Re: [RFC PATCH] generic_entry: Add stackleak support

[Mark Rutland]

On Tue, Sep 06, 2022 at 09:48:09PM -0400, guoren@xxxxxxxxxx wrote:
> From: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
> 
> Make generic_entry supports basic STACKLEAK, and no arch custom
> code is needed.

IIUC, this change is going to cause redundant work to be done on x86 (since it
erases the stack in its entry assembly). It also means any arch relying upon
this will not clear some stack contents that could be cleared from assembly
later in the return to userspace path, after the C entry code stack frames are
gone.

I assume you're adding this so that riscv can use stackleak? WHy can't it call
stackleak_erase*() later in the return-to-userspce path?

> Signed-off-by: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
> Signed-off-by: Guo Ren <guoren@xxxxxxxxxx>
> ---
>  drivers/firmware/efi/libstub/Makefile | 4 +++-
>  include/linux/stackleak.h             | 3 +++
>  kernel/entry/common.c                 | 5 +++++
>  security/Kconfig.hardening            | 2 +-
>  4 files changed, 12 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/firmware/efi/libstub/Makefile b/drivers/firmware/efi/libstub/Makefile
> index d0537573501e..bb6ad37a9690 100644
> --- a/drivers/firmware/efi/libstub/Makefile
> +++ b/drivers/firmware/efi/libstub/Makefile
> @@ -19,7 +19,7 @@ cflags-$(CONFIG_X86)		+= -m$(BITS) -D__KERNEL__ \
>  # arm64 uses the full KBUILD_CFLAGS so it's necessary to explicitly
>  # disable the stackleak plugin
>  cflags-$(CONFIG_ARM64)		:= $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
> -				   -fpie $(DISABLE_STACKLEAK_PLUGIN) \
> +				   -fpie \
>  				   $(call cc-option,-mbranch-protection=none)
>  cflags-$(CONFIG_ARM)		:= $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
>  				   -fno-builtin -fpic \
> @@ -27,6 +27,8 @@ cflags-$(CONFIG_ARM)		:= $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
>  cflags-$(CONFIG_RISCV)		:= $(subst $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
>  				   -fpic
>  
> +cflags-$(CONFIG_GCC_PLUGIN_STACKLEAK) += $(DISABLE_STACKLEAK_PLUGIN)
> +
>  cflags-$(CONFIG_EFI_GENERIC_STUB) += -I$(srctree)/scripts/dtc/libfdt
>  
>  KBUILD_CFLAGS			:= $(cflags-y) -Os -DDISABLE_BRANCH_PROFILING \

Huh; is there a latent bug here where x86's EFI stub is instrumented with
stackleak?

Thanks,
Mark.

> diff --git a/include/linux/stackleak.h b/include/linux/stackleak.h
> index c36e7a3b45e7..9890802a5868 100644
> --- a/include/linux/stackleak.h
> +++ b/include/linux/stackleak.h
> @@ -76,8 +76,11 @@ static inline void stackleak_task_init(struct task_struct *t)
>  # endif
>  }
>  
> +void noinstr stackleak_erase(void);
> +
>  #else /* !CONFIG_GCC_PLUGIN_STACKLEAK */
>  static inline void stackleak_task_init(struct task_struct *t) { }
> +static inline void stackleak_erase(void) {}
>  #endif
>  
>  #endif
> diff --git a/kernel/entry/common.c b/kernel/entry/common.c
> index 063068a9ea9b..6acb1d6a1396 100644
> --- a/kernel/entry/common.c
> +++ b/kernel/entry/common.c
> @@ -8,6 +8,7 @@
>  #include <linux/livepatch.h>
>  #include <linux/audit.h>
>  #include <linux/tick.h>
> +#include <linux/stackleak.h>
>  
>  #include "common.h"
>  
> @@ -194,6 +195,10 @@ static void exit_to_user_mode_prepare(struct pt_regs *regs)
>  
>  	lockdep_assert_irqs_disabled();
>  
> +#ifndef CONFIG_HAVE_ARCH_STACKLEAK
> +	stackleak_erase();
> +#endif
> +
>  	/* Flush pending rcuog wakeup before the last need_resched() check */
>  	tick_nohz_user_enter_prepare();
>  
> diff --git a/security/Kconfig.hardening b/security/Kconfig.hardening
> index bd2aabb2c60f..3329482beb8d 100644
> --- a/security/Kconfig.hardening
> +++ b/security/Kconfig.hardening
> @@ -152,7 +152,7 @@ config GCC_PLUGIN_STRUCTLEAK_VERBOSE
>  config GCC_PLUGIN_STACKLEAK
>  	bool "Poison kernel stack before returning from syscalls"
>  	depends on GCC_PLUGINS
> -	depends on HAVE_ARCH_STACKLEAK
> +	depends on HAVE_ARCH_STACKLEAK || GENERIC_ENTRY
>  	help
>  	  This option makes the kernel erase the kernel stack before
>  	  returning from system calls. This has the effect of leaving
> -- 
> 2.36.1
>