1/*-
2 * SPDX-License-Identifier: BSD-2-Clause
3 *
4 * Copyright (c) 2016-2020 Netflix, Inc.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28#ifndef __tcp_log_buf_h__
29#define __tcp_log_buf_h__
30
31#define TCP_LOG_REASON_LEN 32
32#define TCP_LOG_TAG_LEN 32
33#define TCP_LOG_BUF_VER (9)
34
35/*
36 * Because the (struct tcp_log_buffer) includes 8-byte uint64_t's, it requires
37 * 8-byte alignment to work properly on all platforms. Therefore, we will
38 * enforce 8-byte alignment for all the structures that may appear by
39 * themselves (instead of being embedded in another structure) in a data
40 * stream.
41 */
42#define ALIGN_TCP_LOG __aligned(8)
43
44/* Information about the socketbuffer state. */
45struct tcp_log_sockbuf
46{
47 uint32_t tls_sb_acc; /* available chars (sb->sb_acc) */
48 uint32_t tls_sb_ccc; /* claimed chars (sb->sb_ccc) */
49 uint32_t tls_sb_spare; /* spare */
50};
51
52/* Optional, verbose information that may be appended to an event log. */
53struct tcp_log_verbose
54{
55#define TCP_FUNC_LEN 32
56 char tlv_snd_frm[TCP_FUNC_LEN]; /* tcp_output() caller */
57 char tlv_trace_func[TCP_FUNC_LEN]; /* Function that
58 generated trace */
59 uint32_t tlv_trace_line; /* Line number that generated trace */
60 uint8_t _pad[4];
61} ALIGN_TCP_LOG;
62
63struct tcp_log_bbr {
64 uint64_t cur_del_rate;
65 uint64_t delRate;
66 uint64_t rttProp;
67 uint64_t bw_inuse;
68 uint32_t inflight;
69 uint32_t applimited;
70 uint32_t delivered;
71 uint32_t timeStamp;
72 uint32_t epoch;
73 uint32_t lt_epoch;
74 uint32_t pkts_out;
75 uint32_t flex1;
76 uint32_t flex2;
77 uint32_t flex3;
78 uint32_t flex4;
79 uint32_t flex5;
80 uint32_t flex6;
81 uint32_t lost;
82 uint16_t pacing_gain;
83 uint16_t cwnd_gain;
84 uint16_t flex7;
85 uint8_t bbr_state;
86 uint8_t bbr_substate;
87 uint8_t inhpts;
88 uint8_t __spare;
89 uint8_t use_lt_bw;
90 uint8_t flex8;
91 uint32_t pkt_epoch;
92};
93
94/* shadows tcp_log_bbr struct element sizes */
95struct tcp_log_raw {
96 uint64_t u64_flex[4];
97 uint32_t u32_flex[14];
98 uint16_t u16_flex[3];
99 uint8_t u8_flex[6];
100 uint32_t u32_flex2[1];
101};
102
103struct tcp_log_uint64 {
104 uint64_t u64_flex[13];
105};
106
107struct tcp_log_sendfile {
108 uint64_t offset;
109 uint64_t length;
110 uint32_t flags;
111};
112
113/*
114 * tcp_log_stackspecific is currently being used as "event specific" log
115 * info by all stacks (i.e. struct tcp_log_bbr is used for generic event
116 * logging). Until this is cleaned up more generically and throughout,
117 * allow events to use the same space in the union.
