1The existing interfaces for getting network packages time stamped are:
2
3* SO_TIMESTAMP
4  Generate time stamp for each incoming packet using the (not necessarily
5  monotonous!) system time. Result is returned via recv_msg() in a
6  control message as timeval (usec resolution).
7
8* SO_TIMESTAMPNS
9  Same time stamping mechanism as SO_TIMESTAMP, but returns result as
10  timespec (nsec resolution).
11
12* IP_MULTICAST_LOOP + SO_TIMESTAMP[NS]
13  Only for multicasts: approximate send time stamp by receiving the looped
14  packet and using its receive time stamp.
15
16The following interface complements the existing ones: receive time
17stamps can be generated and returned for arbitrary packets and much
18closer to the point where the packet is really sent. Time stamps can
19be generated in software (as before) or in hardware (if the hardware
20has such a feature).
21
22SO_TIMESTAMPING:
23
24Instructs the socket layer which kind of information is wanted. The
25parameter is an integer with some of the following bits set. Setting
26other bits is an error and doesn't change the current state.
27
28SOF_TIMESTAMPING_TX_HARDWARE:  try to obtain send time stamp in hardware
29SOF_TIMESTAMPING_TX_SOFTWARE:  if SOF_TIMESTAMPING_TX_HARDWARE is off or
30                               fails, then do it in software
31SOF_TIMESTAMPING_RX_HARDWARE:  return the original, unmodified time stamp
32                               as generated by the hardware
33SOF_TIMESTAMPING_RX_SOFTWARE:  if SOF_TIMESTAMPING_RX_HARDWARE is off or
34                               fails, then do it in software
35SOF_TIMESTAMPING_RAW_HARDWARE: return original raw hardware time stamp
36SOF_TIMESTAMPING_SYS_HARDWARE: return hardware time stamp transformed to
37                               the system time base
38SOF_TIMESTAMPING_SOFTWARE:     return system time stamp generated in
39                               software
40
41SOF_TIMESTAMPING_TX/RX determine how time stamps are generated.
42SOF_TIMESTAMPING_RAW/SYS determine how they are reported in the
43following control message:
44
45struct scm_timestamping {
46	struct timespec systime;
47	struct timespec hwtimetrans;
48	struct timespec hwtimeraw;
49};
50
51recvmsg() can be used to get this control message for regular incoming
52packets. For send time stamps the outgoing packet is looped back to
53the socket's error queue with the send time stamp(s) attached. It can
54be received with recvmsg(flags=MSG_ERRQUEUE). The call returns the
55original outgoing packet data including all headers preprended down to
56and including the link layer, the scm_timestamping control message and
57a sock_extended_err control message with ee_errno==ENOMSG and
58ee_origin==SO_EE_ORIGIN_TIMESTAMPING. A socket with such a pending
59bounced packet is ready for reading as far as select() is concerned.
60If the outgoing packet has to be fragmented, then only the first
61fragment is time stamped and returned to the sending socket.
62
63All three values correspond to the same event in time, but were
64generated in different ways. Each of these values may be empty (= all
65zero), in which case no such value was available. If the application
66is not interested in some of these values, they can be left blank to
67avoid the potential overhead of calculating them.
68
69systime is the value of the system time at that moment. This
70corresponds to the value also returned via SO_TIMESTAMP[NS]. If the
71time stamp was generated by hardware, then this field is
72empty. Otherwise it is filled in if SOF_TIMESTAMPING_SOFTWARE is
73set.
74
75hwtimeraw is the original hardware time stamp. Filled in if
76SOF_TIMESTAMPING_RAW_HARDWARE is set. No assumptions about its
77relation to system time should be made.
78
79hwtimetrans is the hardware time stamp transformed so that it
80corresponds as good as possible to system time. This correlation is
81not perfect; as a consequence, sorting packets received via different
82NICs by their hwtimetrans may differ from the order in which they were
83received. hwtimetrans may be non-monotonic even for the same NIC.
84Filled in if SOF_TIMESTAMPING_SYS_HARDWARE is set. Requires support
85by the network device and will be empty without that support.
