mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH v3 0/3] rust: time: fix Delta arithmetic and add math helpers
@ 2026-10-17 11:11 chenhan
  2026-10-17 11:11 ` [PATCH v3 1/3] rust: time: rename the rem_nanos() divisor chenhan
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: chenhan @ 2026-10-17 11:11 UTC (permalink / raw)
  To: rust-for-linux
  Cc: a.hindborg, boqun, fujita.tomonori, frederic, lyude, tglx,
	anna-maria, jstultz, sboyd, ojeda, gary, bjorn3_gh, lossin,
	aliceryhl, tmgross, dakr, daniel.almeida, tamird, acourbot, work,
	miguel.ojeda.sandonis, tomo, georgeandrout13, linux-kernel

Hi,

This revision separates the Delta arithmetic fix from establishing the
math module, following the v2 discussion.

Patch 1 corrects the rem_nanos() parameter name and documentation, as
Miguel requested. Patch 2 adds the panic checks in the existing methods.
Patch 3 moves that implementation to public math::div64_s64() and
math::rem_s64() helpers and uses them from Delta.

For stable, patch 2 depends on patch 1; both carry stable Cc, and patch 1
is marked as a prerequisite. The math refactor is not marked for stable.
The first two patches can be taken without the third. Backport suitability
remains for the maintainers to decide; I have not audited all stable trees.

Each signed helper still has one production call site, in Delta. The
current pwm_th1520 conversions use unsigned arithmetic and are not users
of these helpers. Patch 3 follows Andreas' request to establish math.rs.
Tomonori, it adds no unsigned helpers or PWM changes; if your math module
lands first, I can rebase this patch onto it. The helpers are public so
they are available to driver crates as well as the kernel crate.

Changes since v2:
- Split the parameter correction, behavior fix and math extraction.
- Explain in SAFETY why the conditions hold: the checks above exclude
  zero and i64::MIN / -1, and &mut rem supplies the valid writable pointer.
- Drop the redundant issue-comment Link trailers, retaining Closes.
- Remove the unrelated "a number of" wording, C implementation details
  from public documentation, and the "even though ..." panic explanation.

No in-tree calls to Delta's division or rem_nanos() were found at the
base or rust-next 7c306f00835a. This covered tracked Rust files, named
calls and division expressions in Delta/Instant users, including elapsed()
results. Fixed-divisor time conversions, unsigned jiffies arithmetic and
generic bounded-integer operations are not extra users of these helpers.

Testing used a built-in Rust module whose init function passes black_box
operands to the actual Delta APIs (a and b are i64, d is i32):

    Delta::from_nanos(black_box(a)) / Delta::from_nanos(black_box(b))
    Delta::from_nanos(black_box(a)).rem_nanos(black_box(d)).as_nanos()

A boot-parameter selector chooses the test. One boot compares 26 valid
divisions and 21 valid remainders against literal expected values; four
separate boots exercise the invalid inputs because each panic stops its
test. Core expectations are:

    dividend    divisor    division    remainder (nanoseconds)
    10          0          panic       panic
    i64::MIN    -1         panic       panic
    10          3          3           1
    -10         3          -3          -1

Other valid cases cover signed extrema, i32::MIN divisors and full-width
i64 division operands. Patches 1-2 passed on x86-64 and ARMv7 QEMU with
CONFIG_RUST_OVERFLOW_CHECKS=y. The full series passed on both architectures
with that option enabled and disabled: 30 boots, 282 valid comparisons and
24 expected panics in total. Kernel builds and rustdoc passed on both
architectures. The test module is validation code, not part of this series.

Base: timekeeping-next 2ea0119f72dba597aa8a98cbdb72c564bfc5cb38.
Rechecked on October 17: this is still the timekeeping-next tip, and
neither it nor rust-next has a math module. When applying patch 3 to
rust-next 7c306f00835a, the new "pub mod mem;" line overlaps the module
registration hunk. Keep both modules, with math between maple_tree and
mem. Patches 1 and 2 apply there unchanged. Runtime testing above is on
the stated timekeeping-next base.

V2: https://lore.kernel.org/rust-for-linux/20261001042942.109012-1-chenhan0017.work@gmail.com/

Best regards,
chenhan

chenhan (3):
  rust: time: rename the rem_nanos() divisor
  rust: time: make Delta division and remainder fail consistently
  rust: math: add signed division and remainder helpers

 rust/kernel/lib.rs  |  1 +
 rust/kernel/math.rs | 64 +++++++++++++++++++++++++++++++++++++++++++++
 rust/kernel/time.rs | 46 +++++++++++---------------------
 3 files changed, 80 insertions(+), 31 deletions(-)
 create mode 100644 rust/kernel/math.rs


base-commit: 2ea0119f72dba597aa8a98cbdb72c564bfc5cb38
-- 
2.34.1


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2026-10-17 11:12 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-10-17 11:11 [PATCH v3 0/3] rust: time: fix Delta arithmetic and add math helpers chenhan
2026-10-17 11:11 ` [PATCH v3 1/3] rust: time: rename the rem_nanos() divisor chenhan
2026-10-17 11:11 ` [PATCH v3 2/3] rust: time: make Delta division and remainder fail consistently chenhan
2026-10-17 11:11 ` [PATCH v3 3/3] rust: math: add signed division and remainder helpers chenhan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®