Skip to contentSkip to conversation

AARCH64: Add gcc Shadow Call Stack support

PATCH Superseded Dan Li 6 participants 12 messages Feb 22, 2022 – Feb 25, 2022 arm64

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Dan Li

author
#0 lore raw

PATCH 1/1 AARCH64: Add gcc Shadow Call Stack support

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.

Note that the implementation in GCC is slightly different from Clang. With SCS enabled, functions will only pop x30 once in the epilogue, like:

str x30, [x18], #8
stp x29, x30, [sp, #-16]!
......

-  ldp     x29, x30, [sp], #16	  //clang
+  ldr     x29, [sp], #16	  //GCC
   ldr     x30, [x18, #-8]!

---
FYI:
This function can be used to test if the shadow call stack works:
//noinline void __noscs scs_test(void)
noinline void scs_test(void)

{
    register unsigned long *sp asm("sp");
    unsigned long * lr = sp + 1;

    asm volatile("":::"x30");
    *lr = 0;
}
ffff800008012704:       d503233f        paciasp
ffff800008012708:       f800865e        str     x30, [x18], #8
ffff80000801270c:       a9bf7bfd        stp     x29, x30, [sp, #-16]!
ffff800008012710:       910003fd        mov     x29, sp
ffff800008012714:       910003e0        mov     x0, sp
ffff800008012718:       f900041f        str     xzr, [x0, #8]
ffff80000801271c:       f85f8e5e        ldr     x30, [x18, #-8]!
ffff800008012720:       f84107fd        ldr     x29, [sp], #16
ffff800008012724:       d50323bf        autiasp
ffff800008012728:       d65f03c0        ret

If SCS protection is enabled, this function will return normally. If the function has __noscs attribute (scs disabled), it will crash due to 0 address access.

 arch/Kconfig                 | 6 +++---
 arch/arm64/Kconfig           | 2 +-
 include/linux/compiler-gcc.h | 4 ++++
 3 files changed, 8 insertions(+), 4 deletions(-)
arch/Kconfig +3 -3
index 678a80713b21..35db7b72bdb0 100644
@@ -604,11 +604,11 @@ config ARCH_SUPPORTS_SHADOW_CALL_STACK
 	  switching.
 
 config SHADOW_CALL_STACK
-	bool "Clang Shadow Call Stack"
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
 	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
 	help
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC'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 in
 	  Clang's documentation:
arch/arm64/Kconfig +1 -1
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
 config ARCH_HAS_FILTER_PGPROT
 	def_bool y
 
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0
 config CC_HAVE_SHADOW_CALL_STACK
 	def_bool $(cc-option, -fsanitize=shadow-call-stack -ffixed-x18)
 
include/linux/compiler-gcc.h +4 -0
index ccbbd31b3aae..deff5b308470 100644
@@ -97,6 +97,10 @@
 #define KASAN_ABI_VERSION 4
 #endif
 
+#ifdef CONFIG_SHADOW_CALL_STACK
+#define __noscs __attribute__((__no_sanitize__("shadow-call-stack")))
+#endif
+
 #if __has_attribute(__no_sanitize_address__)
 #define __no_sanitize_address __attribute__((no_sanitize_address))
 #else
Signature · 1 line
2.17.1
Reviewed-by on v1 1/1
#1 lore raw
Quoted 15 lines · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.

Note that the implementation in GCC is slightly different from Clang. With SCS enabled, functions will only pop x30 once in the epilogue, like:

   str     x30, [x18], #8
   stp     x29, x30, [sp, #-16]!
   ......
(quoted diff) +1 -1
-  ldp     x29, x30, [sp], #16	  //clang
+  ldr     x29, [sp], #16	  //GCC
   ldr     x30, [x18, #-8]!
Quoted 3 lines

A few open-ended comments below.

