[PATCH 0/2] Fix objtool warnings if LTO is enabled for LoongArch

[PATCH 0/2] Fix objtool warnings if LTO is enabled for LoongArch

[Tiezhu Yang]

The patch #1 should be a preparation for patch #2, that is to say,
the patch #2 is dependent on the patch #1, otherwise there is build
error if LTO is enabled after only applying patch #2.

With this series, most of warnings have been silenced, only remains
the following warning by now, it needs more analysis:

  vmlinux.o: warning: objtool: __efistub_efi_boot_kernel() falls through
  to next function __efistub_exit_boot_func()

How to reproduce:

  $ make ARCH=loongarch LLVM=1 clean defconfig
  $ scripts/config -d LTO_NONE -e LTO_CLANG_THIN
  $ make ARCH=loongarch LLVM=1 olddefconfig all

Tiezhu Yang (2):
  objtool/LoongArch: Get table size correctly if LTO is enabled
  LoongArch: Pass annotate-tablejump option if LTO is enabled

 arch/loongarch/Makefile                |  6 ++++++
 tools/objtool/arch/loongarch/special.c | 23 +++++++++++++++++++++++
 2 files changed, 29 insertions(+)

-- 
2.42.0

[PATCH 1/2] objtool/LoongArch: Get table size correctly if LTO is enabled

[Tiezhu Yang]

When compiling with LLVM and CONFIG_LTO_CLANG is set, there exist
many objtool warnings "sibling call from callable instruction with
modified stack frame".

For this special case, the related object file shows that there is
no generated relocation section '.rela.discard.tablejump_annotate'
for the table jump instruction jirl, thus objtool can not know that
what is the actual destination address.

It needs to do something on the LLVM side to make sure that there is
the relocation section '.rela.discard.tablejump_annotate' if LTO is
enabled, but in order to maintain compatibility for the current LLVM
compiler, this can be done in the kernel Makefile for now. Ensure it
is aware of linker with LTO, '--loongarch-annotate-tablejump' needs
to be passed via '-mllvm' to ld.lld.

Before doing the above changes, it should handle the special case of
the relocation section '.rela.discard.tablejump_annotate' to get the
correct table size first, otherwise there are many objtool warnings
and errors if LTO is enabled.

There are many different rodata for each function if LTO is enabled,
it is necessary to enhance get_rodata_table_size_by_table_annotate().

Reported-by: Nathan Chancellor <nathan@kernel.org>
Closes: https://lore.kernel.org/loongarch/20250731175655.GA1455142@ax162/
Fixes: b95f852d3af2 ("objtool/LoongArch: Add support for switch table")
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
---
 tools/objtool/arch/loongarch/special.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

diff --git a/tools/objtool/arch/loongarch/special.c b/tools/objtool/arch/loongarch/special.c
index e39f86d97002..a80b75f7b061 100644
--- a/tools/objtool/arch/loongarch/special.c
+++ b/tools/objtool/arch/loongarch/special.c
@@ -27,6 +27,7 @@ static void get_rodata_table_size_by_table_annotate(struct objtool_file *file,
 	struct table_info *next_table;
 	unsigned long tmp_insn_offset;
 	unsigned long tmp_rodata_offset;
+	bool is_valid_list = false;
 
 	rsec = find_section_by_name(file->elf, ".rela.discard.tablejump_annotate");
 	if (!rsec)
@@ -35,6 +36,12 @@ static void get_rodata_table_size_by_table_annotate(struct objtool_file *file,
 	INIT_LIST_HEAD(&table_list);
 
 	for_each_reloc(rsec, reloc) {
+		if (reloc->sym->sec->rodata)
+			continue;
+
+		if (strcmp(insn->sec->name, reloc->sym->sec->name))
+			continue;
+
 		orig_table = malloc(sizeof(struct table_info));
 		if (!orig_table) {
 			WARN("malloc failed");
@@ -49,6 +56,22 @@ static void get_rodata_table_size_by_table_annotate(struct objtool_file *file,
 
 		if (reloc_idx(reloc) + 1 == sec_num_entries(rsec))
 			break;
+
+		if (strcmp(insn->sec->name, (reloc + 1)->sym->sec->name)) {
+			list_for_each_entry(orig_table, &table_list, jump_info) {
+				if (orig_table->insn_offset == insn->offset) {
+					is_valid_list = true;
+					break;
+				}
+			}
+
+			if (!is_valid_list) {
+				list_del_init(&table_list);
+				continue;
+			}
+
+			break;
+		}
 	}
 
 	list_for_each_entry(orig_table, &table_list, jump_info) {
-- 
2.42.0

[PATCH 2/2] LoongArch: Pass annotate-tablejump option if LTO is enabled

[Tiezhu Yang]

When compiling with LLVM and CONFIG_LTO_CLANG is set, there exist
many objtool warnings "sibling call from callable instruction with
modified stack frame".

For this special case, the related object file shows that there is
no generated relocation section '.rela.discard.tablejump_annotate'
for the table jump instruction jirl, thus objtool can not know that
what is the actual destination address.

It needs to do something on the LLVM side to make sure that there is
the relocation section '.rela.discard.tablejump_annotate' if LTO is
enabled, but in order to maintain compatibility for the current LLVM
compiler, this can be done in the kernel Makefile for now. Ensure it
is aware of linker with LTO, '--loongarch-annotate-tablejump' needs
to be passed via '-mllvm' to ld.lld.

