[PATCH] ARM: module.lds: fix unwind metadata for merged .text sections

[PATCH] ARM: module.lds: fix unwind metadata for merged .text sections

[Egg12138]

Commit 1ba9f8979426 ("vmlinux.lds: Unify TEXT_MAIN, DATA_MAIN,
and related macros") made scripts/module.lds.S merge module input
.text.* sections into the output .text section.

On ARM, the paired unwind input sections keep their original names.  A
module can therefore contain .ARM.exidx.text.unlikely with sh_link
pointing at .text, while .text.unlikely no longer exists.

This is a valid ELF relationship, but ARM module_finalize() does not use
sh_link when registering module unwind tables.  It derives the target
text section from the exidx section name instead:

  .ARM.exidx.text.unlikely -> .text.unlikely

The lookup fails and the unwind table is not registered for the actual
.text range.  This can make module stack unwinding fail with:

  unwind: Index not found

Keep the ARM module unwind output names in sync with the text sections
that scripts/module.lds.S now produces.  Coalesce the .ARM.exidx/.ARM.extab
sections associated with .text.*, into the
stable output names expected by the existing ARM module unwind code.

Fixes: 1ba9f8979426 ("vmlinux.lds: Unify TEXT_MAIN, DATA_MAIN, and related macros")
Signed-off-by: Egg12138 <egg12138@foxmail.com>
---
 arch/arm/include/asm/module.lds.h | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/arm/include/asm/module.lds.h b/arch/arm/include/asm/module.lds.h
index 0e7cb4e314b4..dc9adf8fa50e 100644
--- a/arch/arm/include/asm/module.lds.h
+++ b/arch/arm/include/asm/module.lds.h
@@ -1,4 +1,12 @@
 /* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef CONFIG_ARCH_WANTS_MODULES_TEXT_SECTIONS
+SECTIONS {
+	.ARM.extab		0 : { *(.ARM.extab .ARM.extab.text .ARM.extab.text.[0-9a-zA-Z_]*) }
+	.ARM.exidx		0 : { *(.ARM.exidx .ARM.exidx.text .ARM.exidx.text.[0-9a-zA-Z_]*) }
+}
+#endif
+
 #ifdef CONFIG_ARM_MODULE_PLTS
 SECTIONS {
 	.plt : { BYTE(0) }
-- 
2.43.0

Re: [PATCH] ARM: module.lds: fix unwind metadata for merged .text sections

[Petr Pavlu]

On 5/21/26 5:57 PM, Egg12138 wrote:
> Commit 1ba9f8979426 ("vmlinux.lds: Unify TEXT_MAIN, DATA_MAIN,
> and related macros") made scripts/module.lds.S merge module input
> .text.* sections into the output .text section.
> 
> On ARM, the paired unwind input sections keep their original names.  A
> module can therefore contain .ARM.exidx.text.unlikely with sh_link
> pointing at .text, while .text.unlikely no longer exists.
> 
> This is a valid ELF relationship, but ARM module_finalize() does not use
> sh_link when registering module unwind tables.  It derives the target
> text section from the exidx section name instead:
> 
>   .ARM.exidx.text.unlikely -> .text.unlikely
> 
> The lookup fails and the unwind table is not registered for the actual
> .text range.  This can make module stack unwinding fail with:
> 
>   unwind: Index not found
> 
> Keep the ARM module unwind output names in sync with the text sections
> that scripts/module.lds.S now produces.  Coalesce the .ARM.exidx/.ARM.extab
> sections associated with .text.*, into the
> stable output names expected by the existing ARM module unwind code.
> 
> Fixes: 1ba9f8979426 ("vmlinux.lds: Unify TEXT_MAIN, DATA_MAIN, and related macros")
> Signed-off-by: Egg12138 <egg12138@foxmail.com>

A Signed-off-by tag should use a known identity, see
Documentation/process/submitting-patches.rst, Developer's Certificate of
Origin 1.1 [1].

> ---
>  arch/arm/include/asm/module.lds.h | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/arch/arm/include/asm/module.lds.h b/arch/arm/include/asm/module.lds.h
> index 0e7cb4e314b4..dc9adf8fa50e 100644
> --- a/arch/arm/include/asm/module.lds.h
> +++ b/arch/arm/include/asm/module.lds.h
> @@ -1,4 +1,12 @@
>  /* SPDX-License-Identifier: GPL-2.0 */
> +
> +#ifndef CONFIG_ARCH_WANTS_MODULES_TEXT_SECTIONS

The Arm architecture support doesn't select
CONFIG_ARCH_WANTS_MODULES_TEXT_SECTIONS, so this #ifndef is not really
necessary. I suggest leaving these linker definitions unconditional, or
alternatively use '#ifdef CONFIG_ARM_UNWIND'.

> +SECTIONS {
> +	.ARM.extab		0 : { *(.ARM.extab .ARM.extab.text .ARM.extab.text.[0-9a-zA-Z_]*) }
> +	.ARM.exidx		0 : { *(.ARM.exidx .ARM.exidx.text .ARM.exidx.text.[0-9a-zA-Z_]*) }
> +}
> +#endif
> +
>  #ifdef CONFIG_ARM_MODULE_PLTS
>  SECTIONS {
>  	.plt : { BYTE(0) }

Besides the above nits, the patch looks ok to me. Please don't forget to
put it in Russell's patch tracker [2].

