PING^2: [PATCH] elf: Avoid nested functions in the loader (all ports) [BZ #27220]

[Fāng-ruì Sòng]

On 2021-09-22, Carlos O'Donell wrote:
>On 9/22/21 14:09, Joseph Myers wrote:
>> On Wed, 22 Sep 2021, Carlos O'Donell via Libc-alpha wrote:
>>
>>> - Split this patch up into more logical changes.
>>
>> I'm not convinced there's much scope for splitting up this patch more; it
>> is a logical change (versions that changed only some architectures were a
>> lot uglier).  My comment about logical changes was a more generic one, not
>> suggesting that this particular patch should be split up further.
>
>Thanks for the clarification. In which case the more detailed review of the boostrap
>relocations is the remaining point.

I really appreciate that Clang buildability is well-received :)
I went ahead and filed https://sourceware.org/bugzilla/show_bug.cgi?id=28376 "[meta] Build glibc with Clang"
to collect miscellaneous small fixes.

---

I agree that this patch is difficult to split.
Modifying the elf/ macro parameters requires fixing all use sites.

The latest version is at https://github.com/MaskRay/glibc/tree/unnest
(I use a different repo instead of the canonical sourceware repo in the fear of pushing bad commits to sourceware :)

About the global variables (cur_*): they are in elf/dl-reloc.c

-    /* String table object symbols.  */
-    const char *strtab = (const void *) D_PTR (l, l_info[DT_STRTAB]);
+    cur_l = l;
+    cur_scope = scope;
+    cur_strtab = (const void *) D_PTR (cur_l, l_info[DT_STRTAB]);

removing them requires more parameters to ELF_DYNAMIC_RELOCATE and RESOLVE_MAP.
Is that the direction you are imagining?

BTW: I am not familiar with the concept of bootstrap relocations.
Does it refer to relocations in ld.so itself   and   all relocations in a static pie?

---

FWIW I initially sent
https://sourceware.org/pipermail/libc-alpha/2021-August/130340.html
(https://github.com/MaskRay/glibc/tree/nesting)
which is a more literal take from
https://sourceware.org/git/?p=glibc.git;a=shortlog;h=refs/heads/google/grte/v5-2.27/master

But I slightly dislike the version myself because it duplicates lots of macros.
(This branch can give some insight for needed Clang fixes but many commits need rewrite for master.)