TOR(1)
TOR(1) — the Tor Manual 💅🧅
okay so the manual name is Tor Manual, and the page is TOR(1). We're talking about tor, the second-generation onion router, and bestie we are going to get through this together. No cap. ✨
NAME 💖
tor — The second-generation onion router.
SYNOPSIS ✨
tor [OPTION value]...DESCRIPTION 🧅💖
Tor is a connection-oriented anonymizing communication service. Users choose a source-routed path through a set of nodes, and negotiate a "virtual circuit" through the network. Each node in a virtual circuit knows its predecessor and successor nodes, but no other nodes. Traffic flowing down the circuit is unwrapped by a symmetric key at each node, which reveals the downstream node. 🔥
Basically, Tor provides a distributed network of servers or relays ("onion routers"). Users bounce their TCP streams, including web traffic, ftp, ssh, etc., around the network, so that recipients, observers, and even the relays themselves have difficulty tracking the source of the stream. 🌈
Note 💀 By default,toracts as a client only. To help the network by providing bandwidth as a relay, change theORPortconfiguration option as mentioned below. Please also consult the documentation on the Tor Project’s website.
COMMAND-LINE OPTIONS 🎀
Tor has a powerful command-line interface. This section lists optional arguments you can specify at the command line using the tor command. Periodt.
Configuration options can be specified on the command line in the format --OptionName OptionValue, on the command line in the format OptionName OptionValue, or in a configuration file. For instance, you can tell Tor to start listening for SOCKS connections on port 9999 by passing either --SocksPort 9999 or SocksPort 9999 on the command line, or by specifying SocksPort 9999 in the configuration file. On the command line, quote option values that contain spaces. For instance, if you want Tor to log all debugging messages to debug.log, you must specify --Log "debug file debug.log". 🫶
Note 📝 Configuration options on the command line override those in configuration files. See THE CONFIGURATION FILE FORMAT for more information.
The following options in this section are only recognized on the tor command line, not in a configuration file. iykyk 💅
-h,--help— display a short help message and exit. 👋-f,--torrc-file FILE— specify a new configuration file to contain further Tor configuration options, or pass-to make Tor read its configuration from standard input. (Default:/etc/tor/torrc, or$HOME/.torrcif that file is not found.) 💖--allow-missing-torrc— allow the configuration file specified by-fto be missing, if thedefaults-torrcfile (see below) is accessible. 🫶--defaults-torrc FILE— specify a file in which to find default values for Tor options. The contents of this file are overridden by those in the regular configuration file, and by those on the command line. (Default:/etc/tor/torrc-defaults.) ✨--ignore-missing-torrc— specify that Tor should treat a missingtorrcfile as though it were empty. Ordinarily, Tor does this for missing default torrc files, but not for those specified on the command line. 😭--hash-password PASSWORD— generate a hashed password for control port access. 🔐--list-fingerprint [key type]— generate your keys and output your nickname and fingerprint. Optionally, you can specify the key type asrsa(default) ored25519. 🔑--verify-config— verify whether the configuration file is valid. ✅--dump-config short|full— write a list of Tor’s configured options to standard output. When theshortflag is selected, only write the options that are different from their default values. Whenfullis selected, write every option. 📝--service install [--options command-line options]— install an instance of Tor as a Windows service, with the provided command-line options. Current instructions can be found athttps://www.torproject.org/docs/faq#NTService. 🪟--service remove|start|stop— remove, start, or stop a configured Tor Windows service. 🛠️--nt-service— used internally to implement a Windows service. 🙈--list-torrc-options— list all valid options. 📋--list-deprecated-options— list all valid options that are scheduled to become obsolete in a future version. (This is a warning, not a promise.) ⚠️--list-modules— list whether each optional module has been compiled into Tor. (Any module not listed is not optional in this version of Tor.) 🧩--version— display Tor version and exit. The output is a single line of the format"Tor version [version number]."(The version number format is as specified inversion-spec.txt.) 👑--quiet|--hush— override the default console logging behavior. By default, Tor starts out logging messages at level"notice"and higher to the console. It stops doing so after it parses its configuration, if the configuration tells it to log anywhere else. These options override the default console logging behavior. Use the--hushoption if you want Tor to log only warnings and errors to the console, or use the--quietoption if you want Tor not to log to the console at all. 🤫--keygen [--newpass]— runningtor --keygencreates a new ed25519 master identity key for a relay, or only a fresh temporary signing key and certificate, if you already have a master key. Optionally, you can encrypt the master identity key with a passphrase. When Tor asks you for a passphrase and you don’t want to encrypt the master key, just don’t enter any passphrase when asked.
Use the --newpass option with --keygen only when you need to add, change, or remove a passphrase on an existing ed25519 master identity key. You will be prompted for the old passphrase (if any), and the new passphrase (if any).
Note 🗝️ When generating a master key, you may want to use--DataDirectoryto control where the keys and certificates will be stored, and--SigningKeyLifetimeto control their lifetimes. SeeSERVER OPTIONSto learn more about the behavior of these options. You must have write access to the specifiedDataDirectory.
To use the generated files, you must copy them to the DataDirectory/keys directory of your Tor daemon, and make sure that they are owned by the user actually running the Tor daemon on your system. 💖
--keygen-family basename— generate a new family ID key inbasename.secret_family_key. To use this key, install it on every relay in your family. (Put it in the relay’sKeyDirectory.) Also, store the corresponding family ID inbasename.public_family_id. Then enable the correspondingFamilyIDoption on your relays. This command overwrites these files if they already exist. Seehttps://community.torproject.org/relay/setup/post-install/family-ids/for more information. 👨👩👧👦--passphrase-fd FILEDES— file descriptor to read the passphrase from. Note that unlike with thetor-gencertprogram, the entire file contents are read and used as the passphrase, including any trailing newlines. If the file descriptor is not specified, the passphrase is read from the terminal by default. 🔑--key-expiration [purpose] [--format iso8601|timestamp]— thepurposespecifies which type of key certificate to determine the expiration of. The only currently recognised purpose is"sign".
Running tor --key-expiration sign will attempt to find your signing key certificate and will output, both in the logs as well as to stdout. The optional --format argument lets you specify the time format. Currently, iso8601 and timestamp are supported. If --format is not specified, the signing key certificate’s expiration time will be in ISO-8601 format. For example, the output sent to stdout will be of the form: "signing-cert-expiry: 2017-07-25 08:30:15 UTC". If --format timestamp is specified, the signing key certificate’s expiration time will be in Unix timestamp format. For example, the output sent to stdout will be of the form: "signing-cert-expiry: 1500971415". ⏰
--dbg-...— Tor may support other options beginning with the string"dbg". These are intended for use by developers to debug and test Tor. They are not supported or guaranteed to be stable, and you should probably not use them. 💀
✨ THE CONFIGURATION FILE FORMAT ✨
okay bestie, so every config option in a config file is written on one single line by default. the shape is: an option name and a value, or an option name and a quoted value. like option value or option "value". literally that's it. 💅
anything after a # character? treated as a comment. gone. ghosted. and the options themselves are case-insensitive — no one is checking your caps lock energy. C-style escaped characters are allowed inside quoted values, slay. 🌸
wanna split ONE config entry across MULTIPLE lines? okay icon. put a single backslash character (\) before the end of the line. you can even drop comments inside those multiline entries, but they must start at the beginning of a line. no exceptions, periodt.
config options can be imported from files or folders using the %include option, where the value is a path. that path can have wildcards 🦋. here's the sequence, no skipping:
- wildcards get expanded first
- then everything is sorted using lexical order
then, for each matching file or folder, the following rules apply:
- if the path is a file, the options from the file get parsed as if they were written right where the
%includeoption is. sneaky! invisible! iconic! - if the path is a folder, all files in that folder get parsed following lexical order. files starting with a dot are ignored. files in subfolders are ignored too. 🫶
%include can be used recursively, because of course it can. and here's the ⚠️: new configuration files or directories cannot be added to an already running Tor instance if Sandbox is enabled.
the supported wildcards are:
*— meaning any number of characters including none?— meaning exactly one character
these characters can be escaped by preceding them with a backslash, except on Windows. also, files starting with a dot are not matched when expanding wildcards unless the starting dot is explicitly in the pattern — again, except on Windows.
now for the precedence drama 🌈. by default:
- an option on the command line overrides an option found in the configuration file
- an option in a configuration file overrides one in the defaults file
simple enough for options that take a single value. but bestie, it gets messy for options that are allowed to occur more than once. so say you specify four SocksPorts in your configuration file, and one more SocksPort on the command line — that command-line option will replace all of the SocksPorts in the configuration file. all four. gone. 😭
if that is not what you want, prefix the option name with a plus sign (+), and it will be appended to the previous set of options instead. watch:
SocksPort 9100uses only port 9100.
+SocksPort 9100uses ports 9100 and 9050 — because 9050 is the default. ✨
alternatively, maybe you want to remove every instance of an option in the configuration file and not replace it at all. you want to say on the command line that you want no SocksPorts whatsoever. to do that, prefix the option name with a forward slash (/). and yes, you can use the plus sign (+) and the forward slash (/) in the configuration file and on the command line. versatility! 💅
👑 GENERAL OPTIONS 👑
AccelDir DIR— specify this option if you're using dynamic hardware acceleration and the engine implementation library lives somewhere other than the OpenSSL default. can not be changed while tor is running.AccelName NAME— when using OpenSSL hardware crypto acceleration, attempt to load the dynamic engine of this name. this must be used for any dynamic hardware engine. names can be verified with theopenssl enginecommand. can not be changed while tor is running.
and here's the twist 🎀: if the engine name is prefixed with a "!", then Tor will exit if the engine cannot be loaded. dramatic, but fair.
AlternateBridgeAuthority [nickname] [flags] ipv4address:port fingerprintAlternateDirAuthority [nickname] [flags] ipv4address:port fingerprint
these two behave as DirAuthority, but they replace fewer of the default directory authorities. using AlternateDirAuthority replaces the default Tor directory authorities, but leaves the default bridge authorities in place. similarly, AlternateBridgeAuthority replaces the default bridge authority, but leaves the directory authorities alone. swap one thing, keep the other. balance. 🧚
AvoidDiskWrites 0|1— if non-zero, try to write to disk less frequently than we would otherwise. useful when you're running on flash memory or other media that only supports a limited number of writes. (Default:0)BandwidthBurst N bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— limit the maximum token bucket size (also known as the burst) to the given number of bytes in each direction. (Default:1 GByte)BandwidthRate N bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— a token bucket limits the average incoming bandwidth usage on this node to the specified number of bytes per second, and the average outgoing bandwidth usage to that same value. if you want to run a relay in the public network, this needs to be at the very least75 KBytesfor a relay (that is,600 kbits) or50 KBytesfor a bridge (400 kbits) — but of course, more is better; we recommend at least250 KBytes(2 mbits) if possible. (Default:1 GByte)
note that this option, and other bandwidth-limiting options, apply to TCP data only: they do not count TCP headers or DNS traffic. 🙈
Tor uses powers of two, not powers of ten, so 1 GByte is 1024*1024*1024 bytes as opposed to 1 billion bytes. math is math. 💀
with this option, and in other options that take arguments in bytes, KBytes, and so on, other formats are also supported. notably, "KBytes" can also be written as "kilobytes" or "kb"; "MBytes" can be written as "megabytes" or "MB"; "kbits" can be written as "kilobits"; and so forth. case doesn't matter. Tor also accepts "byte" and "bit" in the singular. the prefixes "tera" and "T" are also recognized. if no units are given, we default to bytes. to avoid confusion, we recommend writing "bytes" or "bits" explicitly, since it's easy to forget that "B" means bytes, not bits. 🫠
CacheDirectory DIR— store cached directory data in DIR. can not be changed while tor is running. (Default: uses the value ofDataDirectory.)CacheDirectoryGroupReadable 0|1|auto— if this option is set to0, don't allow the filesystem group to read theCacheDirectory. if it's set to1, make theCacheDirectoryreadable by the default GID. if it's"auto", then we use the setting forDataDirectoryGroupReadablewhen theCacheDirectoryis the same as theDataDirectory, and0otherwise. (Default:auto)CircuitPriorityHalflife NUM— if this value is set, we override the default algorithm for choosing which circuit's cell to deliver or relay next. it gets delivered first to the circuit that has the lowest weighted cell count, where cells are weighted exponentially according to this value (in seconds). if the value is-1, it's taken from the consensus if possible, else it will fallback to the default value of30. Minimum:1, Maximum:2147483647. this can be defined as a float value. this is an advanced option; you generally shouldn't have to mess with it. (Default:-1)ClientTransportPlugin transport socks4|socks5 IP:PORTClientTransportPlugin transport exec path-to-binary [options]
in its first form, when set along with a corresponding Bridge line, the Tor client forwards its traffic to a SOCKS-speaking proxy on "IP:PORT". (IPv4 addresses should be written as-is; IPv6 addresses should be wrapped in square brackets.) it's the duty of that proxy to properly forward the traffic to the bridge.
in its second form, when set along with a corresponding Bridge line, the Tor client launches the pluggable transport proxy executable in path-to-binary using options as its command-line options, and forwards its traffic to it. it's the duty of that proxy to properly forward the traffic to the bridge. (Default: none)
ConfluxEnabled 0|1|auto— if this option is set to1, general purpose traffic will use Conflux, which is traffic splitting among multiple legs (circuits). Onion services are not supported at the moment. the default value is set to"auto", meaning the consensus is used to decide unless set. (Default:auto)ConfluxClientUX throughput|latency|throughput_lowmem|latency_lowmem— this option configures the user experience that the client requests from the exit, for data that the exit sends to the client. the default is"throughput", which maximizes throughput."Latency"will tell the exit to only use the circuit with lower latency for all data. the lowmem versions minimize queue usage memory at the client. (Default:"throughput")ConnLimit NUM— the minimum number of file descriptors that must be available to the Tor process before it will start. Tor will ask the OS for as many file descriptors as the OS will allow (you can find this byulimit -H -n). if this number is less thanConnLimit, then Tor will refuse to start. 😭
Tor relays need thousands of sockets, to connect to every other relay. if you're running a private bridge, you can reduce the number of sockets that Tor uses. for example, to limit Tor to 500 sockets, run ulimit -n 500 in a shell. then start tor in the same shell, with ConnLimit 500. you may also need to set DisableOOSCheck 0.
unless you have severely limited sockets, you probably don't need to adjust ConnLimit itself. it has no effect on Windows, since that platform lacks getrlimit(). (Default: 1000)
ConstrainedSockets 0|1— if set, Tor will tell the kernel to attempt to shrink the buffers for all sockets to the size specified inConstrainedSockSize. this is useful for virtual servers and other environments where system level TCP buffers may be limited. if you're on a virtual server and you encounter the"Error creating network socket: No buffer space available"message, you are likely experiencing this problem. 💀
the preferred solution is to have the admin increase the buffer pool for the host itself via /proc/sys/net/ipv4/tcp_mem or equivalent facility; this configuration option is a second-resort. 🫶
the DirPort option should also not be used if TCP buffers are scarce. the cached directory requests consume additional sockets, which exacerbates the problem. 🙈
🚨 you should not enable this feature unless you encounter the"no buffer space available"issue. reducing the TCP buffers affects window size for the TCP stream and will reduce throughput in proportion to round trip time on long paths. (Default:0)
ConstrainedSockSize N bytes|KBytes— whenConstrainedSocketsis enabled, the receive and transmit buffers for all sockets will be set to this limit. must be a value between2048and262144, in1024byte increments. the default of8192is recommended. ✨ControlPort [address:]port|unix:path|auto [flags]— bestie, set this and Tor will accept connections on this port and let all those connections control the whole Tor process using the Tor Control Protocol (it's the thing described incontrol-spec.txtin torspec, go look 👀). Note: unless you also specify one or more ofHashedControlPasswordorCookieAuthentication, setting this option makes Tor let any process on the local host control it. 😭 (If you set both auth methods, then either method alone is sufficient to authenticate to Tor.) This option is required for so many Tor controllers, and most of them use9051. If you're on a unix domain socket, you may quote the path using standard C escape sequences. You can specify this directive multiple times to bind to multiple address/port pairs. Set it to"auto"and Tor picks a port for you. (Default:0)
The recognized flags are, and I'm obsessed with how specific they are:
GroupWritable— unix domain sockets only: makes the socket get created as group-writable.WorldWritable— unix domain sockets only: makes the socket get created as world-writable.RelaxDirModeCheck— unix domain sockets only: do not insist that the directory holding the socket be read-restricted.
ControlPortFileGroupReadable 0|1— set it to0and the filesystem group does not get to read the control port file 🙅♀️; set it to1and the control port file becomes readable by the default GID. (Default:0)ControlPortWriteToFile Path— if set, Tor writes the address and port of any control port it opens to this address. Iconic for controllers that need to learn the actual control port whenControlPortis set to"auto". ✨ControlSocket Path— literallyControlPortbut it listens on a Unix domain socket instead of a TCP socket.0disablesControlSocket. (Unix and Unix-like systems only.) (Default:0)ControlSocketsGroupWritable 0|1— set it to0and the filesystem group can't read and write unix sockets (e.g.ControlSocket); set it to1and the control socket becomes readable and writable by the default GID. (Default:0)CookieAuthentication 0|1— set this to1and Tor allows connections on the control port as long as the connecting process knows the contents of a file named"control_auth_cookie", which Tor creates in its data directory. 🍪 This auth method should only be used on systems with good filesystem security, no cap. (Default:0)CookieAuthFile Path— if set, this option overrides the default location and file name for Tor's cookie file. (SeeCookieAuthentication.)CookieAuthFileGroupReadable 0|1— set it to0and the filesystem group can't read the cookie file; set it to1and the cookie file becomes readable by the default GID. [Making the file readable by other groups is not yet implemented; let us know if you need this for some reason.] (Default:0)CountPrivateBandwidth 0|1— if this option is set, Tor's rate-limiting applies not only to remote connections but also to connections to private addresses like127.0.0.1or10.0.0.1. Mostly useful for debugging rate-limiting, ngl. (Default:0)DataDirectory DIR— store working data in DIR. Can not be changed while tor is running. 🚫 (Default:~/.torif your home directory is not/; otherwise,/var/lib/tor. On Windows, the default is your ApplicationData folder.)DataDirectoryGroupReadable 0|1— set it to0and the filesystem group can't read the DataDirectory; set it to1and the DataDirectory becomes readable by the default GID. (Default:0)DirAuthority [nickname] [flags] ipv4address:dirport fingerprint— use a nonstandard authoritative directory server at the provided address and port, with the specified key fingerprint. You can repeat this option so many times, for multiple authoritative directory servers. Flags are separated by spaces and determine what kind of an authority this directory is. By default, an authority is not authoritative for any directory style or version unless an appropriate flag is given.
Tor will use this authority as a bridge authoritative directory if the "bridge" flag is set. If a flag "orport=orport" is given, Tor uses that given port when opening encrypted tunnels to the dirserver. If a flag "weight=num" is given, the directory server is chosen randomly with probability proportional to that weight (default 1.0). If a flag "v3ident=fp" is given, the dirserver is a v3 directory authority whose v3 long-term signing key has the fingerprint fp. And lastly, if an "ipv6=[ipv6address]:orport" flag is present, then the directory authority is listening for IPv6 connections on the indicated IPv6 address and OR Port. 🦋
Tor will contact the authority at ipv4address to download directory documents. Clients always use the ORPort. Relays usually use the DirPort, but will use the ORPort in some circumstances. If an IPv6 ORPort is supplied, clients will also download directory documents at the IPv6 ORPort, if they are configured to use IPv6.
