QNX Neutrino: exponential backoff when fork/spawn needs a retry

[flba-eb]

Fixes #108594: When retrying, sleep with an exponential duration. When sleep duration is lower than minimum possible sleeping time, yield instead (this will not be often due to the exponential increase of duration).

Minimum possible sleeping time is determined using libc::clock_getres but only when spawn/fork failed the first time in a request. This is cached using a LazyLock.

CC @gh-tr

r? @workingjubilee
@rustbot label +O-neutrino

[rustbot]

Hey! It looks like you've submitted a new PR for the library teams!

If this PR contains changes to any rust-lang/rust public library APIs then please comment with @rustbot label +T-libs-api -T-libs to tag it appropriately. If this PR contains changes to any unstable APIs please edit the PR description to add a link to the relevant API Change Proposal or create one if you haven't already. If you're unsure where your change falls no worries, just leave it as is and the reviewer will take a look and make a decision to forward on if necessary.

Examples of T-libs-api changes:

• Stabilizing library features
• Introducing insta-stable changes such as new implementations of existing stable traits on existing stable types
• Introducing new or changing existing unstable library APIs (excluding permanently unstable features / features without a tracking issue)
• Changing public documentation in ways that create new stability guarantees
• Changing observable runtime behavior of library APIs

[workingjubilee]

Sorry about the long turnaround time, April was apparently not great for thinking.

[workingjubilee]

I would like it if we still bailed for ErrorKind::WouldBlock or ErrorKind::Interrupted after a while, probably if we're getting up to actual seconds between yields.

[workingjubilee]

@rustbot author

[flba-eb]

Thank you very much @workingjubilee for the good feedback! Returning ErrorKind::WouldBlock instead of sleeping for seconds sounds like a good idea.

[flba-eb]

Requested changes have been implemented but should be tested before merging them. Branch is rebased; requested changes are in a separate commit to make the review easier.

[flba-eb]

r? @workingjubilee
Requested changes are done and tested on QNX Neutrino.

[rustbot]

Could not assign reviewer from: workingjubilee.
User(s) workingjubilee are either the PR author or are already assigned, and there are no other candidates.
Use r? to specify someone else to assign.

[flba-eb]

@rustbot review

[flba-eb]

@workingjubilee What do you think about giving up after one second? Does it make sense to document this somewhere as QNX Neutrino specific behavior? Where?

[workingjubilee]

Yeah, 1 second as the upper bar seems fine. If you wish to document it in the platform support (for now? maybe we should have a "platform quirks" page somewhere...) page, then go ahead. It's an unlikely circumstance, even under kernel stress, that I think having this surface in Rust programs as an unhandled "glitch" is fine, i.e. real programs usually won't hit this. However, I feel I should note that EAGAIN is actually specifically defined as a plausible POSIX error for posix_spawn due to it including "all the reasons fork/exec would fail", and fork can EAGAIN.

[workingjubilee]

Given that, it might actually make more sense to document this as a possible error for all platforms? But any changes for documenting spawn's errors for all platforms should be in another PR, and you might want to document it as specifically likely for Neutrino anyways.

[flba-eb]

Yes I think this should be documented in another PR. With this patch I wasn't able at all to reproduce the issue anymore (before the ui test suit always had 4-10 failures). As soon as we're sleeping the problem basically goes away (even when reducing the give-up threshold to milliseconds). After thinking about it, I would not even mention it in nto-qnx.md anymore. Would you agree?

[workingjubilee]

Sure! Then let's go with this.

@bors r+

[bors]

📌 Commit 716cc5a has been approved by workingjubilee

It is now in the queue for this repository.

[bors]

⌛ Testing commit 716cc5a with merge 1e17cef...

[bors]

☀️ Test successful - checks-actions
Approved by: workingjubilee
Pushing 1e17cef to master...

[rust-timer]

Finished benchmarking commit (1e17cef): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.2%	[2.2%, 2.2%]	1
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-	-	0

Cycles

Results

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.3%	[-4.3%, -1.3%]	8
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-3.3%	[-4.3%, -1.3%]	8

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 647.289s -> 647.076s (-0.03%)