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 the BPF Just in Time (JIT) compiler. BPF is a flexible
39 and efficient infrastructure allowing to execute bytecode at various
40 hook points. It is used in a number of Linux kernel subsystems such
41 as networking (e.g. XDP, tc), tracing (e.g. kprobes, uprobes, tracepoints)
42 and security (e.g. seccomp). LLVM has a BPF back end that can compile
43 restricted C into a sequence of BPF instructions. After program load
44 through bpf(2) and passing a verifier in the kernel, a JIT will then
45 translate these BPF proglets into native CPU instructions. There are
46 two flavors of JITs, the newer eBPF JIT currently supported on:
55 And the older cBPF JIT supported on the following archs:
60 eBPF JITs are a superset of cBPF JITs, meaning the kernel will
61 migrate cBPF instructions into eBPF instructions and then JIT
62 compile them transparently. Older cBPF JITs can only translate
63 tcpdump filters, seccomp rules, etc, but not mentioned eBPF
64 programs loaded through bpf(2).
67 0 - disable the JIT (default value)
69 2 - enable the JIT and ask the compiler to emit traces on kernel log.
74 This enables hardening for the BPF JIT compiler. Supported are eBPF
75 JIT backends. Enabling hardening trades off performance, but can
76 mitigate JIT spraying.
78 0 - disable JIT hardening (default value)
79 1 - enable JIT hardening for unprivileged users only
80 2 - enable JIT hardening for all users
85 When BPF JIT compiler is enabled, then compiled images are unknown
86 addresses to the kernel, meaning they neither show up in traces nor
87 in /proc/kallsyms. This enables export of these addresses, which can
88 be used for debugging/tracing. If bpf_jit_harden is enabled, this
91 0 - disable JIT kallsyms export (default value)
92 1 - enable JIT kallsyms export for privileged users only
97 The maximum number of packets that kernel can handle on a NAPI interrupt,
98 it's a Per-CPU variable.
104 RPS (e.g. RFS, aRFS) processing is competing with the registered NAPI poll function
105 of the driver for the per softirq cycle netdev_budget. This parameter influences
106 the proportion of the configured netdev_budget that is spent on RPS based packet
107 processing during RX softirq cycles. It is further meant for making current
108 dev_weight adaptable for asymmetric CPU needs on RX/TX side of the network stack.
109 (see dev_weight_tx_bias) It is effective on a per CPU basis. Determination is based
110 on dev_weight and is calculated multiplicative (dev_weight * dev_weight_rx_bias).
116 Scales the maximum number of packets that can be processed during a TX softirq cycle.
117 Effective on a per CPU basis. Allows scaling of current dev_weight for asymmetric
118 net stack processing needs. Be careful to avoid making TX softirq processing a CPU hog.
119 Calculation is based on dev_weight (dev_weight * dev_weight_tx_bias).
125 The default queuing discipline to use for network devices. This allows
126 overriding the default of pfifo_fast with an alternative. Since the default
127 queuing discipline is created without additional parameters so is best suited
128 to queuing disciplines that work well without configuration like stochastic
129 fair queue (sfq), CoDel (codel) or fair queue CoDel (fq_codel). Don't use
130 queuing disciplines like Hierarchical Token Bucket or Deficit Round Robin
131 which require setting up classes and bandwidths. Note that physical multiqueue
132 interfaces still use mq as root qdisc, which in turn uses this default for its
133 leaves. Virtual devices (like e.g. lo or veth) ignore this setting and instead
139 Low latency busy poll timeout for socket reads. (needs CONFIG_NET_RX_BUSY_POLL)
140 Approximate time in us to busy loop waiting for packets on the device queue.
141 This sets the default value of the SO_BUSY_POLL socket option.
142 Can be set or overridden per socket by setting socket option SO_BUSY_POLL,
143 which is the preferred method of enabling. If you need to enable the feature
144 globally via sysctl, a value of 50 is recommended.
145 Will increase power usage.
150 Low latency busy poll timeout for poll and select. (needs CONFIG_NET_RX_BUSY_POLL)
151 Approximate time in us to busy loop waiting for events.
152 Recommended value depends on the number of sockets you poll on.
153 For several sockets 50, for several hundreds 100.
154 For more than that you probably want to use epoll.
155 Note that only sockets with SO_BUSY_POLL set will be busy polled,
156 so you want to either selectively set SO_BUSY_POLL on those sockets or set
157 sysctl.net.busy_read globally.
158 Will increase power usage.
164 The default setting of the socket receive buffer in bytes.
169 The maximum receive socket buffer size in bytes.
173 Allow processes to receive tx timestamps looped together with the original
174 packet contents. If disabled, transmit timestamp requests from unprivileged
175 processes are dropped unless socket option SOF_TIMESTAMPING_OPT_TSONLY is set.
182 The default setting (in bytes) of the socket send buffer.
187 The maximum send socket buffer size in bytes.
189 message_burst and message_cost
190 ------------------------------
192 These parameters are used to limit the warning messages written to the kernel
193 log from the networking code. They enforce a rate limit to make a
194 denial-of-service attack impossible. A higher message_cost factor, results in
195 fewer messages that will be written. Message_burst controls when messages will
196 be dropped. The default settings limit warning messages to one every five
202 This sysctl is now unused.
204 This was used to control console messages from the networking stack that
205 occur because of problems on the network like duplicate address or bad
208 These messages are now emitted at KERN_DEBUG and can generally be enabled
209 and controlled by the dynamic_debug facility.
