PTP4l(8)

📦
version 4.4 · 8 versions
, in arch rolling, fedora 43, nixpkgs unstable, opensuse-leap 16.0, ubuntu resolute, ubuntu stonking

✨ NAME ✨

okay so bestie, this is the System Manager's Manual page for ptp4l(8) — PTP Boundary/Ordinary/Transparent Clock. 💖

🦋 SYNOPSIS 🦋

ptp4l [ -AEP246HSLmqsv ] [ -f config ] [ -p phc-device ] [ -l print-level ] [ -i interface ] [ long-options ] ...

🌸 DESCRIPTION 🌸

ptp4l is literally an implementation of the Precision Time Protocol (PTP) according to IEEE standard 1588 for Linux. It implements Boundary Clock (BC), Ordinary Clock (OC), and Transparent Clock (TC). no cap.

👑 OPTIONS 👑

  • -A — Select the delay mechanism automatically. Start with E2E and switch to P2P when a peer delay request is received. auto magic ✨
  • -E — Select the delay request-response (E2E) mechanism. This is the default mechanism. All clocks on single PTP communication path must use the same mechanism. A warning will be printed when a peer delay request is received on port using the E2E mechanism. heads up!
  • -P — Select the peer delay (P2P) mechanism. A warning will be printed when a delay request is received on port using the P2P mechanism. another heads up!
  • -2 — Select the IEEE 802.3 network transport.
  • -4 — Select the UDP IPv4 network transport. This is the default transport. default vibe
  • -6 — Select the UDP IPv6 network transport.
  • -H — Select the hardware time stamping. All ports specified by the -i option and in the configuration file must be attached to the same PTP hardware clock (PHC). This is the default time stamping. PHC bestie
  • -S — Select the software time stamping.
  • -L — Select the legacy hardware time stamping.
  • -f config — Read configuration from the specified file. No configuration file is read by default. config file? optional, periodt
  • -i interface — Specify a PTP port, it may be used multiple times. At least one port must be specified by this option or in the configuration file. stack 'em up
  • -p phc-device — (This option is deprecated.) Before Linux kernel v3.5 there was no way to discover the PHC device associated with a network interface. This option specifies the PHC device (e.g. /dev/ptp0) to be used when running on legacy kernels. 🚨 deprecated, bestie 🚨
  • -s — Enable the clientOnly mode.
  • -l print-level — Set the maximum syslog level of messages which should be printed or sent to the system logger. The default is 6 (LOG_INFO). loggy loggy
  • -m — Print messages to the standard output.
  • -q — Don't send messages to the system logger.
  • -v — Prints the software version and exits.
  • -h — Display a help message.

💅 LONG OPTIONS 💅

Each and every configuration file option (see below) may also appear as a "long" style command line argument. For example, the clientOnly option may be set using either of these two forms.

--clientOnly 1   --clientOnly=1

Option values given on the command line override values in the global section of the configuration file. CLI wins, periodt

🎀 CONFIGURATION FILE 🎀

The configuration file is divided into sections. Each section starts with a line containing its name enclosed in brackets and it follows with settings. Each setting is placed on a separate line, it contains the name of the option and the value separated by whitespace characters. Empty lines and lines starting with # are ignored. clean and simple

There are three different section types.

  1. The global section (indicated as [global]) sets the program options, clock options and default port options. Other sections are port specific sections and they override the default port options.
  2. Port sections give the name of the configured port (e.g. [eth0]). Ports specified in the configuration file don't need to be specified by the -i option. An empty port section can be used to replace the command line option.
  3. Tables for configuring unicast discovery begin with [unicast_master_table].

