path: root/man/man7/packet.7

.\" SPDX-License-Identifier: Linux-man-pages-1-para
.\"
.\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
.\"
.\" $Id: packet.7,v 1.13 2000/08/14 08:03:45 ak Exp $
.\"
.TH packet 7 2024-05-02 "Linux man-pages (unreleased)"
.SH NAME
packet \- packet interface on device level
.SH SYNOPSIS
.nf
.B #include <sys/socket.h>
.B #include <linux/if_packet.h>
.B #include <net/ethernet.h>     /* the L2 protocols */
.P
.BI "packet_socket = socket(AF_PACKET, int " socket_type ", int "protocol );
.fi
.SH DESCRIPTION
Packet sockets are used to receive or send raw packets at the device driver
(OSI Layer 2) level.
They allow the user to implement protocol modules in user space
on top of the physical layer.
.P
The
.I socket_type
is either
.B SOCK_RAW
for raw packets including the link-level header or
.B SOCK_DGRAM
for cooked packets with the link-level header removed.
The link-level header information is available in a common format in a
.I sockaddr_ll
structure.
.I protocol
is the IEEE 802.3 protocol number in network byte order.
See the
.I <linux/if_ether.h>
include file for a list of allowed protocols.
When protocol
is set to
.BR htons(ETH_P_ALL) ,
then all protocols are received.
All incoming packets of that protocol type will be passed to the packet
socket before they are passed to the protocols implemented in the kernel.
If
.I protocol
is set to zero,
no packets are received.
.BR bind (2)
can optionally be called with a nonzero
.I sll_protocol
to start receiving packets for the protocols specified.
.P
In order to create a packet socket, a process must have the
.B CAP_NET_RAW
capability in the user namespace that governs its network namespace.
.P
.B SOCK_RAW
packets are passed to and from the device driver without any changes in
the packet data.
When receiving a packet, the address is still parsed and
passed in a standard
.I sockaddr_ll
address structure.
When transmitting a packet, the user-supplied buffer
should contain the physical-layer header.
That packet is then
queued unmodified to the network driver of the interface defined by the
destination address.
Some device drivers always add other headers.
.B SOCK_RAW
is similar to but not compatible with the obsolete
.B AF_INET/SOCK_PACKET
of Linux 2.0.
.P
.B SOCK_DGRAM
operates on a slightly higher level.
The physical header is removed before the packet is passed to the user.
Packets sent through a
.B SOCK_DGRAM
packet socket get a suitable physical-layer header based on the
information in the
.I sockaddr_ll
destination address before they are queued.
.P
By default, all packets of the specified protocol type
are passed to a packet socket.
To get packets only from a specific interface use
.BR bind (2)
specifying an address in a
.I struct sockaddr_ll
to bind the packet socket to an interface.
Fields used for binding are
.I sll_family
(should be
.BR AF_PACKET ),
.IR sll_protocol ,
and
.IR sll_ifindex .
.P
The
.BR connect (2)
operation is not supported on packet sockets.
.P
When the
.B MSG_TRUNC
flag is passed to
.BR recvmsg (2),
.BR recv (2),
or
.BR recvfrom (2),
the real length of the packet on the wire is always returned,
even when it is longer than the buffer.
.SS Address types
The
.I sockaddr_ll
structure is a device-independent physical-layer address.
.P
.in +4n
.EX
struct sockaddr_ll {
    unsigned short sll_family;   /* Always AF_PACKET */
    unsigned short sll_protocol; /* Physical\-layer protocol */
    int            sll_ifindex;  /* Interface number */
    unsigned short sll_hatype;   /* ARP hardware type */
    unsigned char  sll_pkttype;  /* Packet type */
    unsigned char  sll_halen;    /* Length of address */
    unsigned char  sll_addr[8];  /* Physical\-layer address */
};
.EE
.in
.P
The fields of this structure are as follows:
.TP
.I sll_protocol
is the standard ethernet protocol type in network byte order as defined
in the
.I <linux/if_ether.h>
include file.
It defaults to the socket's protocol.
.TP
.I sll_ifindex
is the interface index of the interface
(see
.BR netdevice (7));
0 matches any interface (only permitted for binding).
