[PATCH v2 08/11] arm64/static_call: Fix static call CFI violations

[Mark Rutland]

On Tue, Mar 21, 2023 at 09:00:14PM -0700, Josh Poimboeuf wrote:
> On arm64, with CONFIG_CFI_CLANG, it's trivial to trigger CFI violations
> by running "perf record -e sched:sched_switch -a":
> 
>   CFI failure at perf_misc_flags+0x34/0x70 (target: __static_call_return0+0x0/0xc; expected type: 0x837de525)
>   WARNING: CPU: 3 PID: 32 at perf_misc_flags+0x34/0x70
>   CPU: 3 PID: 32 Comm: ksoftirqd/3 Kdump: loaded Tainted: P                   6.3.0-rc2 #8
>   Hardware name: QEMU KVM Virtual Machine, BIOS 0.0.0 02/06/2015
>   pstate: 904000c5 (NzcV daIF +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>   pc : perf_misc_flags+0x34/0x70
>   lr : perf_event_output_forward+0x74/0xf0
>   sp : ffff80000a98b970
>   x29: ffff80000a98b970 x28: ffff00077bd34d00 x27: ffff8000097d2d00
>   x26: fffffbffeff6a360 x25: ffff800009835a30 x24: ffff0000c2e8dca0
>   x23: 0000000000000000 x22: 0000000000000080 x21: ffff00077bd31610
>   x20: ffff0000c2e8dca0 x19: ffff00077bd31610 x18: ffff800008cd52f0
>   x17: 00000000837de525 x16: 0000000072923c8f x15: 000000000000b67e
>   x14: 000000000178797d x13: 0000000000000004 x12: 0000000070b5b3a8
>   x11: 0000000000000015 x10: 0000000000000048 x9 : ffff80000829e2b4
>   x8 : ffff80000829c6f0 x7 : 0000000000000000 x6 : 0000000000000000
>   x5 : fffffbffeff6a340 x4 : ffff00077bd31610 x3 : ffff00077bd31610
>   x2 : ffff800009833400 x1 : 0000000000000000 x0 : ffff00077bd31610
>   Call trace:
>    perf_misc_flags+0x34/0x70
>    perf_event_output_forward+0x74/0xf0
>    __perf_event_overflow+0x12c/0x1e8
>    perf_swevent_event+0x98/0x1a0
>    perf_tp_event+0x140/0x558
>    perf_trace_run_bpf_submit+0x88/0xc8
>    perf_trace_sched_switch+0x160/0x19c
>    __schedule+0xabc/0x153c
>    dynamic_cond_resched+0x48/0x68
>    run_ksoftirqd+0x3c/0x138
>    smpboot_thread_fn+0x26c/0x2f8
>    kthread+0x108/0x1c4
>    ret_from_fork+0x10/0x20
> 
> The problem is that the __perf_guest_state() static call does an
> indirect branch to __static_call_return0(), which isn't CFI-compliant.

IIUC that'd be broken even with the old CFI mechanism, since commit:

  87b940a0675e2526 ("perf/core: Use static_call to optimize perf_guest_info_callbacks")

If so, we probably want a Fixes tag?

> Fix that by generating custom CFI-compliant ret0 functions for each
> defined static key.
> 
> Signed-off-by: Josh Poimboeuf <jpoimboe at kernel.org>
> ---
>  arch/Kconfig                            |  4 ++
>  arch/arm64/include/asm/static_call.h    | 29 +++++++++++
>  include/linux/static_call.h             | 64 +++++++++++++++++++++----
>  include/linux/static_call_types.h       |  4 ++
>  kernel/Makefile                         |  2 +-
>  kernel/static_call.c                    |  2 +-
>  tools/include/linux/static_call_types.h |  4 ++
>  7 files changed, 97 insertions(+), 12 deletions(-)
>  create mode 100644 arch/arm64/include/asm/static_call.h
> 
> diff --git a/arch/Kconfig b/arch/Kconfig
> index e3511afbb7f2..8800fe80a0f9 100644
> --- a/arch/Kconfig
> +++ b/arch/Kconfig
> @@ -1348,6 +1348,10 @@ config HAVE_STATIC_CALL_INLINE
>  	depends on HAVE_STATIC_CALL
>  	select OBJTOOL
>  
> +config CFI_WITHOUT_STATIC_CALL
> +	def_bool y
> +	depends on CFI_CLANG && !HAVE_STATIC_CALL
> +
>  config HAVE_PREEMPT_DYNAMIC
>  	bool
>  
> diff --git a/arch/arm64/include/asm/static_call.h b/arch/arm64/include/asm/static_call.h
> new file mode 100644
> index 000000000000..b3489cac7742
> --- /dev/null
> +++ b/arch/arm64/include/asm/static_call.h
> @@ -0,0 +1,29 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +#ifndef _ASM_ARM64_STATIC_CALL_H
> +#define _ASM_ARM64_STATIC_CALL_H
> +
> +/*
> + * Make a dummy reference to a function pointer in C to force the compiler to
> + * emit a __kcfi_typeid_ symbol for asm to use.
> + */
> +#define GEN_CFI_SYM(func)						\
> +	static typeof(func) __used __section(".discard.cfi") *__UNIQUE_ID(cfi) = func
> +
> +
> +/* Generate a CFI-compliant static call NOP function */
> +#define __ARCH_DEFINE_STATIC_CALL_CFI(name, insns)			\
> +	asm(".align 4						\n"	\
> +	    ".word __kcfi_typeid_" name "			\n"	\
> +	    ".globl " name "					\n"	\
> +	    name ":						\n"	\
> +	    "bti c						\n"	\
> +	    insns "						\n"	\
> +	    "ret						\n"	\
> +	    ".type " name ", @function				\n"	\
> +	    ".size " name ", . - " name "			\n")
> +
> +#define __ARCH_DEFINE_STATIC_CALL_RET0_CFI(name)			\
> +	GEN_CFI_SYM(STATIC_CALL_RET0_CFI(name));			\
> +	__ARCH_DEFINE_STATIC_CALL_CFI(STATIC_CALL_RET0_CFI_STR(name), "mov x0, xzr")

This looks correct, but given we're generating a regular functions it's
unfortunate we can't have the compiler generate the actual code with something
like:

#define __ARCH_DEFINE_STATIC_CALL_RET0_CFI(rettype, name, args...)	\
rettype name(args)							\
{									\
	return (rettype)0;						\
}

... but I guess passing the rettype and args around is painful.

Regardless, I gave this a spin atop v6.3-rc3 using LLVM 16.0.0 and CFI_CLANG,
and it does seem to work, so:

Tested-by: Mark Rutland <mark.rutland at arm.com>

Thanks,
Mark.