diff mbox series

efi/arm: Disable LPAE PAN when calling EFI runtime services

Message ID 20240610122437.2358778-2-ardb+git@google.com (mailing list archive)
State New, archived
Headers show
Series efi/arm: Disable LPAE PAN when calling EFI runtime services | expand

Commit Message

Ard Biesheuvel June 10, 2024, 12:24 p.m. UTC
From: Ard Biesheuvel <ardb@kernel.org>

EFI runtime services are remapped into the lower 1 GiB of virtual
address space at boot, so they are guaranteed to be able to co-exist
with the kernel virtual mappings without the need to allocate space for
them in the kernel's vmalloc region, which is rather small.

This means those mappings are covered by TTBR0 when LPAE PAN is enabled,
and so 'user' access must be enabled while such calls are in progress.

To avoid the need to refactor the code that is shared between ARM, arm64
and other EFI architectures, fold this into efi_set_pgd(). Given that
EFI runtime services are serialized and not pre-emptible, storing the
flags into a global variable is reasonable here - efi_set_pgd() calls
will always occur in pairs on a single CPU.

Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
---
 arch/arm/include/asm/efi.h | 15 +++++++++++++++
 arch/arm/kernel/efi.c      |  4 ++++
 2 files changed, 19 insertions(+)

Comments

Linus Walleij June 11, 2024, 1:17 p.m. UTC | #1
On Mon, Jun 10, 2024 at 2:24 PM Ard Biesheuvel <ardb+git@google.com> wrote:

> From: Ard Biesheuvel <ardb@kernel.org>
>
> EFI runtime services are remapped into the lower 1 GiB of virtual
> address space at boot, so they are guaranteed to be able to co-exist
> with the kernel virtual mappings without the need to allocate space for
> them in the kernel's vmalloc region, which is rather small.
>
> This means those mappings are covered by TTBR0 when LPAE PAN is enabled,
> and so 'user' access must be enabled while such calls are in progress.
>
> To avoid the need to refactor the code that is shared between ARM, arm64
> and other EFI architectures, fold this into efi_set_pgd(). Given that
> EFI runtime services are serialized and not pre-emptible, storing the
> flags into a global variable is reasonable here - efi_set_pgd() calls
> will always occur in pairs on a single CPU.
>
> Cc: Kees Cook <keescook@chromium.org>
> Cc: Linus Walleij <linus.walleij@linaro.org>
> Signed-off-by: Ard Biesheuvel <ardb@kernel.org>

Makes sense to me! Thanks for looking into this.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>

Yours,
Linus Walleij
Ard Biesheuvel June 11, 2024, 2:08 p.m. UTC | #2
On Tue, 11 Jun 2024 at 15:17, Linus Walleij <linus.walleij@linaro.org> wrote:
>
> On Mon, Jun 10, 2024 at 2:24 PM Ard Biesheuvel <ardb+git@google.com> wrote:
>
> > From: Ard Biesheuvel <ardb@kernel.org>
> >
> > EFI runtime services are remapped into the lower 1 GiB of virtual
> > address space at boot, so they are guaranteed to be able to co-exist
> > with the kernel virtual mappings without the need to allocate space for
> > them in the kernel's vmalloc region, which is rather small.
> >
> > This means those mappings are covered by TTBR0 when LPAE PAN is enabled,
> > and so 'user' access must be enabled while such calls are in progress.
> >
> > To avoid the need to refactor the code that is shared between ARM, arm64
> > and other EFI architectures, fold this into efi_set_pgd(). Given that
> > EFI runtime services are serialized and not pre-emptible, storing the
> > flags into a global variable is reasonable here - efi_set_pgd() calls
> > will always occur in pairs on a single CPU.
> >
> > Cc: Kees Cook <keescook@chromium.org>
> > Cc: Linus Walleij <linus.walleij@linaro.org>
> > Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
>
> Makes sense to me! Thanks for looking into this.
> Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
>

Thanks, I'll queue this up as a EFI fix.

Note that I have to fix an error in the patch: CONFIG_ARM_TTBR0_PAN
does not exist, it should be CONFIG_CPU_TTBR0_PAN (and don't ask me
why it worked because it definitely did - probably forgot to do git
commit --amend)
diff mbox series

Patch

diff --git a/arch/arm/include/asm/efi.h b/arch/arm/include/asm/efi.h
index 78282ced5038..773fb072c040 100644
--- a/arch/arm/include/asm/efi.h
+++ b/arch/arm/include/asm/efi.h
@@ -14,6 +14,7 @@ 
 #include <asm/mach/map.h>
 #include <asm/mmu_context.h>
 #include <asm/ptrace.h>
+#include <asm/uaccess.h>
 
 #ifdef CONFIG_EFI
 void efi_init(void);
@@ -32,6 +33,20 @@  int efi_set_mapping_permissions(struct mm_struct *mm, efi_memory_desc_t *md, boo
 static inline void efi_set_pgd(struct mm_struct *mm)
 {
 	check_and_switch_context(mm, NULL);
+
+	if (IS_ENABLED(CONFIG_ARM_TTBR0_PAN)) {
+		extern unsigned int efi_arm_ttbr0_pan_flags;
+
+		/*
+		 * EFI runtime services are mapped in the lower TTBR0 region,
+		 * so TTBR0 based PAN should be disabled while making a EFI
+		 * runtime service call.
+		 */
+		if (mm != current->active_mm)
+			efi_arm_ttbr0_pan_flags = uaccess_save_and_enable();
+		else
+			uaccess_restore(efi_arm_ttbr0_pan_flags);
+	}
 }
 
 void efi_virtmap_load(void);
diff --git a/arch/arm/kernel/efi.c b/arch/arm/kernel/efi.c
index 6f9ec7d28a71..6864338073e6 100644
--- a/arch/arm/kernel/efi.c
+++ b/arch/arm/kernel/efi.c
@@ -11,6 +11,10 @@ 
 #include <asm/mach/map.h>
 #include <asm/mmu_context.h>
 
+#ifdef CONFIG_ARM_TTBR0_PAN
+unsigned int efi_arm_ttbr0_pan_flags;
+#endif
+
 static int __init set_permissions(pte_t *ptep, unsigned long addr, void *data)
 {
 	efi_memory_desc_t *md = data;