[runtime/iouring] account for linked timeout op space in the SQ #1465
Conversation
|
Should we internally double the ring size when |
runtime/src/iouring/mod.rs
Outdated
| // Try to fill the submission queue with incoming work. | ||
| // Stop if we are at the max number of processing work. | ||
| while waiters.len() < cfg.size as usize { | ||
| // NOTE: When op_timeout is set, each operation uses 2 SQ entries |
That's a good question. I think doubling makes the most sense. |
Yeah that makes the most sense to me too. Updated the PR. |
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## main #1465 +/- ##
==========================================
- Coverage 91.69% 91.69% -0.01%
==========================================
Files 278 278
Lines 70326 70326
==========================================
- Hits 64487 64486 -1
- Misses 5839 5840 +1 see 2 files with indirect coverage changes Continue to review full report in Codecov by Sentry.
🚀 New features to boost your workflow:
|
Timeout operations were not acounted for when submitting to the ring, this could lead to capacity overflow. Contains a regression test to showcase the problem (panics before the last commit).