See UNICAST DISCOVERY OPTIONS, below. iykyk

🌈 PORT OPTIONS 🌈

  • active_key_id — Used in conjunction with spp and sa_file directives to specify which key from the spp defined Security Association should be used for outbound icv calculations. All Security Associations are read from the file specified by sa_file. Requires spp and sa_file directives. Must be in the range of 1 to 2^32-1, inclusive. The default is 0 (disabled). security stuff, bestie
  • allowedLostResponses — The number of missed peer delay responses before the asCapable variable is reset. The default is 3. three strikes
  • announceReceiptTimeout — The number of missed Announce messages before the last Announce messages expires. The default is 3. announce me
  • boundary_clock_jbod — When running as a boundary clock (that is, when more than one network interface is configured), ptp4l performs a sanity check to make sure that all of the ports share the same hardware clock device. This option allows ptp4l to work as a boundary clock using "just a bunch of devices" that are not synchronized to each other. For this mode, the collection of clocks must be synchronized by an external program, for example phc2sys(8) in "automatic" mode. The default is 0 (disabled). JBOD energy
  • cmlds.client_address — Specifies the source address for the UNIX domain socket that receives peer delay measurement when using the "COMMON_P2P" delay mechanism. The default is /var/run/cmlds_client. socket city
  • cmlds.domainNumber — Specifies the domainNumber message field for communications with the local Common Mean Link Delay Service, used with the "COMMON_P2P" delay mechanism. The default is 0.
  • cmlds.majorSdoId — Specifies the transportSpecific message field for communications with the local Common Mean Link Delay Service, used with the "COMMON_P2P" delay mechanism. The default is 2.
  • cmlds.port — Specifies the port number of local Common Mean Link Delay Service from which to accept delay measurements. The range of valid port numbers is one to 65535, but the special value zero configures the port number of the port requesting the measurements. The default is 0, meaning the requesting port number.
  • cmlds.server_address — Specifies the UNIX domain socket address of the local Common Mean Link Delay Service, for use with the "COMMON_P2P" delay mechanism. The default is /var/run/cmlds_server.
  • delayAsymmetry — The time difference in nanoseconds of the transmit and receive paths. This value should be positive when the server-to-client propagation time is longer and negative when the client-to-server time is longer. The default is 0 nanoseconds. asymmetry, but make it fashion
  • delay_filter — Select the algorithm used to filter the measured delay and peer delay. Possible values are moving_average and moving_median. The default is moving_median.
  • delay_filter_length — The length of the delay filter in samples. The default is 10.
  • delay_mechanism — Select the delay mechanism. Possible values are Auto, COMMON_P2P, E2E, P2P, and NONE. The default is E2E. mechanism moment
  • delay_response_timeout — The number of delay response messages that may go missing before triggering a synchronization fault. Setting this option to zero will disable the delay response timeout. The default is 0 or disabled.
  • egressLatency — Specifies the difference in nanoseconds between the actual transmission time at the reference plane and the reported transmit time stamp. This value will be added to egress time stamps obtained from the hardware. The default is 0.
  • fault_badpeernet_interval — The time in seconds between the detection of a peer network misconfiguration and the fault being reset. The port is disabled for the duration of the interval. The value is in seconds and the special key word ASAP will let the fault be reset immediately. The default is 16 seconds. peer network drama
  • fault_reset_interval — The time in seconds between the detection of a port's fault and the fault being reset. This value is expressed as a power of two. Setting this value to -128 or to the special key word "ASAP" will let the fault be reset immediately. The default is 4 (16 seconds). fault reset, but make it ASAP
  • follow_up_info — Include the 802.1AS data in the Follow_Up messages if enabled. The default is 0 (disabled).
  • G.8275.portDS.localPriority — The Telecom Profiles (ITU-T G.8275.1 and G.8275.2) specify an alternate Best Master Clock Algorithm (BMCA) with a unique data set comparison algorithm. The value of this option is associated with Announce messages arriving on a particular port and is used as a tie breaker whenever clockClass, clockAccuracy, offsetScaledLogVariance, and priority2 are equal. This option is only used when "dataset_comparison" is set to "G.8275.x". The default value is 128.
Warning: the BMCA is guaranteed to produce a spanning tree (that is, a timing network without loops) only when using the default values of G.8275.defaultDS.localPriority and G.8275.portDS.localPriority. Careful network engineering is needed when using non-default values. do not play with this unless you know what you're doing, bestie
