cpu: Fix default mitigation behavior
Subsystems (4)
v2 cover cpu: Fix default mitigation behavior
Linus, I Cc'd you on this as patch 1 fixes a goof that causes mitigations to be completely disabled on all !x86 architectures, and it'd be nice to fix that in rc5. There was a decent bit of discussion on how exactly to juggle the Kconfigs, and so I don't expect anyone to include this in a pull request for rc5.
The discussion didn't fully resolve, i.e. this hasn't gotten a thumbs up from the affected parties, but I think/hope my approach here is minimal enough for other architectures (just restores previous behavior), and shouldn't result in a huge amount of churn if we decide to go in a different direction.
TL;DR: please grab patch 1 directly if you think it's worth squeezing into rc5, and isn't completely crazy.
Thanks!
Patch 2 is probably 6.9 material, but is definitely not rc5 material. It disallows retroactively enabling mitigations via command line if the kernel was built with CPU_MITIGATIONS=n, as it's infeasible for x86 to provide sane, predictable behavior for this scenario.
v2:
- Rework the Kconfigs so that there's a single user-visible CPU_MITIGATION config. [Everyone]
- Define CPU_MITIGATIONS in arch/Kconfig. [Josh]
- Completely compile out the cpu_mitigations code if CPU_MITIGATIONS=n, e.g. to make impossible to end up in a half-baked state where cpu_mitigations ends up enabled but the kernel wasn't compiled with mitigations enabled.
v1: https://lore.kernel.org/all/[email protected]
Sean Christopherson (2):
cpu: Re-enable CPU mitigations by default for !X86 architectures
cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
.../admin-guide/kernel-parameters.txt | 3 +++ arch/Kconfig | 8 ++++++++ arch/x86/Kconfig | 19 ++++++++++++------- include/linux/cpu.h | 11 +++++++++++ kernel/cpu.c | 13 ++++++++++--- 5 files changed, 44 insertions(+), 10 deletions(-)
base-commit: 96fca68c4fbf77a8185eb10f7557e23352732ea2
Signature · 1 line
2.44.0.769.g3c40516874-goog
PATCH v2 1/2 cpu: Re-enable CPU mitigations by default for !X86 architectures
Rename x86's to CPU_MITIGATIONS, define it in generic code, and force it on for all architectures exception x86. A recent commit to turn mitigations off by default if SPECULATION_MITIGATIONS=n kinda sorta missed that "cpu_mitigations" is completely generic, whereas SPECULATION_MITIGATIONS is x86 specific.
Rename x86's SPECULATIVE_MITIGATIONS instead of keeping both and have it select CPU_MITIGATIONS, as having two configs for the same thing is unnecessary and confusing. This will also allow x86 to use the knob to manage mitigations that aren't strictly related to speculative execution.
Use another Kconfig to communicate to common code that CPU_MITIGATIONS is already defined instead of having x86's menu depend on the common CPU_MITIGATIONS. This allows keeping a single point of contact for all of x86's mitigations, and it's not clear that other architectures *want* to allow disabling mitigations at compile-time.
