public inbox for binutils@sourceware.org
 help / color / mirror / Atom feed
From: Alan Modra <amodra@gmail.com>
To: Fangrui Song <maskray@google.com>
Cc: Joseph Myers <josmyers@redhat.com>,
	Alexandre Oliva <oliva@adacore.com>,
	binutils@sourceware.org, Fangrui Song <maskray@gcc.gnu.org>
Subject: Re: [PATCH] sh: Make protected symbols local for FDPIC -shared
Date: Wed, 21 Feb 2024 08:40:21 +1030	[thread overview]
Message-ID: <ZdUjTVLDoH7s/Qr1@squeak.grove.modra.org> (raw)
In-Reply-To: <20240219070941.2254076-1-maskray@google.com>

On Sun, Feb 18, 2024 at 11:09:40PM -0800, Fangrui Song wrote:
> From: Fangrui Song <maskray@gcc.gnu.org>
> 
>     __attribute__((visibility("protected"))) void protected_fun() {}
>     void *get_protected_fun() { return protected_fun; }
> 
> links fine in the non-FDPIC mode but not in the FDPIC mode:
> 
>     R_SH_GOTOFFFUNCDESC relocation against external symbol "protected_fun"
> 

Right.
$ sh4-linux-gnu-gcc -fPIC -mfdpic -c fundesc.c
$ ld/ld-new -m shlelf_fd -shared -o fundesc.so fundesc.o
ld/ld-new: fundesc.o(.text+0x24): R_SH_GOTOFFFUNCDESC relocation against external symbol "protected_fun"
ld/ld-new: final link failed

> The issue is similar to the x86 issue fixed by commit
> d19a265487eda186b6977d9d15648cda9fad3298 ("x86: Make protected symbols
> local for -shared").
> 
> SYMBOL_FUNCDESC_LOCAL in the R_SH_GOTOFFFUNCDESC check unnecessarily
> considers protected symbols non-local.  elf32-frv.c:FRVFDPIC_SYM_LOCAL
> avoids the issue by setting local_protected.  Set local_protected for sh
> as well. It seems that the condition `! elf_hash_table
> (INFO)->dynamic_sections_created` is redundant as `h->dynindx` should be
> -1.  Let's remove the condition.

Are you sure the dynamic_sections_created test is redundant?
_bfd_elf_symbol_refs_local_p returns false for undefined syms.  BTW, I
think it is better to write SYMBOL_CALLS_LOCAL rather than expanding
to _bfd_elf_symbol_refs_local_p.  The name conveys meaning.

This isn't an ack or a nack on the patch, I'll leave that to Alexandre
since I really don't know enough about the sh fdpic ABI without
spending quite a lot of time checking, but I suspect the patch isn't
complete due to code like

	    else if (SYMBOL_CALLS_LOCAL (info, h)
		     && ! SYMBOL_FUNCDESC_LOCAL (info, h))
	      {
		/* If the symbol needs a non-local function descriptor
		   but binds locally (i.e., its visibility is
		   protected), emit a dynamic relocation decayed to
		   section+offset.  This is an optimization; the dynamic
		   linker would resolve our function descriptor request
		   to our copy of the function anyway.  */
		dynindx = elf_section_data (h->root.u.def.section
					    ->output_section)->dynindx;
		relocation += h->root.u.def.section->output_offset
		  + h->root.u.def.value;
	      }

That would be an "else if (0)" after your patch.

> ---
>  bfd/elf32-sh.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
> 
> diff --git a/bfd/elf32-sh.c b/bfd/elf32-sh.c
> index bf4cb2a8242..7d35144cdc3 100644
> --- a/bfd/elf32-sh.c
> +++ b/bfd/elf32-sh.c
> @@ -64,8 +64,7 @@ static bfd_vma tpoff
>     not.  If the symbol is protected, we want the local address, but
>     its function descriptor must be assigned by the dynamic linker.  */
>  #define SYMBOL_FUNCDESC_LOCAL(INFO, H) \
> -  (SYMBOL_REFERENCES_LOCAL (INFO, H) \
> -   || ! elf_hash_table (INFO)->dynamic_sections_created)
> +  _bfd_elf_symbol_refs_local_p (H, INFO, 1)
>  \f
>  #define SH_PARTIAL32 true
>  #define SH_SRC_MASK32 0xffffffff
> -- 
> 2.44.0.rc0.258.g7320e95886-goog

-- 
Alan Modra
Australia Development Lab, IBM

      reply	other threads:[~2024-02-20 22:10 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-19  7:09 Fangrui Song
2024-02-20 22:10 ` Alan Modra [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=ZdUjTVLDoH7s/Qr1@squeak.grove.modra.org \
    --to=amodra@gmail.com \
    --cc=binutils@sourceware.org \
    --cc=josmyers@redhat.com \
    --cc=maskray@gcc.gnu.org \
    --cc=maskray@google.com \
    --cc=oliva@adacore.com \
    /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).