Re: [PATCH v5 1/2] kasan: introduce ARCH_DEFER_KASAN and unify static key across modes

[Sabyrzhan Tasbolatov <snovitoll@gmail.com>]

On Fri, Aug 8, 2025 at 10:03 PM Christophe Leroy
<christophe.leroy@csgroup.eu> wrote:
>
>
>
> Le 08/08/2025 à 17:33, Sabyrzhan Tasbolatov a écrit :
> > On Fri, Aug 8, 2025 at 10:03 AM Christophe Leroy
> > <christophe.leroy@csgroup.eu> wrote:
> >>
> >>
> >>
> >> Le 07/08/2025 à 21:40, Sabyrzhan Tasbolatov a écrit :
> >>> Introduce CONFIG_ARCH_DEFER_KASAN to identify architectures [1] that need
> >>> to defer KASAN initialization until shadow memory is properly set up,
> >>> and unify the static key infrastructure across all KASAN modes.
> >>
> >> That probably desserves more details, maybe copy in informations from
> >> the top of cover letter.
> >>
> >> I think there should also be some exeplanations about
> >> kasan_arch_is_ready() becoming kasan_enabled(), and also why
> >> kasan_arch_is_ready() completely disappear from mm/kasan/common.c
> >> without being replaced by kasan_enabled().
> >>
> >>>
> >>> [1] PowerPC, UML, LoongArch selects ARCH_DEFER_KASAN.
> >>>
> >>> Closes: https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbugzilla.kernel.org%2Fshow_bug.cgi%3Fid%3D217049&data=05%7C02%7Cchristophe.leroy%40csgroup.eu%7Cfe4f5a759ad6452b047408ddd691024a%7C8b87af7d86474dc78df45f69a2011bb5%7C0%7C0%7C638902640503259176%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=UM4uvQihJdeWwcC6DIiJXbn4wGsrijjRcHc55uCMErI%3D&reserved=0
> >>> Signed-off-by: Sabyrzhan Tasbolatov <snovitoll@gmail.com>
> >>> ---
> >>> Changes in v5:
> >>> - Unified patches where arch (powerpc, UML, loongarch) selects
> >>>     ARCH_DEFER_KASAN in the first patch not to break
> >>>     bisectability
> >>> - Removed kasan_arch_is_ready completely as there is no user
> >>> - Removed __wrappers in v4, left only those where it's necessary
> >>>     due to different implementations
> >>>
> >>> Changes in v4:
> >>> - Fixed HW_TAGS static key functionality (was broken in v3)
> >>> - Merged configuration and implementation for atomicity
> >>> ---
> >>>    arch/loongarch/Kconfig                 |  1 +
> >>>    arch/loongarch/include/asm/kasan.h     |  7 ------
> >>>    arch/loongarch/mm/kasan_init.c         |  8 +++----
> >>>    arch/powerpc/Kconfig                   |  1 +
> >>>    arch/powerpc/include/asm/kasan.h       | 12 ----------
> >>>    arch/powerpc/mm/kasan/init_32.c        |  2 +-
> >>>    arch/powerpc/mm/kasan/init_book3e_64.c |  2 +-
> >>>    arch/powerpc/mm/kasan/init_book3s_64.c |  6 +----
> >>>    arch/um/Kconfig                        |  1 +
> >>>    arch/um/include/asm/kasan.h            |  5 ++--
> >>>    arch/um/kernel/mem.c                   | 10 ++++++--
> >>>    include/linux/kasan-enabled.h          | 32 ++++++++++++++++++--------
> >>>    include/linux/kasan.h                  |  6 +++++
> >>>    lib/Kconfig.kasan                      |  8 +++++++
> >>>    mm/kasan/common.c                      | 17 ++++++++++----
> >>>    mm/kasan/generic.c                     | 19 +++++++++++----
> >>>    mm/kasan/hw_tags.c                     |  9 +-------
> >>>    mm/kasan/kasan.h                       |  8 ++++++-
> >>>    mm/kasan/shadow.c                      | 12 +++++-----
> >>>    mm/kasan/sw_tags.c                     |  1 +
> >>>    mm/kasan/tags.c                        |  2 +-
> >>>    21 files changed, 100 insertions(+), 69 deletions(-)
> >>>
> >>> diff --git a/arch/loongarch/Kconfig b/arch/loongarch/Kconfig
> >>> index f0abc38c40a..cd64b2bc12d 100644
> >>> --- a/arch/loongarch/Kconfig
> >>> +++ b/arch/loongarch/Kconfig
> >>> @@ -9,6 +9,7 @@ config LOONGARCH
> >>>        select ACPI_PPTT if ACPI
> >>>        select ACPI_SYSTEM_POWER_STATES_SUPPORT if ACPI
> >>>        select ARCH_BINFMT_ELF_STATE
> >>> +     select ARCH_DEFER_KASAN if KASAN
> >>
> >> Instead of adding 'if KASAN' in all users, you could do in two steps:
> >>
> >> Add a symbol ARCH_NEEDS_DEFER_KASAN.
> >>
> >> +config ARCH_NEEDS_DEFER_KASAN
> >> +       bool
> >>
> >> And then:
> >>
> >> +config ARCH_DEFER_KASAN
> >> +       def_bool
> >> +       depends on KASAN
> >> +       depends on ARCH_DEFER_KASAN
> >> +       help
> >> +         Architectures should select this if they need to defer KASAN
> >> +         initialization until shadow memory is properly set up. This
> >> +         enables runtime control via static keys. Otherwise, KASAN uses
> >> +         compile-time constants for better performance.
> >>
> >
> > Actually, I don't see the benefits from this option. Sorry, have just
> > revisited this again.
> > With the new symbol, arch (PowerPC, UML, LoongArch) still needs select
> > 2 options:
> >
> > select ARCH_NEEDS_DEFER_KASAN
> > select ARCH_DEFER_KASAN
>
> Sorry, my mistake, ARCH_DEFER_KASAN has to be 'def_bool y'. Missing the
> 'y'. That way it is automatically set to 'y' as long as KASAN and
> ARCH_NEEDS_DEFER_KASAN are selected. Should be:
>
> config ARCH_DEFER_KASAN
>         def_bool y
>         depends on KASAN
>         depends on ARCH_NEEDS_DEFER_KASAN
>
>
> >
> > and the oneline with `if` condition is cleaner.
> > select ARCH_DEFER_KASAN if KASAN

Hello,

Have just had a chance to test this.

lib/Kconfig.kasan:
        config ARCH_NEEDS_DEFER_KASAN
                bool

        config ARCH_DEFER_KASAN
                def_bool y
                depends on KASAN
                depends on ARCH_NEEDS_DEFER_KASAN

It works for UML defconfig where arch/um/Kconfig is:

config UML
        bool
        default y
        select ARCH_NEEDS_DEFER_KASAN
        select ARCH_DEFER_KASAN if STATIC_LINK

But it prints warnings for PowerPC, LoongArch:

config LOONGARCH
        bool
        ...
        select ARCH_NEEDS_DEFER_KASAN
        select ARCH_DEFER_KASAN

$ make defconfig ARCH=loongarch
*** Default configuration is based on 'loongson3_defconfig'

WARNING: unmet direct dependencies detected for ARCH_DEFER_KASAN
  Depends on [n]: KASAN [=n] && ARCH_NEEDS_DEFER_KASAN [=y]
  Selected by [y]:
  - LOONGARCH [=y]


config PPC
        bool
        default y
        select ARCH_DEFER_KASAN if PPC_RADIX_MMU
        select ARCH_NEEDS_DEFER_KASAN

$ make ppc64_defconfig

WARNING: unmet direct dependencies detected for ARCH_DEFER_KASAN
  Depends on [n]: KASAN [=n] && ARCH_NEEDS_DEFER_KASAN [=y]
  Selected by [y]:
  - PPC [=y] && PPC_RADIX_MMU [=y]


> >
>
> I don't think so because it requires all architectures to add 'if KASAN'
> which is not convenient.
>
> Christophe