Re: [PATCH] EDAC/mce_amd: Reduce unnecessary spew in dmesg if SMCA feature bit is not exposed

[Yazen Ghannam]

On Tue, Jun 15, 2021 at 05:18:50PM +0200, Borislav Petkov wrote:
> On Tue, Jun 15, 2021 at 11:08:46AM -0400, Yazen Ghannam wrote:
> > How about adding the the SMCA feature to the amd64_cpuids[] table in
> > amd64_edac.c?
> > 
> > We can use X86_MATCH_VENDOR_FEATURE to match on AMD (and Hygon) systems
> > with SMCA. And we can remove the X86_MATCH_VENDOR_FAM entries for
> > families 17h-19h.
> 
> Sure. That'll alleviate the need to add new families which support SMCA
> too.
> 
> > I'm assuming the issue is that amd64_edac_mod is autoloading due to the
> > family-based device table, and this will load edac_mce_amd as a
> > dependency.
> 
> Is it?
> 
> We have
> 
> early_initcall(mce_amd_init);
> 
> in mce_amd.c which attempts to load this thing unconditionally.
>

I think edac_mce_amd is usually built as a module by distro configs, so
early_initcall() would be replaced by module_init(). But the default
option is built-in, so you're right about the early_initcall().

Also, you bring up a good point. We can't say to people "don't load the
module" if it's builtin. And I don't think it's fair to say "don't
build-in the module" if the default is "y".

So I think we can downgrade this warning to a debug message, if the
module stays builtin. And/or we change the default config option to
module, and we make sure the module only autoloads in the proper cases.

What do you think?

Thanks,
Yazen