.I sll_hatype
is an ARP type as defined in the
.I <linux/if_arp.h>
include file.
.TP
.I sll_pkttype
contains the packet type.
Valid types are
.B PACKET_HOST
for a packet addressed to the local host,
.B PACKET_BROADCAST
for a physical-layer broadcast packet,
.B PACKET_MULTICAST
for a packet sent to a physical-layer multicast address,
.B PACKET_OTHERHOST
for a packet to some other host that has been caught by a device driver
in promiscuous mode, and
.B PACKET_OUTGOING
for a packet originating from the local host that is looped back to a packet
socket.
These types make sense only for receiving.
.TP
.I sll_addr
.TQ
.I sll_halen
contain the physical-layer (e.g., IEEE 802.3) address and its length.
The exact interpretation depends on the device.
.P
When you send packets, it is enough to specify
.IR sll_family ,
.IR sll_addr ,
.IR sll_halen ,
.IR sll_ifindex ,
and
.IR sll_protocol .
The other fields should be 0.
.I sll_hatype
and
.I sll_pkttype
are set on received packets for your information.
.SS Socket options
Packet socket options are configured by calling
.BR setsockopt (2)
with level
.BR SOL_PACKET .
.TP
.B PACKET_ADD_MEMBERSHIP
.PD 0
.TP
.B PACKET_DROP_MEMBERSHIP
.PD
Packet sockets can be used to configure physical-layer multicasting
and promiscuous mode.
.B PACKET_ADD_MEMBERSHIP
adds a binding and
.B PACKET_DROP_MEMBERSHIP
drops it.
They both expect a
.I packet_mreq
structure as argument:
.IP
.in +4n
.EX
struct packet_mreq {
    int            mr_ifindex;    /* interface index */
    unsigned short mr_type;       /* action */
    unsigned short mr_alen;       /* address length */
    unsigned char  mr_address[8]; /* physical\-layer address */
};
.EE
.in
.IP
.I mr_ifindex
contains the interface index for the interface whose status
should be changed.
The
.I mr_type
field specifies which action to perform.
.B PACKET_MR_PROMISC
enables receiving all packets on a shared medium (often known as
"promiscuous mode"),
.B PACKET_MR_MULTICAST
binds the socket to the physical-layer multicast group specified in
.I mr_address
and
.IR mr_alen ,
and
.B PACKET_MR_ALLMULTI
sets the socket up to receive all multicast packets arriving at
the interface.
.IP
In addition, the traditional ioctls
.BR SIOCSIFFLAGS ,
.BR SIOCADDMULTI ,
.B SIOCDELMULTI
can be used for the same purpose.
.TP
.BR PACKET_AUXDATA " (since Linux 2.6.21)"
.\" commit 8dc4194474159660d7f37c495e3fc3f10d0db8cc
If this binary option is enabled, the packet socket passes a metadata
structure along with each packet in the
.BR recvmsg (2)
control field.
The structure can be read with
.BR cmsg (3).
It is defined as
.IP
.in +4n
.EX
struct tpacket_auxdata {
    __u32 tp_status;
    __u32 tp_len;      /* packet length */
    __u32 tp_snaplen;  /* captured length */
    __u16 tp_mac;
    __u16 tp_net;
    __u16 tp_vlan_tci;
    __u16 tp_vlan_tpid; /* Since Linux 3.14; earlier, these
                           were unused padding bytes */
.\" commit a0cdfcf39362410d5ea983f4daf67b38de129408 added tp_vlan_tpid
};
.EE
.in
.TP
.BR PACKET_FANOUT " (since Linux 3.1)"
.\" commit dc99f600698dcac69b8f56dda9a8a00d645c5ffc
To scale processing across threads, packet sockets can form a fanout
group.
In this mode, each matching packet is enqueued onto only one
socket in the group.
A socket joins a fanout group by calling
.BR setsockopt (2)
with level
.B SOL_PACKET
and option
.BR PACKET_FANOUT .
Each network namespace can have up to 65536 independent groups.
A socket selects a group by encoding the ID in the first 16 bits of
the integer option value.
The first packet socket to join a group implicitly creates it.
To successfully join an existing group, subsequent packet sockets
must have the same protocol, device settings, fanout mode, and
flags (see below).
