From: Carlos O'Donell <carlos@redhat.com>
To: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>,
Xi Ruoyao <xry111@xry111.site>,
libc-alpha@sourceware.org
Cc: caiyinyu <caiyinyu@loongson.cn>, Wang Xuerui <i@xen0n.name>
Subject: Re: [PATCH] LoongArch: ldconfig: Ignore EF_LARCH_OBJABI_V1 in shared objects
Date: Wed, 29 Mar 2023 09:12:56 -0400 [thread overview]
Message-ID: <4b1411d2-b9ab-1395-9360-ced8d12c6ddc@redhat.com> (raw)
In-Reply-To: <d2d239aa-75ed-847d-d97b-465a85bd47d4@linaro.org>
On 3/27/23 12:54, Adhemerval Zanella Netto wrote:
>
>
> On 27/03/23 11:57, Xi Ruoyao wrote:
>> On Mon, 2023-03-27 at 09:44 -0400, Carlos O'Donell wrote:
>>> On 3/26/23 07:13, Xi Ruoyao via Libc-alpha wrote:
>>>> Binutils 2.40 sets EF_LARCH_OBJABI_V1 for shared objects:
>>>>
>>>> $ ld --version | head -n1
>>>> GNU ld (GNU Binutils) 2.40
>>>> $ echo 'int dummy;' > dummy.c
>>>> $ cc dummy.c -shared
>>>> $ readelf -h a.out | grep Flags
>>>> Flags: 0x43, DOUBLE-FLOAT, OBJ-v1
>>>>
>>>> We need to ignore it in ldconfig or ldconfig will consider all shared
>>>> objects linked by Binutils 2.40 "unsupported". Maybe we should stop
>>>> setting EF_LARCH_OBJABI_V1 for shared objects, but Binutils 2.40 is
>>>> already released and we cannot change it.
>>>> ---
>>>> sysdeps/unix/sysv/linux/loongarch/readelflib.c | 2 +-
>>>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/sysdeps/unix/sysv/linux/loongarch/readelflib.c b/sysdeps/unix/sysv/linux/loongarch/readelflib.c
>>>> index bcaff86b36..ceba355959 100644
>>>> --- a/sysdeps/unix/sysv/linux/loongarch/readelflib.c
>>>> +++ b/sysdeps/unix/sysv/linux/loongarch/readelflib.c
>>>> @@ -40,7 +40,7 @@ process_elf_file (const char *file_name, const char *lib, int *flag,
>>>>
>>>> ret = process_elf64_file (file_name, lib, flag, isa_level, soname,
>>>> file_contents, file_length);
>>>> - flags = elf64_header->e_flags;
>>>> + flags = elf64_header->e_flags & ~EF_LARCH_OBJABI_V1;
>>>
>>> Are such objects ABI compatible?
>>
>> This flag was designed for indicating the relocation type usage in a .o
>> file:
OK, so the final form of the shared objects is compatible, this is only about static
link compatibility and non-shared relocations.
I would suggest reviewing ".gnu_attribute" for an extension to Ehdr->ef_flags that
other architectures use, like ppc64le, to do such compatibility checking and error
detection.
>> If EF_LARCH_OBJABI_V1 is set, it's allowed to use relocation types with
>> ID in [64, 100], but it's prohibited to use relocation types with ID in
>> [22, 46].
>>
>> If EF_LARCH_OBJABI_V1 is not set, it's allowed to use relocation types
>> with ID in [22, 46], but it's prohibited to use relocation types with ID
>> in [64, 100].
>>
>> A linker may only support the .o files with EF_LARCH_OBJABI_V1 unset
>> (for example, GNU ld 2.38), or only support the .o files with
>> EF_LARCH_OBJABI_V1 set (for example, LLD under review at
>> https://reviews.llvm.org/D138135), or support both (for example, GNU ld
>> 2.40).
>>
>> If a linker supports both, it's OK to link a EF_LARCH_OBJABI_V1 .o file
>> together with a non-EF_LARCH_OBJABI_V1 .o file into an executable or
>> shared object.
>>
>> But none of relocation type ID in [22, 46] or [64, 100] is runtime
>> relocation. I. e. those relocation types should not show up in a shared
>> object at all (if one shows up, the linker is buggy). So
>> EF_LARCH_OBJABI_V1 makes no difference for shared objects.
>
> Since the boat is already sailed with 2.40, I think the proposed patch
> is fine (it would be better if you indeed fix it on 2.41). I would
> just like to add a comment on why this is required:
>
> /* The EF_LARCH_OBJABI_V1 flag indicate which set of static relocations
> the object might use and it only considered during static linking,
> it does not reflect in runtime relocations. However some binutils
> version might set it on dynamic shared object, so clear it to avoid
> see the SO as unsupported. */
Agreed.
commit 07dd75589ecbedec5162a5645d57f8bd093a45db
Author: caiyinyu <caiyinyu@loongson.cn>
Date: Tue Mar 28 09:19:53 2023 +0800
LoongArch: ldconfig: Add comments for using EF_LARCH_OBJABI_V1
We added Adhemerval Zanella's comment to explain the reason for
using EF_LARCH_OBJABI_V1.
And the comment is now in the sources.
This is better, but I have a post-commit review comment here about the overall
usage of ELF flags.
While it is possible to use Ehdr->e_flags for such purposes as marking static
relocation compatibility, you also have the broader use of ".gnu_attribute"
via:
GNU Object Attributes:
https://sourceware.org/binutils/docs/as/GNU-Object-Attributes.html
Which allows a richer set of non-flag-based attributes to be used for object
files and for static link diagnostics.
--
Cheers,
Carlos.
prev parent reply other threads:[~2023-03-29 13:13 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-26 11:13 Xi Ruoyao
2023-03-27 13:44 ` Carlos O'Donell
2023-03-27 14:57 ` Xi Ruoyao
2023-03-27 16:54 ` Adhemerval Zanella Netto
2023-03-27 17:46 ` Xi Ruoyao
2023-03-29 13:12 ` Carlos O'Donell [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4b1411d2-b9ab-1395-9360-ced8d12c6ddc@redhat.com \
--to=carlos@redhat.com \
--cc=adhemerval.zanella@linaro.org \
--cc=caiyinyu@loongson.cn \
--cc=i@xen0n.name \
--cc=libc-alpha@sourceware.org \
--cc=xry111@xry111.site \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).