Re: [PATCH v7 09/11] riscv: cpufeature: Introduce ISA bases bitmap and rva23u64 detection

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] riscv: RVA23U64 detection incorrectly requires legacy B extension
- [Medium] riscv: riscv_set_isa_bases couples independent profile evaluations
--

commit e0026c8e3b6f0de96cacfa903bced9ce7509616a
Author: Guodong Xu <guodong.xu@oss.qualcomm.com>

riscv: cpufeature: Introduce ISA bases bitmap and rva23u64 detection

Introduce a per-hart and host-wide bitmap of conformant ISA "bases", computed
at init time by riscv_init_isa_bases(). Register it as an arch_initcall to
ensure it is evaluated after core_initcall sets up PMLEN features. This allows
subsequent consumers like hwprobe and /proc/cpuinfo to read a consistent cache
of base profiles.

> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> --- a/arch/riscv/kernel/cpufeature.c
> +++ b/arch/riscv/kernel/cpufeature.c
[ ... ]
> +static void riscv_set_isa_bases(unsigned long *bases, const unsigned long *isa_bitmap)
> +{
> +	const unsigned long *isa = isa_bitmap ? isa_bitmap : riscv_isa;
> +	DECLARE_BITMAP(ext_mask, RISCV_ISA_EXT_MAX) = { 0 };

[Severity: Medium]
Since ext_mask is initialized once and never cleared, will this cause bits to
accumulate across independent profile checks?

For example, if a future orthogonal base is added later in this function, it
would inherit stale bits like RISCV_ISA_EXT_I and could falsely fail the
bitmap_subset() check.

> +
> +	/* IMA */
> +	__set_bit(RISCV_ISA_EXT_I, ext_mask);
> +	__set_bit(RISCV_ISA_EXT_M, ext_mask);
> +	__set_bit(RISCV_ISA_EXT_A, ext_mask);
> +
> +	if (!bitmap_subset(ext_mask, isa, RISCV_ISA_EXT_MAX))
> +		return;
> +
> +	set_bit(RISCV_ISA_BASE_IMA, bases);
> +
> +	/* RVA23U64 */
> +
> +	/* Supm with PMLEN=7 */
> +	if (!riscv_have_user_pmlen(7))
> +		return;

[Severity: Medium]
Will this early return prevent subsequent profile checks from being evaluated?

If riscv_have_user_pmlen(7) evaluates to false, the function aborts
completely. This seems like it will inadvertently skip the evaluation of any
future bases added after RVA23U64.

> +
> +	__set_bit(RISCV_ISA_EXT_F, ext_mask);
> +	__set_bit(RISCV_ISA_EXT_D, ext_mask);
> +	__set_bit(RISCV_ISA_EXT_C, ext_mask);
> +	__set_bit(RISCV_ISA_EXT_B, ext_mask);

[Severity: High]
Does this incorrectly require the legacy, unratified 'b' extension for
RVA23U64 detection?

When the kernel parses a DT containing the mandated individual Zba, Zbb, and
Zbs extensions but not the legacy 'b' extension, the isa bitmap will not
contain RISCV_ISA_EXT_B. 

Consequently, the bitmap_subset() check below will evaluate to false,
preventing RVA23U64 from being detected on compliant hardware.

> +	__set_bit(RISCV_ISA_EXT_ZICSR, ext_mask);
[ ... ]
> +	__set_bit(RISCV_ISA_EXT_SUPM, ext_mask);
> +
> +	if (!bitmap_subset(ext_mask, isa, RISCV_ISA_EXT_MAX))
> +		return;
> +
> +	set_bit(RISCV_ISA_BASE_RVA23U64, bases);
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260915-rva23u64-hwprobe-v2-v7-0-aa35ebd9ae96@oss.qualcomm.com?part=9