  • hybrid_e2e — Enables the "hybrid" delay mechanism from the draft Enterprise Profile. When enabled, ports in the client state send their delay request messages to the unicast address taken from the server's announce message. Ports in the server state will reply to unicast delay requests using unicast delay responses. This option has no effect if the delay_mechanism is set to P2P. The default is 0 (disabled). hybrid era
  • ignore_transport_specific — By default, incoming messages are dropped if their transportSpecific field does not match the configured value. However, many of the transports specified in the 1588 standard mandate ignoring this field. Moreover, some equipment is known to set the reserved bits. Configuring this option as 1 causes this field to be ignored completely on receive. The default is 0.
  • ingressLatency — Specifies the difference in nanoseconds between the reported receive time stamp and the actual reception time at reference plane. This value will be subtracted from ingress time stamps obtained from the hardware. The default is 0.
  • inhibit_delay_req — Don't send any delay requests. This will need the asCapable config option to be set to 'true'. This is useful when running as a designated server who does not need to calculate offset from client. The default is 0 (disabled). no delay reqs, no problems
  • inhibit_multicast_service — bestie, some unicast mode profiles are so dramatic and insist that no multicast message ever gets sent. Periodt. Set this and multicast transmission is inhibited. Default is 0 (multicast enabled). 💅
  • logAnnounceInterval — the mean time interval between Announce messages. A shorter interval makes ptp4l react faster to changes in the client/server hierarchy (main character reactivity ✨). It should be the same across the whole domain, okay? It's specified as a power of two in seconds. Default is 1 (2 seconds).
  • logMinDelayReqInterval — the minimum permitted mean time interval between Delay_Req messages. Shorter = ptp4l reacts faster to path delay changes. Specified as a power of two in seconds. Default is 0 (1 second).
  • logMinPdelayReqInterval — the minimum permitted mean time interval between Pdelay_Req messages. Specified as a power of two in seconds. Default is 0 (1 second).
  • logSyncInterval — the mean time interval between Sync messages. A shorter interval may improve accuracy of the local clock 👑. Specified as a power of two in seconds. Default is 0 (1 second).
  • masterOnly — ⚠️ okay so this one is deprecated and will be removed in a future release. Use "serverOnly" instead. We don't use it anymore, bestie.
  • min_neighbor_prop_delay — lower limit for peer delay, in nanoseconds. If the estimated peer delay is smaller than this value, the port gets marked as not 802.1AS capable. 😭
  • neighborPropDelayThresh — upper limit for peer delay, in nanoseconds. If the estimated peer delay is greater than this value, the port gets marked as not 802.1AS capable.
  • network_transport — picks the network transport, the whole vibe of how messages travel. Possible values are UDPv4, UDPv6 and L2. Default is UDPv4.
  • net_sync_monitor — enables the NetSync Monitor (NSM) protocol 🦋. The NSM protocol lets a station measure how well another node is synchronized. The monitor sends a unicast delay request to the node, which replies unconditionally with unicast delay response, sync, and follow up messages. If the monitor is synchronized to the GM, it can use the time stamps in the message to estimate the node's offset. This option requires that the 'hybrid_e2e' option be enabled as well. Default is 0 (disabled).
  • operLogPdelayReqInterval — the Pdelay Request messages interval to be used once the clock enters the SERVO_LOCKED_STABLE state. If the 'msg_interval_request' option is set, then the local client port will adopt this rate when the local clock enters the "locked stable" state. Specified as a power of two in seconds, and the default value is 0 (1 second).
  • operLogSyncInterval — the Sync message interval to be requested once the clock enters the SERVO_LOCKED_STABLE state. If the 'msg_interval_request' option is set, then the local client port will request the remote server to switch to the given message rate via a signaling message containing a Message interval request TLV. Specified as a power of two in seconds, and the default value is 0 (1 second).
  • path_trace_enabled — enables the mechanism used to trace the route of the Announce messages 🧚. Default is 0 (disabled).
  • phc_index — specifies the index of the PHC to be used for synchronization with hardware timestamping. This option is useful with virtual clocks running on top of a free-running physical clock (created by writing to /sys/class/ptp/ptp*/n_vclocks). Default is -1, which means the index will be set to the PHC associated with the interface, or the device specified by the -p option.