If no DirAuthority line is given, Tor uses the default directory authorities. NOTE: this option is intended for setting up a private Tor network with its own directory authorities. If you use it, you will be distinguishable from other users, because you won't believe the same authorities they do. 👀
DirAuthorityFallbackRate NUM— when configured to use both directory authorities and fallback directories, the directory authorities also work as fallbacks. They're chosen with their regular weights, multiplied by this number, which should be1.0or less. The default is less than 1, to reduce load on authorities. (Default:0.1)DisableAllSwap 0|1— set it to1and Tor attempts to lock all current and future memory pages so that memory cannot be paged out. 🔒 Windows, OS X and Solaris are currently not supported. We believe this feature works on modern Gnu/Linux distributions, and that it should work on *BSD systems (untested). This option requires that you start your Tor as root, and you should use theUseroption to properly reduce Tor's privileges. Can not be changed while tor is running. (Default:0)DisableDebuggerAttachment 0|1— set it to1and Tor attempts to prevent basic debugging attachment attempts by other processes. This may also keep Tor from generating core files if it crashes. It has no impact for users who wish to attach if they haveCAP_SYS_PTRACEor if they are root. We believe this feature works on modern Gnu/Linux distributions, and that it may also work on *BSD systems (untested). Some modern Gnu/Linux systems like Ubuntu have thekernel.yama.ptrace_scopesysctl and by default enable it as an attempt to limit the PTRACE scope for all user processes by default. This feature attempts to limit the PTRACE scope for Tor specifically — it will not attempt to alter the system wide ptrace scope, because it may not even exist. If you wish to attach to Tor with a debugger such asgdborstraceyou will want to set this to0for the duration of your debugging. Normal users should leave it on. Disabling this option while Tor is running is prohibited. 🚫 (Default:1)DisableNetwork 0|1— when this option is set, we don't listen for or accept any connections other than controller connections, and we close (and don't reattempt) any outbound connections. Controllers sometimes use this option to avoid using the network until Tor is fully configured. Tor will still make certain network-related calls (like DNS lookups) as a part of its configuration process, even ifDisableNetworkis set. 🙈 (Default:0)ExtendByEd25519ID 0|1|auto— set it to1and we always try to include a relay's Ed25519 ID when telling the preceding relay in a circuit to extend to it. Set it to0and we never include Ed25519 IDs when extending circuits. Set it to"auto"and we obey a parameter in the consensus document. (Default:auto)ExtORPort [address:]port|auto— open this port to listen for Extended ORPort connections from your pluggable transports. 🌈
(Default: DataDirectory/extended_orport_auth_cookie)
ExtORPortCookieAuthFile Path— if set, this option overrides the default location and file name for the Extended ORPort's cookie file — and that cookie file is needed for pluggable transports to communicate through the Extended ORPort. 🍪ExtORPortCookieAuthFileGroupReadable 0|1— set it to0and the filesystem group can't read the Extended OR Port cookie file; set it to1and the cookie file becomes readable by the default GID. [Making the file readable by other groups is not yet implemented; let us know if you need this for some reason.] (Default:0)FallbackDir ipv4address:dirport orport=orport id=fingerprint [weight=num] [ipv6=[ipv6address]:orport]— when tor is unable to connect to any directory cache for directory info (usually because it doesn't know about any yet) it tries a hard-coded directory. Relays try one directory authority at a time. Clients try multiple directory authorities and FallbackDirs, so they don't hang on startup if a hard-coded directory is down. Clients wait a few seconds between each attempt, and retry FallbackDirs more often than directory authorities, to reduce the load on the directory authorities. 🫶
FallbackDirs should be stable relays with stable IP addresses, ports, and identity keys. They must have a DirPort.
By default, the directory authorities are also FallbackDirs. Specifying a FallbackDir replaces Tor's default hard-coded FallbackDirs (if any). (See DirAuthority for an explanation of each flag.)
FetchDirInfoEarly 0|1— set it to1and Tor will always fetch directory information like other directory caches do, even if you don't meet the normal criteria for fetching early. Normal users should leave it off. (Default:0)FetchDirInfoExtraEarly 0|1— set it to1and Tor will fetch directory information before other directory caches. It will attempt to download directory information closer to the start of the consensus period. Normal users should leave it off. (Default:0)FetchHidServDescriptors 0|1— set this to0and Tor will literally never fetch hidden service descriptors from the rendezvous directories, not a single one, not even a little. This one is only useful if you're running a Tor controller that handles the hidden service fetches for you, like it's doing your homework. (Default:1) 💅FetchServerDescriptors 0|1— same energy but for the directory side:0means Tor will never fetch network status summaries or server descriptors from the directory servers. Again, only useful if a Tor controller is out here handling directory fetches for you. (Default:1)FetchUselessDescriptors 0|1— okay so1is the maximalist setting: Tor will fetch every consensus flavor, plus all server descriptors and authority certificates referenced by those consensuses, except for extra info descriptors. And when this is on, Tor also keeps fetching descriptors even when it's idle — no rest, main character never sleeps. 🫡
Set it to 0 and Tor avoids the useless stuff: flavors it's not using to build circuits, and authority certificates it doesn't trust. When Tor hasn't built any application circuits, it goes idle and stops fetching descriptors. Very "I'm off the clock" of it.
This option is useful if you're using a tor client with an external parser that uses a full consensus. For the record, here's how the family fits together: this option fetches all documents except extrainfo descriptors, DirCache fetches and serves all documents except extrainfo descriptors, DownloadExtraInfo* fetches extrainfo documents, and serves them if DirCache is on, and UseMicrodescriptors changes the flavor of consensuses and descriptors that is fetched and used for building circuits. (Default: 0)
HardwareAccel 0|1— if non-zero, Tor tries to use built-in (static) crypto hardware acceleration when it's available. 🚀 Can not be changed while tor is running. (Default:0)HashedControlPassword hashed_password— this lets connections on the control port through if they present the password whose one-way hash ishashed_password. You compute that hash by running:
tor --hash-password passwordWant several acceptable passwords? Just use more than one HashedControlPassword line. Multiple besties, multiple keys. 🗝️
HTTPProxy host[:port]— Tor will make all its directory requests through thishost:port— orhost:80if you didn't specify a port — instead of connecting directly to any directory servers. (DEPRECATED: As of0.3.1.0-alphayou should useHTTPSProxy.)HTTPProxyAuthenticator username:password— if you define this, Tor uses thisusername:passwordfor Basic HTTP proxy authentication, as inRFC 2617. This is currently the only form of HTTP proxy authentication that Tor supports; feel free to submit a patch if you want it to support others. 👀 (DEPRECATED: As of0.3.1.0-alphayou should useHTTPSProxyAuthenticator.)HTTPSProxy host[:port]— Tor will make all its OR (SSL) connections through thishost:port— orhost:443if no port is given — via HTTP CONNECT rather than connecting directly to servers. Pro tip: you may want to setFascistFirewallto restrict the set of ports you might try to connect to, if your HTTPS proxy only allows connecting to certain ports. ✨HTTPSProxyAuthenticator username:password— if defined, Tor uses thisusername:passwordfor Basic HTTPS proxy authentication, as inRFC 2617. Again, currently the only form of HTTPS proxy authentication Tor supports — patches welcome if you want more. 🙈KeepalivePeriod NUM— to keep firewalls from letting connections expire on us, send a padding keepalive cell everyNUMseconds on open connections that are in use. (Default:5 minutes) 💓KeepBindCapabilities 0|1|auto— Linux-only lore 🐧. When Tor is started as root and then switches its identity using theUseroption, this tells us whether to try to retain our ability to bind to low ports:1→ we try to keep the capability0→ we do notauto→ we keep the capability only if we're configured to listen on a low port
Can not be changed while tor is running. (Default: auto.)
Log minSeverity[-maxSeverity] stderr|stdout|syslog— send all messages betweenminSeverityandmaxSeverityto the standard output stream, the standard error stream, or to the system log. (The"syslog"value is only supported on Unix.) The recognized severity levels aredebug,info,notice,warn, anderr. We advise using"notice"in most cases, since anything more verbose may provide sensitive information to an attacker who obtains the logs. 😭 If only one severity level is given, all messages of that level or higher will be sent to the listed destination.
Heads up though: some low-level logs may be sent from signal handlers, so their destination logs must be signal-safe. These low-level logs include backtraces, logging function errors, and errors in code called by logging functions. Signal-safe logs are always sent to stderr or stdout. They're also sent to a limited number of log files that are configured to log messages at error severity from the bug or general domains. They are never sent as syslogs, control port log events, or to any API-based log destinations. 🚫
Log minSeverity[-maxSeverity] file FILENAME— as above, but send log messages to the listedFILENAME. The"Log"option may appear more than once in a configuration file. Messages are sent to all the logs that match their severity level. 💖Log [domain,...]minSeverity[-maxSeverity] ... file FILENAMELog [domain,...]minSeverity[-maxSeverity] ... stderr|stdout|syslog— as above, but now you're selecting messages by range of log severity and by a set of "logging domains". Each logging domain corresponds to an area of functionality inside Tor. You can specify any number of severity ranges for a single log statement, each of them prefixed by a comma-separated list of logging domains. Prefix a domain with~to indicate negation, and use*to indicate "all domains". If you specify a severity range without a list of domains, it matches all domains. ✨
This is an advanced feature which is most useful for debugging one or two of Tor's subsystems at a time.
The currently recognized domains are: general, crypto, net, config, fs, protocol, mm, http, app, control, circ, rend, bug, dir, dirserv, or, edge, acct, hist, handshake, heartbeat, channel, sched, guard, consdiff, dos, process, pt, btrack, and mesg. Domain names are case-insensitive.
For example, this line:
Log [handshake]debug [~net,~mm]info notice stdoutsends to stdout: all handshake messages of any severity, all info-and-higher messages from domains other than networking and memory management, and all messages of severity notice or higher. Slay. 👑
LogMessageDomains 0|1— if1, Tor includes message domains with each log message. Every log message currently has at least one domain; most currently have exactly one. This doesn't affect controller log messages. (Default:0)LogTimeGranularity NUM— sets the resolution of timestamps in Tor's logs toNUMmilliseconds.NUMmust be positive and either a divisor or a multiple of 1 second. Note that this option only controls the granularity written by Tor to a file or console log. Tor does not (for example) "batch up" log messages to affect times logged by a controller, times attached to syslog messages, or the mtime fields on log files. (Default:1 second) ⏱️MaxAdvertisedBandwidth N bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— if set, we will not advertise more than this amount of bandwidth for ourBandwidthRate. Server operators who want to reduce the number of clients who ask to build circuits through them (since this is proportional to advertised bandwidth rate) can thus reduce the CPU demands on their server without impacting network performance. Iconic efficiency. 💅MaxUnparseableDescSizeToLog N bytes|KBytes|MBytes|GBytes|TBytes— unparseable descriptors (e.g. for votes, consensuses, routers) are logged in separate files by hash, up to the specified size in total. Note that only files logged during the lifetime of this Tor process count toward the total; this is intended to be used to debug problems without opening live servers to resource exhaustion attacks. (Default:10 MBytes)MetricsPort [address:]port [format]— WARNING: Before enabling this, it is important to understand that exposing tor metrics publicly is dangerous to the Tor network users. Please take extra precaution and care when opening this port. Set a very strict access policy withMetricsPortPolicyand consider using your operating systems firewall features for defense in depth. 🛡️
For the prometheus format, we recommend that the only address that can access this port should be the Prometheus server itself. Remember that the connection is unencrypted (HTTP), hence consider using a tool like stunnel to secure the link from this port to the server. 🔐
If set, open this port to listen for an HTTP GET request to "/metrics". Upon a request, the collected metrics in the the tor instance are formatted for the given format and then sent back. If this is set, MetricsPortPolicy must be defined else every request will be rejected.
Supported format is "prometheus", which is also the default if not set. The Prometheus data model can be found here: https://prometheus.io/docs/concepts/data_model/
The tor metrics are constantly collected and they solely consists of counters. Thus, asking for those metrics is very lightweight on the tor process. (Default: None)
As an example, here only 5.6.7.8 will be allowed to connect: 🎀
MetricsPort 1.2.3.4:9035
MetricsPortPolicy accept 5.6.7.8MetricsPortPolicy policy,policy,...— set an entrance policy for theMetricsPort, to limit who can access it. The policies have the same form as exit policies below, except that port specifiers are ignored. For multiple entries, this line can be used multiple times. It is a reject-all-by-default policy. (Default:None)
Please, keep in mind here that if the server collecting metrics on the MetricsPort is behind a NAT, then everything behind it can access it. Same situation if you allow localhost — every user on the server will be able to access it. Again, strongly consider using a tool like stunnel to secure the link or to strengthen access control. 😭🙏
NoExec0|1— set this to1and Tor will never ever launch another executable, no matter whatClientTransportPluginorServerTransportPluginare saying. 💀 And once you flip it to1, you literally can not flip it back to0without restarting Tor. That's the rules bestie. (Default:0)OutboundBindAddressIP — makes every single outbound connection originate from the IP you specify. This is only useful when you've got multiple network interfaces and you want all of Tor's outgoing connections using just one of them. 🫶 You can use this option twice — once with an IPv4 address and once with an IPv6 address. IPv6 addresses should be wrapped in square brackets. This setting gets ignored for connections to the loopback addresses (127.0.0.0/8and::1), and ngl it's not used for DNS requests either.OutboundBindAddressExitIP — all your outbound exit connections originate from the IP you give it. This one overridesOutboundBindAddressfor the same IP version. ✨ May be used twice, once IPv4 once IPv6. IPv6 goes in square brackets. Ignored for loopback (127.0.0.0/8and::1).OutboundBindAddressORIP — all outbound non-exit (relay and other) connections come from the IP you specify. Also overridesOutboundBindAddressfor the same IP version. 🌸 Twice again — one IPv4, one IPv6, IPv6 in square brackets. Ignored for loopback (127.0.0.0/8and::1).__OwningControllerProcessPID — makes the Tor instance periodically check whether a controller process with the given PID is still alive, and terminate itself if it's not. 👑 Polling interval is15 seconds.PerConnBWBurstN bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits — if you set this manually, or it comes in via the"perconnbwburst"consensus field, Tor uses it for separate rate limiting on each connection from a non-relay. (Default:0)PerConnBWRateN bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits — same energy as above but for rate: set manually or via the"perconnbwrate"consensus field, and Tor uses it for separate rate limiting for each connection from a non-relay. (Default:0)OutboundBindAddressPTIP — this one is a request to pluggable transports to make all their outbound connections originate from the IP you specify. But here's the tea: because outgoing connections are handled by the pluggable transport itself, it is not possible for Tor to enforce whether the pluggable transport honors this option. 😭 It overridesOutboundBindAddressfor the same IP version. May be used twice — one IPv4, one IPv6, IPv6 wrapped in square brackets. Ignored for loopback (127.0.0.0/8and::1).PidFileFILE — on startup, write our PID to FILE. On a clean shutdown, remove FILE. Can not be changed while tor is running. 🎀ProtocolWarnings0|1— if1, Tor logs with severity'warn'all the various cases of other parties not following the Tor specification. Otherwise, those get logged with severity'info'. (Default:0)RelayBandwidthBurstN bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits — if not0, this limits the maximum token bucket size (aka the burst) for _relayed traffic_ to the given number of bytes in each direction. 🌈 Heads up tho: this does not include directory fetches by the relay (from authority or other relays), because that counts as"client"activity. (Default:0)RelayBandwidthBurstdefaults to the value ofRelayBandwidthRateif unset.RelayBandwidthRateN bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits — if not0, a separate token bucket limits the average incoming bandwidth usage for _relayed traffic_ on this node to the specified number of bytes per second, and the average outgoing bandwidth usage to that same value. 🔥 Fyi, relayed traffic currently is calculated to include answers to directory requests, but that may change in future versions. Same as above — no directory fetches by the relay (from authority or other relays), because that's"client"activity. (Default:0)RelayBandwidthRatedefaults to the value ofRelayBandwidthBurstif unset.RephistTrackTimeN seconds|minutes|hours|days|weeks — tells an authority, or any other node tracking node reliability and history, that fine-grained info about nodes can be thrown out once it hasn't changed for the given amount of time. 🗑️ (Default:24 hours)RunAsDaemon0|1— if1, Tor forks and daemonizes to the background. This does absolutely nothing on Windows; instead you should use the--servicecommand-line option. Can not be changed while tor is running. (Default:0)SafeLogging0|1|relay— Tor can scrub potentially sensitive strings from log messages (addresses, for example) by swapping them out for the string[scrubbed]. This way your logs can still be useful, but they don't leave behind personally identifying info about what sites a user might have visited. 💅
If you set this to 0, Tor does no scrubbing at all. Set it to 1 and all potentially sensitive strings get replaced. Set it to relay and all log messages generated when acting as a relay are sanitized, but everything generated when acting as a client is not. Note: Tor may not heed this option when logging at log levels more verbose than Notice. (Default: 1)
Sandbox0|1— if set to1, Tor runs securely through the use of a syscall sandbox. Otherwise the sandbox is disabled. 🧚 This option only works on Linux-based operating systems, and only when Tor has been built with the libseccomp library. Note that it may be incompatible with some versions of libc, and some kernel versions. Can not be changed while tor is running.
When the Sandbox is 1, these options can not be changed when tor is running: Address, ConnLimit, CookieAuthFile, DirPortFrontPage, ExtORPortCookieAuthFile, Logs, ServerDNSResolvConfFile, ClientOnionAuthDir (and any files in it won't reload on HUP signal). 🫠
Launching new Onion Services through the control port is not supported with the current syscall sandboxing implementation.
Tor must remain in client or server mode (some changes to ClientOnly and ORPort are not allowed). Currently, if Sandbox is 1, the ControlPort command "GETINFO address" will not work. 💀
When using %include in the tor configuration files, reloading the tor configuration is not supported after adding new configuration files or directories.
(Default: 0)
SchedulersKIST|KISTLite|Vanilla— pick the scheduler type tor should use. The scheduler is the thing responsible for moving data around within a Tor process. ✨ This is an ordered list by priority, which means the first value gets tried first, and if it's unavailable the second one is tried, and so on. You can change these values at runtime. This option mostly effects relays, and honestly most operators should just leave it at its default value. (Default:KIST,KISTLite,Vanilla)
The possible scheduler types are:
KIST: Kernel-Informed Socket Transport. Tor uses TCP information from the kernel to make informed decisions about how much data to send and when to send it. KIST also handles traffic in batches (seeKISTSchedRunInterval) to improve traffic prioritization decisions. As implemented, KIST will only work on Linux kernel version2.6.39or higher.KISTLite: Same as KIST but without the kernel support. Tor uses all the same mechanics as with KIST, including the batching, but its decisions about how much data to send won't be as good. KISTLite works on all kernels and operating systems, and the majority of the benefits of KIST are still realized with KISTLite. 🙌Vanilla: the scheduler that Tor used before KIST was implemented. It sends as much data as possible, as soon as possible. Vanilla works on all kernels and operating systems.
KISTSchedRunIntervalNUM msec — ifKISTorKISTLiteis used in theSchedulersoption, this controls at which interval the scheduler tick happens. ⏰ If the value is0 msec, the value is taken from the consensus if possible, else it falls back to the default10 msec. Maximum possible value is100 msec. (Default:0 msec)KISTSockBufSizeFactorNUM — ifKISTis used inSchedulers, this is a multiplier of the per-socket limit calculation of the KIST algorithm. (Default:1.0)Socks4Proxyhost[:port] — Tor makes all OR connections through the SOCKS 4 proxy at host:port (or host:1080if the port isn't specified).Socks5Proxyhost[:port] — Tor makes all OR connections through the SOCKS 5 proxy at host:port (or host:1080if the port isn't specified).Socks5ProxyUsernameusernameSocks5ProxyPasswordpassword — if defined, authenticate to the SOCKS 5 server using username and password in accordance to RFC 1929. 🔐 Both username and password must be between1and255characters.SyslogIdentityTagtag — when logging to syslog, this adds a tag to the syslog identity so that log entries are marked with"Tor-tag". Can not be changed while tor is running. (Default: none)TCPProxyprotocol host:port — Tor uses the given protocol to make all its OR (SSL) connections through a TCP proxy on host:port, rather than connecting directly to servers. You might wanna setFascistFirewallto restrict the set of ports you might try to connect to, if your proxy only allows connecting to certain ports. There's no equivalent option for directory connections, because all Tor client versions that support this option download directory documents via OR connections. 🌸
The only protocol supported right now is 'haproxy'. This option is only for clients. (Default: none) +
The HAProxy version 1 proxy protocol is described in detail at https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt +
Both source IP address and source port will be set to zero.
TruncateLogFile0|1— if1, Tor overwrites logs at startup and in response to aHUPsignal, instead of appending to them. (Default:0)UnixSocksGroupWritable 0|1— okay so if you set this to0, the filesystem group does not get to read + write your unix sockets (like the one fromSocksPort unix:). Set it to1and the socket becomes readable and writable by the default GID. Bestie, choose wisely. (Default:0)UseDefaultFallbackDirs 0|1— basically: do we use Tor's default hard-codedFallbackDirs(if there even are any)? Tiny plot twist tho — the moment aFallbackDirline shows up, it replaces the hard-codedFallbackDirsno matter what this option says. So this one's kinda overruled. 🙈 (Default:1)User Username— on startup, Tor does asetuidto this user and asetgidto their primary group. And no, you literally can not change this while tor is running. It's locked in, periodt. 🔒
✨ CLIENT OPTIONS ✨
These ones only do anything for clients — aka if SocksPort, HTTPTunnelPort, TransPort, DNSPort, or NATDPort is non-zero. Everyone else can keep scrolling. 💅
AllowNonRFC953Hostnames 0|1— when this is off, Tor straight up blocks hostnames with illegal characters in them (think@and:) instead of yeeting them to an exit node to be resolved. It's a trap card for accidental attempts to resolve URLs and whatever else. (Default:0)AutomapHostsOnResolve 0|1— turn this on and when someone asks us to resolve an address ending in one of the suffixes fromAutomapHostsSuffixes, we map an unused virtual address to that address and hand back the new virtual one. It's giving "make.onionaddresses work with apps that resolve an address and then connect to it." Iconic behavior tbh. (Default:0)AutomapHostsSuffixes SUFFIX,SUFFIX,...— a comma-separated list of suffixes to use withAutomapHostsOnResolve. The"."suffix is equivalent to"all addresses."Yes, every single one. (Default:.exit,.onion).Bridge [transport] IP:ORPort [fingerprint]— set this alongsideUseBridgesand you're telling Tor to use the relay at"IP:ORPort"as a "bridge" relaying into the Tor network. If a"fingerprint"is provided (same format as forDirAuthority), we verify that the relay chillin' at that location actually has the right fingerprint. We also use that fingerprint to look up the bridge descriptor at the bridge authority, if it's provided and ifUpdateBridgesFromAuthorityis set too. Double conditions, we love a crossover.
