[PATCH v3] riscv: probes: reject kprobes inside LR/SC sequences

[Nam Cao]

Xiaofeng Yuan <yuanxiaofeng at eswincomputing.com> writes:
> +	while (tmp > scan_end && count < MAX_ATOMIC_CONTEXT_INSN) {
> +		u16 halfword = get_unaligned((u16 *)(tmp - 2));
> +		unsigned int len = (halfword & 0x3) == 0x3 ? 4 : 2;
> +
> +		if (tmp - len < scan_end)
> +			break;
> +
> +		if (len == 4) {
> +			u32 insn = get_unaligned((u32 *)(tmp - 4));
> +
> +			if (riscv_insn_is_lr(insn))
> +				in_atomic = true;
> +			else if (riscv_insn_is_sc(insn))
> +				in_atomic = false;
> +		}
> +
> +		tmp -= len;
> +		count++;
> +	}

This does not look right. If we have

    lr
    another instruction
    sc
    probed instruction

Then riscv_probe_insn_in_atomic() return true, which is clearly wrong.

Have you even tested this patch?

Nam