From: Nathan Chancellor <nathan@kernel.org>
To: Kees Cook <keescook@chromium.org>
Cc: Marco Elver <elver@google.com>,
Andrey Konovalov <andreyknvl@gmail.com>,
Andrey Ryabinin <ryabinin.a.a@gmail.com>,
Nick Desaulniers <ndesaulniers@google.com>,
Bill Wendling <morbo@google.com>,
Justin Stitt <justinstitt@google.com>,
llvm@lists.linux.dev, kasan-dev@googlegroups.com,
linux-hardening@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] ubsan: Avoid i386 UBSAN handler crashes with Clang
Date: Wed, 24 Apr 2024 12:26:52 -0700 [thread overview]
Message-ID: <20240424192652.GA3341665@dev-arch.thelio-3990X> (raw)
In-Reply-To: <20240424162942.work.341-kees@kernel.org>
Hi Kees,
On Wed, Apr 24, 2024 at 09:29:43AM -0700, Kees Cook wrote:
> When generating Runtime Calls, Clang doesn't respect the -mregparm=3
> option used on i386. Hopefully this will be fixed correctly in Clang 19:
> https://github.com/llvm/llvm-project/pull/89707
> but we need to fix this for earlier Clang versions today. Force the
> calling convention to use non-register arguments.
>
> Reported-by: ernsteiswuerfel
FWIW, I think this can be
Reported-by: Erhard Furtner <erhard_f@mailbox.org>
since it has been used in the kernel before, the reporter is well known
:)
> Closes: https://github.com/KSPP/linux/issues/350
> Signed-off-by: Kees Cook <keescook@chromium.org>
> ---
> Cc: Marco Elver <elver@google.com>
> Cc: Andrey Konovalov <andreyknvl@gmail.com>
> Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
> Cc: Nathan Chancellor <nathan@kernel.org>
> Cc: Nick Desaulniers <ndesaulniers@google.com>
> Cc: Bill Wendling <morbo@google.com>
> Cc: Justin Stitt <justinstitt@google.com>
> Cc: llvm@lists.linux.dev
> Cc: kasan-dev@googlegroups.com
> Cc: linux-hardening@vger.kernel.org
> ---
> lib/ubsan.h | 41 +++++++++++++++++++++++++++--------------
> 1 file changed, 27 insertions(+), 14 deletions(-)
>
> diff --git a/lib/ubsan.h b/lib/ubsan.h
> index 50ef50811b7c..978828f6099d 100644
> --- a/lib/ubsan.h
> +++ b/lib/ubsan.h
> @@ -124,19 +124,32 @@ 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);
> -void __ubsan_handle_out_of_bounds(void *_data, void *index);
> -void __ubsan_handle_shift_out_of_bounds(void *_data, void *lhs, void *rhs);
> -void __ubsan_handle_builtin_unreachable(void *_data);
> -void __ubsan_handle_load_invalid_value(void *_data, void *val);
> -void __ubsan_handle_alignment_assumption(void *_data, unsigned long ptr,
> - unsigned long align,
> - unsigned long offset);
> +/*
> + * When generating Runtime Calls, Clang doesn't respect the -mregparm=3
> + * option used on i386. Hopefully this will be fixed correctly in Clang 19:
> + * https://github.com/llvm/llvm-project/pull/89707
> + * but we need to fix this for earlier Clang versions today. Force the
It may be better to link to the tracking issue upstream instead of the
pull request just in case someone comes up with an alternative fix (not
that I think your change is wrong or anything but it seems like that
happens every so often).
I also get leary of the version information in the comment, even though
I don't doubt this will be fixed in clang 19.
> + * calling convention to use non-register arguments.
> + */
> +#if defined(__clang__) && defined(CONFIG_X86_32)
While __clang__ is what causes CONFIG_CC_IS_CLANG to get set and there
is some existing use of it throughout the kernel, I think
CONFIG_CC_IS_CLANG makes it easier to audit the workarounds that we
have, plus this will be presumably covered to
CONFIG_CLANG_VERSION < 190000
when the fix actually lands. This file is not expected to be used
outside of the kernel, right? That is the only thing I could think of
where this distinction would actually matter.
> +# define ubsan_linkage asmlinkage
Heh, clever...
> +#else
> +# define ubsan_linkage /**/
Why is this defined as a comment rather than just nothing?
> +#endif
> +
> +void ubsan_linkage __ubsan_handle_add_overflow(void *data, void *lhs, void *rhs);
> +void ubsan_linkage __ubsan_handle_sub_overflow(void *data, void *lhs, void *rhs);
> +void ubsan_linkage __ubsan_handle_mul_overflow(void *data, void *lhs, void *rhs);
> +void ubsan_linkage __ubsan_handle_negate_overflow(void *_data, void *old_val);
> +void ubsan_linkage __ubsan_handle_divrem_overflow(void *_data, void *lhs, void *rhs);
> +void ubsan_linkage __ubsan_handle_type_mismatch(struct type_mismatch_data *data, void *ptr);
> +void ubsan_linkage __ubsan_handle_type_mismatch_v1(void *_data, void *ptr);
> +void ubsan_linkage __ubsan_handle_out_of_bounds(void *_data, void *index);
> +void ubsan_linkage __ubsan_handle_shift_out_of_bounds(void *_data, void *lhs, void *rhs);
> +void ubsan_linkage __ubsan_handle_builtin_unreachable(void *_data);
> +void ubsan_linkage __ubsan_handle_load_invalid_value(void *_data, void *val);
> +void ubsan_linkage __ubsan_handle_alignment_assumption(void *_data, unsigned long ptr,
> + unsigned long align,
> + unsigned long offset);
>
> #endif
> --
> 2.34.1
>
>
next prev parent reply other threads:[~2024-04-24 19:26 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-04-24 16:29 [PATCH] ubsan: Avoid i386 UBSAN handler crashes with Clang Kees Cook
2024-04-24 19:26 ` Nathan Chancellor [this message]
2024-04-24 22:32 ` Kees Cook
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240424192652.GA3341665@dev-arch.thelio-3990X \
--to=nathan@kernel.org \
--cc=andreyknvl@gmail.com \
--cc=elver@google.com \
--cc=justinstitt@google.com \
--cc=kasan-dev@googlegroups.com \
--cc=keescook@chromium.org \
--cc=linux-hardening@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=llvm@lists.linux.dev \
--cc=morbo@google.com \
--cc=ndesaulniers@google.com \
--cc=ryabinin.a.a@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).