2014-01-10 15:49:49

by Taras Kondratiuk

[permalink] [raw]
Subject: Re: [PATCH 3/5] ARM: kprobes-test: use <asm/opcodes.h> for ARM instruction building

On 23 December 2013 18:19, Taras Kondratiuk <[email protected]> wrote:
> From: Ben Dooks <[email protected]>
>
> The kprobes test will build certain instructions incorrectly if building
> big endian as .word output gets endian-swapped by the linker. Change to
> using <asm/opcodes.h> and __inst_arm() to produce instructions.
>
> Signed-off-by: Ben Dooks <[email protected]>
> [[email protected]: fixed a few missed instructions]
> Signed-off-by: Taras Kondratiuk <[email protected]>
> ---
> arch/arm/kernel/kprobes-test-arm.c | 595 ++++++++++++++++++------------------
> 1 file changed, 298 insertions(+), 297 deletions(-)
>
> diff --git a/arch/arm/kernel/kprobes-test-arm.c b/arch/arm/kernel/kprobes-test-arm.c
> index 8393129..56be041 100644
> --- a/arch/arm/kernel/kprobes-test-arm.c
> +++ b/arch/arm/kernel/kprobes-test-arm.c

[snip]

> @@ -1180,43 +1181,43 @@ void kprobe_arm_test_cases(void)
> \
> TEST_COPROCESSOR( "stc"two" 0, cr0, [r15, #4]") \
> TEST_COPROCESSOR( "stc"two" 0, cr0, [r15, #-4]") \
> - TEST_UNSUPPORTED(".word 0x"cc"daf0001 @ stc"two" 0, cr0, [r15, #4]!") \
> - TEST_UNSUPPORTED(".word 0x"cc"d2f0001 @ stc"two" 0, cr0, [r15, #-4]!") \
> - TEST_UNSUPPORTED(".word 0x"cc"caf0001 @ stc"two" 0, cr0, [r15], #4") \
> - TEST_UNSUPPORTED(".word 0x"cc"c2f0001 @ stc"two" 0, cr0, [r15], #-4") \
> + TEST_UNSUPPORTED(__inst_arm(0x##ccdaf0001) " @ stc"two" 0, cr0, [r15, #4]!") \
> + TEST_UNSUPPORTED(__inst_arm(0x##ccd2f0001) " @ stc"two" 0, cr0, [r15, #-4]!") \
> + TEST_UNSUPPORTED(__inst_arm(0x##cccaf0001) " @ stc"two" 0, cr0, [r15], #4") \
> + TEST_UNSUPPORTED(__inst_arm(0x##ccc2f0001) " @ stc"two" 0, cr0, [r15], #-4") \

Oops. Should be something like __inst_arm(0x##cc##c2f0001).
My compiler truncated it to 32 bits, so these tests passed anyway.
I'll repost a fixed series.

--
Regards,
Taras Kondratiuk