Re: [PATCH v3 05/17] add support for Clang's Shadow Call Stack (SCS)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Oct 31, 2019 at 09:46:25AM -0700, samitolvanen@xxxxxxxxxx wrote:
> This change adds generic support for Clang's Shadow Call Stack,
> which uses a shadow stack to protect return addresses from being
> overwritten by an attacker. Details are available here:
> 
>   https://clang.llvm.org/docs/ShadowCallStack.html
> 
> Note that security guarantees in the kernel differ from the
> ones documented for user space. The kernel must store addresses
> of shadow stacks used by other tasks and interrupt handlers in
> memory, which means an attacker capable reading and writing
> arbitrary memory may be able to locate them and hijack control
> flow by modifying shadow stacks that are not currently in use.
> 
> Signed-off-by: Sami Tolvanen <samitolvanen@xxxxxxxxxx>
> ---
>  Makefile                       |   6 ++
>  arch/Kconfig                   |  33 +++++++
>  include/linux/compiler-clang.h |   6 ++
>  include/linux/compiler_types.h |   4 +
>  include/linux/scs.h            |  54 +++++++++++
>  init/init_task.c               |   8 ++
>  kernel/Makefile                |   1 +
>  kernel/fork.c                  |   9 ++
>  kernel/sched/core.c            |   2 +
>  kernel/sched/sched.h           |   1 +
>  kernel/scs.c                   | 169 +++++++++++++++++++++++++++++++++
>  11 files changed, 293 insertions(+)
>  create mode 100644 include/linux/scs.h
>  create mode 100644 kernel/scs.c
> 
> diff --git a/Makefile b/Makefile
> index 79be70bf2899..e6337314f8fb 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -846,6 +846,12 @@ ifdef CONFIG_LIVEPATCH
>  KBUILD_CFLAGS += $(call cc-option, -flive-patching=inline-clone)
>  endif
>  
> +ifdef CONFIG_SHADOW_CALL_STACK
> +CC_FLAGS_SCS	:= -fsanitize=shadow-call-stack
> +KBUILD_CFLAGS	+= $(CC_FLAGS_SCS)
> +export CC_FLAGS_SCS
> +endif
> +
>  # arch Makefile may override CC so keep this after arch Makefile is included
>  NOSTDINC_FLAGS += -nostdinc -isystem $(shell $(CC) -print-file-name=include)
>  
> diff --git a/arch/Kconfig b/arch/Kconfig
> index 5f8a5d84dbbe..5e34cbcd8d6a 100644
> --- a/arch/Kconfig
> +++ b/arch/Kconfig
> @@ -521,6 +521,39 @@ config STACKPROTECTOR_STRONG
>  	  about 20% of all kernel functions, which increases the kernel code
>  	  size by about 2%.
>  
> +config ARCH_SUPPORTS_SHADOW_CALL_STACK
> +	bool
> +	help
> +	  An architecture should select this if it supports Clang's Shadow
> +	  Call Stack, has asm/scs.h, and implements runtime support for shadow
> +	  stack switching.
> +
> +config SHADOW_CALL_STACK_VMAP
> +	bool
> +	depends on SHADOW_CALL_STACK
> +	help
> +	  Use virtually mapped shadow call stacks. Selecting this option
> +	  provides better stack exhaustion protection, but increases per-thread
> +	  memory consumption as a full page is allocated for each shadow stack.
> +
> +config SHADOW_CALL_STACK
> +	bool "Clang Shadow Call Stack"
> +	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
> +	help
> +	  This option enables Clang's Shadow Call Stack, which uses a
> +	  shadow stack to protect function return addresses from being
> +	  overwritten by an attacker. More information can be found from
> +	  Clang's documentation:
> +
> +	    https://clang.llvm.org/docs/ShadowCallStack.html
> +
> +	  Note that security guarantees in the kernel differ from the ones
> +	  documented for user space. The kernel must store addresses of shadow
> +	  stacks used by other tasks and interrupt handlers in memory, which
> +	  means an attacker capable reading and writing arbitrary memory may
> +	  be able to locate them and hijack control flow by modifying shadow
> +	  stacks that are not currently in use.
> +
>  config HAVE_ARCH_WITHIN_STACK_FRAMES
>  	bool
>  	help
> diff --git a/include/linux/compiler-clang.h b/include/linux/compiler-clang.h
> index 333a6695a918..18fc4d29ef27 100644
> --- a/include/linux/compiler-clang.h
> +++ b/include/linux/compiler-clang.h
> @@ -42,3 +42,9 @@
>   * compilers, like ICC.
>   */
>  #define barrier() __asm__ __volatile__("" : : : "memory")
> +
> +#if __has_feature(shadow_call_stack)
> +# define __noscs	__attribute__((__no_sanitize__("shadow-call-stack")))
> +#else
> +# define __noscs
> +#endif
> diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
> index 72393a8c1a6c..be5d5be4b1ae 100644
> --- a/include/linux/compiler_types.h
> +++ b/include/linux/compiler_types.h
> @@ -202,6 +202,10 @@ struct ftrace_likely_data {
>  # define randomized_struct_fields_end
>  #endif
>  
> +#ifndef __noscs
> +# define __noscs
> +#endif
> +
>  #ifndef asm_volatile_goto
>  #define asm_volatile_goto(x...) asm goto(x)
>  #endif
> diff --git a/include/linux/scs.h b/include/linux/scs.h
> new file mode 100644
> index 000000000000..0b70aff3846a
> --- /dev/null
> +++ b/include/linux/scs.h
> @@ -0,0 +1,54 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Shadow Call Stack support.
> + *
> + * Copyright (C) 2019 Google LLC
> + */
> +
> +#ifndef _LINUX_SCS_H
> +#define _LINUX_SCS_H
> +
> +#include <linux/gfp.h>
> +#include <linux/sched.h>
> +#include <asm/page.h>
> +
> +#ifdef CONFIG_SHADOW_CALL_STACK
> +
> +/*
> + * In testing, 1 KiB shadow stack size (i.e. 128 stack frames on a 64-bit
> + * architecture) provided ~40% safety margin on stack usage while keeping
> + * memory allocation overhead reasonable.
> + */
> +#define SCS_SIZE	1024
> +#define GFP_SCS		(GFP_KERNEL | __GFP_ZERO)
> +
> +/* A random number to mark the end of the shadow stack. */
> +#define SCS_END_MAGIC	0xaf0194819b1635f6UL