If "transport" is provided, it has to match a ClientTransportPlugin line. Then we use that pluggable transport's proxy to move data to the bridge instead of connecting to the bridge directly. Some transports use a transport-specific method to figure out the remote address to connect to — those transports typically just ignore the "IP:ORPort" from the bridge line. Rude but valid.
Tor passes any "key=val" settings to the pluggable transport proxy as per-connection arguments when connecting to the bridge. For what arguments it actually supports, go read that pluggable transport's own docs. Not our problem bestie. 🫶
CircuitPadding 0|1— set to0and Tor won't pad client circuits with extra cover traffic. Only clients may set this one. It should be offered via the UI to mobile users for when bandwidth might be expensive (data is money, okay?). Set to1and padding gets negotiated as per the consensus and relay support — and heads up, unlikeConnectionPadding,CircuitPaddingcannot be force-enabled. (Default:1)ReducedCircuitPadding 0|1— set to1and Tor only uses circuit padding algorithms with low overhead. Only clients may set this. Also should be offered via the UI to mobile users where bandwidth may be expensive. (Default:0)ClientBootstrapConsensusAuthorityDownloadInitialDelay N— the initial delay, in seconds, for when clients should download consensuses from authorities while they're bootstrapping (meaning: they don't have a usable, reasonably live consensus yet). Only used by clients fetching from a list of fallback directory mirrors. And lowkey important: this schedule is advanced by (potentially concurrent) connection attempts, unlike other schedules which get advanced by connection failures. Not the same thing!! (Default:6)ClientBootstrapConsensusAuthorityOnlyDownloadInitialDelay N— initial delay in seconds for when clients should download consensuses from authorities while bootstrapping (still no usable, reasonably live consensus). This one's only used by clients which don't have — or won't fetch from — a list of fallback directory mirrors. Same deal as above: advanced by (potentially concurrent) connection attempts, not by connection failures. (Default:0)ClientBootstrapConsensusFallbackDownloadInitialDelay N— initial delay in seconds for when clients should download consensuses from fallback directory mirrors while bootstrapping (yep, no usable, reasonably live consensus). Only used by clients fetching from a list of fallback directory mirrors. And again: advanced by (potentially concurrent) connection attempts, unlike other schedules which are advanced by connection failures. (Default:0)ClientBootstrapConsensusMaxInProgressTries NUM— try this many simultaneous connections to download a consensus before waiting for one of them to complete, timeout, or error out. Multitasking queen. (Default:3)ClientDNSRejectInternalAddresses 0|1— if true, Tor doesn't believe any anonymously retrieved DNS answer claiming an address resolves to an internal address (like127.0.0.1or192.168.0.1). This option shuts down certain browser-based attacks, and it is not allowed to be set on the default network. (Default:1)ClientOnionAuthDir path— path to the directory holding v3 hidden service authorization files. Each file is for a single onion address, and the files MUST have the suffix".auth_private"(i.e."bob_onion.auth_private"). The content format MUST be:
<onion-address>:descriptor:x25519:<base32-encoded-privkey>The <onion-address> MUST NOT have the ".onion" suffix. The <base32-encoded-privkey> is the base32 representation of the raw key bytes only (32 bytes for x25519). Want more info? See Appendix G in the rend-spec-v3.txt file of torspec. ✨
ClientOnly 0|1— set to1and Tor won't run as a relay or serve directory requests, even ifORPort,ExtORPort, orDirPortare set. (This config option is mostly unnecessary — we added it back when we were considering having Tor clients auto-promote themselves to relays if they were stable and fast enough. Current behavior is simply that Tor is a client unlessORPort,ExtORPort, orDirPortare configured.) Nostalgia option fr. (Default:0)ClientPreferIPv6DirPort 0|1|auto— set to1and Tor prefers a directory port with an IPv6 address over one with IPv4 for direct connections, if a given directory server has both. (Tor also prefers an IPv6 DirPort ifIPv4Clientis set to0.) Set toautoand clients prefer IPv4. Other things may influence the choice too — this option just breaks a tie in favor of IPv6. (Default:auto) (DEPRECATED: This option has had no effect for some time.) Yeah, it's a ghost. 💀ClientPreferIPv6ORPort 0|1|auto— set to1and Tor prefers an OR port with an IPv6 address over IPv4 if a given entry node has both. (Tor also prefers an IPv6 ORPort ifIPv4Clientis set to0.) Set toautoand Tor bridge clients prefer the configured bridge address while other clients prefer IPv4. Again, other things may influence the choice — this one breaks a tie toward IPv6. (Default:auto)ClientRejectInternalAddresses 0|1— if true, Tor doesn't try to fulfill requests to connect to an internal address (like127.0.0.1or192.168.0.1unless an exit node is specifically requested (for example, via a.exithostname, or a controller request). And if true, multicast DNS hostnames for machines on the local network (of the form*.local) also get rejected. (Default:1)ClientUseIPv4 0|1— set to0and Tor avoids connecting to directory servers and entry nodes over IPv4. But bestie, note this: clients with an IPv4 address in aBridge, proxy, or pluggable transport line will still try connecting over IPv4 even ifClientUseIPv4is0. (Default:1)ClientUseIPv6 0|1— set to1and Tor might connect to directory servers or entry nodes over IPv6. For IPv6-only hosts, you need to also setClientUseIPv4to0to disable IPv4. And heads up: clients configured with an IPv6 address in aBridge, proxy, or pluggable transportline will try connecting over IPv6 even ifClientUseIPv6is set to0. Sneaky. (Default:1)ConnectionPadding 0|1|auto— this option governs Tor's use of padding to defend against some forms of traffic analysis. Set toautoand Tor sends padding only if both the client and the relay support it. Set to0and Tor sends zero padding cells. Set to1and Tor will still send padding for client connections regardless of relay support. Only clients may set this. It should be offered via the UI to mobile users for use where bandwidth may be expensive. (Default:auto)ReducedConnectionPadding 0|1— set to1and Tor will not not hold OR connections open for very long, and will send less padding on these connections. (Yes, the original says "not not" — we are quoting it exactly, no editing the source bestie. 😭) Only clients may set this. Should be offered via the UI to mobile users for use where bandwidth may be expensive. (Default:0)DNSPort [address:]port|auto [isolation flags]— if non-zero, open this port to listen for UDP DNS requests and resolve them anonymously. This port only handlesA,AAAA, andPTRrequests — it doesn't handle arbitrary DNS request types, so don't expect the whole menu. Set the port to"auto"to have Tor pick a port for you. This directive can be specified multiple times to bind to multiple addresses/ports. For what the isolation flags mean, go look atSocksPort. (Default:0)DownloadExtraInfo 0|1— if true, Tor downloads and caches"extra-info"documents. These bad boys contain information about servers beyond what's in their regular server descriptors. Tor doesn't actually use this info for anything itself, so to save bandwidth, leave this option turned off. (Default:0)EnforceDistinctSubnets 0|1— if1, Tor won't put two servers whose IP addresses are "too close" on the same circuit. Currently, two addresses count as "too close" if they lie in the same/16range. (Default:1)
FascistFirewall 0|1
Okay so if you set this to 1, Tor is only gonna make outgoing connections to ORs that are running on the ports your firewall actually lets through (defaults to 80 and 443; peep FirewallPorts). This lets you run Tor as a client behind a firewall with restrictive policies — but bestie, it will not let you run as a server behind such a firewall, no cap. If you want more fine-grained control, use ReachableAddresses instead. ✨
FirewallPorts PORTS
A list of ports that your firewall allows you to connect to. Only used when FascistFirewall is set. Heads up: this option is deprecated; use ReachableAddresses instead. (Default: 80, 443)
HTTPTunnelPort [address:]port|auto [isolation flags]
Open this port to listen for proxy connections using the "HTTP CONNECT" protocol instead of SOCKS. Set this to 0 if you don't want to allow "HTTP CONNECT" connections. Set the port to "auto" to have Tor pick a port for you. This directive can be specified multiple times to bind to multiple addresses/ports. If multiple entries of this option are present in your configuration file, Tor will perform stream isolation between listeners by default. See SocksPort for an explanation of isolation flags. (Default: 0)
LongLivedPorts PORTS
A list of ports for services that tend to have long-running connections (e.g. chat and interactive shells — the yappers 🗣️). Circuits for streams that use these ports will contain only high-uptime nodes, to reduce the chance that a node will go down before the stream is finished. Note that the list is also honored for circuits (both client and service side) involving hidden services whose virtual port is in this list. (Default: 21, 22, 706, 1863, 5050, 5190, 5222, 5223, 6523, 6667, 6697, 8300)
MapAddress address newaddress
When a request for address arrives to Tor, it will transform to newaddress before processing it. For example, if you always want connections to www.example.com to exit via torserver (where torserver is the fingerprint of the server), use "MapAddress www.example.com www.example.com.torserver.exit". If the value is prefixed with a "*.", it matches an entire domain. For example, if you always want connections to example.com and any of its subdomains to exit via torserver (where torserver is the fingerprint of the server), use "MapAddress *.example.com *.example.com.torserver.exit". (Note the leading "*." in each part of the directive.) You can also redirect all subdomains of a domain to a single address. For example, "MapAddress *.example.com www.example.com". If the specified exit is not available, or the exit can not connect to the site, Tor will fail any connections to the mapped address.
NOTES:
- When evaluating MapAddress expressions Tor stops when it hits the most recently added expression that matches the requested address. So if you have the following in your torrc,
www.torproject.orgwill map to198.51.100.1:
MapAddress www.torproject.org 192.0.2.1
MapAddress www.torproject.org 198.51.100.1- Tor evaluates the MapAddress configuration until it finds no matches. So if you have the following in your torrc,
www.torproject.orgwill map to203.0.113.1:
MapAddress 198.51.100.1 203.0.113.1
MapAddress www.torproject.org 198.51.100.1- The following MapAddress expression is invalid (and will be ignored) because you cannot map from a specific address to a wildcard address:
MapAddress www.torproject.org *.torproject.org.torserver.exit- Using a wildcard to match only part of a string (as in
*ample.com) is also invalid. - Tor maps hostnames and IP addresses separately. If you MapAddress a DNS name, but use an IP address to connect, then Tor will ignore the DNS name mapping.
- MapAddress does not apply to redirects in the application protocol. For example, HTTP redirects and alt-svc headers will ignore mappings for the original address. You can use a wildcard mapping to handle redirects within the same site.
MaxCircuitDirtiness NUM
Feel free to reuse a circuit that was first used at most NUM seconds ago, but never attach a new stream to a circuit that is too old. For hidden services, this applies to the last time a circuit was used, not the first. Circuits with streams constructed with SOCKS authentication via SocksPorts that have KeepAliveIsolateSOCKSAuth also remain alive for MaxCircuitDirtiness seconds after carrying the last such stream. (Default: 10 minutes)
MaxClientCircuitsPending NUM
Do not allow more than NUM circuits to be pending at a time for handling client streams. A circuit is "pending" if we have begun constructing it, but it has not yet been completely constructed. Basically it's still cooking 🍳. (Default: 32)
NATDPort [address:]port|auto [isolation flags]
Open this port to listen for connections from old versions of ipfw (as included in old versions of FreeBSD, etc) using the NATD protocol. Use 0 if you don't want to allow NATD connections. Set the port to "auto" to have Tor pick a port for you. This directive can be specified multiple times to bind to multiple addresses/ports. If multiple entries of this option are present in your configuration file, Tor will perform stream isolation between listeners by default. See SocksPort for an explanation of isolation flags.
This option is only for people who cannot use TransPort. (Default: 0)
NewCircuitPeriod NUM
Every NUM seconds, consider whether to build a new circuit. Fresh circuit check-in, periodt. (Default: 30 seconds)
PathBiasCircThreshold NUM
PathBiasDropGuards NUM
PathBiasExtremeRate NUM
PathBiasNoticeRate NUM
PathBiasWarnRate NUM
PathBiasScaleThreshold NUM
These options override the default behavior of Tor’s (currently experimental) path bias detection algorithm. To try to find broken or misbehaving guard nodes, Tor looks for nodes where more than a certain fraction of circuits through that guard fail to get built.
The PathBiasCircThreshold option controls how many circuits we need to build through a guard before we make these checks. The PathBiasNoticeRate, PathBiasWarnRate and PathBiasExtremeRate options control what fraction of circuits must succeed through a guard so we won’t write log messages. If less than PathBiasExtremeRate circuits succeed and PathBiasDropGuards is set to 1, we disable use of that guard.
When we have seen more than PathBiasScaleThreshold circuits through a guard, we scale our observations by 0.5 (governed by the consensus) so that new observations don’t get swamped by old ones.
By default, or if a negative value is provided for one of these options, Tor uses reasonable defaults from the networkstatus consensus document. If no defaults are available there, these options default to 150, .70, .50, .30, 0, and 300 respectively.
PathBiasUseThreshold NUM
PathBiasNoticeUseRate NUM
PathBiasExtremeUseRate NUM
PathBiasScaleUseThreshold NUM
Similar to the above options, these options override the default behavior of Tor’s (currently experimental) path use bias detection algorithm.
Where as the path bias parameters govern thresholds for successfully building circuits, these four path use bias parameters govern thresholds only for circuit usage. Circuits which receive no stream usage are not counted by this detection algorithm. A used circuit is considered successful if it is capable of carrying streams or otherwise receiving well-formed responses to RELAY cells.
By default, or if a negative value is provided for one of these options, Tor uses reasonable defaults from the networkstatus consensus document. If no defaults are available there, these options default to 20, .80, .60, and 100, respectively.
PathsNeededToBuildCircuits NUM
Tor clients don’t build circuits for user traffic until they know about enough of the network so that they could potentially construct enough of the possible paths through the network. If this option is set to a fraction between 0.25 and 0.95, Tor won’t build circuits until it has enough descriptors or microdescriptors to construct that fraction of possible paths. Note that setting this option too low can make your Tor client less anonymous, and setting it too high can prevent your Tor client from bootstrapping. If this option is negative, Tor will use a default value chosen by the directory authorities. If the directory authorities do not choose a value, Tor will default to 0.6. (Default: -1)
ReachableAddresses IP[/MASK][:PORT]...
A comma-separated list of IP addresses and ports that your firewall allows you to connect to. The format is as for the addresses in ExitPolicy, except that "accept" is understood unless "reject" is explicitly provided. For example, 'ReachableAddresses 99.0.0.0/8, reject 18.0.0.0/8:80, accept *:80' means that your firewall allows connections to everything inside net 99, rejects port 80 connections to net 18, and accepts connections to port 80 otherwise. (Default: 'accept *:*'.)
ReachableDirAddresses IP[/MASK][:PORT]...
Like ReachableAddresses, a list of addresses and ports. Tor will obey these restrictions when fetching directory information, using standard HTTP GET requests. If not set explicitly then the value of ReachableAddresses is used. If HTTPProxy is set then these connections will go through that proxy. (DEPRECATED: This option has had no effect for some time.) 🫠
ReachableORAddresses IP[/MASK][:PORT]...
Like ReachableAddresses, a list of addresses and ports. Tor will obey these restrictions when connecting to Onion Routers, using TLS/SSL. If not set explicitly then the value of ReachableAddresses is used. If HTTPSProxy is set then these connections will go through that proxy.
The separation between ReachableORAddresses and ReachableDirAddresses is only interesting when you are connecting through proxies (see HTTPProxy and HTTPSProxy). Most proxies limit TLS connections (which Tor uses to connect to Onion Routers) to port 443, and some limit HTTP GET requests (which Tor uses for fetching directory information) to port 80. 💅
SafeSocks 0|1
Flip this on and Tor will reject any application connection that uses an unsafe variant of the socks protocol 💀 — i.e. the ones that only hand over an IP address, meaning the app already did a DNS resolve by itself first. Specifically, these are socks4 and socks5 when not doing remote DNS. (Default: 0)
TestSocks 0|1
Turn this on and Tor drops a notice-level log entry for every single connection to the Socks port, telling you whether the request used a safe socks protocol or an unsafe one (peep SafeSocks for what "unsafe" means 👀). This is lowkey the vibe-check tool for figuring out whether an app using Tor might be leaking DNS requests. (Default: 0)
WarnPlaintextPorts port,port,...
Tells Tor to yell a warning whenever the user tries to make an anonymous connection to one of these ports 🚨. This option exists to alert users about services that risk sending passwords in the clear, which is so not it. (Default: 23,109,110,143)
RejectPlaintextPorts port,port,...
Same energy as WarnPlaintextPorts, but instead of just warning you about risky port usage, Tor will flat-out refuse to make the connection. Periodt. (Default: None)
SocksPolicy policy,policy,...
Sets an entrance policy for this server, limiting who can connect to the SocksPort and DNSPort ports 🚪. The policies have the same form as exit policies below, except that port specifiers are ignored. Any address not matched by some entry in the policy is accepted. (No gatekeeping by default, bestie 🫶)
SocksPort [address:]port|unix:path|auto [flags] [isolation flags]
Opens this port to listen for connections from SOCKS-speaking applications ✨. Set it to 0 if you don't want to allow application connections via SOCKS. Set it to "auto" to have Tor pick a port for you (iconic, no decisions required). This directive can be specified multiple times to bind to multiple addresses/ports. If a unix domain socket is used, you may quote the path using standard C escape sequences. Most flags are off by default, except where specified. Flags that are on by default can be disabled by putting "No" before the flag name. (Default: 9050)
NOTE: Although this option allows you to specify an IP address other than localhost, you should do so only with extreme caution 😭. The SOCKS protocol is unencrypted and (as we use it) unauthenticated, so exposing it in this way could leak your information to anybody watching your network, and allow anybody to use your computer as an open proxy. Please don't do this to yourself 💀.
If multiple entries of this option are present in your configuration file, Tor will perform stream isolation between listeners by default 🦋. The isolation flags arguments give Tor rules for which streams received on this SocksPort are allowed to share circuits with one another. Recognized isolation flags are:
IsolateClientAddr— Don't share circuits with streams from a different client address. (On by default and strongly recommended when supported; you can disable it withNoIsolateClientAddr. Unsupported and force-disabled when using Unix domain sockets.)IsolateSOCKSAuth— Don't share circuits with streams for which different SOCKS authentication was provided. (ForHTTPTunnelPortconnections, this option looks at theProxy-AuthorizationandX-Tor-Stream-Isolationheaders. On by default; you can disable it withNoIsolateSOCKSAuth.)IsolateClientProtocol— Don't share circuits with streams using a different protocol. (SOCKS 4, SOCKS 5,HTTPTunnelPortconnections,TransPortconnections,NATDPortconnections, andDNSPortrequests are all considered to be different protocols.)IsolateDestPort— Don't share circuits with streams targeting a different destination port.IsolateDestAddr— Don't share circuits with streams targeting a different destination address.KeepAliveIsolateSOCKSAuth— IfIsolateSOCKSAuthis enabled, keep alive circuits while they have at least one stream with SOCKS authentication active. After such a circuit is idle for more thanMaxCircuitDirtinessseconds, it can be closed.SessionGroup=INT— If no other isolation rules would prevent it, allow streams on this port to share circuits with streams from every other port with the same session group. (By default, streams received on differentSocksPorts,TransPorts, etc are always isolated from one another. This option overrides that behavior.)
