Re: [PATCH] cfi: fix conditions in HAVE_CFI_ICALL_NORMALIZE_INTEGERS

[Sami Tolvanen]

Hi Alice,

On Tue, Oct 8, 2024 at 10:46 AM Matthew Maurer <mmaurer@xxxxxxxxxx> wrote:
>
> This makes sense, as some folks have a Rust compiler they know has the
> fix, but build system detection for it isn't there yet. This lets them
> override availability if needed.
>
> That said, we should definitely be sure to get this back to a
> non-configurable toggle once the LLVM version detection is in.
>
> Reviewed-By: Matthew Maurer <mmaurer@xxxxxxxxxx>
>
> On Tue, Oct 8, 2024 at 10:42 AM Alice Ryhl <aliceryhl@xxxxxxxxxx> wrote:
> >
> > 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`.)
> >
> > Additionally, the option is configured to have a prompt and gated behind
> > EXPERT. The previous method for a user override of the option did not
> > actually work. This is expected to be temporary, as I intend to add a
> > precise detection check for 6.13 - I did not included that here to avoid
> > adding a RUSTC_LLVM_VERSION config in a fix.

This sounds reasonable to me.

Reviewed-by: Sami Tolvanen <samitolvanen@xxxxxxxxxx>

Sami