d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Document support for alternate TCP stacks.

Browse Source 
Differential Revision:	https://reviews.freebsd.org/D6940
Reviewed by:	hiren
Approved by:	re (gjb)
Sponsored by:	Juniper Networks
This commit is contained in:
Jonathan T. Looney 2016-06-28 13:37:01 +00:00
parent 3e04b0a043
commit cf3c688cc9
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=302247
4 changed files with 314 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
share/man/man4/tcp.4
View File

@ -34,7 +34,7 @@
.\" From: @(#)tcp.4 8.1 (Berkeley) 6/5/93
.\" $FreeBSD$
.\"
.Dd May 19, 2016
.Dd June 28, 2016
.Dt TCP 4
.Os
.Sh NAME
@ -119,7 +119,7 @@ supports a number of socket options which can be set with
.Xr setsockopt 2
and tested with
.Xr getsockopt 2 :
.Bl -tag -width ".Dv TCP_CONGESTION"
.Bl -tag -width ".Dv TCP_FUNCTION_BLK"
.It Dv TCP_INFO
Information about a socket's underlying TCP session may be retrieved
by passing the read-only option
@ -148,6 +148,20 @@ connection.
See
.Xr mod_cc 4
for details.
.It Dv TCP_FUNCTION_BLK
Select or query the set of functions that TCP will use for this connection.
This allows a user to select an alternate TCP stack.
The alternate TCP stack must already be loaded in the kernel.
To list the available TCP stacks, see
.Va functions_available
in the
.Sx MIB Variables
section further down.
To list the default TCP stack, see
.Va functions_default
in the
.Sx MIB Variables
section.
.It Dv TCP_KEEPINIT
This
.Xr setsockopt 2
@ -568,6 +582,10 @@ Number of times default MSS was used in an attempt to downshift.
.It Va pmtud_blackhole_failed
Number of connections for which retransmits continued even after MSS
downshift.
.It Va functions_available
List of available TCP function blocks (TCP stacks).
.It Va functions_default
The default TCP function block (TCP stack).
.El
.Sh ERRORS
A socket operation may fail with one of the following errors returned:
@ -599,6 +617,10 @@ exists;
.It Bq Er EAFNOSUPPORT
when an attempt is made to bind or connect a socket to a multicast
address.
.It Bq Er EINVAL
when trying to change TCP function blocks at an invalid point in the session;
.It Bq Er ENOENT
when trying to use a TCP function block that is not available;
.El
.Sh SEE ALSO
.Xr getsockopt 2 ,

3
share/man/man9/Makefile
View File

@ -284,6 +284,7 @@ MAN= accept_filter.9 \
sysctl_ctx_init.9 \
SYSINIT.9 \
taskqueue.9 \
tcp_functions.9 \
thread_exit.9 \
time.9 \
timeout.9 \
@ -1734,6 +1735,8 @@ MLINKS+=taskqueue.9 TASK_INIT.9 \
taskqueue.9 taskqueue_start_threads_pinned.9 \
taskqueue.9 taskqueue_unblock.9 \
taskqueue.9 TIMEOUT_TASK_INIT.9
MLINKS+=tcp_functions.9 register_tcp_functions.9 \
tcp_functions.9 deregister_tcp_functions.9
MLINKS+=time.9 boottime.9 \
time.9 time_second.9 \
time.9 time_uptime.9