Quoted 39 lines

---
FYI:
This function can be used to test if the shadow call stack works:
//noinline void __noscs scs_test(void)
noinline void scs_test(void)

{
    register unsigned long *sp asm("sp");
    unsigned long * lr = sp + 1;

    asm volatile("":::"x30");
    *lr = 0;
}
ffff800008012704:       d503233f        paciasp
ffff800008012708:       f800865e        str     x30, [x18], #8
ffff80000801270c:       a9bf7bfd        stp     x29, x30, [sp, #-16]!
ffff800008012710:       910003fd        mov     x29, sp
ffff800008012714:       910003e0        mov     x0, sp
ffff800008012718:       f900041f        str     xzr, [x0, #8]
ffff80000801271c:       f85f8e5e        ldr     x30, [x18, #-8]!
ffff800008012720:       f84107fd        ldr     x29, [sp], #16
ffff800008012724:       d50323bf        autiasp
ffff800008012728:       d65f03c0        ret

If SCS protection is enabled, this function will return normally. If the function has __noscs attribute (scs disabled), it will crash due to 0 address access.

 arch/Kconfig                 | 6 +++---
 arch/arm64/Kconfig           | 2 +-
 include/linux/compiler-gcc.h | 4 ++++
 3 files changed, 8 insertions(+), 4 deletions(-)
arch/Kconfig +0 -0
index 678a80713b21..35db7b72bdb0 100644
@@ -604,11 +604,11 @@ config ARCH_SUPPORTS_SHADOW_CALL_STACK
 	  switching.

arch/Kconfig config ARCH_SUPPORTS_SHADOW_CALL_STACK · line 611

Quoted 10 lines
arch/Kconfig +3 -3
 
 config SHADOW_CALL_STACK
-	bool "Clang Shadow Call Stack"
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
 	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
 	help
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC's Shadow Call Stack, which uses a

I wonder if we want to just ditch the mention of the compiler if both support it?

arch/arm64/Kconfig config HW_PERF_EVENTS · line 1258

Quoted 13 lines
arch/Kconfig +0 -0
 	  shadow stack to protect function return addresses from being
 	  overwritten by an attacker. More information can be found in
 	  Clang's documentation:
arch/arm64/Kconfig +1 -1
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
 config ARCH_HAS_FILTER_PGPROT
 	def_bool y
 
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0

Same thing here, although eventually there may be a minimum GCC version bump to something newer than 11.2.0, which would allow us to just drop CONFIG_CC_HAVE_SHADOW_CALL_STACK altogether. No strong opinion.

Quoted 21 lines
arch/arm64/Kconfig +0 -0
 config CC_HAVE_SHADOW_CALL_STACK
 	def_bool $(cc-option, -fsanitize=shadow-call-stack -ffixed-x18)
 
include/linux/compiler-gcc.h +4 -0
index ccbbd31b3aae..deff5b308470 100644
@@ -97,6 +97,10 @@
 #define KASAN_ABI_VERSION 4
 #endif
 
+#ifdef CONFIG_SHADOW_CALL_STACK
+#define __noscs __attribute__((__no_sanitize__("shadow-call-stack")))
+#endif
+
 #if __has_attribute(__no_sanitize_address__)
 #define __no_sanitize_address __attribute__((no_sanitize_address))
 #else
Signature · 1 line
2.17.1
#2 lore raw
Quoted 2 lines · Nathan Chancellor
Quoted 1 line · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes

The above suggests that the option will be available with gcc 11.3.0. Information available in public suggests that it will be introduced with gcc 12.0.

Quoted 86 lines
Quoted 14 lines

the corresponding kernel configuration available when compiling the kernel with gcc.

Note that the implementation in GCC is slightly different from Clang. With SCS enabled, functions will only pop x30 once in the epilogue, like:

    str     x30, [x18], #8
    stp     x29, x30, [sp, #-16]!
    ......
(quoted diff) +1 -1
-  ldp     x29, x30, [sp], #16	  //clang
+  ldr     x29, [sp], #16	  //GCC
    ldr     x30, [x18, #-8]!
Quoted 3 lines

A few open-ended comments below.

Quoted 39 lines

---
FYI:
This function can be used to test if the shadow call stack works:
//noinline void __noscs scs_test(void)
noinline void scs_test(void)

{
     register unsigned long *sp asm("sp");
     unsigned long * lr = sp + 1;

     asm volatile("":::"x30");
     *lr = 0;
}
ffff800008012704:       d503233f        paciasp
ffff800008012708:       f800865e        str     x30, [x18], #8
ffff80000801270c:       a9bf7bfd        stp     x29, x30, [sp, #-16]!
ffff800008012710:       910003fd        mov     x29, sp
ffff800008012714:       910003e0        mov     x0, sp
ffff800008012718:       f900041f        str     xzr, [x0, #8]
ffff80000801271c:       f85f8e5e        ldr     x30, [x18, #-8]!
ffff800008012720:       f84107fd        ldr     x29, [sp], #16
ffff800008012724:       d50323bf        autiasp
ffff800008012728:       d65f03c0        ret

