On 02/11/18 15:55, Wei Liu wrote:
> Skip building x86_64/compat/entry.S and put CONFIG_PV in
> x86_64/entry.S.
>
> Signed-off-by: Wei Liu <[email protected]>
> ---
> v3:
> 1. make CR4_PV32_RESTORE expand to nothing when !PV
> 2. use unconditional jmp and add assertions
>
> v2: new
> ---
> xen/arch/x86/x86_64/Makefile | 2 +-
> xen/arch/x86/x86_64/entry.S | 39 +++++++++++++++++++++++++++++++++-
> xen/include/asm-x86/asm_defns.h | 4 +++-
> 3 files changed, 43 insertions(+), 2 deletions(-)
>
> diff --git a/xen/arch/x86/x86_64/Makefile b/xen/arch/x86/x86_64/Makefile
> index f336a6a..4bfa148 100644
> --- a/xen/arch/x86/x86_64/Makefile
> +++ b/xen/arch/x86/x86_64/Makefile
> @@ -1,4 +1,4 @@
> -subdir-y += compat
> +subdir-$(CONFIG_PV) += compat
>
> obj-bin-y += entry.o
> obj-y += traps.o
> diff --git a/xen/arch/x86/x86_64/entry.S b/xen/arch/x86/x86_64/entry.S
> index 9b02899..975ed6a 100644
> --- a/xen/arch/x86/x86_64/entry.S
> +++ b/xen/arch/x86/x86_64/entry.S
> @@ -15,6 +15,18 @@
> #include <public/xen.h>
> #include <irq_vectors.h>
>
> +#ifndef NDEBUG
> +/* %rsp: struct cpu_user_regs */
> +#define ASSERT_INTERRUPTED_XEN_CONTEXT \
Could I suggest ASSERT_CONTEXT_IS_XEN which I think fits better in some
of the cases below?
> + testb $3, UREGS_cs(%rsp); \
> + jz 1f; \
> + ASSERT_FAILED("INTERRUPTED XEN CONTEXT"); \
This probably wants to be:
.macro ASSERT_CONTEXT_IS_XEN
#ifndef NDEBUG
testb $3, UREGS_cs(%rsp);
UNLIKELY_START(nz, ASSERT_XEN_\@);
ASSERT_FAILED("CONTEXT IS XEN");
__UNLIKELY_END(ASSERT_XEN_\@);
#endif
.endm
which will assemble in the fastpath to:
testb $3, UREGS_cs(%rsp);
jne <unlikely>
jmp restore_all_xen
And avoid a conditional jump in the expected path. (Code completely
untested.)
~Andrew
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
