Thanks for publishing this package!
I did stumble upon an issue, though: printing an utmp record when SSH'ing into my computer via IPv6, I get:
&{Type:7 Pid:1130555 Device:pts/45 Id: User:michael Host:2a02:168:9300:0:6d9:f5ff:fe1f:9de3 Exit:{Termination:0 Exit:0} Session:0 Time:Sun, 27 Oct 2019 11:56:55 CET Addr:22a:6801:93:0:d906:fff5:1ffe:e39d}
You can see that the Addr
field (parsed by utmpdump) does not match the Host
field (string).
This is because the ut_addr_v6
utmp field is stored in Network Byte Order (big endian), but utmpdump is reading the whole utmp file in little endian:
|
err := binary.Read(file, binary.LittleEndian, u) |
(Side note: always using little endian to read utmp seems incorrect—isn’t the file stored in native endianness, i.e. in big endian on big endian architectures?)
A simple fix is to swap the bytes like so:
first := uint32(0x6801022a)
swapped := (first&0xFF)<<24 | ((first&0xFF00)>>8)<<16 | ((first&0xFF0000)>>16)<<8 | (first&0xFF000000)>>24
fmt.Printf("%x:%x\n", (swapped>>16)&0xffff, swapped&0xffff)
A cleaner, yet backwards compatibility-breaking fix is to change Addr [4]int32
into AddrV6 [16]byte
and return a net.IP from an Addr()
accessor, like so:
ip := make(net.IP, 16)
if err := binary.Read(bytes.NewReader(r.AddrV6[:]), binary.BigEndian, ip); err != nil {
return err
}
if bytes.Equal(ip[4:], net.IPv6zero[4:]) {
// IPv4 address, shorten the slice so that net.IP behaves correctly:
ip = ip[:4]
}
I’m not sure whether you’re following Semantic Import Versioning in this package (you should! :), so I have not sent a pull request. Feel free to use the above in a fix, or let me know if you’d prefer a PR.
A more tedious, yet backwards-compatible way to fix this issue would be to retain the Addr field, add the AddrV6 field, add an accessor, and instead of using binary.Read, read each field separately so that we can populate both Addr and AddrV6.