Re: [PATCH RESEND 1/1] X86: Handle Hyper-V vmbus interrupts as special hypervisor interrupts

["H. Peter Anvin" <hpa@zytor.com>]

Sounds great in theory at least, as long as it doesn't mean pushing a bunch of #ifdefs into other code.

Borislav Petkov <bp@alien8.de> wrote:

>On Wed, Jan 23, 2013 at 05:56:09PM -0800, K. Y. Srinivasan wrote:
>> diff --git a/arch/x86/kernel/entry_64.S b/arch/x86/kernel/entry_64.S
>> index 1975122..803ca69 100644
>> --- a/arch/x86/kernel/entry_64.S
>> +++ b/arch/x86/kernel/entry_64.S
>> @@ -1446,11 +1446,16 @@ ENTRY(xen_failsafe_callback)
>>  	CFI_ENDPROC
>>  END(xen_failsafe_callback)
>>  
>> -apicinterrupt XEN_HVM_EVTCHN_CALLBACK \
>> +apicinterrupt HYPERVISOR_CALLBACK_VECTOR \
>>  	xen_hvm_callback_vector xen_evtchn_do_upcall
>>  
>>  #endif /* CONFIG_XEN */
>>  
>> +#if IS_ENABLED(CONFIG_HYPERV)
>> +apicinterrupt HYPERVISOR_CALLBACK_VECTOR \
>> +	hyperv_callback_vector hyperv_vector_handler
>> +#endif /* CONFIG_HYPERV */
>
>arch/x86/built-in.o: In function `_set_gate':
>/w/kernel/linux-2.6/arch/x86/include/asm/desc.h:328: undefined
>reference to `hyperv_callback_vector'
>make: *** [vmlinux] Error 1
>
>because, of course:
>
># CONFIG_HYPERV is not set
>
>But, I have a more serious pet-peeve with the whole hypervisors
>detection stuff: we're building arch/x86/kernel/cpu/hypervisor.c
>unconditionally and yet, we have CONFIG_PARAVIRT_GUEST to ask the user
>whether she wants to enable some options for running linux as a guest.
>
>And actually, it would be better to put all that virt-related stuff
>under a config option called HYPERVISOR or whatever, under "Processor
>type and features" which opens a menu with all virt stuff for people
>and
>distros to select.
>
>This way, init_hypervisor_platform and the rest of hypervisors stuff
>won't run needlessly on baremetal and setups who don't want that.
>
>Any non-starter reasons for not doing that?

-- 
Sent from my mobile phone. Please excuse brevity and lack of formatting.