Commit 8b8ca9c2 authored by Alice Ryhl's avatar Alice Ryhl Committed by Miguel Ojeda

cfi: fix conditions for HAVE_CFI_ICALL_NORMALIZE_INTEGERS

The HAVE_CFI_ICALL_NORMALIZE_INTEGERS option has some tricky conditions
when KASAN or GCOV are turned on, as in that case we need some clang and
rustc fixes [1][2] to avoid boot failures. The intent with the current
setup is that you should be able to override the check and turn on the
option if your clang/rustc has the fix. However, this override does not
work in practice. Thus, use the new RUSTC_LLVM_VERSION to correctly
implement the check for whether the fix is available.

Additionally, remove KASAN_HW_TAGS from the list of incompatible
options. The CFI_ICALL_NORMALIZE_INTEGERS option is incompatible with
KASAN because LLVM will emit some constructors when using KASAN that are
assigned incorrect CFI tags. These constructors are emitted due to use
of -fsanitize=kernel-address or -fsanitize=kernel-hwaddress that are
respectively passed when KASAN_GENERIC or KASAN_SW_TAGS are enabled.
However, the KASAN_HW_TAGS option relies on hardware support for MTE
instead and does not pass either flag. (Note also that KASAN_HW_TAGS
does not `select CONSTRUCTORS`.)

Link: https://github.com/llvm/llvm-project/pull/104826 [1]
Link: https://github.com/rust-lang/rust/pull/129373 [2]
Fixes: 4c66f830 ("cfi: encode cfi normalized integers + kasan/gcov bug in Kconfig")
Signed-off-by: default avatarAlice Ryhl <aliceryhl@google.com>
Reviewed-by: default avatarSami Tolvanen <samitolvanen@google.com>
Link: https://lore.kernel.org/r/20241010-icall-detect-vers-v1-2-8f114956aa88@google.comSigned-off-by: default avatarMiguel Ojeda <ojeda@kernel.org>
parent af0121c2
......@@ -838,7 +838,7 @@ config CFI_CLANG
config CFI_ICALL_NORMALIZE_INTEGERS
bool "Normalize CFI tags for integers"
depends on CFI_CLANG
depends on HAVE_CFI_ICALL_NORMALIZE_INTEGERS
depends on HAVE_CFI_ICALL_NORMALIZE_INTEGERS_CLANG
help
This option normalizes the CFI tags for integer types so that all
integer types of the same size and signedness receive the same CFI
......@@ -851,21 +851,19 @@ config CFI_ICALL_NORMALIZE_INTEGERS
This option is necessary for using CFI with Rust. If unsure, say N.
config HAVE_CFI_ICALL_NORMALIZE_INTEGERS
def_bool !GCOV_KERNEL && !KASAN
depends on CFI_CLANG
config HAVE_CFI_ICALL_NORMALIZE_INTEGERS_CLANG
def_bool y
depends on $(cc-option,-fsanitize=kcfi -fsanitize-cfi-icall-experimental-normalize-integers)
help
Is CFI_ICALL_NORMALIZE_INTEGERS supported with the set of compilers
currently in use?
# With GCOV/KASAN we need this fix: https://github.com/llvm/llvm-project/pull/104826
depends on CLANG_VERSION >= 190000 || (!GCOV_KERNEL && !KASAN_GENERIC && !KASAN_SW_TAGS)
This option defaults to false if GCOV or KASAN is enabled, as there is
an LLVM bug that makes normalized integers tags incompatible with
KASAN and GCOV. Kconfig currently does not have the infrastructure to
detect whether your rustc compiler contains the fix for this bug, so
it is assumed that it doesn't. If your compiler has the fix, you can
explicitly enable this option in your config file. The Kconfig logic
needed to detect this will be added in a future kernel release.
config HAVE_CFI_ICALL_NORMALIZE_INTEGERS_RUSTC
def_bool y
depends on HAVE_CFI_ICALL_NORMALIZE_INTEGERS_CLANG
depends on RUSTC_VERSION >= 107900
# With GCOV/KASAN we need this fix: https://github.com/rust-lang/rust/pull/129373
depends on (RUSTC_LLVM_VERSION >= 190000 && RUSTC_VERSION >= 108200) || \
(!GCOV_KERNEL && !KASAN_GENERIC && !KASAN_SW_TAGS)
config CFI_PERMISSIVE
bool "Use CFI in permissive mode"
......
......@@ -1950,7 +1950,7 @@ config RUST
depends on !GCC_PLUGIN_RANDSTRUCT
depends on !RANDSTRUCT
depends on !DEBUG_INFO_BTF || PAHOLE_HAS_LANG_EXCLUDE
depends on !CFI_CLANG || RUSTC_VERSION >= 107900 && HAVE_CFI_ICALL_NORMALIZE_INTEGERS
depends on !CFI_CLANG || HAVE_CFI_ICALL_NORMALIZE_INTEGERS_RUSTC
select CFI_ICALL_NORMALIZE_INTEGERS if CFI_CLANG
depends on !CALL_PADDING || RUSTC_VERSION >= 108100
depends on !KASAN_SW_TAGS
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment