iproute2/man/man8/tc-mqprio.8

.TH MQPRIO 8 "24 Sept 2013" "iproute2" "Linux"
.SH NAME
MQPRIO \- Multiqueue Priority Qdisc (Offloaded Hardware QOS)
.SH SYNOPSIS
.B tc qdisc ... dev
dev
.B  ( parent
classid
.B | root) [ handle
major:
.B ] mqprio [ numtc
tcs
.B ] [ map
P0 P1 P2...
.B ] [ queues
count1@offset1 count2@offset2 ...
.B ] [ hw
1|0
.B ]

.SH DESCRIPTION
The MQPRIO qdisc is a simple queuing discipline that allows mapping
traffic flows to hardware queue ranges using priorities and a configurable
priority to traffic class mapping. A traffic class in this context is
a set of contiguous qdisc classes which map 1:1 to a set of hardware
exposed queues.

By default the qdisc allocates a pfifo qdisc (packet limited first in, first
out queue) per TX queue exposed by the lower layer device. Other queuing
disciplines may be added subsequently. Packets are enqueued using the
.B map
parameter and hashed across the indicated queues in the
.B offset
and
.B count.
By default these parameters are configured by the hardware
driver to match the hardware QOS structures.

Enabled hardware can provide hardware QOS with the ability to steer
traffic flows to designated traffic classes provided by this qdisc.
Configuring the hardware based QOS mechanism is outside the scope of
this qdisc. Tools such as
.B lldpad
and
.B ethtool
exist to provide this functionality. Also further qdiscs may be added
to the classes of MQPRIO to create more complex configurations.

.SH ALGORITHM
On creation with 'tc qdisc add', eight traffic classes are created mapping
priorities 0..7 to traffic classes 0..7 and priorities greater than 7 to
traffic class 0. This requires base driver support and the creation will
fail on devices that do not support hardware QOS schemes.

These defaults can be overridden using the qdisc parameters. Providing
the 'hw 0' flag allows software to run without hardware coordination.

If hardware coordination is being used and arguments are provided that
the hardware can not support then an error is returned. For many users
hardware defaults should work reasonably well.

As one specific example numerous Ethernet cards support the 802.1Q
link strict priority transmission selection algorithm (TSA). MQPRIO
enabled hardware in conjunction with the classification methods below
can provide hardware offloaded support for this TSA.

.SH CLASSIFICATION
Multiple methods are available to set the SKB priority which MQPRIO
uses to select which traffic class to enqueue the packet.
.TP
From user space
A process with sufficient privileges can encode the destination class
directly with SO_PRIORITY, see
.BR socket(7).
.TP
with iptables/nftables
An iptables/nftables rule can be created to match traffic flows and
set the priority.
.BR iptables(8)
.TP
with net_prio cgroups
The net_prio cgroup can be used to set the priority of all sockets
belong to an application. See kernel and cgroup documentation for details.

.SH QDISC PARAMETERS
.TP
num_tc
Number of traffic classes to use. Up to 16 classes supported.

.TP
map
The priority to traffic class map. Maps priorities 0..15 to a specified
traffic class.

.TP
queues
Provide count and offset of queue range for each traffic class. In the
format,
.B count@offset.
Queue ranges for each traffic classes cannot overlap and must be a
contiguous range of queues.

.TP
hw
Set to
.B 1
to use hardware QOS defaults. Set to
.B 0
to override hardware defaults with user specified values.

.SH AUTHORS
John Fastabend, <john.r.fastabend@intel.com>