| 1 |
2024-09-14 Dmitry Butskoy <[email protected]> - 2.1.6 |
| 2 |
|
| 3 |
* Let getaddrinfo(3) select the default IPv4 or IPv6 protocol version |
| 4 |
when it is not explicitly specified on the command line |
| 5 |
(Jan Macku, SF bug #16) |
| 6 |
|
| 7 |
No more mandatory default to IPv4, follow rfc3484 and |
| 8 |
the similar ping(1) behaviour. Drop DEF_AF macro. |
| 9 |
|
| 10 |
|
| 11 |
2023-12-18 Dmitry Butskoy <[email protected]> - 2.1.5 |
| 12 |
|
| 13 |
* Fix rfc5837 parsing (Francois Rigault) |
| 14 |
|
| 15 |
|
| 16 |
2023-12-13 Dmitry Butskoy <[email protected]> - 2.1.4 |
| 17 |
|
| 18 |
* Parse interface information (rfc5837) for ICMP extensions |
| 19 |
|
| 20 |
* Add `fastopen' tcp module option (cookie negotiation only) |
| 21 |
|
| 22 |
* Complete tcp module option `mss' to discover possible mss clamping |
| 23 |
along the path being traced (idea and testing from Francois Rigault). |
| 24 |
The argument is optional now. |
| 25 |
|
| 26 |
Changed mss is printed once in a form of `M=NUM' at the first probe |
| 27 |
it was detected on. (Actually, the mss clamping performed by |
| 28 |
some previous hop). |
| 29 |
|
| 30 |
Note, some routers may return too short original fragment |
| 31 |
in the time exceeded message, making the check impossible. |
| 32 |
Besides that the responses may come in a different order. |
| 33 |
All this can lead to a later place of the report |
| 34 |
(using -N 1 can help for the order). |
| 35 |
|
| 36 |
* Complete tcp module option `info' to print returned tcp header options too |
| 37 |
(all those that can be set or altered by `-O' for tcp module). |
| 38 |
|
| 39 |
|
| 40 |
2023-08-30 Dmitry Butskoy <[email protected]> - 2.1.3 |
| 41 |
|
| 42 |
* Fix command line parsing in wrappers. |
| 43 |
|
| 44 |
|
| 45 |
2023-02-13 Dmitry Butskoy <[email protected]> - 2.1.2 |
| 46 |
|
| 47 |
* Fix unprivileged ICMP tracerouting with Linux kernel >= 6.1 |
| 48 |
(Eric Dumazet, SF bug #14) |
| 49 |
|
| 50 |
|
| 51 |
2022-12-27 Dmitry Butskoy <[email protected]> - 2.1.1 |
| 52 |
|
| 53 |
* Interpret ipv4-mapped ipv6 addresses (::ffff:A.B.C.D) as true ipv4. |
| 54 |
|
| 55 |
There are no ipv4-mapped addresses in the real network which we |
| 56 |
operate on, so use just ipv4 in such cases, but allow users |
| 57 |
to specify it this way for convenience. |
| 58 |
|
| 59 |
* Return back more robast poll(2) loop handling. |
| 60 |
|
| 61 |
|
| 62 |
2016-03-08 Dmitry Butskoy <[email protected]> - 2.1.0 |
| 63 |
|
| 64 |
* Improve the main loop for better interactivity. |
| 65 |
|
| 66 |
Instead of waiting silently for maximum expiration time of probes |
| 67 |
in progress, use timeout of the first probe (which will be printed |
| 68 |
first from now) only. |
| 69 |
|
| 70 |
* Speedup wait mechanism. |
| 71 |
|
| 72 |
Traditional traceroute implementation always waited the whole timeout |
| 73 |
for any probe. But if we already have some replies from the same hop, |
| 74 |
or even from some next hop, we can use the round trip time |
| 75 |
of such a reply as a hint to determine the actual reasonable |
| 76 |
amount of time to wait. |
| 77 |
|
| 78 |
Now the `-w' option has a form of three (in general) float values |
| 79 |
separated by a comma (or a slash): `-w MAX_SECS,HERE,NEAR' . |
| 80 |
