1 ================================
2 Documentation for /proc/sys/net/
3 ================================
9 - Terrehon Bowden <terrehon@pacbell.net>
10 - Bodo Bauer <bb@ricochet.net>
14 - Jorge Nerin <comandante@zaralinux.com>
18 - Shen Feng <shen@cn.fujitsu.com>
20 For general info and legal blurb, please look in index.rst.
22 ------------------------------------------------------------------------------
24 This file contains the documentation for the sysctl files in
27 The interface to the networking parts of the kernel is located in
28 /proc/sys/net. The following table shows all possible subdirectories. You may
29 see only some of them, depending on your kernel's configuration.
32 Table : Subdirectories in /proc/sys/net
34 ========= =================== = ========== ===================
35 Directory Content Directory Content
36 ========= =================== = ========== ===================
37 802 E802 protocol mptcp Multipath TCP
38 appletalk Appletalk protocol netfilter Network Filter
39 ax25 AX25 netrom NET/ROM
40 bridge Bridging rose X.25 PLP layer
41 core General parameter tipc TIPC
42 ethernet Ethernet protocol unix Unix domain sockets
43 ipv4 IP version 4 x25 X.25 protocol
45 ========= =================== = ========== ===================
47 1. /proc/sys/net/core - Network core options
48 ============================================
53 This enables the BPF Just in Time (JIT) compiler. BPF is a flexible
54 and efficient infrastructure allowing to execute bytecode at various
55 hook points. It is used in a number of Linux kernel subsystems such
56 as networking (e.g. XDP, tc), tracing (e.g. kprobes, uprobes, tracepoints)
57 and security (e.g. seccomp). LLVM has a BPF back end that can compile
58 restricted C into a sequence of BPF instructions. After program load
59 through bpf(2) and passing a verifier in the kernel, a JIT will then
60 translate these BPF proglets into native CPU instructions. There are
61 two flavors of JITs, the newer eBPF JIT currently supported on:
75 And the older cBPF JIT supported on the following archs:
80 eBPF JITs are a superset of cBPF JITs, meaning the kernel will
81 migrate cBPF instructions into eBPF instructions and then JIT
82 compile them transparently. Older cBPF JITs can only translate
83 tcpdump filters, seccomp rules, etc, but not mentioned eBPF
84 programs loaded through bpf(2).
88 - 0 - disable the JIT (default value)
90 - 2 - enable the JIT and ask the compiler to emit traces on kernel log.
95 This enables hardening for the BPF JIT compiler. Supported are eBPF
96 JIT backends. Enabling hardening trades off performance, but can
97 mitigate JIT spraying.
101 - 0 - disable JIT hardening (default value)
102 - 1 - enable JIT hardening for unprivileged users only
103 - 2 - enable JIT hardening for all users
108 When BPF JIT compiler is enabled, then compiled images are unknown
109 addresses to the kernel, meaning they neither show up in traces nor
110 in /proc/kallsyms. This enables export of these addresses, which can
111 be used for debugging/tracing. If bpf_jit_harden is enabled, this
116 - 0 - disable JIT kallsyms export (default value)
117 - 1 - enable JIT kallsyms export for privileged users only
122 This enforces a global limit for memory allocations to the BPF JIT
123 compiler in order to reject unprivileged JIT requests once it has
124 been surpassed. bpf_jit_limit contains the value of the global limit
130 The maximum number of packets that kernel can handle on a NAPI interrupt,
131 it's a Per-CPU variable. For drivers that support LRO or GRO_HW, a hardware
132 aggregated packet is counted as one packet in this context.
139 RPS (e.g. RFS, aRFS) processing is competing with the registered NAPI poll function
140 of the driver for the per softirq cycle netdev_budget. This parameter influences
141 the proportion of the configured netdev_budget that is spent on RPS based packet
142 processing during RX softirq cycles. It is further meant for making current
143 dev_weight adaptable for asymmetric CPU needs on RX/TX side of the network stack.
144 (see dev_weight_tx_bias) It is effective on a per CPU basis. Determination is based
145 on dev_weight and is calculated multiplicative (dev_weight * dev_weight_rx_bias).
152 Scales the maximum number of packets that can be processed during a TX softirq cycle.
153 Effective on a per CPU basis. Allows scaling of current dev_weight for asymmetric
154 net stack processing needs. Be careful to avoid making TX softirq processing a CPU hog.