--- arch/Kconfig | 8 ++++++++ arch/x86/Kconfig | 11 ++++++----- kernel/cpu.c | 4 ++-- 3 files changed, 16 insertions(+), 7 deletions(-)
arch/Kconfig +8 -0
@@ -9,6 +9,14 @@ # source "arch/$(SRCARCH)/Kconfig" +config ARCH_CONFIGURES_CPU_MITIGATIONS + bool + +if !ARCH_CONFIGURES_CPU_MITIGATIONS +config CPU_MITIGATIONS + def_bool y +endif + menu "General architecture-dependent options" config ARCH_HAS_SUBPAGE_FAULTS
arch/x86/Kconfig +6 -5
@@ -62,6 +62,7 @@ config X86 select ACPI_HOTPLUG_CPU if ACPI_PROCESSOR && HOTPLUG_CPU select ARCH_32BIT_OFF_T if X86_32 select ARCH_CLOCKSOURCE_INIT + select ARCH_CONFIGURES_CPU_MITIGATIONS select ARCH_CORRECT_STACKTRACE_ON_KRETPROBE select ARCH_ENABLE_HUGEPAGE_MIGRATION if X86_64 && HUGETLB_PAGE && MIGRATION select ARCH_ENABLE_MEMORY_HOTPLUG if X86_64
@@ -2488,17 +2489,17 @@ config PREFIX_SYMBOLS def_bool y depends on CALL_PADDING && !CFI_CLANG -menuconfig SPECULATION_MITIGATIONS - bool "Mitigations for speculative execution vulnerabilities" +menuconfig CPU_MITIGATIONS + bool "Mitigations for CPU vulnerabilities" default y help - Say Y here to enable options which enable mitigations for - speculative execution hardware vulnerabilities. + Say Y here to enable options which enable mitigations for hardware + vulnerabilities (usually related to speculative execution). If you say N, all mitigations will be disabled. You really should know what you are doing to say so. -if SPECULATION_MITIGATIONS +if CPU_MITIGATIONS config MITIGATION_PAGE_TABLE_ISOLATION bool "Remove the kernel mapping in user mode"
kernel/cpu.c +2 -2
@@ -3207,8 +3207,8 @@ enum cpu_mitigations { };static enum cpu_mitigations cpu_mitigations__ro_after_init = -IS_ENABLED (CONFIG_SPECULATION_MITIGATIONS ) ?CPU_MITIGATIONS_AUTO : -CPU_MITIGATIONS_OFF ; +IS_ENABLED (CONFIG_CPU_MITIGATIONS ) ?CPU_MITIGATIONS_AUTO : +CPU_MITIGATIONS_OFF ;static int __init mitigations_parse_cmdline (char *arg) {
Signature · 1 line
2.44.0.769.g3c40516874-goog
Re: [PATCH v2 1/2] cpu: Re-enable CPU mitigations by default for !X86 architectures
Quoted 17 lines · Sean Christopherson
Rename x86's to
CPU_MITIGATIONS, define it in generic code, and force it on for all architectures exception x86. A recent commit to turn mitigations off by default if SPECULATION_MITIGATIONS=n kinda sorta missed that "cpu_mitigations" is completely generic, whereasSPECULATION_MITIGATIONSis x86 specific.Rename x86's
SPECULATIVE_MITIGATIONSinstead of keeping both and have it selectCPU_MITIGATIONS, as having two configs for the same thing is unnecessary and confusing. This will also allow x86 to use the knob to manage mitigations that aren't strictly related to speculative execution.Use another
Kconfigto communicate to common code thatCPU_MITIGATIONSis already defined instead of having x86's menu depend on the commonCPU_MITIGATIONS. This allows keeping a single point of contact for all of x86's mitigations, and it's not clear that other architectures *want* to allow disabling mitigations at compile-time.
Quoted 10 lines
Reported-by Stephen Rothwell <[email protected]>Reported-by Michael Ellerman <[email protected]>Reported-by Geert Uytterhoeven <[email protected]>Fixes f337a6a21e2f ("x86/cpu: Actually turn off mitigations by default for SPECULATION_MITIGATIONS=n")Cc Josh Poimboeuf <[email protected]>Cc Will Deacon <[email protected]>Cc Linus Torvalds <[email protected]>Signed-off-by Sean Christopherson <[email protected]>
Signature · 1 line
Josh
[tip: x86/urgent] cpu: Re-enable CPU mitigations by default for !X86 architectures
In reply to Sean ChristophersonThe following commit has been merged into the x86/urgent branch of tip:
The following commit has been merged into the x86/urgent branch of tip:
Commit-ID: fe42754b94a42d08cf9501790afc25c4f6a5f631 Gitweb: https://git.kernel.org/tip/fe42754b94a42d08cf9501790afc25c4f6a5f631 Author: Sean Christopherson <[email protected]> AuthorDate: Fri, 19 Apr 2024 17:05:54 -07:00 Committer: Borislav Petkov (AMD) <[email protected]> CommitterDate: Thu, 25 Apr 2024 15:47:35 +02:00
cpu: Re-enable CPU mitigations by default for !X86 architectures
Rename x86's to CPU_MITIGATIONS, define it in generic code, and force it on for all architectures exception x86. A recent commit to turn mitigations off by default if SPECULATION_MITIGATIONS=n kinda sorta missed that "cpu_mitigations" is completely generic, whereas SPECULATION_MITIGATIONS is x86-specific.