[1] https://github.com/torvalds/linux/blob/v7.1-rc5/Documentation/process/submitting-patches.rst#developers-certificate-of-origin-11
[2] https://www.arm.linux.org.uk/developer/patches/

-- 
Thanks,
Petr

[PATCH v2] ARM: module.lds: fix unwind metadata for merged .text sections

[Egg12138]

From: Xiao Junzhe <egg12138@foxmail.com>

Commit 1ba9f8979426 ("vmlinux.lds: Unify TEXT_MAIN, DATA_MAIN,
and related macros") made scripts/module.lds.S merge module input
.text.* sections into the output .text section.

On ARM, the paired unwind input sections keep their original names.  A
module can therefore contain .ARM.exidx.text.unlikely with sh_link
pointing at .text, while .text.unlikely no longer exists.

This is a valid ELF relationship, but ARM module_finalize() does not use
sh_link when registering module unwind tables.  It derives the target
text section from the exidx section name instead:

  .ARM.exidx.text.unlikely -> .text.unlikely

The lookup fails and the unwind table is not registered for the actual
.text range.  This can make module stack unwinding fail with:

  unwind: Index not found

Keep the ARM module unwind output names in sync with the text sections
that scripts/module.lds.S now produces.  Coalesce the .ARM.exidx/.ARM.extab
section associated with .text.*, into the
stable output names expected by the existing ARM module unwind code.

Fixes: 1ba9f8979426 ("vmlinux.lds: Unify TEXT_MAIN, DATA_MAIN, and related macros")
Signed-off-by: Xiao Junzhe <egg12138@foxmail.com>
---
Changes in v2:
  - Quote ARM unwind input section patterns and split them across separate
    linker-script entries to avoid checkpatch false positives on wildcard
    syntax.
  - No functional change: the script still folds .ARM.extab.text.* and
    .ARM.exidx.text.* into the stable .ARM.extab/.ARM.exidx output sections.

 arch/arm/include/asm/module.lds.h | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/arch/arm/include/asm/module.lds.h b/arch/arm/include/asm/module.lds.h
index 0e7cb4e314b4..19ceb8cf235f 100644
--- a/arch/arm/include/asm/module.lds.h
+++ b/arch/arm/include/asm/module.lds.h
@@ -1,4 +1,20 @@
 /* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef CONFIG_ARM_UNWIND
+SECTIONS {
+	.ARM.extab		0 : {
+		*(.ARM.extab)
+		* (".ARM.extab.text")
+		* (".ARM.extab.text.[0-9a-zA-Z_]*")
+	}
+	.ARM.exidx		0 : {
+		*(.ARM.exidx)
+		* (".ARM.exidx.text")
+		* (".ARM.exidx.text.[0-9a-zA-Z_]*")
+	}
+}
+#endif
+
 #ifdef CONFIG_ARM_MODULE_PLTS
 SECTIONS {
 	.plt : { BYTE(0) }
-- 
2.43.0

Re: [PATCH v2] ARM: module.lds: fix unwind metadata for merged .text sections

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [High] The linker script fix is wrapped in an inverted configuration conditional (`#ifndef CONFIG_ARM_UNWIND`), applying the fix only when unwinding is disabled.
- [High] The wildcard patterns are enclosed in double quotes, which disables wildcard matching in GNU ld.
- [Low] Non-standard linker script syntax with a space between the wildcard asterisk and the section name list.
--

commit 165574ad9cf92e77eac788d811caa72c990e0b3d
Author: Xiao Junzhe <egg12138@foxmail.com>
Subject: ARM: module.lds: fix unwind metadata for merged .text sections

This commit attempts to fix a module stack unwinding failure on ARM by
coalescing .ARM.exidx and .ARM.extab input sections that belong to merged
.text.* sections. This is intended to keep the output names in sync with
what the ARM module unwind code expects.

> diff --git a/arch/arm/include/asm/module.lds.h b/arch/arm/include/asm/module.lds.h
> --- a/arch/arm/include/asm/module.lds.h
> +++ b/arch/arm/include/asm/module.lds.h
> @@ -1,4 +1,20 @@
>  /* SPDX-License-Identifier: GPL-2.0 */
> +
> +#ifndef CONFIG_ARM_UNWIND

[Severity: High]
Does this exclude the fix in the exact configuration it is intended for?

The patch aims to fix module stack unwinding, which is controlled by
CONFIG_ARM_UNWIND. Because this block is wrapped in an #ifndef, the section
coalescing will be skipped when unwinding is enabled, leaving the unwind
tables unmerged and the issue unresolved.

> +SECTIONS {
> +	.ARM.extab		0 : {
> +		*(.ARM.extab)
> +		* (".ARM.extab.text")
> +		* (".ARM.extab.text.[0-9a-zA-Z_]*")

[Severity: Low]
This isn't a bug, but there is a space between the asterisk wildcard and
the section name list. While GNU ld ignores this whitespace, standard kernel
linker script conventions typically omit it.

[Severity: High]
Will GNU ld expand the wildcard patterns when they are enclosed in double
quotes?

Enclosing a section name in double quotes causes wildcard characters like
the asterisk and brackets to lose their special meaning and be treated as
literal characters. The linker will try to match a section literally named
".ARM.extab.text.[0-9a-zA-Z_]*" instead of expanding the glob to match
sections like ".ARM.extab.text.unlikely".

> +	}
> +	.ARM.exidx		0 : {
> +		*(.ARM.exidx)
> +		* (".ARM.exidx.text")
> +		* (".ARM.exidx.text.[0-9a-zA-Z_]*")
> +	}
> +}
> +#endif

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/tencent_D9CDCA221D0AF5EDF260559734464D39B708@qq.com?part=1