On Mon, 22 Apr 2024 10:13:59 +0100, Mark Rutland mark.rutland@arm.com wrote:
On Mon, Apr 22, 2024 at 02:04:43PM +0530, Naresh Kamboju wrote:
The arm64 defconfig build failed with gcc-8 and passed with gcc-13.
Reported-by: Linux Kernel Functional Testing lkft@linaro.org
Build log:
/tmp/ccSUNNZy.s: Assembler messages: /tmp/ccSUNNZy.s:3159: Error: unknown architectural extension `pauth' make[5]: *** [scripts/Makefile.build:244: arch/arm64/kvm/pauth.o] Error 1
Steps to reproduce:
# tuxmake --runtime podman --target-arch arm64 --toolchain gcc-8 --kconfig defconfig
I think the key thing here is GCC 8; the associated assembler won't necessarily have ARMv8.3-A support, since all the relevant bits got added around GCC 9.
Looking at the commits, I think this is broken since its introduction in commit:
6ccc971ee2c61a1f ("KVM: arm64: nv: Add emulation for ERETAx instructions")
... where the pauth.c file only depends on ARM64_PTR_AUTH (which doesn't imply AS_HAS_ARMV8_3), but in the file we do:
asm volatile(ARM64_ASM_PREAMBLE ".arch_extension pauth\n" "pacga %0, %1, %2" : "=r" (pac) : "r" (ptr), "r" (mod));
Given the minimum supported toolchain comes with an assembler that doesn't necessarily support ARMv8.3, I reckon we'll either have to make NV pauth support depend upon AS_HAS_ARMV8_3, or manually assemble the PACGA instruction.
I suspect the latter is the better option.
This is what I've done [1]. It certainly isn't pretty, but this counts as an incentive to drop some of the old stuff in the near-ish future.
I've pushed this out to -next, and hopefully the breakage will be solved tomorrow.
M.
[1] https://lore.kernel.org/r/20240422224849.2238222-1-maz@kernel.org