118 */
119union tcp_log_stackspecific
120{
121 struct tcp_log_bbr u_bbr;
122 struct tcp_log_sendfile u_sf;
123 struct tcp_log_raw u_raw; /* "raw" log access */
124 struct tcp_log_uint64 u64_raw; /* just u64's - used by process info */
125};
126
127typedef union tcp_log_stackspecific tcp_log_eventspecific_t;
128
129struct tcp_log_buffer
130{
131 /* Event basics */
132 struct timeval tlb_tv; /* Timestamp of trace */
133 uint32_t tlb_ticks; /* Timestamp of trace */
134 uint32_t tlb_sn; /* Serial number */
135 uint8_t tlb_stackid; /* Stack ID */
136 uint8_t tlb_eventid; /* Event ID */
137 uint16_t tlb_eventflags; /* Flags for the record */
138#define TLB_FLAG_RXBUF 0x0001 /* Includes receive buffer info */
139#define TLB_FLAG_TXBUF 0x0002 /* Includes send buffer info */
140#define TLB_FLAG_HDR 0x0004 /* Includes a TCP header */
141#define TLB_FLAG_VERBOSE 0x0008 /* Includes function/line numbers */
142#define TLB_FLAG_STACKINFO 0x0010 /* Includes stack-specific info */
143 int tlb_errno; /* Event error (if any) */
144
145 /* Internal session state */
146 struct tcp_log_sockbuf tlb_rxbuf; /* Receive buffer */
147 struct tcp_log_sockbuf tlb_txbuf; /* Send buffer */
148
149 int tlb_state; /* TCPCB t_state */
150 uint32_t tlb_starttime; /* TCPCB t_starttime */
151 uint32_t tlb_iss; /* TCPCB iss */
152 uint32_t tlb_flags; /* TCPCB flags */
153 uint32_t tlb_snd_una; /* TCPCB snd_una */
154 uint32_t tlb_snd_max; /* TCPCB snd_max */
155 uint32_t tlb_snd_cwnd; /* TCPCB snd_cwnd */
156 uint32_t tlb_snd_nxt; /* TCPCB snd_nxt */
157 uint32_t tlb_snd_recover;/* TCPCB snd_recover */
158 uint32_t tlb_snd_wnd; /* TCPCB snd_wnd */
159 uint32_t tlb_snd_ssthresh; /* TCPCB snd_ssthresh */
160 uint32_t tlb_srtt; /* TCPCB t_srtt */
161 uint32_t tlb_rttvar; /* TCPCB t_rttvar */
162 uint32_t tlb_rcv_up; /* TCPCB rcv_up */
163 uint32_t tlb_rcv_adv; /* TCPCB rcv_adv */
164 uint32_t tlb_flags2; /* TCPCB t_flags2 */
165 uint32_t tlb_rcv_nxt; /* TCPCB rcv_nxt */
166 uint32_t tlb_rcv_wnd; /* TCPCB rcv_wnd */
167 uint32_t tlb_dupacks; /* TCPCB t_dupacks */
168 int tlb_segqlen; /* TCPCB segqlen */
169 int tlb_snd_numholes; /* TCPCB snd_numholes */
170 uint32_t tlb_flex1; /* Event specific information */
171 uint32_t tlb_flex2; /* Event specific information */
172 uint32_t tlb_fbyte_in; /* TCPCB first byte in time */
173 uint32_t tlb_fbyte_out; /* TCPCB first byte out time */
174 uint8_t tlb_snd_scale:4, /* TCPCB snd_scale */
175 tlb_rcv_scale:4; /* TCPCB rcv_scale */
176 uint8_t _pad[3]; /* Padding */
177 /* Per-stack info */
178 union tcp_log_stackspecific tlb_stackinfo;
179
180 /* The packet */
181 uint32_t tlb_len; /* The packet's data length */
182 struct tcphdr tlb_th; /* The TCP header */
183 uint8_t tlb_opts[TCP_MAXOLEN]; /* The TCP options */
184
185 /* Verbose information (optional) */
186 struct tcp_log_verbose tlb_verbose[0];
187} ALIGN_TCP_LOG;
188
189enum tcp_log_events {
190 TCP_LOG_IN = 1, /* Incoming packet 1 */
191 TCP_LOG_OUT, /* Transmit (without other event) 2 */
192 TCP_LOG_RTO, /* Retransmit timeout 3 */
193 TCP_LOG_SB_WAKE, /* Awaken socket buffer 4 */
194 TCP_UNUSED_5, /* Detected bad retransmission 5 */
195 TCP_LOG_PRR, /* Doing PRR 6 */
