516 lines · plain
1================================2Documentation for /proc/sys/net/3================================4 5Copyright6 7Copyright (c) 19998 9 - Terrehon Bowden <terrehon@pacbell.net>10 - Bodo Bauer <bb@ricochet.net>11 12Copyright (c) 200013 14 - Jorge Nerin <comandante@zaralinux.com>15 16Copyright (c) 200917 18 - Shen Feng <shen@cn.fujitsu.com>19 20For general info and legal blurb, please look in index.rst.21 22------------------------------------------------------------------------------23 24This file contains the documentation for the sysctl files in25/proc/sys/net26 27The interface to the networking parts of the kernel is located in28/proc/sys/net. The following table shows all possible subdirectories. You may29see only some of them, depending on your kernel's configuration.30 31 32Table : Subdirectories in /proc/sys/net33 34 ========= =================== = ========== ===================35 Directory Content Directory Content36 ========= =================== = ========== ===================37 802 E802 protocol mptcp Multipath TCP38 appletalk Appletalk protocol netfilter Network Filter39 ax25 AX25 netrom NET/ROM40 bridge Bridging rose X.25 PLP layer41 core General parameter tipc TIPC42 ethernet Ethernet protocol unix Unix domain sockets43 ipv4 IP version 4 x25 X.25 protocol44 ipv6 IP version 645 ========= =================== = ========== ===================46 471. /proc/sys/net/core - Network core options48============================================49 50bpf_jit_enable51--------------52 53This enables the BPF Just in Time (JIT) compiler. BPF is a flexible54and efficient infrastructure allowing to execute bytecode at various55hook points. It is used in a number of Linux kernel subsystems such56as networking (e.g. XDP, tc), tracing (e.g. kprobes, uprobes, tracepoints)57and security (e.g. seccomp). LLVM has a BPF back end that can compile58restricted C into a sequence of BPF instructions. After program load59through bpf(2) and passing a verifier in the kernel, a JIT will then60translate these BPF proglets into native CPU instructions. There are61two flavors of JITs, the newer eBPF JIT currently supported on:62 63 - x86_6464 - x86_3265 - arm6466 - arm3267 - ppc6468 - ppc3269 - sparc6470 - mips6471 - s390x72 - riscv6473 - riscv3274 - loongarch6475 - arc76 77And the older cBPF JIT supported on the following archs:78 79 - mips80 - sparc81 82eBPF JITs are a superset of cBPF JITs, meaning the kernel will83migrate cBPF instructions into eBPF instructions and then JIT84compile them transparently. Older cBPF JITs can only translate85tcpdump filters, seccomp rules, etc, but not mentioned eBPF86programs loaded through bpf(2).87 88Values:89 90 - 0 - disable the JIT (default value)91 - 1 - enable the JIT92 - 2 - enable the JIT and ask the compiler to emit traces on kernel log.93 94bpf_jit_harden95--------------96 97This enables hardening for the BPF JIT compiler. Supported are eBPF98JIT backends. Enabling hardening trades off performance, but can99mitigate JIT spraying.100 101Values:102 103 - 0 - disable JIT hardening (default value)104 - 1 - enable JIT hardening for unprivileged users only105 - 2 - enable JIT hardening for all users106 107where "privileged user" in this context means a process having108CAP_BPF or CAP_SYS_ADMIN in the root user name space.109 110bpf_jit_kallsyms111----------------112 113When BPF JIT compiler is enabled, then compiled images are unknown114addresses to the kernel, meaning they neither show up in traces nor115in /proc/kallsyms. This enables export of these addresses, which can116be used for debugging/tracing. If bpf_jit_harden is enabled, this117feature is disabled.118 119Values :120 121 - 0 - disable JIT kallsyms export (default value)122 - 1 - enable JIT kallsyms export for privileged users only123 124bpf_jit_limit125-------------126 127This enforces a global limit for memory allocations to the BPF JIT128compiler in order to reject unprivileged JIT requests once it has129been surpassed. bpf_jit_limit contains the value of the global limit130in bytes.131 132dev_weight133----------134 135The maximum number of packets that kernel can handle on a NAPI interrupt,136it's a Per-CPU variable. For drivers that support LRO or GRO_HW, a hardware137aggregated packet is counted as one packet in this context.138 139Default: 64140 141dev_weight_rx_bias142------------------143 144RPS (e.g. RFS, aRFS) processing is competing with the registered NAPI poll function145of the driver for the per softirq cycle netdev_budget. This parameter influences146the