Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1336820 > unrolled thread

[PATCH] ARM: move __kuser_cmpxchg{32,64} into .kprobes.text

Started byArnd Bergmann <arnd@arndb.de>
First post2016-02-17 23:30 +0100
Last post2016-02-18 00:20 +0100
Articles 2 — 2 participants

Back to article view | Back to linux.kernel


Contents

  [PATCH] ARM: move __kuser_cmpxchg{32,64} into .kprobes.text Arnd Bergmann <arnd@arndb.de> - 2016-02-17 23:30 +0100
    Re: [PATCH] ARM: move __kuser_cmpxchg{32,64} into .kprobes.text Nicolas Pitre <nicolas.pitre@linaro.org> - 2016-02-18 00:20 +0100

#1336820 — [PATCH] ARM: move __kuser_cmpxchg{32,64} into .kprobes.text

FromArnd Bergmann <arnd@arndb.de>
Date2016-02-17 23:30 +0100
Subject[PATCH] ARM: move __kuser_cmpxchg{32,64} into .kprobes.text
Message-ID<r3iLE-4sg-13@gated-at.bofh.it>
When kprobes is used, most of the entry-armv.S file is put into the .kprobes.text
section rather than .text, but the kuser_cmpxchg64_fixup and kuser_cmpxchg32_fixup
code is not, which can lead to a link error when extremely large kernels get
built (typically for randconfig):

arch/arm/kernel/built-in.o: In function `__dabt_usr':
:(.kprobes.text+0x47c): relocation truncated to fit: R_ARM_JUMP24 against `.text'
arch/arm/kernel/built-in.o: In function `__irq_usr':
:(.kprobes.text+0x4e4): relocation truncated to fit: R_ARM_JUMP24 against `.text'
arch/arm/kernel/built-in.o: In function `__fiq_usr':
:(.kprobes.text+0x740): relocation truncated to fit: R_ARM_JUMP24 against `.text'

This moves the two remaining functions into the same section as the rest,
to avoid the link error.

The current behavior has existed for many years and has not caused problems in
practice except for extreme randconfig builds, so the patch should not need to
get backported.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: b49c0f24cf67 ("[ARM] 4659/1: remove possibilities for spurious false negative with __kuser_cmpxchg")
Fixes: 785d3cd286f0 ("ARM kprobes: prevent some functions involved with kprobes from being probed")
---
 arch/arm/kernel/entry-armv.S | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index e2550500486d..030e43a0dce8 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -895,7 +895,11 @@ __kuser_cmpxchg64:				@ 0xffff0f60
 	rsbs	r0, r3, #0			@ set return val and C flag
 	ldmfd	sp!, {r4, r5, r6, pc}
 
+#ifdef CONFIG_KPROBES
+	.section	.kprobes.text,"ax",%progbits
+#else
 	.text
+#endif
 kuser_cmpxchg64_fixup:
 	@ Called from kuser_cmpxchg_fixup.
 	@ r4 = address of interrupted insn (must be preserved).
@@ -953,7 +957,11 @@ __kuser_cmpxchg:				@ 0xffff0fc0
 	rsbs	r0, r3, #0			@ set return val and C flag
 	usr_ret	lr
 
+#ifdef CONFIG_KPROBES
+	.section	.kprobes.text,"ax",%progbits
+#else
 	.text
+#endif
 kuser_cmpxchg32_fixup:
 	@ Called from kuser_cmpxchg_check macro.
 	@ r4 = address of interrupted insn (must be preserved).
-- 
2.7.0

[toc] | [next] | [standalone]


#1336870

FromNicolas Pitre <nicolas.pitre@linaro.org>
Date2016-02-18 00:20 +0100
Message-ID<r3jy1-527-5@gated-at.bofh.it>
In reply to#1336820
On Wed, 17 Feb 2016, Arnd Bergmann wrote:

> When kprobes is used, most of the entry-armv.S file is put into the .kprobes.text
> section rather than .text, but the kuser_cmpxchg64_fixup and kuser_cmpxchg32_fixup
> code is not, which can lead to a link error when extremely large kernels get
> built (typically for randconfig):
> 
> arch/arm/kernel/built-in.o: In function `__dabt_usr':
> :(.kprobes.text+0x47c): relocation truncated to fit: R_ARM_JUMP24 against `.text'
> arch/arm/kernel/built-in.o: In function `__irq_usr':
> :(.kprobes.text+0x4e4): relocation truncated to fit: R_ARM_JUMP24 against `.text'
> arch/arm/kernel/built-in.o: In function `__fiq_usr':
> :(.kprobes.text+0x740): relocation truncated to fit: R_ARM_JUMP24 against `.text'
> 
> This moves the two remaining functions into the same section as the rest,
> to avoid the link error.
> 
> The current behavior has existed for many years and has not caused problems in
> practice except for extreme randconfig builds, so the patch should not need to
> get backported.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> Fixes: b49c0f24cf67 ("[ARM] 4659/1: remove possibilities for spurious false negative with __kuser_cmpxchg")
> Fixes: 785d3cd286f0 ("ARM kprobes: prevent some functions involved with kprobes from being probed")

Stylewise it might be worth having something like:

#ifdef CONFIG_KPROBES
#define KPROBE_TEXT  .section .kprobes.text,"ax",%progbits
#else
#define KPROBE_TEXT .text
#endif

and then use that throughout.  Then...

Acked-by: Nicolas Pitre <nico@linaro.org>


> ---
>  arch/arm/kernel/entry-armv.S | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
> index e2550500486d..030e43a0dce8 100644
> --- a/arch/arm/kernel/entry-armv.S
> +++ b/arch/arm/kernel/entry-armv.S
> @@ -895,7 +895,11 @@ __kuser_cmpxchg64:				@ 0xffff0f60
>  	rsbs	r0, r3, #0			@ set return val and C flag
>  	ldmfd	sp!, {r4, r5, r6, pc}
>  
> +#ifdef CONFIG_KPROBES
> +	.section	.kprobes.text,"ax",%progbits
> +#else
>  	.text
> +#endif
>  kuser_cmpxchg64_fixup:
>  	@ Called from kuser_cmpxchg_fixup.
>  	@ r4 = address of interrupted insn (must be preserved).
> @@ -953,7 +957,11 @@ __kuser_cmpxchg:				@ 0xffff0fc0
>  	rsbs	r0, r3, #0			@ set return val and C flag
>  	usr_ret	lr
>  
> +#ifdef CONFIG_KPROBES
> +	.section	.kprobes.text,"ax",%progbits
> +#else
>  	.text
> +#endif
>  kuser_cmpxchg32_fixup:
>  	@ Called from kuser_cmpxchg_check macro.
>  	@ r4 = address of interrupted insn (must be preserved).
> -- 
> 2.7.0
> 
> 

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web