[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v1] x86: make Viridian support optional


  • To: Alejandro Vallejo <alejandro.vallejo@xxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Thu, 20 Mar 2025 16:22:57 +0100
  • Autocrypt: addr=jbeulich@xxxxxxxx; keydata= xsDiBFk3nEQRBADAEaSw6zC/EJkiwGPXbWtPxl2xCdSoeepS07jW8UgcHNurfHvUzogEq5xk hu507c3BarVjyWCJOylMNR98Yd8VqD9UfmX0Hb8/BrA+Hl6/DB/eqGptrf4BSRwcZQM32aZK 7Pj2XbGWIUrZrd70x1eAP9QE3P79Y2oLrsCgbZJfEwCgvz9JjGmQqQkRiTVzlZVCJYcyGGsD /0tbFCzD2h20ahe8rC1gbb3K3qk+LpBtvjBu1RY9drYk0NymiGbJWZgab6t1jM7sk2vuf0Py O9Hf9XBmK0uE9IgMaiCpc32XV9oASz6UJebwkX+zF2jG5I1BfnO9g7KlotcA/v5ClMjgo6Gl MDY4HxoSRu3i1cqqSDtVlt+AOVBJBACrZcnHAUSuCXBPy0jOlBhxPqRWv6ND4c9PH1xjQ3NP nxJuMBS8rnNg22uyfAgmBKNLpLgAGVRMZGaGoJObGf72s6TeIqKJo/LtggAS9qAUiuKVnygo 3wjfkS9A3DRO+SpU7JqWdsveeIQyeyEJ/8PTowmSQLakF+3fote9ybzd880fSmFuIEJldWxp Y2ggPGpiZXVsaWNoQHN1c2UuY29tPsJgBBMRAgAgBQJZN5xEAhsDBgsJCAcDAgQVAggDBBYC AwECHgECF4AACgkQoDSui/t3IH4J+wCfQ5jHdEjCRHj23O/5ttg9r9OIruwAn3103WUITZee e7Sbg12UgcQ5lv7SzsFNBFk3nEQQCACCuTjCjFOUdi5Nm244F+78kLghRcin/awv+IrTcIWF hUpSs1Y91iQQ7KItirz5uwCPlwejSJDQJLIS+QtJHaXDXeV6NI0Uef1hP20+y8qydDiVkv6l IreXjTb7DvksRgJNvCkWtYnlS3mYvQ9NzS9PhyALWbXnH6sIJd2O9lKS1Mrfq+y0IXCP10eS FFGg+Av3IQeFatkJAyju0PPthyTqxSI4lZYuJVPknzgaeuJv/2NccrPvmeDg6Coe7ZIeQ8Yj t0ARxu2xytAkkLCel1Lz1WLmwLstV30g80nkgZf/wr+/BXJW/oIvRlonUkxv+IbBM3dX2OV8 AmRv1ySWPTP7AAMFB/9PQK/VtlNUJvg8GXj9ootzrteGfVZVVT4XBJkfwBcpC/XcPzldjv+3 HYudvpdNK3lLujXeA5fLOH+Z/G9WBc5pFVSMocI71I8bT8lIAzreg0WvkWg5V2WZsUMlnDL9 mpwIGFhlbM3gfDMs7MPMu8YQRFVdUvtSpaAs8OFfGQ0ia3LGZcjA6Ik2+xcqscEJzNH+qh8V m5jjp28yZgaqTaRbg3M/+MTbMpicpZuqF4rnB0AQD12/3BNWDR6bmh+EkYSMcEIpQmBM51qM EKYTQGybRCjpnKHGOxG0rfFY1085mBDZCH5Kx0cl0HVJuQKC+dV2ZY5AqjcKwAxpE75MLFkr wkkEGBECAAkFAlk3nEQCGwwACgkQoDSui/t3IH7nnwCfcJWUDUFKdCsBH/E5d+0ZnMQi+G0A nAuWpQkjM1ASeQwSHEeAWPgskBQL
  • Cc: Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>, Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Roger Pau Monné <roger.pau@xxxxxxxxxx>, Stefano Stabellini <sstabellini@xxxxxxxxxx>, Sergiy Kibrik <sergiy_kibrik@xxxxxxxx>
  • Delivery-date: Thu, 20 Mar 2025 15:23:07 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 20.03.2025 16:18, Alejandro Vallejo wrote:
> On Thu Mar 20, 2025 at 9:39 AM GMT, Sergiy Kibrik wrote:
>> hi Alejandro,
>>
>> 17.03.25 11:19, Alejandro Vallejo:
>> [..]
>>>       endif
>>>
>>>     +config HVM_VIRIDIAN
>>>     +       bool "Viridian enlightenments support" if EXPERT
>>>     +       depends on HVM
>>>     +       default y
>>>     +
>>>
>>>
>>>
>>> I don't see why this should be gated by EXPERT, provided a
>>> suitable (now absent) help message was to exist explaining
>>> what it does in plain simple words.
>>
>> The option is intended primarily for fine-tuned systems optimized for 
>> particular platform and mode of operation. As for generic systems (e.g. 
>> distributions) whey wouldn't want to disable it anyway.
> 
> 
> 
>>>
>>> For the title, I'd say it needs to properly state it refers to
>>> enlightenments for guests, rather than enlightenments for
>>> Xen itself when running under Hyper-V. As it is, it sounds
