Correctly handle some extreme edge cases in the ratelimiter implementation #3706
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
This is due to the constructor attempting to convert from milliseconds to nanoseconds without checking whether the nanosecond value fits into a u64. Signed-off-by: Patrick Roy <roypat@amazon.co.uk>
Due to rounding down instead of up when computing the time that it would take to generate "x" tokens, the fractional part of a nanosecond is used "twice" once per call to auto_replenish. This can only happen if the time since the last auto_replenish call is enough to generate at least one token, e.g. (refill_time / bucket_size) milliseconds. To accumulate enough "double used" time to generate an extra token, at least (refill_time * 1_000_000 / bucket_size) calls to auto_replenish are needed, with the calls timed in a way that causes the "double used" time to be as close as possible to 1ns. Note that since time resolution is at the nanosecond level, at least one nanosecond has to pass between these calls. For example, if a TokenBucket is setup for 1GB/s, we could allow one extra byte of traffic every (1000 / 1_000_000_000) * 1_000_000 = 1ms, e.g. excess throughput of 1KB/s (or 0.0001% above the target) Additionally, in extremely unrealistic scenarios (either refill_time of thousands of years or the guest waiting thousands of years), an integer overflow in a multiplication in auto_replenish can leads to less tokens being replenished than should be. Signed-off-by: Patrick Roy <roypat@amazon.co.uk>
Due to a missing clamping overation, it used to be able to increment one_time_burst beyond initial_one_time_burst. Additionally, integer overflow during addition could cause the budget/one_time_burst to actually decrease if they were close to u64::MAX priot to calling force_replenish (a very unrealistic scenario) Signed-off-by: Patrick Roy <roypat@amazon.co.uk>
roypat
added
the
Status: Awaiting review
bchalios
reviewed
May 22, 2023
bchalios
approved these changes
May 22, 2023
Spoke offline with @roypat. We will be adding tests for these cases with subsequent contributions. |
mattschlebusch
approved these changes
May 24, 2023
3 tasks
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.
Changes
This series of commits fixes some edge cases in the ratelimiter implementation related to insanely large timespans (on the magnitude of hundreds of thousands of years)
License Acceptance
By submitting this pull request, I confirm that my contribution is made under
the terms of the Apache 2.0 license. For more information on following
Developer Certificate of Origin and signing off your commits, please check
CONTRIBUTING.md.PR Checklist
CHANGELOG.md.TODOs link to an issue.rust-vmm.