196 TCP_UNUSED_7, /* Detected reorder 7 */
197 TCP_LOG_HPTS, /* Hpts sending a packet 8 */
198 BBR_LOG_BBRUPD, /* We updated BBR info 9 */
199 BBR_LOG_BBRSND, /* We did a slot calculation and sending is done 10 */
200 BBR_LOG_ACKCLEAR, /* A ack clears all outstanding 11 */
201 TCP_UNUSED_12, /* The tcb had a packet input to it 12 */
202 BBR_LOG_TIMERSTAR, /* Start a timer 13 */
203 BBR_LOG_TIMERCANC, /* Cancel a timer 14 */
204 BBR_LOG_ENTREC, /* Entered recovery 15 */
205 BBR_LOG_EXITREC, /* Exited recovery 16 */
206 BBR_LOG_CWND, /* Cwnd change 17 */
207 BBR_LOG_BWSAMP, /* LT B/W sample has been made 18 */
208 BBR_LOG_MSGSIZE, /* We received a EMSGSIZE error 19 */
209 BBR_LOG_BBRRTT, /* BBR RTT is updated 20 */
210 BBR_LOG_JUSTRET, /* We just returned out of output 21 */
211 BBR_LOG_STATE, /* A BBR state change occurred 22 */
212 BBR_LOG_PKT_EPOCH, /* A BBR packet epoch occurred 23 */
213 BBR_LOG_PERSIST, /* BBR changed to/from a persists 24 */
214 TCP_LOG_FLOWEND, /* End of a flow 25 */
215 BBR_LOG_RTO, /* BBR's timeout includes BBR info 26 */
216 BBR_LOG_DOSEG_DONE, /* hpts do_segment completes 27 */
217 BBR_LOG_EXIT_GAIN, /* hpts do_segment completes 28 */
218 BBR_LOG_THRESH_CALC, /* Doing threshold calculation 29 */
219 TCP_LOG_MAPCHG, /* Map Changes to the sendmap 30 */
220 TCP_LOG_USERSEND, /* User level sends data 31 */
221 BBR_RSM_CLEARED, /* RSM cleared of ACK flags 32 */
222 BBR_LOG_STATE_TARGET, /* Log of target at state 33 */
223 BBR_LOG_TIME_EPOCH, /* A timed based Epoch occurred 34 */
224 BBR_LOG_TO_PROCESS, /* A to was processed 35 */
225 BBR_LOG_BBRTSO, /* TSO update 36 */
226 BBR_LOG_HPTSDIAG, /* Hpts diag insert 37 */
227 BBR_LOG_LOWGAIN, /* Low gain accounting 38 */
228 BBR_LOG_PROGRESS, /* Progress timer event 39 */
229 TCP_LOG_SOCKET_OPT, /* A socket option is set 40 */
230 BBR_LOG_TIMERPREP, /* A BBR var to debug out TLP issues 41 */
231 BBR_LOG_ENOBUF_JMP, /* We had a enobuf jump 42 */
232 BBR_LOG_HPTSI_CALC, /* calc the hptsi time 43 */
233 BBR_LOG_RTT_SHRINKS, /* We had a log reduction of rttProp 44 */
234 BBR_LOG_BW_RED_EV, /* B/W reduction events 45 */
235 BBR_LOG_REDUCE, /* old bbr log reduce for 4.1 and earlier 46*/
236 TCP_LOG_RTT, /* A rtt (in useconds) is being sampled and applied to the srtt algo 47 */
237 BBR_LOG_SETTINGS_CHG, /* Settings changed for loss response 48 */
238 TCP_UNUSED_49, /* SRTT gaining -- now not used 49 */
239 TCP_LOG_REASS, /* Reassembly buffer logging 50 */
240 TCP_HDWR_PACE_SIZE, /* TCP pacing size set (rl and rack uses this) 51 */
241 BBR_LOG_HDWR_PACE, /* TCP Hardware pacing log 52 */
242 BBR_LOG_TSTMP_VAL, /* Temp debug timestamp validation 53 */
243 TCP_LOG_CONNEND, /* End of connection 54 */
244 TCP_LOG_LRO, /* LRO entry 55 */
245 TCP_SACK_FILTER_RES, /* Results of SACK Filter 56 */
246 TCP_UNUSED_57, /* Sack Attack Detection 57 */
247 TCP_TIMELY_WORK, /* Logs regarding Timely CC tweaks 58 */
248 TCP_UNUSED_59, /* User space event data 59 */
249 TCP_LOG_SENDFILE, /* sendfile() logging for TCP connections 60 */
250 TCP_LOG_REQ_T, /* logging of request tracking 61 */
251 TCP_LOG_ACCOUNTING, /* Log of TCP Accounting data 62 */
252 TCP_LOG_FSB, /* FSB information 63 */
