bpf, arm64: Support 32-bit offset jmp instruction

Add support for 32-bit offset jmp instructions. Given the arm64 direct jump
range is +-128MB, which is large enough for BPF prog, jumps beyond this range
are not supported.

Signed-off-by: Xu Kuohai <xukuohai@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Florent Revest <revest@chromium.org>
Acked-by: Florent Revest <revest@chromium.org>
Link: https://lore.kernel.org/bpf/20230815154158.717901-6-xukuohai@huaweicloud.com

Author: Xu Kuohai

arch/arm64/net/bpf_jit_comp.c

@@ -982,7 +982,11 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx,
 
 	/* JUMP off */
 	case BPF_JMP | BPF_JA:
-		jmp_offset = bpf2a64_offset(i, off, ctx);
+	case BPF_JMP32 | BPF_JA:
+		if (BPF_CLASS(code) == BPF_JMP)
+			jmp_offset = bpf2a64_offset(i, off, ctx);
+		else
+			jmp_offset = bpf2a64_offset(i, imm, ctx);
 		check_imm26(jmp_offset);
 		emit(A64_B(jmp_offset), ctx);
 		break;