Re: [PATCH v3 3/3] x86/entry: Clear registers for compat syscalls

From: Ingo Molnar
Date: Tue Feb 06 2018 - 02:26:25 EST



* Dan Williams <dan.j.williams@xxxxxxxxx> wrote:

> From: Andi Kleen <ak@xxxxxxxxxxxxxxx>
>
> At entry userspace may have populated registers with values that could
> be useful in a speculative execution attack. Clear them to minimize the
> kernel's attack surface.
>
> [djbw: interleave the clearing with setting up the stack ]
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: "H. Peter Anvin" <hpa@xxxxxxxxx>
> Cc: x86@xxxxxxxxxx
> Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> Signed-off-by: Andi Kleen <ak@xxxxxxxxxxxxxxx>
> Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
> ---
> arch/x86/entry/entry_64_compat.S | 30 ++++++++++++++++++++++++++++++
> 1 file changed, 30 insertions(+)
>
> diff --git a/arch/x86/entry/entry_64_compat.S b/arch/x86/entry/entry_64_compat.S
> index 98d5358e4041..fd65e016e413 100644
> --- a/arch/x86/entry/entry_64_compat.S
> +++ b/arch/x86/entry/entry_64_compat.S
> @@ -85,15 +85,25 @@ ENTRY(entry_SYSENTER_compat)
> pushq %rcx /* pt_regs->cx */
> pushq $-ENOSYS /* pt_regs->ax */
> pushq $0 /* pt_regs->r8 = 0 */
> + xorq %r8, %r8 /* nospec r8 */
> pushq $0 /* pt_regs->r9 = 0 */
> + xorq %r9, %r9 /* nospec r9 */
> pushq $0 /* pt_regs->r10 = 0 */
> + xorq %r10, %r10 /* nospec r10 */
> pushq $0 /* pt_regs->r11 = 0 */
> + xorq %r11, %r11 /* nospec r11 */
> pushq %rbx /* pt_regs->rbx */
> + xorl %ebx, %ebx /* nospec rbx */
> pushq %rbp /* pt_regs->rbp (will be overwritten) */
> + xorl %ebp, %ebp /* nospec rbp */
> pushq $0 /* pt_regs->r12 = 0 */
> + xorq %r12, %r12 /* nospec r12 */
> pushq $0 /* pt_regs->r13 = 0 */
> + xorq %r13, %r13 /* nospec r13 */
> pushq $0 /* pt_regs->r14 = 0 */
> + xorq %r14, %r14 /* nospec r14 */
> pushq $0 /* pt_regs->r15 = 0 */
> + xorq %r15, %r15 /* nospec r15 */
> cld
>
> /*
> @@ -214,15 +224,25 @@ GLOBAL(entry_SYSCALL_compat_after_hwframe)
> pushq %rbp /* pt_regs->cx (stashed in bp) */
> pushq $-ENOSYS /* pt_regs->ax */
> pushq $0 /* pt_regs->r8 = 0 */
> + xorq %r8, %r8 /* nospec r8 */
> pushq $0 /* pt_regs->r9 = 0 */
> + xorq %r9, %r9 /* nospec r9 */
> pushq $0 /* pt_regs->r10 = 0 */
> + xorq %r10, %r10 /* nospec r10 */
> pushq $0 /* pt_regs->r11 = 0 */
> + xorq %r11, %r11 /* nospec r11 */
> pushq %rbx /* pt_regs->rbx */
> + xorl %ebx, %ebx /* nospec rbx */
> pushq %rbp /* pt_regs->rbp (will be overwritten) */
> + xorl %ebp, %ebp /* nospec rbp */
> pushq $0 /* pt_regs->r12 = 0 */
> + xorq %r12, %r12 /* nospec r12 */
> pushq $0 /* pt_regs->r13 = 0 */
> + xorq %r13, %r13 /* nospec r13 */
> pushq $0 /* pt_regs->r14 = 0 */
> + xorq %r14, %r14 /* nospec r14 */
> pushq $0 /* pt_regs->r15 = 0 */
> + xorq %r15, %r15 /* nospec r15 */

I really love it how you've aligned the comment fields vertically - nice!

> /*
> * User mode is traced as though IRQs are on, and SYSENTER
> @@ -338,15 +358,25 @@ ENTRY(entry_INT80_compat)
> pushq %rcx /* pt_regs->cx */
> pushq $-ENOSYS /* pt_regs->ax */
> pushq $0 /* pt_regs->r8 = 0 */
> + xorq %r8, %r8 /* nospec r8 */
> pushq $0 /* pt_regs->r9 = 0 */
> + xorq %r9, %r9 /* nospec r9 */
> pushq $0 /* pt_regs->r10 = 0 */
> + xorq %r10, %r10 /* nospec r10 */
> pushq $0 /* pt_regs->r11 = 0 */
> + xorq %r11, %r11 /* nospec r11 */
> pushq %rbx /* pt_regs->rbx */
> + xorl %ebx, %ebx /* nospec rbx */
> pushq %rbp /* pt_regs->rbp */
> + xorl %ebp, %ebp /* nospec rbp */
> pushq %r12 /* pt_regs->r12 */
> + xorq %r12, %r12 /* nospec r12 */
> pushq %r13 /* pt_regs->r13 */
> + xorq %r13, %r13 /* nospec r13 */
> pushq %r14 /* pt_regs->r14 */
> + xorq %r14, %r14 /* nospec r14 */
> pushq %r15 /* pt_regs->r15 */
> + xorq %r15, %r15 /* nospec r15 */
> cld

BTW., these last two patches have changed *significantly* from Andi's original
patches that were submitted originally, so I changed them over to:

From: Dan Williams <dan.j.williams@xxxxxxxxx>
...
Originally-From: Andi Kleen <ak@xxxxxxxxxxxxxxx>
Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>

... to better show authorship history.

Please let me know if that's not OK.

Thanks,

Ingo