  • power_profile.2011.grandmasterTimeInaccuracy — specifies the time inaccuracy of the GM in nanoseconds. Relevant only when power_profile.version is 2011. This value may be changed dynamically using the POWER_PROFILE_SETTINGS_NP management message. Default is -1 meaning unknown inaccuracy.
  • power_profile.2011.networkTimeInaccuracy — specifies the time inaccuracy of the network in nanoseconds. Relevant only when power_profile.version is 2011. This value may be changed dynamically using the POWER_PROFILE_SETTINGS_NP management message. Default is -1 meaning unknown inaccuracy.
  • power_profile.2017.totalTimeInaccuracy — specifies the sum of the GM, network, and local node inaccuracies in nanoseconds. Relevant only when power_profile.version is 2017. This value may be changed dynamically using the POWER_PROFILE_SETTINGS_NP management message. Default is -1 meaning unknown inaccuracy.
  • power_profile.grandmasterID — specifies an optional, non-zero identification code for the GM. Heads up though: the code is an arbitrary, power profile specific integer, not necessarily related to the clockIdentity in any way. This value may be changed dynamically using the POWER_PROFILE_SETTINGS_NP management message. Default is 0 meaning unused.
  • power_profile.version — specifies the power profile version to be used. Valid values are "none", "2011", or "2017". This value may be changed dynamically using the POWER_PROFILE_SETTINGS_NP management message. Default is "none".
  • ptp_dst_ipv4 — the IPv4 address to which PTP messages should be sent. Relevant only with UDPv4 transport. Default is 224.0.1.129.
  • p2p_dst_ipv4 — the IPv4 address to which peer delay messages should be sent. Relevant only with UDPv4 transport. Default is 224.0.0.107.
  • ptp_dst_ipv6 — the IPv6 address to which PTP messages should be sent. The second byte of the address is substituted with udp6_scope. Relevant only with UDPv6 transport. Default is FF0E:0:0:0:0:0:0:181.
  • p2p_dst_ipv6 — the IPv6 address to which peer delay messages should be sent. Relevant only with UDPv6 transport. Default is FF02:0:0:0:0:0:0:6B.
  • ptp_dst_mac — the MAC address to which PTP messages should be sent. Relevant only with L2 transport. Default is 01:1B:19:00:00:00.
  • p2p_dst_mac — the MAC address to which peer delay messages should be sent. Relevant only with L2 transport. Default is 01:80:C2:00:00:0E.
  • serverOnly — setting this to one (1) prevents the port from entering the client state. On top of that, the local clock will ignore Announce messages received on this port. The intended use is to support the Telecom Profiles according to ITU-T G.8265.1, G.8275.1, and G.8275.2. Default value is zero or false.
  • spp — specifies the Security Parameters Pointer of the desired Security Association to be used for Authentication TLV support for a given port 🔐. Any port with an assigned spp will attach Authentication TLVs to all outbound messages and check for Authentication TLVs on all inbound messages in accordance to the corresponding security association sourced via the sa_file directive. Outbound Authentication TLVs are generated using the key specified by active_key_id. Not compatible with one step ports or advertised versions less then PTPv2.1. Requires sa_file and active_key_id directives. Must be in the range of 0 to 255, inclusive. Default is -1 (disabled).
  • syncReceiptTimeout — the number of sync/follow up messages that may go missing before triggering a Best Master Clock election 🗳️. This option is used for running in gPTP mode according to the 802.1AS-2011 standard. Setting this to zero will disable the sync message timeout. Default is 0 or disabled.
  • transportSpecific — the transport specific field. Must be in the range 0 to 255. Default is 0.
  • tsproc_mode — selects the time stamp processing mode used to calculate offset and delay. Possible values are filter, raw, filter_weight, raw_weight. Raw modes perform well when the rate of sync messages (logSyncInterval) is similar to the rate of delay messages (logMinDelayReqInterval or logMinPdelayReqInterval). Weighting is useful with larger network jitters (e.g. software time stamping). Default is filter.
  • udp_ttl — specifies the Time to live (TTL) value for IPv4 multicast messages and the hop limit for IPv6 multicast messages. Only relevant with the IPv4 and IPv6 UDP transports. Default is 1 to restrict the messages sent by ptp4l to the same subnet.
  • unicast_listen — when enabled, this option allows the port to grant unicast message contracts. Incoming requests will be granted limited only by the amount of memory available. Default is 0 (disabled).
  • unicast_master_table — when set to a positive integer, this option specifies the table id to be used for unicast discovery. Each table lives in its own section and has a unique, positive numerical ID. Entries in the table are a pair of transport type and protocol address. Default is 0 (unicast discovery disabled).
  • unicast_req_duration — the service time in seconds to be requested during unicast discovery. Note that the remote node is free to grant a different duration — it does what it wants, bestie. Default is 3600 seconds or one hour.

