Re: [PATCH v3 5/8] linux/log2.h: Add 64-bit roundup/down_pow_of_two64() variants

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] log2: 64-bit roundup/down_pow_of_two64() truncates and UBs on 32-bit archs
--

commit 56bd761df1693290739ce316beb7a19a30f72c63
Author: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
linux/log2.h: Add 64-bit roundup/down_pow_of_two64() variants

This commit introduces new 64-bit variants of the power-of-two rounding
functions and macros to prevent unexpected truncation of 64-bit inputs
that occurs with the existing helpers.

[Severity: High]
In include/linux/log2.h, the newly added 64-bit functions
__roundup_pow_of_two64 at line 69, __rounddown_pow_of_two64 at line 89,
and their macro equivalents roundup_pow_of_two64 at line 198 and
rounddown_pow_of_two64 at line 231 all use 1UL as the base for their bit
shift operations.

Since these functions and macros are explicitly designed to process u64
values, can this cause undefined behavior and truncation on 32-bit
architectures?

On platforms where sizeof(unsigned long) is 4, shifting 1UL by 32 or
greater will result in undefined behavior, and the upper 32 bits will be
truncated before the result is promoted to u64.

Would it be safer to use 1ULL as the shift base to ensure correct 64-bit
math across all architectures?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013130948.6231-1-ilpo.jarvinen@linux.intel.com?part=5