hardening: Enable i386 FORTIFY_SOURCE on Clang 16+

PATCH Applied Kees Cook 2 participants 6 messages Mar 3, 2025 – Mar 7, 2025

Subsystems (2)

Security (LSM) Kernel hardening (not covered by other areas)

lore

Share & tools

lore

Kees Cook

author maintainer

2025-03-03 21:49 UTC #0 lore raw

PATCH 1/1 hardening: Enable i386 FORTIFY_SOURCE on Clang 16+

The i386 regparm bug exposed with FORTIFY_SOURCE with Clang was fixed in Clang 16[1].

Link https://github.com/llvm/llvm-project/commit/c167c0a4dcdb998affb2756ce76903a12f7d8ca5 [1]

Signed-off-by Kees Cook <[email protected]>

---
Cc: Nathan Chancellor <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Bill Wendling <[email protected]>
Cc: Justin Stitt <[email protected]>
Cc: "Gustavo A. R. Silva" <[email protected]>
Cc: [email protected]
Cc: [email protected]

---
 security/Kconfig.hardening | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

security/Kconfig.hardening +1 -1

index 23ffb0d7c845..c17366ce8224 100644
@@ -286,7 +286,7 @@ config FORTIFY_SOURCE
 	bool "Harden common str/mem functions against buffer overflows"
 	depends on ARCH_HAS_FORTIFY_SOURCE
 	# https://github.com/llvm/llvm-project/issues/53645
-	depends on !CC_IS_CLANG || !X86_32
+	depends on !X86_32 || !CC_IS_CLANG || CLANG_VERSION >= 160000
 	help
 	  Detect overflows of buffers in common string and memory functions
 	  where the compiler can determine and validate the buffer sizes.

Signature · 1 line

2.34.1

Nathan Chancellor

Reviewed-by on v1 1/1

2025-03-04 10:14 UTC #1 lore raw

Quoted 5 lines · Kees Cook

The i386 regparm bug exposed with FORTIFY_SOURCE with Clang was fixed in Clang 16[1].
Link https://github.com/llvm/llvm-project/commit/c167c0a4dcdb998affb2756ce76903a12f7d8ca5 [1]
Signed-off-by Kees Cook <[email protected]>

Yes, thank you for catching this!

Reviewed-by Nathan Chancellor <[email protected]>

I assume you'll take this?

Quoted 28 lines

---
Cc: Nathan Chancellor <[email protected]>
Cc: Nick Desaulniers <[email protected]>
Cc: Bill Wendling <[email protected]>
Cc: Justin Stitt <[email protected]>
Cc: "Gustavo A. R. Silva" <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
 security/Kconfig.hardening | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
security/Kconfig.hardening +1 -1
index 23ffb0d7c845..c17366ce8224 100644
@@ -286,7 +286,7 @@ config FORTIFY_SOURCE
 	bool "Harden common str/mem functions against buffer overflows"
 	depends on ARCH_HAS_FORTIFY_SOURCE
 	# https://github.com/llvm/llvm-project/issues/53645
-	depends on !CC_IS_CLANG || !X86_32
+	depends on !X86_32 || !CC_IS_CLANG || CLANG_VERSION >= 160000
 	help
 	  Detect overflows of buffers in common string and memory functions
 	  where the compiler can determine and validate the buffer sizes.
Signature · 1 line
2.34.1

Kees Cook

author maintainer Applied

2025-03-04 17:50 UTC #2 lore raw

In reply to Kees Cook

Quoted 4 lines · Kees Cook

The i386 regparm bug exposed with FORTIFY_SOURCE with Clang was fixed in Clang 16[1].

Applied to for-next/hardening, thanks!