Other recognized flags for a SocksPort are:
NoIPv4Traffic— Tell exits to not connect to IPv4 addresses in response to SOCKS requests on this connection.IPv6Traffic— Tell exits to allow IPv6 addresses in response to SOCKS requests on this connection, so long as SOCKS5 is in use. (SOCKS4 can't handle IPv6. L + no IPv6 for you, SOCKS4 🍓)PreferIPv6— Tells exits that, if a host has both an IPv4 and an IPv6 address, we would prefer to connect to it via IPv6. (IPv4 is the default.)NoDNSRequest— Do not ask exits to resolve DNS addresses in SOCKS5 requests. Tor will connect to IPv4 addresses, IPv6 addresses (ifIPv6Trafficis set) and.onionaddresses.NoOnionTraffic— Do not connect to.onionaddresses in SOCKS5 requests.OnionTrafficOnly— Tell the tor client to only connect to.onionaddresses in response to SOCKS5 requests on this connection. This is equivalent toNoDNSRequest,NoIPv4Traffic,NoIPv6Traffic. The correspondingNoOnionTrafficOnlyflag is not supported 🙈.CacheIPv4DNS— Tells the client to remember IPv4 DNS answers we receive from exit nodes via this connection.CacheIPv6DNS— Tells the client to remember IPv6 DNS answers we receive from exit nodes via this connection.GroupWritable— Unix domain sockets only: makes the socket get created as group-writable.WorldWritable— Unix domain sockets only: makes the socket get created as world-writable.CacheDNS— Tells the client to remember all DNS answers we receive from exit nodes via this connection.UseIPv4Cache— Tells the client to use any cached IPv4 DNS answers we have when making requests via this connection. (NOTE: This option, orUseIPv6CacheorUseDNSCache, can harm your anonymity, and probably won't help performance as much as you might expect. Use with care! 😬)UseIPv6Cache— Tells the client to use any cached IPv6 DNS answers we have when making requests via this connection.UseDNSCache— Tells the client to use any cached DNS answers we have when making requests via this connection.NoPreferIPv6Automap— When serving a hostname lookup request on this port that should get automapped (according toAutomapHostsOnResolve), if we could return either an IPv4 or an IPv6 answer, prefer an IPv4 answer. (Tor prefers IPv6 by default.)PreferSOCKSNoAuth— Ordinarily, when an application offers both"username/password authentication"and"no authentication"to Tor via SOCKS5, Tor selects username/password authentication so thatIsolateSOCKSAuthcan work. This can confuse some applications, if they offer a username/password combination then get confused when asked for one 💀. You can disable this behavior, so that Tor will select"No authentication"whenIsolateSOCKSAuthis disabled, or when this option is set.ExtendedErrors— Return extended error code in the SOCKS reply. So far, the possible errors are:X'F0'Onion Service Descriptor Can Not be Found — The requested onion service descriptor can't be found on the hashring and thus not reachable by the client. (v3 only)X'F1'Onion Service Descriptor Is Invalid — The requested onion service descriptor can't be parsed or signature validation failed. (v3 only)X'F2'Onion Service Introduction Failed — All introduction attempts failed either due to a combination of NACK by the intro point or time out. (v3 only)X'F3'Onion Service Rendezvous Failed — Every rendezvous circuit has timed out and thus the client is unable to rendezvous with the service. (v3 only)X'F4'Onion Service Missing Client Authorization — Client was able to download the requested onion service descriptor but is unable to decrypt its content because it is missing client authorization information. (v3 only)X'F5'Onion Service Wrong Client Authorization — Client was able to download the requested onion service descriptor but is unable to decrypt its content using the client authorization information it has. This means the client access were revoked. (v3 only)X'F6'Onion Service Invalid Address — The given.onionaddress is invalid. In one of these cases this error is returned: address checksum doesn't match, ed25519 public key is invalid or the encoding is invalid. (v3 only)X'F7'Onion Service Introduction Timed Out — Similar toX'F2'code but in this case, all introduction attempts have failed due to a time out. (v3 only)
Flags are processed left to right. If flags conflict, the last flag on the line is used, and all earlier flags are ignored. No error is issued for conflicting flags. (So order matters, bestie — it's giving "last one wins" 👑)
TokenBucketRefillInterval NUM [msec|second]
Sets the refill delay interval of Tor's token bucket to NUM milliseconds ⏱️. NUM must be between 1 and 1000, inclusive. When Tor is out of bandwidth, on a connection or globally, it will wait up to this long before it tries to use that connection again. Note that bandwidth limits are still expressed in bytes per second: this option only affects the frequency with which Tor checks to see whether previously exhausted connections may read again. Can not be changed while tor is running. (Default: 100 msec)
TrackHostExitshost,.domain,...— okay so for every single value in that comma-separated list, Tor will track your recent connections to hosts that match it and attempt to reuse the same exit node for each one. If a value is prepended with a., it's treated as matching an entire domain. And if one of the values is just a., that means match everything. So when is this lowkey a lifesaver? When you frequently connect to sites that will expire all your authentication cookies (i.e. log you out 😭) the second your IP address changes. Heads up though — this option does have the disadvantage of making it more clear that a given history is associated with a single user. Which, 💀. But most people who would wish to observe this will observe it through cookies or other protocol-specific means anyhow, so. whatever.TrackHostExitsExpireNUM— since exit servers go up and down (they do, they're so dramatic), it is desirable to expire the association between host and exit server afterNUMseconds. Default is1800seconds, aka 30 minutes. ⏰ periodt.TransPort[address:]port|auto[isolation flags]— open this port to listen for transparent proxy connections. Set it to0if you don't want to allow transparent proxy connections. Set the port to"auto"to have Tor pick a port for you ✨. This directive can be specified multiple times to bind to multiple addresses/ports. If multiple entries of this option are present in your configuration file, Tor will perform stream isolation between listeners by default. SeeSocksPortfor an explanation of isolation flags.
TransPort requires OS support for transparent proxies, such as BSDs' pf or Linux's IPTables. If you're planning to use Tor as a transparent proxy for a network, you'll want to examine and change VirtualAddrNetwork from the default setting. (Default: 0)
TransProxyTypedefault|TPROXY|ipfw|pf-divert—TransProxyTypemay only be enabled when there is a transparent proxy listener enabled. no cap.
Set this to "TPROXY" if you wish to be able to use the TPROXY Linux module to transparently proxy connections that are configured using the TransPort option. Detailed information on how to configure the TPROXY feature can be found in the Linux kernel source tree in the file Documentation/networking/tproxy.txt.
Set this option to "ipfw" to use the FreeBSD ipfw interface.
On *BSD operating systems when using pf, set this to "pf-divert" to take advantage of divert-to rules, which do not modify the packets like rdr-to rules do. Detailed information on how to configure pf to use divert-to rules can be found in the pf.conf(5) manual page. On OpenBSD, divert-to is available to use on versions greater than or equal to OpenBSD 4.4.
Set this to "default", or leave it unconfigured, to use regular IPTables on Linux, or to use pf rdr-to rules on *BSD systems.
(Default: "default")
UpdateBridgesFromAuthority0|1— when set (along withUseBridges), Tor will try to fetch bridge descriptors from the configured bridge authorities when feasible. It will fall back to a direct request if the authority responds with a404. (Default:0)UseBridges0|1— when set, Tor will fetch descriptors for each bridge listed in the"Bridge"config lines, and use these relays as both entry guards and directory guards. Double duty, main character energy. (Default:0)UseEntryGuards0|1— if this option is set to1, we pick a few long-term entry servers, and try to stick with them. This is desirable because constantly changing servers increases the odds that an adversary who owns some servers will observe a fraction of your paths. yikes. Entry Guards can not be used by Directory Authorities or Single Onion Services. In these cases, this option is ignored. (Default:1)UseGuardFraction0|1|auto— this option specifies whether clients should use the guardfraction information found in the consensus during path selection. If it's set toauto, clients will do what theUseGuardFractionconsensus parameter tells them to do. (Default:auto)GuardLifetimeN days|weeks|months— ifUseEntryGuardsis set, this is the minimum time to keep a guard on our guard list before picking a new one. If less than one day, we use defaults from the consensus directory. (Default:0)NumDirectoryGuardsNUM— ifUseEntryGuardsis set to1, we try to make sure we have at leastNUMrouters to use as directory guards. If this option is set to0, use the value from theguard-n-primary-dir-guards-to-useconsensus parameter, and default to3if the consensus parameter isn't set. (Default:0)NumEntryGuardsNUM— ifUseEntryGuardsis set to1, we will try to pick a total ofNUMrouters as long-term entries for our circuits. IfNUMis0, we try to learn the number from theguard-n-primary-guards-to-useconsensus parameter, and default to1if the consensus parameter isn't set. (Default:0)NumPrimaryGuardsNUM— ifUseEntryGuardsis set to1, we will try to pickNUMrouters for our primary guard list, which is the set of routers we strongly prefer when connecting to the Tor network. IfNUMis0, we try to learn the number from theguard-n-primary-guardsconsensus parameter, and default to3if the consensus parameter isn't set. (Default:0)VanguardsLiteEnabled0|1|auto— this option specifies whether clients should use the vanguards-lite subsystem to protect against guard discovery attacks. If it's set toauto, clients will do what thevanguards-lite-enabledconsensus parameter tells them to do, and will default to enable the subsystem if the consensus parameter isn't set. (Default:auto)UseMicrodescriptors0|1|auto— microdescriptors are a smaller version of the information that Tor needs in order to build its circuits. Using microdescriptors makes Tor clients download less directory information, thus saving bandwidth 💚. Directory caches need to fetch regular descriptors and microdescriptors, so this option doesn't save any bandwidth for them. sad for them tbh. For legacy reasons,autois accepted, but it has the same effect as1. (Default:auto)VirtualAddrNetworkIPv4IPv4Address/bits, andVirtualAddrNetworkIPv6[IPv6Address]/bits— when Tor needs to assign a virtual (unused) address because of aMAPADDRESScommand from the controller or theAutomapHostsOnResolvefeature, Tor picks an unassigned address from this range. (Defaults:127.192.0.0/10and[FE80::]/10respectively.)
When providing proxy server service to a network of computers using a tool like dns-proxy-tor, change the IPv4 network to "10.192.0.0/10" or "172.16.0.0/12" and change the IPv6 network to "[FC00::]/7". The default VirtualAddrNetwork address ranges on a properly configured machine will route to the loopback or link-local interface. The maximum number of bits for the network prefix is set to 104 for IPv6 and 16 for IPv4. However, a larger network (that is, one with a smaller prefix length) is preferable, since it reduces the chances for an attacker to guess the used IP. For local use, no change to the default VirtualAddrNetwork setting is needed. we love an easy setup.
✨ CIRCUIT TIMEOUT OPTIONS ✨
The following options are useful for configuring timeouts related to building Tor circuits and using them: 🦋 buckle up bestie, we're timing things.
CircuitsAvailableTimeoutNUM— Tor will attempt to keep at least one open, unused circuit available for this amount of time. This option governs how long idle circuits are kept open, as well as the amount of time Tor will keep a circuit open to each of the recently used ports. This way when the Tor client is entirely idle, it can expire all of its circuits, and then expire its TLS connections. Note that the actual timeout value is uniformly randomized from the specified value to twice that amount. (Default: 30 minutes; Max: 24 hours) ⏳LearnCircuitBuildTimeout0|1— if0,CircuitBuildTimeoutadaptive learning is disabled. (Default:1)
If adaptive history contains only abandoned measurements, Tor clears that history and restarts learning with conservative timeout values. This does not change guard selection or guarantee that a stalled path will recover. During continuing circuit-build failures, Tor logs a process-wide expiry summary at most once every five minutes. Hop-stage and channel-state counts are separate totals, not correlated observations. Failed first-hop connections are counted separately as connection_failed; they have not yet started circuit build timing.
To diagnose persistent failures with fixed timeouts, restart with LearnCircuitBuildTimeout 0 and CircuitBuildTimeout 60 while preserving the existing state file. Purpose-specific timeout rules still apply; a path already failing after 60 seconds might remain unusable. Changing these options in a running process is not equivalent to this restart diagnostic. When reporting failures, share the aggregate expiry/recovery summaries and, if requested, only TotalBuildTimes and CircuitBuildAbandonedCount from state. Do not share the full state file, histogram bins, or circuit path logs by default. fr fr, privacy first.
CircuitBuildTimeoutNUM— try for at mostNUMseconds when building circuits. If the circuit isn't open in that time, give up on it. IfLearnCircuitBuildTimeoutis1, this value serves as the initial value to use before a timeout is learned. IfLearnCircuitBuildTimeoutis0, this value is the only value used. (Default:60seconds)CircuitStreamTimeoutNUM— if non-zero, this option overrides our internal timeout schedule for how many seconds until we detach a stream from a circuit and try a new circuit. If your network is particularly slow (😭 we've all been there), you might want to set this to a number like60. (Default:0)SocksTimeoutNUM— let a socks connection waitNUMseconds handshaking, andNUMseconds unattached waiting for an appropriate circuit, before we fail it. (Default: 2 minutes) patience is a virtue, etc.
✨ DORMANT MODE OPTIONS ✨
okay so like, Tor can go into dormancy to save power and bandwidth, bestie 💤. These settings control when it slips into and out of that sleepy vibe:
DormantCanceledByStartup0|1— By default, Tor boots up active if it was active when it last shut down, and dormant if it was dormant. But if you set this to true, Tor treats every startup as user activity, so it'll never start in dormant mode, even if it's been chilling unused for ages. (Default:0)
Note: Packagers and application devs, please change this only with extreme caution — it can create spurious traffic on the network. Only use this if Tor is started by an actual user action (like clicking an app or running a command), not if it's launched for some other reason (like a startup process or an app that launches itself on every login). 🚨
DormantClientTimeoutN minutes|hours|days|weeks— If Tor goes this long without any client activity, it enters a dormant state where it stops building automatic circuits and doesn't fetch directory info. Servers and onion services are not affected. Must be at least 10 minutes. (Default:24 hours)DormantOnFirstStartup0|1— If true, the first time Tor starts with a fresh DataDirectory, it starts dormant and does nothing until the user makes a request. (This is recommended if you're installing a Tor client for someone who might not actually use it.) If false, Tor bootstraps on first start, whether it sees a user request or not.
After that first start, Tor begins dormant if it was dormant before, and not otherwise. (Default: 0)
DormantTimeoutDisabledByIdleStreams0|1— If true, any open client stream (even one not reading or writing) counts as client activity forDormantClientTimeout. If false, only network activity counts. (Default:1)DormantTimeoutEnabled0|1— If false, no amount of idleness will make Tor go dormant. Setting this to zero is only recommended for special-purpose apps that need the Tor binary for something other than sending/receiving Tor traffic. (Default:1)
🌈 NODE SELECTION OPTIONS 🌈
These options let you restrict which nodes a tor client (or onion service) can use when building a circuit. Heads up: they can weaken your anonymity by making your client behave differently from other Tor clients, so use with care 👀:
EntryNodesnode,node,... — A list of identity fingerprints and country codes for nodes to use as the first hop in your normal circuits. Normal circuits are all circuits except direct connections to directory servers. TheBridgeoption overrides this; if you've set up bridges andUseBridgesis 1, the Bridges become your entry nodes.
You can list this multiple times — values from multiple lines get spliced together.
The ExcludeNodes option overrides this: any node in both EntryNodes and ExcludeNodes is treated as excluded. See ExcludeNodes for how to specify nodes.
ExcludeNodesnode,node,... — A list of identity fingerprints, country codes, and address patterns of nodes to avoid when building a circuit. Country codes are 2-letter ISO3166 codes and must be in braces; fingerprints can have a dollar sign before them. (Example:ExcludeNodes ABCD1234CDEF5678ABCD1234CDEF5678ABCD1234, {cc}, 255.254.0.0/8)
You can list this multiple times — values get spliced together.
By default, this is treated as a preference that Tor can override to keep working. For example, if you try to connect to a hidden service but you've excluded all its introduction points, Tor will connect to one anyway. If you don't want that, set StrictNodes (documented below).
Also note: if you're a relay, this (and the other node selection options) only affects the circuits Tor builds for you. Clients can still build circuits through you to any node. Controllers can tell Tor to build circuits through any node.
Country codes are case-insensitive. The code {??} means nodes whose country can't be identified. No country code, including {??}, works if no GeoIPFile can be loaded. See also GeoIPExcludeUnknown below.
ExcludeExitNodesnode,node,... — A list of identity fingerprints, country codes, and address patterns of nodes to never use as an exit node — that is, a node that delivers traffic for you outside the Tor network. Any node listed inExcludeNodesis automatically in this list too. SeeExcludeNodesfor how to specify nodes. Also check the caveats onExitNodesbelow.
You can list this multiple times — values get spliced together.
ExitNodesnode,node,... — A list of identity fingerprints, country codes, and address patterns of nodes to use as an exit node — that is, a node that delivers traffic for you outside the Tor network. SeeExcludeNodesfor how to specify nodes.
You can list this multiple times — values get spliced together.
Note: if you list too few nodes here, or exclude too many exits with ExcludeExitNodes, you can degrade functionality. For example, if none of your exits allow traffic on port 80 or 443, you won't be able to browse the web.
Also, not every circuit delivers traffic outside the Tor network. It's normal to see non-exit circuits (like those for hidden services, directory fetches, relay reachability self-tests, etc.) that end at a non-exit node. To keep a node from being used entirely, see ExcludeNodes and StrictNodes.
The ExcludeNodes option overrides this: any node in both ExitNodes and ExcludeNodes is treated as excluded.
The .exit address notation, if enabled via MapAddress, overrides this option.
GeoIPExcludeUnknown0|1|auto— If set toauto, then whenever any country code is set inExcludeNodesorExcludeExitNodes, all nodes with unknown country ({??}and possibly{A1}) are treated as excluded too. If set to1, all unknown countries are excluded inExcludeNodesandExcludeExitNodes. This option does nothing if a GeoIP file isn't configured or can't be found. (Default:auto)HSLayer2Nodesnode,node,... — A list of identity fingerprints, nicknames, country codes, and address patterns of nodes allowed as the second hop in all client or service-side Onion Service circuits. This mitigates attacks where an adversary runs middle nodes and tricks your client or service into creating many circuits to discover your primary guard node. (Default: Any node in the network may be used in the second hop.)
(Example: HSLayer2Nodes ABCD1234CDEF5678ABCD1234CDEF5678ABCD1234, {cc}, 255.254.0.0/8)
You can list this multiple times — values get spliced together.
When this is set, the resulting hidden service paths look like:
C - G - L2 - M - Rend
C - G - L2 - M - HSDir
C - G - L2 - M - Intro
S - G - L2 - M - Rend
S - G - L2 - M - HSDir
S - G - L2 - M - Introwhere C is this client, S is the service, G is the Guard node, L2 is a node from this option, and M is a random middle node. Rend, HSDir, and Intro point selection is not affected by this option.
This can be combined with HSLayer3Nodes to create paths like:
C - G - L2 - L3 - Rend
C - G - L2 - L3 - M - HSDir
C - G - L2 - L3 - M - Intro
S - G - L2 - L3 - M - Rend
S - G - L2 - L3 - HSDir
S - G - L2 - L3 - IntroExcludeNodes have higher priority than HSLayer2Nodes, so nodes in ExcludeNodes will not be picked.
When either this option or HSLayer3Nodes is set, the /16 subnet and node family restrictions are removed for hidden service circuits. Also, we allow the guard node to be the Rend, HSDir, and IP node, and the hop before it. This prevents the adversary from inferring info about our guard, layer2, and layer3 node choices later in the path.
This option is meant to be managed by a Tor controller like https://github.com/mikeperry-tor/vanguards that selects and updates the node set for you. So it doesn't do load balancing if fewer than 20 nodes are selected, and if no nodes in HSLayer2Nodes are available, Tor will not work. Please use extreme care if setting this manually.
HSLayer3Nodesnode,node,... — A list of identity fingerprints, nicknames, country codes, and address patterns of nodes allowed as the third hop in all client and service-side Onion Service circuits. This mitigates attacks where an adversary runs middle nodes and tricks your client or service into creating many circuits to discover your primary or Layer2 guard nodes. (Default: Any node in the network may be used in the third hop.)
(Example: HSLayer3Nodes ABCD1234CDEF5678ABCD1234CDEF5678ABCD1234, {cc}, 255.254.0.0/8)
You can list this multiple times — values get spliced together.
When set by itself, the resulting hidden service paths look like:
C - G - M - L3 - Rend
C - G - M - L3 - M - HSDir
C - G - M - L3 - M - Intro
S - G - M - L3 - M - Rend
S - G - M - L3 - HSDir
S - G - M - L3 - Introwhere C is this client, S is the service, G is the Guard node, L2 is a node from this option, and M is a random middle node. Rend, HSDir, and Intro point selection is not affected by this option.
While it's possible to use this option alone, it should be combined with HSLayer2Nodes to create paths like:
C - G - L2 - L3 - RendAlright bestie, here's the actual layout of the hidden-service circuits this option is shaping — the letters are hops in the path (C = client, S = service, G = guard, L2 = layer 2, L3 = layer 3, M = middle, and the last one is the role of that final hop):
C - G - L2 - L3 - M - HSDir
C - G - L2 - L3 - M - Intro
S - G - L2 - L3 - M - Rend
S - G - L2 - L3 - HSDir
S - G - L2 - L3 - IntroNow the tea ☕: ExcludeNodes has higher priority than HSLayer3Nodes — so like, if a node is chillin' in ExcludeNodes, it is never getting picked, no matter how cute it looks in your layer-3 list. Periodt.