Rename x86's SPECULATIVE_MITIGATIONS instead of keeping both and have it select CPU_MITIGATIONS, as having two configs for the same thing is unnecessary and confusing. This will also allow x86 to use the knob to manage mitigations that aren't strictly related to speculative execution.
Use another Kconfig to communicate to common code that CPU_MITIGATIONS is already defined instead of having x86's menu depend on the common CPU_MITIGATIONS. This allows keeping a single point of contact for all of x86's mitigations, and it's not clear that other architectures *want* to allow disabling mitigations at compile-time.
--- arch/Kconfig | 8 ++++++++ arch/x86/Kconfig | 11 ++++++----- kernel/cpu.c | 4 ++-- 3 files changed, 16 insertions(+), 7 deletions(-)
arch/Kconfig +8 -0
@@ -9,6 +9,14 @@ # source "arch/$(SRCARCH)/Kconfig" +config ARCH_CONFIGURES_CPU_MITIGATIONS + bool + +if !ARCH_CONFIGURES_CPU_MITIGATIONS +config CPU_MITIGATIONS + def_bool y +endif + menu "General architecture-dependent options" config ARCH_HAS_SUBPAGE_FAULTS
arch/x86/Kconfig +6 -5
@@ -62,6 +62,7 @@ config X86 select ACPI_HOTPLUG_CPU if ACPI_PROCESSOR && HOTPLUG_CPU select ARCH_32BIT_OFF_T if X86_32 select ARCH_CLOCKSOURCE_INIT + select ARCH_CONFIGURES_CPU_MITIGATIONS select ARCH_CORRECT_STACKTRACE_ON_KRETPROBE select ARCH_ENABLE_HUGEPAGE_MIGRATION if X86_64 && HUGETLB_PAGE && MIGRATION select ARCH_ENABLE_MEMORY_HOTPLUG if X86_64
@@ -2488,17 +2489,17 @@ config PREFIX_SYMBOLS def_bool y depends on CALL_PADDING && !CFI_CLANG -menuconfig SPECULATION_MITIGATIONS - bool "Mitigations for speculative execution vulnerabilities" +menuconfig CPU_MITIGATIONS + bool "Mitigations for CPU vulnerabilities" default y help - Say Y here to enable options which enable mitigations for - speculative execution hardware vulnerabilities. + Say Y here to enable options which enable mitigations for hardware + vulnerabilities (usually related to speculative execution). If you say N, all mitigations will be disabled. You really should know what you are doing to say so. -if SPECULATION_MITIGATIONS +if CPU_MITIGATIONS config MITIGATION_PAGE_TABLE_ISOLATION bool "Remove the kernel mapping in user mode"
kernel/cpu.c +2 -2
@@ -3207,8 +3207,8 @@ enum cpu_mitigations { };static enum cpu_mitigations cpu_mitigations__ro_after_init = -IS_ENABLED (CONFIG_SPECULATION_MITIGATIONS ) ?CPU_MITIGATIONS_AUTO : -CPU_MITIGATIONS_OFF ; +IS_ENABLED (CONFIG_CPU_MITIGATIONS ) ?CPU_MITIGATIONS_AUTO : +CPU_MITIGATIONS_OFF ;static int __init mitigations_parse_cmdline (char *arg) {
PATCH v2 2/2 cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
In reply to Sean ChristophersonExplicitly disallow enabling mitigations at runtime for kernels that were built with CONFIG_CPU_MITIGATIONS=n, as some architectures may omit code entirely if mitigations are disabled at compile time.
E.g. on x86, a large pile of Kconfigs are buried behind CPU_MITIGATIONS, and trying to provide sane behavior for retroactively enabling mitigations is extremely difficult, bordering on impossible. E.g. page table isolation and call depth tracking requrie build-time support, BHI mitigations will still be off without additional kernel parameters, etc.
