x86: Don't allow HVM alongside PV_SHIM_EXCLUSIVE

[PATCH] x86: Don't allow HVM alongside PV_SHIM_EXCLUSIVE

Posted by Alejandro Vallejo 3 months, 2 weeks ago

Otherwise compile-time errors ensue. It's a nonsensical configuration,
but it's supriously triggered in randconfig jobs.

Fixes: 8b5b49ceb3d9("x86: don't include domctl and alike in shim-excl...")
Signed-off-by: Alejandro Vallejo <alejandro.garciavallejo@amd.com>
---
 xen/arch/x86/hvm/Kconfig | 1 +
 1 file changed, 1 insertion(+)

diff --git a/xen/arch/x86/hvm/Kconfig b/xen/arch/x86/hvm/Kconfig
index b903764bda..f57a87bca3 100644
--- a/xen/arch/x86/hvm/Kconfig
+++ b/xen/arch/x86/hvm/Kconfig
@@ -1,6 +1,7 @@
 menuconfig HVM
 	bool "HVM support"
 	default !PV_SHIM
+	depends on !PV_SHIM_EXCLUSIVE
 	select COMPAT
 	select IOREQ_SERVER
 	select MEM_ACCESS_ALWAYS_ON

base-commit: 6efafbf13351cd4ca0b01407a27570a43bc9f15f
-- 
2.43.0

Re: [PATCH] x86: Don't allow HVM alongside PV_SHIM_EXCLUSIVE

Posted by Andrew Cooper 3 months, 2 weeks ago

On 18/07/2025 4:12 pm, Alejandro Vallejo wrote:
> Otherwise compile-time errors ensue. It's a nonsensical configuration,
> but it's supriously triggered in randconfig jobs.
>
> Fixes: 8b5b49ceb3d9("x86: don't include domctl and alike in shim-excl...")
> Signed-off-by: Alejandro Vallejo <alejandro.garciavallejo@amd.com>
> ---
>  xen/arch/x86/hvm/Kconfig | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/xen/arch/x86/hvm/Kconfig b/xen/arch/x86/hvm/Kconfig
> index b903764bda..f57a87bca3 100644
> --- a/xen/arch/x86/hvm/Kconfig
> +++ b/xen/arch/x86/hvm/Kconfig
> @@ -1,6 +1,7 @@
>  menuconfig HVM
>  	bool "HVM support"
>  	default !PV_SHIM
> +	depends on !PV_SHIM_EXCLUSIVE
>  	select COMPAT
>  	select IOREQ_SERVER
>  	select MEM_ACCESS_ALWAYS_ON

I deeply regret not nacking this originally and blocking the patch you
listed as fixed.

This depends was explicitly reverted in 568f806cba4c because it breaks a
lot of things, notably `make allyesconfig`.

We're going to need to find a different way to fix RANDCONFIG.

~Andrew

Re: [PATCH] x86: Don't allow HVM alongside PV_SHIM_EXCLUSIVE

Posted by Alejandro Vallejo 3 months, 2 weeks ago

On Fri Jul 18, 2025 at 5:41 PM CEST, Andrew Cooper wrote:
> On 18/07/2025 4:12 pm, Alejandro Vallejo wrote:
>> Otherwise compile-time errors ensue. It's a nonsensical configuration,
>> but it's supriously triggered in randconfig jobs.
>>
>> Fixes: 8b5b49ceb3d9("x86: don't include domctl and alike in shim-excl...")
>> Signed-off-by: Alejandro Vallejo <alejandro.garciavallejo@amd.com>
>> ---
>>  xen/arch/x86/hvm/Kconfig | 1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff --git a/xen/arch/x86/hvm/Kconfig b/xen/arch/x86/hvm/Kconfig
>> index b903764bda..f57a87bca3 100644
>> --- a/xen/arch/x86/hvm/Kconfig
>> +++ b/xen/arch/x86/hvm/Kconfig
>> @@ -1,6 +1,7 @@
>>  menuconfig HVM
>>  	bool "HVM support"
>>  	default !PV_SHIM
>> +	depends on !PV_SHIM_EXCLUSIVE
>>  	select COMPAT
>>  	select IOREQ_SERVER
>>  	select MEM_ACCESS_ALWAYS_ON
>
> I deeply regret not nacking this originally and blocking the patch you
> listed as fixed.
>
> This depends was explicitly reverted in 568f806cba4c because it breaks a
> lot of things, notably `make allyesconfig`.
>
> We're going to need to find a different way to fix RANDCONFIG.
>
> ~Andrew

It can be done the other way around. PV_SHIM_EXCLUSIVE depending on !CONFIG_HVM,
but that won't help with allyesconfig.

PV_SHIM_EXCLUSIVE could be replaced by a PVSHIM_BOOT Kconfig. Combined with
DOM0LESS_BOOT and DOM0_BOOT this gives the full freedom of booting in shim mode,
in dom0 mode, or in dom0less/hlaunch mode; and is compatible with allyesconfig.

PV_SHIM_EXCLUSIVE would be unsetting DOM0_BOOT and DOM0LESS_BOOT.

Not that I want to do it right now. But it sounds like a nicer solution than the
current mess.

Cheers,
Alejandro

Re: [PATCH] x86: Don't allow HVM alongside PV_SHIM_EXCLUSIVE

Posted by Jan Beulich 3 months, 1 week ago

On 18.07.2025 18:12, Alejandro Vallejo wrote:
> On Fri Jul 18, 2025 at 5:41 PM CEST, Andrew Cooper wrote:
>> On 18/07/2025 4:12 pm, Alejandro Vallejo wrote:
>>> Otherwise compile-time errors ensue. It's a nonsensical configuration,
>>> but it's supriously triggered in randconfig jobs.
>>>
>>> Fixes: 8b5b49ceb3d9("x86: don't include domctl and alike in shim-excl...")
>>> Signed-off-by: Alejandro Vallejo <alejandro.garciavallejo@amd.com>
>>> ---
>>>  xen/arch/x86/hvm/Kconfig | 1 +
>>>  1 file changed, 1 insertion(+)
>>>
>>> diff --git a/xen/arch/x86/hvm/Kconfig b/xen/arch/x86/hvm/Kconfig
>>> index b903764bda..f57a87bca3 100644
>>> --- a/xen/arch/x86/hvm/Kconfig
>>> +++ b/xen/arch/x86/hvm/Kconfig
>>> @@ -1,6 +1,7 @@
>>>  menuconfig HVM
>>>  	bool "HVM support"
>>>  	default !PV_SHIM
>>> +	depends on !PV_SHIM_EXCLUSIVE
>>>  	select COMPAT
>>>  	select IOREQ_SERVER
>>>  	select MEM_ACCESS_ALWAYS_ON
>>
>> I deeply regret not nacking this originally and blocking the patch you
>> listed as fixed.
>>
>> This depends was explicitly reverted in 568f806cba4c because it breaks a
>> lot of things, notably `make allyesconfig`.
>>
>> We're going to need to find a different way to fix RANDCONFIG.
>>
>> ~Andrew
> 
> It can be done the other way around. PV_SHIM_EXCLUSIVE depending on !CONFIG_HVM,
> but that won't help with allyesconfig.

And it wouldn't be right either, in the model we're trying to follow now.

Jan