Specialize sleep_until implementation for unix (except mac) #141829

dvdsk · 2025-05-31T19:38:39Z

related tracking issue: #113752
Supersedes #118480 for the reasons see: #113752 (comment)

Replaces the generic catch all implementation with target_os specific ones for: linux/netbsd/freebsd/android/solaris/illumos etc. Other platforms like wasi, macos/ios/tvos/watchos and windows will follow in later separate PR's (once this is merged).

rustbot · 2025-05-31T19:38:43Z

r? @tgross35

rustbot has assigned @tgross35.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

dvdsk · 2025-05-31T19:39:32Z

r? @cuviper

dvdsk · 2025-05-31T22:38:40Z

This can be done without touching all pal's, ill be doing that tomorrow, now its bed time 😴

edit: I was mistaken, tidy does not allow #[cfg(...)] in strc/thread/mod.rs probably for a good reason. Therefore we need a trivial not platform specific sleep_until in every pal that does not have a specialized one.

bors · 2025-07-03T02:32:55Z

⌛ Testing commit dd0a1e6 with merge 94817f8...

Specialize sleep_until implementation for unix (except mac) related tracking issue: #113752 Supersedes #118480 for the reasons see: #113752 (comment) Replaces the generic catch all implementation with target_os specific ones for: linux/netbsd/freebsd/android/solaris/illumos etc. Other platforms like wasi, macos/ios/tvos/watchos and windows will follow in later separate PR's (once this is merged).

bors · 2025-07-03T04:15:13Z

💔 Test failed - checks-actions

dvdsk · 2025-07-03T10:27:18Z

The job x86_64-gnu-aux failed! Check out the build log: (web) (plain enhanced) (plain)
Click to see the possible cause of the failure (guessed by this bot)

Miri does not understand clock_nanosleep. To fix that I think I need to add a shim for it to miri/src/shims/time.rs. Seems doable, I'll see if I can adjust whats done for nanosleep. I'll get on that now.

rustbot · 2025-07-03T15:28:24Z

The Miri subtree was changed

cc @rust-lang/miri

dvdsk · 2025-07-03T15:34:04Z

First time adding anything to MIRI.

It does not make sense to me to support all argument combinations of clock_nanosleep. Instead I have implemented those needed for sleep_until and those trivial to add.
Is it correct to panic when an unimplemented argument combination is passed?
I am not sure about my usage of callback!, I just copied what nanosleep had without fully understanding it. I've added a comment to reflect that. Please review that carefully.

It was fun to see some MIRI code but I understand if the additional code to MIRI for supporting clock_nanosleep is not worth it at this stage.

Let me also thank you all again for your reviews and mentoring me on how to get this PR up to spec.

RalfJung

Thanks for working on the Miri shims! However, they don't quite match what we'd usually expect. If you need more Miri-specific help, please ask on Zulip.

src/tools/miri/src/shims/unix/foreign_items.rs

src/tools/miri/src/shims/time.rs

RalfJung · 2025-07-03T15:42:31Z

src/tools/miri/src/shims/time.rs

Please also add a test in tests/pass-dep/libc/libc-time.rs for the new shim. In particular, it seems there are two codepaths (relative and absolute?), they should both be tested.

Also, please add a call to sleep_until in tests/pass/shims/time.rs.

RalfJung · 2025-07-03T15:47:13Z

src/tools/miri/src/shims/time.rs

+        &mut self,
+        clock_id: &OpTy<'tcx>,
+        flags: &OpTy<'tcx>,
+        req_op: &OpTy<'tcx>,


Why do you call this req_op? It seems to be the timeout?

I agree the name could be better. I've mirrored what was there for nanosleep as the argument for clock_nanosleep has a similar function and they are called similar in the man pages.

nanosleep:

int nanosleep(const struct timespec *req, struct timespec *rem);

clock_nanosleep:

int clock_nanosleep(clockid_t clockid, int flags, const struct timespec *request, struct timespec *remain);

I don't know why they call it req/request. Timeout doesn't quite work as its either a deadline or a timeout depending on if the TIMER_ABSTIME flag is set.