--- Documentation/admin-guide/kernel-parameters.txt | 3 +++ arch/x86/Kconfig | 8 ++++++-- include/linux/cpu.h | 11 +++++++++++ kernel/cpu.c | 13 ++++++++++--- 4 files changed, 30 insertions(+), 5 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +3 -0
@@ -3423,6 +3423,9 @@ arch-independent options, each of which is an aggregation of existing arch-specific options. + Note, "mitigations" is supported if and only if the + kernel was built with CPU_MITIGATIONS=y. + off Disable all optional CPU mitigations. This improves system performance, but it may also
arch/x86/Kconfig +6 -2
@@ -2495,9 +2495,13 @@ menuconfig CPU_MITIGATIONS help Say Y here to enable options which enable mitigations for hardware vulnerabilities (usually related to speculative execution). + Mitigations can be disabled or restricted to SMT systems at runtime + via the "mitigations" kernel parameter. - If you say N, all mitigations will be disabled. You really - should know what you are doing to say so. + If you say N, all mitigations will be disabled. This CANNOT be + overridden at runtime. + + Say 'Y', unless you really know what you are doing. if CPU_MITIGATIONS
include/linux/cpu.h +11 -0
@@ -221,7 +221,18 @@ void cpuhp_report_idle_dead(void);static inline void cpuhp_report_idle_dead (void ) { }#endif /* #ifdef CONFIG_HOTPLUG_CPU */ +#ifdef CONFIG_CPU_MITIGATIONS extern bool cpu_mitigations_off (void );extern bool cpu_mitigations_auto_nosmt (void ); +#else +static inline bool cpu_mitigations_off (void ) +{ +return false; +} +static inline bool cpu_mitigations_auto_nosmt (void ) +{ +return false; +} +#endif #endif /* _LINUX_CPU_H_ */
kernel/cpu.c +10 -3
@@ -3196,6 +3196,7 @@ void __init boot_cpu_hotplug_init(void)this_cpu_write (cpuhp_state.target,CPUHP_ONLINE ); } +#ifdef CONFIG_CPU_MITIGATIONS /* * These are used for a global "mitigations=" cmdline option for toggling * optional CPU mitigations.
@@ -3207,8 +3208,7 @@ enum cpu_mitigations { };static enum cpu_mitigations cpu_mitigations__ro_after_init = -IS_ENABLED (CONFIG_CPU_MITIGATIONS ) ?CPU_MITIGATIONS_AUTO : -CPU_MITIGATIONS_OFF ; +CPU_MITIGATIONS_AUTO ;static int __init mitigations_parse_cmdline (char *arg) {
@@ -3224,7 +3224,6 @@ static int __init mitigations_parse_cmdline(char *arg)return 0 ; } -early_param ("mitigations" , mitigations_parse_cmdline);/* mitigations=off */ bool cpu_mitigations_off (void )
@@ -3239,3 +3238,11 @@ bool cpu_mitigations_auto_nosmt(void) return cpu_mitigations == CPU_MITIGATIONS_AUTO_NOSMT; }EXPORT_SYMBOL_GPL (cpu_mitigations_auto_nosmt); +#else +static int __init mitigations_parse_cmdline (char *arg) +{ +pr_crit ("Kernel compiled without mitigations, ignoring 'mitigations'; system may still be vulnerable\n" ); +return 0 ; +} +#endif +early_param ("mitigations" , mitigations_parse_cmdline);
Signature · 1 line
2.44.0.769.g3c40516874-goog
Borislav Petkov
maintainerRe: [PATCH v2 2/2] cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
Quoted 5 lines · Sean Christopherson
arch/x86/Kconfig +0
-0
arch/x86/Kconfig menuconfig CPU_MITIGATIONS · line 2502
Quoted 10 lines
arch/x86/Kconfig +4
-2@@ -2495,9 +2495,13 @@ menuconfig CPU_MITIGATIONS help Say Y here to enable options which enable mitigations for hardware vulnerabilities (usually related to speculative execution). + Mitigations can be disabled or restricted to SMT systems at runtime + via the "mitigations" kernel parameter. - If you say N, all mitigations will be disabled. You really - should know what you are doing to say so. + If you say N, all mitigations will be disabled. This CANNOT be + overridden at runtime.
