Difference between revisions of "TCP connection"

From CRIU
Jump to: navigation, search
(Initial draft)
 
m (Grammar fixes)
 
(19 intermediate revisions by 7 users not shown)
Line 1: Line 1:
This page describes how we handle established TCP connections
+
This page describes how we handle established TCP connections.
  
 
== TCP repair mode in kernel ==
 
== TCP repair mode in kernel ==
  
The sockoption called TCP_REPAIR was recently added to the kernel and helps with doing C/R for TCP sockets.
+
The <code>TCP_REPAIR</code> socket option was added to the kernel 3.5 to help with C/R for TCP sockets.
  
When set this option turn the socket into a special state in which any action performed on it doesn't
+
When this option is used, the socket is switched into a special mode, in which any action performed on it
result in any defined by protocol actions, but instead directly puts the socket into a state, which
+
does not result in anything defined by an appropriate protocol actions, but rather directly puts the socket
should be at the end of the successfully finished operation.
+
into the state that the socket is expected to be in at the end of a successfully finished operation.
  
E.g. calling connect() on a repaired socket just switches one to the ESTABLISHED state with the peer set as requested.
+
For example, calling <code>connect()</code> on a repaired socket just changes its state to <code>ESTABLISHED</code>,
The bind() call forcibly binds the socket to a given address (ignoring any potential conflicts). Close()-ing the
+
with the peer address set as requested.
socket under repair happens without any transient FIN_WAIT/TIME_WAIT/etc states. Socket is silently killed.
+
The <code>bind()</code> call forcibly binds the socket to a given address (ignoring any potential conflicts).
 +
The <code>close()</code> call closes the socket without any transient <code>FIN_WAIT</code>/<code>TIME_WAIT</code>/etc states,
 +
socket is silently killed.
  
 
=== Sequences ===
 
=== Sequences ===
  
In order to restore the connection properly only binding and connecting it is not enough. One also needs to restore the
+
To restore the connection properly, bind() and connect() is not enough. One also needs to restore the
TCP sequence numbers. To do so the TCP_REPAIR_QUEUE and TCP_QUEUE_SEQ options were introduced.
+
TCP sequence numbers. To do so, the <code>TCP_REPAIR_QUEUE</code> and <code>TCP_QUEUE_SEQ</code> options were introduced.
  
The former one selects which queue (input or output) will be repaired and the latter gets/sets the sequence. Note, that
+
The former one selects which queue (input or output) will be repaired and the latter gets/sets the sequence. Note
 
setting the sequence is only possible on CLOSE-d socket.
 
setting the sequence is only possible on CLOSE-d socket.
  
Line 25: Line 27:
 
When set the queue to repair as described above, one can call recv or send syscalls on a repaired socket. Both calls
 
When set the queue to repair as described above, one can call recv or send syscalls on a repaired socket. Both calls
 
result on peeking or poking data from/to the respective queue. This sounds funny, but yes, for repaired socket one
 
result on peeking or poking data from/to the respective queue. This sounds funny, but yes, for repaired socket one
can receve the outgoing and send the incoming queues. Using the MSG_PEEK flag for recv is required.
+
can receve the outgoing and send the incoming queues. Using the <code>MSG_PEEK</code> flag for <code>recv()</code> is required.
  
 
=== Options ===
 
=== Options ===
Line 36: Line 38:
 
* tstamp -- whether timestamps on packets are supported
 
* tstamp -- whether timestamps on packets are supported
  
All four can be read with getsockopt calls to a socket and in order to restore them the TCP_REPAIR_OPTIONS sockoption
+
All four can be read with <code>getsockopt()</code> calls to a socket and in order to restore them the <code>TCP_REPAIR_OPTIONS</code> sockoption is introduced.
is introduced.
+
 
 +
== Timestamp ==
 +
"The sender's timestamp clock is used as a source of monotonic non-decreasing values to stamp the segments"(rfc7323). The Linux kernel uses the jiffies counter as the tcp timestamp.
 +
 
 +
<code>#define tcp_time_stamp          ((__u32)(jiffies))</code>
 +
 
 +
We add the <code>TCP_TIMESTAMP</code> options to be able to compensate a difference between jiffies counters, when a connection is migrated on another host. When a connection is dumped, criu calls <code>getsockopt(TCP_TIMESTAMP)</code> to get a current timestamp, then on restore it calls <code>setsockopt(TCP_TIMESTAMP)</code> to set this timestamp as a starting point.
  
