[v3] ubsan: Reintroduce signed overflow sanitizer

Message ID 20240205093725.make.582-kees@kernel.org
State New
Headers
Series [v3] ubsan: Reintroduce signed overflow sanitizer |

Commit Message

Kees Cook Feb. 5, 2024, 9:37 a.m. UTC
  In order to mitigate unexpected signed wrap-around[1], bring back the
signed integer overflow sanitizer. It was removed in commit 6aaa31aeb9cf
("ubsan: remove overflow checks") because it was effectively a no-op
when combined with -fno-strict-overflow (which correctly changes signed
overflow from being "undefined" to being explicitly "wrap around").

Compilers are adjusting their sanitizers to trap wrap-around and to
detecting common code patterns that should not be instrumented
(e.g. "var + offset < var"). Prepare for this and explicitly rename
the option from "OVERFLOW" to "WRAP".

To annotate intentional wrap-around arithmetic, the add/sub/mul_wrap()
helpers can be used for individual statements. At the function level,
the __signed_wrap attribute can be used to mark an entire function as
expecting its signed arithmetic to wrap around. For a single object file
the Makefile can use "UBSAN_WRAP_SIGNED_target.o := n" to mark it as
wrapping, and for an entire directory, "UBSAN_WRAP_SIGNED := n" can be
used.

Additionally keep these disabled under CONFIG_COMPILE_TEST for now.

Link: https://github.com/KSPP/linux/issues/26 [1]
Cc: Justin Stitt <justinstitt@google.com>
Cc: Marco Elver <elver@google.com>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Hao Luo <haoluo@google.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
---
v3:
 - split out signed overflow sanitizer so we can do each separately
v2: https://lore.kernel.org/all/20240202101311.it.893-kees@kernel.org/
v1: https://lore.kernel.org/all/20240129175033.work.813-kees@kernel.org/
---
 include/linux/compiler_types.h |  9 ++++-
 lib/Kconfig.ubsan              | 14 +++++++
 lib/test_ubsan.c               | 37 ++++++++++++++++++
 lib/ubsan.c                    | 68 ++++++++++++++++++++++++++++++++++
 lib/ubsan.h                    |  4 ++
 scripts/Makefile.lib           |  3 ++
 scripts/Makefile.ubsan         |  3 ++
 7 files changed, 137 insertions(+), 1 deletion(-)
  

Comments

Kees Cook Feb. 5, 2024, 12:51 p.m. UTC | #1
On Mon, Feb 05, 2024 at 12:29:21PM +0100, Marco Elver wrote:
> On Mon, 5 Feb 2024 at 10:37, Kees Cook <keescook@chromium.org> wrote:
> >
> > In order to mitigate unexpected signed wrap-around[1], bring back the
> > signed integer overflow sanitizer. It was removed in commit 6aaa31aeb9cf
> > ("ubsan: remove overflow checks") because it was effectively a no-op
> > when combined with -fno-strict-overflow (which correctly changes signed
> > overflow from being "undefined" to being explicitly "wrap around").
> >
> > Compilers are adjusting their sanitizers to trap wrap-around and to
> > detecting common code patterns that should not be instrumented
> > (e.g. "var + offset < var"). Prepare for this and explicitly rename
> > the option from "OVERFLOW" to "WRAP".
> >
> > To annotate intentional wrap-around arithmetic, the add/sub/mul_wrap()
> > helpers can be used for individual statements. At the function level,
> > the __signed_wrap attribute can be used to mark an entire function as
> > expecting its signed arithmetic to wrap around. For a single object file
> > the Makefile can use "UBSAN_WRAP_SIGNED_target.o := n" to mark it as
> > wrapping, and for an entire directory, "UBSAN_WRAP_SIGNED := n" can be
> > used.
> >
> > Additionally keep these disabled under CONFIG_COMPILE_TEST for now.
> >
> > Link: https://github.com/KSPP/linux/issues/26 [1]
> > Cc: Justin Stitt <justinstitt@google.com>
> > Cc: Marco Elver <elver@google.com>
> > Cc: Miguel Ojeda <ojeda@kernel.org>
> > Cc: Nathan Chancellor <nathan@kernel.org>
> > Cc: Peter Zijlstra <peterz@infradead.org>
> > Cc: Hao Luo <haoluo@google.com>
> > Signed-off-by: Kees Cook <keescook@chromium.org>
> 
> Looks good.
> 
> Reviewed-by: Marco Elver <elver@google.com>

Thanks!