Packet sockets can leave a fanout group only by closing the socket.
The group is deleted when the last socket is closed.
.IP
Fanout supports multiple algorithms to spread traffic between sockets,
as follows:
.RS
.IP \[bu] 3
The default mode,
.BR PACKET_FANOUT_HASH ,
sends packets from the same flow to the same socket to maintain
per-flow ordering.
For each packet, it chooses a socket by taking the packet flow hash
modulo the number of sockets in the group, where a flow hash is a hash
over network-layer address and optional transport-layer port fields.
.IP \[bu]
The load-balance mode
.B PACKET_FANOUT_LB
implements a round-robin algorithm.
.IP \[bu]
.B PACKET_FANOUT_CPU
selects the socket based on the CPU that the packet arrived on.
.IP \[bu]
.B PACKET_FANOUT_ROLLOVER
processes all data on a single socket, moving to the next when one
becomes backlogged.
.IP \[bu]
.B PACKET_FANOUT_RND
selects the socket using a pseudo-random number generator.
.IP \[bu]
.B PACKET_FANOUT_QM
.\" commit 2d36097d26b5991d71a2cf4a20c1a158f0f1bfcd
(available since Linux 3.14)
selects the socket using the recorded queue_mapping of the received skb.
.RE
.IP
Fanout modes can take additional options.
IP fragmentation causes packets from the same flow to have different
flow hashes.
The flag
.BR PACKET_FANOUT_FLAG_DEFRAG ,
if set, causes packets to be defragmented before fanout is applied, to
preserve order even in this case.
Fanout mode and options are communicated in the second 16 bits of the
integer option value.
The flag
.B PACKET_FANOUT_FLAG_ROLLOVER
enables the roll over mechanism as a backup strategy: if the
original fanout algorithm selects a backlogged socket, the packet
rolls over to the next available one.
.TP
.BR PACKET_LOSS " (with " PACKET_TX_RING )
When a malformed packet is encountered on a transmit ring,
the default is to reset its
.I tp_status
to
.B TP_STATUS_WRONG_FORMAT
and abort the transmission immediately.
The malformed packet blocks itself and subsequently enqueued packets from
being sent.
The format error must be fixed, the associated
.I tp_status
reset to
.BR TP_STATUS_SEND_REQUEST ,
and the transmission process restarted via
.BR send (2).
However, if
.B PACKET_LOSS
is set, any malformed packet will be skipped, its
.I tp_status
reset to
.BR TP_STATUS_AVAILABLE ,
and the transmission process continued.
.TP
.BR PACKET_RESERVE " (with " PACKET_RX_RING )
By default, a packet receive ring writes packets immediately following the
metadata structure and alignment padding.
This integer option reserves additional headroom.
.TP
.B PACKET_RX_RING
Create a memory-mapped ring buffer for asynchronous packet reception.
The packet socket reserves a contiguous region of application address
space, lays it out into an array of packet slots and copies packets
(up to
.IR tp_snaplen )
into subsequent slots.
Each packet is preceded by a metadata structure similar to
.IR tpacket_auxdata .
The protocol fields encode the offset to the data
from the start of the metadata header.
.I tp_net
stores the offset to the network layer.
If the packet socket is of type
.BR SOCK_DGRAM ,
then
.I tp_mac
is the same.
If it is of type
.BR SOCK_RAW ,
then that field stores the offset to the link-layer frame.
Packet socket and application communicate the head and tail of the ring
through the
.I tp_status
field.
The packet socket owns all slots with
.I tp_status
equal to
.BR TP_STATUS_KERNEL .
After filling a slot, it changes the status of the slot to transfer
ownership to the application.
During normal operation, the new
.I tp_status
value has at least the
.B TP_STATUS_USER
bit set to signal that a received packet has been stored.
When the application has finished processing a packet, it transfers
ownership of the slot back to the socket by setting
.I tp_status
equal to
.BR TP_STATUS_KERNEL .
.IP
Packet sockets implement multiple variants of the packet ring.
The implementation details are described in
.I Documentation/networking/packet_mmap.rst
in the Linux kernel source tree.
