On Sun, Jan 18, 2026 at 01:25:52PM -0500, Mimi Zohar wrote:
On Fri, 2026-01-16 at 18:27 +0100, Ard Biesheuvel wrote:
Hi Ard and Mimi,
Thanks for your discussion on improving the patch!
On Fri, 16 Jan 2026 at 17:39, Mimi Zohar <[email protected]> wrote:
>
> On Fri, 2026-01-16 at 14:18 +0100, Ard Biesheuvel wrote:
> > On Fri, 16 Jan 2026 at 14:11, Mimi Zohar <[email protected]> wrote:
> > >
> > > On Fri, 2026-01-16 at 10:41 +0100, Ard Biesheuvel wrote:
> > > > On Thu, 15 Jan 2026 at 01:43, Coiby Xu <[email protected]> wrote:
> > > > >
> > > > > EVM and other LSMs need the ability to query the secure boot status of
> > > > > the system, without directly calling the IMA arch_ima_get_secureboot
> > > > > function. Refactor the secure boot status check into a general,
> > > > > integrity-wide function named arch_integrity_get_secureboot.
> > > > >
> > > > > Define a new Kconfig option CONFIG_INTEGRITY_SECURE_BOOT, which is
> > > > > automatically configured by the supported architectures. The existing
> > > > > IMA_SECURE_AND_OR_TRUSTED_BOOT Kconfig loads the architecture specific
> > > > > IMA policy based on the refactored secure boot status code.
> > > > >
> > > > > Reported-and-suggested-by: Mimi Zohar <[email protected]>
> > > > > Suggested-by: Roberto Sassu <[email protected]>
> > > > > Signed-off-by: Coiby Xu <[email protected]>
> > > > > ---
> > > > > arch/arm64/Kconfig | 1 +
> > > > > arch/powerpc/Kconfig | 1 +
> > > > > arch/powerpc/kernel/Makefile | 2 +-
> > > > > arch/powerpc/kernel/ima_arch.c | 5 --
> > > > > arch/powerpc/kernel/integrity_sb_arch.c | 13 +++++
> > > > > arch/s390/Kconfig | 1 +
> > > > > arch/s390/kernel/Makefile | 1 +
> > > > > arch/s390/kernel/ima_arch.c | 6 --
> > > > > arch/s390/kernel/integrity_sb_arch.c | 9 +++
> > > > > arch/x86/Kconfig | 1 +
> > > > > arch/x86/include/asm/efi.h | 4 +-
> > > > > arch/x86/platform/efi/efi.c | 2 +-
> > > > > include/linux/ima.h | 7 +--
> > > > > include/linux/integrity.h | 8 +++
> > > > > security/integrity/Kconfig | 6 ++
> > > > > security/integrity/Makefile | 3 +
> > > > > security/integrity/efi_secureboot.c | 56
+++++++++++++++++++
> > > > > security/integrity/ima/ima_appraise.c | 2 +-
> > > > > security/integrity/ima/ima_efi.c | 47 +---------------
> > > > > security/integrity/ima/ima_main.c | 4 +-
> > > > > security/integrity/platform_certs/load_uefi.c | 2 +-
> > > > > 21 files changed, 111 insertions(+), 70 deletions(-)
> > > > > create mode 100644 arch/powerpc/kernel/integrity_sb_arch.c
> > > > > create mode 100644 arch/s390/kernel/integrity_sb_arch.c
> > > > > create mode 100644 security/integrity/efi_secureboot.c
> > > > >
> > > > > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > > > > index 93173f0a09c7..4c265b7386bb 100644
> > > > > --- a/arch/arm64/Kconfig
> > > > > +++ b/arch/arm64/Kconfig
> > > > > @@ -2427,6 +2427,7 @@ config EFI
> > > > > select EFI_STUB
> > > > > select EFI_GENERIC_STUB
> > > > > imply IMA_SECURE_AND_OR_TRUSTED_BOOT
> > > > > + imply INTEGRITY_SECURE_BOOT
> > > >
> > > > This allows both to be en/disabled individually, which I don't think
> > > > is what we want. It also results in more churn across the
> > > > arch-specific Kconfigs than needed.
> > > >
> > > > Wouldn't it be better if IMA_SECURE_AND_OR_TRUSTED_BOOT 'select'ed
> > > > INTEGRITY_SECURE_BOOT in its Kconfig definition?
> > >
> > > As much as possible, EVM (and other LSMs) shouldn't be dependent on
another LSM,
> > > in this case IMA, being configured.
> >
> > Sure, but that is not my point.
> >
> > This arrangement allows for IMA_SECURE_AND_OR_TRUSTED_BOOT to be
> > enabled without INTEGRITY_SECURE_BOOT, resulting in the stub
> > implementation of arch_integrity_get_secureboot() being used, which
> > always returns false.
