public inbox for binutils@sourceware.org
 help / color / mirror / Atom feed
From: Fangrui Song <i@maskray.me>
To: binutils@sourceware.org,
	 Adhemerval Zanella <adhemerval.zanella@linaro.org>,
	Szabolcs Nagy <szabolcs.nagy@arm.com>,
	Nick Clifton <nickc@redhat.com>
Subject: Re: [PATCH] aarch64: Disallow copy relocations on protected data
Date: Fri, 27 May 2022 15:43:46 -0700	[thread overview]
Message-ID: <DS7PR12MB576504ED5C63D87C233B58D6CBD89@DS7PR12MB5765.namprd12.prod.outlook.com> (raw)
In-Reply-To: <20220524051017.2244288-1-i@maskray.me>

On Mon, May 23, 2022 at 10:10 PM Fangrui Song <i@maskray.me> wrote:
>
> The behavior matches ld.lld which has been used on many aarch64 OSes.
>
> Rationale: copy relocations indicate that the executable and the shared object
> would access different copies, if the shared object does not reference the
> variable with GLOB_DAT.
>
> Note: x86 requires GNU_PROPERTY_NO_COPY_ON_PROTECTED to have the error.
> This is to largely due to GCC 5's
> "x86-64: Optimize access to globals in PIE with copy reloc" which started to use
> direct access relocations for external data symbols in -fpie mode.
>
> GCC's aarch64 port does not have the change. Nowadays with most builds switching
> to -fpie/-fpic, aarch64 mostly doesn't need to worry about copy relocations. So
> for aarch64 we simply don't check GNU_PROPERTY_NO_COPY_ON_PROTECTED.

The diagnostic is requested on
https://sourceware.org/pipermail/libc-alpha/2022-May/139014.html

