Re: [PATCH 1/3] ACPI: CPPC: Refactor boost ratio handling

[Mario Limonciello]

On 8/31/26 23:41, K Prateek Nayak wrote:
> Hello Mario,
>
> On 9/1/2026 1:14 AM, Mario Limonciello wrote:
> > +static const struct amd_max_freq max_freq_zen6_client __initconst = {
> > +	.performance = 5025,
> > +	.efficiency = 3524,
> > +	.low_power = 2399,
> > +};
> > +
> > +static const struct x86_cpu_id amd_zen6_client_ids[] __initconst = {
> > +	X86_MATCH_VENDOR_FAM_MODEL(AMD, 0x1A, 0x80, &max_freq_zen6_client),
> > +	X86_MATCH_VENDOR_FAM_MODEL(AMD, 0x1A, 0x81, &max_freq_zen6_client),
> > +	X86_MATCH_VENDOR_FAM_MODEL(AMD, 0x1A, 0x84, &max_freq_zen6_client),
> > +	X86_MATCH_VENDOR_FAM_MODEL(AMD, 0x1A, 0x85, &max_freq_zen6_client),
> > +	X86_MATCH_VENDOR_FAM_MODEL(AMD, 0x1A, 0xe0, &max_freq_zen6_client),
> > +	{}
> > +};
>
> So I see:
>
>    WARNING: modpost: vmlinux: section mismatch in reference: amd_get_max_frequency+0x27 (section: .text) -> amd_zen6_client_ids (section: .init.rodata)
>
> Since amd_get_boost_ratio() can be called post init too, I think
> dropping the __initconst annotation makes sense for these constants
> otherwise we may have issues after initmem is freed.

Thanks, will get this fixed.  I didn't catch it because I usually step 
out during my kernel builds and it didn't make enough noise to stop the 
build :P.