This is a LLVM and Clang compiler toolchain built for kernel development. Builds are always made from the latest LLVM sources rather than stable releases, so complete stability cannot be guaranteed.
This toolchain targets the AArch32, AArch64, and x86 architectures. It is built with LTO and PGO to reduce compile times as much as possible. Polly, LLVM's polyhedral loop optimizer, is also included for users who want to experiment with additional optimization. Note that this toolchain is not suitable for anything other than bare-metal development; it has not been built with support for any libc or userspace development in mind.
binutils is also included for convenience. Unlike LLVM, however, the latest stable version of binutils is always used to reduce the opportunity for breakage because this project is primarily focused on cutting-edge Clang, not binutils. This means that users do not need to download separate GCC toolchains to build the Linux kernel.
This toolchain is built on Ubuntu 22.04 LTS, which uses glibc 2.35. Compatibility with older distributions cannot be guaranteed. Other libc implementations (such as musl) are not supported.
Make sure you have this toolchain in your PATH
:
export PATH="$HOME/toolchains/gonon-clang/bin:$PATH"
For an AArch64 cross-compilation setup, you must set the following variables. Some of them can be environment variables, but some must be passed directly to make
as a command-line argument. It is recommended to pass all of them as make
arguments to avoid confusing errors:
CC=clang
(must be passed directly tomake
)CROSS_COMPILE=aarch64-linux-gnu-
- If your kernel has a 32-bit vDSO:
CROSS_COMPILE_ARM32=arm-linux-gnueabi-
Optionally, you can also choose to use as many LLVM tools as possible to reduce reliance on binutils. All of these must be passed directly to make
:
AR=llvm-ar
NM=llvm-nm
OBJCOPY=llvm-objcopy
OBJDUMP=llvm-objdump
STRIP=llvm-strip
Note, however, that additional kernel patches may be required for these LLVM tools to work. It is also possible to replace the binutils linkers (lf.bfd
and ld.gold
) with lld
and use Clang's integrated assembler for inline assembly in C code, but that will require many more kernel patches and it is currently impossible to use the integrated assembler for all assembly code in the kernel.
Android kernels older than 4.14 will require patches for compiling with any Clang toolchain to work; those patches are out of the scope of this project. See android-kernel-clang for more information.
Android kernels 4.19 and newer use the upstream variable CROSS_COMPILE_COMPAT
. When building these kernels, replace CROSS_COMPILE_ARM32
in your commands and scripts with CROSS_COMPILE_COMPAT
.
Gonon Clang has been designed to be easy-to-use compared to other toolchains, such as AOSP Clang. The differences are as follows:
CLANG_TRIPLE
does not need to be set because we don't use AOSP binutilsLD_LIBRARY_PATH
does not need to be set because we set library load paths in the toolchain- No separate GCC/binutils toolchains are necessary; all tools are bundled