214 Maximum number of packets taken from all interfaces in one polling cycle (NAPI
215 poll). In one polling cycle interfaces which are registered to polling are
216 probed in a round-robin manner. Also, a polling cycle may not exceed
217 netdev_budget_usecs microseconds, even if netdev_budget has not been
221 ---------------------
223 Maximum number of microseconds in one NAPI polling cycle. Polling
224 will exit when either netdev_budget_usecs have elapsed during the
225 poll cycle or the number of packets processed reaches netdev_budget.
230 Maximum number of packets, queued on the INPUT side, when the interface
231 receives packets faster than kernel can process them.
236 RSS (Receive Side Scaling) enabled drivers use a 40 bytes host key that is
238 Some user space might need to gather its content even if drivers do not
239 provide ethtool -x support yet.
241 myhost:~# cat /proc/sys/net/core/netdev_rss_key
242 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)
244 File contains nul bytes if no driver ever called netdev_rss_key_fill() function.
246 /proc/sys/net/core/netdev_rss_key contains 52 bytes of key,
247 but most drivers only use 40 bytes of it.
249 myhost:~# ethtool -x eth0
250 RX flow hash indirection table for eth0 with 8 RX ring(s):
253 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
255 netdev_tstamp_prequeue
256 ----------------------
258 If set to 0, RX packet timestamps can be sampled after RPS processing, when
259 the target CPU processes packets. It might give some delay on timestamps, but
260 permit to distribute the load on several cpus.
262 If set to 1 (default), timestamps are sampled as soon as possible, before
268 Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence
269 of struct cmsghdr structures with appended data.
271 2. /proc/sys/net/unix - Parameters for Unix domain sockets
272 -------------------------------------------------------
274 There is only one file in this directory.
275 unix_dgram_qlen limits the max number of datagrams queued in Unix domain
276 socket's buffer. It will not take effect unless PF_UNIX flag is specified.
279 3. /proc/sys/net/ipv4 - IPV4 settings
280 -------------------------------------------------------
281 Please see: Documentation/networking/ip-sysctl.txt and ipvs-sysctl.txt for
282 descriptions of these entries.
286 -------------------------------------------------------
288 The /proc/sys/net/appletalk directory holds the Appletalk configuration data
289 when Appletalk is loaded. The configurable parameters are:
294 The amount of time we keep an ARP entry before expiring it. Used to age out
300 The amount of time we will spend trying to resolve an Appletalk address.
302 aarp-retransmit-limit
303 ---------------------
305 The number of times we will retransmit a query before giving up.
310 Controls the rate at which expires are checked.
312 The directory /proc/net/appletalk holds the list of active Appletalk sockets
315 The fields indicate the DDP type, the local address (in network:node format)
316 the remote address, the size of the transmit pending queue, the size of the
317 received queue (bytes waiting for applications to read) the state and the uid
320 /proc/net/atalk_iface lists all the interfaces configured for appletalk.It
321 shows the name of the interface, its Appletalk address, the network range on
322 that address (or network number for phase 1 networks), and the status of the
325 /proc/net/atalk_route lists each known network route. It lists the target
326 (network) that the route leads to, the router (may be directly connected), the
327 route flags, and the device the route is using.
331 -------------------------------------------------------
333 The IPX protocol has no tunable values in proc/sys/net.
335 The IPX protocol does, however, provide proc/net/ipx. This lists each IPX
336 socket giving the local and remote addresses in Novell format (that is
337 network:node:port). In accordance with the strange Novell tradition,
338 everything but the port is in hex. Not_Connected is displayed for sockets that
339 are not tied to a specific remote address. The Tx and Rx queue sizes indicate
340 the number of bytes pending for transmission and reception. The state
341 indicates the state the socket is in and the uid is the owning uid of the
344 The /proc/net/ipx_interface file lists all IPX interfaces. For each interface
345 it gives the network number, the node number, and indicates if the network is
346 the primary network. It also indicates which device it is bound to (or
347 Internal for internal networks) and the Frame Type if appropriate. Linux
348 supports 802.3, 802.2, 802.2 SNAP and DIX (Blue Book) ethernet framing for
351 The /proc/net/ipx_route table holds a list of IPX routes. For each route it
352 gives the destination network, the router node (or Directly) and the network
353 address of the router (or Connected) for internal networks.
356 -------------------------------------------------------
361 The TIPC protocol now has a tunable for the receive memory, similar to the
362 tcp_rmem - i.e. a vector of 3 INTEGERs: (min, default, max)
364 # cat /proc/sys/net/tipc/tipc_rmem
365 4252725 34021800 68043600
368 The max value is set to CONN_OVERLOAD_LIMIT, and the default and min values
369 are scaled (shifted) versions of that same value. Note that the min value
370 is not at this point in time used in any meaningful way, but the triplet is
371 preserved in order to be consistent with things like tcp_rmem.
376 TIPC name table updates are distributed asynchronously in a cluster, without
377 any form of transaction handling. This means that different race scenarios are
378 possible. One such is that a name withdrawal sent out by one node and received
379 by another node may arrive after a second, overlapping name publication already
380 has been accepted from a third node, although the conflicting updates
381 originally may have been issued in the correct sequential order.
382 If named_timeout is nonzero, failed topology updates will be placed on a defer
383 queue until another event arrives that clears the error, or until the timeout
384 expires. Value is in milliseconds.