You probably wanna highlight the fact here that saying N means it'll simply not even build in the mitigations code, leading to the physical inability :) to enable them later, at run time.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [PATCH v2 2/2] cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
In reply to Sean Christophersoninclude/linux/cpu.h void cpuhp_report_idle_dead(void); · line 231
Quoted 8 lines · Sean Christopherson
include/linux/cpu.h +6
-0+#ifdef CONFIG_CPU_MITIGATIONS extern bool cpu_mitigations_off (void );extern bool cpu_mitigations_auto_nosmt (void ); +#else +static inline bool cpu_mitigations_off (void ) +{ +return false; +}
This should probably return true?
Signature · 1 line
Josh
Borislav Petkov
maintainerRe: [PATCH v2 2/2] cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
Quoted 11 lines · Josh Poimboeuf
include/linux/cpu.h void cpuhp_report_idle_dead(void); · line 231
Quoted 8 lines · Sean Christopherson
include/linux/cpu.h +6
-0+#ifdef CONFIG_CPU_MITIGATIONS extern bool cpu_mitigations_off (void );extern bool cpu_mitigations_auto_nosmt (void ); +#else +static inline bool cpu_mitigations_off (void ) +{ +return false; +}This should probably return true?
Right, I'll fix it up while applying and send them linuswards this weekend so that 6.9 releases fixed.
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [PATCH v2 2/2] cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
Quoted 12 lines · Borislav Petkov
Quoted 11 lines · Josh Poimboeuf
include/linux/cpu.h void cpuhp_report_idle_dead(void); · line 231
Quoted 8 lines · Sean Christopherson
include/linux/cpu.h +6
-0+#ifdef CONFIG_CPU_MITIGATIONS extern bool cpu_mitigations_off (void );extern bool cpu_mitigations_auto_nosmt (void ); +#else +static inline bool cpu_mitigations_off (void ) +{ +return false; +}This should probably return true?
/facepalm
Glad you were paying attention, as I was clearly not. I double checked that flipping that to true does indeed force off mitigations.
Quoted 2 lines
Right, I'll fix it up while applying and send them linuswards this weekend so that 6.9 releases fixed.
Thanks!
[tip: x86/urgent] cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
In reply to Sean ChristophersonThe following commit has been merged into the x86/urgent branch of tip:
The following commit has been merged into the x86/urgent branch of tip:
Commit-ID: ce0abef6a1d540acef85068e0e82bdf1fbeeb0e9 Gitweb: https://git.kernel.org/tip/ce0abef6a1d540acef85068e0e82bdf1fbeeb0e9 Author: Sean Christopherson <[email protected]> AuthorDate: Fri, 19 Apr 2024 17:05:55 -07:00 Committer: Borislav Petkov (AMD) <[email protected]> CommitterDate: Thu, 25 Apr 2024 15:47:39 +02:00
cpu: Ignore "mitigations" kernel parameter if CPU_MITIGATIONS=n
Explicitly disallow enabling mitigations at runtime for kernels that were built with CONFIG_CPU_MITIGATIONS=n, as some architectures may omit code entirely if mitigations are disabled at compile time.
E.g. on x86, a large pile of Kconfigs are buried behind CPU_MITIGATIONS, and trying to provide sane behavior for retroactively enabling mitigations is extremely difficult, bordering on impossible. E.g. page table isolation and call depth tracking require build-time support, BHI mitigations will still be off without additional kernel parameters, etc.