253 RACK_DSACK_HANDLING, /* Handling of DSACK in rack for reordering window 64 */
254 TCP_HYSTART, /* TCP Hystart logging 65 */
255 TCP_CHG_QUERY, /* Change query during fnc_init() 66 */
256 TCP_RACK_LOG_COLLAPSE, /* Window collapse by peer 67 */
257 TCP_RACK_TP_TRIGGERED, /* A rack tracepoint is triggered 68 */
258 TCP_HYBRID_PACING_LOG, /* Hybrid pacing log 69 */
259 TCP_LOG_PRU, /* TCP protocol user request 70 */
260 TCP_UNUSED_71, /* old TCP Policer detectionn, not used 71 */
261 TCP_PCM_MEASURE, /* TCP Path Capacity Measurement 72 */
262 TCP_LOG_END /* End (keep at end) 73 */
263};
264
265enum tcp_log_states {
266 TCP_LOG_STATE_RATIO_OFF = -2, /* Log ratio evaluation yielded an OFF
267 result. Only used for tlb_logstate */
268 TCP_LOG_STATE_CLEAR = -1, /* Deactivate and clear tracing. Passed
269 to tcp_log_state_change() but never
270 stored in any logstate variable */
271 TCP_LOG_STATE_OFF = 0, /* Pause */
272
273 /* Positively numbered states represent active logging modes */
274 TCP_LOG_STATE_TAIL=1, /* Keep the trailing events */
275 TCP_LOG_STATE_HEAD=2, /* Keep the leading events */
276 TCP_LOG_STATE_HEAD_AUTO=3, /* Keep the leading events, and
277 automatically dump them to the
278 device */
279 TCP_LOG_STATE_CONTINUAL=4, /* Continually dump the data when full */
280 TCP_LOG_STATE_TAIL_AUTO=5, /* Keep the trailing events, and
281 automatically dump them when the
282 session ends */
283 TCP_LOG_VIA_BBPOINTS=6 /* Log only if the BB point has been configured */
284};
285
286/* Use this if we don't know whether the operation succeeded. */
287#define ERRNO_UNK (-1)
288
289/*
290 * If the user included dev/tcp_log/tcp_log_dev.h, then include our private
291 * headers. Otherwise, there is no reason to pollute all the files with an
292 * additional include.
293 *
294 * This structure is aligned to an 8-byte boundary to match the alignment
295 * requirements of (struct tcp_log_buffer).
296 */
297#ifdef __tcp_log_dev_h__
298struct tcp_log_header {
299 struct tcp_log_common_header tlh_common;
300#define tlh_version tlh_common.tlch_version
301#define tlh_type tlh_common.tlch_type
302#define tlh_length tlh_common.tlch_length
303 struct in_endpoints tlh_ie;
304 struct timeval tlh_offset; /* Uptime -> UTC offset */
305 char tlh_id[TCP_LOG_ID_LEN];
306 char tlh_reason[TCP_LOG_REASON_LEN];
307 char tlh_tag[TCP_LOG_TAG_LEN];
308 uint8_t tlh_af;
309 uint8_t _pad[7];
310} ALIGN_TCP_LOG;
311
312#ifdef _KERNEL
313struct tcp_log_dev_log_queue {
314 struct tcp_log_dev_queue tldl_common;
315 char tldl_id[TCP_LOG_ID_LEN];
316 char tldl_reason[TCP_LOG_REASON_LEN];
317 char tldl_tag[TCP_LOG_TAG_LEN];
318 struct in_endpoints tldl_ie;
319 struct tcp_log_stailq tldl_entries;
320 int tldl_count;
321 uint8_t tldl_af;
322};
323#endif /* _KERNEL */
324#endif /* __tcp_log_dev_h__ */
325
326/*
327 * Defined BBPOINTS that can be used
328 * with TCP_LOG_VIA_BBPOINTS.
329 */
330#define TCP_BBPOINT_NONE 0
331#define TCP_BBPOINT_REQ_LEVEL_LOGGING 1
332
333/*********************/
334/* TCP Trace points */
335/*********************/
336/*
337 * TCP trace points are interesting points within
338 * the TCP code that the author/debugger may want
339 * to have BB logging enabled if we hit that point.