(last two are optional). MAX_SECS specifies the maximum time |
| 81 |
(in seconds) to wait, in any case. |
| 82 |
|
| 83 |
The optional HERE specifies a factor to multiply the round trip time |
| 84 |
of an already received response from the same hop. |
| 85 |
The resulting value is used as a timeout for the probe, instead of |
| 86 |
(but no more than) MAX_SECS. The optional NEAR specifies a similar |
| 87 |
factor for a response from some next hop. |
| 88 |
The time of the first found result is used in both cases. |
| 89 |
|
| 90 |
First, we look for the same hop (of the probe which will be printed |
| 91 |
first from now). If nothing found, then look for some next hop. |
| 92 |
If nothing found, use MAX_SECS. If HERE and/or NEAR have zero values, |
| 93 |
the corresponding computation is skipped. |
| 94 |
|
| 95 |
HERE and NEAR are always set to zero if only MAX_SECS is specified |
| 96 |
(which provides compatibility with previous versions). Thus, if your |
| 97 |
scripts use `-w SECS', then nothing changed for you, since |
| 98 |
the lonely SECS implies `-w SECS,0,0' . |
| 99 |
|
| 100 |
Defaults are 5.0 seconds for MAX_SECS, 3.0 times for HERE and |
| 101 |
10.0 times for NEAR. |
| 102 |
|
| 103 |
Certainly, the new algorithm can lead to premature expiry |
| 104 |
(especially when response times differ at times) and printing "*" |
| 105 |
instead of a time. Anyway, you can always switch this algorithm off, |
| 106 |
just by specifying `-w' with the desired timeout only (fe. `-w 5'). |
| 107 |
|
| 108 |
We continue to wait whole MAX_SECS when one probe per time |
| 109 |
must be sent (`--sport', `-P proto'), because it seems more harmful |
| 110 |
rather than helpful to try to wait less in such cases. |
| 111 |
|
| 112 |
To provide compatibility with 2.0.x versions, use: |
| 113 |
|
| 114 |
traceroute -w 5 |
| 115 |
|
| 116 |
(or any other desired `-w' value). |
| 117 |
|
| 118 |
* Hint people to use the system traceroute(8) instead of |
| 119 |
tcptraceroute wrapper (by providing a stderr header). |
| 120 |
|
| 121 |
The using of this wrapper is a little bit harmful, since it has |
| 122 |
less possibilities and a little different set of options. |
| 123 |
|
| 124 |
For those who are used to use tcptraceroute in cmdline, |
| 125 |
just create a link with that name to the system traceroute. |
| 126 |
When invoked as "tcp*", it then behaves as `traceroute -T'. |
| 127 |
(The simple manual page added for this case in the wrapper subdir). |
| 128 |
|
| 129 |
The original tcptraceroute had some options differ ("lpNSAE"), |
| 130 |
but they was rare used. Most common "dnFifmqwst" was just the same. |
| 131 |
Therefore it should be painless to use the system binary directly, |
| 132 |
instead of the limited wrapper (which is still provided indeed). |
| 133 |
|
| 134 |
|
| 135 |
2016-02-15 Dmitry Butskoy <[email protected]> - 2.0.22 |
| 136 |
|
| 137 |
* Some portability fixing and improvements (Felix Janda) |
| 138 |
|
| 139 |
* Require clear numbers for options and arguments (Sergey Salnikov) |
| 140 |
|
| 141 |
* Drop compilation date from the version string (Debian #774365) |
| 142 |
|
| 143 |
* New tcp module option `reuse', which utilize SO_REUSEADDR |
| 144 |
to reuse local port numbers for the huge workloads (Richard Sheehan) |
| 145 |
|
| 146 |
