From: Liu Song <[email protected]>
In our environment, it was found that the mitigation BHB has a great
impact on the benchmark performance. For example, in the lmbench test,
the "process fork && exit" test performance drops by 20%.
So it is necessary to have the ability to turn off the mitigation
individually through cmdline, thus avoiding having to compile the
kernel by adjusting the config.
Signed-off-by: Liu Song <[email protected]>
---
Documentation/admin-guide/kernel-parameters.txt | 5 +++++
arch/arm64/kernel/proton-pack.c | 10 +++++++++-
2 files changed, 14 insertions(+), 1 deletion(-)
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index d7f3090..1edc9a6 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -3207,6 +3207,7 @@
spectre_v2_user=off [X86]
spec_store_bypass_disable=off [X86,PPC]
ssbd=force-off [ARM64]
+ nospectre_bhb [ARM64]
l1tf=off [X86]
mds=off [X86]
tsx_async_abort=off [X86]
@@ -3631,6 +3632,10 @@
vulnerability. System may allow data leaks with this
option.
+ nospectre_bhb [ARM64] Disable all mitigations for Spectre-BHB (branch
+ history injection) vulnerability. System may allow data leaks
+ with this option.
+
nospec_store_bypass_disable
[HW] Disable all mitigations for the Speculative Store Bypass vulnerability
diff --git a/arch/arm64/kernel/proton-pack.c b/arch/arm64/kernel/proton-pack.c
index 40be3a7..bd16903 100644
--- a/arch/arm64/kernel/proton-pack.c
+++ b/arch/arm64/kernel/proton-pack.c
@@ -988,6 +988,14 @@ static void this_cpu_set_vectors(enum arm64_bp_harden_el1_vectors slot)
isb();
}
+static bool __read_mostly __nospectre_bhb;
+static int __init parse_spectre_bhb_param(char *str)
+{
+ __nospectre_bhb = true;
+ return 0;
+}
+early_param("nospectre_bhb", parse_spectre_bhb_param);
+
void spectre_bhb_enable_mitigation(const struct arm64_cpu_capabilities *entry)
{
bp_hardening_cb_t cpu_cb;
@@ -1001,7 +1009,7 @@ void spectre_bhb_enable_mitigation(const struct arm64_cpu_capabilities *entry)
/* No point mitigating Spectre-BHB alone. */
} else if (!IS_ENABLED(CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY)) {
pr_info_once("spectre-bhb mitigation disabled by compile time option\n");
- } else if (cpu_mitigations_off()) {
+ } else if (cpu_mitigations_off() || __nospectre_bhb) {
pr_info_once("spectre-bhb mitigation disabled by command line option\n");
} else if (supports_ecbhb(SCOPE_LOCAL_CPU)) {
state = SPECTRE_MITIGATED;
--
1.8.3.1
On Fri, Aug 26, 2022 at 07:40:50PM +0800, Liu Song wrote:
> From: Liu Song <[email protected]>
>
> In our environment, it was found that the mitigation BHB has a great
> impact on the benchmark performance. For example, in the lmbench test,
> the "process fork && exit" test performance drops by 20%.
> So it is necessary to have the ability to turn off the mitigation
> individually through cmdline, thus avoiding having to compile the
> kernel by adjusting the config.
>
> Signed-off-by: Liu Song <[email protected]>
If people want to disable this mitigation and know what they are doing,
I have no objection:
Acked-by: Catalin Marinas <[email protected]>
(I guess that's more like 6.1 material)
--
Catalin
On Wed, Sep 07, 2022 at 03:29:05PM +0100, Catalin Marinas wrote:
> On Fri, Aug 26, 2022 at 07:40:50PM +0800, Liu Song wrote:
> > From: Liu Song <[email protected]>
> >
> > In our environment, it was found that the mitigation BHB has a great
> > impact on the benchmark performance. For example, in the lmbench test,
> > the "process fork && exit" test performance drops by 20%.
> > So it is necessary to have the ability to turn off the mitigation
> > individually through cmdline, thus avoiding having to compile the
> > kernel by adjusting the config.
> >
> > Signed-off-by: Liu Song <[email protected]>
>
> If people want to disable this mitigation and know what they are doing,
> I have no objection:
>
> Acked-by: Catalin Marinas <[email protected]>
Thanks!
> (I guess that's more like 6.1 material)
Yup, agreed. I take it you'll pick it up, then?
Will
On Fri, 26 Aug 2022 19:40:50 +0800, Liu Song wrote:
> From: Liu Song <[email protected]>
>
> In our environment, it was found that the mitigation BHB has a great
> impact on the benchmark performance. For example, in the lmbench test,
> the "process fork && exit" test performance drops by 20%.
> So it is necessary to have the ability to turn off the mitigation
> individually through cmdline, thus avoiding having to compile the
> kernel by adjusting the config.
>
> [...]
Applied to arm64 (for-next/misc), thanks!
[1/1] arm64: spectre: increase parameters that can be used to turn off bhb mitigation individually
https://git.kernel.org/arm64/c/877ace9eab7d
--
Catalin