Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1631657 > unrolled thread

[PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode

Started byThomas Garnier <thgarnie@google.com>
First post2017-04-26 20:40 +0200
Last post2017-04-28 08:40 +0200
Articles 7 — 2 participants

Back to article view | Back to linux.kernel


Contents

  [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode Thomas Garnier <thgarnie@google.com> - 2017-04-26 20:40 +0200
    [PATCH v8 3/4] arm/syscalls: Optimize address limit check Thomas Garnier <thgarnie@google.com> - 2017-04-26 20:40 +0200
    Re: [PATCH v8 1/4] syscalls: Verify address limit before returning  to user-mode Ingo Molnar <mingo@kernel.org> - 2017-04-27 08:50 +0200
      Re: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode Thomas Garnier <thgarnie@google.com> - 2017-04-27 16:20 +0200
        [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode Thomas Garnier <thgarnie@google.com> - 2017-04-27 16:50 +0200
          Re: [PATCH v8 1/4] syscalls: Verify address limit before returning  to user-mode Ingo Molnar <mingo@kernel.org> - 2017-04-28 08:40 +0200
        Re: [PATCH v8 1/4] syscalls: Verify address limit before returning  to user-mode Ingo Molnar <mingo@kernel.org> - 2017-04-28 08:40 +0200

#1631657 — [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode

FromThomas Garnier <thgarnie@google.com>
Date2017-04-26 20:40 +0200
Subject[PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode
Message-ID<tAA14-61F-13@gated-at.bofh.it>
Ensure that a syscall does not return to user-mode with a kernel address
limit. If that happens, a process can corrupt kernel-mode memory and
elevate privileges [1].

The CONFIG_ADDR_LIMIT_CHECK option disables the generic check so each
architecture can create optimized versions.

[1] https://bugs.chromium.org/p/project-zero/issues/detail?id=990

Signed-off-by: Thomas Garnier <thgarnie@google.com>
Tested-by: Kees Cook <keescook@chromium.org>
---
Based on next-20170426
---
 arch/s390/Kconfig        |  1 +
 include/linux/syscalls.h | 27 ++++++++++++++++++++++++++-
 init/Kconfig             |  6 ++++++
 kernel/sys.c             | 13 +++++++++++++
 4 files changed, 46 insertions(+), 1 deletion(-)

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index d25435d94b6e..164de1d24e92 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -103,6 +103,7 @@ config S390
 	select ARCH_INLINE_WRITE_UNLOCK_BH
 	select ARCH_INLINE_WRITE_UNLOCK_IRQ
 	select ARCH_INLINE_WRITE_UNLOCK_IRQRESTORE
+	select ADDR_LIMIT_CHECK
 	select ARCH_SAVE_PAGE_KEYS if HIBERNATION
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index 980c3c9b06f8..ebde64f1622c 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -191,6 +191,28 @@ extern struct trace_event_functions exit_syscall_print_funcs;
 	SYSCALL_METADATA(sname, x, __VA_ARGS__)			\
 	__SYSCALL_DEFINEx(x, sname, __VA_ARGS__)
 
+
+/*
+ * Called before coming back to user-mode. Returning to user-mode with an
+ * address limit different than USER_DS can allow to overwrite kernel memory.
+ */
+static inline void addr_limit_check_syscall(void)
+{
+	BUG_ON(!segment_eq(get_fs(), USER_DS));
+}
+
+#ifndef CONFIG_ADDR_LIMIT_CHECK
+#define __CHECK_USERMODE_SYSCALL() \
+	bool user_caller = segment_eq(get_fs(), USER_DS)
+#define __VERIFY_ADDR_LIMIT() \
+	if (user_caller) addr_limit_check_syscall()
+#else
+#define __CHECK_USERMODE_SYSCALL()
+#define __VERIFY_ADDR_LIMIT()
+asmlinkage void addr_limit_check_failed(void) __noreturn;
+#endif
+
+
 #define __PROTECT(...) asmlinkage_protect(__VA_ARGS__)
 #define __SYSCALL_DEFINEx(x, name, ...)					\
 	asmlinkage long sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))	\
@@ -199,7 +221,10 @@ extern struct trace_event_functions exit_syscall_print_funcs;
 	asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__));	\
 	asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__))	\
 	{								\
-		long ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
+		long ret;						\
+		__CHECK_USERMODE_SYSCALL();				\
+		ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
+		__VERIFY_ADDR_LIMIT();					\
 		__MAP(x,__SC_TEST,__VA_ARGS__);				\
 		__PROTECT(x, ret,__MAP(x,__SC_ARGS,__VA_ARGS__));	\
 		return ret;						\
diff --git a/init/Kconfig b/init/Kconfig
index 42a346b0df43..599d9fe30703 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1961,6 +1961,12 @@ config PROFILING
 config TRACEPOINTS
 	bool
 
