this by re-borrowing the ref instead:
let io = SockRef::from(&*io);
Author
There was a problem hiding this comment.
Yeah I opted for & because it's feels less magic, but IDK.
kixelated
force-pushed
the
codex/gso-disable-retry
branch
from
July 15, 2026 17:16
adc90b9 to
3535538
Compare
thomaseizinger
approved these changes
Jul 16, 2026
thomaseizinger
requested changes
Jul 16, 2026
| let segment_size = transmit | ||
| .segment_size | ||
| .unwrap_or(transmit.contents.len()) | ||
| .max(1); |
Collaborator
There was a problem hiding this comment.
What is that for?
If contents.len is 0, we can just return Ok(()), right? And if segment_size is 0, we should probably fail?
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
kixelated
force-pushed
the
codex/gso-disable-retry
branch
from
July 17, 2026 03:50
3535538 to
c6ad76d
Compare
Collaborator
|
Note that this also gets fixed in #2748 which may be preferable to merge because it also fixes another issue. |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Fixes #2399
The first batch that causes a GSO failure is often the initial packet, which takes forever to retry sending at the QUIC layer because of the default RTT. This should improves connection time dramatically on some Linux devices (apparently Android).