Message-ID: <aJ2jgqKYL2C6bUjC@e129823.arm.com>
Date: Thu, 14 Aug 2025 09:51:14 +0100
From: Yeoreum Yun <yeoreum.yun@....com>
To: Andrey Konovalov <andreyknvl@...il.com>
Cc: glider@...gle.com, Marco Elver <elver@...gle.com>,
	ryabinin.a.a@...il.com, dvyukov@...gle.com,
	vincenzo.frascino@....com, corbet@....net, catalin.marinas@....com,
	will@...nel.org, akpm@...ux-foundation.org,
	scott@...amperecomputing.com, jhubbard@...dia.com,
	pankaj.gupta@....com, leitao@...ian.org, kaleshsingh@...gle.com,
	maz@...nel.org, broonie@...nel.org, oliver.upton@...ux.dev,
	james.morse@....com, ardb@...nel.org,
	hardevsinh.palaniya@...iconsignals.io, david@...hat.com,
	yang@...amperecomputing.com, kasan-dev@...glegroups.com,
	workflows@...r.kernel.org, linux-doc@...r.kernel.org,
	linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
	linux-mm@...ck.org
Subject: Re: [PATCH v2 1/2] kasan/hw-tags: introduce kasan.store_only option

Hi Andrey,
> On Wed, Aug 13, 2025 at 7:53 PM Yeoreum Yun <yeoreum.yun@....com> wrote:
> >
> > Since Armv8.9, FEATURE_MTE_STORE_ONLY feature is introduced to restrict
> > raise of tag check fault on store operation only.
> > Introcude KASAN store only mode based on this feature.
> >
> > KASAN store only mode restricts KASAN checks operation for store only and
> > omits the checks for fetch/read operation when accessing memory.
> > So it might be used not only debugging enviroment but also normal
> > enviroment to check memory safty.
> >
> > This features can be controlled with "kasan.store_only" arguments.
> > When "kasan.store_only=on", KASAN checks store only mode otherwise
> > KASAN checks all operations.
>
> I'm thinking if we should name this "kasan.write_only" instead of
> "kasan.store_only". This would align the terms with the
> "kasan.fault=panic_on_write" parameter we already have. But then it
> would be different from "FEATURE_MTE_STORE_ONLY", which is what Arm
> documentation uses (right?).

Yes. it uses "MTE_STORE_ONLY". but, write seems fine for me too.