>>> ambiguous (Maybe "Hyper-V enlighnments for guests"?).
>>>
>>
>> Agree, "Hyper-V enlighnments for guests" is better title.
>> As for help message, would the one below be sufficient?:
>>
>>   help
>>     Support optimizations for Hyper-V guests such as faster hypercalls,
>>     efficient timer and interrupt handling, and enhanced paravirtualized
>>     I/O. This is to improve performance and compatibility of Windows VMs.
>>
>>     If unsure, say Y.
> 
> Sounds good enough to me.
> 
>>
>>
>>> On a personal nitpicky preference note, I'd say HVM_VIRIDIAN sounds
>>> rather redundant, and I think just VIRIDIAN works just as well
>>> while being shorter.
>>>
>>
>> this is rather to highlight the fact that the code is part of HVM 
>> support, a bit of self-documenting
>>
>> [..]
> 
> That's fair enough.
> 
>>>     diff --git a/xen/arch/x86/include/asm/hvm/vcpu.h
>>>     b/xen/arch/x86/include/asm/hvm/vcpu.h
>>>     index 196fed6d5d..bac35ec47a 100644
>>>     --- a/xen/arch/x86/include/asm/hvm/vcpu.h
>>>     +++ b/xen/arch/x86/include/asm/hvm/vcpu.h
>>>     @@ -171,8 +171,9 @@ struct hvm_vcpu {
>>>
>>>           /* Pending hw/sw interrupt (.vector = -1 means nothing
>>>     pending). */
>>>           struct x86_event     inject_event;
>>>     -
>>>     +#ifdef CONFIG_HVM_VIRIDIAN
>>>           struct viridian_vcpu *viridian;
>>>     +#endif
>>>       };
>>>
>>>       #endif /* __ASM_X86_HVM_VCPU_H__ */
>>>
>>>
>>> nit: I suspect the code would be far less cluttered with "if 
>>> viridian..." if the
>>> init/deinit/etc functions had dummy versions of those functions when
>>> !HVM_VIRIDIAN in the header.
>>>
>>
>> as Jan explained some time ago [1] it's preferable to compile as much as 
>> possible in all build configuration. Besides most of calls to viridian 
>> code are already guarded by is_viridian_domain() & not actually require 
>> stubs.
>>
>>   -Sergiy
>>
>> [1] 
>> https://lore.kernel.org/xen-devel/36708a3f-2664-4b04-9f5d-f115d362d100@xxxxxxxx/
> 
> That answer seems to state a preference for...
> 
>     if ( foo_enabled() )
>         rc = foo();
> 
> ... against...
> 
>     #ifdef CONFIG_FOO
>     rc = foo();
>     #endif
> 
> ... where foo() in the header looks like...
> 
>     #ifdef CONFIG_FOO
>     int foo(void);
>     #else /* CONFIG_FOO */
>     static inline int foo(void)
>     {
>         return /*some-error*/;
>     }
>     #endif /* CONFIG_FOO */
> 
> But that's not what's going on here, I think? I didn't initially notice the
> subtlety of the change. On more careful look, it seems to rely on the compiler
> doing dead-code-elimination. The functions missing in the linking stage don't
> cause a compile-time error because the conditionals are completely gone by
> then. Neat as it is, it sounds a bit fragile. Can we really rely on this
> behaviour not changing? Furthermore, does MISRA have views about having dead
> code calls to unimplemented functions?

We use DCE like this in many places, so we already rely on this behavior not
changing.

Jan



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.