proportion of the configured netdev_budget that is spent on RPS based packet147processing during RX softirq cycles. It is further meant for making current148dev_weight adaptable for asymmetric CPU needs on RX/TX side of the network stack.149(see dev_weight_tx_bias) It is effective on a per CPU basis. Determination is based150on dev_weight and is calculated multiplicative (dev_weight * dev_weight_rx_bias).151 152Default: 1153 154dev_weight_tx_bias155------------------156 157Scales the maximum number of packets that can be processed during a TX softirq cycle.158Effective on a per CPU basis. Allows scaling of current dev_weight for asymmetric159net stack processing needs. Be careful to avoid making TX softirq processing a CPU hog.160 161Calculation is based on dev_weight (dev_weight * dev_weight_tx_bias).162 163Default: 1164 165default_qdisc166-------------167 168The default queuing discipline to use for network devices. This allows169overriding the default of pfifo_fast with an alternative. Since the default170queuing discipline is created without additional parameters so is best suited171to queuing disciplines that work well without configuration like stochastic172fair queue (sfq), CoDel (codel) or fair queue CoDel (fq_codel). Don't use173queuing disciplines like Hierarchical Token Bucket or Deficit Round Robin174which require setting up classes and bandwidths. Note that physical multiqueue175interfaces still use mq as root qdisc, which in turn uses this default for its176leaves. Virtual devices (like e.g. lo or veth) ignore this setting and instead177default to noqueue.178 179Default: pfifo_fast180 181busy_read182---------183 184Low latency busy poll timeout for socket reads. (needs CONFIG_NET_RX_BUSY_POLL)185Approximate time in us to busy loop waiting for packets on the device queue.186This sets the default value of the SO_BUSY_POLL socket option.187Can be set or overridden per socket by setting socket option SO_BUSY_POLL,188which is the preferred method of enabling. If you need to enable the feature189globally via sysctl, a value of 50 is recommended.190 191Will increase power usage.192 193Default: 0 (off)194 195busy_poll196----------------197Low latency busy poll timeout for poll and select. (needs CONFIG_NET_RX_BUSY_POLL)198Approximate time in us to busy loop waiting for events.199Recommended value depends on the number of sockets you poll on.200For several sockets 50, for several hundreds 100.201For more than that you probably want to use epoll.202Note that only sockets with SO_BUSY_POLL set will be busy polled,203so you want to either selectively set SO_BUSY_POLL on those sockets or set204sysctl.net.busy_read globally.205 206Will increase power usage.207 208Default: 0 (off)209 210mem_pcpu_rsv211------------212 213Per-cpu reserved forward alloc cache size in page units. Default 1MB per CPU.214 215rmem_default216------------217 218The default setting of the socket receive buffer in bytes.219 220rmem_max221--------222 223The maximum receive socket buffer size in bytes.224 225rps_default_mask226----------------227 228The default RPS CPU mask used on newly created network devices. An empty229mask means RPS disabled by default.230 231tstamp_allow_data232-----------------233Allow processes to receive tx timestamps looped together with the original234packet contents. If disabled, transmit timestamp requests from unprivileged235processes are dropped unless socket option SOF_TIMESTAMPING_OPT_TSONLY is set.236 237Default: 1 (on)238 239 240wmem_default241------------242 243The default setting (in bytes) of the socket send buffer.244 245wmem_max246--------247 248The maximum send socket buffer size in bytes.249 250message_burst and message_cost251------------------------------252 253These parameters are used to limit the warning messages written to the kernel254log from the networking code. They enforce a rate limit to make a255denial-of-service attack impossible. A higher message_cost factor, results in256fewer messages that will be written. Message_burst controls when messages will257be dropped. The default settings limit warning messages to one every five258seconds.259 260warnings261--------262 263This sysctl is now unused.264 265This was used to control console messages from the networking stack that266occur because of problems on the network like duplicate address or bad267checksums.268 269These messages are now emitted at KERN_DEBUG and can generally be enabled270and controlled by the dynamic_debug facility.271 272netdev_budget273-------------274 275Maximum number of packets