tpcpr/src/parser/mod.rs

extern crate bitfield;
extern crate byteorder;
extern crate eui48;
mod packet_handler;
use pcap::{Capture, Linktype};
use regex::bytes::Regex;
use std::convert::TryInto;
use std::str;
//use std::thread::{spawn, JoinHandle};
//use std::sync::mpsc::{channel, Receiver};

/* protocol ids, LittleEndian */
const ETH_P_IPV6: usize = 0xDD86;
const ETH_P_IP: usize = 0x08;
const TCP: usize = 0x06;
const UDP: usize = 0x11;
const ETH_P_ARP: usize = 0x0608;
const ETH_P_RARP: usize = 0x3580;

/* Protocol header sizes */
const ETHER_HDRLEN: usize = 0xE;
const NO_PREDECESSOR: usize = 0x0;
const IPV6_HDRLEN: u32 = 0xA; // I know, this will get changed. It works for now.

/* random constants */
const IPV4: usize = 0x4;
const IPV6: usize = 0x6;

/*
 QryData could be written in the sense of QryData{ ... frame: .., packet: .., segment:.. }
 On the one hand, only the actual type of frame/packet/segment would be contained in the resulting struct.
 So, increased benefit in serialization/cpu time, could result in less data to be serialized, depending on layout.
 On the other hand, each datagram::type needs to implement traits which would need to be dynamically dispatched by returning any of these types per iso level from a single function each. The result would be a performance decrease.
 See: https://doc.rust-lang.org/book/ch10-02-traits.html#returning-types-that-implement-traits
 See: https://doc.rust-lang.org/book/ch17-02-trait-objects.html#trait-objects-perform-dynamic-dispatch
 Then again, parser logic would be fewer lines + more unified using the latter method. Maybe better optimizable as well? Maybe this is a nice tradeoff?
 TODO: Implement and benchmark dynamically dispatched packet data in conjunction with restructured QryData.
*/
#[derive(Debug, Clone)]
pub struct QryData {
    pub id: i32,
    pub time: f64,
    pub data: Option<Vec<u8>>,
    pub ether_header: Option<packet_handler::EtherHeader>,
    pub ipv4_header: Option<packet_handler::IpV4Header>,
    pub ipv6_header: Option<packet_handler::IpV6Header>,
    pub tcp_header: Option<packet_handler::TcpHeader>,
    pub udp_header: Option<packet_handler::UdpHeader>,
    pub arp_header: Option<packet_handler::ArpHeader>,
    pub reg_res: Option<String>,
}

#[allow(dead_code)]
enum EncapsulationType {
    // pcap::Linktype::get_name() is unsafe.
    EN10MB = 1, // See: https://docs.rs/pcap/0.7.0/src/pcap/lib.rs.html#247-261
    RAW = 101,  // Would this be an issue?
}

impl QryData {
    // This is not cool!
    // Implementing objectoriented is slower by 3-10%. Variance is all over the place. It's awful but modular!
    // Guess I'll do a roolback and do a different design

    fn new() -> QryData {
        QryData {
            id: 0,
            time: 0.0,
            data: None,
            ether_header: None::<packet_handler::EtherHeader>,
            ipv4_header: None::<packet_handler::IpV4Header>,
            ipv6_header: None::<packet_handler::IpV6Header>,
            tcp_header: None::<packet_handler::TcpHeader>,
            udp_header: None::<packet_handler::UdpHeader>,
            arp_header: None::<packet_handler::ArpHeader>,
            reg_res: None::<String>,
        }
    }

    fn encap_en10mb(&mut self, packet_data: &[u8]) -> Result<(), core::fmt::Error> {
        self.ether_header = Some(packet_handler::ethernet_handler(packet_data)).unwrap();
        match self.ether_header.unwrap().ether_type as usize {
            ETH_P_IP => {
                self.ipv4_header =
                    Some(packet_handler::ip_handler(packet_data, ETHER_HDRLEN)).unwrap();
                self.transport_layer(
                    packet_data,
                    self.ipv4_header.unwrap().ip_protocol as usize,
                    self.ipv4_header.unwrap().ip_ihl,
                    ETHER_HDRLEN,
                )
                .unwrap();
            }
            ETH_P_IPV6 => {
                self.ipv6_header =
                    Some(packet_handler::ipv6_handler(packet_data, ETHER_HDRLEN)).unwrap();
                self.transport_layer(
                    packet_data,
                    self.ipv6_header.unwrap().next_header as usize,
                    IPV6_HDRLEN,
                    ETHER_HDRLEN,
                )
                .unwrap();
            }
            ETH_P_ARP | ETH_P_RARP => {
                self.arp_header =
                    Some(packet_handler::arp_handler(packet_data, ETHER_HDRLEN)).unwrap();
            }
            _ => println!("Network protocol not implemented"),
        }
        Ok(())
    }

    fn encap_raw(&mut self, packet_data: &[u8]) -> Result<(), core::fmt::Error> {
        let ip_version: usize = ((packet_data[0] & 0xf0) >> 4).try_into().unwrap();
        match ip_version {
            IPV4 => {
                self.ipv4_header =
                    Some(packet_handler::ip_handler(packet_data, NO_PREDECESSOR)).unwrap();
                self.transport_layer(
                    packet_data,
                    self.ipv4_header.unwrap().ip_protocol as usize,
                    self.ipv4_header.unwrap().ip_ihl,
                    NO_PREDECESSOR,
                )
                .unwrap();
            }
            IPV6 => {
                self.ipv6_header =
                    Some(packet_handler::ipv6_handler(packet_data, NO_PREDECESSOR)).unwrap();
                self.transport_layer(
                    packet_data,
                    self.ipv6_header.unwrap().next_header as usize,
                    IPV6_HDRLEN,
                    NO_PREDECESSOR,
                )
                .unwrap();
            }
            _ => println!("Network Protocol not implemented"),
        }
        Ok(())
    }

