Re: [PATCH v1 3/3] x86/msr: Convert a native_wrmsr() use to native_wrmsrq()

[Xin Li]

On 5/15/2025 10:54 AM, Xin Li wrote:
> On 5/15/2025 8:27 AM, Ingo Molnar wrote:
> > * Xin Li (Intel) <xin@xxxxxxxxx> wrote:
> >
> > > Convert a native_wrmsr() use to native_wrmsrq() to zap meaningless type
> > > conversions when a u64 MSR value is splitted into two u32.
> >
> > BTW., at this point we should probably just replace
> > sev_es_wr_ghcb_msr() calls with direct calls to:
> >
> >     native_wrmsrq(MSR_AMD64_SEV_ES_GHCB, ...);
> >
> > as sev_es_wr_ghcb_msr() is now basically an open-coded native_wrmsrq().
>
> I thought about it, however it looks to me that current code prefers not
> to spread MSR_AMD64_SEV_ES_GHCB in 17 callsites.  And anyway it's a 
> __always_inline function.
>
> But as you have asked, I will make the change unless someone objects.

Hi Ingo,

I took a further look and found that we can't simply replace
sev_es_wr_ghcb_msr() with native_wrmsrq(MSR_AMD64_SEV_ES_GHCB, ...).

There are two sev_es_wr_ghcb_msr() definitions.  One is defined in
arch/x86/boot/compressed/sev.h and it references boot_wrmsr() defined in
arch/x86/boot/msr.h to do MSR write.

The other one is defined in arch/x86/include/asm/sev-internal.h, which
uses native_wrmsrq() from arch/x86/include/asm/msr.h to write MSR.

Because:
1) arch/x86/boot/startup/sev-shared.c is included in both
        arch/x86/boot/compressed/sev.c
   and
        arch/x86/boot/startup/sev-startup.c

2) arch/x86/boot/startup/sev-shared.c has several references to
   sev_es_wr_ghcb_msr(),

sev_es_wr_ghcb_msr() is converted to boot_wrmsr() when included in
arch/x86/boot/compressed/sev.c or native_wrmsrq() when included in
arch/x86/boot/startup/sev-startup.c.

It would change the compressed code to use native_wrmsrq() if we remove
sev_es_wr_ghcb_msr() from arch/x86/include/asm/sev-internal.h and use 
native_wrmsrq() directly in the startup code.

We probably should get rid of boot_wrmsr() and use native_wrmsrq() in
the compressed code because they are indeed the same thing.  But as we
are so close to the v6.16 merge window, I don't think it's a good idea
to make the change right now.

So maybe I should just drop this patch and we can do the job after the 
coming merge window.

But if you think it's not a bad idea to replace native_wrmsr() with
native_wrmsrq() right now, I can keep this original patch.

Thanks!
    Xin