 
== Checkpoint and restore TCP connection ==
 
== Checkpoint and restore TCP connection ==
  
With the above sockoptions dumping and restoring TCP connection becomes possible. The crtools just reads the socket
+
With the above sockoptions dumping and restoring TCP connection becomes possible. The criu just reads the socket
 
state and restores it back letting the protocol resurrect the data sequence.
 
state and restores it back letting the protocol resurrect the data sequence.
  
One thing to note here -- while the socket is closed between dump and restore the connection should be "locked", i.e.
+
One thing to note here while the socket is closed between dump and restore the connection should be "locked", i.e.
 
no packets from peer should enter the stack, otherwise the RST will be sent by a kernel. In order to do so a simple
 
no packets from peer should enter the stack, otherwise the RST will be sent by a kernel. In order to do so a simple
 
netfilter rule is configured that drops all the packets from peer to a socket we're dealing with. This rule sits
 
netfilter rule is configured that drops all the packets from peer to a socket we're dealing with. This rule sits
in the host netfilter tables after the crtools dump command finishes and it should be there when you issue the
+
in the host netfilter tables after the criu dump command finishes and it should be there when you issue the
crtools restore one.
+
criu restore one.
  
That said, the command line option --tcp-established should be used when calling crtools to explicitly state, that the
+
Another thing to note is -- on restore there should be available the IP address, that was used by the connection.
 +
This is automatically so if restore happens on the same box as dump. In case of hand-made live migration the
 +
IP address should be copied too.
 +
 
 +
That said, the command line option <code>--tcp-established</code> should be used when calling criu to explicitly state, that the
 
caller is aware of this "transitional" state of the netfilter.
 
caller is aware of this "transitional" state of the netfilter.
 +
 +
In case the target process lives in NET namespace the connection locking happens the other way. Instead of
 +
per-connection iptables rules the "network-lock"/"network-unlock" [[action scripts]] are called so that the user
 +
could isolate the whole netns from network. Typically this is done by downing the respective veth pair end.
 +
 +
== States ==
 +
=== TCP_SYN_SENT ===
 +
There is only one difference with TCP_ESTABLISHED, we have to restore a socket and disable the repair mode before calling <code>connect()</code>. The kernel will send a one syn-sent packet with the same initial sequence number and sets the TCP_SYN_SENT state for the socket.
 +
 +
=== Half-closed sockets ===
 +
A socket is half-closed when it sent or received a fin packet. These sockets are in one for these states: TCP_FIN_WAIT1, TCP_FIN_WAIT2, TCP_CLOSING, TCP_LAST_ACL, TCP_CLOSE_WAIT. To restore these states, we restore a socket into the TCP_ESTABLISHED state and then we call shutfown(SHUT_WR), if a socket has sent a fin packet and we send a fake fin packet, if a socket has received it before. For example, if we want to restore the TCP_FIN_WAIT1 state, we have to call shutfown(SHUT_WR) and we can send a fake ack to the fin packet to restore the TCP_FIN_WAIT2 state.
 +
 +
== See also ==
 +
* [[Simple TCP pair]]
 +
* [[TCP repair TODO]]
 +
* [[CLI/opt/--tcp-close|Dropping the connection]]
 +
 +
== External links ==
 +
* http://lwn.net/Articles/495304/
 +
 +
[[Category:Under the hood]]
 +
[[Category:Sockets]]
 +
[[Category: Editor help needed]]

Latest revision as of 18:39, 11 March 2019

This page describes how we handle established TCP connections.

TCP repair mode in kernel[edit]

The TCP_REPAIR socket option was added to the kernel 3.5 to help with C/R for TCP sockets.

When this option is used, the socket is switched into a special mode, in which any action performed on it does not result in anything defined by an appropriate protocol actions, but rather directly puts the socket into the state that the socket is expected to be in at the end of a successfully finished operation.