.TP
.B PACKET_STATISTICS
Retrieve packet socket statistics in the form of a structure
.IP
.in +4n
.EX
struct tpacket_stats {
    unsigned int tp_packets;  /* Total packet count */
    unsigned int tp_drops;    /* Dropped packet count */
};
.EE
.in
.IP
Receiving statistics resets the internal counters.
The statistics structure differs when using a ring of variant
.BR TPACKET_V3 .
.TP
.BR PACKET_TIMESTAMP " (with " PACKET_RX_RING "; since Linux 2.6.36)"
.\" commit 614f60fa9d73a9e8fdff3df83381907fea7c5649
The packet receive ring always stores a timestamp in the metadata header.
By default, this is a software generated timestamp generated when the
packet is copied into the ring.
This integer option selects the type of timestamp.
Besides the default, it support the two hardware formats described in
.I Documentation/networking/timestamping.rst
in the Linux kernel source tree.
.TP
.BR PACKET_TX_RING " (since Linux 2.6.31)"
.\" commit 69e3c75f4d541a6eb151b3ef91f34033cb3ad6e1
Create a memory-mapped ring buffer for packet transmission.
This option is similar to
.B PACKET_RX_RING
and takes the same arguments.
The application writes packets into slots with
.I tp_status
equal to
.B TP_STATUS_AVAILABLE
and schedules them for transmission by changing
.I tp_status
to
.BR TP_STATUS_SEND_REQUEST .
When packets are ready to be transmitted, the application calls
.BR send (2)
or a variant thereof.
The
.I buf
and
.I len
fields of this call are ignored.
If an address is passed using
.BR sendto (2)
or
.BR sendmsg (2),
then that overrides the socket default.
On successful transmission, the socket resets
.I tp_status
to
.BR TP_STATUS_AVAILABLE .
It immediately aborts the transmission on error unless
.B PACKET_LOSS
is set.
.TP
.BR PACKET_VERSION " (with " PACKET_RX_RING "; since Linux 2.6.27)"
.\" commit bbd6ef87c544d88c30e4b762b1b61ef267a7d279
By default,
.B PACKET_RX_RING
creates a packet receive ring of variant
.BR TPACKET_V1 .
To create another variant, configure the desired variant by setting this
integer option before creating the ring.
.TP
.BR PACKET_QDISC_BYPASS " (since Linux 3.14)"
.\" commit d346a3fae3ff1d99f5d0c819bf86edf9094a26a1
By default, packets sent through packet sockets pass through the kernel's
qdisc (traffic control) layer, which is fine for the vast majority of use
cases.
For traffic generator appliances using packet sockets
that intend to brute-force flood the network\[em]for example,
to test devices under load in a similar
fashion to pktgen\[em]this layer can be bypassed by setting
this integer option to 1.
A side effect is that packet buffering in the qdisc layer is avoided,
which will lead to increased drops when network
device transmit queues are busy;
therefore, use at your own risk.
.SS Ioctls
.B SIOCGSTAMP
can be used to receive the timestamp of the last received packet.
Argument is a
.I struct timeval
variable.
.\" FIXME Document SIOCGSTAMPNS
.P
In addition, all standard ioctls defined in
.BR netdevice (7)
and
.BR socket (7)
are valid on packet sockets.
.SS Error handling
Packet sockets do no error handling other than errors occurred
while passing the packet to the device driver.
They don't have the concept of a pending error.
.SH ERRORS
.TP
.B EADDRNOTAVAIL
Unknown multicast group address passed.
.TP
.B EFAULT
User passed invalid memory address.
.TP
.B EINVAL
Invalid argument.
.TP
.B EMSGSIZE
Packet is bigger than interface MTU.
.TP
.B ENETDOWN
Interface is not up.
.TP
.B ENOBUFS
Not enough memory to allocate the packet.
.TP
.B ENODEV
Unknown device name or interface index specified in interface address.
.TP
.B ENOENT
No packet received.
.TP
.B ENOTCONN
No interface address passed.
.TP
.B ENXIO
Interface address contained an invalid interface index.
.TP
.B EPERM
User has insufficient privileges to carry out this operation.
.P
In addition, other errors may be generated by the low-level driver.
.SH VERSIONS
.B AF_PACKET
is a new feature in Linux 2.2.