    // TODO: impl correct Err type and use in Result
    fn transport_layer(
        &mut self,
        packet_data: &[u8],
        protocol_type: usize,
        l3_header_length: u32,
        ether_hdrlen: usize,
    ) -> Result<(), core::fmt::Error> {
        match protocol_type {
            TCP => {
                self.tcp_header = Some(packet_handler::tcp_handler(
                    l3_header_length,
                    packet_data,
                    ether_hdrlen,
                ))
                .unwrap();
                self.data = Some(packet_handler::payload_handler(
                    l3_header_length,
                    self.tcp_header.unwrap().data_offset,
                    packet_data,
                    ether_hdrlen,
                ))
                .unwrap();
            }
            UDP => {
                self.udp_header = Some(packet_handler::udp_handler(
                    l3_header_length,
                    packet_data,
                    ether_hdrlen,
                ))
                .unwrap();
                self.data = Some(packet_handler::payload_handler(
                    l3_header_length,
                    7,
                    packet_data,
                    ether_hdrlen,
                ))
                .unwrap();
            }
            _ => println!("Transport layer protocol not implemented"),
        }
        Ok(())
    }
}

/* Regex parse _complete_ package */
fn flag_carnage(re: &Regex, payload: &[u8]) -> Option<String> {
    let mut flags: String = String::new();
    for mat in re.find_iter(payload) {
        // TODO:  Test benchmark format! vs. push_str()
        // flags.push_str(&format!("{} ",std::str::from_utf8(mat.as_bytes()).unwrap()));
        // See: https://github.com/hoodie/concatenation_benchmarks-rs
        flags.push_str(std::str::from_utf8(mat.as_bytes()).unwrap());
        flags.push_str(";");
    }
    match 0 < flags.len() {
        false => None,
        true => Some(flags),
    }
}

pub fn parse(parse_file: &std::path::Path, filter_str: &str, regex_filter: &str) -> Vec<QryData> {
    let mut v: Vec<QryData> = Vec::new();
    let mut cap = Capture::from_file(parse_file).unwrap();
    Capture::filter(&mut cap, &filter_str).unwrap();
    let linktype = cap.get_datalink();
    println!("{:?}", &linktype);
    let re = Regex::new(regex_filter).unwrap();
    while let Ok(packet) = cap.next() {
        let mut me = QryData::new();
        match linktype {
            Linktype(1) => me.encap_en10mb(packet.data).unwrap(), //me = QryData::encap_en10mb(packet.data).unwrap(),        // EN10MB
            Linktype(12) => me.encap_raw(packet.data).unwrap(), //me = QryData::encap_raw(packet.data).unwrap(),     // RAW
            _ => (),
        };

        me.time = (packet.header.ts.tv_usec as f64 / 1000000.0) + packet.header.ts.tv_sec as f64;
        me.reg_res = flag_carnage(&re, packet.data); // Regex overhead is between 4-9% --single threaded-- on complete packet [u8] data
                                                     //v.push(me.clone());
        v.push(QryData {
            id: 0,
            time: me.time,
            data: me.data,
            ether_header: me.ether_header,
            ipv4_header: me.ipv4_header,
            ipv6_header: me.ipv6_header,
            tcp_header: me.tcp_header,
            udp_header: me.udp_header,
            arp_header: me.arp_header,
            reg_res: me.reg_res,
        });
    }
    v
}

/* This could need some love */
pub fn parse_device(
    parse_device: &str,
    filter_str: &str,
    insert_max: &usize,
    regex_filter: &str,
) -> Vec<QryData> {
    //let mut me:  QryData = QryData::new ( );
    let mut v: Vec<QryData> = Vec::new();
    let mut cap = Capture::from_device(parse_device).unwrap().open().unwrap();
    Capture::filter(&mut cap, &filter_str).unwrap();
    let linktype = cap.get_datalink();
    let re = Regex::new(regex_filter).unwrap();
    'parse: while let Ok(packet) = cap.next() {
        let mut me = QryData::new();
        match linktype {
            Linktype(1) => me.encap_en10mb(packet.data).unwrap(), //me = QryData::encap_en10mb(packet.data).unwrap(),
            Linktype(101) => me.encap_raw(packet.data).unwrap(), //me = QryData::encap_raw(packet.data).unwrap(),
            _ => (),
        };

        me.time = (packet.header.ts.tv_usec as f64 / 1000000.0) + packet.header.ts.tv_sec as f64;
        me.reg_res = flag_carnage(&re, packet.data);

        v.push(me.clone());

        if &v.len() >= insert_max {
            break 'parse;
        }
    }
    v
}