freebsd-nq/share/man/man4/ng_pipe.4
Christian Brueffer 17019bff0c Add a manpage for ng_pipe(4).
Submitted by:	Lutz Donnerhacke <lutz_donnerhacke.de>
Reviewed by:	bcr (previous version)
MFC after:	1 week
Differential Revision:	https://reviews.freebsd.org/D22067
2019-10-20 20:57:57 +00:00

216 lines
6.8 KiB
Groff

.\" Copyright (c) 2019 Lutz Donnerhacke
.\" Copyright (c) 2004-2008 University of Zagreb
.\" Copyright (c) 2007-2008 FreeBSD Foundation
.\" 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 October 17, 2019
.Dt NG_PIPE 4
.Os
.Sh NAME
.Nm ng_pipe
.Nd Traffic manipulating netgraph node type
.Sh SYNOPSIS
.In netgraph/ng_pipe.h
.Sh DESCRIPTION
The
.Nm pipe
node type manipulates traffic by emulating bandwidth and delay, as well as
random packet losses.
.Sh HOOKS
This node type supports the following hooks:
.Bl -tag -width ".Va upper"
.It Va upper
Hook leading to upper layer protocols.
.It Va lower
Hook leading to lower layer protocols.
.El
.Pp
Traffic flowing from
.Va upper
to
.Va lower
is considered
.Sy downstream
traffic.
Traffic flowing from
.Va lower
to
.Va upper
is considered
.Sy upstream
traffic.
.Sh MODE OF OPERATION
Data received on a hook - both in upstream and downstream direction -
is put into an inbound queue.
If inbound queue is full, discard one frame
depending on dropping policy (from the head or from the tail of the
queue).
.Pp
There are three mutually exclusive modes for the input queue:
.Bl -tag -width foo
.It Dv "First In First Out (FIFO)"
A single queue holds packets in chronological order.
.It Dv Weighted fair queuing (WFQ)
There are multiple queues for different traffic flows (based on IPv4
IPs).
The longest queue is truncated if necessary.
This approach
assumes that the stalling flow is the flow with the most packets currently
on hold.
.It Dv Deficit Round Robin (DRR)
This mode is similar to WFQ, but packets are not taken out in
strict chronological order.
In principle oldest packets come first,
but not too many packets from the same flow.
.El
.Pp
It is possible to configure a duplication probability.
As the dice
decides, the currently active packet stays in the queue while a copy
of the packet is sent out.
Nothing prevents a packet from being
duplicated multiple times.
.Pp
Packets are dropped with an increasing probability depending on the
size of the packet, if a
.Va ber
(bit error rate) is configured.
.Pp
Surviving packets are delayed by the time the packet would need to
travel through a link of the configured bandwidth.
If this outbound
queue is full, the packet is dropped.
.Sh CONTROL MESSAGES
This node type supports the generic control messages and the following
specific messages.
.Bl -tag -width foo
.It Dv NGM_PIPE_SET_CFG Pq Ic setcfg
Set node configuration to the one specified in
.Vt "struct ng_pipe_cfg"
.Pp
Note: To set a value to zero, specify -1 instead.
This allows omitting configuration values, which should not be
modified.
.It Dv NGM_PIPE_GET_CFG Pq Ic getcfg
Return current node configuration as
.Vt "struct ng_pipe_cfg"
.Bd -literal
struct ng_pipe_cfg {
u_int64_t bandwidth; /* bits per second */
u_int64_t delay; /* additional delay, usec */
u_int32_t header_offset; /* offset of IP header in bytes */
u_int32_t overhead; /* assumed L2 overhead in bytes */
struct ng_pipe_hookcfg downstream;
struct ng_pipe_hookcfg upstream;
};
/* Config structure for one hook */
struct ng_pipe_hookcfg {
u_int64_t bandwidth; /* bits per second */
u_int64_t ber; /* errors per 2^48 bits */
u_int32_t qin_size_limit; /* number of queue items */
u_int32_t qout_size_limit; /* number of queue items */
u_int32_t duplicate; /* probability in % */
u_int32_t fifo; /* 0 = off, 1 = on */
u_int32_t drr; /* 0 = off, 1 = 2048 bytes, or x bytes */
u_int32_t wfq; /* 0 = off, 1 = on */
u_int32_t droptail; /* 0 = off, 1 = on */
u_int32_t drophead; /* 0 = off, 1 = on */
};
.Ed
.It Dv NGM_PIPE_GET_STATS Pq Ic getstats
Return node statistics as
.Vt "struct ng_pipe_stats"
.Bd -literal
/* Statistics structure for one hook */
struct ng_pipe_hookstat {
u_int64_t fwd_octets;
u_int64_t fwd_frames;
u_int64_t in_disc_octets;
u_int64_t in_disc_frames;
u_int64_t out_disc_octets;
u_int64_t out_disc_frames;
};
/* Statistics structure returned by NGM_PIPE_GET_STATS */
struct ng_pipe_stats {
struct ng_pipe_hookstat downstream;
struct ng_pipe_hookstat upstream;
};
.Ed
.It Dv NGM_PIPE_CLR_STATS Pq Ic clrstats
Clear node statistics.
.It Dv NGM_PIPE_GETCLR_STATS Pq Ic getclrstats
Atomically return and clear node statistics.
.It Dv NGM_PIPE_GET_RUN Pq Ic getrun
Return node statistics as
.Vt "struct ng_pipe_run"
.Bd -literal
/* Runtime structure for one hook */
struct ng_pipe_hookrun {
u_int32_t fifo_queues;
u_int32_t qin_octets;
u_int32_t qin_frames;
u_int32_t qout_octets;
u_int32_t qout_frames;
};
/* Runtime structure returned by NGM_PIPE_GET_RUN */
struct ng_pipe_run {
struct ng_pipe_hookrun downstream;
struct ng_pipe_hookrun upstream;
};
.Ed
.El
.Sh SHUTDOWN
This node shuts down upon receipt of a
.Dv NGM_SHUTDOWN
control message, or when all hooks have been disconnected.
.Sh EXAMPLES
Limit outgoing data rate over fxp0 Ethernet interface to 20Mbps in
fifo mode and incoming to 50kbps in drr mode and 2% duplicate
probability.
.Bd -literal -offset indent
/usr/sbin/ngctl -f- <<-SEQ
mkpeer fxp0: pipe lower lower
name fxp0:lower fxp0_pipe
connect fxp0: fxp0_pipe: upper upper
msg fxp0_pipe: setcfg { downstream={ bandwidth=20000000 fifo=1 } }
msg fxp0_pipe: setcfg { upstream={ bandwidth=500000 drr=1 duplicate=2 } }
SEQ
.Ed
.Sh SEE ALSO
.Xr netgraph 4 ,
.Xr ngctl 8
.Sh AUTHORS
.An Lutz Donnerhacke Aq Mt lutz@donnerhacke.de
.Pq man page
.Sh BUGS
Error handling for memory issues is missing.
If kernel memory cannot be allocated immediately, a kernel panic will
be triggered.
Same happens if an mbuf is fragmented within the transport headers.