When either this option or HSLayer2Nodes is set, the /16 subnet and node family restrictions get removed for hidden service circuits. And on top of that, the guard node is allowed to also be the Rend, HSDir, and IP node, plus the hop right before it. Why tho? So an adversary can't sniff out info about our guard, layer2, and layer3 node choices at later points in the path. Sneaky-proofing, basically. 🕵️♀️
This option is meant to be managed by a Tor controller — like https://github.com/mikeperry-tor/vanguards — which selects and updates this set of nodes for you. Which is why it does not do load balancing if fewer than 20 nodes are selected, and if no nodes in HSLayer3Nodes are currently available for use, Tor will not work. So please, please use extreme care if you are setting this option manually. 🙏
MiddleNodesnode,node,... — a list of identity fingerprints and country codes of nodes to use for the "middle" hops in your normal circuits. Normal circuits include all circuits except for direct connections to directory servers, and middle hops are all hops other than exit and entry.
This option can appear multiple times: the values from multiple lines are spliced together.
Heads up — this is an experimental feature that's meant for researchers and developers to test new features in the Tor network safely. Using it without care will strongly influence your anonymity. Other tor features may not work with MiddleNodes. And this feature might get removed in the future. 💀
- The
HSLayer2NodeandHSLayer3Nodeoptions override this option for onion service circuits, if they are set. The vanguards addon will read this option, and if set, it will setHSLayer2NodesandHSLayer3Nodesto nodes from this set. - The
ExcludeNodesoption overrides this option: any node listed in bothMiddleNodesandExcludeNodesis treated as excluded. See<<ExcludeNodes,ExcludeNodes>>for more information on how to specify nodes.
NodeFamilynode,node,... — the Tor servers, defined by their identity fingerprints, constitute a "family" of similar or co-administered servers, so never use any two of them in the same circuit. Defining aNodeFamilyis only needed when a server doesn't list the family itself (withMyFamily). You can use this option multiple times; each instance defines a separate family. And not just nodes — you can also list IP addresses and ranges and country codes in{curly braces}. SeeExcludeNodesfor more info on how to specify nodes.StrictNodes0|1 — if it's set to1, Tor will treat solely theExcludeNodesoption as a requirement to follow for all the circuits you generate, even if doing so will break functionality for you. (Note:StrictNodesdoes not apply toExcludeExitNodes,ExitNodes,MiddleNodes, orMapAddress.) If it's set to0, Tor will still try to avoid nodes in theExcludeNodeslist, but it'll err on the side of avoiding unexpected errors. Specifically,StrictNodes 0tells Tor it's okay to use an excluded node when it's necessary to perform relay reachability self-tests, connect to a hidden service, provide a hidden service to a client, fulfill a.exitrequest, upload directory information, or download directory information. (Default:0)
🌸 SERVER OPTIONS 🌸
These ones only matter if you're running a server — that is, if ORPort is non-zero. Relay girlies, this is your section. 📡✨
AccountingMaxN bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits — limits the max number of bytes sent and received within a set time period, using a given calculation rule (peepAccountingStartandAccountingRule). Super useful if you need to stay under a specific bandwidth. By default, the number used for calculation is the max of either the bytes sent or received. So like, withAccountingMaxset to1 TByte, a server could send900 GBytesand receive800 GBytesand keep on running — it'll only hibernate once one of the two reaches1 TByte. You can switch this to use the sum of both bytes received and sent by setting theAccountingRuleoption to"sum"(total bandwidth in/out). When the number of bytes remaining gets low, Tor will stop accepting new connections and circuits. When the number of bytes is exhausted, Tor hibernates until some time in the next accounting period. And to prevent all servers from waking at the same time, Tor also waits until a random point in each period before waking up. If you've got bandwidth cost issues, enabling hibernation is preferable to setting a low bandwidth — because it gives users a collection of fast servers that are up some of the time, which is way more useful than a set of slow servers that are always "available".
Note that (as also described in the Bandwidth section) Tor uses powers of two, not powers of ten: 1 GByte is 1024*1024*1024, not one billion. Be careful — some internet service providers might count GBytes differently. 😭
AccountingRulesum|max|in|out — how we determine when ourAccountingMaxhas been reached (aka when we should hibernate) during a time interval. Set to"max"to calculate using the higher of either the sent or received bytes (this is the default functionality). Set to"sum"to calculate using the sent plus received bytes. Set to"in"to calculate using only the received bytes. Set to"out"to calculate using only the sent bytes. (Default:max)AccountingStartday|week|month [day] HH:MM — specify how long accounting periods last. Ifmonthis given, each accounting period runs from the timeHH:MMon the dayth day of one month to the same day and time of the next. The relay will go at full speed, use all the quota you specify, then hibernate for the rest of the period. (The day must be between1and28.) Ifweekis given, each accounting period runs from the timeHH:MMof the dayth day of one week to the same day and time of the next week, with Monday as day1and Sunday as day7. Ifdayis given, each accounting period runs from the timeHH:MMeach day to the same time on the next day. All times are local, and given in 24-hour time. (Default:"month 1 0:00")Addressaddress — the address of this server, or a fully qualified domain name of this server that resolves to an address. You can leave this unset and Tor will try to guess your address. If a domain name is provided, Tor will attempt to resolve it and use the underlying IPv4/IPv6 address as its publish address (taking precedence over theORPortconfiguration). The publish address is the one used to tell clients and other servers where to find your Tor server — it doesn't affect the address that your server binds to. To bind to a different address, use theORPortandOutboundBindAddressoptions.AddressDisableIPv60|1 — by default, Tor will attempt to find the IPv6 of the relay if there is noIPv4OnlyORPort. If set, this option disables IPv6 auto discovery. This disables IPv6 address resolution, IPv6 ORPorts, and IPv6 reachability checks. Also, the relay won't publish an IPv6 ORPort in its descriptor. (Default:0)AssumeReachable0|1 — this option is used when bootstrapping a new Tor network. If set to1, don't do self-reachability testing; just upload your server descriptor immediately. (Default:0)AssumeReachableIPv60|1|auto — likeAssumeReachable, but affects only the relay's own IPv6 ORPort. If this value is set to"auto", then Tor will look atAssumeReachableinstead. (Default:auto)BridgeRelay0|1 — sets the relay to act as a "bridge" with respect to relaying connections from bridge users to the Tor network. It mainly causes Tor to publish a server descriptor to the bridge database, rather than to the public directory authorities.
Note: make sure that no MyFamily lines are present in your torrc when relay is configured in bridge mode. 👀
BridgeDistributionstring — if set along withBridgeRelay, Tor will include a new line in its bridge descriptor which indicates to the BridgeDB service how it would like its bridge address to be given out. Set it to"none"if you want BridgeDB to avoid distributing your bridge address, or"any"to let BridgeDB decide. See https://bridges.torproject.org/info for a more up-to-date list of options. (Default:any)ContactInfoemail_address — administrative contact information for this relay or bridge. This line can be used to contact you if your relay or bridge is misconfigured or something else goes wrong. Note that we archive and publish all descriptors containing these lines and that Google indexes them, so spammers might also collect them. You may want to obscure the fact that it's an email address and/or generate a new address for this purpose. 🙈
ContactInfo must be set to a working address if you run more than one relay or bridge. (Really, everybody running a relay or bridge should set it.)
DisableOOSCheck0|1 — this option disables the code that closes connections when Tor notices that it is running low on sockets. Right now, it's on by default, since the existing out-of-sockets mechanism tends to kill OR connections more than it should. (Default:1)ExitPolicypolicy,policy,... — set an exit policy for this server. Each policy is of the form"accept[6]|reject[6] ADDR[/MASK][:PORT]". If/MASKis omitted then this policy just applies to the host given. Instead of giving a host or network you can also use"*"to denote the universe (0.0.0.0/0and::/0), or*4to denote all IPv4 addresses, and*6to denote all IPv6 addresses.PORTcan be a single port number, an interval of ports"FROM_PORT-TO_PORT", or"*". IfPORTis omitted, that means"*".
For example, "accept 18.7.22.69:*,reject 18.0.0.0/8:*,accept *:*" would reject any IPv4 traffic destined for MIT except for web.mit.edu, and accept any other IPv4 or IPv6 traffic. Slay. 💅
Okay so Tor also does IPv6 exit policy entries, bestie, and they're lowkey iconic. Like, "reject6 [FC00::]/7:*" rejects every destination that shares its 7 most significant bit prefix with the address FC00::. And respectively, "accept6 [C000::]/3:*" accepts all the destinations sharing 3 most significant bit prefix with C000::. 💅
But listen — accept6 and reject6 ONLY spit out IPv6 exit policy entries. If you feed them an IPv4 address it gets straight up ignored and you get a warning, bestie. Not the vibe. Meanwhile plain accept/reject takes either IPv4 or IPv6, flexible queen. Use *4 as the IPv4 wildcard address, and *6 as the IPv6 wildcard address. And accept/reject * expands out to matching IPv4 and IPv6 wildcard address rules. periodt.
So if you wanna specify all the IPv4 and IPv6 internal and link-local networks — and we mean 0.0.0.0/8, 169.254.0.0/16, 127.0.0.0/8, 192.168.0.0/16, 10.0.0.0/8, 172.16.0.0/12, [::]/8, [FC00::]/7, [FE80::]/10, [FEC0::]/10, [FF00::]/8, and [::]/127 — you can just use the "private" alias instead of typing an address. 🫶 And yes, "private" always produces rules for IPv4 and IPv6 addresses, even when you use it with accept6/reject6. consistent queen.
Private addresses get rejected by default — like, at the very beginning of your exit policy — along with any configured primary public IPv4 and IPv6 addresses. They stay rejected unless you set the ExitPolicyRejectPrivate config option to 0. So, for example, once you've done that little edit, you could allow HTTP to 127.0.0.1 and block every other connection to internal networks with "accept 127.0.0.1:80,reject private:*" — though ngl that may also allow connections to your own computer that are addressed to its public (external) IP address. 🙈 See RFC 1918 and RFC 3330 for more details about internal and reserved IP address space. And if you want to block every address on the relay, even the ones that aren't advertised in the descriptor, go check out ExitPolicyRejectLocalInterfaces. iykyk.
Also — this directive can be specified multiple times, so you don't have to cram it all onto one line. your paragraph breaks are safe here. 🎀
Policies get considered first to last, and the first match wins. So if you want the same ports open on IPv4 and IPv6, write your rules using accept/reject *. If you want different ports on IPv4 and IPv6, write your IPv6 rules using accept6/reject6 *6, and your IPv4 rules using accept/reject *4. And if you wanna _replace_ the default exit policy, end your exit policy with either a reject *:* or an accept *:*. Otherwise you're _augmenting_ it — literally prepending to the default exit policy. 👑
Wanna run a reduced exit policy instead of the default one? Set "ReducedExitPolicy 1". Wanna fully replace the default exit policy with your own custom one? End your exit policy with either a reject : or an accept :. Otherwise, you guessed it, you're augmenting (prepending to) the default or reduced exit policy.
And here's the default exit policy, serving looks:
reject *:25
reject *:119
reject *:135-139
reject *:445
reject *:563
reject *:1214
reject *:4661-4666
reject *:6346-6429
reject *:6699
reject *:6881-6999
accept *:*Since the default exit policy uses accept/reject *, it applies to both IPv4 and IPv6 addresses. balance. ✨
ExitPolicyRejectLocalInterfaces 0|1— reject every IPv4 and IPv6 address the relay knows about, right at the beginning of your exit policy. That includes anyOutboundBindAddress, the bind addresses of any port options (likeControlPortorDNSPort), and any public IPv4 and IPv6 addresses on any interface on the relay. (Side note: ifIPv6Exitisn't set, all IPv6 addresses will get rejected anyway.) See the entry above onExitPolicy. This one is off by default because it lists all public relay IP addresses in theExitPolicy— even the ones relay operators might prefer not to disclose. 👀 (Default:0)ExitPolicyRejectPrivate 0|1— reject all private (local) networks, plus the relay's advertised public IPv4 and IPv6 addresses, at the beginning of your exit policy. See the entry above onExitPolicy. (Default:1)ExitRelay 0|1|auto— tells Tor whether to run as an exit relay. If Tor is running as a non-bridge server andExitRelayis1, then Tor lets traffic exit according to theExitPolicyoption, theReducedExitPolicyoption, or the defaultExitPolicy(if no other exit policy option is specified) 🚪✨
If ExitRelay is 0, no traffic is allowed to exit, and the ExitPolicy, ReducedExitPolicy, and IPv6Exit options are all just ignored.
If ExitRelay is "auto", Tor goes and checks the ExitPolicy, ReducedExitPolicy, and IPv6Exit options. If at least one of them is set, Tor behaves as if ExitRelay were 1. If none of those exit policy options are set, Tor behaves as if ExitRelay were 0. (Default: auto)
ReevaluateExitPolicy 0|1— if set, re-evaluate the exit policy on existing connections when reloading configuration. 🔄
When the exit policy of an exit node changes while reloading config, connections made before that change could violate the new policy. By setting this to 1, Tor checks whether such connections exist and marks them for termination. bye bestie 💀 (Default: 0)
ExtendAllowPrivateAddresses 0|1— when this is enabled, Tor will connect to relays onlocalhost, RFC1918 addresses, and so on. Specifically: Tor makes direct OR connections, and Tor routers allow EXTEND requests, to these private addresses. (Tor will always allow connections to bridges, proxies, and pluggable transports configured on private addresses, no matter what.) Enabling this option can create security issues, so you should probably leave it off, fr fr. ⚠️ (Default:0)GeoIPFile filename— a filename containing IPv4 GeoIP data, for use with by-country statistics. 🗺️GeoIPv6File filename— a filename containing IPv6 GeoIP data, for use with by-country statistics.HeartbeatPeriod N minutes|hours|days|weeks— log a heartbeat message everyHeartbeatPeriodseconds. It's a log level notice message, designed to let you know your Tor server is still alive and doing useful things. 💓 Setting this to0disables the heartbeat entirely. Otherwise, it must be at least 30 minutes. (Default:6 hours)IPv6Exit 0|1— if set, and we're an exit node, allow clients to use us for IPv6 traffic. When this option is set andExitRelayisauto, we act as ifExitRelayis1. (Default:0)KeyDirectory DIR— store secret keys inDIR. Can not be changed while tor is running. 🙅 (Default: the"keys"subdirectory ofDataDirectory.)KeyDirectoryGroupReadable 0|1|auto— if this is set to0, don't let the filesystem group read theKeyDirectory. If it's set to1, make theKeyDirectoryreadable by the default GID. If it's"auto", then we use the setting forDataDirectoryGroupReadablewhen theKeyDirectoryis the same as theDataDirectory, and0otherwise. (Default:auto)MainloopStats 0|1— log main loop statistics everyHeartbeatPeriodseconds. It's a log level notice message designed to help developers instrumenting Tor's main event loop, so it's giving dev-tools energy. 🛠️ (Default:0)MaxHSDirCacheBytes N bytes|KBytes|MBytes|GBytes— this option sets a threshold for Hidden Service Directory memory consumption, above which your Tor relay starts pruning the least-frequently accessed hidden service descriptors from the relay's HSDir cache. 🌸 If set to0, it defaults to 20% ofMaxMemInQueues. (Default:0)
This pruning used to be done as part of MaxMemInQueues, but it's been decoupled so you can get more fine-grained control of descriptor cache size under DDoS conditions. glow-up moment.
MaxMemInQueues N bytes|KBytes|MBytes|GBytes— this option sets a threshold above which Tor assumes it needs to stop queueing or buffering data, because it's about to run out of memory. 😭 If it hits this threshold, it starts killing circuits until it has recovered at least 10% of this memory. Do not set this option too low, or your relay may be unreliable under load. This option only affects some queues, so the actual process size will be larger than this (sneaky). If it's set to0, Tor tries to pick a reasonable default based on your system's physical memory. (Default:0)MaxOnionQueueDelay NUM [msec|second]— if we have more onionskins queued for processing than we can get through in this amount of time, reject the new ones. 🧅🚫 (Default:1750 msec)MyFamily fingerprint,fingerprint,...— declare that this Tor relay is controlled or administered by a group or organization that's identical or similar to that of the other relays, defined by their (possibly$-prefixed) identity fingerprints. 👯 This option can be repeated many times, for convenience when defining large families: all fingerprints in allMyFamilylines get merged into one list. When two relays both declare they're in the same 'family', Tor clients will not use them in the same circuit. (Each relay only needs to list the other servers in its family; it doesn't need to list itself, but it won't hurt if it does. self-love 🫶) Do not list any bridge relay as it would compromise its concealment. 🤫
If you run more than one relay, the MyFamily option on each relay must list all the other relays, as described above.
Note: do not use MyFamily when configuring your Tor instance as a bridge.
FamilyId ident— configure this relay to be part of a family identified by a shared secret family key with the given key identity. A corresponding family key must be stored in the relay's key directory, with a filename ending with".secret_family_key". 🔑 This option can appear multiple times. Family keys are generated with"--keygen-family"; this also generates the value you should use in theidentfield in a file ending with".public_family_id". For info on generating and installing a family key, seehttps://community.torproject.org/relay/setup/post-install/family-ids/
In the future, this will be the preferred way for relays to advertise family membership. But for now, relay families should configure both this option and MyFamily, so older clients will still recognize the relays' family membership. gotta be backwards compatible bestie 💖
(Note that if the seccomp2 Sandbox feature is enabled, it is not possible to change the key filenames while Tor is running.)
FamilyId *— okay so this one's the ✨automatic✨ version: your relay joins every family whose family ID key shows up in the family key directory. Only files ending in.secret_family_keyget counted, bestie, nothing else. The slay part? If the family key gets rotated, you don't have to touch your config at all. The not-so-slay part? You're wayyy more likely to accidentally be using a different set of family keys than the ones you thought you signed up for. 🙈 So like, convenient but spicy.FamilyKeyDirectory directory:— when Tor goes looking for family ID keys, use thisdirectoryinstead of the normal key directory. Simple as. 💅Nickname name— sets your server's nickname toname. Rules are strict, bestie: 1 to 19 characters inclusive, and only these characters allowed:[a-zA-Z0-9]. If you don't set one, it just defaults toUnnamed. Also don't stress about uniqueness, because relays can always be uniquely identified by their identity fingerprints. 👑NumCPUs num— how many processes to run at once for decrypting onionskins and all the other parallelizable stuff. If you set it to0, Tor tries to sniff out how many CPUs you've got, and if it literally cannot tell, it defaults to1. (Default:0)OfflineMasterKey 0|1— if this is non-zero, your Tor relay will never generate or load its master secret key. Not once. Instead you have to go usetor --keygento manage the permanent ed25519 master identity key, plus the matching temporary signing keys and certificates. (Default:0)ORPort [address:]PORT|auto [flags]— this advertises the port you're listening on for connections from Tor clients and servers. It's required if you wanna be a Tor server, no way around it. Set it to"auto"and Tor will pick a port for you. Set it to0and you run no ORPort at all. This option can occur more than once, so go off. (Default:0)
Tor recognizes these flags on each ORPort:
NoAdvertise— by default we bind to a port and tell our users about it. WithNoAdvertise, we don't tell anybody, but we still listen. Super useful when the port everyone actually connects to (say, one opened on your firewall) lives somewhere else entirely.NoListen— flip of the above: by default we bind to a port and tell our users about it. WithNoListen, we don't bind, but we advertise anyway. Useful if something else (like a firewall's port forwarding config) is the thing making connections reach us.IPv4Only— if the address is absent, or resolves to both an IPv4 and an IPv6 address, only listen on the IPv4 one.IPv6Only— same deal: if the address is absent, or resolves to both an IPv4 and an IPv6 address, only listen on the IPv6 one.
For obvious reasons 🌸, NoAdvertise and NoListen are mutually exclusive, and IPv4Only and IPv6Only are mutually exclusive. You can't have both, periodt.
PublishServerDescriptor 0|1|v3|bridge,...— this decides which descriptors Tor publishes when it's acting as a relay. You can pick multiple arguments, comma-separated. 🎀
Set it to 0 and Tor publishes its descriptors to no directories. (Great if you're just testing your server out, or if you've got a Tor controller handling directory publishing for you.) Otherwise Tor publishes descriptors of all the type(s) you specified. The default is "1", which means "if running as a relay or bridge, publish descriptors to the appropriate authorities". The other options are "v3", meaning "publish as if you're a relay", and "bridge", meaning "publish as if you're a bridge".
ReducedExitPolicy 0|1— if set, use a reduced exit policy instead of the default one.