156 Calculation is based on dev_weight (dev_weight * dev_weight_tx_bias).
163 The default queuing discipline to use for network devices. This allows
164 overriding the default of pfifo_fast with an alternative. Since the default
165 queuing discipline is created without additional parameters so is best suited
166 to queuing disciplines that work well without configuration like stochastic
167 fair queue (sfq), CoDel (codel) or fair queue CoDel (fq_codel). Don't use
168 queuing disciplines like Hierarchical Token Bucket or Deficit Round Robin
169 which require setting up classes and bandwidths. Note that physical multiqueue
170 interfaces still use mq as root qdisc, which in turn uses this default for its
171 leaves. Virtual devices (like e.g. lo or veth) ignore this setting and instead
179 Low latency busy poll timeout for socket reads. (needs CONFIG_NET_RX_BUSY_POLL)
180 Approximate time in us to busy loop waiting for packets on the device queue.
181 This sets the default value of the SO_BUSY_POLL socket option.
182 Can be set or overridden per socket by setting socket option SO_BUSY_POLL,
183 which is the preferred method of enabling. If you need to enable the feature
184 globally via sysctl, a value of 50 is recommended.
186 Will increase power usage.
192 Low latency busy poll timeout for poll and select. (needs CONFIG_NET_RX_BUSY_POLL)
193 Approximate time in us to busy loop waiting for events.
194 Recommended value depends on the number of sockets you poll on.
195 For several sockets 50, for several hundreds 100.
196 For more than that you probably want to use epoll.
197 Note that only sockets with SO_BUSY_POLL set will be busy polled,
198 so you want to either selectively set SO_BUSY_POLL on those sockets or set
199 sysctl.net.busy_read globally.
201 Will increase power usage.
208 The default setting of the socket receive buffer in bytes.
213 The maximum receive socket buffer size in bytes.
217 Allow processes to receive tx timestamps looped together with the original
218 packet contents. If disabled, transmit timestamp requests from unprivileged
219 processes are dropped unless socket option SOF_TIMESTAMPING_OPT_TSONLY is set.
227 The default setting (in bytes) of the socket send buffer.
232 The maximum send socket buffer size in bytes.
234 message_burst and message_cost
235 ------------------------------
237 These parameters are used to limit the warning messages written to the kernel
238 log from the networking code. They enforce a rate limit to make a
239 denial-of-service attack impossible. A higher message_cost factor, results in
240 fewer messages that will be written. Message_burst controls when messages will
241 be dropped. The default settings limit warning messages to one every five
247 This sysctl is now unused.
249 This was used to control console messages from the networking stack that
250 occur because of problems on the network like duplicate address or bad
253 These messages are now emitted at KERN_DEBUG and can generally be enabled
254 and controlled by the dynamic_debug facility.
259 Maximum number of packets taken from all interfaces in one polling cycle (NAPI
260 poll). In one polling cycle interfaces which are registered to polling are
261 probed in a round-robin manner. Also, a polling cycle may not exceed
262 netdev_budget_usecs microseconds, even if netdev_budget has not been
266 ---------------------
268 Maximum number of microseconds in one NAPI polling cycle. Polling
269 will exit when either netdev_budget_usecs have elapsed during the
270 poll cycle or the number of packets processed reaches netdev_budget.
275 Maximum number of packets, queued on the INPUT side, when the interface
276 receives packets faster than kernel can process them.
281 RSS (Receive Side Scaling) enabled drivers use a 40 bytes host key that is
283 Some user space might need to gather its content even if drivers do not
284 provide ethtool -x support yet.
288 myhost:~# cat /proc/sys/net/core/netdev_rss_key
289 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)
291 File contains nul bytes if no driver ever called netdev_rss_key_fill() function.
294 /proc/sys/net/core/netdev_rss_key contains 52 bytes of key,
295 but most drivers only use 40 bytes of it.
299 myhost:~# ethtool -x eth0
300 RX flow hash indirection table for eth0 with 8 RX ring(s):
303 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
305 netdev_tstamp_prequeue
306 ----------------------
308 If set to 0, RX packet timestamps can be sampled after RPS processing, when
309 the target CPU processes packets. It might give some delay on timestamps, but
310 permit to distribute the load on several cpus.
312 If set to 1 (default), timestamps are sampled as soon as possible, before
315 netdev_unregister_timeout_secs
316 ------------------------------
318 Unregister network device timeout in seconds.
319 This option controls the timeout (in seconds) used to issue a warning while
320 waiting for a network device refcount to drop to 0 during device
321 unregistration. A lower value may be useful during bisection to detect
322 a leaked reference faster. A larger value may be useful to prevent false
323 warnings on slow/loaded systems.
324 Default value is 10, minimum 1, maximum 3600.
329 Max size (in skbs) of the per-cpu list of skbs being freed
330 by the cpu which allocated them. Used by TCP stack so far.
337 Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence
338 of struct cmsghdr structures with appended data.
340 fb_tunnels_only_for_init_net
341 ----------------------------
343 Controls if fallback tunnels (like tunl0, gre0, gretap0, erspan0,
344 sit0, ip6tnl0, ip6gre0) are automatically created. There are 3 possibilities
345 (a) value = 0; respective fallback tunnels are created when module is
346 loaded in every net namespaces (backward compatible behavior).