+config ADDR_LIMIT_CHECK
+	bool
+	help
+	  Disable the generic address limit check. Allow each architecture to
+	  optimize how and when the verification is done.
+
 source "arch/Kconfig"
 
 endmenu		# General setup
diff --git a/kernel/sys.c b/kernel/sys.c
index 8a94b4eabcaa..a1cbcd715d62 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -2458,3 +2458,16 @@ COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
 	return 0;
 }
 #endif /* CONFIG_COMPAT */
+
+#ifdef CONFIG_ADDR_LIMIT_CHECK
+/*
+ * Used when an architecture specific implementation detects an invalid address
+ * limit. This function does not return.
+ */
+asmlinkage void addr_limit_check_failed(void)
+{
+	/* Try to fail on the generic address limit check */
+	addr_limit_check_syscall();
+	panic("Invalid address limit before returning to user-mode");
+}
+#endif
-- 
2.13.0.rc0.306.g87b477812d-goog

[toc] | [next] | [standalone]


#1631661 — [PATCH v8 3/4] arm/syscalls: Optimize address limit check

FromThomas Garnier <thgarnie@google.com>
Date2017-04-26 20:40 +0200
Subject[PATCH v8 3/4] arm/syscalls: Optimize address limit check
Message-ID<tAA14-61F-25@gated-at.bofh.it>
In reply to#1631657
Disable the generic address limit check in favor of an architecture
specific optimized implementation.

The address limit is checked on each syscall return path to user-mode
path as well as the irq user-mode return function. If the address limit
was changed, a generic handler is called to stop the kernel on an
explicit check.

Signed-off-by: Thomas Garnier <thgarnie@google.com>
---
Based on next-20170426
---
 arch/arm/Kconfig               |  1 +
 arch/arm/kernel/entry-common.S | 10 +++++++++-
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index 4c1a35f15838..f17f30084a4e 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -12,6 +12,7 @@ config ARM
 	select ARCH_HAVE_CUSTOM_GPIO_H
 	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_MIGHT_HAVE_PC_PARPORT
+	select ADDR_LIMIT_CHECK
 	select ARCH_OPTIONAL_KERNEL_RWX if ARCH_HAS_STRICT_KERNEL_RWX
 	select ARCH_OPTIONAL_KERNEL_RWX_DEFAULT if CPU_V7
 	select ARCH_SUPPORTS_ATOMIC_RMW
diff --git a/arch/arm/kernel/entry-common.S b/arch/arm/kernel/entry-common.S
index eb5cd77bf1d8..c83927498f40 100644
--- a/arch/arm/kernel/entry-common.S
+++ b/arch/arm/kernel/entry-common.S
@@ -12,6 +12,7 @@
 #include <asm/unistd.h>
 #include <asm/ftrace.h>
 #include <asm/unwind.h>
+#include <asm/memory.h>
 #ifdef CONFIG_AEABI
 #include <asm/unistd-oabi.h>
 #endif
@@ -27,7 +28,6 @@
 
 #include "entry-header.S"
 