* Avoid poll(2) call with spurious zero timeout in some rare cases |
| 147 |
by rounding the value properly using ceil(3) |
| 148 |
|
| 149 |
|
| 150 |
2014-11-12 Dmitry Butskoy <[email protected]> - 2.0.21 |
| 151 |
|
| 152 |
* Fix `--mtu' and `-F' working on kernels >= 3.13 |
| 153 |
|
| 154 |
* Some manual page improving (Christopher Mann) |
| 155 |
|
| 156 |
|
| 157 |
2014-06-14 Dmitry Butskoy <[email protected]> - 2.0.20 |
| 158 |
|
| 159 |
* Describe all complementary long options in the man page (Jan Synacek) |
| 160 |
|
| 161 |
* Use correct service name for AS lookups (Frederic Mangano) |
| 162 |
|
| 163 |
* Avoid some rare case null dereference ([email protected]) |
| 164 |
|
| 165 |
* Improve expiration check for simultaneous probes |
| 166 |
|
| 167 |
|
| 168 |
2012-11-19 Dmitry Butskoy <[email protected]> - 2.0.19 |
| 169 |
|
| 170 |
* DCCP protocol support (rfc4340), by Samuel Jero |
| 171 |
|
| 172 |
Use "-D" option for it (the protocol-specific options |
| 173 |
are available too). |
| 174 |
|
| 175 |
* Update COPYING and COPYING.LIB license files to the latest |
| 176 |
published ones (due to FSF address changes etc.) (Jan Synacek) |
| 177 |
|
| 178 |
* Add mention of "-l" option to manual (Filip Holec) |
| 179 |
|
| 180 |
|
| 181 |
2011-08-16 Dmitry Butskoy <[email protected]> - 2.0.18 |
| 182 |
|
| 183 |
* Handle new dgram icmp sockets ("echo ping sockets"), |
| 184 |
appeared in kernel 3.0 . |
| 185 |
|
| 186 |
Now unprivileged users may perform ICMP tracerouting |
| 187 |
without any special rights of the executable |
| 188 |
(neither setuid bits nor cap_net_raw settings). |
| 189 |
It is allowed if any group of a user matches sysctl range |
| 190 |
of "net/ipv4/ping_group_range". |
| 191 |
|
| 192 |
The support for dgram icmp way (and whether it is allowed) |
| 193 |
is auto-detected at runtime. First, the traditional raw socket |
| 194 |
is tried (for full compatibility reasons), then new dgram |
| 195 |
socket as a fallback. |
| 196 |
The icmp module now has two additional options "raw" and "dgram", |
| 197 |
which cause to try one particular way only. |
| 198 |
|
| 199 |
Note, that there is no IPv6 implementation for dgram icmp sockets |
| 200 |
in kernels 3.0 yet, but new traceroute is ready for it anyway. |
| 201 |
|
| 202 |
* New tcp module option `info' ("-T -O info"), |
| 203 |
which prints all tcp flags of tcp reply from the reached |
| 204 |
target host. |
| 205 |
|
| 206 |
The flags are shown comma-separated in the same place |
| 207 |
where icmp extensions is printed (ie. in `<>' brackets) |
| 208 |
|
| 209 |
This feature is utilized by tcptraceroute wrapper now, |
| 210 |
and allow it to be completely functional replacement |
| 211 |
of the original tcptraceroute. |
| 212 |
|
| 213 |
* Fix determination of system-wide ECN setings for tcp module. |
| 214 |
|
| 215 |
Since the kernel 2.6.31 the default sysctl net/ipv4/tcp_ecn |
| 216 |
was changed from zero to '2', whereas the actual value |
| 217 |
for ecn to be set is still '1' |
| 218 |
|
| 219 |
* Allow different packet sizes for `--mtu'. |
| 220 |
Suport `-l' option for tracepath wrapper. |
| 221 |
|
| 222 |
* Some code and manual cleanups |
| 223 |
|
| 224 |
|
| 225 |
2010-12-14 Dmitry Butskoy <[email protected]> - 2.0.17 |
| 226 |
|
| 227 |
* Adapt code to make possible the use of Linux capabilities |
| 228 |
(for raw sockets etc.) instead of superuser privileges only. |
| 229 |
|
| 230 |
On modern systems the capabilities can be stored as |
| 231 |
file attributes, ie.: |
| 232 |
|
| 233 |
"setcap cap_net_raw=pe /usr/bin/traceroute" |
| 234 |
|
| 235 |
|
| 236 |
2010-09-13 Dmitry Butskoy <[email protected]> - 2.0.16 |
| 237 |
|
| 238 |
* A little work-around in the build system |
| 239 |
for the new (buggy?) make 3.82 |
| 240 |
|
| 241 |
* Add `--fwmark=num' option for firewall mark (for kernel >= 2.6.25). |
| 242 |
Idea comes from an anonymous SF patch #3042539 |
| 243 |
|
| 244 |
|
| 245 |
2010-07-14 Dmitry Butskoy <[email protected]> - 2.0.15 |
| 246 |
|
| 247 |
* Use string routines more safely (fix SF bug #3029216) |
| 248 |
|
| 249 |
* Provide help for lft wrapper |
| 250 |
|
| 251 |
|
| 252 |
2010-04-21 Dmitry Butskoy <[email protected]> - 2.0.14 |
| 253 |
|
| 254 |
* Fix support for IPv6's flow_labels and tclass. |
| 255 |
Thanks to Peter Bieringer for testing |
| 256 |
|
| 257 |
* Use route header "type 2" instead of deprecated "type 0" |
| 258 |
for `-g' option for IPv6. The default value can be changed |
| 259 |
by specifying a number in the place of the first `-g' address. |
| 260 |
|
| 261 |
|
| 262 |
2009-11-02 Dmitry Butskoy <[email protected]> - 2.0.13 |
| 263 |
|
| 264 |
* Check for first_hop is not zero value ([email protected]) |
| 265 |
|
| 266 |
* Always fill unresolved IP address by its numeric interpretation, |
| 267 |
even if getnameinfo(3) leaves it untouched (as it does for ipv6 |
| 268 |
in some glibc versions, whereas always fills for ipv4) |
| 269 |
|
| 270 |
* Cosmetic changes for man page (Andreas Mohr) |
| 271 |
|
| 272 |
|
| 273 |
2008-09-15 Dmitry Butskoy <[email protected]> - 2.0.12 |
| 274 |
|
| 275 |
* Use common recv_reply() routine for all modules which |
| 276 |
do recvmsg(2) call. Method-specific things go to callbacks. |
| 277 |
|
| 278 |
Pass to init methods pointer to datalen instead of the value. |
| 279 |
|
| 280 |
* Implement ICMP Extension support (rfc4884), `-e' option. |
| 281 |
Parse MPLS info (rfc4950) to be more readable (Kaj Niemi) |
| 282 |
|
| 283 |
* Implement Path MTU Discovery (similar to tracepath(1)), |
| 284 |
with `--mtu' option. Changed mtu is printed once in a form |
| 285 |
of `F=NUM' at the first probe of a hop which requires |
| 286 |
such mtu to be reached. (Actually, the correspond "frag needed" |
| 287 |
icmp message is normally sent by the previous hop). |
| 288 |
|
| 289 |
* Print the number of backward hops when it differs with forward, |
| 290 |
by `--back' option. The backward hops is guessed by a technique |
| 291 |
similar to tracepath(1), there is no reliable way to obtain |
| 292 |
such info though. |
| 293 |
|
| 294 |
* The optional second argument (packet_len) now is the full length |
| 295 |
of the packet, including IP headers. (It is obvious enough due to |
| 296 |
the nature of this feature, and this is the behaviour of the |
| 297 |
original traceroute). Particular trace methods can ignore this |
| 298 |
(fe. tcp), or increase it up to the minimal value (udp, icmp). |
| 299 |
The actual packet's size is alvays reported in the output header. |
| 300 |
|
| 301 |
* Add tracepath(1)/tracepath6(1) shell wrapper. |
| 302 |
|
| 303 |