Maybe deadline_or_timeout would be a better name.

and they are called similar in the man pages.

Hm, odd, in my manpages it's

int nanosleep(const struct timespec *duration, struct timespec *_Nullable rem);

int clock_nanosleep(clockid_t clockid, int flags, const struct timespec *t, struct timespec *_Nullable remain);

My OS release is kinda old (Pop!_OS seems to be stuck on 22.04 until they finish cosmic).

I'm going to rename the the argument to instant_or_timeout, though those words are not used in the man pages it best describes what the argument is for.

edit: let me know if you disagree

I would suggest naming it after the type, timespec.

thats better 👍

(unresolved since that rename didn't happen yet)

RalfJung · 2025-07-03T15:56:01Z

(The build fails on Windows because you're using libc.)

Co-authored-by: Jonas Böttiger <jonasboettiger@icloud.com> - use the internal platform specific Instant::CLOCK_ID - skip allow(unused) on on platform that uses it such that it can not become dead code - sleep_until: remove leftover links

- In contradiction to nanosleep clock_nanosleep returns the error directly and does not require a call to `os::errno()`. - The last argument to clock_nanosleep can be NULL removing the need for mutating the timespec. - Missed an `allow(unused)` that could be made conditional.

This is intended to support the std's new sleep_until. Only the clocks REALTIME and MONOTONIC are supported. The first because it was trivial to add the second as its needed for sleep_until. Only passing no flags or passing only TIMER_ABSTIME is supported. If an unsupported flags or clocks are passed this implementation panics.

src/tools/miri/src/shims/time.rs

RalfJung · 2025-07-03T19:31:42Z

src/tools/miri/src/shims/time.rs

+        let this = self.eval_context_mut();
+
+        let clockid_t_size = this.libc_ty_layout("clockid_t").size;


Suggested change

let this = self.eval_context_mut();

let clockid_t_size = this.libc_ty_layout("clockid_t").size;

let this = self.eval_context_mut();

let clockid_t_size = this.libc_ty_layout("clockid_t").size;

making code more readable by having empty lines between semantically relevant blocks, like paragraphs in a text

Just want to make sure I understand correctly. You would like two empty lines between semantically relevant blocks?

That is not what I see for other shims in time.rs.

That's not what the diff does. What it does is it moves the clockid_t_size out of the block that should consist of exactly one line per argument, converting that argument into a more typed representation.

Oh I read the diff wrong.

Its getting late, thank you very much for the quick and extensive feedback! I'm gonna call it a day here and add the tests tomorrow when I've got some more sleep 💤

Have great day!

src/tools/miri/src/shims/time.rs

dvdsk · 2025-07-03T19:57:34Z

rebased to:

fix commit formatting unrelated MIRI code.
fix spell errors for months ago without adding new useless commit

edit: I just realized rebasing all the time makes this rather hard to review. I'll rebase/squash at the end, for now I'll focus on short descriptive commits.

RalfJung · 2025-07-03T20:25:45Z

src/tools/miri/src/shims/time.rs

@@ -396,14 +393,14 @@ pub trait EvalContextExt<'tcx>: crate::MiriInterpCxExt<'tcx> {
            // No flags set, the timespec should be interperted as a duration
            // to sleep for
            TimeoutAnchor::Relative
-        } else if flag == this.eval_libc("TIMER_ABSTIME").to_int(int_size) {
+        } else if flags == this.eval_libc("TIMER_ABSTIME").to_i32()? {


Suggested change

} else if flags == this.eval_libc("TIMER_ABSTIME").to_i32()? {

} else if flags == this.eval_libc_i32("TIMER_ABSTIME") {

rustbot assigned tgross35 May 31, 2025

rustbot assigned cuviper and unassigned tgross35 May 31, 2025

dvdsk mentioned this pull request May 31, 2025

Specialize sleep_until implementation #118480

Closed

workingjubilee changed the title ~~Specialize sleep_until implementation for unix (expect mac)~~ May 31, 2025

This comment has been minimized.

