[PATCH] x86: keep PLT32 relocation for local symbols instead of converting to PC32

[Jan Beulich]

On 18.02.2026 19:05, Fangrui Song wrote:
> On Wed, Feb 18, 2026 at 1:36 AM Jan Beulich <jbeulich@suse.com> wrote:
>> On 17.02.2026 19:12, Fangrui Song wrote:
>>> We should accept that the @plt notation in assembly and the PLT32
>>> relocation type names are misleading. They suggest that a PLT entry is
>>> involved, but that is often not the case.
>>
>> Effectively what you're suggesting is that we deprecate @plt, properly
>> documenting that its absence on CALL/JMP operands still means PLT32 will be
>> used as relocation. That'll need properly spelling out in the doc then.
>> Which may be not as easy as it sounds, as it looks as if none of the @...
>> operand modifiers were documented anywhere.
> 
> I agree that the @plt behavior on x86-64 deserves documentation. One
> choice is for the x86-64 psABI to document @plt .

The psABI isn't the place to talk about assembly syntax. It could be put
there in more general terms, sure.

> Alternatively, x86-64 could have a convention doc like RISC-V's
> https://github.com/riscv-non-isa/riscv-toolchain-conventions

No idea whether introducing something like that would make sense this late.
For AT&T syntax there used to be an assembler spec (maintained first by Sun,
then Oracle), but I don't think that's being kept up-to-date.

Jan