If code compiled with -fPIC -mtls-dialect=desc is linked into a PDE or
PIE, and the code refers to external DSO symbols, we can produce broken
link unit as check_relocs expects DESC => IE transition to happen and
emits a TLS IE entry in the GOT, but a too early "continue" in
relax_section actually jumps over the DESC => IE transition so the code
sequence is unchanged and still expecting a TLS descriptor (instead of
an IE entry) in the GOT.
The bug is already fixed in master branch by commit 5c3d09c1855b
("LoongArch: Optimize the relaxation process") so this fix is only
needed for the 2.43 branch.
Reported-by: Icenowy Zheng <uwu@icenowy.me>
Closes: https://gcc.gnu.org/PR118114
Tested-by: Icenowy Zheng <uwu@icenowy.me>
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
The inline plt code emitted by gcc is incompatible with the
linker/ld.so --tls-get-addr-optimize scheme. This is the runtime
optimisation where the first call to __tls_get_addr results in
__tls_get_addr updating the tls_index pair, then the special linker
stub using that to short-circuit second and subsequent calls for a
given tls symbol. Enabled by default when the linker sees
__tls_get_addr_opt is preseent, and enabled in ld.so when DT_PPC64_OPT
has PPC64_OPT_TLS set. Note that this is distinct from link-time tls
optimisation.
PR 32387
* elf64-ppc.c (ppc64_elf_check_relocs): Disable tls_get_addr_opt
on detecting inline plt calls to __tls_get_addr.
(cherry picked from commit 1686dc7079f1c03bdaffd2f779b92aa2b7ad97b5)
GCC trunk now defaults to -std=gnu23. We return false in a few places
which can't work when true/false are a proper type (_Bool). Return NULL
where appropriate instead of false. All callers handle this appropriately.
ChangeLog:
PR ld/32372
* pdb.c (add_stream): Return NULL.
(cherry picked from commit bf447eec6d7b69cba652127e2f38f102de38de39)
GCC trunk now defaults to -std=gnu23. We return false in a few places
which can't work when true/false are a proper type (_Bool). Return NULL
where appropriate instead of false. All callers handle this appropriately.
ChangeLog:
PR ld/32372
* prdbg.c (visibility_name): Return NULL.
(cherry picked from commit eeff15bc88b89abed1cdb4d3b1b2cc7b4cec6fe7)
static_assert is declared in C23 so we can't reuse that identifier:
* Define our own static_assert conditionally;
* Rename "static assert" hacks to _N as we do already in some places
to avoid a conflict.
ChangeLog:
PR ld/32372
* i386-gen.c (static_assert): Define conditionally.
* mips-formats.h (MAPPED_INT): Rename identifier.
(MAPPED_REG): Rename identifier.
(OPTIONAL_MAPPED_REG): Rename identifier.
* s390-opc.c (static_assert): Define conditionally.
(cherry picked from commit 8ebe62f3f0d27806b1bf69f301f5e188b4acd2b4)
GCC trunk now defaults to -std=gnu23. We return false in a few places
which can't work when true/false are a proper type (_Bool). Return NULL
where appropriate instead of false. All callers handle this appropriately.
ChangeLog:
PR ld/32372
* elf32-ppc.c (ppc_elf_tls_setup): Return NULL.
* elf32-xtensa.c (translate_reloc_bfd_fix): Ditto.
(translate_reloc): Ditto.
* elf64-ppc.c (update_local_sym_info): Ditto.
* mach-o.c (bfd_mach_o_lookup_uuid_command): Ditto.
* xsym.c (bfd_sym_read_name_table): Ditto.
(cherry picked from commit 931e4f9b09512bb9012204bb2e4060ab6d9b9b46)