If SCS protection is enabled, this function will return normally. If the function has __noscs attribute (scs disabled), it will crash due to 0 address access.

  arch/Kconfig                 | 6 +++---
  arch/arm64/Kconfig           | 2 +-
  include/linux/compiler-gcc.h | 4 ++++
  3 files changed, 8 insertions(+), 4 deletions(-)
arch/Kconfig +0 -0
index 678a80713b21..35db7b72bdb0 100644
@@ -604,11 +604,11 @@ config ARCH_SUPPORTS_SHADOW_CALL_STACK
  	  switching.

arch/Kconfig config ARCH_SUPPORTS_SHADOW_CALL_STACK · line 611

Quoted 10 lines
arch/Kconfig +3 -3
 
  config SHADOW_CALL_STACK
-	bool "Clang Shadow Call Stack"
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
  	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
  	help
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC's Shadow Call Stack, which uses a

I wonder if we want to just ditch the mention of the compiler if both support it?

Quoted 11 lines
arch/Kconfig +0 -0
  	  shadow stack to protect function return addresses from being
  	  overwritten by an attacker. More information can be found in
  	  Clang's documentation:
arch/arm64/Kconfig +0 -0
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
  config ARCH_HAS_FILTER_PGPROT
  	def_bool y
 
Quoted 4 lines

arch/arm64/Kconfig config HW_PERF_EVENTS · line 1258

Quoted 2 lines
arch/arm64/Kconfig +1 -1
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0

Same thing here, although eventually there may be a minimum GCC version

The point here, I think, is to list the minimum gcc version. It is going to be a long time until gcc 12.0 is the minimum version, so I think it makes sense to list the minimum version number for each compiler here.

However, it may make sense to add some reference indicating that support will indeed be added with gcc 11.3.0, and not only starting with gcc 12.0 (and maybe wait with applying this patch until it is actually available in gcc and can be confirmed to work as intended).

Thanks,
Guenter

Quoted 24 lines

bump to something newer than 11.2.0, which would allow us to just drop CONFIG_CC_HAVE_SHADOW_CALL_STACK altogether. No strong opinion.

Quoted 21 lines
arch/arm64/Kconfig +0 -0
  config CC_HAVE_SHADOW_CALL_STACK
  	def_bool $(cc-option, -fsanitize=shadow-call-stack -ffixed-x18)
 
include/linux/compiler-gcc.h +4 -0
index ccbbd31b3aae..deff5b308470 100644
@@ -97,6 +97,10 @@
  #define KASAN_ABI_VERSION 4
  #endif
 
+#ifdef CONFIG_SHADOW_CALL_STACK
+#define __noscs __attribute__((__no_sanitize__("shadow-call-stack")))
+#endif
+
  #if __has_attribute(__no_sanitize_address__)
  #define __no_sanitize_address __attribute__((no_sanitize_address))
  #else
Signature · 1 line
2.17.1
#3 lore raw
Quoted 5 lines · Guenter Roeck

The point here, I think, is to list the minimum gcc version. It is going to be a long time until gcc 12.0 is the minimum version, so I think it makes sense to list the minimum version number for each compiler here.

Yeah, please list the minimum version (ideally >= instead of >, with the actual version where it first appeared). It makes it easy to then remove things that are not needed anymore when the minimum GCC version is upgraded.

Cheers,
Miguel

author
#4 lore raw
Quoted 7 lines · Miguel Ojeda
Quoted 5 lines · Guenter Roeck

The point here, I think, is to list the minimum gcc version. It is going to be a long time until gcc 12.0 is the minimum version, so I think it makes sense to list the minimum version number for each compiler here.

Quoted 5 lines

Yeah, please list the minimum version (ideally >= instead of >, with the actual version where it first appeared). It makes it easy to then remove things that are not needed anymore when the minimum GCC version is upgraded.

Got it, I'll use "gcc >= 12.0.0" in the next patch.

Thanks,
Dan.

author
#5 lore raw
In reply to Guenter Roeck
Quoted 8 lines · Guenter Roeck
Quoted 2 lines · Nathan Chancellor
Quoted 1 line · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes

The above suggests that the option will be available with gcc 11.3.0. Information available in public suggests that it will be introduced with gcc 12.0.

Ah, yes, I think we could use "gcc >= 12.0.0" here.

Quoted 7 lines

The point here, I think, is to list the minimum gcc version. It is going to be a long time until gcc 12.0 is the minimum version, so I think it makes sense to list the minimum version number for each compiler here.

