1Documentation for /proc/sys/net/* 2 (c) 1999 Terrehon Bowden <terrehon@pacbell.net> 3 Bodo Bauer <bb@ricochet.net> 4 (c) 2000 Jorge Nerin <comandante@zaralinux.com> 5 (c) 2009 Shen Feng <shen@cn.fujitsu.com> 6 7For general info and legal blurb, please look in README. 8 9============================================================== 10 11This file contains the documentation for the sysctl files in 12/proc/sys/net 13 14The interface to the networking parts of the kernel is located in 15/proc/sys/net. The following table shows all possible subdirectories. You may 16see only some of them, depending on your kernel's configuration. 17 18 19Table : Subdirectories in /proc/sys/net 20.............................................................................. 21 Directory Content Directory Content 22 core General parameter appletalk Appletalk protocol 23 unix Unix domain sockets netrom NET/ROM 24 802 E802 protocol ax25 AX25 25 ethernet Ethernet protocol rose X.25 PLP layer 26 ipv4 IP version 4 x25 X.25 protocol 27 ipx IPX token-ring IBM token ring 28 bridge Bridging decnet DEC net 29 ipv6 IP version 6 tipc TIPC 30.............................................................................. 31 321. /proc/sys/net/core - Network core options 33------------------------------------------------------- 34 35bpf_jit_enable 36-------------- 37 38This enables Berkeley Packet Filter Just in Time compiler. 39Currently supported on x86_64 architecture, bpf_jit provides a framework 40to speed packet filtering, the one used by tcpdump/libpcap for example. 41Values : 42 0 - disable the JIT (default value) 43 1 - enable the JIT 44 2 - enable the JIT and ask the compiler to emit traces on kernel log. 45 46dev_weight 47-------------- 48 49The maximum number of packets that kernel can handle on a NAPI interrupt, 50it's a Per-CPU variable. 51Default: 64 52 53default_qdisc 54-------------- 55 56The default queuing discipline to use for network devices. This allows 57overriding the default of pfifo_fast with an alternative. Since the default 58queuing discipline is created without additional parameters so is best suited 59to queuing disciplines that work well without configuration like stochastic 60fair queue (sfq), CoDel (codel) or fair queue CoDel (fq_codel). Don't use 61queuing disciplines like Hierarchical Token Bucket or Deficit Round Robin 62which require setting up classes and bandwidths. Note that physical multiqueue 63interfaces still use mq as root qdisc, which in turn uses this default for its 64leaves. Virtual devices (like e.g. lo or veth) ignore this setting and instead 65default to noqueue. 66Default: pfifo_fast 67 68busy_read 69---------------- 70Low latency busy poll timeout for socket reads. (needs CONFIG_NET_RX_BUSY_POLL) 71Approximate time in us to busy loop waiting for packets on the device queue. 72This sets the default value of the SO_BUSY_POLL socket option. 73Can be set or overridden per socket by setting socket option SO_BUSY_POLL, 74which is the preferred method of enabling. If you need to enable the feature 75globally via sysctl, a value of 50 is recommended. 76Will increase power usage. 77Default: 0 (off) 78 79busy_poll 80---------------- 81Low latency busy poll timeout for poll and select. (needs CONFIG_NET_RX_BUSY_POLL) 82Approximate time in us to busy loop waiting for events. 83Recommended value depends on the number of sockets you poll on. 84For several sockets 50, for several hundreds 100. 85For more than that you probably want to use epoll. 86Note that only sockets with SO_BUSY_POLL set will be busy polled, 87so you want to either selectively set SO_BUSY_POLL on those sockets or set 88sysctl.net.busy_read globally. 89Will increase power usage. 90Default: 0 (off) 91 92rmem_default 93------------ 94 95The default setting of the socket receive buffer in bytes. 96 97rmem_max 98-------- 99 100The maximum receive socket buffer size in bytes. 101 102tstamp_allow_data 103----------------- 104Allow processes to receive tx timestamps looped together with the original 105packet contents. If disabled, transmit timestamp requests from unprivileged 106processes are dropped unless socket option SOF_TIMESTAMPING_OPT_TSONLY is set. 107Default: 1 (on) 108 109 110wmem_default 111------------ 112 113The default setting (in bytes) of the socket send buffer. 114 115wmem_max 116-------- 117 118The maximum send socket buffer size in bytes. 119 120message_burst and message_cost 121------------------------------ 122 123These parameters are used to limit the warning messages written to the kernel 124log from the networking code. They enforce a rate limit to make a 125denial-of-service attack impossible. A higher message_cost factor, results in 126fewer messages that will be written. Message_burst controls when messages will 127be dropped. The default settings limit warning messages to one every five 128seconds. 129 130warnings 131-------- 132 133This sysctl is now unused. 134 135This was used to control console messages from the networking stack that 136occur because of problems on the network like duplicate address or bad 137checksums. 138 139These messages are now emitted at KERN_DEBUG and can generally be enabled 140and controlled by the dynamic_debug facility. 141 142netdev_budget 143------------- 144 145Maximum number of packets taken from all interfaces in one polling cycle (NAPI 146poll). In one polling cycle interfaces which are registered to polling are 147probed in a round-robin manner. 148 149netdev_max_backlog 150------------------ 151 152Maximum number of packets, queued on the INPUT side, when the interface 153receives packets faster than kernel can process them. 154 155netdev_rss_key 156-------------- 157 158RSS (Receive Side Scaling) enabled drivers use a 40 bytes host key that is 159randomly generated. 