Problems with examples/new_rule.rs

[RussellStrong]

netlink-packet-route examples/new_rule.rs

running as is:

NetlinkMessage { header: NetlinkHeader { length: 60, message_type: 32, flags: 1541, sequence_number: 0, port_number: 0 }, payload: InnerMessage(NewRule(RuleMessage { header: RuleHeader { family: Inet, dst_len: 0, src_len: 0, tos: 0, table: 254, action: ToTable, flags: RuleFlags(0x0) }, attributes: [Table(254), SuppressPrefixLen(4294967295), Priority(1000), Protocol(Kernel)] })) }
<<< Err(DecodeError { inner: invalid netlink buffer: length field says 0 but netlink packets are at least 4132 bytes })

First problem: send is sending too much data.

It is sending the entire buffer rather than just the serialized netlink message.

strace shows:
sendto(3, [[{nlmsg_len=60, nlmsg_type=RTM_NEWRULE, nlmsg_flags=NLM_F_REQUEST|NLM_F_ACK|NLM_F_EXCL|NLM_F_CREATE, nlmsg_seq=0, nlmsg_pid=0}, {family=AF_INET, dst_len=0, src_len=0, tos=0, table=RT_TABLE_MAIN, action=FR_ACT_TO_TBL, flags=0}, [[{nla_len=8, nla_type=FRA_TABLE}, RT_TABLE_MAIN], [{nla_len=8, nla_type=FRA_SUPPRESS_PREFIXLEN}, 4294967295], [{nla_len=8, nla_type=FRA_PRIORITY}, 1000], [{nla_len=5, nla_type=FRA_PROTOCOL}, RTPROT_KERNEL]]], {nlmsg_len=0, nlmsg_type=0 /* NLMSG_??? */, nlmsg_flags=0, nlmsg_seq=0, nlmsg_pid=0}], 8192, 0, NULL, 0) = 8192

changed statement as line 49:
socket
.send(&buf, 0)
.expect("failed to send netlink message");

to:
socket
.send(&buf[..msg.buffer_len()], 0)
.expect("failed to send netlink message");

strace now shows:
sendto(3, [{nlmsg_len=60, nlmsg_type=RTM_NEWRULE, nlmsg_flags=NLM_F_REQUEST|NLM_F_ACK|NLM_F_EXCL|NLM_F_CREATE, nlmsg_seq=0, nlmsg_pid=0}, {family=AF_INET, dst_len=0, src_len=0, tos=0, table=RT_TABLE_MAIN, action=FR_ACT_TO_TBL, flags=0}, [[{nla_len=8, nla_type=FRA_TABLE}, RT_TABLE_MAIN], [{nla_len=8, nla_type=FRA_SUPPRESS_PREFIXLEN}, 4294967295], [{nla_len=8, nla_type=FRA_PRIORITY}, 1000], [{nla_len=5, nla_type=FRA_PROTOCOL}, RTPROT_KERNEL]]], 60, 0, NULL, 0) = 60

Second problem: received size does not match header size

socket.recv is appending data to a buffer full of zeros giving a received buffer length
of 4132 instead of something smaller. _size on line 55 has a value of 80.

changing:
let mut receive_buffer = vec![0; 4096];
to
let mut receive_buffer = vec![];

This now has _size on line 55 with a value of 64. Why is this not 80 as before?

Decoding still fails:
<<< Err(DecodeError { inner: invalid netlink buffer: length field says 80 the buffer is 64 bytes long })

Interesting that the header length is 80....

Trying to override the header length to 64...

It now decodes

NetlinkMessage { header: NetlinkHeader { length: 60, message_type: 32, flags: 1541, sequence_number: 0, port_number: 0 }, payload: InnerMessage(NewRule(RuleMessage { header: RuleHeader { family: Inet, dst_len: 0, src_len: 0, tos: 0, table: 254, action: ToTable, flags: RuleFlags(0x0) }, attributes: [Table(254), SuppressPrefixLen(4294967295), Priority(1000), Protocol(Kernel)] })) }
64
<<< Ok(NetlinkMessage { header: NetlinkHeader { length: 64, message_type: 2, flags: 0, sequence_number: 0, port_number: 6481 }, payload: Error(ErrorMessage { code: Some(-17), header: [60, 0, 0, 0, 32, 0, 5, 6, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 254, 0, 0, 1, 0, 0, 0, 0, 8, 0, 15, 0, 254, 0, 0, 0, 8, 0, 14, 0, 255, 255, 255, 255] }) })
ErrorMessage { code: Some(-17), header: [60, 0, 0, 0, 32, 0, 5, 6, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 254, 0, 0, 1, 0, 0, 0, 0, 8, 0, 15, 0, 254, 0, 0, 0, 8, 0, 14, 0, 255, 255, 255, 255] }

What is happening here?
Is 64 the correct length here and the kernel has a bug?

[RussellStrong]

New clue

Parsing of the response works when the response is not an Error

to fix the changed line 30 from:

    RuleAttribute::SuppressPrefixLen(4294967295),

to:

    RuleAttribute::SuppressPrefixLen(32),

Causes the kernel to accept the new rule ( no error response ). In this case the response has the correct length and parses correctly.

 Running `/home/russell/rules/target/debug/rules`

NetlinkMessage { header: NetlinkHeader { length: 60, message_type: 32, flags: 1541, sequence_number: 0, port_number: 0 }, payload: InnerMessage(NewRule(RuleMessage { header: RuleHeader { family: Inet, dst_len: 0, src_len: 0, tos: 0, table: 254, action: Unspec, flags: RuleFlags(0x0) }, attributes: [Table(254), SuppressPrefixLen(32), Priority(1000), Protocol(Kernel)] })) }
36
<<< Ok(NetlinkMessage { header: NetlinkHeader { length: 36, message_type: 2, flags: 256, sequence_number: 0, port_number: 3631 }, payload: Error(ErrorMessage { code: None, header: [60, 0, 0, 0, 32, 0, 5, 6, 0, 0, 0, 0, 0, 0, 0, 0] }) })
ErrorMessage { code: None, header: [60, 0, 0, 0, 32, 0, 5, 6, 0, 0, 0, 0, 0, 0, 0, 0] }

[cathay4t]

Regardless the correct SuppressPrefixLen or not, we should parse the error correctly.

Let me investigate more.