347 (b) value = 1; [kcmd value: initns] respective fallback tunnels are
348 created only in init net namespace and every other net namespace will
350 (c) value = 2; [kcmd value: none] fallback tunnels are not created
351 when a module is loaded in any of the net namespace. Setting value to
352 "2" is pointless after boot if these modules are built-in, so there is
353 a kernel command-line option that can change this default. Please refer to
354 Documentation/admin-guide/kernel-parameters.txt for additional details.
356 Not creating fallback tunnels gives control to userspace to create
357 whatever is needed only and avoid creating devices which are redundant.
359 Default : 0 (for compatibility reasons)
361 devconf_inherit_init_net
362 ------------------------
364 Controls if a new network namespace should inherit all current
365 settings under /proc/sys/net/{ipv4,ipv6}/conf/{all,default}/. By
366 default, we keep the current behavior: for IPv4 we inherit all current
367 settings from init_net and for IPv6 we reset all settings to default.
369 If set to 1, both IPv4 and IPv6 settings are forced to inherit from
370 current ones in init_net. If set to 2, both IPv4 and IPv6 settings are
371 forced to reset to their default values. If set to 3, both IPv4 and IPv6
372 settings are forced to inherit from current ones in the netns where this
373 new netns has been created.
375 Default : 0 (for compatibility reasons)
380 Controls default hash rethink behaviour on listening socket when SO_TXREHASH
381 option is set to SOCK_TXREHASH_DEFAULT (i. e. not overridden by setsockopt).
383 If set to 1 (default), hash rethink is performed on listening socket.
384 If set to 0, hash rethink is not performed.
389 Maximum number of the segments to batch up on output of GRO. When a packet
390 exits GRO, either as a coalesced superframe or as an original packet which
391 GRO has decided not to coalesce, it is placed on a per-NAPI list. This
392 list is then passed to the stack when the number of segments reaches the
393 gro_normal_batch limit.
395 high_order_alloc_disable
396 ------------------------
398 By default the allocator for page frags tries to use high order pages (order-3
399 on x86). While the default behavior gives good results in most cases, some users
400 might have hit a contention in page allocations/freeing. This was especially
401 true on older kernels (< 5.14) when high-order pages were not stored on per-cpu
402 lists. This allows to opt-in for order-0 allocation instead but is now mostly of
403 historical importance.
407 2. /proc/sys/net/unix - Parameters for Unix domain sockets
408 ----------------------------------------------------------
410 There is only one file in this directory.
411 unix_dgram_qlen limits the max number of datagrams queued in Unix domain
412 socket's buffer. It will not take effect unless PF_UNIX flag is specified.
415 3. /proc/sys/net/ipv4 - IPV4 settings
416 -------------------------------------
417 Please see: Documentation/networking/ip-sysctl.rst and
418 Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.
424 The /proc/sys/net/appletalk directory holds the Appletalk configuration data
425 when Appletalk is loaded. The configurable parameters are:
430 The amount of time we keep an ARP entry before expiring it. Used to age out
436 The amount of time we will spend trying to resolve an Appletalk address.
438 aarp-retransmit-limit
439 ---------------------
441 The number of times we will retransmit a query before giving up.
446 Controls the rate at which expires are checked.
448 The directory /proc/net/appletalk holds the list of active Appletalk sockets
451 The fields indicate the DDP type, the local address (in network:node format)
452 the remote address, the size of the transmit pending queue, the size of the
453 received queue (bytes waiting for applications to read) the state and the uid
456 /proc/net/atalk_iface lists all the interfaces configured for appletalk.It
457 shows the name of the interface, its Appletalk address, the network range on
458 that address (or network number for phase 1 networks), and the status of the
461 /proc/net/atalk_route lists each known network route. It lists the target
462 (network) that the route leads to, the router (may be directly connected), the
463 route flags, and the device the route is using.
471 The TIPC protocol now has a tunable for the receive memory, similar to the
472 tcp_rmem - i.e. a vector of 3 INTEGERs: (min, default, max)
476 # cat /proc/sys/net/tipc/tipc_rmem
477 4252725 34021800 68043600
480 The max value is set to CONN_OVERLOAD_LIMIT, and the default and min values
481 are scaled (shifted) versions of that same value. Note that the min value
482 is not at this point in time used in any meaningful way, but the triplet is
483 preserved in order to be consistent with things like tcp_rmem.
488 TIPC name table updates are distributed asynchronously in a cluster, without
489 any form of transaction handling. This means that different race scenarios are
490 possible. One such is that a name withdrawal sent out by one node and received
491 by another node may arrive after a second, overlapping name publication already
492 has been accepted from a third node, although the conflicting updates
493 originally may have been issued in the correct sequential order.
494 If named_timeout is nonzero, failed topology updates will be placed on a defer
495 queue until another event arrives that clears the error, or until the timeout
496 expires. Value is in milliseconds.
projects / platform / kernel / linux-rpi.git / blob