Re: [PATCH] x86: avoid SORT_BY_INIT_PRIORITY with old GNU ld

[Jan Beulich <jbeulich@suse.com>]

On 07.03.2022 16:52, Roger Pau Monné wrote:
> On Mon, Mar 07, 2022 at 04:29:22PM +0100, Jan Beulich wrote:
>> On 07.03.2022 16:05, Roger Pau Monné wrote:
>>> On Mon, Mar 07, 2022 at 02:49:32PM +0100, Jan Beulich wrote:
>>>> Support for this construct was added in 2.22 only. Avoid the need to
>>>> introduce logic to probe for linker script capabilities by (ab)using the
>>>> probe for a command line option having appeared at about the same time.
>>>>
>>>> Fixes: 4b7fd8153ddf ("x86: fold sections in final binaries")
>>>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
>>>> ---
>>>> Obviously this doesn't take care of old LLVM ld yet, if any care is
>>>> needed there in the first place.
>>>>
>>>> --- a/xen/arch/x86/arch.mk
>>>> +++ b/xen/arch/x86/arch.mk
>>>> @@ -70,6 +70,11 @@ ifeq ($(CONFIG_UBSAN),y)
>>>>  $(call cc-option-add,CFLAGS_UBSAN,CC,-fno-sanitize=alignment)
>>>>  endif
>>>>  
>>>> +# While not much better than going by raw GNU ld version, utilize that the
>>>> +# feature we're after has appeared in the same release as the
>>>> +# --print-output-format command line option.
>>>> +AFLAGS-$(call ld-option,--print-output-format) += -DHAVE_LD_SORT_BY_INIT_PRIORITY
>>>
>>> LLVM ld doesn't have --print-output-format:
>>>
>>> % ld --print-output-format
>>> ld: error: unknown argument '--print-output-format'
>>>
>>> So it would be always defaulting to SORT. I'm not sure what to
>>> recommend.
>>
>> Do you know if all versions we support know of SORT_BY_INIT_PRIORITY?
> 
> Hm, I don't think we can assume that we support LLVM LD in 3.5. I'm
> not even sure if LLVM 3.5 had LLD in the first place.
> 
> The first version that we care about and that we test is LLVM LD 6,
> anything below that version is of unknown state.
> 
> I've tested you change with SORT_BY_INIT_PRIORITY on it and it seems
> to be fine. This was on FreeBSD 12.3 version of LLD, not sure how
> many changes have been backported from newer versions there.

I'm inclined to suggest then that we unconditionally enable use of
this, in the hope that we'll never see a bug report. But of course
this then again gets me into the business of needing to determine
the which ld variant we're working with. Looks like I won't be able
to escape this anymore ...

Jan