taken from all interfaces in one polling cycle (NAPI276poll). In one polling cycle interfaces which are registered to polling are277probed in a round-robin manner. Also, a polling cycle may not exceed278netdev_budget_usecs microseconds, even if netdev_budget has not been279exhausted.280 281netdev_budget_usecs282---------------------283 284Maximum number of microseconds in one NAPI polling cycle. Polling285will exit when either netdev_budget_usecs have elapsed during the286poll cycle or the number of packets processed reaches netdev_budget.287 288netdev_max_backlog289------------------290 291Maximum number of packets, queued on the INPUT side, when the interface292receives packets faster than kernel can process them.293 294netdev_rss_key295--------------296 297RSS (Receive Side Scaling) enabled drivers use a 40 bytes host key that is298randomly generated.299Some user space might need to gather its content even if drivers do not300provide ethtool -x support yet.301 302::303 304 myhost:~# cat /proc/sys/net/core/netdev_rss_key305 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)306 307File contains nul bytes if no driver ever called netdev_rss_key_fill() function.308 309Note:310 /proc/sys/net/core/netdev_rss_key contains 52 bytes of key,311 but most drivers only use 40 bytes of it.312 313::314 315 myhost:~# ethtool -x eth0316 RX flow hash indirection table for eth0 with 8 RX ring(s):317 0: 0 1 2 3 4 5 6 7318 RSS hash key:319 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:f1:42:89320 321netdev_tstamp_prequeue322----------------------323 324If set to 0, RX packet timestamps can be sampled after RPS processing, when325the target CPU processes packets. It might give some delay on timestamps, but326permit to distribute the load on several cpus.327 328If set to 1 (default), timestamps are sampled as soon as possible, before329queueing.330 331netdev_unregister_timeout_secs332------------------------------333 334Unregister network device timeout in seconds.335This option controls the timeout (in seconds) used to issue a warning while336waiting for a network device refcount to drop to 0 during device337unregistration. A lower value may be useful during bisection to detect338a leaked reference faster. A larger value may be useful to prevent false339warnings on slow/loaded systems.340Default value is 10, minimum 1, maximum 3600.341 342skb_defer_max343-------------344 345Max size (in skbs) of the per-cpu list of skbs being freed346by the cpu which allocated them. Used by TCP stack so far.347 348Default: 64349 350optmem_max351----------352 353Maximum ancillary buffer size allowed per socket. Ancillary data is a sequence354of struct cmsghdr structures with appended data. TCP tx zerocopy also uses355optmem_max as a limit for its internal structures.356 357Default : 128 KB358 359fb_tunnels_only_for_init_net360----------------------------361 362Controls if fallback tunnels (like tunl0, gre0, gretap0, erspan0,363sit0, ip6tnl0, ip6gre0) are automatically created. There are 3 possibilities364(a) value = 0; respective fallback tunnels are created when module is365loaded in every net namespaces (backward compatible behavior).366(b) value = 1; [kcmd value: initns] respective fallback tunnels are367created only in init net namespace and every other net namespace will368not have them.369(c) value = 2; [kcmd value: none] fallback tunnels are not created370when a module is loaded in any of the net namespace. Setting value to371"2" is pointless after boot if these modules are built-in, so there is372a kernel command-line option that can change this default. Please refer to373Documentation/admin-guide/kernel-parameters.txt for additional details.374 375Not creating fallback tunnels gives control to userspace to create376whatever is needed only and avoid creating devices which are redundant.377 378Default : 0 (for compatibility reasons)379 380devconf_inherit_init_net381------------------------382 383Controls if a new network namespace should inherit all current384settings under /proc/sys/net/{ipv4,ipv6}/conf/{all,default}/. By385default, we keep the current behavior: for IPv4 we inherit all current386settings from init_net and for IPv6 we reset all settings to default.387 388If set to 1, both IPv4 and IPv6 settings are forced to inherit from389current ones in init_net. If set to 2, both IPv4 and IPv6 settings are390forced to reset to their default values. If set to 3, both IPv4 and IPv6391settings are forced to inherit from current ones in the netns where this392new netns has been