⏰ PROGRAM AND CLOCK OPTIONS ⏰

  • asCapable — if set to 'true', all the checks which can unset the asCapable variable (as described in Section 10.2.4.1 of 802.1AS) are skipped. If set to 'auto', asCapable is initialized to 'false' and will be set to 'true' after the relevant checks have passed ✅. Default value is 'auto'.
  • assume_two_step — treat one-step responses as two-step if enabled. It's used to work around buggy 802.1AS switches 🙈. Default is 0 (disabled).
  • BMCA — bestie, this one lets you use static roles for server and client devices instead of running the best master clock algorithm (BMCA) described in the 1588 profile. It's giving "I already know who's who so let's skip the audition" energy, and it speeds up the start time for servers and clients when you know the roles of the devices in advance ✨. When set to noop, the traditional BMCA algorithm used by 1588 is skipped. masterOnly and clientOnly will be used to determine the server or client role for the device. In a bridge, clientOnly (which is a global option) can be set to make all ports assume the client role. masterOnly (which is a per-port config option) can then be used to set individual ports to take on the server role. The default value is ptp which runs the BMCA related state machines. 🎀
  • check_fup_sync — okay so packet reordering can happen in the network, in the hardware, or in the networking stack, and when it does, a follow up message can appear to arrive in the application before the matching sync message 😭. This is a normal occurrence, and the sequenceID message field ensures proper matching, so the ptp4l program accepts out of order packets. This option adds an additional check using the software time stamps from the networking stack to verify that the sync message did arrive first. This option is only useful if you do not trust the sequence IDs generated by the server. The default is 0 (disabled).
  • clientOnly — the local clock is a client-only clock if enabled. The default is 0 (disabled). 💅
  • clockAccuracy — the clockAccuracy attribute of the local clock. It is used in the PTP server selection algorithm. The default is 0xFE.
  • clockClass — the clockClass attribute of the local clock. It denotes the traceability of the time distributed by the grandmaster clock. The default is 248.
  • clock_class_threshold — the maximum clock class value from master, acceptible to subordinate clock beyond which it moves out of lock state. The default value is 248.
  • clockIdentity — the clockIdentity attribute of the local clock. The clockIdentity is an 8-octet array and should in this configuration be written in textual form, see default. It should be unique since it is used to identify the specific clock. If default is used or if not set at all, the clockIdentity will be automtically generated. The default is "000000.0000.000000".
  • clock_servo — the servo which is used to synchronize the local clock. Valid values are "pi" for a PI controller, "linreg" for an adaptive controller using linear regression, "ntpshm" and "refclock_sock" for the NTP SHM and chrony SOCK reference clocks respectively to allow another process to synchronize the local clock, and "nullf" for a servo that always dials frequency offset zero (for use in SyncE nodes). The default is "pi." 🦋
  • clock_type — specifies the kind of PTP clock. Valid values are "OC" for ordinary clock, "BC" for boundary clock, "P2P_TC" for peer to peer transparent clock, and "E2E_TC" for end to end transparent clock. An multi-port ordinary clock will automatically be configured as a boundary clock. The default is "OC".
  • dataset_comparison — specifies the method to be used when comparing data sets during the Best Master Clock Algorithm. The possible values are "ieee1588" and "G.8275.x". The default is "ieee1588".
  • domainNumber — the domain attribute of the local clock. The default is 0.
  • dscp_event — defines the Differentiated Services Codepoint (DSCP) to be used for PTP event messages. Must be a value between 0 and 63. There are several media streaming standards out there that require specific values for this option. For example 46 (EF PHB) in AES67 or 48 (CS6 PHB) in RAVENNA. The default is 0. 🌈
  • dscp_general — defines the Differentiated Services Codepoint (DSCP) to be used for PTP general messages. Must be a value between 0 and 63. There are several media streaming standards out there that recommend specific values for this option. For example 34 (AF41 PHB) in AES67 or 46 (EF PHB) in RAVENNA. The default is 0.
  • first_step_threshold — the maximum offset the servo will correct by changing the clock frequency (phase when using nullf servo) instead of stepping the clock. This is only applied on the first update. It's specified in seconds. When set to 0.0, the servo won't step the clock on start. The default is 0.00002 (20 microseconds). This option used to be called pi_f_offset_const.
  • free_running — don't adjust the local clock if enabled. The default is 0 (disabled).
  • freq_est_interval — the time interval over which is estimated the ratio of the local and peer clock frequencies. It is specified as a power of two in seconds. The default is 1 (2 seconds).
  • G.8275.defaultDS.localPriority — the Telecom Profiles (ITU-T G.8275.1 and G.8275.2) specify an alternate Best Master Clock Algorithm (BMCA) with a unique data set comparison algorithm. The value of this option is associated with the local clock and is used as a tie breaker whenever clockClass, clockAccuracy, offsetScaledLogVariance, and priority2 are equal. This option is only used when "dataset_comparison" is set to "G.8275.x". The default value is 128.
⚠️ Warning: the BMCA is guaranteed to produce a spanning tree (that is, a timing network without loops) only when using the default values of G.8275.defaultDS.localPriority and G.8275.portDS.localPriority. Careful network engineering is needed when using non-default values. no cap, this one's serious 💀
  • gmCapable — if this option is enabled, then the local clock is able to become grand master. This is only for use with 802.1AS clocks and has no effect on 1588 clocks. The default is 1 (enabled). 👑
  • ignore_source_id — this will disable source port identity checking for Sync and Follow_Up messages. This is useful when the announce messages are disabled in the server and the client does not have any way to know the server's identity. The default is 0 (disabled).