[1/1] hardening: Enable i386 FORTIFY_SOURCE on Clang 16+

      https://git.kernel.org/kees/c/3e5820429980

Take care,

Signature · 1 line

Kees Cook

Nathan Chancellor

2025-03-07 21:47 UTC #3 lore raw

Quoted 10 lines · Kees Cook

Quoted 4 lines · Kees Cook
The i386 regparm bug exposed with FORTIFY_SOURCE with Clang was fixed in Clang 16[1].
Applied to for-next/hardening, thanks!
[1/1] hardening: Enable i386 FORTIFY_SOURCE on Clang 16+
      https://git.kernel.org/kees/c/3e5820429980

Turns out this is actually incomplete based on my testing, I see the following warnings with ARCH=i386 allmodconfig with all supported clang versions:

warning: unsafe strcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-strcpy-lit.c
warning: unsafe strcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-strcpy.c

We also need to drop '-ffreestanding' from arch/x86/Makefile (which Nick has mentioned in [1]). Time to revive [2]? :) or just do it in this patch, since it sounds like there was no regression with GCC?

[1]: https://github.com/ClangBuiltLinux/linux/issues/1583#issuecomment-1123016466
[2]: https://lore.kernel.org/[email protected]/

Cheers,
Nathan

Kees Cook

author maintainer

2025-03-07 22:57 UTC #4 lore raw

Quoted 19 lines · Nathan Chancellor

Quoted 10 lines · Kees Cook
Quoted 4 lines · Kees Cook
The i386 regparm bug exposed with FORTIFY_SOURCE with Clang was fixed in Clang 16[1].
Applied to for-next/hardening, thanks!
[1/1] hardening: Enable i386 FORTIFY_SOURCE on Clang 16+
      https://git.kernel.org/kees/c/3e5820429980
Turns out this is actually incomplete based on my testing, I see the following warnings with ARCH=i386 allmodconfig with all supported clang versions:
warning: unsafe strcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-strcpy-lit.c
warning: unsafe strcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-strcpy.c

Quoted 6 lines

We also need to drop '-ffreestanding' from arch/x86/Makefile (which Nick has mentioned in [1]). Time to revive [2]? :) or just do it in this patch, since it sounds like there was no regression with GCC?
[1]: https://github.com/ClangBuiltLinux/linux/issues/1583#issuecomment-1123016466
[2]: https://lore.kernel.org/[email protected]/

Oh good catch! Probably I need to do this in two patches just to be safe:
- make -ffreestanding be Clang only
- switch to version checks for FORTIFY and freestanding

That will split the changes for GCC and Clang into separate patches. Or maybe that's overkill?

-Kees

Signature · 1 line

Kees Cook

Nathan Chancellor

2025-03-07 23:10 UTC #5 lore raw

Quoted 26 lines · Kees Cook

Quoted 23 lines · Nathan Chancellor
Quoted 10 lines · Kees Cook
Quoted 4 lines · Kees Cook
The i386 regparm bug exposed with FORTIFY_SOURCE with Clang was fixed in Clang 16[1].
Applied to for-next/hardening, thanks!
[1/1] hardening: Enable i386 FORTIFY_SOURCE on Clang 16+
      https://git.kernel.org/kees/c/3e5820429980
Turns out this is actually incomplete based on my testing, I see the following warnings with ARCH=i386 allmodconfig with all supported clang versions:
warning: unsafe strcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-strcpy-lit.c
warning: unsafe strcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-strcpy.c
We also need to drop '-ffreestanding' from arch/x86/Makefile (which Nick has mentioned in [1]). Time to revive [2]? :) or just do it in this patch, since it sounds like there was no regression with GCC?

Quoted 8 lines

Quoted 2 lines
[1]: https://github.com/ClangBuiltLinux/linux/issues/1583#issuecomment-1123016466
[2]: https://lore.kernel.org/[email protected]/
Oh good catch! Probably I need to do this in two patches just to be safe:
- make -ffreestanding be Clang only
- switch to version checks for FORTIFY and freestanding
That will split the changes for GCC and Clang into separate patches. Or maybe that's overkill?

No, I think that is reasonable, as any regressions with GCC would be noticed by the first change, rather than being mixed in with the second. It should also make it easier to drop that when the minimum version of clang gets bumped to 16 (either for the whole kernel or just x86, like the bump to 15 this dev cycle for the recent stack protector fixes in -top).

Cheers,
Nathan