However, it may make sense to add some reference indicating that support will indeed be added with gcc 11.3.0, and not only starting

I took a quick look at the gcc description, and it seems like the y in x.y.z is usually used to fix bugs, and new features should be added directly to the trunk.

Quoted 3 lines

with gcc 12.0 (and maybe wait with applying this patch until it is actually available in gcc and can be confirmed to work as intended).

It's also fine to wait for gcc 12 to be released, and I thought maybe I could submit the "final" version of this patch to the community (or mailing list) first so maybe more people would test it and if there were any issues, it could be fixed before GCC 12 is released :)

Thanks,
Dan.

author
#6 lore raw
In reply to Nathan Chancellor
Quoted 17 lines · Nathan Chancellor

arch/Kconfig config ARCH_SUPPORTS_SHADOW_CALL_STACK · line 611

Quoted 12 lines · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.
config SHADOW_CALL_STACK

arch/Kconfig +3 -3
-	bool "Clang Shadow Call Stack"
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
  	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
  	help
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC's Shadow Call Stack, which uses a

I wonder if we want to just ditch the mention of the compiler if both support it?

My intention is to remind users that this is a compiler feature. But since there is also a hint in CC_HAVE_SHADOW_CALL_STACK:
+# Supported by clang >= 7.0 or GCC ...

Removing the specific compiler here also looks fine to me. Would this look better?

"This option enables Shadow Call Stack, which uses a ..."

or maybe:

"This option enables compiler's Shadow Call Stack, which uses a ..."

Quoted 11 lines
Quoted 11 lines

shadow stack to protect function return addresses from being overwritten by an attacker. More information can be found in Clang's documentation:

arch/arm64/Kconfig +0 -0
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
  config ARCH_HAS_FILTER_PGPROT
  	def_bool y
 
Quoted 7 lines

arch/arm64/Kconfig config HW_PERF_EVENTS · line 1258

Quoted 2 lines
arch/arm64/Kconfig +1 -1
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0

Same thing here, although eventually there may be a minimum GCC version bump to something newer than 11.2.0, which would allow us to just drop CONFIG_CC_HAVE_SHADOW_CALL_STACK altogether. No strong opinion.

As Guenter said, I thought maybe we could mark the minimum available version for users :)

Thanks,
Dan.

#7 lore raw
Quoted 25 lines · Dan Li
Quoted 17 lines · Nathan Chancellor

arch/Kconfig config ARCH_SUPPORTS_SHADOW_CALL_STACK · line 611

Quoted 12 lines · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.
config SHADOW_CALL_STACK

arch/Kconfig +3 -3
-	bool "Clang Shadow Call Stack"
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
  	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
  	help
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC's Shadow Call Stack, which uses a

I wonder if we want to just ditch the mention of the compiler if both support it?

My intention is to remind users that this is a compiler feature. But since there is also a hint in CC_HAVE_SHADOW_CALL_STACK:
+# Supported by clang >= 7.0 or GCC ...

Quoted 8 lines

Removing the specific compiler here also looks fine to me. Would this look better?

"This option enables Shadow Call Stack, which uses a ..."

or maybe:

"This option enables compiler's Shadow Call Stack, which uses a ..."

I do not honestly have a strong opinion around removing mention of the compiler so either looks fine to me (might be better to say "the compiler's Shadow ..." in the second one).

Quoted 18 lines
Quoted 8 lines
Quoted 8 lines

shadow stack to protect function return addresses from being overwritten by an attacker. More information can be found in Clang's documentation:

arch/arm64/Kconfig +0 -0
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
Quoted 9 lines

arch/arm64/Kconfig config HW_PERF_EVENTS · line 1258

Quoted 4 lines
arch/arm64/Kconfig +1 -1
  config ARCH_HAS_FILTER_PGPROT
  	def_bool y
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0

Same thing here, although eventually there may be a minimum GCC version bump to something newer than 11.2.0, which would allow us to just drop CONFIG_CC_HAVE_SHADOW_CALL_STACK altogether. No strong opinion.

Quoted 2 lines

As Guenter said, I thought maybe we could mark the minimum available version for users :)

Yes, that is what I was getting at with the "minimum version" comment. It should remain around.

Cheers,
Nathan

author
#8 lore raw
Quoted 10 lines · Nathan Chancellor
Quoted 9 lines · Dan Li

My intention is to remind users that this is a compiler feature. But since there is also a hint in CC_HAVE_SHADOW_CALL_STACK:
+# Supported by clang >= 7.0 or GCC ...

