All of lore.kernel.org
 help / color / mirror / Atom feed
From: Nick Desaulniers <ndesaulniers@google.com>
To: ashimida <ashimida@linux.alibaba.com>
Cc: masahiroy@kernel.org, michal.lkml@markovi.net, nathan@kernel.org,
	linux-kbuild@vger.kernel.org, linux-kernel@vger.kernel.org,
	clang-built-linux@googlegroups.com,
	Alistair Delva <adelva@google.com>
Subject: Re: [PATCH] [RFC] kbuild: add CLANG_TRIPLE to prevent clang from compiling with wrong --target
Date: Thu, 9 Sep 2021 10:19:52 -0700	[thread overview]
Message-ID: <CAKwvOdnuiV3mHxxCpWbMaZn9vggL4B+PPrMtuX=QOO-yUQj2mA@mail.gmail.com> (raw)
In-Reply-To: <1631173363-40160-1-git-send-email-ashimida@linux.alibaba.com>

On Thu, Sep 9, 2021 at 12:42 AM ashimida <ashimida@linux.alibaba.com> wrote:
>
> Kernel compiled with tool chain CROSS_COMPILE=aarch64-linux-android-
> will panic during the startup phase.
>
> Clang's --target option comes from $(CROSS_COMPILE). At the time
> -fstack-protector-strong is enabled, and compiled with command:
> make CC=clang HOSTCC=clang ARCH=arm64 CROSS_COMPILE=aarch64-linux-android-
>
> clang will insert code like:
>    mrs     x8, TPIDR_EL0        //default value is zero
>    str     x8, [sp]
>    ldr     x8, [x8, #40]        //access addr 0x40
>
> instead of the code that accesses __stack_chk_guard to get the
> canary, which will cause the kernel to crash due to 0x40
> address access.
>
> This patch (from android) is used to remind the user that current
> tool chain cannot be used as the "--target" of clang, the user
> should specify an additional "--target" through CLANG_TRIPLE.

Hi Ashimida,
Thanks for sending this patch; I recognize it from Android, which we
had to carry for years due to:
1. reliance on GNU `as` ie. "GAS"
2. not distributing binary prefixes of GNU binutils with a target
triple that clang recognized. (ie. Android's binutils were prefixed
aarch64-linux-android- while Clang expected something more like
aarch64-linux-gnu for --target=)

We solved this by working out the issues in clang's assembler.  With
LLVM=1 LLVM_IAS=1, we no longer rely on GNU binutils, and no longer
need such patch.  You'll find it's been dropped from Android Common
Kernels now.  With mainline, LLVM_IAS=1 is now the default when
building with LLVM=1, and CROSS_COMPILE is now inferred from ARCH for
LLVM=1 as well.

So all you should need is:
$ ARCH=arm64 make LLVM=1 -j$(nproc)

Is there a reason why the above doesn't work for you?  I do not wish
to see this patch upstream (or downstream; it should be unnecessary).

>
> Signed-off-by: ashimida <ashimida@linux.alibaba.com>
> ---
>  Makefile                 | 6 +++++-
>  scripts/clang-android.sh | 4 ++++
>  2 files changed, 9 insertions(+), 1 deletion(-)
>  create mode 100755 scripts/clang-android.sh
>
> diff --git a/Makefile b/Makefile
> index 61741e9..09bb314 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -586,7 +586,11 @@ CC_VERSION_TEXT = $(subst $(pound),,$(shell $(CC) --version 2>/dev/null | head -
>
>  ifneq ($(findstring clang,$(CC_VERSION_TEXT)),)
>  ifneq ($(CROSS_COMPILE),)
> -CLANG_FLAGS    += --target=$(notdir $(CROSS_COMPILE:%-=%))
> +CLANG_TRIPLE    ?= $(CROSS_COMPILE)
> +CLANG_FLAGS     += --target=$(notdir $(CLANG_TRIPLE:%-=%))
> +ifeq ($(shell $(srctree)/scripts/clang-android.sh $(CC) $(CLANG_FLAGS)), y)
> +$(error "Clang with Android --target detected. Did you specify CLANG_TRIPLE?")
> +endif
>  endif
>  ifeq ($(LLVM_IAS),1)
>  CLANG_FLAGS    += -integrated-as
> diff --git a/scripts/clang-android.sh b/scripts/clang-android.sh
> new file mode 100755
> index 0000000..9186c4f
> --- /dev/null
> +++ b/scripts/clang-android.sh
> @@ -0,0 +1,4 @@
> +#!/bin/sh
> +# SPDX-License-Identifier: GPL-2.0
> +
> +$* -dM -E - </dev/null 2>&1 | grep -q __ANDROID__ && echo "y"
> --
> 2.7.4
>


-- 
Thanks,
~Nick Desaulniers

  reply	other threads:[~2021-09-09 17:20 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-09  7:42 [PATCH] [RFC] kbuild: add CLANG_TRIPLE to prevent clang from compiling with wrong --target ashimida
2021-09-09 17:19 ` Nick Desaulniers [this message]
2021-09-11 11:55   ` ashimida

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='CAKwvOdnuiV3mHxxCpWbMaZn9vggL4B+PPrMtuX=QOO-yUQj2mA@mail.gmail.com' \
    --to=ndesaulniers@google.com \
    --cc=adelva@google.com \
    --cc=ashimida@linux.alibaba.com \
    --cc=clang-built-linux@googlegroups.com \
    --cc=linux-kbuild@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=masahiroy@kernel.org \
    --cc=michal.lkml@markovi.net \
    --cc=nathan@kernel.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.