> And just to double check, you don't think we need 'depends on EXPERT'
> (or DEBUG_KERNEL) to keep the noise down initially?

Not for signed, no. It's almost a no-op like this. Once Clang and GCC
support the wrap version (which will likely require changing the
command line argument), we can re-evaluate. So far in my testing, I've
not been able to trip it. I'm planning to get a local syzbot running
with the wrap sanitizer later this week to see how noisy it gets (if at
all).
  
Kees Cook Feb. 5, 2024, 12:59 p.m. UTC | #2
On Mon, Feb 05, 2024 at 01:54:24PM +0100, Andrey Ryabinin wrote:
> 
> 
> On 2/5/24 10:37, Kees Cook wrote:
> 
> > ---
> >  include/linux/compiler_types.h |  9 ++++-
> >  lib/Kconfig.ubsan              | 14 +++++++
> >  lib/test_ubsan.c               | 37 ++++++++++++++++++
> >  lib/ubsan.c                    | 68 ++++++++++++++++++++++++++++++++++
> >  lib/ubsan.h                    |  4 ++
> >  scripts/Makefile.lib           |  3 ++
> >  scripts/Makefile.ubsan         |  3 ++
> >  7 files changed, 137 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
> > index 6f1ca49306d2..ee9d272008a5 100644
> > --- a/include/linux/compiler_types.h
> > +++ b/include/linux/compiler_types.h
> > @@ -282,11 +282,18 @@ struct ftrace_likely_data {
> >  #define __no_sanitize_or_inline __always_inline
> >  #endif
> >  
> > +/* Do not trap wrapping arithmetic within an annotated function. */
> > +#ifdef CONFIG_UBSAN_SIGNED_WRAP
> > +# define __signed_wrap __attribute__((no_sanitize("signed-integer-overflow")))
> > +#else
> > +# define __signed_wrap
> > +#endif
> > +
> >  /* Section for code which can't be instrumented at all */
> >  #define __noinstr_section(section)					\
> >  	noinline notrace __attribute((__section__(section)))		\
> >  	__no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage \
> > -	__no_sanitize_memory
> > +	__no_sanitize_memory __signed_wrap
> >  
> 
> Given this disables all kinds of code instrumentations,
> shouldn't we just add __no_sanitize_undefined here?

Yeah, that's a very good point.

> I suspect that ubsan's instrumentation usually doesn't cause problems
> because it calls __ubsan_* functions with all heavy stuff (printk, locks etc)
> only if code has an UB. So the answer to the question above depends on
> whether we want to ignore UBs in "noinstr" code or to get some weird side effect,
> possibly without proper UBSAN report in dmesg.

I think my preference would be to fail safe (i.e. leave in the
instrumentation), but the intent of noinstr is pretty clear. :P I wonder
if, instead, we could adjust objtool to yell about cases where calls are
made in noinstr functions (like it does for UACCESS)... maybe it already
does?

-Kees
  
Kees Cook Feb. 6, 2024, 11:09 a.m. UTC | #3
On Mon, Feb 05, 2024 at 02:10:26PM +0100, Marco Elver wrote:
> On Mon, 5 Feb 2024 at 13:59, Kees Cook <keescook@chromium.org> wrote:
> >
> > On Mon, Feb 05, 2024 at 01:54:24PM +0100, Andrey Ryabinin wrote:
> > >
> > >
> > > On 2/5/24 10:37, Kees Cook wrote:
> > >
> > > > ---
> > > >  include/linux/compiler_types.h |  9 ++++-
> > > >  lib/Kconfig.ubsan              | 14 +++++++
> > > >  lib/test_ubsan.c               | 37 ++++++++++++++++++
> > > >  lib/ubsan.c                    | 68 ++++++++++++++++++++++++++++++++++
> > > >  lib/ubsan.h                    |  4 ++
> > > >  scripts/Makefile.lib           |  3 ++
> > > >  scripts/Makefile.ubsan         |  3 ++
> > > >  7 files changed, 137 insertions(+), 1 deletion(-)
> > > >
> > > > diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
> > > > index 6f1ca49306d2..ee9d272008a5 100644
> > > > --- a/include/linux/compiler_types.h
> > > > +++ b/include/linux/compiler_types.h
> > > > @@ -282,11 +282,18 @@ struct ftrace_likely_data {
> > > >  #define __no_sanitize_or_inline __always_inline
> > > >  #endif
> > > >
> > > > +/* Do not trap wrapping arithmetic within an annotated function. */
> > > > +#ifdef CONFIG_UBSAN_SIGNED_WRAP
> > > > +# define __signed_wrap __attribute__((no_sanitize("signed-integer-overflow")))
> > > > +#else
> > > > +# define __signed_wrap
> > > > +#endif
> > > > +
> > > >  /* Section for code which can't be instrumented at all */
> > > >  #define __noinstr_section(section)                                 \
> > > >     noinline notrace __attribute((__section__(section)))            \
> > > >     __no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage \
> > > > -   __no_sanitize_memory
> > > > +   __no_sanitize_memory __signed_wrap
> > > >
> > >
> > > Given this disables all kinds of code instrumentations,
> > > shouldn't we just add __no_sanitize_undefined here?
> >
> > Yeah, that's a very good point.
> >
> > > I suspect that ubsan's instrumentation usually doesn't cause problems
> > > because it calls __ubsan_* functions with all heavy stuff (printk, locks etc)
> > > only if code has an UB. So the answer to the question above depends on
> > > whether we want to ignore UBs in "noinstr" code or to get some weird side effect,
> > > possibly without proper UBSAN report in dmesg.
> >
> > I think my preference would be to fail safe (i.e. leave in the
> > instrumentation), but the intent of noinstr is pretty clear. :P I wonder
> > if, instead, we could adjust objtool to yell about cases where calls are
> > made in noinstr functions (like it does for UACCESS)... maybe it already
> > does?
> 
> It already does, see CONFIG_NOINSTR_VALIDATION (yes by default on x86).

This is actually a reason to not include the ubsan disabling in
__noinstr_section just to see what ends up in there so we can fix it
immediately....
  
Kees Cook Feb. 7, 2024, 11:04 a.m. UTC | #4
On Wed, Feb 07, 2024 at 01:45:28AM +0000, Justin Stitt wrote:
> I wouldn't mind also seeing a test_ubsan_div_overflow test case here.
> 
> It has some quirky behavior and it'd be nice to test that the sanitizers
> properly capture it.
> 
> Check out this Godbolt: https://godbolt.org/z/qG5f1j6n1
> 
> tl;dr: with -fsanitize=signed-integer-overflow division (/) and
> remainder (%) operators still instrument arithmetic even with
> -fno-strict-overflow on.
> 
> This makes sense as division by 0 and INT_MIN/-1 are UBs that are not
> influenced by -fno-strict-overflow.

There is actually already a test_ubsan_divrem_overflow, but because the
failure modes result in a trap even without the sanitizer, it's disabled
in the test. For testing a crashing mode, it might be interesting to add
it to LKDTM, which is the crash tester...

> 
> Really though, the patch is fine and the above test case is optional and
> can be shipped later -- as such:
> 
> Reviewed-by: Justin Stitt <justinstitt@google.com>

Thanks!

-Kees
  

Patch

diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
index 6f1ca49306d2..ee9d272008a5 100644
--- a/include/linux/compiler_types.h
+++ b/include/linux/compiler_types.h
@@ -282,11 +282,18 @@  struct ftrace_likely_data {
 #define __no_sanitize_or_inline __always_inline
 #endif
 
+/* Do not trap wrapping arithmetic within an annotated function. */
+#ifdef CONFIG_UBSAN_SIGNED_WRAP
+# define __signed_wrap __attribute__((no_sanitize("signed-integer-overflow")))
+#else
+# define __signed_wrap
+#endif
+
 /* Section for code which can't be instrumented at all */
 #define __noinstr_section(section)					\
 	noinline notrace __attribute((__section__(section)))		\
 	__no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage \
-	__no_sanitize_memory
+	__no_sanitize_memory __signed_wrap
 
 #define noinstr __noinstr_section(".noinstr.text")
 
diff --git a/lib/Kconfig.ubsan b/lib/Kconfig.ubsan
index 56d7653f4941..129e9bc21877 100644
--- a/lib/Kconfig.ubsan
+++ b/lib/Kconfig.ubsan
@@ -116,6 +116,20 @@  config UBSAN_UNREACHABLE
 	  This option enables -fsanitize=unreachable which checks for control
 	  flow reaching an expected-to-be-unreachable position.
 