160Some user space might need to gather its content even if drivers do not 161provide ethtool -x support yet. 162 163myhost:~# cat /proc/sys/net/core/netdev_rss_key 16484:50:f4:00:a8:15:d1:a7:e9:7f:1d:60:35:c7:47:25:42:97:74:ca:56:bb:b6:a1:d8: ... (52 bytes total) 165 166File contains nul bytes if no driver ever called netdev_rss_key_fill() function. 167Note: 168/proc/sys/net/core/netdev_rss_key contains 52 bytes of key, 169but most drivers only use 40 bytes of it. 170 171myhost:~# ethtool -x eth0 172RX flow hash indirection table for eth0 with 8 RX ring(s): 173 0: 0 1 2 3 4 5 6 7 174RSS hash key: 17584:50:f4:00:a8:15:d1:a7:e9:7f:1d:60:35:c7:47:25:42:97:74:ca:56:bb:b6:a1:d8:43:e3:c9:0c:fd:17:55:c2:3a:4d:69:ed:f1:42:89 176 177netdev_tstamp_prequeue 178---------------------- 179 180If set to 0, RX packet timestamps can be sampled after RPS processing, when 181the target CPU processes packets. It might give some delay on timestamps, but 182permit to distribute the load on several cpus. 183 184If set to 1 (default), timestamps are sampled as soon as possible, before 185queueing. 186 187optmem_max 188---------- 189 190Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence 191of struct cmsghdr structures with appended data. 192 1932. /proc/sys/net/unix - Parameters for Unix domain sockets 194------------------------------------------------------- 195 196There is only one file in this directory. 197unix_dgram_qlen limits the max number of datagrams queued in Unix domain 198socket's buffer. It will not take effect unless PF_UNIX flag is specified. 199 200 2013. /proc/sys/net/ipv4 - IPV4 settings 202------------------------------------------------------- 203Please see: Documentation/networking/ip-sysctl.txt and ipvs-sysctl.txt for 204descriptions of these entries. 205 206 2074. Appletalk 208------------------------------------------------------- 209 210The /proc/sys/net/appletalk directory holds the Appletalk configuration data 211when Appletalk is loaded. The configurable parameters are: 212 213aarp-expiry-time 214---------------- 215 216The amount of time we keep an ARP entry before expiring it. Used to age out 217old hosts. 218 219aarp-resolve-time 220----------------- 221 222The amount of time we will spend trying to resolve an Appletalk address. 223 224aarp-retransmit-limit 225--------------------- 226 227The number of times we will retransmit a query before giving up. 228 229aarp-tick-time 230-------------- 231 232Controls the rate at which expires are checked. 233 234The directory /proc/net/appletalk holds the list of active Appletalk sockets 235on a machine. 236 237The fields indicate the DDP type, the local address (in network:node format) 238the remote address, the size of the transmit pending queue, the size of the 239received queue (bytes waiting for applications to read) the state and the uid 240owning the socket. 241 242/proc/net/atalk_iface lists all the interfaces configured for appletalk.It 243shows the name of the interface, its Appletalk address, the network range on 244that address (or network number for phase 1 networks), and the status of the 245interface. 246 247/proc/net/atalk_route lists each known network route. It lists the target 248(network) that the route leads to, the router (may be directly connected), the 249route flags, and the device the route is using. 250 251 2525. IPX 253------------------------------------------------------- 254 255The IPX protocol has no tunable values in proc/sys/net. 256 257The IPX protocol does, however, provide proc/net/ipx. This lists each IPX 258socket giving the local and remote addresses in Novell format (that is 259network:node:port). In accordance with the strange Novell tradition, 260everything but the port is in hex. Not_Connected is displayed for sockets that 261are not tied to a specific remote address. The Tx and Rx queue sizes indicate 262the number of bytes pending for transmission and reception. The state 263indicates the state the socket is in and the uid is the owning uid of the 264socket. 265 266The /proc/net/ipx_interface file lists all IPX interfaces. For each interface 267it gives the network number, the node number, and indicates if the network is 268the primary network. It also indicates which device it is bound to (or 269Internal for internal networks) and the Frame Type if appropriate. Linux 270supports 802.3, 802.2, 802.2 SNAP and DIX (Blue Book) ethernet framing for 271IPX. 272 273The /proc/net/ipx_route table holds a list of IPX routes. For each route it 274gives the destination network, the router node (or Directly) and the network 275address of the router (or Connected) for internal networks. 276 2776. TIPC 278------------------------------------------------------- 279 280tipc_rmem 281---------- 282 283The TIPC protocol now has a tunable for the receive memory, similar to the 284tcp_rmem - i.e. a vector of 3 INTEGERs: (min, default, max) 285 286 # cat /proc/sys/net/tipc/tipc_rmem 287 4252725 34021800 68043600 288 # 289 290The max value is set to CONN_OVERLOAD_LIMIT, and the default and min values 291are scaled (shifted) versions of that same value. Note that the min value 292is not at this point in time used in any meaningful way, but the triplet is 293preserved in order to be consistent with things like tcp_rmem. 294 295named_timeout 296-------------- 297 298TIPC name table updates are distributed asynchronously in a cluster, without 299any form of transaction handling. This means that different race scenarios are 300possible. One such is that a name withdrawal sent out by one node and received 301by another node may arrive after a second, overlapping name publication already 302has been accepted from a third node, although the conflicting updates 303originally may have been issued in the correct sequential order. 304If named_timeout is nonzero, failed topology updates will be placed on a defer 305queue until another event arrives that clears the error, or until the timeout 306expires. Value is in milliseconds. 307