285
share/man/man9/tcp_functions.9 Normal file
View File

@ -0,0 +1,285 @@
.\"
.\" Copyright (c) 2016 Jonathan Looney <jtl@FreeBSD.org>
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
.\" ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd June 28, 2016
.Dt TCP_FUNCTIONS 9
.Os
.Sh NAME
.Nm tcp_functions
.Nd Alternate TCP Stack Framework
.Sh SYNOPSIS
.In netinet/tcp.h
.In netinet/tcp_var.h
.Ft int
.Fn register_tcp_functions "struct tcp_function_block *blk" "int wait"
.Ft int
.Fn deregister_tcp_functions "struct tcp_function_block *blk"
.Sh DESCRIPTION
The
.Nm
framework allows a kernel developer to implement alternate TCP stacks.
The alternate stacks can be compiled in the kernel or can be implemented in
loadable kernel modules.
This functionality is intended to encourage experimentation with the TCP stack
and to allow alternate behaviors to be deployed for different TCP connections
on a single system.
.Pp
A system administrator can set a system default stack.
By default, all TCP connections will use the system default stack.
Additionally, users can specify a particular stack to use on a per-connection
basis.
(See
.Xr tcp 4
for details on setting the system default stack, or selecting a specific stack
for a given connection.)
.Pp
This man page treats "TCP stacks" as synonymous with "function blocks".
This is intentional.
A "TCP stack" is a collection of functions that implement a set of behavior.
Therefore, an alternate "function block" defines an alternate "TCP stack".
.Pp
.Nm
modules must call the
.Fn register_tcp_functions
function during initialization and successfully call the
.Fn deregister_tcp_functions
function prior to allowing the module to be unloaded.
.Pp
The
.Fn register_tcp_functions
function requests that the system add a specified function block to the system.
.Pp
The
.Fn deregister_tcp_functions
function requests that the system remove a specified function block from the
system.
If the call fails because sockets are still using the specified function block,
the system will mark the function block as being in the process of being
removed.
This will prevent additional sockets from using the specified function block.
However, it will not impact sockets that are already using the function block.
.Pp
The
.Fa blk
argument is a pointer to a
.Vt "struct tcp_function_block" ,
which is explained below (see
.Sx Function Block Structure ) .
The
.Fa wait
argument is used as the
.Fa flags
argument to
.Xr malloc 9 ,
and must be set to one of the valid values defined in that man page.
.Ss Function Block Structure
The
.Fa blk argument is a pointer to a
.Vt "struct tcp_function_block" ,
which has the following members:
.Bd -literal -offset indent
struct tcp_function_block {
char tfb_tcp_block_name[TCP_FUNCTION_NAME_LEN_MAX];
int (*tfb_tcp_output)(struct tcpcb *);
void (*tfb_tcp_do_segment)(struct mbuf *, struct tcphdr *,
struct socket *, struct tcpcb *,
int, int, uint8_t,
int);
int (*tfb_tcp_ctloutput)(struct socket *so,
struct sockopt *sopt,
struct inpcb *inp, struct tcpcb *tp);
/* Optional memory allocation/free routine */
void (*tfb_tcp_fb_init)(struct tcpcb *);
void (*tfb_tcp_fb_fini)(struct tcpcb *);
/* Optional timers, must define all if you define one */
int (*tfb_tcp_timer_stop_all)(struct tcpcb *);
void (*tfb_tcp_timer_activate)(struct tcpcb *,
uint32_t, u_int);
int (*tfb_tcp_timer_active)(struct tcpcb *, uint32_t);
void (*tfb_tcp_timer_stop)(struct tcpcb *, uint32_t);
void (*tfb_tcp_rexmit_tmr)(struct tcpcb *);
volatile uint32_t tfb_refcnt;
uint32_t tfb_flags;
};
.Ed
.Pp
The
.Va tfb_tcp_block_name
field identifies the unique name of the TCP stack, and should be no longer than
TCP_FUNCTION_NAME_LEN_MAX-1 characters in length.
.Pp
The
.Va tfb_tcp_output ,
.Va tfb_tcp_do_segment ,
and
.Va tfb_tcp_ctloutput
fields are pointers to functions that perform the equivalent actions
as the default
.Fn tcp_output ,
.Fn tcp_do_segment ,
and
.Fn tcp_default_ctloutput
functions, respectively.
Each of these function pointers must be non-NULL.
.Pp
If a TCP stack needs to initialize data when a socket first selects the TCP
stack (or, when the socket is first opened), it should set a non-NULL
pointer in the
.Va tfb_tcp_fb_init
field.
Likewise, if a TCP stack needs to cleanup data when a socket stops using the
TCP stack (or, when the socket is closed), it should set a non-NULL pointer
in the
.Va tfb_tcp_fb_fini
field.
.Pp
If the TCP stack implements additional timers, the TCP stack should set a
non-NULL pointer in the
.Va tfb_tcp_timer_stop_all ,
.Va tfb_tcp_timer_activate ,
.Va tfb_tcp_timer_active ,
and
.Va tfb_tcp_timer_stop
fields.
These fields should all be
.Dv NULL
or should all contain pointers to functions.
The
.Va tfb_tcp_timer_activate ,
.Va tfb_tcp_timer_active ,
and
.Va tfb_tcp_timer_stop
functions will be called when the
.Fn tcp_timer_activate ,
.Fn tcp_timer_active ,
and
.Fn tcp_timer_stop
functions, respectively, are called with a timer type other than the standard
types.
The functions defined by the TCP stack have the same semantics (both for
arguments and return values) as the normal timer functions they supplement.
.Pp
Additionally, a stack may define its own actions to take when the retransmit
timer fires by setting a non-NULL function pointer in the
.Va tfb_tcp_rexmit_tmr
field.
This function is called very early in the process of handling a retransmit
timer.
However, care must be taken to ensure the retransmit timer leaves the
TCP control block in a valid state for the remainder of the retransmit
timer logic.
.Pp
The
.Va tfb_refcnt
and
.Va tfb_flags
fields are used by the kernel's TCP code and will be initialized when the
TCP stack is registered.
.Ss Requirements for Alternate TCP Stacks
If the TCP stack needs to store data beyond what is stored in the default
TCP control block, the TCP stack can initialize its own per-connection storage.
The
.Va t_fb_ptr
field in the
.Vt "struct tcpcb"
control block structure has been reserved to hold a pointer to this
per-connection storage.
If the TCP stack uses this alternate storage, it should understand that the
value of the
.Va t_fb_ptr
pointer may not be initialized to
.Dv NULL .
Therefore, it should use a
.Va tfb_tcp_fb_init
function to initialize this field.
Additionally, it should use a
.Va tfb_tcp_fb_fini
function to deallocate storage when the socket is closed.
.Pp
It is understood that alternate TCP stacks may keep different sets of data.
However, in order to ensure that data is available to both the user and the
rest of the system in a standardized format, alternate TCP stacks must
update all fields in the TCP control block to the greatest extent practical.
.Sh RETURN VALUES
The
.Fn register_tcp_functions
and
.Fn deregister_tcp_functions
functions return zero on success and non-zero on failure.
In particular, the
.Fn deregister_tcp_functions
will return
.Er EBUSY
until no more connections are using the specified TCP stack.
A module calling
.Fn deregister_tcp_functions
must be prepared to wait until all connections have stopped using the
specified TCP stack.
.Sh ERRORS
The
.Fn register_tcp_functions
function will fail if:
.Bl -tag -width Er
.It Bq Er EINVAL
Any of the members of the
.Fa blk
argument are set incorrectly.
.It Bq Er ENOMEM
The function could not allocate memory for its internal data.
.It Bq Er EALREADY
A function block is already registered with the same name.
.El
The
.Fn deregister_tcp_functions
function will fail if:
.Bl -tag -width Er
.It Bq Er EPERM
The
.Fa blk
argument references the kernel's compiled-in default function block.
.It Bq Er EBUSY
The function block is still in use by one or more sockets, or is defined as
the current default function block.
.It Bq Er ENOENT
The
.Fa blk
argument references a function block that is not currently registered.
.Sh SEE ALSO
.Xr malloc 9 ,
.Xr tcp 4
.Sh HISTORY
This framework first appeared in
.Fx 11.0 .
.Sh AUTHORS
.An -nosplit
The
.Nm
framework was written by
.An Randall Stewart Aq Mt rrs@FreeBSD.org .
.Pp
This manual page was written by
.An Jonathan Looney Aq Mt jtl@FreeBSD.org .

2
tools/build/options/WITH_EXTRA_TCP_STACKS Normal file
View File

@ -0,0 +1,2 @@
.\" $FreeBSD$
Set to build extra TCP stack modules.