The reduced exit policy is the alternative to the default exit policy. It allows as many Internet services as possible while still blocking the majority of TCP ports. Currently it allows approximately 65 ports. This lowers the odds your node gets used for peer-to-peer applications. 🦋
The reduced exit policy is:
accept *:20-21
accept *:22
accept *:23
accept *:43
accept *:53
accept *:79
accept *:80-81
accept *:88
accept *:110
accept *:143
accept *:194
accept *:220
accept *:389
accept *:443
accept *:464
accept *:465
accept *:531
accept *:543-544
accept *:554
accept *:563
accept *:587
accept *:636
accept *:706
accept *:749
accept *:873
accept *:902-904
accept *:981
accept *:989-990
accept *:991
accept *:992
accept *:993
accept *:994
accept *:995
accept *:1194
accept *:1220
accept *:1293
accept *:1500
accept *:1533
accept *:1677
accept *:1723
accept *:1755
accept *:1863
accept *:2082
accept *:2083
accept *:2086-2087
accept *:2095-2096
accept *:2102-2104
accept *:3128
accept *:3389
accept *:3690
accept *:4321
accept *:4643
accept *:5050
accept *:5190
accept *:5222-5223
accept *:5228
accept *:5900
accept *:6660-6669
accept *:6679
accept *:6697
accept *:8000
accept *:8008
accept *:8074
accept *:8080
accept *:8082
accept *:8087-8088
accept *:8232-8233
accept *:8332-8333
accept *:8443
accept *:8888
accept *:9418
accept *:9999
accept *:10000
accept *:11371
accept *:19294
accept *:19638
accept *:50002
accept *:64738
reject *:*(Default: 0)
RefuseUnknownExits 0|1|auto— stops nodes that don't show up in the consensus from exiting through this relay. If the option is1, we always block exit attempts from those nodes. If it's0, we never do. And if it's"auto", we do whatever the authorities suggest in the consensus, and we block if the consensus is quiet about it. (Default:auto)ServerDNSAllowBrokenConfig 0|1— if this is false, Tor exits immediately the second there are problems parsing the system DNS config or connecting to nameservers. Otherwise, Tor keeps periodically retrying the system nameservers until it eventually gets through. (Default:1)ServerDNSAllowNonRFC953Hostnames 0|1— turn this off and Tor won't try to resolve hostnames that contain illegal characters (like@and:), instead of shipping them off to an exit node to be resolved. This helps trap accidental attempts to resolve URLs and such. Heads up: this option only affects the name lookups your server does on behalf of clients. (Default:0)ServerDNSDetectHijacking 0|1— set this to1and we'll periodically test whether your local nameservers have been configured to hijack failing DNS requests (usually straight to some advertising site 😭). If they are, we'll try to fix it. Again: only affects the name lookups your server does on behalf of clients. (Default:1)ServerDNSRandomizeCase 0|1— when this is set, Tor randomizes the case of each character in outgoing DNS requests and makes sure the case matches in DNS replies. This so-called "0x20 hack" helps resist some types of DNS poisoning attack. For more info, see "Increased DNS Forgery Resistance through 0x20-Bit Encoding". And yep, only affects the name lookups your server does on behalf of clients. (Default:1)ServerDNSResolvConfFile filename— overrides the default DNS configuration with whatever's infilename. The file format is the same as the standard Unixresolv.conffile (7). This option, like all other ServerDNS options, only affects name lookups your server does on behalf of clients. (Defaults to using the system DNS configuration, or a localhost DNS service if no nameservers are found in a given configuration.)ServerDNSSearchDomains 0|1— set to1and we'll search for addresses in the local search domain. Example: if this system is configured to believe it's in"example.com", and a client tries to connect to"www", that client gets connected to"www.example.com". This option only affects name lookups your server does on behalf of clients. (Default:0)ServerDNSTestAddresses hostname,hostname,...— while we're detecting DNS hijacking, make sure these valid addresses aren't getting redirected. If they are, then our DNS is completely useless 💀 and we'll reset our exit policy to"reject *:*". This option only affects name lookups your server does on behalf of clients. (Default:"www.google.com, www.mit.edu, www.yahoo.com, www.slashdot.org")ServerTransportListenAddr transport IP:PORT— when this is set, Tor will suggestIP:PORTas the listening address of any pluggable transport proxy that tries to launchtransport. (IPv4 addresses get written as-is; IPv6 addresses need to be wrapped in square brackets.) (Default: none)ServerTransportOptions transport k=v k=v ...— when this is set, Tor passes thek=vparameters along to any pluggable transport proxy that tries to launchtransport.
(Example: ServerTransportOptions obfs45 shared-secret=bridgepasswd cache=/var/lib/tor/cache) (Default: none)
ServerTransportPlugin transport exec path-to-binary [options]— the Tor relay launches the pluggable transport proxy atpath-to-binary, usingoptionsas its command-line options, and expects to receive proxied client traffic from it. (Default: none)ShutdownWaitLength NUM— when we get a SIGINT and we're a server, we start shutting down: we close listeners and start refusing new circuits. AfterNUMseconds, we exit. Get a second SIGINT? We exit immediately, no waiting. 💅 (Default: 30 seconds)SigningKeyLifetime N days|weeks|months— how long should each Ed25519 signing key stay valid? Tor uses a permanent master identity key that can be kept offline, and it periodically generates new "signing" keys that it uses online. This option configures their lifetime. (Default: 30 days)SSLKeyLifetime N minutes|hours|days|weeks— okay so when we're making a link certificate for our outermost SSL handshake, this sets how long that cert lives. If you set it to0, Tor just picks some reasonable random defaults for you. (Default:0)
📊 STATISTICS OPTIONS 📊
Relays publish most of their stats in a lil doc called the extra-info document. The options below control which different types of statistics Tor relays collect and publish — iykyk 💅
BridgeRecordUsageByCountry 0|1— when this is on andBridgeRelayis also on, and we have GeoIP data, Tor keeps a per-country count of how many client addresses have hit it up, so it can help the bridge authority guess which countries have blocked access to it. IfExtraInfoStatisticsis enabled, this gets published as part of the extra-info document. (Default:1)CellStatistics 0|1— relays only 🫶 when this is on, Tor collects statistics about cell processing (that's the mean time a cell spends in a queue, the mean number of cells in a queue, and the mean number of processed cells per circuit) and writes them to disk every 24 hours. Onion router operators can use these for performance monitoring. IfExtraInfoStatisticsis enabled, it gets published as part of the extra-info document. (Default:0)ConnDirectionStatistics 0|1— relays only ✨ when this is on, Tor writes stats about the amounts of traffic it passes between itself and other relays to disk every 24 hours. Lets relay operators monitor how much their relay is being used as a middle node in the circuit. IfExtraInfoStatisticsis enabled, it gets published as part of the extra-info document. (Default:0)DirReqStatistics 0|1— relays and bridges only 💖 when this is on, a Tor directory writes stats on the number and response time of network status requests to disk every 24 hours. Lets relay and bridge operators monitor how much their server is being used by clients to learn about the Tor network. IfExtraInfoStatisticsis enabled, it gets published as part of the extra-info document. (Default:1)EntryStatistics 0|1— relays only 🌸 when this is on, Tor writes stats on the number of directly connecting clients to disk every 24 hours. Lets relay operators monitor how much inbound traffic that originates from Tor clients passes through their server to go further down the Tor network. IfExtraInfoStatisticsis enabled, it gets published as part of the extra-info document. (Default:0)ExitPortStatistics 0|1— exit relays only 👑 when this is on, Tor writes stats on the number of relayed bytes and opened streams per exit port to disk every 24 hours. Lets exit relay operators measure and monitor the amounts of traffic leaving the Tor network through their exit node. IfExtraInfoStatisticsis enabled, it gets published as part of the extra-info document. (Default:0)ExtraInfoStatistics 0|1— when this is on, Tor includes previously gathered statistics in the extra-info documents it uploads to the directory authorities. Turning this off also removes bandwidth usage statistics, plus theGeoIPFileandGeoIPv6Filehashes, from the extra-info file. BridgeServerTransportPluginlines are always included in the extra-info file, because BridgeDB requires them fr fr. (Default:1)HiddenServiceStatistics 0|1— relays and bridges only 🦋 when this is on, a Tor relay writes obfuscated statistics about its role as hidden-service directory, introduction point, or rendezvous point to disk every 24 hours. IfExtraInfoStatisticsis enabled, it gets published as part of the extra-info document. (Default:1)OverloadStatistics 0|1*— relays and bridges only 💀 when this is on, a Tor relay will write anoverload generalline in the server descriptor if the relay is considered overloaded. (Default:1)
A relay is considered overloaded if at least one of these conditions is met:
- A certain ratio of ntor onionskins are dropped.
- The OOM was invoked.
- TCP Port exhaustion.
If ExtraInfoStatistics is enabled, it can also put two more specific overload lines in the extra-info document, if at least one of these conditions is met: - Connection rate limits have been reached (read and write side). - File descriptors are exhausted.
PaddingStatistics 0|1— relays and bridges only 🎀 when this is on, Tor collects stats for padding cells sent and received by this relay, on top of total cell counts. These statistics get rounded, and are omitted if traffic is low. This info matters for load balancing decisions related to padding. IfExtraInfoStatisticsis enabled, it gets published as a part of the extra-info document. (Default:1)
🗂️ DIRECTORY SERVER OPTIONS 🗂️
These ones are useful only for directory servers, bestie. (Relays with enough bandwidth automatically become directory servers — peep DirCache for the deets.)
DirCache 0|1— when this is set, Tor caches all current directory documents except extra info documents, and accepts client requests for them. IfDownloadExtraInfois set, cached extra info documents get cached too. SettingDirPortis not required forDirCache, because clients connect via theORPortby default. Setting eitherDirPortorBridgeRelaywhile settingDirCacheto0? Not supported, sorry babe. (Default:1)DirPolicy policy,policy,...— sets an entrance policy for this server, to limit who can connect to the directory ports. The policies have the same form as exit policies above, except that port specifiers are ignored. Any address not matched by some entry in the policy is accepted.DirPort [address:]PORT|auto [flags]— if this is nonzero, advertise the directory service on this port. Set it to"auto"and Tor will pick a port for you ✨ This option can occur more than once, but only one advertisedDirPortis supported: all but oneDirPortmust have theNoAdvertiseflag set. (Default:0)
The same flags are supported here as are supported by ORPort. This port can only be IPv4.
As of Tor 0.4.6.1-alpha, non-authoritative relays (see AuthoritativeDirectory) will not publish the DirPort but will still listen on it. Clients don't use the DirPorts on relays, so it is safe for you to remove the DirPort from your torrc configuration. 💅
DirPortFrontPage FILENAME— when this is set, it takes an HTML file and publishes it as"/"on theDirPort. Now relay operators can provide a disclaimer without needing to set up a separate webserver. There's a sample disclaimer incontrib/operator-tools/tor-exit-notice.html. 🫶MaxConsensusAgeForDiffs N minutes|hours|days|weeks— when this is nonzero, Tor caches will not try to generate consensus diffs for any consensus older than this amount of time. If this is set to zero, Tor will pick a reasonable default from the current networkstatus document. You should not set this option unless your cache is severely low on disk space or CPU. And if you really gotta set it, keeping it above3or4hours will help clients way more than setting it to zero. (Default:0)
🛡️ DENIAL OF SERVICE MITIGATION OPTIONS 🛡️
Tor has a series of built-in denial of service mitigation options that you can individually enable/disable and fine-tune, but by default Tor directory authorities will define reasonable values for the network, so like... no explicit configuration is required to make use of these protections. Slay 💅
Below is a series of config options for relays, and then options for onion services and how they work.
The mitigations take place at relays, and are as follows:
- If a single client address makes too many concurrent connections (this is configurable via
DoSConnectionMaxConcurrentCount), hang up on further connections. - If a single client IP address (v4 or v6) makes circuits too quickly (default values are more than 3 per second, with an allowed burst of 90, see
DoSCircuitCreationRateandDoSCircuitCreationBurst) while also having too many connections open (default is 3, seeDoSCircuitCreationMinConnections), tor will refuse any new circuit (CREATE cells) for the next while (random value between 1 and 2 hours). - If a client asks to establish a rendezvous point to you directly (ex: Tor2Web client), ignore the request.
These defenses can be manually controlled by torrc options, but relays will also take guidance from consensus parameters using these same names, so there's no need to configure anything manually. In doubt, do not change those values. 😭
The values set by the consensus, if any, can be found here: https://consensus-health.torproject.org/#consensusparams
If any of the DoS mitigations are enabled, a heartbeat message will appear in your log at NOTICE level which looks like:
DoS mitigation since startup: 429042 circuits rejected, 17 marked addresses.
2238 connections closed. 8052 single hop clients refused.The following options are useful only for a public relay. They control the Denial of Service mitigation subsystem described above. 👑
DoSCircuitCreationEnabled 0|1|auto— enable circuit creation DoS mitigation. If set to1(enabled), tor will cache client IPs along with statistics in order to detect circuit DoS attacks. If an address is positively identified, tor will activate defenses against the address. See theDoSCircuitCreationDefenseTypeoption for more details. This is a client to relay detection only."auto"means use the consensus parameter. If not defined in the consensus, the value is0. (Default:auto)DoSCircuitCreationBurst NUM— the allowed circuit creation burst per client IP address. If the circuit rate and the burst are reached, a client is marked as executing a circuit creation DoS."0"means use the consensus parameter. If not defined in the consensus, the value is90. (Default:0)DoSCircuitCreationDefenseTimePeriod N seconds|minutes|hours— the base time period in seconds that the DoS defense is activated for. The actual value is selected randomly for each activation fromN+1to3/2 * N."0"means use the consensus parameter. If not defined in the consensus, the value is3600seconds (1 hour). (Default:0)DoSCircuitCreationDefenseTypeNUM — what kind of defense do we throw at a client address we caught doing circuit-creation nonsense 💅 here's the menu:1: literally no defense, we chill.2: refuse circuit creation for theDoSCircuitCreationDefenseTimePeriodstretch of time.
Setting it to "0" means "use the consensus parameter bestie." If the consensus doesn't define it, the value is 2. (Default: 0)
DoSCircuitCreationMinConnectionsNUM — the minimum threshold of concurrent connections before a client address can even be flagged as doing a circuit creation DoS. So like, once a client address hits the circuit rate and has a minimum of NUM concurrent connections, that's a positive detection 🔍"0"means use the consensus parameter. If it's not defined in the consensus, the value is 3. (Default: 0)DoSCircuitCreationRateNUM — the allowed circuit creation rate per second, applied per client IP address. If this option is 0, it obeys a consensus parameter. If it's not defined in the consensus, the value is 3. (Default: 0)DoSConnectionEnabled0|1|auto — turns on the connection DoS mitigation. If you set it to 1 (enabled), then for a client address only, tor gets to mitigate against a huge number of concurrent connections made by one single IP address 🛑"auto"means use the consensus parameter. If it's not defined in the consensus, the value is 0. (Default: auto)DoSConnectionDefenseTypeNUM — the type of defense applied to a detected client address for the connection mitigation. Your options are:1: No defense.2: Immediately close new connections.
"0" means use the consensus parameter. If it's not defined in the consensus, the value is 2. (Default: 0)
DoSConnectionMaxConcurrentCountNUM — the maximum threshold of concurrent connections from a client IP address. Go above this limit and a defense selected byDoSConnectionDefenseTypegets applied."0"means use the consensus parameter. If it's not defined in the consensus, the value is 100. (Default: 0)DoSConnectionConnectRateNUM — the allowed rate of client connection from a single address per second. Paired up with the burst (peep below 👇), if the limit gets reached, the address is marked and a defense is applied (DoSConnectionDefenseType) for a period of time defined byDoSConnectionConnectDefenseTimePeriod. If it's not defined or set to 0, it's controlled by a consensus parameter. (Default: 0)DoSConnectionConnectBurstNUM — the allowed burst of client connection from a single address per second. Check theDoSConnectionConnectRatefor more deets on this detection. If it's not defined or set to 0, it's controlled by a consensus parameter. (Default: 0)DoSConnectionConnectDefenseTimePeriodN seconds|minutes|hours — the base time period in seconds that the client connection defense stays activated for. Heads up: the actual value is selected randomly for each activation, somewhere fromN+1to3/2 * N🎲 If it's not defined or set to 0, it's controlled by a consensus parameter. (Default: 24 hours)DoSRefuseSingleHopClientRendezvous0|1|auto — refuses the establishment of rendezvous points for single hop clients. Basically, if a client directly connects to the relay and sends anESTABLISH_RENDEZVOUScell, it's silently dropped 🙈"auto"means use the consensus parameter. If it's not defined in the consensus, the value is 0. (Default: auto)
the next little batch is useful only for an exit relay, okay? 🫶
DoSStreamCreationEnabled0|1|auto — turns on the stream DoS mitigation. If you set it to 1 (enabled), tor will apply a rate limit on the creation of new streams and dns requests per circuit."auto"means use the consensus parameter. If it's not defined in the consensus, the value is 0. (Default: auto)DoSStreamCreationDefenseTypeNUM — the type of defense applied to a detected circuit or stream for the stream mitigation. Pick your fighter:1: No defense.2: Reject the stream or resolve request.3: Close the circuit creating too many streams.
"0" means use the consensus parameter. If it's not defined in the consensus, the value is 2. (Default: 0)
DoSStreamCreationRateNUM — the allowed rate of stream creation from a single circuit per second. Coupled with the burst (see below 👀), if the limit is reached, actions can be taken against the stream or circuit (DoSStreamCreationDefenseType). If it's not defined or set to 0, it's controlled by a consensus parameter. If it's not defined in the consensus, the value is 100. (Default: 0)DoSStreamCreationBurstNUM — the allowed burst of stream creation from a circuit per second. See theDoSStreamCreationRatefor more deets on this detection. If it's not defined or set to 0, it's controlled by a consensus parameter. If it's not defined in the consensus, the value is 300. (Default: 0)
and this batch? useful only for a directory cache 🌸
DoSBeginDirRateEnabled0|1|auto — turns on theBEGIN_DIRrate limit. If you set it to 1 (enabled), tor will apply a rate limit on the directory streams opened per circuit."auto"means use the consensus parameter. If it's not defined in the consensus, the value is 1. (Default: auto)DoSBeginDirDefenseTypeNUM — the type of defense applied to a circuit going over theBEGIN_DIRrate limit. The possible values are:1: No defense.2: Reject the directory stream.3: Close the circuit.
"0" means use the consensus parameter. If it's not defined in the consensus, the value is 3. (Default: 0)
DoSBeginDirRateNUM — the allowed rate ofBEGIN_DIRfrom a single circuit per second. Coupled with the burst (see below bestie), if the limit is reached, actions can be taken against the stream or circuit (DoSBeginDirDefenseType). If it's not defined or set to 0, it's controlled by a consensus parameter. If it's not defined in the consensus, the value is 10. (Default: 0)DoSBeginDirBurstNUM — the allowed burst ofBEGIN_DIRfrom a circuit. See theDoSBeginDirRatefor more deets on this detection. If it's not defined or set to 0, it's controlled by a consensus parameter. If it's not defined in the consensus, the value is 60. (Default: 0)
For onion services, mitigations are a work in progress and multiple options are currently available 🧚 No cap.
The introduction point defense is a rate limit on the number of introduction requests that will get forwarded to a service by each of its honest introduction point routers. This can prevent some types of overwhelming floods from reaching the service, but plot twist — it will also prevent legitimate clients from establishing new connections 😭
the following options are per onion service 💜
HiddenServiceEnableIntroDoSDefense0|1 — enables DoS defense at the intropoint level. When this is enabled, the rate and burst parameter (see below) will be sent to the intro point, which will then use them to apply rate limiting for introduction requests to this service.
The introduction point honors the consensus parameters except if this is specifically set by the service operator using this option. The service never looks at the consensus parameters in order to enable or disable this defense. (Default: 0)
HiddenServiceEnableIntroDoSBurstPerSecNUM — the allowed client introduction burst per second at the introduction point. If this option is 0, it's considered infinite, so ifHiddenServiceEnableIntroDoSDefenseis set, it then effectively disables the defenses 🙃 (Default: 200)HiddenServiceEnableIntroDoSRatePerSecNUM — the allowed client introduction rate per second at the introduction point. If this option is 0, it's considered infinite, so ifHiddenServiceEnableIntroDoSDefenseis set, it then effectively disables the defenses. (Default: 25)
The rate is the maximum number of clients a service will ask its introduction points to allow every seconds. And the burst is a parameter that allows that many within one second.
For example, the default values of 25 and 200 respectively means that for every introduction points a service has (default 3 but can be configured with HiddenServiceNumIntroductionPoints), 25 clients per seconds will be allowed to reach the service and 200 at most within 1 second as a burst. This means that if 200 clients are seen within 1 second, it will take 8 seconds (200/25) for another client to be able to be allowed to introduce due to the rate of 25 per second 🔥
This might be too much for your use case or not, fine tuning these values is hard and are likely different for each service operator. it's giving ✨artisanal configuration✨
Why is this not helping reachability of the service? Because the defenses are at the introduction point, an attacker can easily flood all introduction point rendering the service unavailable due to no client being able to pass through. But, the service itself is not overwhelmed with connections, allowing it to function properly for the few clients that were able to go through or other any services running on the same tor instance.
The bottom line is that this protects the network by preventing an onion service to flood the network with new rendezvous circuits that is reducing load on the network 😌 iconic behavior honestly.