-
 	.align	5
 #if !(IS_ENABLED(CONFIG_TRACE_IRQFLAGS) || IS_ENABLED(CONFIG_CONTEXT_TRACKING))
 /*
@@ -40,9 +40,12 @@ ret_fast_syscall:
  UNWIND(.fnstart	)
  UNWIND(.cantunwind	)
 	disable_irq_notrace			@ disable interrupts
+	ldr	r2, [tsk, #TI_ADDR_LIMIT]
 	ldr	r1, [tsk, #TI_FLAGS]		@ re-check for syscall tracing
 	tst	r1, #_TIF_SYSCALL_WORK | _TIF_WORK_MASK
 	bne	fast_work_pending
+	cmp	r2, #TASK_SIZE
+	blne	addr_limit_check_failed
 
 	/* perform architecture specific actions before user return */
 	arch_ret_to_user r1, lr
@@ -66,6 +69,7 @@ ret_fast_syscall:
  UNWIND(.cantunwind	)
 	str	r0, [sp, #S_R0 + S_OFF]!	@ save returned r0
 	disable_irq_notrace			@ disable interrupts
+	ldr	r2, [tsk, #TI_ADDR_LIMIT]
 	ldr	r1, [tsk, #TI_FLAGS]		@ re-check for syscall tracing
 	tst	r1, #_TIF_SYSCALL_WORK | _TIF_WORK_MASK
 	beq	no_work_pending
@@ -82,6 +86,7 @@ slow_work_pending:
 	mov	r2, why				@ 'syscall'
 	bl	do_work_pending
 	cmp	r0, #0
+	ldreq   r2, [tsk, #TI_ADDR_LIMIT]
 	beq	no_work_pending
 	movlt	scno, #(__NR_restart_syscall - __NR_SYSCALL_BASE)
 	ldmia	sp, {r0 - r6}			@ have to reload r0 - r6
@@ -99,9 +104,12 @@ ret_slow_syscall:
 	disable_irq_notrace			@ disable interrupts
 ENTRY(ret_to_user_from_irq)
 	ldr	r1, [tsk, #TI_FLAGS]
+	ldr	r2, [tsk, #TI_ADDR_LIMIT]
 	tst	r1, #_TIF_WORK_MASK
 	bne	slow_work_pending
 no_work_pending:
+	cmp	r2, #TASK_SIZE
+	blne	addr_limit_check_failed
 	asm_trace_hardirqs_on save = 0
 
 	/* perform architecture specific actions before user return */
-- 
2.13.0.rc0.306.g87b477812d-goog

[toc] | [prev] | [next] | [standalone]


#1631900 — Re: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode

FromIngo Molnar <mingo@kernel.org>
Date2017-04-27 08:50 +0200
SubjectRe: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode
Message-ID<tALpv-5dO-11@gated-at.bofh.it>
In reply to#1631657
* Thomas Garnier <thgarnie@google.com> wrote:

> +
> +/*
> + * Called before coming back to user-mode. Returning to user-mode with an
> + * address limit different than USER_DS can allow to overwrite kernel memory.
> + */
> +static inline void addr_limit_check_syscall(void)
> +{
> +	BUG_ON(!segment_eq(get_fs(), USER_DS));
> +}
> +
> +#ifndef CONFIG_ADDR_LIMIT_CHECK
> +#define __CHECK_USERMODE_SYSCALL() \
> +	bool user_caller = segment_eq(get_fs(), USER_DS)
> +#define __VERIFY_ADDR_LIMIT() \
> +	if (user_caller) addr_limit_check_syscall()
> +#else
> +#define __CHECK_USERMODE_SYSCALL()
> +#define __VERIFY_ADDR_LIMIT()
> +asmlinkage void addr_limit_check_failed(void) __noreturn;
> +#endif

_Please_ harmonize all the externally exposed names and symbols.

There's no reason for this mismash of names:

	CONFIG_ADDR_LIMIT_CHECK

	__CHECK_USERMODE_SYSCALL
	__VERIFY_ADDR_LIMIT

When we could just as easily name them consistently, along the existing pattern:

	CONFIG_ADDR_LIMIT_CHECK

	__SYSCALL_ADDR_LIMIT_CHECK
	__ADDR_LIMIT_CHECK

which should fit into existing nomenclature:

>  #define __SYSCALL_DEFINEx(x, name, ...)					\

But even with that fixed, the whole construct still looks pretty weird:

>  	{								\
> -		long ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
> +		long ret;						\
> +		__CHECK_USERMODE_SYSCALL();				\
> +		ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
> +		__ADDR_LIMIT_CHECK();					\
>  		__MAP(x,__SC_TEST,__VA_ARGS__);				\
>  		__PROTECT(x, ret,__MAP(x,__SC_ARGS,__VA_ARGS__));	\
>  		return ret;						\

I think something like this would be more natural to read:

> +		ADDR_LIMIT_CHECK_PRE();					\
> +		ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
> +		ADDR_LIMIT_CHECK_POST();				\

it's a clear pre/post construct. Also note the lack of double underscores.

BTW., a further simplification would be:

#ifndef ADDR_LIMIT_CHECK_PRE
# define ADDR_LIMIT_CHECK_PRE ...
#endif

This way architectures could override this generic functionality simply by 
defining the helpers. Architectures that don't do that get the generic version. 

Thanks,

	Ingo

[toc] | [prev] | [next] | [standalone]


#1632124

FromThomas Garnier <thgarnie@google.com>
Date2017-04-27 16:20 +0200
Message-ID<tASr0-1Iz-15@gated-at.bofh.it>
In reply to#1631900
On Wed, Apr 26, 2017 at 11:49 PM, Ingo Molnar <mingo@kernel.org> wrote:
>
> * Thomas Garnier <thgarnie@google.com> wrote:
>
>> +
>> +/*
>> + * Called before coming back to user-mode. Returning to user-mode with an
>> + * address limit different than USER_DS can allow to overwrite kernel memory.
>> + */
>> +static inline void addr_limit_check_syscall(void)
>> +{
>> +     BUG_ON(!segment_eq(get_fs(), USER_DS));
>> +}
>> +
>> +#ifndef CONFIG_ADDR_LIMIT_CHECK
>> +#define __CHECK_USERMODE_SYSCALL() \
>> +     bool user_caller = segment_eq(get_fs(), USER_DS)
>> +#define __VERIFY_ADDR_LIMIT() \
>> +     if (user_caller) addr_limit_check_syscall()
>> +#else
>> +#define __CHECK_USERMODE_SYSCALL()
>> +#define __VERIFY_ADDR_LIMIT()
>> +asmlinkage void addr_limit_check_failed(void) __noreturn;
>> +#endif
>
> _Please_ harmonize all the externally exposed names and symbols.
>
> There's no reason for this mismash of names:
>
>         CONFIG_ADDR_LIMIT_CHECK
>
>         __CHECK_USERMODE_SYSCALL
>         __VERIFY_ADDR_LIMIT
>
> When we could just as easily name them consistently, along the existing pattern:
>
>         CONFIG_ADDR_LIMIT_CHECK
>
>         __SYSCALL_ADDR_LIMIT_CHECK
>         __ADDR_LIMIT_CHECK
>
> which should fit into existing nomenclature:
>
>>  #define __SYSCALL_DEFINEx(x, name, ...)                                      \
>
> But even with that fixed, the whole construct still looks pretty weird:
>
>>       {                                                               \
>> -             long ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));  \
>> +             long ret;                                               \
>> +             __CHECK_USERMODE_SYSCALL();                             \
>> +             ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));       \
>> +             __ADDR_LIMIT_CHECK();                                   \
>>               __MAP(x,__SC_TEST,__VA_ARGS__);                         \
>>               __PROTECT(x, ret,__MAP(x,__SC_ARGS,__VA_ARGS__));       \
>>               return ret;                                             \
>
> I think something like this would be more natural to read:
>
>> +             ADDR_LIMIT_CHECK_PRE();                                 \
>> +             ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));       \
>> +             ADDR_LIMIT_CHECK_POST();                                \
>
> it's a clear pre/post construct. Also note the lack of double underscores.

I think this construct makes more sense because the first macro check
if the syscall was called by user-mode. I will send an update for this
on this thread.

>
> BTW., a further simplification would be:
>
> #ifndef ADDR_LIMIT_CHECK_PRE
> # define ADDR_LIMIT_CHECK_PRE ...
> #endif
>
> This way architectures could override this generic functionality simply by
> defining the helpers. Architectures that don't do that get the generic version.

I don't think architectures need to do that. The optimizations are
embedding the checks on their architecture-specific code to make it
faster and remove the size impact. The pre/post is fine for the rest.

>
> Thanks,
>
>         Ingo



-- 
Thomas

[toc] | [prev] | [next] | [standalone]


#1632147

FromThomas Garnier <thgarnie@google.com>
Date2017-04-27 16:50 +0200
Message-ID<tASU1-1Wr-5@gated-at.bofh.it>
In reply to#1632124
Ensure that a syscall does not return to user-mode with a kernel address
limit. If that happens, a process can corrupt kernel-mode memory and
elevate privileges [1].

The CONFIG_ADDR_LIMIT_CHECK option disables the generic check so each
architecture can create optimized versions.

[1] https://bugs.chromium.org/p/project-zero/issues/detail?id=990

Signed-off-by: Thomas Garnier <thgarnie@google.com>
Tested-by: Kees Cook <keescook@chromium.org>
---
Based on next-20170426
---
 arch/s390/Kconfig        |  1 +
 include/linux/syscalls.h | 27 ++++++++++++++++++++++++++-
 init/Kconfig             |  6 ++++++
 kernel/sys.c             | 13 +++++++++++++
 4 files changed, 46 insertions(+), 1 deletion(-)

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index d25435d94b6e..164de1d24e92 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -103,6 +103,7 @@ config S390
 	select ARCH_INLINE_WRITE_UNLOCK_BH
 	select ARCH_INLINE_WRITE_UNLOCK_IRQ
 	select ARCH_INLINE_WRITE_UNLOCK_IRQRESTORE
+	select ADDR_LIMIT_CHECK
 	select ARCH_SAVE_PAGE_KEYS if HIBERNATION
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
index 980c3c9b06f8..ebde64f1622c 100644
--- a/include/linux/syscalls.h
+++ b/include/linux/syscalls.h
@@ -191,6 +191,28 @@ extern struct trace_event_functions exit_syscall_print_funcs;
 	SYSCALL_METADATA(sname, x, __VA_ARGS__)			\
 	__SYSCALL_DEFINEx(x, sname, __VA_ARGS__)
 
+
+/*
+ * Called before coming back to user-mode. Returning to user-mode with an
+ * address limit different than USER_DS can allow to overwrite kernel memory.
+ */
+static inline void addr_limit_check_syscall(void)
+{
+	BUG_ON(!segment_eq(get_fs(), USER_DS));
+}
+
+#ifndef CONFIG_ADDR_LIMIT_CHECK
+#define ADDR_LIMIT_CHECK_PRE() \
+	bool user_caller = segment_eq(get_fs(), USER_DS)
+#define ADDR_LIMIT_CHECK_POST() \
+	if (user_caller) addr_limit_check_syscall()
+#else
+#define ADDR_LIMIT_CHECK_PRE()
+#define ADDR_LIMIT_CHECK_POST()
+asmlinkage void addr_limit_check_failed(void) __noreturn;
+#endif
+
+
 #define __PROTECT(...) asmlinkage_protect(__VA_ARGS__)
 #define __SYSCALL_DEFINEx(x, name, ...)					\
 	asmlinkage long sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))	\
@@ -199,7 +221,10 @@ extern struct trace_event_functions exit_syscall_print_funcs;
 	asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__));	\
 	asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__))	\
 	{								\
-		long ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
+		long ret;						\
+		ADDR_LIMIT_CHECK_PRE();					\
+		ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
+		ADDR_LIMIT_CHECK_POST();				\
 		__MAP(x,__SC_TEST,__VA_ARGS__);				\
 		__PROTECT(x, ret,__MAP(x,__SC_ARGS,__VA_ARGS__));	\
 		return ret;						\
diff --git a/init/Kconfig b/init/Kconfig
index 42a346b0df43..599d9fe30703 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -1961,6 +1961,12 @@ config PROFILING
 config TRACEPOINTS
 	bool
 
+config ADDR_LIMIT_CHECK
+	bool
+	help
+	  Disable the generic address limit check. Allow each architecture to
+	  optimize how and when the verification is done.
+
 source "arch/Kconfig"
 
 endmenu		# General setup
diff --git a/kernel/sys.c b/kernel/sys.c
index 8a94b4eabcaa..a1cbcd715d62 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -2458,3 +2458,16 @@ COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
 	return 0;
 }
 #endif /* CONFIG_COMPAT */
+
+#ifdef CONFIG_ADDR_LIMIT_CHECK
+/*
+ * Used when an architecture specific implementation detects an invalid address
+ * limit. This function does not return.
+ */
+asmlinkage void addr_limit_check_failed(void)
+{
+	/* Try to fail on the generic address limit check */
+	addr_limit_check_syscall();
+	panic("Invalid address limit before returning to user-mode");
+}
+#endif
-- 
2.13.0.rc0.306.g87b477812d-goog

[toc] | [prev] | [next] | [standalone]


#1632557 — Re: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode

FromIngo Molnar <mingo@kernel.org>
Date2017-04-28 08:40 +0200
SubjectRe: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode
Message-ID<tB7Jn-3F8-11@gated-at.bofh.it>
In reply to#1632147
* Thomas Garnier <thgarnie@google.com> wrote:

> Ensure that a syscall does not return to user-mode with a kernel address
> limit. If that happens, a process can corrupt kernel-mode memory and
> elevate privileges [1].
> 
> The CONFIG_ADDR_LIMIT_CHECK option disables the generic check so each
> architecture can create optimized versions.
> 
> [1] https://bugs.chromium.org/p/project-zero/issues/detail?id=990
> 
> Signed-off-by: Thomas Garnier <thgarnie@google.com>
> Tested-by: Kees Cook <keescook@chromium.org>
> ---
> Based on next-20170426
> ---
>  arch/s390/Kconfig        |  1 +
>  include/linux/syscalls.h | 27 ++++++++++++++++++++++++++-
>  init/Kconfig             |  6 ++++++
>  kernel/sys.c             | 13 +++++++++++++
>  4 files changed, 46 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
> index d25435d94b6e..164de1d24e92 100644
> --- a/arch/s390/Kconfig
> +++ b/arch/s390/Kconfig
> @@ -103,6 +103,7 @@ config S390
>  	select ARCH_INLINE_WRITE_UNLOCK_BH
>  	select ARCH_INLINE_WRITE_UNLOCK_IRQ
>  	select ARCH_INLINE_WRITE_UNLOCK_IRQRESTORE
> +	select ADDR_LIMIT_CHECK
>  	select ARCH_SAVE_PAGE_KEYS if HIBERNATION
>  	select ARCH_SUPPORTS_ATOMIC_RMW
>  	select ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT
> diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h
> index 980c3c9b06f8..ebde64f1622c 100644
> --- a/include/linux/syscalls.h
> +++ b/include/linux/syscalls.h
> @@ -191,6 +191,28 @@ extern struct trace_event_functions exit_syscall_print_funcs;
>  	SYSCALL_METADATA(sname, x, __VA_ARGS__)			\
>  	__SYSCALL_DEFINEx(x, sname, __VA_ARGS__)
>  
> +
> +/*
> + * Called before coming back to user-mode. Returning to user-mode with an
> + * address limit different than USER_DS can allow to overwrite kernel memory.
> + */
> +static inline void addr_limit_check_syscall(void)
> +{
> +	BUG_ON(!segment_eq(get_fs(), USER_DS));
> +}
> +
> +#ifndef CONFIG_ADDR_LIMIT_CHECK
> +#define ADDR_LIMIT_CHECK_PRE() \
> +	bool user_caller = segment_eq(get_fs(), USER_DS)
> +#define ADDR_LIMIT_CHECK_POST() \
> +	if (user_caller) addr_limit_check_syscall()
> +#else
> +#define ADDR_LIMIT_CHECK_PRE()
> +#define ADDR_LIMIT_CHECK_POST()
> +asmlinkage void addr_limit_check_failed(void) __noreturn;
> +#endif
> +
> +
>  #define __PROTECT(...) asmlinkage_protect(__VA_ARGS__)
>  #define __SYSCALL_DEFINEx(x, name, ...)					\
>  	asmlinkage long sys##name(__MAP(x,__SC_DECL,__VA_ARGS__))	\
> @@ -199,7 +221,10 @@ extern struct trace_event_functions exit_syscall_print_funcs;
>  	asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__));	\
>  	asmlinkage long SyS##name(__MAP(x,__SC_LONG,__VA_ARGS__))	\
>  	{								\
> -		long ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
> +		long ret;						\
> +		ADDR_LIMIT_CHECK_PRE();					\
> +		ret = SYSC##name(__MAP(x,__SC_CAST,__VA_ARGS__));	\
> +		ADDR_LIMIT_CHECK_POST();				\
>  		__MAP(x,__SC_TEST,__VA_ARGS__);				\
>  		__PROTECT(x, ret,__MAP(x,__SC_ARGS,__VA_ARGS__));	\
>  		return ret;						\
> diff --git a/init/Kconfig b/init/Kconfig
> index 42a346b0df43..599d9fe30703 100644
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -1961,6 +1961,12 @@ config PROFILING
>  config TRACEPOINTS
>  	bool
>  
> +config ADDR_LIMIT_CHECK
> +	bool
> +	help
> +	  Disable the generic address limit check. Allow each architecture to
> +	  optimize how and when the verification is done.
> +
>  source "arch/Kconfig"
>  
>  endmenu		# General setup
> diff --git a/kernel/sys.c b/kernel/sys.c
> index 8a94b4eabcaa..a1cbcd715d62 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -2458,3 +2458,16 @@ COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
>  	return 0;
>  }
>  #endif /* CONFIG_COMPAT */
> +
> +#ifdef CONFIG_ADDR_LIMIT_CHECK
> +/*
> + * Used when an architecture specific implementation detects an invalid address
> + * limit. This function does not return.
> + */
> +asmlinkage void addr_limit_check_failed(void)
> +{
> +	/* Try to fail on the generic address limit check */
> +	addr_limit_check_syscall();
> +	panic("Invalid address limit before returning to user-mode");
> +}
> +#endif

Ok, this version looks pretty good to me. Could you (re-)send a full series?

I assume some of these changes need to be propagated into the followup patches but 
even if not it's better to pick up a clean series.

Thanks,

	Ingo

[toc] | [prev] | [next] | [standalone]


#1632558 — Re: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode

FromIngo Molnar <mingo@kernel.org>
Date2017-04-28 08:40 +0200
SubjectRe: [PATCH v8 1/4] syscalls: Verify address limit before returning to user-mode
Message-ID<tB7Jn-3F8-13@gated-at.bofh.it>
In reply to#1632124
* Thomas Garnier <thgarnie@google.com> wrote:

> > BTW., a further simplification would be:
> >
> > #ifndef ADDR_LIMIT_CHECK_PRE
> > # define ADDR_LIMIT_CHECK_PRE ...
> > #endif
> >
> > This way architectures could override this generic functionality simply by
> > defining the helpers. Architectures that don't do that get the generic version.
> 
> I don't think architectures need to do that. The optimizations are
> embedding the checks on their architecture-specific code to make it
> faster and remove the size impact. The pre/post is fine for the rest.

Indeed, only the generic code needs to turn off that code - architectures will 
place these callbacks elsewhere.

Thanks,

	Ingo

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web