340 * In order to enable a trace point you set the
341 * sysctl var net.inet.tcp.bb.tp.number to
342 * one of the numbers listed below. You also
343 * must make sure net.inet.tcp.bb.tp.bbmode is
344 * non-zero, the default is 4 for continuous tracing.
345 * You also set in the number of connections you want
346 * have get BB logs in net.inet.tcp.bb.tp.count.
347 *
348 * Count will decrement every time BB logging is assigned
349 * to a connection that hit your tracepoint.
350 *
351 * You can enable all trace points by setting the number
352 * to 0xffffffff. You can disable all trace points by
353 * setting number to zero (or count to 0).
354 *
355 * Below are the enumerated list of tracepoints that
356 * have currently been defined in the code. Add more
357 * as you add a call to rack_trace_point(rack, <name>);
358 * where <name> is defined below.
359 */
360#define TCP_TP_HWENOBUF 0x00000001 /* When we are doing hardware pacing and hit enobufs */
361#define TCP_TP_ENOBUF 0x00000002 /* When we hit enobufs with software pacing */
362#define TCP_TP_COLLAPSED_WND 0x00000003 /* When a peer to collapses its rwnd on us */
363#define TCP_TP_COLLAPSED_RXT 0x00000004 /* When we actually retransmit a collapsed window rsm */
364#define TCP_TP_REQ_LOG_FAIL 0x00000005 /* We tried to allocate a Request log but had no space */
365#define TCP_TP_RESET_RCV 0x00000006 /* Triggers when we receive a RST */
366#define TCP_TP_POLICER_DET 0x00000007 /* When we detect a policer */
367#define TCP_TP_EXCESS_RXT TCP_TP_POLICER_DET /* alias */
368#define TCP_TP_SAD_TRIGGERED 0x00000008 /* Sack Attack Detection triggers */
369#define TCP_TP_SAD_SUSPECT 0x0000000a /* A sack has supicious information in it */
370#define TCP_TP_PACED_BOTTOM 0x0000000b /* We have paced at the bottom */
371
372#ifdef _KERNEL
373
374extern uint32_t tcp_trace_point_config;
375extern uint32_t tcp_trace_point_bb_mode;
376extern int32_t tcp_trace_point_count;
377
378/*
379 * Returns true if any sort of BB logging is enabled,
380 * commonly used throughout the codebase.
381 */
382static inline int
383tcp_bblogging_on(struct tcpcb *tp)
384{
385 if (tp->_t_logstate <= TCP_LOG_STATE_OFF)
386 return (0);
387 if (tp->_t_logstate == TCP_LOG_VIA_BBPOINTS)
388 return (0);
389 return (1);
390}
391
392/*
393 * Returns true if we match a specific bbpoint when
394 * in TCP_LOG_VIA_BBPOINTS, but also returns true
395 * for all the other logging states.
396 */
397static inline int
398tcp_bblogging_point_on(struct tcpcb *tp, uint8_t bbpoint)
399{
400 if (tp->_t_logstate <= TCP_LOG_STATE_OFF)
401 return (0);
402 if ((tp->_t_logstate == TCP_LOG_VIA_BBPOINTS) &&
403 (tp->_t_logpoint == bbpoint))
404 return (1);
405 else if (tp->_t_logstate == TCP_LOG_VIA_BBPOINTS)
406 return (0);
407 return (1);
408}
409
410static inline void
411tcp_set_bblog_state(struct tcpcb *tp, uint8_t ls, uint8_t bbpoint)
412{
413 if ((ls == TCP_LOG_VIA_BBPOINTS) &&
414 (tp->_t_logstate == TCP_LOG_STATE_OFF)){
415 /*
416 * We don't allow a BBPOINTS set to override
417 * other types of BB logging set by other means such
418 * as the bb_ratio/bb_state URL parameters. In other
419 * words BBlogging must be *off* in order to turn on
420 * a BBpoint.
421 */
422 tp->_t_logpoint = bbpoint;
423 tp->_t_logstate = ls;
424 } else if (ls < TCP_LOG_VIA_BBPOINTS) {
425 tp->_t_logpoint = TCP_BBPOINT_NONE;
426 tp->_t_logstate = ls;
427 }
428}
429
430static inline uint32_t
431tcp_get_bblog_state(struct tcpcb *tp)
432{
433 return (tp->_t_logstate);
434}
435
436static inline void
437tcp_trace_point(struct tcpcb *tp, int num)
438{
439#ifdef TCP_BLACKBOX
440 if (((tcp_trace_point_config == num) ||
441 (tcp_trace_point_config == 0xffffffff)) &&
442 (tcp_trace_point_bb_mode != 0) &&
443 (tcp_trace_point_count > 0) &&
444 (tcp_bblogging_on(tp) == 0)) {
445 int res;
446 res = atomic_fetchadd_int(&tcp_trace_point_count, -1);
447 if (res > 0) {
448 tcp_set_bblog_state(tp, tcp_trace_point_bb_mode, TCP_BBPOINT_NONE);
449 } else {
450 /* Loss a race assure its zero now */
451 tcp_trace_point_count = 0;
452 }
453 }
454#endif
455}
456
457#define TCP_LOG_BUF_DEFAULT_SESSION_LIMIT 5000
458#define TCP_LOG_BUF_DEFAULT_GLOBAL_LIMIT 5000000
459
460/*
461 * TCP_LOG_EVENT_VERBOSE: The same as TCP_LOG_EVENT, except it always
462 * tries to record verbose information.
463 */
464#define TCP_LOG_EVENT_VERBOSE(tp, th, rxbuf, txbuf, eventid, errornum, len, stackinfo, th_hostorder, tv) \
465 do { \
466 if (tcp_bblogging_on(tp)) \
467 tcp_log_event(tp, th, rxbuf, txbuf, eventid, \
468 errornum, len, stackinfo, th_hostorder, \
469 tp->t_output_caller, __func__, __LINE__, tv);\
470 } while (0)
471
472/*
473 * TCP_LOG_EVENT: This is a macro so we can capture function/line
474 * information when needed. You can use the macro when you are not
475 * doing a lot of prep in the stack specific information i.e. you
476 * don't add extras (stackinfo). If you are adding extras which
477 * means filling out a stack variable instead use the tcp_log_event()
478 * function but enclose the call to the log (and all the setup) in a
479 * if (tcp_bblogging_on(tp)) {
480 * ... setup and logging call ...
481 * }
482 *
483 * Always use the macro tcp_bblogging_on() since sometimes the defintions
484 * do change.
485 *
486 * BBlogging also supports the concept of a BBpoint. The idea behind this
487 * is that when you set a specific BBpoint on and turn the logging into
488 * the BBpoint mode (TCP_LOG_VIA_BBPOINTS) you will be defining very very
489 * few of these points to come out. The point is specific to a code you
490 * want tied to that one BB logging. This allows you to turn on a much broader
491 * scale set of limited logging on more connections without overwhelming the
492 * I/O system with too much BBlogs. This of course means you need to be quite
493 * careful on how many BBlogs go with each point, but you can have multiple points
494 * only one of which is active at a time.
495 *
496 * To define a point you add it above under the define for TCP_BBPOINT_NONE (which
497 * is the default i.e. no point is defined. You then, for your point use the
498 * tcp_bblogging_point_on(struct tcpcb *tp, uint8_t bbpoint) inline to enclose
499 * your call to tcp_log_event. Do not use one of the TCP_LOGGING macros else
500 * your point will never come out. You specify your defined point in the bbpoint
501 * side of the inline. An example of this you can find in rack where the
502 * TCP_BBPOINT_REQ_LEVEL_LOGGING is used. There a specific set of logs are generated
503 * for each request that tcp is tracking.
504 *
505 * When turning on BB logging use the inline:
506 * tcp_set_bblog_state(struct tcpcb *tp, uint8_t ls, uint8_t bbpoint)
507 * the ls field is the logging state TCP_LOG_STATE_CONTINUAL etc. The
508 * bbpoint field is ignored unless the ls field is set to TCP_LOG_VIA_BBPOINTS.
509 * Currently there is only a socket option that turns on the non-BBPOINT
510 * logging.
511 *
512 * Prototype:
513 * TCP_LOG_EVENT(struct tcpcb *tp, struct tcphdr *th, struct sockbuf *rxbuf,
514 * struct sockbuf *txbuf, uint8_t eventid, int errornum,
515 * union tcp_log_stackspecific *stackinfo)
516 *
517 * tp is mandatory and must be write locked.
518 * th is optional; if present, it will appear in the record.
519 * rxbuf and txbuf are optional; if present, they will appear in the record.
520 * eventid is mandatory.
521 * errornum is mandatory (it indicates the success or failure of the
522 * operation associated with the event).
523 * len indicates the length of the packet. If no packet, use 0.
524 * stackinfo is optional; if present, it will appear in the record.
525 */
526struct tcpcb;
527#ifdef TCP_LOG_FORCEVERBOSE
528#define TCP_LOG_EVENT TCP_LOG_EVENT_VERBOSE
529#else
530#define TCP_LOG_EVENT(tp, th, rxbuf, txbuf, eventid, errornum, len, stackinfo, th_hostorder) \
531 do { \
532 if (tcp_log_verbose) \
533 TCP_LOG_EVENT_VERBOSE(tp, th, rxbuf, txbuf, \
534 eventid, errornum, len, stackinfo, \
535 th_hostorder, NULL); \
536 else if (tcp_bblogging_on(tp)) \
537 tcp_log_event(tp, th, rxbuf, txbuf, eventid, \
538 errornum, len, stackinfo, th_hostorder, \
539 NULL, NULL, 0, NULL); \
540 } while (0)
541#endif /* TCP_LOG_FORCEVERBOSE */
542/* Assumes/requires the caller has already checked tcp_bblogging_on(tp). */
543#define TCP_LOG_EVENTP(tp, th, rxbuf, txbuf, eventid, errornum, len, stackinfo, th_hostorder, tv) \
544 do { \
545 KASSERT(tcp_bblogging_on(tp), ("bblogging is off")); \
546 tcp_log_event(tp, th, rxbuf, txbuf, eventid, errornum, len, \
547 stackinfo, th_hostorder, NULL, NULL, 0, tv); \
548 } while (0)
549
550#ifdef TCP_BLACKBOX
551extern bool tcp_log_verbose;
552void tcp_log_drain(struct tcpcb *tp);
553int tcp_log_dump_tp_logbuf(struct tcpcb *tp, char *reason, int how, bool force);
554void tcp_log_dump_tp_bucket_logbufs(struct tcpcb *tp, char *reason);
555struct tcp_log_buffer *tcp_log_event(struct tcpcb *tp, struct tcphdr *th, struct sockbuf *rxbuf,
556 struct sockbuf *txbuf, uint8_t eventid, int errornum, uint32_t len,
557 union tcp_log_stackspecific *stackinfo, int th_hostorder,
558 const char *output_caller, const char *func, int line, const struct timeval *tv);
559size_t tcp_log_get_id(struct tcpcb *tp, char *buf);
560size_t tcp_log_get_tag(struct tcpcb *tp, char *buf);
561u_int tcp_log_get_id_cnt(struct tcpcb *tp);
562int tcp_log_getlogbuf(struct sockopt *sopt, struct tcpcb *tp);
563void tcp_log_init(void);
564int tcp_log_set_id(struct tcpcb *tp, char *id);
565int tcp_log_set_tag(struct tcpcb *tp, char *tag);
566int tcp_log_state_change(struct tcpcb *tp, int state);
567void tcp_log_tcpcbinit(struct tcpcb *tp);
568void tcp_log_tcpcbfini(struct tcpcb *tp);
569void tcp_log_flowend(struct tcpcb *tp);
570void tcp_log_sendfile(struct socket *so, off_t offset, size_t nbytes,
571 int flags);
572int tcp_log_apply_ratio(struct tcpcb *tp, int ratio);
573#ifdef DDB
574void db_print_bblog_entries(struct tcp_log_stailq *log_entries, int indent);
575#endif
576#else /* !TCP_BLACKBOX */
577#define tcp_log_verbose (false)
578
579static inline struct tcp_log_buffer *
580tcp_log_event(struct tcpcb *tp, struct tcphdr *th, struct sockbuf *rxbuf,
581 struct sockbuf *txbuf, uint8_t eventid, int errornum, uint32_t len,
582 union tcp_log_stackspecific *stackinfo, int th_hostorder,
583 const char *output_caller, const char *func, int line,
584 const struct timeval *tv)
585{
586
587 return (NULL);
588}
589#endif /* TCP_BLACKBOX */
590
591#endif /* _KERNEL */
592#endif /* __tcp_log_buf_h__ */