This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: 32-bit PowerPC sdata linker problem
- From: Will Newton <will dot newton at linaro dot org>
- To: "binutils at sourceware dot org" <binutils at sourceware dot org>, Alan Modra <amodra at gmail dot com>
- Date: Mon, 16 Jun 2014 12:21:23 +0100
- Subject: Re: 32-bit PowerPC sdata linker problem
- Authentication-results: sourceware.org; auth=none
- References: <53918356 dot 3040102 at embedded-brains dot de> <20140606105420 dot GH5592 at bubble dot grove dot modra dot org> <5391A4BF dot 2030308 at embedded-brains dot de> <20140606121659 dot GJ5592 at bubble dot grove dot modra dot org> <20140607124711 dot GO5592 at bubble dot grove dot modra dot org>
On 7 June 2014 13:47, Alan Modra <amodra@gmail.com> wrote:
Hi Alan,
> On Fri, Jun 06, 2014 at 09:46:59PM +0930, Alan Modra wrote:
>> Ah, light dawns. I'm guessing you still have a definition for
>> _SDA_BASE_ in your linker script, but using PROVIDE. Due to the way I
>> implemented the automatic define of _SDA_BASE_, PROVIDE in a linker
>> script won't override the automatic define. That's a bug.
>
> Fixed like this. Will fail to work if crt1.o happens to have a
> .sdata section, due to the necessity of defining _SDA_BASE_ on the
> first .sdata section.
>
> * ldexp.c (exp_fold_tree_1 <etree_provide>): Make PROVIDEd
> linker script symbol value override a built-in linker symbol.
> diff --git a/ld/ldexp.c b/ld/ldexp.c
> index d573fb7..5c4f8dd 100644
> --- a/ld/ldexp.c
> +++ b/ld/ldexp.c
> @@ -1001,7 +1001,10 @@ exp_fold_tree_1 (etree_type *tree)
> if (h == NULL
> || (h->type != bfd_link_hash_new
> && h->type != bfd_link_hash_undefined
> - && h->type != bfd_link_hash_common))
> + && h->type != bfd_link_hash_common
> + && !(h->type == bfd_link_hash_defined
> + && (h->u.def.section->flags
> + & SEC_LINKER_CREATED) != 0)))
> {
> /* Do nothing. The symbol was never referenced, or was
> defined by some object. */
This patch seems to break static ifunc links on ARM. I believe it is
caused by the default linker script providing symbols twice
(__rel_iplt_start, __rel_iplt_end) and the second definition now
overriding the first. At first glance that would appear to be sane
behaviour, but with the current linker script the result is a broken
binary. Any ideas on what the best fix would be?
Thanks,
.rel.dyn :
{
*(.rel.init)
*(.rel.text .rel.text.* .rel.gnu.linkonce.t.*)
*(.rel.fini)
*(.rel.rodata .rel.rodata.* .rel.gnu.linkonce.r.*)
*(.rel.data.rel.ro .rel.data.rel.ro.* .rel.gnu.linkonce.d.rel.ro.*)
*(.rel.data .rel.data.* .rel.gnu.linkonce.d.*)
*(.rel.tdata .rel.tdata.* .rel.gnu.linkonce.td.*)
*(.rel.tbss .rel.tbss.* .rel.gnu.linkonce.tb.*)
*(.rel.ctors)
*(.rel.dtors)
*(.rel.got)
*(.rel.bss .rel.bss.* .rel.gnu.linkonce.b.*)
PROVIDE_HIDDEN (__rel_iplt_start = .);
*(.rel.iplt)
PROVIDE_HIDDEN (__rel_iplt_end = .);
PROVIDE_HIDDEN (__rela_iplt_start = .);
PROVIDE_HIDDEN (__rela_iplt_end = .);
}
.rela.dyn :
{
*(.rela.init)
*(.rela.text .rela.text.* .rela.gnu.linkonce.t.*)
*(.rela.fini)
*(.rela.rodata .rela.rodata.* .rela.gnu.linkonce.r.*)
*(.rela.data .rela.data.* .rela.gnu.linkonce.d.*)
*(.rela.tdata .rela.tdata.* .rela.gnu.linkonce.td.*)
*(.rela.tbss .rela.tbss.* .rela.gnu.linkonce.tb.*)
*(.rela.ctors)
*(.rela.dtors)
*(.rela.got)
*(.rela.bss .rela.bss.* .rela.gnu.linkonce.b.*)
PROVIDE_HIDDEN (__rel_iplt_start = .);
PROVIDE_HIDDEN (__rel_iplt_end = .);
PROVIDE_HIDDEN (__rela_iplt_start = .);
*(.rela.iplt)
PROVIDE_HIDDEN (__rela_iplt_end = .);
--
Will Newton
Toolchain Working Group, Linaro