public inbox for binutils@sourceware.org
 help / color / mirror / Atom feed
From: "H.J. Lu" <hjl.tools@gmail.com>
To: Alan Modra <amodra@gmail.com>
Cc: Binutils <binutils@sourceware.org>
Subject: Re: [PATCH] x86: Ignore protected visibility in shared libraries on Solaris
Date: Wed, 24 Aug 2022 07:36:52 -0700	[thread overview]
Message-ID: <CAMe9rOoKGWuMBHCgt-U+Cu8eW6AOWHiG9qCS2rVYfXr1a9y_Mg@mail.gmail.com> (raw)
In-Reply-To: <YwWBqL3ndlXIrkHw@squeak.grove.modra.org>

On Tue, Aug 23, 2022 at 6:41 PM Alan Modra <amodra@gmail.com> wrote:
>
> On Wed, Aug 24, 2022 at 09:46:32AM +0930, Alan Modra wrote:
> > On Tue, Aug 23, 2022 at 10:34:32AM -0700, H.J. Lu via Binutils wrote:
> > > On x86, the PLT entry in executable may be used as function address for
> > > functions in shared libraries.  If functions are protected, the function
> > > address used in executable can be different from the function address
> > > used in shared library.  This will lead to incorrect run-time behavior
> > > if function pointer equality is needed.  By default, x86 linker issues
> > > an error in this case.
> > >
> > > On Solaris, linker issued an error for
> > >
> > > struct tm *tb = (kind == CPP_time_kind::FIXED ? gmtime : localtime) (&tt);
> > >
> > > where gmtime is a protected function in libc.so.  Use gmtime's PLT entry
> > > in executable as function address is safe since function pointer equality
> > > isn't needed.
> >
> > I'm curious as to how pointer_equality_needed came to be set for
> > gmtime.
>
> Hmm, I figure it was -fno-PIC code with a R_X86_64_32 or R_X86_64_64
> referencing gmtime.  And yes, -fno-PIC -mcmodel=medium will generate
> R_X86_64_32 in a testcase like
>
> extern int f1 (int);
> extern int f2 (int);
> int foo (int what, int val) { return (what ? f1 : f2) (val); }
>
> So why exclude R_X86_64_32 with an ABI_64_P test before setting
> func_pointer_ref in elf_x86_64_scan_relocs?

The whole condition is

                 /* At run-time, R_X86_64_64 can be resolved for both
                     x86-64 and x32. But R_X86_64_32 and R_X86_64_32S
                     can only be resolved for x32.  Function pointer
                     reference doesn't need PLT for pointer equality.  */
                  if ((sec->flags & SEC_READONLY) == 0
                      && (r_type == R_X86_64_64
                          || (!ABI_64_P (abfd)
                              && (r_type == R_X86_64_32
                                  || r_type == R_X86_64_32S))))
                    func_pointer_ref = true;

This is for relocations in data section (not readonly).   The function
pointer reference must use R_X86_64_64 if ABI_64_P is true.

-- 
H.J.

      reply	other threads:[~2022-08-24 14:37 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-23 17:34 H.J. Lu
2022-08-24  0:16 ` Alan Modra
2022-08-24  1:40   ` Alan Modra
2022-08-24 14:36     ` H.J. Lu [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=CAMe9rOoKGWuMBHCgt-U+Cu8eW6AOWHiG9qCS2rVYfXr1a9y_Mg@mail.gmail.com \
    --to=hjl.tools@gmail.com \
    --cc=amodra@gmail.com \
    --cc=binutils@sourceware.org \
    /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).