A secondary mitigation is available, based on prioritized dispatch of rendezvous circuits for new connections. The queue is ordered based on effort a client chooses to spend at computing a proof-of-work function 👑
the following options are per onion service 💅
HiddenServicePoWDefensesEnabled0|1 — enables proof-of-work based service DoS mitigation. If set to 1 (enabled), tor will include parameters for an optional client puzzle in the encrypted portion of this hidden service's descriptor. Incoming rendezvous requests will be prioritized based on the amount of effort a client chooses to make when computing a solution to the puzzle. The service will periodically update a suggested amount of effort, based on attack load, and disable the puzzle entirely when the service is not overloaded. (Default: 0)HiddenServicePoWQueueRateNUM — the sustained rate of rendezvous requests to dispatch per second from the priority queue. Has no effect when proof-of-work is disabled. If this is set to 0 there's no explicit limit and we will process requests as quickly as possible ⚡ (Default: 250)HiddenServicePoWQueueBurstNUM — the maximum burst size for rendezvous requests handled from the priority queue at once. (Default: 2500)
These options are applicable to both onion services and their clients:
CompiledProofOfWorkHash 0|1|auto— so when proof-of-work DoS mitigation is turned on, both the services themselves and the clients connecting to them use a dynamically generated hash function as part of the puzzle computation. That's the whole vibe.
Here's the breakdown bestie:
- Set it to
1and puzzles only get solved and verified using the compiled implementation (which is about 20x faster), and we choose to fail rather than fall back to something slower. - Set it to
0and the compiler is never used. - By default, the compiler is always tried if possible, but the interpreter is available as a fallback.
(Default: auto)
See also--list-modules— heads up, these proof of work options do nothing unless the"pow"module is enabled at compile time. 💀
👑 DIRECTORY AUTHORITY SERVER OPTIONS 👑
Okay so these options let Tor operate as a directory authority, and they control how Tor behaves when it's being one. If you're just running a regular relay or exit server on the public Tor network, you should not need to adjust any of them. Like, genuinely. Don't touch. 🙈
AuthoritativeDirectory 0|1— set this to1and Tor operates as an authoritative directory server. Instead of caching the directory, it generates its own list of good servers, signs it, and sends that to the clients. Unless the clients already have you listed as a trusted directory, you probably do not want to set this option. 👑BridgeAuthoritativeDir 0|1— when you set this in addition toAuthoritativeDirectory, Tor accepts and serves server descriptors, but it caches and serves the main networkstatus documents rather than generating its own. (Default:0)V3AuthoritativeDirectory 0|1— when you set this in addition toAuthoritativeDirectory, Tor generates version 3 network statuses and serves descriptors, etc, as described in thedir-spec.txtfile of torspec (for Tor clients and servers running at least0.2.0.x). ✨AuthDirBadExit AddressPattern...— authoritative directories only. A set of address patterns for servers that will be listed as bad exits in any network status document this authority publishes, ifAuthDirListBadExitsis set.
(The address pattern syntax here and in the options below is the same as for exit policies, except that you don't need to say "accept" or "reject", and ports are not needed.)
AuthDirMiddleOnly AddressPattern...— authoritative directories only. A set of address patterns for servers that will be listed as middle-only in any network status document this authority publishes, ifAuthDirListMiddleOnlyis set.AuthDirFastGuarantee N bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— authoritative directories only. If non-zero, always vote the Fast flag for any relay advertising this amount of capacity or more. (Default:100 KBytes)AuthDirGuardBWGuarantee N bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— authoritative directories only. If non-zero, this advertised capacity or more is always sufficient to satisfy the bandwidth requirement for the Guard flag. (Default:2 MBytes)AuthDirHasIPv6Connectivity 0|1— authoritative directories only. Set it to0and OR ports with an IPv6 address are not included in the authority's votes. Set it to1and IPv6 OR ports get tested for reachability just like IPv4 OR ports. If the reachability test succeeds, the authority votes for the IPv6 ORPort, and votes Running for the relay. If the reachability test fails, the authority does not vote for the IPv6 ORPort, and does not vote Running. (Default:0)
The content of the consensus depends on the number of voting authorities that set AuthDirHasIPv6Connectivity:
- If no authorities set
AuthDirHasIPv6Connectivity 1, there will be no IPv6 ORPorts in the consensus. - If a minority of authorities set
AuthDirHasIPv6Connectivity 1, unreachable IPv6 ORPorts will be removed from the consensus. But the majority of IPv4-only authorities will still vote the relay as Running. Reachable IPv6 ORPort lines will be included in the consensus. - If a majority of voting authorities set
AuthDirHasIPv6Connectivity 1, relays with unreachable IPv6 ORPorts will not be listed as Running. Reachable IPv6 ORPort lines will be included in the consensus. (To ensure that any valid majority will vote relays with unreachable IPv6 ORPorts not Running, 75% of authorities must setAuthDirHasIPv6Connectivity 1.)
😭 that 75% number is doing so much heavy lifting, bestie. remember it.
AuthDirInvalid AddressPattern...— authoritative directories only. A set of address patterns for servers that will never be listed as"valid"in any network status document that this authority publishes.AuthDirListBadExits 0|1— authoritative directories only. If set to1, this directory has some opinion about which nodes are unsuitable as exit nodes. (Do not set this to1unless you plan to list non-functioning exits as bad; otherwise you are effectively voting in favor of every declared exit as an exit.) 🫠AuthDirListMiddleOnly 0|1— authoritative directories only. If set to1, this directory has some opinion about which nodes should only be used in the middle position. (Do not set this to1unless you plan to list questionable relays as"middle only"; otherwise you are effectively voting against middle-only status for every relay.)AuthDirMaxServersPerAddr NUM— authoritative directories only. The maximum number of servers that we will list as acceptable on a single IP address. Only servers that we have found reachable at that address count toward the limit, so it is only enforced whenAuthDirTestReachabilityis set, and for IPv6 addresses only whenAuthDirHasIPv6Connectivityis set. Set this to"0"for "no limit". (Default:2)AuthDirPinKeys 0|1— authoritative directories only. If non-zero, do not allow any relay to publish a descriptor if any other relay has reserved its<Ed25519,RSA>identity keypair. In all cases, Tor records every keypair it accepts in a journal if it is new, or if it differs from the most recently accepted pinning for one of the keys it contains. (Default:1)AuthDirReject AddressPattern...— authoritative directories only. A set of address patterns for servers that will never be listed at all in any network status document that this authority publishes, or accepted as an OR address in any descriptor submitted for publication by this authority.AuthDirRejectRequestsUnderLoad 0|1— if set, the directory authority will start rejecting directory requests from non relay connections by sending a 503 error code if it is under bandwidth pressure (reaching the configured limit if any). Relays will always be answered even if this is on. (Default:1)AuthDirBadExitCCs CC,...AuthDirInvalidCCs CC,...AuthDirMiddleOnlyCCs CC,...AuthDirRejectCCs CC,...
^^ those four are a little family, and here's their shared description: authoritative directories only. These options contain a comma-separated list of country codes such that any server in one of those country codes will be marked as a bad exit/invalid for use, or rejected entirely. 🌍
AuthDirSharedRandomness 0|1— authoritative directories only. Switch for the shared random protocol. If zero, the authority won't participate in the protocol. If non-zero (default), the flag"shared-rand-participate"is added to the authority vote indicating participation in the protocol. (Default:1)AuthDirSupport048Clients 0|1— authoritative directory only. If true, we do not accept any router descriptors or generate any consensuses that would confuse clients running0.4.8and earlier. If false,0.4.8clients will get a somewhat degraded view of the network, but will generally continue working. (Default:0)AuthDirTestEd25519LinkKeys 0|1— authoritative directories only. If this option is set to0, then we treat relays as"Running"if their RSA key is correct when we probe them, regardless of their Ed25519 key. We should only ever set this option to0if there is some major bug in Ed25519 link authentication that causes us to label all the relays as not Running. (Default:1)AuthDirTestReachability 0|1— authoritative directories only. If set to1, then we periodically check every relay we know about to see whether it is running. If set to0, we vote Running for every relay, and don't perform these tests. (Default:1)AuthDirVoteGuard node,node,...— a list of identity fingerprints or country codes or address patterns of nodes to vote Guard for regardless of their uptime and bandwidth. SeeExcludeNodesfor more information on how to specify nodes. 👑AuthDirVoteGuardBwThresholdFraction FRACTION— the Guard flag bandwidth performance threshold fraction, which is the fraction representing who gets the Guard flag out of all measured bandwidth. (Default:0.75)AuthDirVoteGuardGuaranteeTimeKnown N seconds|minutes|hours|days|weeks— a relay with at least this much weighted time known can be considered familiar enough to be a guard. (Default:8 days)AuthDirVoteGuardGuaranteeWFU FRACTION— a level of weighted fractional uptime (WFU) is that is sufficient to be a Guard. (Default:0.98)AuthDirVoteStableGuaranteeMinUptime N seconds|minutes|hours|days|weeks— if a relay's uptime is at least this value, then it is always considered stable, regardless of the rest of the network. (Default:30 days)AuthDirVoteStableGuaranteeMTBF N seconds|minutes|hours|days|weeks— if a relay's mean time between failures (MTBF) is least this value, then it will always be considered stable. (Default:5 days)BridgePassword Password— if set, contains an HTTP authenticator that tells a bridge authority to serve all requested bridge information. Used by the (only partially implemented)"bridge community"design, where a community of bridge relay operators all use an alternate bridge directory authority, and their target user audience can periodically fetch the list of available community bridges to stay up-to-date. (Default: not set)ConsensusParams STRING—STRINGis a space-separated list ofkey=valuepairs that Tor will include in the"params"line of its networkstatus vote. This directive can be specified multiple times so you don't have to put it all on one line. 🌈DirAllowPrivateAddresses 0|1— set this to1and Tor will happily accept server descriptors that have literally any old thing in theirAddresselement 🫶. Leave it at0and it's pickier: if that address isn't a real IP address, or if it's a private one, the server descriptor gets rejected, no debate. Bonus behavior when it's1: Tor will also let exit policies for private networks count toward fulfilling Exit flag requirements. (Default:0)GuardfractionFile FILENAME— V3 authoritative directories ONLY, bestie 👑. This tells the authority where to find the guardfraction file, which is the file holding the tea on how long relays have been guards. (Default: unset)MinMeasuredBWsForAuthToIgnoreAdvertised N— a total value in abstract bandwidth units, iykyk. It describes how much measured total bandwidth an authority should have observed across the whole network before it starts treating advertised bandwidths as completely and utterly unreliable. (Default:500)MinUptimeHidServDirectoryV2 N seconds|minutes|hours|days|weeks— the minimum uptime a relay needs before directory authorities will accept it as a hidden service directory. One little exception: a relay that already has the HSDir flag in a recent consensus doesn't have to meet this bar 🎀. (Default:96 hours)RecommendedClientVersions STRING—STRINGis a comma-separated list of Tor versions that are currently believed to be safe for clients to use ✨. This info gets included in version 2 directories. If you don't set it, then the value ofRecommendedVersionsis used instead. And heads up: when you do set this, you should setVersioningAuthoritativeDirectorytoo.RecommendedServerVersions STRING— same energy, server edition 🦋.STRINGis a comma-separated list of Tor versions currently believed safe for servers. It's included in version 2 directories. Not set? ThenRecommendedVersionssteps in. And again: setting this means you should also setVersioningAuthoritativeDirectory.RecommendedVersions STRING—STRINGis a comma-separated list of Tor versions currently believed to be safe, periodt 🌸. The list goes into each directory, and nodes that pull the directory down learn from it whether they need to upgrade. This option can show up multiple times, and the values from all those lines get spliced together 💅. When this is set,VersioningAuthoritativeDirectoryshould be set too.MinimalAcceptedServerVersion STRING—STRINGis the oldest Tor version that the directory authority will accept for relays and bridges 🧚. Anything older than that? Rejected, sorry not sorry. (Default:0.4.9.1-alpha)V3AuthDistDelay N seconds|minutes|hours— V3 authoritative directories only 👑. Configures the server's preferred delay between publishing its consensus and signature, and then assuming it's got all the signatures from all the other authorities. Huge caveat though: the time actually used is not the server's preferred time, it's the consensus of everyone's preferences. (Default:5 minutes)V3AuthNIntervalsValid NUM— V3 authoritative directories only. This sets the number ofVotingIntervalsthat each consensus should be valid for 🗳️. Going high on this number raises network partitioning risks; going low raises directory traffic. And same deal as before — the actual number of intervals used isn't the server's preferred number, it's the consensus of all the preferences. Must be at least2. (Default:3)V3AuthUseLegacyKey 0|1— turn this on and the directory authority signs consensuses not just with its own signing key, but also with a"legacy"key and certificate that carry a different identity 🔑. The whole point of this feature is migrating directory authority keys in the event of a compromise. (Default:0)V3AuthVoteDelay N seconds|minutes|hours— V3 authoritative directories only. The server's preferred delay between publishing its vote and assuming it's collected all the votes from all the other authorities. But (you knew this was coming) the actual time used is not the server's preferred time — it's the consensus of all preferences ✨. (Default:5 minutes)V3AuthVotingInterval N minutes|hours— V3 authoritative directories only. Configures the server's preferred voting interval ⏰. Real talk: voting will actually happen at an interval chosen by consensus from all the authorities' preferred intervals. And this time SHOULD divide evenly into a day, no cap. (Default:1 hour)V3BandwidthsFile FILENAME— V3 authoritative directories only. Points at the location of the bandwidth-authority generated file that stores info on relays' measured bandwidth capacities 📊. Pro tip for avoiding inconsistent reads: bandwidth data should be written to a temporary file, and then renamed to the configured filename.VersioningAuthoritativeDirectory 0|1— set this to1and Tor adds information about which versions of Tor are still believed safe for use into the published directory 🫶. Every version 1 authority is automatically a versioning authority; version 2 authorities get to provide this service optionally. SeeRecommendedVersions,RecommendedClientVersions, andRecommendedServerVersions.
✨ PER SERVICE OPTIONS ✨
These ones only work after a HiddenServiceDir line, bestie — they're per-service, not per-whole-tor. Got it? Good. 💖
HiddenServiceAllowUnknownPorts0|1— set it to1and connections hitting ports nobody recognizes will not make the current hidden service tear down its rendezvous circuits. Setting it to0is NOT an authorization mechanism, okay?? It's just a mild inconvenience aimed at port-scanners. That's it. (Default:0)HiddenServiceDirDIRECTORY — this is where you stash the data files for a hidden service, in DIRECTORY. Every single hidden service needs its own separate directory, no sharing. You can drop this option multiple times to set up multiple services. If DIRECTORY doesn't exist yet, Tor will just create it for you. ✨ Heads up though: you cannot add a new Onion Service to an already-running Tor instance if Sandbox is enabled. (Note: in current versions of Tor, if DIRECTORY is a relative path it's relative to the current working directory of the Tor instance, not to its DataDirectory. Don't rely on this behavior — it's not guaranteed to stay the same in future versions.)HiddenServiceDirGroupReadable0|1— set to1and the filesystem group gets to read the hidden service directory and the hostname file. Set to0and only the owner can read it. (Default:0) Has no effect on Windows. 🪟HiddenServiceExportCircuitIDprotocol — the onion service uses this protocol to expose the global circuit identifier of each inbound client circuit. The only protocol supported right now is'haproxy'. This option is for v3 services only. (Default:none)
So here's how the haproxy option actually works: when you flip the feature on, the Tor process writes a header line whenever a client connects to the onion service. That header looks exactly like this:
"PROXY TCP6 fc00:dead:beef:4dad::ffff:ffff ::1 65535 42\r\n"We encode the "global circuit identifier" as the last 32-bits of that first IPv6 address. Every other value in the header can safely be ignored, fr. You can compute the global circuit identifier with this formula, given the IPv6 address "fc00:dead:beef:4dad::AABB:CCDD":
global_circuit_id = (0xAA << 24) + (0xBB << 16) + (0xCC << 8) + 0xDD;In the case above, where the last 32-bits are 0xffffffff, the global circuit identifier would be 4294967295. You can take that value and, together with Tor's control port, terminate particular circuits using their global circuit identifiers. More on that in control-spec.txt.
The HAProxy version 1 protocol is described in detail at https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt — go read it if you want the whole lore. 📖
HiddenServiceOnionBalanceInstance0|1— set to1and this onion service becomes an OnionBalance instance, meaning it'll accept client connections destined for an OnionBalance frontend. In that case Tor expects to find a file named"ob_config"inside theHiddenServiceDirdirectory, with content like:
MasterOnionAddress <frontend_onion_address>where <frontend_onion_address> is the onion address of the OnionBalance frontend (e.g. wrxdvcaqpuzakbfww5sxs6r2uybczwijzfn2ezy2osaj7iox7kl7nhad.onion). Yes that's a whole address, no I'm not retyping it, it's right there 💅
HiddenServiceMaxStreamsN — the maximum number of simultaneous streams (connections) per rendezvous circuit. The biggest value allowed is65535. (Setting this to0allows an unlimited number of simultaneous streams.) (Default:0)HiddenServiceMaxStreamsCloseCircuit0|1— set to1and going overHiddenServiceMaxStreamswill cause the offending rendezvous circuit to be torn down, as opposed to stream creation requests that exceed the limit just being silently ignored. (Default:0)HiddenServiceNumIntroductionPointsNUM — how many introduction points your hidden service gets. You can't have more than20. (Default:3)HiddenServicePortVIRTPORT [TARGET] — configure a virtual port VIRTPORT for a hidden service. You can use this option multiple times; each time it applies to the service using the most recentHiddenServiceDir. By default this maps the virtual port to the same port on127.0.0.1over TCP. You can override the target port, address, or both by specifying a target of addr, port, addr:port, orunix:path. (You can specify an IPv6 target as[addr]:port. Unix paths may be quoted, and may use standard C escapes.) You can also have multiple lines with the same VIRTPORT: when someone connects to that VIRTPORT, one of the *TARGET*s from those lines gets chosen at random. Note that address-port pairs have to be comma-separated.HiddenServiceVersion3— a list of rendezvous service descriptor versions to publish for the hidden service. Currently only version3is supported. (Default:3)
🦋 PER INSTANCE OPTIONS 🦋
These ones hit the whole tor instance, not just one service. Big main-character energy, whole-brain decisions. 👑
HiddenServiceSingleHopMode0|1— Experimental! Non Anonymous Hidden Services on a tor instance inHiddenServiceSingleHopModemake one-hop (direct) circuits between the onion service server, and the introduction and rendezvous points. (Onion service descriptors are still posted using 3-hop paths, so onion service directories can't block the service.) This option makes every hidden service instance hosted by a tor instance a Single Onion Service. One-hop circuits make Single Onion servers easily locatable, but clients stay location-anonymous. However, the fact that a client is accessing a Single Onion rather than a Hidden Service may be statistically distinguishable. 👀
WARNING: Once a hidden service directory has been used by a tor instance in HiddenServiceSingleHopMode, it can NEVER be used again for a hidden service. Best practice is to create a new hidden service directory, key, and address for each new Single Onion Service and each Hidden Service. It is not possible to run Single Onion Services and Hidden Services from the same tor instance — they should be run on different servers with different IP addresses. Periodt.
HiddenServiceSingleHopMode requires HiddenServiceNonAnonymousMode to be set to 1. And since a Single Onion service is non-anonymous, you can't configure a SOCKSPort on a tor instance that's running in HiddenServiceSingleHopMode. Can not be changed while tor is running. (Default: 0)
HiddenServiceNonAnonymousMode0|1— makes hidden services non-anonymous on this tor instance. Allows the non-anonymousHiddenServiceSingleHopMode, and enables direct connections in the server-side hidden service protocol. If you're using this, you need to disable all client-side services on your Tor instance,SOCKSPortset to"0"included. Can not be changed while tor is running. (Default:0)PublishHidServDescriptors0|1— set to0and Tor will still run any hidden services you configure, but it won't advertise them to the rendezvous directory. Only useful if you're using a Tor controller that handles hidserv publishing for you. (Default:1)
🔐 CLIENT AUTHORIZATION 🔐
Okay so client auth is how you gatekeep your onion service. Let's get into it. 🎀
🧚 Service side
To set up client authorization on the service side, the "<HiddenServiceDir>/authorized_clients/" directory needs to exist. Each file in that directory should be suffixed with .auth (i.e. "alice.auth"; the file name is irrelevant) and its content format MUST be:
<auth-type>:<key-type>:<base32-encoded-public-key>The supported <auth-type> is: "descriptor". The supported <key-type> is: "x25519". The <base32-encoded-public-key> is the base32 representation of the raw key bytes only (32 bytes for x25519).
Each file MUST contain one line only. Any malformed file will be ignored. Client authorization only gets enabled for the service if tor successfully loads at least one authorization file.
Note that once you've configured client authorization, anyone else with the address won't be able to access it from that point on. If no authorization is configured, the service will be accessible to anyone with the onion address. 😬 So like... decide carefully.
Revoking a client can be done by removing their .auth file, however the revocation will be in effect only after the tor process gets restarted or if a SIGHUP takes place. So don't think deleting the file is instant, bestie.
