On Fri, Oct 20, 2023 at 04:13:13PM -0700, Sean Christopherson wrote: > On Fri, Oct 20, 2023, Pawan Gupta wrote: > > diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h > > index c55cc243592e..e1b623a27e1b 100644 > > --- a/arch/x86/include/asm/nospec-branch.h > > +++ b/arch/x86/include/asm/nospec-branch.h > > @@ -111,6 +111,24 @@ > > #define RESET_CALL_DEPTH_FROM_CALL > > #endif > > > > +/* > > + * Macro to execute VERW instruction to mitigate transient data sampling > > + * attacks such as MDS. On affected systems a microcode update overloaded VERW > > + * instruction to also clear the CPU buffers. > > + * > > + * Note: Only the memory operand variant of VERW clears the CPU buffers. To > > + * handle the case when VERW is executed after user registers are restored, use > > + * RIP to point the memory operand to a part NOPL instruction that contains > > + * __KERNEL_DS. > > + */ > > +#define __EXEC_VERW(m) verw _ASM_RIP(m) > > + > > +#define EXEC_VERW \ > > + __EXEC_VERW(551f); \ > > + /* nopl __KERNEL_DS(%rax) */ \ > > + .byte 0x0f, 0x1f, 0x80, 0x00, 0x00; \ > > +551: .word __KERNEL_DS; \ > > Why are there so many macro layers? Nothing jumps out to justfying two layers, > let alone three. I can't remember the exact reason, but I think the reason I added __EXEC_VERW() was because using EXEC_VERW() in a C function was leading to build error in the internal draft version. This version is not calling it from C, so yes I can get rid of the extra layer. > > /* > > * Fill the CPU return stack buffer. > > * > > @@ -329,6 +347,13 @@ > > #endif > > .endm > > > > +/* Clear CPU buffers before returning to user */ > > +.macro USER_CLEAR_CPU_BUFFERS > > + ALTERNATIVE "jmp .Lskip_verw_\@;", "", X86_FEATURE_USER_CLEAR_CPU_BUF > > + EXEC_VERW > > Rather than a NOP after VERW, why not something like this? > > /* Clear CPU buffers before returning to user */ > .macro USER_CLEAR_CPU_BUFFERS > ALTERNATIVE "jmp .Lskip_verw_\@;", "jmp .Ldo_verw_\@;", X86_FEATURE_USER_CLEAR_CPU_BUF > 551: .word __KERNEL_DS > .Ldo_verw_\@: verw _ASM_RIP(551b) > .Lskip_verw_\@: > .endm I wasn't comfortable adding a variable directly in the instruction stream because the CPU may interpret it wrongly. With NOP it is bound to ignore the data part.
