Re: [PATCH v3 3/3] arm64/ima: add ima_arch support

[Mimi Zohar <zohar@linux.ibm.com>]

On Mon, 2020-11-02 at 15:20 +0800, Chester Lin wrote:
> On Fri, Oct 30, 2020 at 12:53:25PM +0100, Ard Biesheuvel wrote:
> > On Fri, 30 Oct 2020 at 07:09, Chester Lin <clin@suse.com> wrote:
> > >
> > > Add arm64 IMA arch support. The code and arch policy is mainly inherited
> > > from x86.
> > >
> > > Signed-off-by: Chester Lin <clin@suse.com>
> > > ---
> > >  arch/arm64/Kconfig           |  1 +
> > >  arch/arm64/kernel/Makefile   |  2 ++
> > >  arch/arm64/kernel/ima_arch.c | 43 ++++++++++++++++++++++++++++++++++++
> > >  3 files changed, 46 insertions(+)
> > >  create mode 100644 arch/arm64/kernel/ima_arch.c
> > >
> > > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > > index a42e8d13cc88..496a4a26afc6 100644
> > > --- a/arch/arm64/Kconfig
> > > +++ b/arch/arm64/Kconfig
> > > @@ -201,6 +201,7 @@ config ARM64
> > >         select SWIOTLB
> > >         select SYSCTL_EXCEPTION_TRACE
> > >         select THREAD_INFO_IN_TASK
> > > +       imply IMA_SECURE_AND_OR_TRUSTED_BOOT if EFI
> > >         help
> > >           ARM 64-bit (AArch64) Linux support.
> > >
> > > diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile
> > > index bbaf0bc4ad60..0f6cbb50668c 100644
> > > --- a/arch/arm64/kernel/Makefile
> > > +++ b/arch/arm64/kernel/Makefile
> > > @@ -69,3 +69,5 @@ extra-y                                       += $(head-y) vmlinux.lds
> > >  ifeq ($(CONFIG_DEBUG_EFI),y)
> > >  AFLAGS_head.o += -DVMLINUX_PATH="\"$(realpath $(objtree)/vmlinux)\""
> > >  endif
> > > +
> > > +obj-$(CONFIG_IMA_SECURE_AND_OR_TRUSTED_BOOT)   += ima_arch.o
> > > diff --git a/arch/arm64/kernel/ima_arch.c b/arch/arm64/kernel/ima_arch.c
> > > new file mode 100644
> > > index 000000000000..564236d77adc
> > > --- /dev/null
> > > +++ b/arch/arm64/kernel/ima_arch.c
> > > @@ -0,0 +1,43 @@
> > > +// SPDX-License-Identifier: GPL-2.0+
> > > +/*
> > > + * Copyright (C) 2018 IBM Corporation
> > > + */
> > > +#include <linux/efi.h>
> > > +#include <linux/module.h>
> > > +#include <linux/ima.h>
> > > +
> > > +bool arch_ima_get_secureboot(void)
> > > +{
> > > +       static bool sb_enabled;
> > > +       static bool initialized;
> > > +
> > > +       if (!initialized & efi_enabled(EFI_BOOT)) {
> > > +               sb_enabled = ima_get_efi_secureboot();
> > > +               initialized = true;
> > > +       }
> > > +
> > > +       return sb_enabled;
> > > +}
> > > +
> > > +/* secure and trusted boot arch rules */
> > > +static const char * const sb_arch_rules[] = {
> > > +#if !IS_ENABLED(CONFIG_KEXEC_SIG)
> > > +       "appraise func=KEXEC_KERNEL_CHECK appraise_type=imasig",
> > > +#endif /* CONFIG_KEXEC_SIG */
> > > +       "measure func=KEXEC_KERNEL_CHECK",
> > > +#if !IS_ENABLED(CONFIG_MODULE_SIG)
> > > +       "appraise func=MODULE_CHECK appraise_type=imasig",
> > > +#endif
> > > +       "measure func=MODULE_CHECK",
> > > +       NULL
> > > +};
> > > +
> > > +const char * const *arch_get_ima_policy(void)
> > > +{
> > > +       if (IS_ENABLED(CONFIG_IMA_ARCH_POLICY) && arch_ima_get_secureboot()) {
> > > +               if (IS_ENABLED(CONFIG_MODULE_SIG))
> > > +                       set_module_sig_enforced();
> > > +               return sb_arch_rules;
> > > +       }
> > > +       return NULL;
> > > +}
> > > --
> > > 2.28.0
> > >
> > 
> > Can we move all this stuff into security/integrity/ima/ima_efi.c instead?
> >
> Actually I hesitated to move all this stuff into ima_efi.c when coding v3
> because I haven't figured out a clear picture to achieve it. Since each
> architecture could still have different details to trigger secure boot detection
> and define their arch-specific rules [e.g. Having boot_params in x86_64 creates
> more conditions that need to be determined before calling get_sb_mode()], moving
> all this stuff seems to decrease the flexibility. Besides, it might also affect
> the consistency of ima_arch as well, for example, ppc and s390 still use these
> function prototypes defined in ima.h. Since these functions are already referred
> by non-EFI architectures, why don't we still reuse these prototypes? For example,
> we could remain a smaller arch_ima_get_secureboot() and the arch-specific rules
> but move the major part of arch_get_ima_policy() into ima_efi.c. For example,
> we could implement an efi_ima_policy() for arch_get_ima_policy() to call so that
> the arch_get_ima_policy() doesn't have to know some details such as checking
> conditions or calling set_module_sig_enforced().
> 
> Please feel free to let me know if any suggestions.

Yes, that is the point and the reason for defining ima_efi.c and
conditionally including it only for EFI systems.  The existing ppc and
s390 code should remain unaffected by this change.

thanks,

Mimi