Sign in to view

dvdsk force-pushed the sleep_until_linux branch from 8010ec8 to 295e8d1 Compare May 31, 2025 19:54

This comment has been minimized.

Sign in to view

dvdsk force-pushed the sleep_until_linux branch from 295e8d1 to 1e7d958 Compare May 31, 2025 20:04

This comment has been minimized.

Sign in to view

dvdsk force-pushed the sleep_until_linux branch from 1e7d958 to 406e32b Compare May 31, 2025 21:31

This comment has been minimized.

Sign in to view

dvdsk marked this pull request as draft May 31, 2025 22:37

rustbot added the O-wasm Target: WASM (WebAssembly), http://webassembly.org/ label Jun 1, 2025

dvdsk force-pushed the sleep_until_linux branch from 4bcfd27 to f657ec1 Compare June 1, 2025 07:08

This comment has been minimized.

Sign in to view

dvdsk force-pushed the sleep_until_linux branch from f657ec1 to 1127a50 Compare June 1, 2025 07:26

This comment has been minimized.

Sign in to view

dvdsk force-pushed the sleep_until_linux branch from 1127a50 to 60edd5a Compare June 1, 2025 07:35

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Jul 2, 2025

This comment has been minimized.

Sign in to view

bors added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Jul 3, 2025

RalfJung reviewed Jul 3, 2025

View reviewed changes

This comment has been minimized.

Sign in to view

dvdsk force-pushed the sleep_until_linux branch from a7c539f to f65530d Compare July 3, 2025 19:02

This comment has been minimized.

Sign in to view

dvdsk and others added 6 commits July 3, 2025 21:20

sleep_until: handle 32bit time_t targets

c1df306

sleep_until: improve maintainability

fd20d20

Co-authored-by: Jonas Böttiger <jonasboettiger@icloud.com> - use the internal platform specific Instant::CLOCK_ID - skip allow(unused) on on platform that uses it such that it can not become dead code - sleep_until: remove leftover links

fixes missing import for platform unsupported

e69a490

Adhere to Miri code standards, fixes using host constants/types

dcb3593

dvdsk force-pushed the sleep_until_linux branch from d91e6a9 to dcb3593 Compare July 3, 2025 19:20

This comment has been minimized.

Sign in to view

RalfJung reviewed Jul 3, 2025

View reviewed changes

dvdsk added 2 commits July 3, 2025 22:16

fix read_scalar/read_pointer

f07f99c

rename clock_nanosleep argument

7c039f4

RalfJung reviewed Jul 3, 2025

View reviewed changes

		let this = self.eval_context_mut();

		let clockid_t_size = this.libc_ty_layout("clockid_t").size;

	} else if flags == this.eval_libc("TIMER_ABSTIME").to_i32()? {
	} else if flags == this.eval_libc_i32("TIMER_ABSTIME") {

Specialize sleep_until implementation for unix (except mac) #141829

Are you sure you want to change the base?

Specialize sleep_until implementation for unix (except mac) #141829

Conversation

dvdsk commented May 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

rustbot commented May 31, 2025

dvdsk commented May 31, 2025

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

dvdsk commented May 31, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

This comment has been minimized.

This comment has been minimized.

bors commented Jul 3, 2025

This comment has been minimized.

bors commented Jul 3, 2025

dvdsk commented Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

rustbot commented Jul 3, 2025

dvdsk commented Jul 3, 2025

RalfJung left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dvdsk Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment has been minimized.

RalfJung commented Jul 3, 2025

This comment has been minimized.

This comment has been minimized.

Uh oh!

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dvdsk Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

dvdsk commented Jul 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

dvdsk commented May 31, 2025 •

edited

Loading

dvdsk commented May 31, 2025 •

edited

Loading

dvdsk commented Jul 3, 2025 •

edited

Loading

RalfJung left a comment •

edited

Loading

dvdsk Jul 3, 2025 •

edited

Loading

dvdsk Jul 3, 2025 •

edited

Loading

dvdsk commented Jul 3, 2025 •

edited

Loading