Since both INTEGRITY_SECURE_BOOT and IMA_SECURE_AND_OR_TRUSTED_BOOT
don't define a prompt, they are not user-configurable and will always be
enable/disabled together with arch-specific secure boot feature. So
despite the "imply" key word, the case where
IMA_SECURE_AND_OR_TRUSTED_BOOT is enabled whereas INTEGRITY_SECURE_BOOT
is disabled won't happen.
But I agree an arch may not care much about INTEGRITY_SECURE_BOOT so it
may be a churn. So limiting it to the scope of the integrity subsystem
can be a better idea.
>
> I understand your concern, but instead of "select"ing INTEGRITY_SECURE_BOOT
from
> IMA_SECURE_AND_OR_TRUSTED_BOOT, how making IMA_SECURE_AND_OR_TRUSTED_BOOT
> dependent on both IMA_ARCH_POLICY and INTEGRITY_SECURE_BOOT.
>
Given that INTEGRITY_SECURE_BOOT has no dependencies of its own,
afaict, selecting it is the least disruptive option, as otherwise,
existing configs will disable IMA_SECURE_AND_OR_TRUSTED_BOOT as the
kernel is being upgraded. But conceptually, I agree that they are
equivalent.
As already pointed out on by Mimi, INTEGRITY_SECURE_BOOT depend on
arch-specific secure boot feature. So we can't say INTEGRITY_SECURE_BOOT
has no dependencies.
> Including the "imply INTEGRITY_SECURE_BOOT" here in the arch Kconfig allows
EVM
> to query the secure boot state without relying on
IMA_SECURE_AND_OR_TRUSTED_BOOT
> being configured.
Yes, I understand that this is the whole point of the exercise. But
'imply' should be used with care, and in this case, implying both from
CONFIG_EFI really makes little sense. INTEGRITY_SECURE_BOOT should be
selected by options that need the functionality, not 'implied' by
options that might provide it.
But again I agree INTEGRITY_SECURE_BOOT should "not 'implied' by options
that might provide it".
As not all arch's implement arch_integrity_get_secureboot, the definition in
include/linux/integrity.h would need to be updated. Something like:
-#ifdef CONFIG_INTEGRITY_SECURE_BOOT
+#if (defined(CONFIG_INTEGRITY_SECURE_BOOT) && \
+ (defined(CONFIG_X86) && defined(CONFIG_EFI)) || defined(CONFIG_S390) \
+ || defined(CONFIG_PPC_SECURE_BOOT))
Then IMA_SECURE_AND_OR_TRUSTED_BOOT and EVM could select INTEGRITY_SECURE_BOOT,
as suggested.
Since INTEGRITY_SECURE_BOOT has a dependency, select doesn't seem to be
a good choice. If EVM does select INTEGRITY_SECURE_BOOT,
INTEGRITY_SECURE_BOOT will be enabled even if arch-specific secure boot
feature is disabled and this can lead to a building failure. How about
always enabling INTEGRITY_SECURE_BOOT when secure boot feature is
enabled and also making IMA_SECURE_AND_OR_TRUSTED_BOOT depend on
INTEGRITY_SECURE_BOOT?
diff --git a/security/integrity/Kconfig b/security/integrity/Kconfig
index 916d4f2bfc44..cd44b46d0325 100644
--- a/security/integrity/Kconfig
+++ b/security/integrity/Kconfig
@@ -97,6 +97,13 @@ config INTEGRITY_CA_MACHINE_KEYRING_MAX
will not be loaded. The remaining MOK keys are loaded into the
.platform keyring.
+config INTEGRITY_SECURE_BOOT
+ def_bool y
+ depends on EFI || PPC_SECURE_BOOT || S390
+ help
+ Provide secure boot related helper functions like querying the
+ secure boot status.
+
diff --git a/security/integrity/ima/Kconfig b/security/integrity/ima/Kconfig
index 976e75f9b9ba..5dce572192d6 100644
--- a/security/integrity/ima/Kconfig
+++ b/security/integrity/ima/Kconfig
@@ -311,6 +311,7 @@ config IMA_QUEUE_EARLY_BOOT_KEYS
config IMA_SECURE_AND_OR_TRUSTED_BOOT
bool
depends on IMA_ARCH_POLICY
+ depends on INTEGRITY_SECURE_BOOT
Another idea is make a tree-wide arch_get_secureboot i.e. to move
current arch_ima_get_secureboot code to arch-specific secure boot
implementation. By this way, there will no need for a new Kconfig option
INTEGRITY_SECURE_BOOT. But I'm not sure if there is any unforeseen
concern.
--
Best regards,
Coiby