86
87
88SIOCSHWTSTAMP:
89
90Hardware time stamping must also be initialized for each device driver
91that is expected to do hardware time stamping. The parameter is defined in
92/include/linux/net_tstamp.h as:
93
94struct hwtstamp_config {
95	int flags;	/* no flags defined right now, must be zero */
96	int tx_type;	/* HWTSTAMP_TX_* */
97	int rx_filter;	/* HWTSTAMP_FILTER_* */
98};
99
100Desired behavior is passed into the kernel and to a specific device by
101calling ioctl(SIOCSHWTSTAMP) with a pointer to a struct ifreq whose
102ifr_data points to a struct hwtstamp_config. The tx_type and
103rx_filter are hints to the driver what it is expected to do. If
104the requested fine-grained filtering for incoming packets is not
105supported, the driver may time stamp more than just the requested types
106of packets.
107
108A driver which supports hardware time stamping shall update the struct
109with the actual, possibly more permissive configuration. If the
110requested packets cannot be time stamped, then nothing should be
111changed and ERANGE shall be returned (in contrast to EINVAL, which
112indicates that SIOCSHWTSTAMP is not supported at all).
113
114Only a processes with admin rights may change the configuration. User
115space is responsible to ensure that multiple processes don't interfere
116with each other and that the settings are reset.
117
118/* possible values for hwtstamp_config->tx_type */
119enum {
120	/*
121	 * no outgoing packet will need hardware time stamping;
122	 * should a packet arrive which asks for it, no hardware
123	 * time stamping will be done
124	 */
125	HWTSTAMP_TX_OFF,
126
127	/*
128	 * enables hardware time stamping for outgoing packets;
129	 * the sender of the packet decides which are to be
130	 * time stamped by setting SOF_TIMESTAMPING_TX_SOFTWARE
131	 * before sending the packet
132	 */
133	HWTSTAMP_TX_ON,
134};
135
136/* possible values for hwtstamp_config->rx_filter */
137enum {
138	/* time stamp no incoming packet at all */
139	HWTSTAMP_FILTER_NONE,
140
141	/* time stamp any incoming packet */
142	HWTSTAMP_FILTER_ALL,
143
144	/* return value: time stamp all packets requested plus some others */
145	HWTSTAMP_FILTER_SOME,
146
147	/* PTP v1, UDP, any kind of event packet */
148	HWTSTAMP_FILTER_PTP_V1_L4_EVENT,
149
150	/* for the complete list of values, please check
151	 * the include file /include/linux/net_tstamp.h
152	 */
153};
154
155
156DEVICE IMPLEMENTATION
157
158A driver which supports hardware time stamping must support the
159SIOCSHWTSTAMP ioctl and update the supplied struct hwtstamp_config with
160the actual values as described in the section on SIOCSHWTSTAMP.
161
162Time stamps for received packets must be stored in the skb. To get a pointer
163to the shared time stamp structure of the skb call skb_hwtstamps(). Then
164set the time stamps in the structure:
165
166struct skb_shared_hwtstamps {
167	/* hardware time stamp transformed into duration
168	 * since arbitrary point in time
169	 */
170	ktime_t	hwtstamp;
171	ktime_t	syststamp; /* hwtstamp transformed to system time base */
172};
173
174Time stamps for outgoing packets are to be generated as follows:
175- In hard_start_xmit(), check if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)
176  is set no-zero. If yes, then the driver is expected to do hardware time
177  stamping.
178- If this is possible for the skb and requested, then declare
179  that the driver is doing the time stamping by setting the flag
180  SKBTX_IN_PROGRESS in skb_shinfo(skb)->tx_flags , e.g. with
181
182      skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
183
184  You might want to keep a pointer to the associated skb for the next step
185  and not free the skb. A driver not supporting hardware time stamping doesn't
186  do that. A driver must never touch sk_buff::tstamp! It is used to store
187  software generated time stamps by the network subsystem.
188- As soon as the driver has sent the packet and/or obtained a
189  hardware time stamp for it, it passes the time stamp back by
190  calling skb_hwtstamp_tx() with the original skb, the raw
191  hardware time stamp. skb_hwtstamp_tx() clones the original skb and
192  adds the timestamps, therefore the original skb has to be freed now.
193  If obtaining the hardware time stamp somehow fails, then the driver
194  should not fall back to software time stamping. The rationale is that
195  this would occur at a later time in the processing pipeline than other
196  software time stamping and therefore could lead to unexpected deltas
197  between time stamps.
198- If the driver did not set the SKBTX_IN_PROGRESS flag (see above), then
199  dev_hard_start_xmit() checks whether software time stamping
200  is wanted as fallback and potentially generates the time stamp.
201