+config UBSAN_SIGNED_WRAP
+	bool "Perform checking for signed arithmetic wrap-around"
+	default UBSAN
+	depends on !COMPILE_TEST
+	depends on $(cc-option,-fsanitize=signed-integer-overflow)
+	help
+	  This option enables -fsanitize=signed-integer-overflow which checks
+	  for wrap-around of any arithmetic operations with signed integers.
+	  This currently performs nearly no instrumentation due to the
+	  kernel's use of -fno-strict-overflow which converts all would-be
+	  arithmetic undefined behavior into wrap-around arithmetic. Future
+	  sanitizer versions will allow for wrap-around checking (rather than
+	  exclusively undefined behavior).
+
 config UBSAN_BOOL
 	bool "Perform checking for non-boolean values used as boolean"
 	default UBSAN
diff --git a/lib/test_ubsan.c b/lib/test_ubsan.c
index f4ee2484d4b5..276c12140ee2 100644
--- a/lib/test_ubsan.c
+++ b/lib/test_ubsan.c
@@ -11,6 +11,39 @@  typedef void(*test_ubsan_fp)(void);
 			#config, IS_ENABLED(config) ? "y" : "n");	\
 	} while (0)
 
+static void test_ubsan_add_overflow(void)
+{
+	volatile int val = INT_MAX;
+
+	UBSAN_TEST(CONFIG_UBSAN_SIGNED_WRAP);
+	val += 2;
+}
+
+static void test_ubsan_sub_overflow(void)
+{
+	volatile int val = INT_MIN;
+	volatile int val2 = 2;
+
+	UBSAN_TEST(CONFIG_UBSAN_SIGNED_WRAP);
+	val -= val2;
+}
+
+static void test_ubsan_mul_overflow(void)
+{
+	volatile int val = INT_MAX / 2;
+
+	UBSAN_TEST(CONFIG_UBSAN_SIGNED_WRAP);
+	val *= 3;
+}
+
+static void test_ubsan_negate_overflow(void)
+{
+	volatile int val = INT_MIN;
+
+	UBSAN_TEST(CONFIG_UBSAN_SIGNED_WRAP);
+	val = -val;
+}
+
 static void test_ubsan_divrem_overflow(void)
 {
 	volatile int val = 16;
@@ -90,6 +123,10 @@  static void test_ubsan_misaligned_access(void)
 }
 
 static const test_ubsan_fp test_ubsan_array[] = {
+	test_ubsan_add_overflow,
+	test_ubsan_sub_overflow,
+	test_ubsan_mul_overflow,
+	test_ubsan_negate_overflow,
 	test_ubsan_shift_out_of_bounds,
 	test_ubsan_out_of_bounds,
 	test_ubsan_load_invalid_value,
diff --git a/lib/ubsan.c b/lib/ubsan.c
index df4f8d1354bb..5fc107f61934 100644
--- a/lib/ubsan.c
+++ b/lib/ubsan.c
@@ -222,6 +222,74 @@  static void ubsan_epilogue(void)
 	check_panic_on_warn("UBSAN");
 }
 
