1 Documentation 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>
7 For general info and legal blurb, please look in README.
9 ==============================================================
11 This file contains the documentation for the sysctl files in
14 The interface to the networking parts of the kernel is located in
15 /proc/sys/net. The following table shows all possible subdirectories. You may
16 see only some of them, depending on your kernel's configuration.
19 Table : 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 ..............................................................................
32 1. /proc/sys/net/core - Network core options
33 -------------------------------------------------------
38 This enables Berkeley Packet Filter Just in Time compiler.
39 Currently supported on x86_64 architecture, bpf_jit provides a framework
40 to speed packet filtering, the one used by tcpdump/libpcap for example.
42 0 - disable the JIT (default value)
44 2 - enable the JIT and ask the compiler to emit traces on kernel log.
49 This enables hardening for the Berkeley Packet Filter Just in Time compiler.
50 Supported are eBPF JIT backends. Enabling hardening trades off performance,
51 but can mitigate JIT spraying.
53 0 - disable JIT hardening (default value)
54 1 - enable JIT hardening for unprivileged users only
55 2 - enable JIT hardening for all users
60 When Berkeley Packet Filter Just in Time compiler is enabled, then compiled
61 images are unknown addresses to the kernel, meaning they neither show up in
62 traces nor in /proc/kallsyms. This enables export of these addresses, which
63 can be used for debugging/tracing. If bpf_jit_harden is enabled, this feature
66 0 - disable JIT kallsyms export (default value)
67 1 - enable JIT kallsyms export for privileged users only
72 The maximum number of packets that kernel can handle on a NAPI interrupt,
73 it's a Per-CPU variable.
79 RPS (e.g. RFS, aRFS) processing is competing with the registered NAPI poll function
80 of the driver for the per softirq cycle netdev_budget. This parameter influences
81 the proportion of the configured netdev_budget that is spent on RPS based packet
82 processing during RX softirq cycles. It is further meant for making current
83 dev_weight adaptable for asymmetric CPU needs on RX/TX side of the network stack.
84 (see dev_weight_tx_bias) It is effective on a per CPU basis. Determination is based
85 on dev_weight and is calculated multiplicative (dev_weight * dev_weight_rx_bias).
91 Scales the maximum number of packets that can be processed during a TX softirq cycle.
92 Effective on a per CPU basis. Allows scaling of current dev_weight for asymmetric
93 net stack processing needs. Be careful to avoid making TX softirq processing a CPU hog.
94 Calculation is based on dev_weight (dev_weight * dev_weight_tx_bias).
100 The default queuing discipline to use for network devices. This allows
101 overriding the default of pfifo_fast with an alternative. Since the default
102 queuing discipline is created without additional parameters so is best suited
103 to queuing disciplines that work well without configuration like stochastic
104 fair queue (sfq), CoDel (codel) or fair queue CoDel (fq_codel). Don't use
105 queuing disciplines like Hierarchical Token Bucket or Deficit Round Robin
106 which require setting up classes and bandwidths. Note that physical multiqueue
107 interfaces still use mq as root qdisc, which in turn uses this default for its
108 leaves. Virtual devices (like e.g. lo or veth) ignore this setting and instead
114 Low latency busy poll timeout for socket reads. (needs CONFIG_NET_RX_BUSY_POLL)
115 Approximate time in us to busy loop waiting for packets on the device queue.
116 This sets the default value of the SO_BUSY_POLL socket option.
117 Can be set or overridden per socket by setting socket option SO_BUSY_POLL,
118 which is the preferred method of enabling. If you need to enable the feature
119 globally via sysctl, a value of 50 is recommended.
120 Will increase power usage.
125 Low latency busy poll timeout for poll and select. (needs CONFIG_NET_RX_BUSY_POLL)
126 Approximate time in us to busy loop waiting for events.
127 Recommended value depends on the number of sockets you poll on.
128 For several sockets 50, for several hundreds 100.
129 For more than that you probably want to use epoll.
130 Note that only sockets with SO_BUSY_POLL set will be busy polled,
131 so you want to either selectively set SO_BUSY_POLL on those sockets or set
132 sysctl.net.busy_read globally.
133 Will increase power usage.
139 The default setting of the socket receive buffer in bytes.
144 The maximum receive socket buffer size in bytes.
148 Allow processes to receive tx timestamps looped together with the original
149 packet contents. If disabled, transmit timestamp requests from unprivileged
150 processes are dropped unless socket option SOF_TIMESTAMPING_OPT_TSONLY is set.
157 The default setting (in bytes) of the socket send buffer.
162 The maximum send socket buffer size in bytes.
164 message_burst and message_cost
165 ------------------------------
167 These parameters are used to limit the warning messages written to the kernel
168 log from the networking code. They enforce a rate limit to make a
169 denial-of-service attack impossible. A higher message_cost factor, results in
170 fewer messages that will be written. Message_burst controls when messages will
171 be dropped. The default settings limit warning messages to one every five
177 This sysctl is now unused.
179 This was used to control console messages from the networking stack that
180 occur because of problems on the network like duplicate address or bad
183 These messages are now emitted at KERN_DEBUG and can generally be enabled
184 and controlled by the dynamic_debug facility.