> ---
>  bfd/elfnn-aarch64.c                           | 28 ++++++++++++++++++-
>  ld/testsuite/ld-aarch64/aarch64-elf.exp       |  9 ++++++
>  .../ld-aarch64/copy-reloc-protected.d         |  2 ++
>  ld/testsuite/ld-aarch64/protected.s           |  8 ++++++
>  4 files changed, 46 insertions(+), 1 deletion(-)
>  create mode 100644 ld/testsuite/ld-aarch64/copy-reloc-protected.d
>  create mode 100644 ld/testsuite/ld-aarch64/protected.s
>
> diff --git a/bfd/elfnn-aarch64.c b/bfd/elfnn-aarch64.c
> index 4926bab9cf2..8896b0f78dd 100644
> --- a/bfd/elfnn-aarch64.c
> +++ b/bfd/elfnn-aarch64.c
> @@ -2579,6 +2579,9 @@ struct elf_aarch64_link_hash_entry
>       this symbol.  */
>    unsigned int got_type;
>
> +  /* TRUE if symbol is defined as a protected symbol.  */
> +  unsigned int def_protected : 1;
> +
>    /* A pointer to the most recently used stub hash entry against this
>       symbol.  */
>    struct elf_aarch64_stub_hash_entry *stub_cache;
> @@ -2855,9 +2858,16 @@ elfNN_aarch64_copy_indirect_symbol (struct bfd_link_info *info,
>  static void
>  elfNN_aarch64_merge_symbol_attribute (struct elf_link_hash_entry *h,
>                                       unsigned int st_other,
> -                                     bool definition ATTRIBUTE_UNUSED,
> +                                     bool definition,
>                                       bool dynamic ATTRIBUTE_UNUSED)
>  {
> +  if (definition)
> +    {
> +      struct elf_aarch64_link_hash_entry *eh
> +       = (struct elf_aarch64_link_hash_entry *) h;
> +      eh->def_protected = ELF_ST_VISIBILITY (st_other) == STV_PROTECTED;
> +    }
> +
>    unsigned int isym_sto = st_other & ~ELF_ST_VISIBILITY (-1);
>    unsigned int h_sto = h->other & ~ELF_ST_VISIBILITY (-1);
>
> @@ -8701,6 +8711,22 @@ elfNN_aarch64_allocate_dynrelocs (struct elf_link_hash_entry *h, void *inf)
>    if (h->dyn_relocs == NULL)
>      return true;
>
> +  for (p = h->dyn_relocs; p != NULL; p = p->next)
> +    if (eh->def_protected)
> +      {
> +       /* Disallow copy relocations against protected symbol.  */
> +       asection *s = p->sec->output_section;
> +       if (s != NULL && (s->flags & SEC_READONLY) != 0)
> +         {
> +           info->callbacks->einfo
> +               /* xgettext:c-format */
> +               (_ ("%F%P: %pB: copy relocation against non-copyable "
> +                   "protected symbol `%s'\n"),
> +                p->sec->owner, h->root.root.string);
> +           return false;
> +         }
> +      }
> +
>    /* In the shared -Bsymbolic case, discard space allocated for
>       dynamic pc-relative relocs against symbols which turn out to be
>       defined in regular objects.  For the normal shared case, discard
> diff --git a/ld/testsuite/ld-aarch64/aarch64-elf.exp b/ld/testsuite/ld-aarch64/aarch64-elf.exp
> index 64476f111e0..31162277bd9 100644
> --- a/ld/testsuite/ld-aarch64/aarch64-elf.exp
> +++ b/ld/testsuite/ld-aarch64/aarch64-elf.exp
> @@ -407,6 +407,8 @@ set aarch64elflinktests {
>      {copy-reloc-exe-2.s} {{objdump -R copy-reloc-2.d}} "copy-reloc-2"}
>    {"ld-aarch64/exe with copy relocation elimination" "-e0 tmpdir/copy-reloc-so.so" "" ""
>      {copy-reloc-exe-eliminate.s} {{objdump -R copy-reloc-eliminate.d}} "copy-reloc-elimination"}
> +  {"Build .so with protected data" "-shared" "" "" {protected.s}
> +    {} "protected.so"}
>    {"ld-aarch64/so with global func" "-shared" "" "" {func-in-so.s}
>      {} "func-in-so.so"}
>    {"ld-aarch64/func sym hash opt for exe"
> @@ -416,8 +418,15 @@ set aarch64elflinktests {
>      {} "libbti-plt-so.so"}
>  }
>
> +set aarch64elfcclinktests [list \
> +  [list "copy relocation on protected data" \
> +    "-no-pie tmpdir/copy-reloc-exe.o tmpdir/protected.so" "" \
> +    {} {{error_output copy-reloc-protected.d}} "copy-reloc-protected"]
> +]
> +
>  if [check_shared_lib_support] {
>      run_ld_link_tests $aarch64elflinktests
> +    run_cc_link_tests $aarch64elfcclinktests
>  }
>
>  run_dump_test "bti-plt-3"
> diff --git a/ld/testsuite/ld-aarch64/copy-reloc-protected.d b/ld/testsuite/ld-aarch64/copy-reloc-protected.d
> new file mode 100644
> index 00000000000..99a356a3df6
> --- /dev/null
> +++ b/ld/testsuite/ld-aarch64/copy-reloc-protected.d
> @@ -0,0 +1,2 @@
> +.*: tmpdir/copy-reloc-exe.o: copy relocation against non-copyable protected symbol `global_a'
> +#...
> diff --git a/ld/testsuite/ld-aarch64/protected.s b/ld/testsuite/ld-aarch64/protected.s
> new file mode 100644
> index 00000000000..eb3fb402dc4
> --- /dev/null
> +++ b/ld/testsuite/ld-aarch64/protected.s
> @@ -0,0 +1,8 @@
> +.global global_a
> +.protected global_a
> +.type global_a, %object
> +.size global_a, 4
> +
> +.data
> +global_a:
> +.word 0xcafedead
> --
> 2.36
>

  reply	other threads:[~2022-05-27 22:51 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-24  5:10 Fangrui Song
2022-05-27 22:43 ` Fangrui Song [this message]
     [not found] ` <CAN30aBHDJM4KJAJJ_=4=vKgJL-yow2roUEUa7LFaAFaivkVswg@mail.gmail.com>
2022-06-01  5:53   ` PING: " Fangrui Song
     [not found]   ` <CAN30aBH68jT8ArEPEx_X-zVzuOmN3u_KV5HKc-Vgc2tVohJADw@mail.gmail.com>
2022-06-15 22:28     ` Fangrui Song
2022-06-21 19:27       ` Fangrui Song
     [not found]       ` <DS7PR12MB57658D646BB4944D4BECAF84CBB39@DS7PR12MB5765.namprd12.prod.outlook.com>
2022-06-22 10:55         ` Nick Clifton
2022-06-22 11:15           ` Szabolcs Nagy
2022-07-01 11:16             ` Matthias Klose

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=DS7PR12MB576504ED5C63D87C233B58D6CBD89@DS7PR12MB5765.namprd12.prod.outlook.com \
    --to=i@maskray.me \
    --cc=adhemerval.zanella@linaro.org \
    --cc=binutils@sourceware.org \
    --cc=nickc@redhat.com \
    --cc=szabolcs.nagy@arm.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).