  • inhibit_announce — this will disable the timer for announce messages (i.e. FD_MANNO_TIMER) and also the announce message timeout timer (i.e. FD_ANNOUNCE_TIMER). This is used by the Automotive profile as part of switching over to a static BMCA. If this option is enabled, ignore_source_id has to be enabled in the client because it has no way to identify the server in the Sync and Follow_Up messages. The default is 0 (disabled).
  • initial_delay — the initial path delay of the clock in nanoseconds used for synchronization of the clock before the delay is measured using the E2E or P2P delay mechanism. If set to 0, the clock will not be updated until the delay is measured. The default is 0.
  • interface_rate_tlv — when the client and server are operating are operating at different interface rate, delay asymmetry caused due to different interface rate needs to be compensated. The server sends its interface rate using interface rate TLV as per G.8275.2 Annex D. The default is 0 (does not support interface rate tlv).
  • hwts_filter — select the hardware time stamp filter setting mode. Possible values are normal, check, full. normal mode set the filters as needed. check mode only check but do not set. full mode set the receive filter to mark all packets with hardware time stamp, so all applications can get them. The default is normal.
  • kernel_leap — when a leap second is announced, let the kernel apply it by stepping the clock instead of correcting the one-second offset with servo, which would correct the one-second offset slowly by changing the clock frequency (unless the step_threshold option is set to correct such offset by stepping). Relevant only with software time stamping. The default is 1 (enabled). ✨
  • logging_level — the maximum logging level of messages which should be printed. The default is 6 (LOG_INFO).
  • manufacturerIdentity — the manufacturer id which should be an OUI owned by the manufacturer. The default is 00:00:00.
  • max_frequency — the maximum allowed frequency adjustment of the clock in parts per billion (ppb). This is an additional limit to the maximum allowed by the hardware. When set to 0, the hardware limit will be used. The default is 900000000 (90%). This option used to be called pi_max_frequency.
  • maxStepsRemoved — when using this option, if the value of stepsRemoved of an Announce message is greater than or equal to the value of maxStepsRemoved the Announce message is not considered in the operation of the BMCA. The default value is 255.
  • message_tag — the tag which is added to all messages printed to the standard output or system log. If the tag contains the string "{level}", it will be replaced with the log level of the message as a number. The default is an empty string (which cannot be set in the configuration file as the option requires an argument).
  • msg_interval_request — this option, when set, will trigger an adjustment to the Sync and peer delay request message intervals when the clock servo transitions into the SERVO_LOCKED_STABLE state. The Sync interval will be adjusted via the signaling mechanism while the pdelay request interval is simply adjusted locally. The values to use for the new Sync and peer delay request intervals are specified by the operLogSyncInterval and operLogPdelayReqInterval options, respectively. The default value of msg_interval_request is 0 (disabled).
  • ntpshm_segment — the number of the SHM segment used by ntpshm servo. The default is 0.
  • offsetScaledLogVariance — the offsetScaledLogVariance attribute of the local clock. It characterizes the stability of the clock. The default is 0xFFFF.
  • pi_integral_const — the integral constant of the PI controller. When set to 0.0, the integral constant will be set by the following formula from the current sync interval. The default is 0.0. 🍓
ki = min(ki_scale * sync^ki_exponent, ki_norm_max / sync)
  • pi_integral_exponent — this is the ki_exponent constant in the formula that sets the integral constant of the PI controller from the sync interval. Default is 0.4. ✨
  • pi_integral_norm_max — the ki_norm_max constant in that same formula, still setting the integral constant of the PI controller from the sync interval. Default is 0.3.
  • pi_integral_scale — the ki_scale constant in the formula setting the integral constant of the PI controller from the sync interval. When you set it to 0.0, the value gets picked from 0.3 and 0.001 for hardware and software time stamping respectively. Default is 0.0.
  • pi_proportional_const — the proportional constant of the PI controller. When set to 0.0, the proportional constant gets set by this formula from the current sync interval. Default is 0.0.
kp = min(kp_scale * sync^kp_exponent, kp_norm_max / sync)
  • pi_proportional_exponent — the kp_exponent constant in the formula that sets the proportional constant of the PI controller from the sync interval. Default is -0.3.
  • pi_proportional_norm_max — the kp_norm_max constant in the formula that sets the proportional constant of the PI controller from the sync interval. Default is 0.7
  • pi_proportional_scale — the kp_scale constant in the formula that sets the proportional constant of the PI controller from the sync interval. When set to 0.0, the value gets picked from 0.7 and 0.1 for hardware and software time stamping respectively. Default is 0.0.
  • productDescription — the product description string. Allowed values must be of the form manufacturerName;modelNumber;instanceIdentifier and hold at most 64 utf8 symbols. Default is ";;".
  • priority1 — the priority1 attribute of the local clock. It's used in the PTP server selection algorithm, and lower values take precedence. Gotta be in the range 0 to 255. Default is 128.
  • priority2 — the priority2 attribute of the local clock. Also used in the PTP server selection algorithm, also lower values take precedence. Also must be in the range 0 to 255. Default is 128.
  • ptp_minor_version — this one sets the minorVersionPTP in the common PTP message header. Default is 1.
  • refclock_sock_address — the address of the UNIX domain socket that the refclock_sock servo uses. Default is /var/run/refclock.ptp.sock.