Earlier Linux versions supported only
.BR SOCK_PACKET .
.SH NOTES
For portable programs it is suggested to use
.B AF_PACKET
via
.BR pcap (3);
although this covers only a subset of the
.B AF_PACKET
features.
.P
The
.B SOCK_DGRAM
packet sockets make no attempt to create or parse the IEEE 802.2 LLC
header for a IEEE 802.3 frame.
When
.B ETH_P_802_3
is specified as protocol for sending the kernel creates the
802.3 frame and fills out the length field; the user has to supply the LLC
header to get a fully conforming packet.
Incoming 802.3 packets are not multiplexed on the DSAP/SSAP protocol
fields; instead they are supplied to the user as protocol
.B ETH_P_802_2
with the LLC header prefixed.
It is thus not possible to bind to
.BR ETH_P_802_3 ;
bind to
.B ETH_P_802_2
instead and do the protocol multiplex yourself.
The default for sending is the standard Ethernet DIX
encapsulation with the protocol filled in.
.P
Packet sockets are not subject to the input or output firewall chains.
.SS Compatibility
In Linux 2.0, the only way to get a packet socket was with the call:
.P
.in +4n
.EX
socket(AF_INET, SOCK_PACKET, protocol)
.EE
.in
.P
This is still supported, but deprecated and strongly discouraged.
The main difference between the two methods is that
.B SOCK_PACKET
uses the old
.I struct sockaddr_pkt
to specify an interface, which doesn't provide physical-layer
independence.
.P
.in +4n
.EX
struct sockaddr_pkt {
    unsigned short spkt_family;
    unsigned char  spkt_device[14];
    unsigned short spkt_protocol;
};
.EE
.in
.P
.I spkt_family
contains
the device type,
.I spkt_protocol
is the IEEE 802.3 protocol type as defined in
.I <sys/if_ether.h>
and
.I spkt_device
is the device name as a null-terminated string, for example, eth0.
.P
This structure is obsolete and should not be used in new code.
.SH BUGS
.SS LLC header handling
The IEEE 802.2/803.3 LLC handling could be considered as a bug.
.SS MSG_TRUNC issues
The
.B MSG_TRUNC
.BR recvmsg (2)
extension is an ugly hack and should be replaced by a control message.
There is currently no way to get the original destination address of
packets via
.BR SOCK_DGRAM .
.SS spkt_device device name truncation
The
.I spkt_device
field of
.I sockaddr_pkt
has a size of 14 bytes,
which is less than the constant
.B IFNAMSIZ
defined in
.I <net/if.h>
which is 16 bytes and describes the system limit for a network interface name.
This means the names of network devices longer than 14 bytes
will be truncated to fit into
.IR spkt_device .
All these lengths include the terminating null byte (\[aq]\e0\[aq])).
.P
Issues from this with old code typically show up with
very long interface names used by the
.B Predictable Network Interface Names
feature enabled by default in many modern Linux distributions.
.P
The preferred solution is to rewrite code to avoid
.BR SOCK_PACKET .
Possible user solutions are to disable
.B Predictable Network Interface Names
or to rename the interface to a name of at most 13 bytes,
for example using the
.BR ip (8)
tool.
.SS Documentation issues
Socket filters are not documented.
.\" .SH CREDITS
.\" This man page was written by Andi Kleen with help from Matthew Wilcox.
.\" AF_PACKET in Linux 2.2 was implemented
.\" by Alexey Kuznetsov, based on code by Alan Cox and others.
.SH SEE ALSO
.BR socket (2),
.BR pcap (3),
.BR capabilities (7),
.BR ip (7),
.BR raw (7),
.BR socket (7),
.BR ip (8),
.P
RFC\ 894 for the standard IP Ethernet encapsulation.
RFC\ 1700 for the IEEE 802.3 IP encapsulation.
.P
The
.I <linux/if_ether.h>
include file for physical-layer protocols.
.P
The Linux kernel source tree.
.I Documentation/networking/filter.rst
describes how to apply Berkeley Packet Filters to packet sockets.
.I tools/testing/selftests/net/psock_tpacket.c
contains example source code for all available versions of
.B PACKET_RX_RING
and
.BR PACKET_TX_RING .