* Allow DEF_AF to be redefined at cmdline (Teran McKinney) |
| 304 |
|
| 305 |
* Do not check the correctness of `sim_probes' value -- it is |
| 306 |
unneeded at all. This also fixes a bug when a value of sim_probes |
| 307 |
appears to be more than the total number of probes. |
| 308 |
Reported by Milos Malik. |
| 309 |
|
| 310 |
* Allow default UDP method to cross zero port boundary (Milos Malik). |
| 311 |
It is a strange corner case, but traditional traceroute |
| 312 |
behaves exactly so. |
| 313 |
|
| 314 |
|
| 315 |
2008-04-25 Dmitry Butskoy <[email protected]> - 2.0.11 |
| 316 |
|
| 317 |
* Use new pmtudisc value "probe" instead of "do" for `-F' option |
| 318 |
(available since the kernel 2.6.22). |
| 319 |
|
| 320 |
For kernels before 2.6.22, the `-F' (dontfragment) option |
| 321 |
seems completely useless for IPv6 and partially useful |
| 322 |
for IPv4 (when a user can flush routing caches some way). |
| 323 |
|
| 324 |
* Fix installation in build system (Mike Frysinger) |
| 325 |
|
| 326 |
* Don't compute checksum for ipv6 icmp packets ourselves, |
| 327 |
the kernel overwrites it anyway by the proper values. |
| 328 |
|
| 329 |
* Don't use explicit path to traceroute in wrapper scripts |
| 330 |
|
| 331 |
|
| 332 |
2008-04-17 Dmitry Butskoy <[email protected]> - 2.0.10 |
| 333 |
|
| 334 |
* raw_can_connect(): ipv6 connected raw sockets |
| 335 |
receive MSG_ERRQUEUE properly only for kernels >= 2.6.25 |
| 336 |
|
| 337 |
* remove useless "host" parameter for init methods |
| 338 |
|
| 339 |
* add probe_by_seq() and probe_by_sk() routines, |
| 340 |
don't pass whole probes' pointer to recv_probe method |
| 341 |
|
| 342 |
* collect all sends in do_send() routine |
| 343 |
|
| 344 |
* Interpret ENOBUFS errors for send(2) as "can retry later". |
| 345 |
|
| 346 |
Slow devices (like ppp) with small tx_queue_len can reject |
| 347 |
the sending of too many packets simultaneously. To handle this, |
| 348 |
do_send() now returns a negate value in a case of ENOBUFS |
| 349 |
and similar (instead of program exit). The send_probe method |
| 350 |
clears the probe and returns immediately in such cases. |
| 351 |
Then, if there is an amount of time to wait for something, |
| 352 |
the failed probe will be attempted again after that time expired. |
| 353 |
If nothing to wait more, the program is exited. |
| 354 |
|
| 355 |
|
| 356 |
2007-09-26 Dmitry Butskoy <[email protected]> - 2.0.9 |
| 357 |
|
| 358 |
* Complete manual page. |
| 359 |
|
| 360 |
* Edit manual page to sound more English, thanks to Chris Ward |
| 361 |
|
| 362 |
|
| 363 |
2007-09-04 Dmitry Butskoy <[email protected]> - 2.0.8 |
| 364 |
|
| 365 |
* Move all wrappers to special "wrappers/" dir. |
| 366 |
Add lft(8) shell wrapper. |
| 367 |
Add traceproto(8) shell wrapper. |
| 368 |
Add traceroute-nanog(8) shell wrapper. |
| 369 |
|
| 370 |
* Interpret first_hop as number, not index |
| 371 |
|
| 372 |
* Build system is re-worked to match more the modern requirements |
| 373 |
(Thanks to Mike Frysinger for testing). |
| 374 |
|
| 375 |
* Check for kernel version >= 2.6.22.2 in raw_can_connect() |
| 376 |
|
| 377 |
* Add generic "raw" method, "-P protonum" option. |
| 378 |
New "one_per_time" flag for tr_module. |
| 379 |
|
| 380 |
|
| 381 |
2007-07-31 Dmitry Butskoy <[email protected]> - 2.0.7 |
