Re: [PATCH] x86/pv: Rework TRY_LOAD_SEG() to use asm goto()

[Nicola Vetrini <nicola.vetrini@bugseng.com>]

On 2025-07-21 11:25, Jan Beulich wrote:
> On 21.07.2025 10:16, Nicola Vetrini wrote:
>> On 2025-07-21 08:41, Jan Beulich wrote:
>>> On 18.07.2025 22:25, Andrew Cooper wrote:
>>>> This moves the exception path to being out-of-line within the
>>>> function, rather
>>>> than in the .fixup section, which improves backtraces.
>>>> 
>>>> Because the macro is used multiple times, the fault label needs
>>>> declaring as
>>>> local.
>>>> 
>>>> No functional change.
>>>> 
>>>> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
>>>> ---
>>>> CC: Jan Beulich <JBeulich@suse.com>
>>>> CC: Roger Pau Monné <roger.pau@citrix.com>
>>>> 
>>>> Slightly RFC.  I haven't checked if Eclair will be happy with
>>>> __label__ yet.
>>> 
>>> Even if it is, I guess you'd need to update the list of extensions we
>>> use (docs/misra/C-language-toolchain.rst)?
>> 
>> Only for using the __label__ token in
>> automation/eclair_analysis/ECLAIR/toolchain.ecl. The extension itself 
>> is
>> already documented in 5590c7e6590d ("eclair: allow and document use of
>> GCC extension for label addresses")
> 
> Except that it's not the address taking that is the point in question 
> here.
> We have meanwhile gained a number of asm-goto (and for the uses there 
> I'm
> not even sure they count as "address taking"). It's really the 
> __label__
> extended keyword (and the thus possible declaration of a 
> scope-restricted
> label) that my remark was about. But yes, toolchain.ecl looks to need a
> change, too.
> 

You're right, it needs also section 6.2 "Locally Declared Labels". Both 
easy to add if needed.

-- 
Nicola Vetrini, B.Sc.
Software Engineer
BUGSENG (https://bugseng.com)
LinkedIn: https://www.linkedin.com/in/nicola-vetrini-a42471253