  • revisionData — the revision description string, holding the revisions for node hardware (HW), firmware (FW) and software (SW). Allowed values are of the form HW;FW;SW and hold at most 32 utf8 symbols. Default is an ";;".
  • sa_file — points at the location of the file holding the Security Associations used for immediate security processing of the Authentication TLV, supporting the optional security mechanism defined in ieee1588-2019 ch 14.16. Peep SECURITY ASSOCIATION OPTIONS for how this file should be formatted. spp and active_key_id should be specifed for each port to say which Security Association from the /fBsa_file/fR should be used. Default is an empty string.
  • sanity_freq_limit — the max allowed frequency offset between the uncorrected clock and the system monotonic clock, in parts per billion (ppb). It's a sanity check on the synchronized clock: when a bigger offset gets measured, a warning message prints and the servo resets. If the frequency correction set by ptp4l changes unexpectedly between updates of the clock (like, say, because another process is trying to control the clock 🙈), a warning message prints. Set it to 0 and the sanity check is off. Default is 200000000 (20%).
  • servo_num_offset_values — how many offset values get considered before transitioning from the SERVO_LOCKED state to the SERVO_LOCKED_STABLE state. The transition happens once the last servo_num_offset_values offsets are all below the servo_offset_threshold value. Default is 10.
  • servo_offset_threshold — the offset threshold used for transitioning from SERVO_LOCKED to SERVO_LOCKED_STABLE. The transition happens once the last servo_num_offset_values offsets are all below the threshold value. Default value of offset_threshold is 0 (disabled).
  • slave_event_monitor — the address of a UNIX domain socket for event monitoring. A local monitoring client bound to this address will receive SLAVE_RX_SYNC_TIMING_DATA and SLAVE_DELAY_TIMING_DATA_NP TLVs. Default is the empty string (disabled).
  • slaveOnly — ⚠️ this option is deprecated and will be removed in a future release. Use clientOnly instead.
  • socket_priority — configures the SO_PRIORITY of sockets. This exists for when you want to route ptp4l traffic using Linux qdiscs for traffic shaping. Heads up: it's only available with the IEEE 802.3 transport (the -2 option) and gets silently ignored when you're using the UDP IPv4/6 network transports. Must be in the range of 0 to 15, inclusive. Default is 0.
  • step_threshold — the maximum offset the servo will correct by changing the clock frequency (phase, when using the nullf servo) instead of stepping the clock. When set to 0.0, the servo will never step the clock except on start. It's specified in seconds. Default is 0.0. Fun fact: this option used to be called pi_offset_const.
  • step_window — when set, this is the number of Sync events after a clock step during which the clock does no frequency or step adjustments at all. It exists for situations where clock stepping can't happen instantaneously, so there's a lag before the timestamps can settle properly and reflect the clock step. Default is 0 (disabled).
  • summary_interval — the time interval in which summary statistics of the clock get printed. Specified as a power of two in seconds. The stats include offset root mean square (RMS), maximum absolute offset, frequency offset mean and standard deviation, and path delay mean and standard deviation. Units are nanoseconds and parts per billion (ppb). If there's only one clock update in the interval, the sample gets printed instead of the statistics. The messages print at the LOG_INFO level. Default is 0 (1 second).
  • tc_spanning_tree — when running as a Transparent Clock, increment the "stepsRemoved" field of Announce messages that pass through the switch. Turning this on means PTP message loops never form, as long as all the switches implement this option together with the BMCA.
  • timeSource — the time source is a single byte code that gives an idea of the kind of local clock in use. The value is purely informational, it has no effect on the outcome of the Best Master Clock algorithm, and it gets advertised when the clock becomes grand master.
  • time_stamping — the time stamping method to be used. Allowed values are hardware, software, legacy, onestep, and p2p1step. Default is hardware.
  • twoStepFlag — enables two-step mode for sync messages. One-step mode can only be used with hardware time stamping. Default is 1 (enabled).
  • tx_timestamp_timeout — the number of milliseconds to poll waiting for the tx time stamp from the kernel when a message has recently been sent. Default is 10.
  • udp6_scope — the desired scope for the IPv6 multicast messages. It gets used as the second byte of the primary address. Only relevant with IPv6 transport. See RFC 4291. Default is 0x0E for the global scope.
  • uds_address — the address of the UNIX domain socket for receiving local management messages. Default is /var/run/ptp4l.
  • uds_file_mode — file mode of the UNIX domain socket used for receiving local management messages. Specify the mode as an octal number, so it should start with a 0 literal. Default mode is 0660.
  • uds_ro_address — the address of the second UNIX domain socket for receiving local management messages, which is restricted to GET actions and doesn't forward messages to other ports. You can hand access to this socket to untrusted applications for monitoring purposes. Default is /var/run/ptp4lro.
  • uds_ro_file_mode — file mode of that second (read-only) UNIX domain socket used for receiving local management messages. Mode should be specified as an octal number, i.e. it should start with a 0 literal. Default mode is 0666.
  • use_syslog — print messages to the system log if enabled. Default is 1 (enabled).
  • userDescription — the user description string. Allowed values are of the form name;location and hold at most 128 utf8 symbols. Default is an empty string.
  • utc_offset — the current offset between TAI and UTC. Default is 37.
  • verbose — print messages to the standard output if enabled. Default is 0 (disabled).
  • write_phase_mode — enables using the "write phase" feature of a PTP Hardware Clock. If the device supports it, this mode uses the hardware's built in phase offset control instead of frequency offset control. Default value is 0 (disabled). 🫶

