Re: [RFC PATCH v2 06/34] x86/msr: Use the alternatives mechanism to read PMC

[Juergen Gross <jgross@suse.com>]

[-- Attachment #1.1.1: Type: text/plain, Size: 2692 bytes --]

On 22.04.25 11:12, Xin Li wrote:
> On 4/22/2025 1:38 AM, Jürgen Groß wrote:
>> On 22.04.25 10:21, Xin Li (Intel) wrote:
>>> To eliminate the indirect call overhead introduced by the pv_ops API,
>>> use the alternatives mechanism to read PMC:
>>
>> Which indirect call overhead? The indirect call is patched via the
>> alternative mechanism to a direct one.
>>
> 
> See below.
> 
> 
>>>
>>>      1) When built with !CONFIG_XEN_PV, X86_FEATURE_XENPV becomes a
>>>         disabled feature, preventing the Xen PMC read code from being
>>>         built and ensuring the native code is executed unconditionally.
>>
>> Without CONFIG_XEN_PV CONFIG_PARAVIRT_XXL is not selected, resulting in
>> native code anyway.
> 
> Yes, this is kept in this patch, but in a little different way.
> 
>>
>>>
>>>      2) When built with CONFIG_XEN_PV:
>>>
>>>         2.1) If not running on the Xen hypervisor (!X86_FEATURE_XENPV),
>>>              the kernel runtime binary is patched to unconditionally
>>>              jump to the native PMC read code.
>>>
>>>         2.2) If running on the Xen hypervisor (X86_FEATURE_XENPV), the
>>>              kernel runtime binary is patched to unconditionally jump
>>>              to the Xen PMC read code.
>>>
>>> Consequently, remove the pv_ops PMC read API.
>>
>> I don't see the value of this patch.
>>
>> It adds more #ifdef and code lines without any real gain.
>>
>> In case the x86 maintainers think it is still worth it, I won't object.
> 
> I think we want to totally bypass pv_ops in the case 2.1).
> 
> Do you mean the indirect call is patched to call native code *directly*
> for 2.1?  I don't know it, can you please elaborate?

All paravirt indirect calls are patched to direct calls via the normal
alternative patch mechanism.

Have a look at alt_replace_call() in arch/x86/kernel/alternative.c

> AFAIK, Xen PV has been the sole user of pv_ops for nearly 20 years. This

Not quite. There was lguest until I ripped it out. :-)

And some use cases are left for KVM and Hyper-V guests (I have kept those
behind CONFIG_PARAVIRT, while the Xen-specific parts are behind
CONFIG_PARAVIRT_XXL now).

> raises significant doubts about whether pv_ops provides Linux with the
> value of being a well-abstracted "CPU" or "Platform".  And the x86
> maintainers have said that it's a maintenance nightmare.

I have worked rather hard to make it less intrusive, especially by removing
the paravirt specific code patching (now all done via alternative patching)
and by removing 32-bit Xen PV mode.


Juergen

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 3743 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 495 bytes --]