For example, calling connect() on a repaired socket just changes its state to ESTABLISHED, with the peer address set as requested. The bind() call forcibly binds the socket to a given address (ignoring any potential conflicts). The close() call closes the socket without any transient FIN_WAIT/TIME_WAIT/etc states, socket is silently killed.

Sequences[edit]

To restore the connection properly, bind() and connect() is not enough. One also needs to restore the TCP sequence numbers. To do so, the TCP_REPAIR_QUEUE and TCP_QUEUE_SEQ options were introduced.

The former one selects which queue (input or output) will be repaired and the latter gets/sets the sequence. Note setting the sequence is only possible on CLOSE-d socket.

Packets in queue[edit]

When set the queue to repair as described above, one can call recv or send syscalls on a repaired socket. Both calls result on peeking or poking data from/to the respective queue. This sounds funny, but yes, for repaired socket one can receve the outgoing and send the incoming queues. Using the MSG_PEEK flag for recv() is required.

Options[edit]

There are 4 options that are negotiated by the socket at the connecting stage. These are

  • mss_clamp -- the maximum size of the segment peer is ready to accept
  • snd _scale -- the scale factor for a window
  • sack -- whether selective acks are permitted or not
  • tstamp -- whether timestamps on packets are supported

All four can be read with getsockopt() calls to a socket and in order to restore them the TCP_REPAIR_OPTIONS sockoption is introduced.

Timestamp[edit]

"The sender's timestamp clock is used as a source of monotonic non-decreasing values to stamp the segments"(rfc7323). The Linux kernel uses the jiffies counter as the tcp timestamp.

#define tcp_time_stamp ((__u32)(jiffies))

We add the TCP_TIMESTAMP options to be able to compensate a difference between jiffies counters, when a connection is migrated on another host. When a connection is dumped, criu calls getsockopt(TCP_TIMESTAMP) to get a current timestamp, then on restore it calls setsockopt(TCP_TIMESTAMP) to set this timestamp as a starting point.

Checkpoint and restore TCP connection[edit]

With the above sockoptions dumping and restoring TCP connection becomes possible. The criu just reads the socket state and restores it back letting the protocol resurrect the data sequence.

One thing to note here — while the socket is closed between dump and restore the connection should be "locked", i.e. no packets from peer should enter the stack, otherwise the RST will be sent by a kernel. In order to do so a simple netfilter rule is configured that drops all the packets from peer to a socket we're dealing with. This rule sits in the host netfilter tables after the criu dump command finishes and it should be there when you issue the criu restore one.

Another thing to note is -- on restore there should be available the IP address, that was used by the connection. This is automatically so if restore happens on the same box as dump. In case of hand-made live migration the IP address should be copied too.

That said, the command line option --tcp-established should be used when calling criu to explicitly state, that the caller is aware of this "transitional" state of the netfilter.

In case the target process lives in NET namespace the connection locking happens the other way. Instead of per-connection iptables rules the "network-lock"/"network-unlock" action scripts are called so that the user could isolate the whole netns from network. Typically this is done by downing the respective veth pair end.

States[edit]

TCP_SYN_SENT[edit]

There is only one difference with TCP_ESTABLISHED, we have to restore a socket and disable the repair mode before calling connect(). The kernel will send a one syn-sent packet with the same initial sequence number and sets the TCP_SYN_SENT state for the socket.

Half-closed sockets[edit]

A socket is half-closed when it sent or received a fin packet. These sockets are in one for these states: TCP_FIN_WAIT1, TCP_FIN_WAIT2, TCP_CLOSING, TCP_LAST_ACL, TCP_CLOSE_WAIT. To restore these states, we restore a socket into the TCP_ESTABLISHED state and then we call shutfown(SHUT_WR), if a socket has sent a fin packet and we send a fake fin packet, if a socket has received it before. For example, if we want to restore the TCP_FIN_WAIT1 state, we have to call shutfown(SHUT_WR) and we can send a fake ack to the fin packet to restore the TCP_FIN_WAIT2 state.

See also[edit]

External links[edit]