Removing the specific compiler here also looks fine to me. Would this look better?

"This option enables Shadow Call Stack, which uses a ..."

Quoted 8 lines
Quoted 3 lines

or maybe:

"This option enables compiler's Shadow Call Stack, which uses a ..."

I do not honestly have a strong opinion around removing mention of the compiler so either looks fine to me (might be better to say "the compiler's Shadow ..." in the second one).

Ah, yes, thanks :)

Dan.

reviewer
#9 lore raw
In reply to Dan Li

Hi,

Quoted 3 lines · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.

Neat!

My local GCC devs told me that means GCC 12.x.x rather than 11.2.x or 11.3.x, so as others have said it'd be clearer to say GCC >= 12.0.0.

I'd like to try this with a GCC binary before I provide an Ack or R-b; but in the mean time I have a few comments below.

Review

Quoted 10 lines

Note that the implementation in GCC is slightly different from Clang. With SCS enabled, functions will only pop x30 once in the epilogue, like:

   str     x30, [x18], #8
   stp     x29, x30, [sp, #-16]!
   ......
(quoted diff) +1 -1
-  ldp     x29, x30, [sp], #16	  //clang
+  ldr     x29, [sp], #16	  //GCC
   ldr     x30, [x18, #-8]!

Given the prologue still pushes both x29 and x30 (which we critically depend upon) that sounds OK to me.

Quoted 9 lines

---
FYI:
This function can be used to test if the shadow call stack works:
//noinline void __noscs scs_test(void)
noinline void scs_test(void)

Quoted 7 lines
{
    register unsigned long *sp asm("sp");
    unsigned long * lr = sp + 1;

    asm volatile("":::"x30");
    *lr = 0;
}

It's probably be better to use __builtin_frame_address(0) to get the address of the frame record rather than assuming that fp==sp in the middle of the function.

Quoted 29 lines
ffff800008012704:       d503233f        paciasp
ffff800008012708:       f800865e        str     x30, [x18], #8
ffff80000801270c:       a9bf7bfd        stp     x29, x30, [sp, #-16]!
ffff800008012710:       910003fd        mov     x29, sp
ffff800008012714:       910003e0        mov     x0, sp
ffff800008012718:       f900041f        str     xzr, [x0, #8]
ffff80000801271c:       f85f8e5e        ldr     x30, [x18, #-8]!
ffff800008012720:       f84107fd        ldr     x29, [sp], #16
ffff800008012724:       d50323bf        autiasp
ffff800008012728:       d65f03c0        ret

If SCS protection is enabled, this function will return normally. If the function has __noscs attribute (scs disabled), it will crash due to 0 address access.

 arch/Kconfig                 | 6 +++---
 arch/arm64/Kconfig           | 2 +-
 include/linux/compiler-gcc.h | 4 ++++
 3 files changed, 8 insertions(+), 4 deletions(-)
arch/Kconfig +0 -1
index 678a80713b21..35db7b72bdb0 100644
@@ -604,11 +604,11 @@ config ARCH_SUPPORTS_SHADOW_CALL_STACK
 	  switching.
 
 config SHADOW_CALL_STACK
-	bool "Clang Shadow Call Stack"

arch/Kconfig config ARCH_SUPPORTS_SHADOW_CALL_STACK · line 614

Quoted 10 lines
arch/Kconfig +3 -2
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
 	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
 	help
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC'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 in
 	  Clang's documentation:

Is there any additional GCC documentation that we can refer to?

arch/arm64/Kconfig config HW_PERF_EVENTS · line 1258

Quoted 10 lines
arch/arm64/Kconfig +1 -1
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
 config ARCH_HAS_FILTER_PGPROT
 	def_bool y
 
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0

As above, I beleive that should be GCC >= 12.0.0.

Thanks,
Mark

Quoted 21 lines
arch/arm64/Kconfig +0 -0
 config CC_HAVE_SHADOW_CALL_STACK
 	def_bool $(cc-option, -fsanitize=shadow-call-stack -ffixed-x18)
 
include/linux/compiler-gcc.h +4 -0
index ccbbd31b3aae..deff5b308470 100644
@@ -97,6 +97,10 @@
 #define KASAN_ABI_VERSION 4
 #endif
 
+#ifdef CONFIG_SHADOW_CALL_STACK
+#define __noscs __attribute__((__no_sanitize__("shadow-call-stack")))
+#endif
+
 #if __has_attribute(__no_sanitize_address__)
 #define __no_sanitize_address __attribute__((no_sanitize_address))
 #else
Signature · 1 line
2.17.1
author
#10 lore raw
Quoted 12 lines · Mark Rutland

Hi,

Quoted 3 lines · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.