🧚 Client side
To access a v3 onion service with client authorization as a client, make sure you have ClientOnionAuthDir set in your torrc. Then, in the <ClientOnionAuthDir> directory, create an .auth_private file for the onion service corresponding to this key (i.e. 'bob_onion.auth_private'). The contents of the <ClientOnionAuthDir>/<user>.auth_private file should look like:
<56-char-onion-addr-without-.onion-part>:descriptor:x25519:<x25519 private key in base32>For more information, please see https://2019.www.torproject.org/docs/tor-onion-service.html.en#ClientAuthorization 💜
✨ TESTING NETWORK OPTIONS ✨
Okay bestie, so these are the options you use when you're setting up a whole entire testing Tor network. Do not touch these on a real one, I'm begging. 🧪
TestingTorNetwork 0|1— Set this to1and Tor goes "say less" and adjusts the default values of all the options listed below, so building a testing Tor network is way easier. You may only set it if a non-default set ofDirAuthoritiesis set. Also — and this is important — it cannot be unset while Tor is running. (Default:0)
DirAllowPrivateAddresses 1
EnforceDistinctSubnets 0
AuthDirMaxServersPerAddr 0
ClientBootstrapConsensusAuthorityDownloadInitialDelay 0
ClientBootstrapConsensusFallbackDownloadInitialDelay 0
ClientBootstrapConsensusAuthorityOnlyDownloadInitialDelay 0
ClientDNSRejectInternalAddresses 0
ClientRejectInternalAddresses 0
CountPrivateBandwidth 1
ExitPolicyRejectPrivate 0
ExtendAllowPrivateAddresses 1
V3AuthVotingInterval 5 minutes
V3AuthVoteDelay 20 seconds
V3AuthDistDelay 20 seconds
TestingV3AuthInitialVotingInterval 150 seconds
TestingV3AuthInitialVoteDelay 20 seconds
TestingV3AuthInitialDistDelay 20 seconds
TestingAuthDirTimeToLearnReachability 0 minutes
MinUptimeHidServDirectoryV2 0 minutes
TestingServerDownloadInitialDelay 0
TestingClientDownloadInitialDelay 0
TestingServerConsensusDownloadInitialDelay 0
TestingClientConsensusDownloadInitialDelay 0
TestingBridgeDownloadInitialDelay 10
TestingBridgeBootstrapDownloadInitialDelay 0
TestingClientMaxIntervalWithoutRequest 5 seconds
TestingDirConnectionMaxStall 30 seconds
TestingEnableConnBwEvent 1
TestingEnableCellStatsEvent 1TestingAuthDirTimeToLearnReachability N seconds|minutes|hours— So like, after you start as an authority, you gotta not make any claims about whether routers areRunninguntil this much time has gone by. Changing this requires thatTestingTorNetworkis set. (Default:30 minutes)TestingAuthKeyLifetime N seconds|minutes|hours|days|weeks|months— Overrides the default lifetime for a signing Ed25519 TLS Link authentication key. (Default:2 days)TestingAuthKeySlop N seconds|minutes|hours— that's it, that's the entry, no description given, she's mysterious like that 💅TestingBridgeBootstrapDownloadInitialDelay N— Initial delay in seconds for how long clients should wait before downloading a bridge descriptor for a new bridge. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingBridgeDownloadInitialDelay N— Once clients have successfully downloaded a bridge descriptor, how long to wait (in seconds) before trying another download for that same bridge. Changing this requires thatTestingTorNetworkis set. (Default:10800)TestingClientConsensusDownloadInitialDelay N— Initial delay in seconds for when clients should download consensuses. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingClientDownloadInitialDelay N— Initial delay in seconds for when clients should download things in general. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingClientMaxIntervalWithoutRequest N seconds|minutes— When directory clients have only a few descriptors to request, they batch them until they have more, or until this amount of time has passed. Changing this requires thatTestingTorNetworkis set. (Default:10 minutes)TestingDirAuthVoteExit node,node,...— A list of identity fingerprints, country codes, and address patterns of nodes to voteExitfor, regardless of their uptime, bandwidth, or exit policy. Wanna know how to specify nodes? Go checkExcludeNodes, it explains it.
For this option to have any effect at all, TestingTorNetwork has to be set. See ExcludeNodes for more info on how to specify nodes. 👑
TestingDirAuthVoteExitIsStrict 0|1— If True (1), a node will never receive theExitflag unless it's specified in theTestingDirAuthVoteExitlist, regardless of its uptime, bandwidth, or exit policy.
For this option to have any effect, TestingTorNetwork has to be set.
TestingDirAuthVoteGuard node,node,...— A list of identity fingerprints and country codes and address patterns of nodes to voteGuardfor, regardless of their uptime and bandwidth. SeeExcludeNodesfor more info on how to specify nodes.
For this option to have any effect, TestingTorNetwork has to be set.
TestingDirAuthVoteGuardIsStrict 0|1— If True (1), a node will never receive theGuardflag unless it's specified in theTestingDirAuthVoteGuardlist, regardless of its uptime and bandwidth.
For this option to have any effect, TestingTorNetwork has to be set.
TestingDirAuthVoteHSDir node,node,...— A list of identity fingerprints and country codes and address patterns of nodes to voteHSDirfor, regardless of their uptime and DirPort. SeeExcludeNodesfor more info on how to specify nodes.
For this option to have any effect, TestingTorNetwork must be set. (Not "has to", must. She said what she said.) 🌸
TestingDirAuthVoteHSDirIsStrict 0|1— If True (1), a node will never receive theHSDirflag unless it's specified in theTestingDirAuthVoteHSDirlist, regardless of its uptime and DirPort.
For this option to have any effect, TestingTorNetwork has to be set.
TestingDirConnectionMaxStall N seconds|minutes— Let a directory connection stall this long before expiring it. Changing this requires thatTestingTorNetworkis set. (Default:5 minutes)TestingEnableCellStatsEvent 0|1— If this option is set, then Tor controllers may register forCELL_STATSevents. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingEnableConnBwEvent 0|1— If this option is set, then Tor controllers may register forCONN_BWevents. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingLinkCertLifetime N seconds|minutes|hours|days|weeks|months— Overrides the default lifetime for the certificates used to authenticate our X509 link cert with our ed25519 signing key. (Default:2 days)TestingLinkKeySlop N seconds|minutes|hours— another one with no description, iconic and elusive 💜TestingMinExitFlagThreshold N KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— Sets a lower-bound for assigning an exit flag when running as an authority on a testing network. Overrides the usual default lower bound of4 KBytes. (Default:0)TestingMinFastFlagThreshold N bytes|KBytes|MBytes|GBytes|TBytes|KBits|MBits|GBits|TBits— Minimum value for theFastflag. Overrides the ordinary minimum taken from the consensus whenTestingTorNetworkis set. (Default:0.)TestingMinTimeToReportBandwidth N seconds|minutes|hours— Do not report our measurements for our maximum observed bandwidth for any time period that has lasted for less than this amount of time. Values over1 dayhave no effect. (Default:1 day)TestingServerConsensusDownloadInitialDelay N— Initial delay in seconds for when servers should download consensuses. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingServerDownloadInitialDelay N— Initial delay in seconds for when servers should download things in general. Changing this requires thatTestingTorNetworkis set. (Default:0)TestingSigningKeySlop N seconds|minutes|hours— How early before the official expiration of a an Ed25519 signing key do we replace it and issue a new key? (Default:3 hoursfor link and auth;1 dayfor signing.)TestingV3AuthInitialDistDelay N seconds|minutes|hours— LikeV3AuthDistDelay, but for initial voting interval before the first consensus has been created. Changing this requires thatTestingTorNetworkis set. (Default:5 minutes)TestingV3AuthInitialVoteDelay N seconds|minutes|hours— LikeV3AuthVoteDelay, but for initial voting interval before the first consensus has been created. Changing this requires thatTestingTorNetworkis set. (Default:5 minutes)TestingV3AuthInitialVotingInterval N seconds|minutes|hours— LikeV3AuthVotingInterval, but for initial voting interval before the first consensus has been created. Changing this requires thatTestingTorNetworkis set. (Default:30 minutes)TestingV3AuthVotingStartOffset N seconds|minutes|hours— Directory authorities offset voting start time by this much. Changing this requires thatTestingTorNetworkis set. (Default:0)
🚫 NON-PERSISTENT OPTIONS 🚫
These ones? They do not get saved to the torrc file by the "SAVECONF" controller command. Other options of this type are documented in control-spec.txt, section 5.4. And honestly, end-users should mostly ignore them, fr fr. 🙈
__ControlPort,__DirPort,__DNSPort,__ExtORPort,__NATDPort,__ORPort,__SocksPort,__TransPort— These underscore-prefixed options are variants of the regular Port options. They behave the same, except they are not saved to the torrc file by the controller'sSAVECONFcommand.
📡 SIGNALS 📡
Tor catches the following signals:
SIGTERM— Tor will catch this, clean up and sync to disk if necessary, and exit. Byeee 👋SIGINT— Tor clients behave as withSIGTERM; but Tor servers will do a controlled slow shutdown, closing listeners and waiting30 secondsbefore exiting. (The delay can be configured with theShutdownWaitLengthconfig option.)SIGHUP— The signal instructs Tor to reload its configuration (including closing and reopening logs), and kill and restart its helper processes if applicable.SIGUSR1— Log statistics about current connections, past connections, and throughput.SIGUSR2— Switch all logs to logleveldebug. You can go back to the old loglevels by sending aSIGHUP.SIGCHLD— Tor receives this signal when one of its helper processes has exited, so it can clean up.SIGPIPE— Tor catches this signal and ignores it.SIGXFSZ— If this signal exists on your platform, Tor catches and ignores it.
📁 FILES 📁
/etc/tor/torrc— Default location of the configuration file. Your home base. 🏠$HOME/.torrc— Fallback location for torrc, if/etc/tor/torrcis not found./var/lib/tor/— The tor process stores keys and other data here.CacheDirectory/cached-certs— Contains downloaded directory key certificates that are used to verify authenticity of documents generated by the Tor directory authorities.CacheDirectory/cached-consensusand/orcached-microdesc-consensus— The most recent consensus network status document we've downloaded.
💾 CacheDirectory/cached-descriptors and cached-descriptors.new
okay so these two files are where all your downloaded router statuses live 💖 Heads up though: some routers might show up more than once, and when that happens Tor just goes with the most recently published descriptor, no drama. Any line starting with an @-sign is an annotation holding extra deets about a given router. And the .new file? It's an append-only journal — it just keeps growing until it gets too big, at which point everything gets merged into a brand new cached-descriptors file. Slay, housekeeping 🧹✨
💾 CacheDirectory/cached-extrainfo and cached-extrainfo.new
Basically the same vibe as cached-descriptors, except these hold the optionally-downloaded "extra-info" documents 🌸 Relays use them to send inessential info about statistics, bandwidth history, and network health over to the authorities. And bestie, note this: they are not fetched by default. Nope. Check DownloadExtraInfo if you wanna know more 👑
💾 CacheDirectory/cached-microdescs and cached-microdescs.new
These hold your downloaded microdescriptors 🦋 Lines beginning with @-signs are annotations carrying more info about a given router. Same journal energy as before — the .new file is append-only, and when it gets too large all the entries get merged into a fresh cached-microdescs file.
💾 DataDirectory/state
This one holds a set of persistent key-value mappings 💅 and it includes:
- the current entry guards and their status.
- the current bandwidth accounting values.
- when the file was last written
- what version of Tor generated the state file
- a short history of bandwidth usage, as produced in the server descriptors.
💾 DataDirectory/sr-state
Authority only 👑 This file records info about the current status of the shared-random-value voting state.
💾 CacheDirectory/diff-cache
Directory cache only 🎀 Holds older consensuses and diffs, going from the oldest all the way to the most recent consensus of each type, compressed in various ways. Each file has a set of key-value arguments describing its contents, then a single NUL byte, then the main file contents. Iconic little sandwich 🥪
💾 DataDirectory/bw_accounting
ngl this file is obsolete — the data now lives in the state file instead. Back in the day it tracked bandwidth accounting values (when the current period starts and ends; how much has been read and written so far this period) ✨
💾 DataDirectory/control_auth_cookie
This file can only be used when cookie authentication is enabled 🔐 It's for cookie authentication with the controller, and its location can be overridden by the CookieAuthFile configuration option. It gets regenerated on startup. See control-spec.txt in torspec for the details, bestie.
💾 DataDirectory/lock
The bouncer of your data directory 🚪 This file stops two Tor instances from using the same data directory. If access to this file is locked, that means the data directory is already in use by Tor. Periodt.
💾 DataDirectory/key-pinning-journal
Used by authorities. A line-based file that records mappings between RSA1024 and Ed25519 identity keys 💜 Authorities enforce these mappings, so that once a relay has picked an Ed25519 key, stealing or factoring the RSA1024 key will no longer let an attacker impersonate the relay. That's the whole point, and it's serving 🫶
💾 KeyDirectory/authority_identity_key
A v3 directory authority's master identity key, used to authenticate its signing key. Tor doesn't use this while it's running — the tor-gencert program uses this instead. And listen: if you're running an authority, you should keep this key offline, and not put it in this file 🙈
💾 KeyDirectory/authority_certificate
Only directory authorities use this one 👑 A v3 directory authority's certificate, which authenticates the authority's current vote- and consensus-signing key using its master identity key.
💾 KeyDirectory/authority_signing_key
Again, only directory authorities 🔥 A v3 directory authority's signing key that's used to sign votes and consensuses. It corresponds to the authority_certificate cert.
💾 KeyDirectory/legacy_certificate
Same as authority_certificate — used only when V3AuthUseLegacyKey is set. See the documentation for V3AuthUseLegacyKey.
💾 KeyDirectory/legacy_signing_key
Same as authority_signing_key — used only when V3AuthUseLegacyKey is set. See the documentation for V3AuthUseLegacyKey.
💾 KeyDirectory/secret_id_key
A relay's RSA1024 permanent identity key, private and public components included 💖 Used to sign router descriptors, and to sign other keys.
💾 KeyDirectory/ed25519_master_id_public_key
The public part of a relay's Ed25519 permanent identity key ✨
💾 KeyDirectory/ed25519_master_id_secret_key
The private part of a relay's Ed25519 permanent identity key 🔐 This key is used to sign the medium-term ed25519 signing key. You can keep this file offline or encrypted — but if you do, Tor won't be able to generate new signing keys automatically, and you'll have to use tor --keygen to do it yourself. No cap.
💾 KeyDirectory/ed25519_signing_secret_key
The private and public components of a relay's medium-term Ed25519 signing key 🦋 This key is authenticated by the Ed25519 master key, which in turn authenticates other keys (and router descriptors). Chain of trust, literally.
💾 KeyDirectory/ed25519_signing_cert
The certificate which authenticates "ed25519_signing_secret_key" as having been signed by the Ed25519 master key 💅
💾 KeyDirectory/secret_onion_key and secret_onion_key.old
A relay's RSA1024 short-term onion key 🧅 Used to decrypt old-style ("TAP") circuit extension requests. The .old file holds the previously generated key, which the relay uses to handle any requests that were made by clients that didn't have the new one. So considerate 🫶
💾 KeyDirectory/secret_onion_key_ntor and secret_onion_key_ntor.old
A relay's Curve25519 short-term onion key 🌈 Used to handle modern ("ntor") circuit extension requests. The .old file holds the previously generated key, which the relay uses to handle any requests that were made by clients that didn't have the new one. Same energy, different curve ✨
💾 KeyDirectory/keyname.secret_family_key
A relay family's family identity key 👑 Used to prove membership in a relay family. See https://community.torproject.org/relay/setup/post-install/family-ids/ for more information.
💾 DataDirectory/fingerprint
Only used by servers. Contains the fingerprint of the server's RSA identity key 💜
💾 DataDirectory/fingerprint-ed25519
Only used by servers. Contains the fingerprint of the server's ed25519 identity key 💜
💾 DataDirectory/hashed-fingerprint
Only used by bridges 🌉 Contains the hashed fingerprint of the bridge's identity key. (That is, the hash of the hash of the identity key.) Yes, hash of a hash. It's giving layers.
💾 DataDirectory/bridgelines
Only used by bridges 🌉 Contains the bridge lines that clients can use to connect using pluggable transports.
💾 DataDirectory/approved-routers
Only used by authoritative directory servers 👑 Each line lists a status and an identity, separated by whitespace. Identities can be hex-encoded RSA fingerprints, or base-64 encoded ed25519 public keys. (See the fingerprint file in a tor relay's DataDirectory for an example fingerprint line.) Here's the status tea:
- If the status is
!reject, then descriptors from the given identity are rejected by this server. - If it is
!invalidthen descriptors are accepted, but marked in the vote as not valid. - If it is
!badexit, then the authority will vote for it to receive aBadExitflag, indicating that it shouldn't be used for traffic leaving the Tor network 🚫 - If it is
!middleonly, then the authority will vote for it to only be used in the middle of circuits.
(Neither rejected nor invalid relays are included in the consensus.)
💾 DataDirectory/v3-status-votes
Only for v3 authoritative directory servers 👑 This file contains status votes from all the authoritative directory servers.
💾 CacheDirectory/unverified-consensus
Contains a network consensus document that has been downloaded, but which we didn't have the right certificates to check yet 🤷 So it just sits there, unverified.
💾 CacheDirectory/unverified-microdesc-consensus
Contains a microdescriptor-flavored network consensus document that has been downloaded, but which we didn't have the right certificates to check yet 🌸 Same situation, microdesc edition.
💾 DataDirectory/unparseable-desc
Onion server descriptors that Tor was unable to parse get dumped into this file 💀 Only used for debugging.
💾 DataDirectory/router-stability
Only used by authoritative directory servers 👑 Tracks measurements for router mean-time-between-failures so that authorities have a fair idea of how to set their Stable flags.
💾 DataDirectory/stats/dirreq-stats
Only used by directory caches and authorities. This file collects directory request statistics 📊
💾 DataDirectory/stats/entry-stats
Only used by servers. Collects incoming connection statistics by Tor entry nodes 🚪✨
💾 DataDirectory/stats/bridge-stats
Only used by servers. Collects incoming connection statistics by Tor bridges 🌉
💾 DataDirectory/stats/exit-stats
Only used by servers. Collects outgoing connection statistics by Tor exit routers 🚪
💾 DataDirectory/stats/buffer-stats
Only used by servers. Collects buffer usage history 📈
💾 DataDirectory/stats/conn-stats
Only used by servers. Collects approximate connection history (number of active connections over time) 📊
💾 DataDirectory/stats/hidserv-stats
Only used by servers. Collects approximate counts of what fraction of the traffic is hidden service rendezvous traffic, and approximately how many hidden services the relay has seen 🧅
💾 DataDirectory/networkstatus-bridges
Only used by authoritative bridge directories 🌉 Contains information about bridges that have self-reported themselves to the bridge authority.
💾 HiddenServiceDirectory/hostname
The <base32-encoded-fingerprint>.onion domain name for this hidden service 🧅 If the hidden service is restricted to authorized clients only, this file also contains authorization data for all clients.
Note The clients will ignore any extra subdomains prepended to a hidden service hostname. Supposing you have"xyz.onion"as your hostname, you can ask your clients to connect to"www.xyz.onion"or"irc.xyz.onion"for virtual-hosting purposes.
💾 HiddenServiceDirectory/private_key
Contains the private key for this hidden service 🔐 keep it safe bestie
💾 HiddenServiceDirectory/client_keys
Contains authorization data for a hidden service that is only accessible by authorized clients 🎀
HiddenServiceDirectory/onion_service_non_anonymous— okay so this lil file is a whole vibe marker</p> 🎀 it shows up inHiddenServiceDirectorywhenever a hidden service key got created while you were inHiddenServiceNonAnonymousMode. if it's there, that key was born non-anonymous, periodt. iykyk 🦋
✨ SEE ALSO ✨
bestie, the lore does not stop at this manpage. 💖 Go read up:
- the Tor Project website at
https://www.torproject.org/🌈 - the Tor specifications at
https://spec.torproject.org📜
Also go say hi to torsocks(1) and torify(1) — they're the side characters that carry the whole plot 🫶
💀 BUGS 💀
ngl, real talk: Tor is still under development, so there may be plenty of bugs. Not a few. Plenty. 😭 it's giving "we're so back" and "it's so over" at the same time fr fr.
If you find one, do not suffer in silence — please report it at https://bugs.torproject.org/ 👑
This isn't a "haha cute bug" situation — reporting is genuinely the move, and the maintainers want to hear from you. 💅
and that's the whole page, bestie! 🎀✨ You just read the entire TOR(1) manual — ate and left no crumbs 👑🌸 This drop of the page is from the Tor project, dated 10/28/2026, and it's filed under TOR(1) section 1. go be safe out there and encrypt everything 💖🦋🍓
<chip>💜</chip>
📦 where this page comes from
- debian trixie: tor 0.4.9.15-0+deb13u1
- 💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms