llvm dev - Nov 2015 - [lld] Hiding original type of GOT related relocations

Hi,

There are more than one MIPS relocations which need GOT entry
creation. Let's consider two of them R_MIPS_GOT16 and R_MIPS_CALL16
[1]. R_MIPS_GOT16 is applicable to local and external symbols and
performs a different calculation in each cases [2]. R_MIPS_CALL16 is
applicable to external symbols only and a linker should show an error
if it finds R_MIPS_CALL16 with a local target. Now LLD in the
`InputSectionBase<ELFT>::relocate` methods hides original type of GOT
relocations and always use the result of `Target->getGotRefReloc()`
call. In case of MIPS it is R_MIPS_GOT16. It makes impossible to
perform target symbol type checking and the error generation in the
`MipsTargetInfo<ELFT>::relocateOne` method.

Is it really necessary to hide original relocation type? If a target
has multiple GOT related relocations it might just add more `cases` to
its `relocateOne` method.

[1] ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf page 4-19
[2] ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf page 4-20

-- 
Simon Atanasyan

On Fri, Nov 20, 2015 at 10:13 PM, Simon Atanasyan <simon at atanasyan.com>
wrote:
> Hi,
>
> There are more than one MIPS relocations which need GOT entry
> creation. Let's consider two of them R_MIPS_GOT16 and R_MIPS_CALL16
> [1]. R_MIPS_GOT16 is applicable to local and external symbols and
> performs a different calculation in each cases [2]. R_MIPS_CALL16 is
> applicable to external symbols only and a linker should show an error
> if it finds R_MIPS_CALL16 with a local target. Now LLD in the
> `InputSectionBase<ELFT>::relocate` methods hides original type of GOT
> relocations and always use the result of `Target->getGotRefReloc()`
> call. In case of MIPS it is R_MIPS_GOT16. It makes impossible to
> perform target symbol type checking and the error generation in the
> `MipsTargetInfo<ELFT>::relocateOne` method.
>
I guess there's no need to hide the type. It's just that because it
works
for x86, x86-64 or aarch64. If mips requires more info, it should be ok to
change that.

> Is it really necessary to hide original relocation type? If a target
> has multiple GOT related relocations it might just add more `cases` to
> its `relocateOne` method.
>
> [1] ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf page 4-19
> [2] ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf page 4-20
>
> --
> Simon Atanasyan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20151120/119d4a8a/attachment.html>

Hi,

On 21.11.2015 12:31, Rui Ueyama via llvm-dev wrote:
> On Fri, Nov 20, 2015 at 10:13 PM, Simon Atanasyan <simon at
atanasyan.com
> <mailto:simon at atanasyan.com>> wrote:
>
>     Hi,
>
>     There are more than one MIPS relocations which need GOT entry
>     creation. Let's consider two of them R_MIPS_GOT16 and R_MIPS_CALL16
>     [1]. R_MIPS_GOT16 is applicable to local and external symbols and
>     performs a different calculation in each cases [2]. R_MIPS_CALL16 is
>     applicable to external symbols only and a linker should show an error
>     if it finds R_MIPS_CALL16 with a local target. Now LLD in the
>     `InputSectionBase<ELFT>::relocate` methods hides original type of
GOT
>     relocations and always use the result of `Target->getGotRefReloc()`
>     call. In case of MIPS it is R_MIPS_GOT16. It makes impossible to
>     perform target symbol type checking and the error generation in the
>     `MipsTargetInfo<ELFT>::relocateOne` method.
>
>
> I guess there's no need to hide the type. It's just that because it
> works for x86, x86-64 or aarch64. If mips requires more info, it 
> should be ok to change that.
>
There was no support for GOT relocations for the AArch64 target. I'm 
adding it with http://reviews.llvm.org/D14917.
I faced the similar problem, so I extended the getGotRefReloc() method 
in this patch.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20151123/2cb83c08/attachment.html>