[ bp: Touchups. ]
--- Documentation/admin-guide/kernel-parameters.txt | 3 +++ arch/x86/Kconfig | 8 ++++++-- include/linux/cpu.h | 11 +++++++++++ kernel/cpu.c | 14 ++++++++++---- 4 files changed, 30 insertions(+), 6 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +3 -0
@@ -3423,6 +3423,9 @@ arch-independent options, each of which is an aggregation of existing arch-specific options. + Note, "mitigations" is supported if and only if the + kernel was built with CPU_MITIGATIONS=y. + off Disable all optional CPU mitigations. This improves system performance, but it may also
arch/x86/Kconfig +6 -2
@@ -2495,9 +2495,13 @@ menuconfig CPU_MITIGATIONS help Say Y here to enable options which enable mitigations for hardware vulnerabilities (usually related to speculative execution). + Mitigations can be disabled or restricted to SMT systems at runtime + via the "mitigations" kernel parameter. - If you say N, all mitigations will be disabled. You really - should know what you are doing to say so. + If you say N, all mitigations will be disabled. This CANNOT be + overridden at runtime. + + Say 'Y', unless you really know what you are doing. if CPU_MITIGATIONS
include/linux/cpu.h +11 -0
@@ -221,7 +221,18 @@ void cpuhp_report_idle_dead(void);static inline void cpuhp_report_idle_dead (void ) { }#endif /* #ifdef CONFIG_HOTPLUG_CPU */ +#ifdef CONFIG_CPU_MITIGATIONS extern bool cpu_mitigations_off (void );extern bool cpu_mitigations_auto_nosmt (void ); +#else +static inline bool cpu_mitigations_off (void ) +{ +return true; +} +static inline bool cpu_mitigations_auto_nosmt (void ) +{ +return false; +} +#endif #endif /* _LINUX_CPU_H_ */
kernel/cpu.c +10 -4
@@ -3196,6 +3196,7 @@ void __init boot_cpu_hotplug_init(void)this_cpu_write (cpuhp_state.target,CPUHP_ONLINE ); } +#ifdef CONFIG_CPU_MITIGATIONS /* * These are used for a global "mitigations=" cmdline option for toggling * optional CPU mitigations.
@@ -3206,9 +3207,7 @@ enum cpu_mitigations { CPU_MITIGATIONS_AUTO_NOSMT, }; -static enum cpu_mitigations cpu_mitigations__ro_after_init = -IS_ENABLED (CONFIG_CPU_MITIGATIONS ) ?CPU_MITIGATIONS_AUTO : -CPU_MITIGATIONS_OFF ; +static enum cpu_mitigations cpu_mitigations__ro_after_init =CPU_MITIGATIONS_AUTO ;static int __init mitigations_parse_cmdline (char *arg) {
@@ -3224,7 +3223,6 @@ static int __init mitigations_parse_cmdline(char *arg)return 0 ; } -early_param ("mitigations" , mitigations_parse_cmdline);/* mitigations=off */ bool cpu_mitigations_off (void )
@@ -3239,3 +3237,11 @@ bool cpu_mitigations_auto_nosmt(void) return cpu_mitigations == CPU_MITIGATIONS_AUTO_NOSMT; }EXPORT_SYMBOL_GPL (cpu_mitigations_auto_nosmt); +#else +static int __init mitigations_parse_cmdline (char *arg) +{ +pr_crit ("Kernel compiled without mitigations, ignoring 'mitigations'; system may still be vulnerable\n" ); +return 0 ; +} +#endif +early_param ("mitigations" , mitigations_parse_cmdline);
Quoted 6 lines · Sean Christopherson
Linus, I Cc'd you on this as patch 1 fixes a goof that causes mitigations to be completely disabled on all !x86 architectures, and it'd be nice to fix that in rc5. There was a decent bit of discussion on how exactly to juggle the Kconfigs, and so I don't expect anyone to include this in a pull request for rc5.
Quoted 8 lines
The discussion didn't fully resolve, i.e. this hasn't gotten a thumbs up from the affected parties, but I think/hope my approach here is minimal enough for other architectures (just restores previous behavior), and shouldn't result in a huge amount of churn if we decide to go in a different direction.
TL;DR: please grab patch 1 directly if you think it's worth squeezing into rc5, and isn't completely crazy.
Agreed, we should have some sort of fix for -rc5.
And I don't see anything wrong with it and the aspect that other arches should get their previous behavior for now makes sense. And we can always bikeshed this at large until it is settled.
So patch 1:
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette