Re: [PATCH v2 5/8] x86/speculation: Add basic support for IBPB

From: Borislav Petkov
Date: Sun Jan 21 2018 - 13:06:32 EST


On Sun, Jan 21, 2018 at 09:49:06AM +0000, David Woodhouse wrote:
> From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
>
> Expose indirect_branch_prediction_barrier() for use in subsequent patches.
>
> [karahmed: remove the special-casing of skylake for using IBPB (wtf?),
> switch to using ALTERNATIVES instead of static_cpu_has]
> [dwmw2: set up ax/cx/dx in the asm too so it gets NOP'd out]
>
> Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Signed-off-by: KarimAllah Ahmed <karahmed@xxxxxxxxx>
> Signed-off-by: David Woodhouse <dwmw@xxxxxxxxxxxx>
> ---
> arch/x86/include/asm/cpufeatures.h | 1 +
> arch/x86/include/asm/nospec-branch.h | 16 ++++++++++++++++
> arch/x86/kernel/cpu/bugs.c | 7 +++++++
> 3 files changed, 24 insertions(+)
>
> diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
> index 8c9e5c0..cf28399 100644
> --- a/arch/x86/include/asm/cpufeatures.h
> +++ b/arch/x86/include/asm/cpufeatures.h
> @@ -207,6 +207,7 @@
> #define X86_FEATURE_RETPOLINE_AMD ( 7*32+13) /* AMD Retpoline mitigation for Spectre variant 2 */
> #define X86_FEATURE_INTEL_PPIN ( 7*32+14) /* Intel Processor Inventory Number */
>
> +#define X86_FEATURE_IBPB ( 7*32+16) /* Using Indirect Branch Prediction Barrier */

Right, and as AMD has a separate bit for this in CPUID_80000008_EBX[12],
we probably don't really need the synthetic bit here but simply use the
one at (13*32+12) - word 13.

> #define X86_FEATURE_AMD_PRED_CMD ( 7*32+17) /* Prediction Command MSR (AMD) */
> #define X86_FEATURE_MBA ( 7*32+18) /* Memory Bandwidth Allocation */
> #define X86_FEATURE_RSB_CTXSW ( 7*32+19) /* Fill RSB on context switches */
> diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h
> index 4ad4108..c333c95 100644
> --- a/arch/x86/include/asm/nospec-branch.h
> +++ b/arch/x86/include/asm/nospec-branch.h
> @@ -218,5 +218,21 @@ static inline void vmexit_fill_RSB(void)
> #endif
> }
>
> +static inline void indirect_branch_prediction_barrier(void)

I like ibp_barrier() better.

> +{
> + unsigned long ax, cx, dx;
> +
> + asm volatile(ALTERNATIVE("",
> + "movl %[msr], %%ecx\n\t"
> + "movl %[val], %%eax\n\t"
> + "movl $0, %%edx\n\t"
> + "wrmsr",
> + X86_FEATURE_IBPB)
> + : "=a" (ax), "=c" (cx), "=d" (dx)
> + : [msr] "i" (MSR_IA32_PRED_CMD),
> + [val] "i" (PRED_CMD_IBPB)
> + : "memory");
> +}

Btw, we can simplify this a bit by dropping the inputs and marking the 3
GPRs as clobbered:

alternative_input("",
"mov $0x49, %%ecx\n\t"
"mov $1, %%eax\n\t"
"xor %%edx, %%edx\n\t"
"wrmsr\n\t",
X86_FEATURE_IBPB,
ASM_NO_INPUT_CLOBBER("eax", "ecx", "edx", "memory"));


The "memory" clobber is probably not really needed but it wouldn't
hurt...

Also, above says:

> switch to using ALTERNATIVES instead of static_cpu_has]

Why?

if (static_cpu_has(X86_FEATURE_IBPB))
wrmsr(MSR_IA32_PRED_CMD, PRED_CMD_IBPB, 0);

It can't get any more readable than this. Why even f*ck with
alternatives?

> +
> #endif /* __ASSEMBLY__ */
> #endif /* __NOSPEC_BRANCH_H__ */
> diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
> index 390b3dc..96548ff 100644
> --- a/arch/x86/kernel/cpu/bugs.c
> +++ b/arch/x86/kernel/cpu/bugs.c
> @@ -249,6 +249,13 @@ static void __init spectre_v2_select_mitigation(void)
> setup_force_cpu_cap(X86_FEATURE_RSB_CTXSW);
> pr_info("Filling RSB on context switch\n");
> }
> +
> + /* Initialize Indirect Branch Prediction Barrier if supported */
> + if (boot_cpu_has(X86_FEATURE_SPEC_CTRL) ||
> + boot_cpu_has(X86_FEATURE_AMD_PRED_CMD)) {
> + setup_force_cpu_cap(X86_FEATURE_IBPB);
> + pr_info("Enabling Indirect Branch Prediction Barrier\n");

We don't really need the pr_info as "ibpb" will appear in /proc/cpuinfo.

--
Regards/Gruss,
Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.