Re: [PATCH 1/3] uprobes/core: Make instruction tables volatile.

From: Ingo Molnar
Date: Wed Feb 22 2012 - 04:47:42 EST



* Srikar Dronamraju <srikar@xxxxxxxxxxxxxxxxxx> wrote:

> From: Srikar Dronamraju <srikar@xxxxxxxxxxxxxxxxxx>
>
> Some versions of gcc spits a warning about the asm operand for test_bit and
> also causes the first long of the instruction table to be output.
>
> Fix is similar to 7115e3fc on arch/x86/kernel/kprobes.c
>
> Signed-off-by: Srikar Dronamraju <srikar@xxxxxxxxxxxxxxxxxx>
> ---
> arch/x86/kernel/uprobes.c | 61 ++++++++++++++++++++++++---------------------
> 1 files changed, 32 insertions(+), 29 deletions(-)
>
> diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
> index cf2a184..13d616d 100644
> --- a/arch/x86/kernel/uprobes.c
> +++ b/arch/x86/kernel/uprobes.c
> @@ -53,34 +53,12 @@
> (bc##UL << 0xc)|(bd##UL << 0xd)|(be##UL << 0xe)|(bf##UL << 0xf)) \
> << (row % 32))
>
> -#ifdef CONFIG_X86_64
> -static u32 good_insns_64[256 / 32] = {
> - /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> - /* ---------------------------------------------- */
> - W(0x00, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) | /* 00 */
> - W(0x10, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) , /* 10 */
> - W(0x20, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) | /* 20 */
> - W(0x30, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) , /* 30 */
> - W(0x40, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) | /* 40 */
> - W(0x50, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 50 */
> - W(0x60, 0, 0, 0, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0) | /* 60 */
> - W(0x70, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 70 */
> - W(0x80, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 80 */
> - W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 90 */
> - W(0xa0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* a0 */
> - W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* b0 */
> - W(0xc0, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0) | /* c0 */
> - W(0xd0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* d0 */
> - W(0xe0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0) | /* e0 */
> - W(0xf0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1) /* f0 */
> - /* ---------------------------------------------- */
> - /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> -};
> -#endif
> -
> -/* Good-instruction tables for 32-bit apps */
> -
> -static u32 good_insns_32[256 / 32] = {
> +/*
> + * Good-instruction tables for 32-bit apps. This is non-const and volatile
> + * to keep gcc from statically optimizing it out, as variable_test_bit makes
> + * some versions of gcc to think only *(unsigned long*) is used.
> + */
> +static volatile u32 good_insns_32[256 / 32] = {
> /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> /* ---------------------------------------------- */
> W(0x00, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 0) | /* 00 */
> @@ -104,7 +82,7 @@ static u32 good_insns_32[256 / 32] = {
> };
>
> /* Using this for both 64-bit and 32-bit apps */
> -static u32 good_2byte_insns[256 / 32] = {
> +static volatile u32 good_2byte_insns[256 / 32] = {
> /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> /* ---------------------------------------------- */
> W(0x00, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1) | /* 00 */
> @@ -127,6 +105,31 @@ static u32 good_2byte_insns[256 / 32] = {
> /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> };
>
> +#ifdef CONFIG_X86_64
> +/* Good-instruction tables for 64-bit apps */
> +static volatile u32 good_insns_64[256 / 32] = {
> + /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> + /* ---------------------------------------------- */
> + W(0x00, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) | /* 00 */
> + W(0x10, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) , /* 10 */
> + W(0x20, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) | /* 20 */
> + W(0x30, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0) , /* 30 */
> + W(0x40, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0) | /* 40 */
> + W(0x50, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 50 */
> + W(0x60, 0, 0, 0, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0) | /* 60 */
> + W(0x70, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 70 */
> + W(0x80, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* 80 */
> + W(0x90, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* 90 */
> + W(0xa0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) | /* a0 */
> + W(0xb0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* b0 */
> + W(0xc0, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0) | /* c0 */
> + W(0xd0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1) , /* d0 */
> + W(0xe0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0) | /* e0 */
> + W(0xf0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1) /* f0 */
> + /* ---------------------------------------------- */
> + /* 0 1 2 3 4 5 6 7 8 9 a b c d e f */
> +};
> +#endif

Hm, why was the table moved around? This makes it hard to see
whether it's just a single volatile that is added, or something
more.

Thanks,

Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/