+static void handle_overflow(struct overflow_data *data, void *lhs,
+			void *rhs, char op)
+{
+
+	struct type_descriptor *type = data->type;
+	char lhs_val_str[VALUE_LENGTH];
+	char rhs_val_str[VALUE_LENGTH];
+
+	if (suppress_report(&data->location))
+		return;
+
+	ubsan_prologue(&data->location, type_is_signed(type) ?
+			"signed-integer-overflow" :
+			"unsigned-integer-overflow");
+
+	val_to_string(lhs_val_str, sizeof(lhs_val_str), type, lhs);
+	val_to_string(rhs_val_str, sizeof(rhs_val_str), type, rhs);
+	pr_err("%s %c %s cannot be represented in type %s\n",
+		lhs_val_str,
+		op,
+		rhs_val_str,
+		type->type_name);
+
+	ubsan_epilogue();
+}
+
+void __ubsan_handle_add_overflow(void *data,
+				void *lhs, void *rhs)
+{
+
+	handle_overflow(data, lhs, rhs, '+');
+}
+EXPORT_SYMBOL(__ubsan_handle_add_overflow);
+
+void __ubsan_handle_sub_overflow(void *data,
+				void *lhs, void *rhs)
+{
+	handle_overflow(data, lhs, rhs, '-');
+}
+EXPORT_SYMBOL(__ubsan_handle_sub_overflow);
+
+void __ubsan_handle_mul_overflow(void *data,
+				void *lhs, void *rhs)
+{
+	handle_overflow(data, lhs, rhs, '*');
+}
+EXPORT_SYMBOL(__ubsan_handle_mul_overflow);
+
+void __ubsan_handle_negate_overflow(void *_data, void *old_val)
+{
+	struct overflow_data *data = _data;
+	char old_val_str[VALUE_LENGTH];
+
+	if (suppress_report(&data->location))
+		return;
+
+	ubsan_prologue(&data->location, "negation-overflow");
+
+	val_to_string(old_val_str, sizeof(old_val_str), data->type, old_val);
+
+	pr_err("negation of %s cannot be represented in type %s:\n",
+		old_val_str, data->type->type_name);
+
+	ubsan_epilogue();
+}
+EXPORT_SYMBOL(__ubsan_handle_negate_overflow);
+
+
 void __ubsan_handle_divrem_overflow(void *_data, void *lhs, void *rhs)
 {
 	struct overflow_data *data = _data;
diff --git a/lib/ubsan.h b/lib/ubsan.h
index 5d99ab81913b..0abbbac8700d 100644
--- a/lib/ubsan.h
+++ b/lib/ubsan.h
@@ -124,6 +124,10 @@  typedef s64 s_max;
 typedef u64 u_max;
 #endif
 
+void __ubsan_handle_add_overflow(void *data, void *lhs, void *rhs);
+void __ubsan_handle_sub_overflow(void *data, void *lhs, void *rhs);
+void __ubsan_handle_mul_overflow(void *data, void *lhs, void *rhs);
+void __ubsan_handle_negate_overflow(void *_data, void *old_val);
 void __ubsan_handle_divrem_overflow(void *_data, void *lhs, void *rhs);
 void __ubsan_handle_type_mismatch(struct type_mismatch_data *data, void *ptr);
 void __ubsan_handle_type_mismatch_v1(void *_data, void *ptr);
diff --git a/scripts/Makefile.lib b/scripts/Makefile.lib
index 52efc520ae4f..7ce8ecccc65a 100644
--- a/scripts/Makefile.lib
+++ b/scripts/Makefile.lib
@@ -177,6 +177,9 @@  ifeq ($(CONFIG_UBSAN),y)
 _c_flags += $(if $(patsubst n%,, \
 		$(UBSAN_SANITIZE_$(basetarget).o)$(UBSAN_SANITIZE)y), \
 		$(CFLAGS_UBSAN))
+_c_flags += $(if $(patsubst n%,, \
+		$(UBSAN_WRAP_SIGNED_$(basetarget).o)$(UBSAN_SANITIZE_$(basetarget).o)$(UBSAN_WRAP_SIGNED)$(UBSAN_SANITIZE)y), \
+		$(CFLAGS_UBSAN_WRAP_SIGNED))
 endif
 
 ifeq ($(CONFIG_KCOV),y)
diff --git a/scripts/Makefile.ubsan b/scripts/Makefile.ubsan
index 7cf42231042b..bc957add0b4d 100644
--- a/scripts/Makefile.ubsan
+++ b/scripts/Makefile.ubsan
@@ -13,3 +13,6 @@  ubsan-cflags-$(CONFIG_UBSAN_ENUM)		+= -fsanitize=enum
 ubsan-cflags-$(CONFIG_UBSAN_TRAP)		+= $(call cc-option,-fsanitize-trap=undefined,-fsanitize-undefined-trap-on-error)
 
 export CFLAGS_UBSAN := $(ubsan-cflags-y)
+
+ubsan-wrap-signed-cflags-$(CONFIG_UBSAN_SIGNED_WRAP)     += -fsanitize=signed-integer-overflow
+export CFLAGS_UBSAN_WRAP_SIGNED := $(ubsan-wrap-signed-cflags-y)