189 Maximum number of packets taken from all interfaces in one polling cycle (NAPI
190 poll). In one polling cycle interfaces which are registered to polling are
191 probed in a round-robin manner. Also, a polling cycle may not exceed
192 netdev_budget_usecs microseconds, even if netdev_budget has not been
196 ---------------------
198 Maximum number of microseconds in one NAPI polling cycle. Polling
199 will exit when either netdev_budget_usecs have elapsed during the
200 poll cycle or the number of packets processed reaches netdev_budget.
205 Maximum number of packets, queued on the INPUT side, when the interface
206 receives packets faster than kernel can process them.
211 RSS (Receive Side Scaling) enabled drivers use a 40 bytes host key that is
213 Some user space might need to gather its content even if drivers do not
214 provide ethtool -x support yet.
216 myhost:~# cat /proc/sys/net/core/netdev_rss_key
217 84:50: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)
219 File contains nul bytes if no driver ever called netdev_rss_key_fill() function.
221 /proc/sys/net/core/netdev_rss_key contains 52 bytes of key,
222 but most drivers only use 40 bytes of it.
224 myhost:~# ethtool -x eth0
225 RX flow hash indirection table for eth0 with 8 RX ring(s):
228 84:50: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
230 netdev_tstamp_prequeue
231 ----------------------
233 If set to 0, RX packet timestamps can be sampled after RPS processing, when
234 the target CPU processes packets. It might give some delay on timestamps, but
235 permit to distribute the load on several cpus.
237 If set to 1 (default), timestamps are sampled as soon as possible, before
243 Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence
244 of struct cmsghdr structures with appended data.
246 2. /proc/sys/net/unix - Parameters for Unix domain sockets
247 -------------------------------------------------------
249 There is only one file in this directory.
250 unix_dgram_qlen limits the max number of datagrams queued in Unix domain
251 socket's buffer. It will not take effect unless PF_UNIX flag is specified.
254 3. /proc/sys/net/ipv4 - IPV4 settings
255 -------------------------------------------------------
256 Please see: Documentation/networking/ip-sysctl.txt and ipvs-sysctl.txt for
257 descriptions of these entries.
261 -------------------------------------------------------
263 The /proc/sys/net/appletalk directory holds the Appletalk configuration data
264 when Appletalk is loaded. The configurable parameters are:
269 The amount of time we keep an ARP entry before expiring it. Used to age out
275 The amount of time we will spend trying to resolve an Appletalk address.
277 aarp-retransmit-limit
278 ---------------------
280 The number of times we will retransmit a query before giving up.
285 Controls the rate at which expires are checked.
287 The directory /proc/net/appletalk holds the list of active Appletalk sockets
290 The fields indicate the DDP type, the local address (in network:node format)
291 the remote address, the size of the transmit pending queue, the size of the
292 received queue (bytes waiting for applications to read) the state and the uid
295 /proc/net/atalk_iface lists all the interfaces configured for appletalk.It
296 shows the name of the interface, its Appletalk address, the network range on
297 that address (or network number for phase 1 networks), and the status of the
300 /proc/net/atalk_route lists each known network route. It lists the target
301 (network) that the route leads to, the router (may be directly connected), the
302 route flags, and the device the route is using.
306 -------------------------------------------------------
308 The IPX protocol has no tunable values in proc/sys/net.
310 The IPX protocol does, however, provide proc/net/ipx. This lists each IPX
311 socket giving the local and remote addresses in Novell format (that is
312 network:node:port). In accordance with the strange Novell tradition,
313 everything but the port is in hex. Not_Connected is displayed for sockets that
314 are not tied to a specific remote address. The Tx and Rx queue sizes indicate
315 the number of bytes pending for transmission and reception. The state
316 indicates the state the socket is in and the uid is the owning uid of the
319 The /proc/net/ipx_interface file lists all IPX interfaces. For each interface
320 it gives the network number, the node number, and indicates if the network is
321 the primary network. It also indicates which device it is bound to (or
322 Internal for internal networks) and the Frame Type if appropriate. Linux
323 supports 802.3, 802.2, 802.2 SNAP and DIX (Blue Book) ethernet framing for
326 The /proc/net/ipx_route table holds a list of IPX routes. For each route it
327 gives the destination network, the router node (or Directly) and the network
328 address of the router (or Connected) for internal networks.
331 -------------------------------------------------------
336 The TIPC protocol now has a tunable for the receive memory, similar to the
337 tcp_rmem - i.e. a vector of 3 INTEGERs: (min, default, max)
339 # cat /proc/sys/net/tipc/tipc_rmem
340 4252725 34021800 68043600
343 The max value is set to CONN_OVERLOAD_LIMIT, and the default and min values
344 are scaled (shifted) versions of that same value. Note that the min value
345 is not at this point in time used in any meaningful way, but the triplet is
346 preserved in order to be consistent with things like tcp_rmem.
351 TIPC name table updates are distributed asynchronously in a cluster, without
352 any form of transaction handling. This means that different race scenarios are
353 possible. One such is that a name withdrawal sent out by one node and received
354 by another node may arrive after a second, overlapping name publication already
355 has been accepted from a third node, although the conflicting updates
356 originally may have been issued in the correct sequential order.
357 If named_timeout is nonzero, failed topology updates will be placed on a defer
358 queue until another event arrives that clears the error, or until the timeout
359 expires. Value is in milliseconds.