Is 0xaf.... non-canonical for arm64? While "random", it should also
likely be an "impossible" value to find on the call stack.

Otherwise, all looks great. :)

Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>

-Kees

> +
> +#define task_scs(tsk)	(task_thread_info(tsk)->shadow_call_stack)
> +
> +static inline void task_set_scs(struct task_struct *tsk, void *s)
> +{
> +	task_scs(tsk) = s;
> +}
> +
> +extern void scs_init(void);
> +extern void scs_task_reset(struct task_struct *tsk);
> +extern int scs_prepare(struct task_struct *tsk, int node);
> +extern bool scs_corrupted(struct task_struct *tsk);
> +extern void scs_release(struct task_struct *tsk);
> +
> +#else /* CONFIG_SHADOW_CALL_STACK */
> +
> +#define task_scs(tsk)	NULL
> +
> +static inline void task_set_scs(struct task_struct *tsk, void *s) {}
> +static inline void scs_init(void) {}
> +static inline void scs_task_reset(struct task_struct *tsk) {}
> +static inline int scs_prepare(struct task_struct *tsk, int node) { return 0; }
> +static inline bool scs_corrupted(struct task_struct *tsk) { return false; }
> +static inline void scs_release(struct task_struct *tsk) {}
> +
> +#endif /* CONFIG_SHADOW_CALL_STACK */
> +
> +#endif /* _LINUX_SCS_H */
> diff --git a/init/init_task.c b/init/init_task.c
> index 9e5cbe5eab7b..cbd40460e903 100644
> --- a/init/init_task.c
> +++ b/init/init_task.c
> @@ -11,6 +11,7 @@
>  #include <linux/mm.h>
>  #include <linux/audit.h>
>  #include <linux/numa.h>
> +#include <linux/scs.h>
>  
>  #include <asm/pgtable.h>
>  #include <linux/uaccess.h>
> @@ -184,6 +185,13 @@ struct task_struct init_task
>  };
>  EXPORT_SYMBOL(init_task);
>  
> +#ifdef CONFIG_SHADOW_CALL_STACK
> +unsigned long init_shadow_call_stack[SCS_SIZE / sizeof(long)] __init_task_data
> +		__aligned(SCS_SIZE) = {
> +	[(SCS_SIZE / sizeof(long)) - 1] = SCS_END_MAGIC
> +};
> +#endif
> +
>  /*
>   * Initial thread structure. Alignment of this is handled by a special
>   * linker map entry.
> diff --git a/kernel/Makefile b/kernel/Makefile
> index daad787fb795..313dbd44d576 100644
> --- a/kernel/Makefile
> +++ b/kernel/Makefile
> @@ -102,6 +102,7 @@ obj-$(CONFIG_TRACEPOINTS) += trace/
>  obj-$(CONFIG_IRQ_WORK) += irq_work.o
>  obj-$(CONFIG_CPU_PM) += cpu_pm.o
>  obj-$(CONFIG_BPF) += bpf/
> +obj-$(CONFIG_SHADOW_CALL_STACK) += scs.o
>  
>  obj-$(CONFIG_PERF_EVENTS) += events/
>  
> diff --git a/kernel/fork.c b/kernel/fork.c
> index bcdf53125210..3fa7ba64c62d 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -94,6 +94,7 @@
>  #include <linux/livepatch.h>
>  #include <linux/thread_info.h>
>  #include <linux/stackleak.h>
> +#include <linux/scs.h>
>  
>  #include <asm/pgtable.h>
>  #include <asm/pgalloc.h>
> @@ -451,6 +452,8 @@ void put_task_stack(struct task_struct *tsk)
>  
>  void free_task(struct task_struct *tsk)
