PR26978, Inconsistency for strong foo@v1 and weak foo@@v1

Prior to this patch
  ld -shared --version-script=pr26979.ver pr26978a.o pr26978b.o
results in
  ld: pr26978b.o: in function `foo_v1':
  (.text+0x0): multiple definition of `foo@v1'
  ld: pr26978b.o:(*IND*+0x0): multiple definition of `foo'
while
  ld -shared --version-script=pr26979.ver pr26978b.o pr26978a.o
results in no error, but some odd dynamic symbols.
  ... 0 NOTYPE  GLOBAL DEFAULT    7 foo@v1
  ... 0 NOTYPE  WEAK   DEFAULT    7 foo@@v1

When linking an undecorated reference to foo against such a shared
library, ld complains about multiple definitions of foo@v1 while gold
creates a dynamic reference to foo@v1.  That results in foo@v1 being
used at runtime.

While we could error in both cases, it is reasonable to say foo@v1 and
foo@@v1 are in fact the same symbol.  (Same name, same version.  The
only real difference is that foo@@v1 satisfies a reference to plain
foo, while foo@v1 does not.)  Just as merging a weak undecorated sym
with a strong sym results in the strong sym prevailing, so should the
strong foo@v1 prevail.  And since there is a definition that satisfies
plain foo, the foo@@v1 variety of dynamic symbol should be emitted at
the foo@v1 value.  That makes the testcase that currently links
continue to produce a shared library, and that shared library can now
be used by both ld and gold with the same runtime behaviour as when
using gold with the odd dynamic symbol library.

bfd/
	PR 26978
	* elflink.c (_bfd_elf_add_default_symbol): Handle the case where
	a new weak sym@@ver should be overridden by an existing sym@ver.
	(elf_link_add_object_symbols): Don't _bfd_elf_add_default_symbol
	for a new weak sym@ver when sym@@ver already exists.
	* linker.c (link_action): Choose MIND for previous indirect,
	current def, rather than MDEF.
	(_bfd_generic_link_add_one_symbol <MIND>): Handle redefinition of
	weak indirect symbol.
ld/
	* testsuite/ld-elf/pr26978a.d,
	* testsuite/ld-elf/pr26978a.s,
	* testsuite/ld-elf/pr26978b.d,
	* testsuite/ld-elf/pr26978b.s: New tests.
This commit is contained in:
Alan Modra 2020-12-02 13:03:23 +10:30
parent 099ee20ac3
commit 726d7d1ecf
8 changed files with 112 additions and 33 deletions

View File

@ -1,3 +1,15 @@
2020-12-04 Alan Modra <amodra@gmail.com>
PR 26978
* elflink.c (_bfd_elf_add_default_symbol): Handle the case where
a new weak sym@@ver should be overridden by an existing sym@ver.
(elf_link_add_object_symbols): Don't _bfd_elf_add_default_symbol
for a new weak sym@ver when sym@@ver already exists.
* linker.c (link_action): Choose MIND for previous indirect,
current def, rather than MDEF.
(_bfd_generic_link_add_one_symbol <MIND>): Handle redefinition of
weak indirect symbol.
2020-12-01 Nelson Chu <nelson.chu@sifive.com>
* elfxx-riscv.c (riscv_parse_prefixed_ext): Use riscv_compare_subsets

View File

@ -2078,9 +2078,26 @@ _bfd_elf_add_default_symbol (bfd *abfd,
return FALSE;
if (skip)
return TRUE;
if (override)
{
if (!dynamic
&& h->root.type == bfd_link_hash_defweak
&& hi->root.type == bfd_link_hash_defined)
{
/* We are handling a weak sym@@ver and attempting to define
a weak sym@ver, but _bfd_elf_merge_symbol said to skip the
new weak sym@ver because there is already a strong sym@ver.
However, sym@ver and sym@@ver are really the same symbol.
The existing strong sym@ver ought to override sym@@ver. */
h->root.type = bfd_link_hash_defined;
h->root.u.def.section = hi->root.u.def.section;
h->root.u.def.value = hi->root.u.def.value;
hi->root.type = bfd_link_hash_indirect;
hi->root.u.i.link = &h->root;
}
else
return TRUE;
}
else if (override)
{
/* Here SHORTNAME is a versioned name, so we don't expect to see
the type of override we do in the case above unless it is
@ -2091,6 +2108,7 @@ _bfd_elf_add_default_symbol (bfd *abfd,
/* xgettext:c-format */
(_("%pB: unexpected redefinition of indirect versioned symbol `%s'"),
abfd, shortname);
return TRUE;
}
else
{
@ -2100,37 +2118,36 @@ _bfd_elf_add_default_symbol (bfd *abfd,
bfd_ind_section_ptr, 0, name, FALSE, collect, &bh)))
return FALSE;
hi = (struct elf_link_hash_entry *) bh;
}
/* If there is a duplicate definition somewhere, then HI may not
point to an indirect symbol. We will have reported an error
to the user in that case. */
/* If there is a duplicate definition somewhere, then HI may not
point to an indirect symbol. We will have reported an error
to the user in that case. */
if (hi->root.type == bfd_link_hash_indirect)
{
(*bed->elf_backend_copy_indirect_symbol) (info, h, hi);
h->ref_dynamic_nonweak |= hi->ref_dynamic_nonweak;
hi->dynamic_def |= h->dynamic_def;
if (hi->root.type == bfd_link_hash_indirect)
/* If we first saw a reference to @VER symbol with
non-default visibility, merge that visibility to the
@@VER symbol. */
elf_merge_st_other (abfd, h, hi->other, sec, TRUE, dynamic);
/* See if the new flags lead us to realize that the symbol
must be dynamic. */
if (! *dynsym)
{
(*bed->elf_backend_copy_indirect_symbol) (info, h, hi);
h->ref_dynamic_nonweak |= hi->ref_dynamic_nonweak;
hi->dynamic_def |= h->dynamic_def;
/* If we first saw a reference to @VER symbol with
non-default visibility, merge that visibility to the
@@VER symbol. */
elf_merge_st_other (abfd, h, hi->other, sec, TRUE, dynamic);
/* See if the new flags lead us to realize that the symbol
must be dynamic. */
if (! *dynsym)
if (! dynamic)
{
if (! dynamic)
{
if (! bfd_link_executable (info)
|| hi->ref_dynamic)
*dynsym = TRUE;
}
else
{
if (hi->ref_regular)
*dynsym = TRUE;
}
if (! bfd_link_executable (info)
|| hi->ref_dynamic)
*dynsym = TRUE;
}
else
{
if (hi->ref_regular)
*dynsym = TRUE;
}
}
}
@ -5060,8 +5077,10 @@ elf_link_add_object_symbols (bfd *abfd, struct bfd_link_info *info)
/* Check to see if we need to add an indirect symbol for
the default name. */
if (definition
|| (!override && h->root.type == bfd_link_hash_common))
if ((definition
|| (!override && h->root.type == bfd_link_hash_common))
&& !(hi != h
&& hi->versioned == versioned_hidden))
if (!_bfd_elf_add_default_symbol (abfd, info, h, name, isym,
sec, value, &old_bfd, &dynsym))
goto error_free_vers;

View File

@ -1301,7 +1301,7 @@ static const enum link_action link_action[8][8] =
/* current\prev new undef undefw def defw com indr warn */
/* UNDEF_ROW */ {UND, NOACT, UND, REF, REF, NOACT, REFC, WARNC },
/* UNDEFW_ROW */ {WEAK, NOACT, NOACT, REF, REF, NOACT, REFC, WARNC },
/* DEF_ROW */ {DEF, DEF, DEF, MDEF, DEF, CDEF, MDEF, CYCLE },
/* DEF_ROW */ {DEF, DEF, DEF, MDEF, DEF, CDEF, MIND, CYCLE },
/* DEFW_ROW */ {DEFW, DEFW, DEFW, NOACT, NOACT, NOACT, NOACT, CYCLE },
/* COMMON_ROW */ {COM, COM, COM, CREF, COM, BIG, REFC, WARNC },
/* INDR_ROW */ {IND, IND, IND, MDEF, IND, CIND, MIND, CYCLE },
@ -1672,6 +1672,17 @@ _bfd_generic_link_add_one_symbol (struct bfd_link_info *info,
case MIND:
/* Multiple indirect symbols. This is OK if they both point
to the same symbol. */
if (h->u.i.link->type == bfd_link_hash_defweak)
{
/* It is also OK to redefine a symbol that indirects to
a weak definition. So for sym@ver -> sym@@ver where
sym@@ver is weak and we have a new strong sym@ver,
redefine sym@@ver. Of course if there exists
sym -> sym@@ver then this also redefines sym. */
h = h->u.i.link;
cycle = TRUE;
break;
}
if (strcmp (h->u.i.link->root.string, string) == 0)
break;
/* Fall through. */

View File

@ -1,3 +1,10 @@
2020-12-04 Alan Modra <amodra@gmail.com>
* testsuite/ld-elf/pr26978a.d,
* testsuite/ld-elf/pr26978a.s,
* testsuite/ld-elf/pr26978b.d,
* testsuite/ld-elf/pr26978b.s: New tests.
2020-12-03 Andreas Krebbel <krebbel@linux.ibm.com>
* testsuite/ld-s390/tlsbin_64.dd: The newly added jgnop mnemonic

View File

@ -0,0 +1,11 @@
#source: pr26978a.s
#source: pr26978b.s
#target: [check_shared_lib_support]
#as:
#ld: -shared --version-script=pr26979.ver
#readelf: -sW
#failif
#...
.*foo@v1
#pass

View File

@ -0,0 +1,4 @@
.weak foo
.symver foo, foo@@@v1
foo:
.octa 0

View File

@ -0,0 +1,11 @@
#source: pr26978b.s
#source: pr26978a.s
#target: [check_shared_lib_support]
#as:
#ld: -shared --version-script=pr26979.ver
#readelf: -sW
#failif
#...
.*foo@v1
#pass

View File

@ -0,0 +1,4 @@
.globl foo_v1
.symver foo_v1, foo@v1
foo_v1:
.octa 0