created.393 394Default : 0 (for compatibility reasons)395 396txrehash397--------398 399Controls default hash rethink behaviour on socket when SO_TXREHASH option is set400to SOCK_TXREHASH_DEFAULT (i. e. not overridden by setsockopt).401 402If set to 1 (default), hash rethink is performed on listening socket.403If set to 0, hash rethink is not performed.404 405gro_normal_batch406----------------407 408Maximum number of the segments to batch up on output of GRO. When a packet409exits GRO, either as a coalesced superframe or as an original packet which410GRO has decided not to coalesce, it is placed on a per-NAPI list. This411list is then passed to the stack when the number of segments reaches the412gro_normal_batch limit.413 414high_order_alloc_disable415------------------------416 417By default the allocator for page frags tries to use high order pages (order-3418on x86). While the default behavior gives good results in most cases, some users419might have hit a contention in page allocations/freeing. This was especially420true on older kernels (< 5.14) when high-order pages were not stored on per-cpu421lists. This allows to opt-in for order-0 allocation instead but is now mostly of422historical importance.423 424Default: 0425 4262. /proc/sys/net/unix - Parameters for Unix domain sockets427----------------------------------------------------------428 429There is only one file in this directory.430unix_dgram_qlen limits the max number of datagrams queued in Unix domain431socket's buffer. It will not take effect unless PF_UNIX flag is specified.432 433 4343. /proc/sys/net/ipv4 - IPV4 settings435-------------------------------------436Please see: Documentation/networking/ip-sysctl.rst and437Documentation/admin-guide/sysctl/net.rst for descriptions of these entries.438 439 4404. Appletalk441------------442 443The /proc/sys/net/appletalk directory holds the Appletalk configuration data444when Appletalk is loaded. The configurable parameters are:445 446aarp-expiry-time447----------------448 449The amount of time we keep an ARP entry before expiring it. Used to age out450old hosts.451 452aarp-resolve-time453-----------------454 455The amount of time we will spend trying to resolve an Appletalk address.456 457aarp-retransmit-limit458---------------------459 460The number of times we will retransmit a query before giving up.461 462aarp-tick-time463--------------464 465Controls the rate at which expires are checked.466 467The directory /proc/net/appletalk holds the list of active Appletalk sockets468on a machine.469 470The fields indicate the DDP type, the local address (in network:node format)471the remote address, the size of the transmit pending queue, the size of the472received queue (bytes waiting for applications to read) the state and the uid473owning the socket.474 475/proc/net/atalk_iface lists all the interfaces configured for appletalk.It476shows the name of the interface, its Appletalk address, the network range on477that address (or network number for phase 1 networks), and the status of the478interface.479 480/proc/net/atalk_route lists each known network route. It lists the target481(network) that the route leads to, the router (may be directly connected), the482route flags, and the device the route is using.483 4845. TIPC485-------486 487tipc_rmem488---------489 490The TIPC protocol now has a tunable for the receive memory, similar to the491tcp_rmem - i.e. a vector of 3 INTEGERs: (min, default, max)492 493::494 495 # cat /proc/sys/net/tipc/tipc_rmem496 4252725 34021800 68043600497 #498 499The max value is set to CONN_OVERLOAD_LIMIT, and the default and min values500are scaled (shifted) versions of that same value. Note that the min value501is not at this point in time used in any meaningful way, but the triplet is502preserved in order to be consistent with things like tcp_rmem.503 504named_timeout505-------------506 507TIPC name table updates are distributed asynchronously in a cluster, without508any form of transaction handling. This means that different race scenarios are509possible. One such is that a name withdrawal sent out by one node and received510by another node may arrive after a second, overlapping name publication already511has been accepted from a third node, although the conflicting updates512originally may have been issued in the correct sequential order.513If named_timeout is nonzero, failed topology updates will be placed on a defer514queue until another event arrives that clears the error, or until the timeout515expires. Value is in milliseconds.516