Neat!

My local GCC devs told me that means GCC 12.x.x rather than 11.2.x or 11.3.x, so as others have said it'd be clearer to say GCC >= 12.0.0.

Ah, yes, I just saw this flag in gcc/BASE-VER.

Quoted 3 lines

I'd like to try this with a GCC binary before I provide an Ack or R-b; but in the mean time I have a few comments below.

Thanks for your test, Mark.
Please let me know if there is any issues :)

Quoted 9 lines
Quoted 9 lines

---
FYI:
This function can be used to test if the shadow call stack works:
//noinline void __noscs scs_test(void)
noinline void scs_test(void)

{
     register unsigned long *sp asm("sp");
     unsigned long * lr = sp + 1;
Quoted 8 lines
Quoted 3 lines
     asm volatile("":::"x30");
     *lr = 0;
}

It's probably be better to use __builtin_frame_address(0) to get the address of the frame record rather than assuming that fp==sp in the middle of the function.

Got it.

Quoted 6 lines
Quoted 6 lines

config SHADOW_CALL_STACK

arch/Kconfig +2 -2
-	bool "Clang Shadow Call Stack"
-	depends on CC_IS_CLANG && ARCH_SUPPORTS_SHADOW_CALL_STACK
+	bool "Shadow Call Stack"
+	depends on ARCH_SUPPORTS_SHADOW_CALL_STACK
  	depends on DYNAMIC_FTRACE_WITH_REGS || !FUNCTION_GRAPH_TRACER
Quoted 9 lines

arch/Kconfig config ARCH_SUPPORTS_SHADOW_CALL_STACK · line 614

Quoted 6 lines

help

arch/Kconfig +1 -1
-	  This option enables Clang's Shadow Call Stack, which uses a
+	  This option enables Clang/GCC'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 in
  	  Clang's documentation:

Is there any additional GCC documentation that we can refer to?

Currently there is only a brief description of
-fsanitize=shadow-call-stack in the user manual.

Since the description of the clang documentation is more detailed, should I add this gcc link to the configuration description?

Quoted 8 lines
Quoted 8 lines
arch/arm64/Kconfig +0 -0
index 09b885cc4db5..a48a604301aa 100644
@@ -1255,7 +1255,7 @@ config HW_PERF_EVENTS
  config ARCH_HAS_FILTER_PGPROT
  	def_bool y
 
Quoted 5 lines

arch/arm64/Kconfig config HW_PERF_EVENTS · line 1258

Quoted 2 lines
arch/arm64/Kconfig +1 -1
-# Supported by clang >= 7.0
+# Supported by clang >= 7.0 or GCC > 11.2.0

As above, I beleive that should be GCC >= 12.0.0.

Yeah.

Thanks,
Dan.

#11 lore raw
In reply to Mark Rutland
Quoted 27 lines · Mark Rutland

Hi,

Quoted 3 lines · Dan Li

Shadow call stack is available in GCC > 11.2.0, this patch makes the corresponding kernel configuration available when compiling the kernel with gcc.

Neat!

My local GCC devs told me that means GCC 12.x.x rather than 11.2.x or 11.3.x, so as others have said it'd be clearer to say GCC >= 12.0.0.

I'd like to try this with a GCC binary before I provide an Ack or R-b; but in the mean time I have a few comments below.

Review

Quoted 10 lines

Note that the implementation in GCC is slightly different from Clang. With SCS enabled, functions will only pop x30 once in the epilogue, like:

   str     x30, [x18], #8
   stp     x29, x30, [sp, #-16]!
   ......
(quoted diff) +1 -1
-  ldp     x29, x30, [sp], #16          //clang
+  ldr     x29, [sp], #16       //GCC
   ldr     x30, [x18, #-8]!
Quoted 3 lines

Given the prologue still pushes both x29 and x30 (which we critically depend upon) that sounds OK to me.

Indeed.

What did come up in the discussion on the GCC side was runtime patching (to avoid the overhead of having both PAC and SCS), but it seems far more likely that we would patch PACIASP/AUTIASP instructions into SCS pushes/pops rather than the other way around, and so loading x30 only once should be fine.