>  {
> +	scs_release(tsk);
> +
>  #ifndef CONFIG_THREAD_INFO_IN_TASK
>  	/*
>  	 * The task is finally done with both the stack and thread_info,
> @@ -834,6 +837,8 @@ void __init fork_init(void)
>  			  NULL, free_vm_stack_cache);
>  #endif
>  
> +	scs_init();
> +
>  	lockdep_init_task(&init_task);
>  	uprobes_init();
>  }
> @@ -893,6 +898,10 @@ static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
>  	if (err)
>  		goto free_stack;
>  
> +	err = scs_prepare(tsk, node);
> +	if (err)
> +		goto free_stack;
> +
>  #ifdef CONFIG_SECCOMP
>  	/*
>  	 * We must handle setting up seccomp filters once we're under
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index dd05a378631a..e7faeb383008 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -6013,6 +6013,8 @@ void init_idle(struct task_struct *idle, int cpu)
>  	raw_spin_lock_irqsave(&idle->pi_lock, flags);
>  	raw_spin_lock(&rq->lock);
>  
> +	scs_task_reset(idle);
> +
>  	__sched_fork(0, idle);
>  	idle->state = TASK_RUNNING;
>  	idle->se.exec_start = sched_clock();
> diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> index 0db2c1b3361e..c153003a011c 100644
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -58,6 +58,7 @@
>  #include <linux/profile.h>
>  #include <linux/psi.h>
>  #include <linux/rcupdate_wait.h>
> +#include <linux/scs.h>
>  #include <linux/security.h>
>  #include <linux/stop_machine.h>
>  #include <linux/suspend.h>
> diff --git a/kernel/scs.c b/kernel/scs.c
> new file mode 100644
> index 000000000000..7c1a40020754
> --- /dev/null
> +++ b/kernel/scs.c
> @@ -0,0 +1,169 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Shadow Call Stack support.
> + *
> + * Copyright (C) 2019 Google LLC
> + */
> +
> +#include <linux/cpuhotplug.h>
> +#include <linux/mm.h>
> +#include <linux/mmzone.h>
> +#include <linux/scs.h>
> +#include <linux/slab.h>
> +#include <linux/vmalloc.h>
> +#include <asm/scs.h>
> +
> +static inline void *__scs_base(struct task_struct *tsk)
> +{
> +	/*
> +	 * We allow architectures to use the shadow_call_stack field in
> +	 * struct thread_info to store the current shadow stack pointer
> +	 * during context switches.
> +	 *
> +	 * This allows the implementation to also clear the field when
> +	 * the task is active to avoid keeping pointers to the current
> +	 * task's shadow stack in memory. This can make it harder for an
> +	 * attacker to locate the shadow stack, but also requires us to
> +	 * compute the base address when needed.
> +	 *
> +	 * We assume the stack is aligned to SCS_SIZE.
> +	 */
> +	return (void *)((uintptr_t)task_scs(tsk) & ~(SCS_SIZE - 1));
> +}
> +
> +#ifdef CONFIG_SHADOW_CALL_STACK_VMAP
> +
> +/* Keep a cache of shadow stacks */
> +#define SCS_CACHE_SIZE 2
> +static DEFINE_PER_CPU(void *, scs_cache[SCS_CACHE_SIZE]);
> +
> +static void *scs_alloc(int node)
> +{
> +	int i;
> +
> +	for (i = 0; i < SCS_CACHE_SIZE; i++) {
> +		void *s;
> +
> +		s = this_cpu_xchg(scs_cache[i], NULL);
> +		if (s) {
> +			memset(s, 0, SCS_SIZE);
> +			return s;
> +		}
> +	}
> +
> +	/*
> +	 * We allocate a full page for the shadow stack, which should be
> +	 * more than we need. Check the assumption nevertheless.
> +	 */
> +	BUILD_BUG_ON(SCS_SIZE > PAGE_SIZE);
> +
> +	return __vmalloc_node_range(PAGE_SIZE, SCS_SIZE,
> +				    VMALLOC_START, VMALLOC_END,
> +				    GFP_SCS, PAGE_KERNEL, 0,
> +				    node, __builtin_return_address(0));
> +}
> +
> +static void scs_free(void *s)
> +{
> +	int i;
> +
> +	for (i = 0; i < SCS_CACHE_SIZE; i++)
> +		if (this_cpu_cmpxchg(scs_cache[i], 0, s) == 0)
> +			return;
> +
> +	vfree_atomic(s);
> +}
> +
> +static int scs_cleanup(unsigned int cpu)
> +{
> +	int i;
> +	void **cache = per_cpu_ptr(scs_cache, cpu);
> +
> +	for (i = 0; i < SCS_CACHE_SIZE; i++) {
> +		vfree(cache[i]);
> +		cache[i] = NULL;
> +	}
> +
> +	return 0;
> +}
> +
> +void __init scs_init(void)
> +{
> +	cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "scs:scs_cache", NULL,
> +		scs_cleanup);
> +}
> +
> +#else /* !CONFIG_SHADOW_CALL_STACK_VMAP */
> +
> +static struct kmem_cache *scs_cache;
> +
> +static inline void *scs_alloc(int node)
> +{
> +	return kmem_cache_alloc_node(scs_cache, GFP_SCS, node);
> +}
> +
> +static inline void scs_free(void *s)
> +{
> +	kmem_cache_free(scs_cache, s);
> +}
> +
> +void __init scs_init(void)
> +{
> +	scs_cache = kmem_cache_create("scs_cache", SCS_SIZE, SCS_SIZE,
> +				0, NULL);
> +	WARN_ON(!scs_cache);
> +}
> +
> +#endif /* CONFIG_SHADOW_CALL_STACK_VMAP */
> +
> +static inline unsigned long *scs_magic(struct task_struct *tsk)
> +{
> +	return (unsigned long *)(__scs_base(tsk) + SCS_SIZE) - 1;
> +}
> +
> +static inline void scs_set_magic(struct task_struct *tsk)
> +{
> +	*scs_magic(tsk) = SCS_END_MAGIC;
> +}
> +
> +void scs_task_reset(struct task_struct *tsk)
> +{
> +	/*
> +	 * Reset the shadow stack to the base address in case the task
> +	 * is reused.
> +	 */
> +	task_set_scs(tsk, __scs_base(tsk));
> +}
> +
> +int scs_prepare(struct task_struct *tsk, int node)
> +{
> +	void *s;
> +
> +	s = scs_alloc(node);
> +	if (!s)
> +		return -ENOMEM;
> +
> +	task_set_scs(tsk, s);
> +	scs_set_magic(tsk);
> +
> +	return 0;
> +}
> +
> +bool scs_corrupted(struct task_struct *tsk)
> +{
> +	return *scs_magic(tsk) != SCS_END_MAGIC;
> +}
> +
> +void scs_release(struct task_struct *tsk)
> +{
> +	void *s;
> +
> +	s = __scs_base(tsk);
> +	if (!s)
> +		return;
> +
> +	WARN_ON(scs_corrupted(tsk));
> +
> +	task_set_scs(tsk, NULL);
> +	scs_free(s);
> +}
> -- 
> 2.24.0.rc0.303.g954a862665-goog
> 

-- 
Kees Cook



[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux