Re: [PATCH v2] lib/raid/xor: x86: Add AVX-512 optimized xor_gen()

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] The AVX-512 XOR implementation relies solely on CPUID hardware checks (`boot_cpu_has(X86_FEATURE_AVX512F)`) and skips verifying if the OS actually enabled AVX-512 state management in the XCR0 register, risking a #UD exception.
--

commit d9094ea6909535dc1fa7dcdb6976401cdf1f90e8
Author: Eric Biggers <[email protected]>

lib/raid/xor: x86: Add AVX-512 optimized xor_gen()

This commit adds an AVX-512 optimized implementation of xor_gen() using
512-bit vectors and the vpternlogq instruction. The optimization is enabled
on x86_64 CPUs with AVX512F support that do not have overly-eager
downclocking (!PREFER_YMM). The author intentionally omitted the
cpu_has_xfeatures() check to avoid build issues on User-Mode Linux.

> diff --git a/lib/raid/xor/x86/xor_arch.h b/lib/raid/xor/x86/xor_arch.h
> index 99fe85a213c66..b5d49376fc974 100644
> --- a/lib/raid/xor/x86/xor_arch.h
> +++ b/lib/raid/xor/x86/xor_arch.h
[ ... ]
>  static __always_inline void __init arch_xor_init(void)
>  {
> -	if (boot_cpu_has(X86_FEATURE_AVX) &&
> -	    boot_cpu_has(X86_FEATURE_OSXSAVE)) {
> +	if (IS_ENABLED(CONFIG_X86_64) && boot_cpu_has(X86_FEATURE_AVX512F) &&
> +	    boot_cpu_has(X86_FEATURE_OSXSAVE) &&
> +	    !boot_cpu_has(X86_FEATURE_PREFER_YMM)) {

[Severity: High]
Could omitting the cpu_has_xfeatures() check in arch_xor_init() result in a
#UD exception?

While CR4.OSXSAVE indicates that XCR0 is enabled, it does not guarantee that
the OS or hypervisor actually enabled the AVX-512 state components (OPMASK,
ZMM_Hi256, Hi16_ZMM) in XCR0. 

If a hypervisor advertises AVX-512 CPUID bits but masks out the corresponding
XCR0 bits, executing EVEX-encoded AVX-512 instructions could trigger an
Invalid Opcode exception and cause a kernel panic.

> +		/* AVX-512 will be the best; no need to try others. */
> +		/* !PREFER_YMM excludes CPUs with overly-eager downclocking. */
> +		xor_force(&xor_block_avx512);
> +	} else if (boot_cpu_has(X86_FEATURE_AVX) &&
> +		   boot_cpu_has(X86_FEATURE_OSXSAVE)) {
> +		/* AVX will be the best; no need to try others. */
>  		xor_force(&xor_block_avx);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1