🌈 UNICAST DISCOVERY OPTIONS 🌈

ok so these live inside a [unicast_master_table] situation, and here's the tea 🍵

  • L2|UDPv4|UDPv6 — every single table entry names the transport type and the network address of a potential remote server. And if you list more than one server? Bestie, unicast negotiation gets performed with each of them. Every. Single. One. 💅
  • logQueryInterval — this one sets how long we wait between unicast negotiation attempts. It's given as a power of two, in seconds. Default is 0, which means 1 second. Yes the default is 0 and yes that means 1 second, that's just how it is, don't fight it 🫶
  • peer_address — this is the unicast address of the peer, for use with the peer-to-peer delay mechanism. If you set it, the port that owns the table will negotiate unicast peer delay responses from the machine at that remote address. If you don't set it, the port just sends multicast peer delay requests instead. your call, queen 👑
  • table_id — every table has to start with a unique, positive table ID. The port claims a given table by sticking that ID in as the value of its unicast_master_table option. matchy matchy, iconic ✨

🔐 SECURITY ASSOCIATION OPTIONS 🔐

these are the keys to the kingdom, literally. lock it down 🔒

  • spp — every security association has to begin with a unique spp. The port claims a given security association by including the spp as the value of its spp option. Has to be in the range 0 to 255, inclusive. no wiggle room, no exceptions 💅
  • seqid_window — this defines how far the sequence id of an incoming sync/follow_up message can advance from the last successfully processed sync/follow_up before it gets treated as a replayed message. sync/follow_up seqid tracking resets on port state change and updates on successful processing. Set it to zero and no seqid checking happens at all (living dangerously, respect 😌). Must be in the range 0 to 32767 (0x7FFF), inclusive. Default value is 3.
  • allow_mutable — this allows mutable correction fields when on-path authentication TLV support just isn't possible. > ⚠️ This option is NOT recommended, because it leaves the correction field unprotected. The default value is 0 (disabled). fr fr, leave this one alone unless you have to 🚩
  • key (format: id type length value) — every security association needs at least one key entry. These entries use the format id type length value.
    • Key id: must be in the range 1 to 2^32-1, inclusive.
    • Key type: allowed are SHA256-128, SHA256, AES128, AES256.
    • Key length: optional verification key length in octets. Used to verify the key value is of the length specified.
    • Key value: can be read as ASCII characters with the ASCII: prefix, as a hexadecimal number with the HEX: prefix, or as a Base64 encoded string with the B64: prefix. If no prefix is included, ASCII is assumed. Key values should be randomly generated if possible. Ciphers (AES) require the key length to match the cipher length (ie AES-128 requires a key of 128 bits or 16 bytes). no short kings in the key department 🔑

Example

The following is an example of a security associations file:

[security_association]
spp 0
allow_mutable 1
1 SHA256-128 HEX:F8ADC6B8B8E9AA709106BA42481EC9E29607334DE2C3C737A11A12931DB27F8C
2 SHA256 32 HEX:EE91D469B3A8ADC6AC8EB28E21794C706E08FDE48863828A7B0281AFCA81B17D

[security_association]
spp 1
seqid_window 20
1 AES128 HEX:FAF48EBA01E7C5966A76CB787AED4E7B
2 AES256 B64:PKTF9VQz94qPaoAzW4eE3JtFoQ8Ov1OTQSojalWyMbs=
3 SHA256 26 abcdefghijklmnopqrstuvwxyz

⏰ TIME SCALE USAGE ⏰

when ptp4l is acting as the domain server, it uses either the PTP or the UTC time scale depending on the time stamping mode. the way that it switches is honestly kind of iconic:

  • In software and legacy time stamping modes, it announces Arbitrary time scale mode — which here is effectively UTC.
  • In hardware time stamping mode, it announces use of PTP time scale.

and when ptp4l is the domain server and using hardware time stamping? it's up to phc2sys to maintain the correct offset between UTC and PTP times. see phc2sys(8) manual page for more details. teamwork makes the dream work 🫶

🧵 KTHREAD PRIORITY 🧵

ok so if you ever see this log:

timed out while polling for tx timestamp increasing
tx_timestamp_timeout or increasing kworker priority may correct
this issue, but a driver bug likely causes it

one possible cause is that the kworker which processes timestamps is being starved. The system admin might try manually increasing the priority of the kworker.

Many device drivers use kworker threads created by the PTP stack. Such kworkers are named:

ptp<decimal number of clock>

The system admin can manually bump the priority of the kworker process using chrt.

Example:

pgrep -f "ptp[0-9]+" | xargs -I {} sudo chrt -f --pid 75 {}

Intel ice driver may create multiple kworkers for one physical NIC and names those processes differently.

Example for Intel E810 card:

pgrep -f ice-ptp | xargs -I {} sudo chrt -f --pid 75 {}
⚠️ Assigning priority needs careful consideration, as assigning too high a priority to any task might make the system unstable. be careful bestie, don't crash the vibe 😭

👀 SEE ALSO 👀

and that's the whole page, you're officially a ptp4l girlie now 💖 this was linuxptp, version from March 2024, and the page you just read is PTP4l(8). go forth and sync, my love 🌸✨👑

📦 where this page comes from