>
> Marco, Alexander, any opinion?
>
> >
> > Signed-off-by: Yeoreum Yun <yeoreum.yun@....com>
> > ---
> >  Documentation/dev-tools/kasan.rst  |  3 ++
> >  arch/arm64/include/asm/memory.h    |  1 +
> >  arch/arm64/include/asm/mte-kasan.h |  6 +++
> >  arch/arm64/kernel/cpufeature.c     |  6 +++
> >  arch/arm64/kernel/mte.c            | 14 ++++++
> >  include/linux/kasan.h              |  2 +
> >  mm/kasan/hw_tags.c                 | 76 +++++++++++++++++++++++++++++-
> >  mm/kasan/kasan.h                   | 10 ++++
> >  8 files changed, 116 insertions(+), 2 deletions(-)
> >
> > diff --git a/Documentation/dev-tools/kasan.rst b/Documentation/dev-tools/kasan.rst
> > index 0a1418ab72fd..fcb70dd821ec 100644
> > --- a/Documentation/dev-tools/kasan.rst
> > +++ b/Documentation/dev-tools/kasan.rst
> > @@ -143,6 +143,9 @@ disabling KASAN altogether or controlling its features:
> >    Asymmetric mode: a bad access is detected synchronously on reads and
> >    asynchronously on writes.
> >
> > +- ``kasan.store_only=off`` or ``kasan.store_only=on`` controls whether KASAN
> > +  checks the store (write) accesses only or all accesses (default: ``off``)
> > +
> >  - ``kasan.vmalloc=off`` or ``=on`` disables or enables tagging of vmalloc
> >    allocations (default: ``on``).
> >
> > diff --git a/arch/arm64/include/asm/memory.h b/arch/arm64/include/asm/memory.h
> > index 5213248e081b..ae29cd3db78d 100644
> > --- a/arch/arm64/include/asm/memory.h
> > +++ b/arch/arm64/include/asm/memory.h
> > @@ -308,6 +308,7 @@ static inline const void *__tag_set(const void *addr, u8 tag)
> >  #define arch_enable_tag_checks_sync()          mte_enable_kernel_sync()
> >  #define arch_enable_tag_checks_async()         mte_enable_kernel_async()
> >  #define arch_enable_tag_checks_asymm()         mte_enable_kernel_asymm()
> > +#define arch_enable_tag_checks_store_only()    mte_enable_kernel_store_only()
> >  #define arch_suppress_tag_checks_start()       mte_enable_tco()
> >  #define arch_suppress_tag_checks_stop()                mte_disable_tco()
> >  #define arch_force_async_tag_fault()           mte_check_tfsr_exit()
> > diff --git a/arch/arm64/include/asm/mte-kasan.h b/arch/arm64/include/asm/mte-kasan.h
> > index 2e98028c1965..3e1cc341d47a 100644
> > --- a/arch/arm64/include/asm/mte-kasan.h
> > +++ b/arch/arm64/include/asm/mte-kasan.h
> > @@ -200,6 +200,7 @@ static inline void mte_set_mem_tag_range(void *addr, size_t size, u8 tag,
> >  void mte_enable_kernel_sync(void);
> >  void mte_enable_kernel_async(void);
> >  void mte_enable_kernel_asymm(void);
> > +int mte_enable_kernel_store_only(void);
> >
> >  #else /* CONFIG_ARM64_MTE */
> >
> > @@ -251,6 +252,11 @@ static inline void mte_enable_kernel_asymm(void)
> >  {
> >  }
> >
> > +static inline int mte_enable_kenrel_store_only(void)
>
> Typo in the function name. Please build/boot test without MTE/KASAN enabled.

Oops... Sorry for mistake :\

>
> > +{
> > +       return -EINVAL;
> > +}
> > +
> >  #endif /* CONFIG_ARM64_MTE */
> >
> >  #endif /* __ASSEMBLY__ */
> > diff --git a/arch/arm64/kernel/cpufeature.c b/arch/arm64/kernel/cpufeature.c
> > index 9ad065f15f1d..7b724fcf20a7 100644
> > --- a/arch/arm64/kernel/cpufeature.c
> > +++ b/arch/arm64/kernel/cpufeature.c
> > @@ -2404,6 +2404,11 @@ static void cpu_enable_mte(struct arm64_cpu_capabilities const *cap)
> >
> >         kasan_init_hw_tags_cpu();
> >  }
> > +
> > +static void cpu_enable_mte_store_only(struct arm64_cpu_capabilities const *cap)
> > +{
> > +       kasan_late_init_hw_tags_cpu();
> > +}
> >  #endif /* CONFIG_ARM64_MTE */
> >
> >  static void user_feature_fixup(void)
> > @@ -2922,6 +2927,7 @@ static const struct arm64_cpu_capabilities arm64_features[] = {
> >                 .capability = ARM64_MTE_STORE_ONLY,
> >                 .type = ARM64_CPUCAP_SYSTEM_FEATURE,
> >                 .matches = has_cpuid_feature,
> > +               .cpu_enable = cpu_enable_mte_store_only,
> >                 ARM64_CPUID_FIELDS(ID_AA64PFR2_EL1, MTESTOREONLY, IMP)
> >         },
> >  #endif /* CONFIG_ARM64_MTE */
> > diff --git a/arch/arm64/kernel/mte.c b/arch/arm64/kernel/mte.c
> > index e5e773844889..8eb1f66f2ccd 100644
> > --- a/arch/arm64/kernel/mte.c
> > +++ b/arch/arm64/kernel/mte.c
> > @@ -157,6 +157,20 @@ void mte_enable_kernel_asymm(void)
> >                 mte_enable_kernel_sync();
> >         }
> >  }
> > +
> > +int mte_enable_kernel_store_only(void)
> > +{
> > +       if (!cpus_have_cap(ARM64_MTE_STORE_ONLY))
> > +               return -EINVAL;
> > +
> > +       sysreg_clear_set(sctlr_el1, SCTLR_EL1_TCSO_MASK,
> > +                        SYS_FIELD_PREP(SCTLR_EL1, TCSO, 1));
> > +       isb();
> > +
> > +       pr_info_once("MTE: enabled stonly mode at EL1\n");
> > +
> > +       return 0;
> > +}
> >  #endif
> >
> >  #ifdef CONFIG_KASAN_HW_TAGS
> > diff --git a/include/linux/kasan.h b/include/linux/kasan.h
> > index 890011071f2b..28951b29c593 100644
> > --- a/include/linux/kasan.h
> > +++ b/include/linux/kasan.h
> > @@ -552,9 +552,11 @@ static inline void kasan_init_sw_tags(void) { }
> >  #ifdef CONFIG_KASAN_HW_TAGS
> >  void kasan_init_hw_tags_cpu(void);
> >  void __init kasan_init_hw_tags(void);
> > +void kasan_late_init_hw_tags_cpu(void);
> >  #else
> >  static inline void kasan_init_hw_tags_cpu(void) { }
> >  static inline void kasan_init_hw_tags(void) { }
> > +static inline void kasan_late_init_hw_tags_cpu(void) { }
> >  #endif
> >
> >  #ifdef CONFIG_KASAN_VMALLOC
> > diff --git a/mm/kasan/hw_tags.c b/mm/kasan/hw_tags.c
> > index 9a6927394b54..c2f90c06076e 100644
> > --- a/mm/kasan/hw_tags.c
> > +++ b/mm/kasan/hw_tags.c
> > @@ -41,9 +41,16 @@ enum kasan_arg_vmalloc {
> >         KASAN_ARG_VMALLOC_ON,
> >  };
> >
> > +enum kasan_arg_store_only {
> > +       KASAN_ARG_STORE_ONLY_DEFAULT,
> > +       KASAN_ARG_STORE_ONLY_OFF,
> > +       KASAN_ARG_STORE_ONLY_ON,
> > +};
> > +
> >  static enum kasan_arg kasan_arg __ro_after_init;
> >  static enum kasan_arg_mode kasan_arg_mode __ro_after_init;
> >  static enum kasan_arg_vmalloc kasan_arg_vmalloc __initdata;
> > +static enum kasan_arg_store_only kasan_arg_store_only __ro_after_init;
> >
> >  /*
> >   * Whether KASAN is enabled at all.
> > @@ -67,6 +74,9 @@ DEFINE_STATIC_KEY_FALSE(kasan_flag_vmalloc);
> >  #endif
> >  EXPORT_SYMBOL_GPL(kasan_flag_vmalloc);
> >
> > +DEFINE_STATIC_KEY_FALSE(kasan_flag_store_only);
>
> Is there a reason to have this as a static key? I think a normal
> global bool would work, just as a normal variable works for
> kasan_mode.

Just for align with the other arguments.
since the kasan_flags_store_only is used only for kunit-test,
not called in any other place, this optimisation is meaningless.
It's fine to change as global bool.

>
> > +EXPORT_SYMBOL_GPL(kasan_flag_store_only);
> > +
> >  #define PAGE_ALLOC_SAMPLE_DEFAULT      1
> >  #define PAGE_ALLOC_SAMPLE_ORDER_DEFAULT        3
> >
> > @@ -141,6 +151,23 @@ static int __init early_kasan_flag_vmalloc(char *arg)
> >  }
> >  early_param("kasan.vmalloc", early_kasan_flag_vmalloc);
> >
> > +/* kasan.store_only=off/on */
> > +static int __init early_kasan_flag_store_only(char *arg)
> > +{
> > +       if (!arg)
> > +               return -EINVAL;
> > +
> > +       if (!strcmp(arg, "off"))
> > +               kasan_arg_store_only = KASAN_ARG_STORE_ONLY_OFF;
> > +       else if (!strcmp(arg, "on"))
> > +               kasan_arg_store_only = KASAN_ARG_STORE_ONLY_ON;
> > +       else
> > +               return -EINVAL;
> > +
> > +       return 0;
> > +}
> > +early_param("kasan.store_only", early_kasan_flag_store_only);
> > +
> >  static inline const char *kasan_mode_info(void)
> >  {
> >         if (kasan_mode == KASAN_MODE_ASYNC)
> > @@ -219,6 +246,20 @@ void kasan_init_hw_tags_cpu(void)
> >         kasan_enable_hw_tags();
> >  }
> >
> > +/*
> > + * kasan_late_init_hw_tags_cpu_post() is called for each CPU after
> > + * all cpus are bring-up at boot.
>
> "CPUs"
> "brought up"
>
> And please spell-check other comments.

Thanks.

>
> > + * Not marked as __init as a CPU can be hot-plugged after boot.
> > + */
> > +void kasan_late_init_hw_tags_cpu(void)
> > +{
> > +       /*
> > +        * Enable stonly mode only when explicitly requested through the command line.
>
> "store-only"
>
> > +        * If system doesn't support, kasan checks all operation.
>
> "If the system doesn't support this mode, KASAN will check both load
> and store operations."

Thanks for suggestion :)

>
> > +        */
> > +       kasan_enable_store_only();
> > +}
> > +
> >  /* kasan_init_hw_tags() is called once on boot CPU. */
> >  void __init kasan_init_hw_tags(void)
> >  {
> > @@ -257,15 +298,28 @@ void __init kasan_init_hw_tags(void)
> >                 break;
> >         }
> >
> > +       switch (kasan_arg_store_only) {
> > +       case KASAN_ARG_STORE_ONLY_DEFAULT:
> > +               /* Default is specified by kasan_flag_store_only definition. */
> > +               break;
> > +       case KASAN_ARG_STORE_ONLY_OFF:
> > +               static_branch_disable(&kasan_flag_store_only);
> > +               break;
> > +       case KASAN_ARG_STORE_ONLY_ON:
> > +               static_branch_enable(&kasan_flag_store_only);
> > +               break;
> > +       }
>
> Let's move this part to kasan_late_init_hw_tags_cpu. Since that's
> where the final decision of whether the store-only mode is enabled is
> taken, we should just set the global flag there.

Okay.

>
> > +
> >         kasan_init_tags();
> >
> >         /* KASAN is now initialized, enable it. */
> >         static_branch_enable(&kasan_flag_enabled);
> >
> > -       pr_info("KernelAddressSanitizer initialized (hw-tags, mode=%s, vmalloc=%s, stacktrace=%s)\n",
> > +       pr_info("KernelAddressSanitizer initialized (hw-tags, mode=%s, vmalloc=%s, stacktrace=%s store_only=%s\n",
>
> Let's put "store_only" here next to "mode".

Hmm, I think it's not a proper place to print store_only in here.
I think it would be better to print log related store_only at
kasan_late_init_hw_tags_cpu() like:
    print_info("KernelAddressSanitizer checks store(write) access only.\n");
When store_only=on.

>
> You're also missing a comma.
>
> >                 kasan_mode_info(),
> >                 str_on_off(kasan_vmalloc_enabled()),
> > -               str_on_off(kasan_stack_collection_enabled()));
> > +               str_on_off(kasan_stack_collection_enabled()),
> > +               str_on_off(kasan_store_only_enabled()));
> >  }
> >
> >  #ifdef CONFIG_KASAN_VMALLOC
> > @@ -394,6 +448,22 @@ void kasan_enable_hw_tags(void)
> >                 hw_enable_tag_checks_sync();
> >  }
> >
> > +void kasan_enable_store_only(void)
>
> Do we need this as a separate function? I think we can just move the
> code to kasan_late_init_hw_tags_cpu.
>
> > +{
> > +       if (kasan_arg_store_only == KASAN_ARG_STORE_ONLY_ON) {
> > +               if (hw_enable_tag_checks_store_only()) {
> > +                       static_branch_disable(&kasan_flag_store_only);
> > +                       kasan_arg_store_only = KASAN_ARG_STORE_ONLY_OFF;
> > +                       pr_warn_once("KernelAddressSanitizer: store only mode isn't supported (hw-tags)\n");
>
> No need for the "KernelAddressSanitizer" prefix, it's already defined
> via pr_fmt().

Okay.

>
> > +               }
> > +       }
> > +}
> > +
> > +bool kasan_store_only_enabled(void)
> > +{
> > +       return static_branch_unlikely(&kasan_flag_store_only);
> > +}
> > +
> >  #if IS_ENABLED(CONFIG_KASAN_KUNIT_TEST)
> >
> >  EXPORT_SYMBOL_IF_KUNIT(kasan_enable_hw_tags);
> > @@ -404,4 +474,6 @@ VISIBLE_IF_KUNIT void kasan_force_async_fault(void)
> >  }
> >  EXPORT_SYMBOL_IF_KUNIT(kasan_force_async_fault);
> >
> > +EXPORT_SYMBOL_IF_KUNIT(kasan_store_only_enabled);
> > +
> >  #endif
> > diff --git a/mm/kasan/kasan.h b/mm/kasan/kasan.h
> > index 129178be5e64..1d853de1c499 100644
> > --- a/mm/kasan/kasan.h
> > +++ b/mm/kasan/kasan.h
> > @@ -33,6 +33,7 @@ static inline bool kasan_stack_collection_enabled(void)
> >  #include "../slab.h"
> >
> >  DECLARE_STATIC_KEY_TRUE(kasan_flag_vmalloc);
> > +DECLARE_STATIC_KEY_FALSE(kasan_flag_stonly);
>
> kasan_flag_store_only
>
> Did you build test this at all?

Yes, But there is no place where directly use kasan_flag_stonly,
I think I miss it. Thanks!

>
>
> >
> >  enum kasan_mode {
> >         KASAN_MODE_SYNC,
> > @@ -428,6 +429,7 @@ static inline const void *arch_kasan_set_tag(const void *addr, u8 tag)
> >  #define hw_enable_tag_checks_sync()            arch_enable_tag_checks_sync()
> >  #define hw_enable_tag_checks_async()           arch_enable_tag_checks_async()
> >  #define hw_enable_tag_checks_asymm()           arch_enable_tag_checks_asymm()
> > +#define hw_enable_tag_checks_store_only()      arch_enable_tag_checks_store_only()
> >  #define hw_suppress_tag_checks_start()         arch_suppress_tag_checks_start()
> >  #define hw_suppress_tag_checks_stop()          arch_suppress_tag_checks_stop()
> >  #define hw_force_async_tag_fault()             arch_force_async_tag_fault()
> > @@ -437,10 +439,18 @@ static inline const void *arch_kasan_set_tag(const void *addr, u8 tag)
> >                         arch_set_mem_tag_range((addr), (size), (tag), (init))
> >
> >  void kasan_enable_hw_tags(void);
> > +void kasan_enable_store_only(void);
> > +bool kasan_store_only_enabled(void);
> >
> >  #else /* CONFIG_KASAN_HW_TAGS */
> >
> >  static inline void kasan_enable_hw_tags(void) { }
> > +static inline void kasan_enable_store_only(void) { }
> > +
> > +static inline bool kasan_store_only_enabled(void)
> > +{
> > +       return false;
> > +}
> >
> >  #endif /* CONFIG_KASAN_HW_TAGS */
> >
> > --
> > LEVI:{C3F47F37-75D8-414A-A8BA-3980EC8A46D7}
> >
> > --
> > You received this message because you are subscribed to the Google Groups "kasan-dev" group.
> > To unsubscribe from this group and stop receiving emails from it, send an email to kasan-dev+unsubscribe@...glegroups.com.
> > To view this discussion visit https://groups.google.com/d/msgid/kasan-dev/20250813175335.3980268-2-yeoreum.yun%40arm.com.

--
Sincerely,
Yeoreum Yun

Powered by blists - more mailing lists