Note that it should also pass the compiler option -mannotate-tablejump
rather than only pass '-mllvm --loongarch-annotate-tablejump' to ld.lld
if LTO is enabled, otherwise there are no jump info for some table jump
instructions.

Reported-by: Nathan Chancellor <nathan@kernel.org>
Closes: https://lore.kernel.org/loongarch/20250731175655.GA1455142@ax162/
Fixes: e20ab7d454ee ("LoongArch: Enable jump table for objtool")
Co-developed-by: WANG Rui <wangrui@loongson.cn>
Signed-off-by: WANG Rui <wangrui@loongson.cn>
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
---
 arch/loongarch/Makefile | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/loongarch/Makefile b/arch/loongarch/Makefile
index b0703a4e02a2..6cdbcbbe730e 100644
--- a/arch/loongarch/Makefile
+++ b/arch/loongarch/Makefile
@@ -103,6 +103,12 @@ KBUILD_CFLAGS			+= $(call cc-option,-mthin-add-sub) $(call cc-option,-Wa$(comma)
 ifdef CONFIG_OBJTOOL
 ifdef CONFIG_CC_HAS_ANNOTATE_TABLEJUMP
 KBUILD_CFLAGS			+= -mannotate-tablejump
+# The annotate-tablejump option can not be passed to LLVM backend when LTO is enabled.
+# Ensure it is aware of linker with LTO, '--loongarch-annotate-tablejump' needs to be
+# passed via '-mllvm' to ld.lld.
+ifdef CONFIG_LTO_CLANG
+KBUILD_LDFLAGS			+= -mllvm --loongarch-annotate-tablejump
+endif
 else
 KBUILD_CFLAGS			+= -fno-jump-tables # keep compatibility with older compilers
 endif
-- 
2.42.0

Re: [PATCH 0/2] Fix objtool warnings if LTO is enabled for LoongArch

[Nathan Chancellor]

On Tue, Aug 12, 2025 at 09:27:14PM +0800, Tiezhu Yang wrote:
> The patch #1 should be a preparation for patch #2, that is to say,
> the patch #2 is dependent on the patch #1, otherwise there is build
> error if LTO is enabled after only applying patch #2.

Thanks, these two patches do indeed resolve most of the warnings that I
see.

Tested-by: Nathan Chancellor <nathan@kernel.org>

> With this series, most of warnings have been silenced, only remains
> the following warning by now, it needs more analysis:
> 
>   vmlinux.o: warning: objtool: __efistub_efi_boot_kernel() falls through
>   to next function __efistub_exit_boot_func()

Yes, I do see this one too. Odd, as efi_boot_kernel() ends in a
__noreturn function...

Cheers,
Nathan

Re: [PATCH 0/2] Fix objtool warnings if LTO is enabled for LoongArch

[Huacai Chen]

On Fri, Aug 15, 2025 at 7:01 AM Nathan Chancellor <nathan@kernel.org> wrote:
>
> On Tue, Aug 12, 2025 at 09:27:14PM +0800, Tiezhu Yang wrote:
> > The patch #1 should be a preparation for patch #2, that is to say,
> > the patch #2 is dependent on the patch #1, otherwise there is build
> > error if LTO is enabled after only applying patch #2.
>
> Thanks, these two patches do indeed resolve most of the warnings that I
> see.
>
> Tested-by: Nathan Chancellor <nathan@kernel.org>
>
> > With this series, most of warnings have been silenced, only remains
> > the following warning by now, it needs more analysis:
> >
> >   vmlinux.o: warning: objtool: __efistub_efi_boot_kernel() falls through
> >   to next function __efistub_exit_boot_func()
>
> Yes, I do see this one too. Odd, as efi_boot_kernel() ends in a
> __noreturn function...
But this one only exists for LTO?

Huacai

>
> Cheers,
> Nathan

Re: [PATCH 0/2] Fix objtool warnings if LTO is enabled for LoongArch

[Tiezhu Yang]

On 2025/8/16 下午10:59, Huacai Chen wrote:
> On Fri, Aug 15, 2025 at 7:01 AM Nathan Chancellor <nathan@kernel.org> wrote:
>>
>> On Tue, Aug 12, 2025 at 09:27:14PM +0800, Tiezhu Yang wrote:
>>> The patch #1 should be a preparation for patch #2, that is to say,
>>> the patch #2 is dependent on the patch #1, otherwise there is build
>>> error if LTO is enabled after only applying patch #2.
>>
>> Thanks, these two patches do indeed resolve most of the warnings that I
>> see.
>>
>> Tested-by: Nathan Chancellor <nathan@kernel.org>
>>
>>> With this series, most of warnings have been silenced, only remains
>>> the following warning by now, it needs more analysis:
>>>
>>>    vmlinux.o: warning: objtool: __efistub_efi_boot_kernel() falls through
>>>    to next function __efistub_exit_boot_func()
>>
>> Yes, I do see this one too. Odd, as efi_boot_kernel() ends in a
>> __noreturn function...
> But this one only exists for LTO?

Yes, this is true.

I have fixed this warning and the other new warnings locally,
will do more testing before sending patches.

Thanks,
Tiezhu