| 382 |
|
| 383 |
* Fix revents checking typo |
| 384 |
|
| 385 |
* Expect normal data reply from udp too. |
| 386 |
|
| 387 |
* Implement udp to port (-U) and udplite (-UL) methods. |
| 388 |
Both available for unprivileged users. |
| 389 |
Add "coverage" option for udplite. |
| 390 |
|
| 391 |
* Allow non-digit service names for `-p' and `--sport' |
| 392 |
|
| 393 |
* Drop period at the end of "SEE ALSO" section, and |
| 394 |
avoid specific distro names in the manual (Mike Frysinger) |
| 395 |
|
| 396 |
* Explicitly mention that this program is licensed |
| 397 |
as "GPL version 2 or any later version". |
| 398 |
(Similar for libsupp subdir: LGPL version 2.1 or any later). |
| 399 |
|
| 400 |
* Always check whether the dest and source port match in |
| 401 |
received packets. Can decrease an amount of (hypothetical) |
| 402 |
garbage received just after the bind() but before connect() |
| 403 |
|
| 404 |
|
| 405 |
2007-07-19 Dmitry Butskoy <[email protected]> - 2.0.6 |
| 406 |
|
| 407 |
* Rename tr_ops to tr_module |
| 408 |
|
| 409 |
* Implement module-specific options (-O opt,...) |
| 410 |
|
| 411 |
* Add TCP specific options (all the tcp header flags, |
| 412 |
ecn, sack, timestamps, window_scaling, mss, sysctl) |
| 413 |
Build tcp probe packet depending on them. |
| 414 |
|
| 415 |
* Add "--sport" option for explicit source port selection. |
| 416 |
Always cause "-N 1" when it is set. |
| 417 |
|
| 418 |
* Add new routine bind_socket(). |
| 419 |
Always (auto)bind sockets in tune_socket(). |
| 420 |
|
| 421 |
* Add tcptraceroute(8) shell wrapper |
| 422 |
|
| 423 |
|
| 424 |
2007-07-16 Dmitry Butskoy <[email protected]> - 2.0.5 |
| 425 |
|
| 426 |
* Use MSG_ERRQUEUE for raw sockets too. |
| 427 |
|
| 428 |
* raw_can_connect () work-around for kernel bug #8747 |
| 429 |
|
| 430 |
* random.c, csum.c: new separate files |
| 431 |
|
| 432 |
* New implementation of tcp method ("-T"), using |
| 433 |
half-open technique. The old implementation module |
| 434 |
renamed to "tcpconn" ("-M tcpconn"). |
| 435 |
|
| 436 |
* Common parse_cmsg() routine |
| 437 |
|
| 438 |
* put ee_info for parse_icmp_res() too, |
| 439 |
handle ICMP6_PACKET_TOO_BIG for IPv6, |
| 440 |
report "!F-num" when "frag needed" (legacy compatibility) |
| 441 |
|
| 442 |
|
| 443 |
2007-07-11 Dmitry Butskoy <[email protected]> - 2.0.4 |
| 444 |
|
| 445 |
* clear includes of unneeded headers |
| 446 |
|
| 447 |
* move poll stuff to separate poll.c |
| 448 |
|
| 449 |
* add module stuff (module.c), options etc. |
| 450 |
Adapt udp/icmp/tcp for this. |
| 451 |
|
| 452 |
* Add common routines use_recverr() and set_ttl() |
| 453 |
|
| 454 |
|
| 455 |
2007-02-28 Dmitry Butskoy <[email protected]> |
| 456 |
|
| 457 |
* fix variable type for getsockname (Mike Frysinger) |
| 458 |
|
| 459 |
|
| 460 |
2007-01-09 Dmitry Butskoy <[email protected]> - 2.0.3 |
| 461 |
|
| 462 |
* version 2.0.3 |
| 463 |
|
| 464 |
* allow option args without separator (add CLIF_MAY_JOIN_ARG flag), |
| 465 |
for compatibility (Benjamin LaHaise) |
| 466 |
|
| 467 |
* no more "tcptraceroute" symlink for rpm packages, because |
| 468 |
it conflicts with the same-name old package anyway (James Ralston) |
| 469 |
|
| 470 |
* fix compilation on glibc < 2.4 (Andy Shevchenko) |
| 471 |
|
| 472 |
|
| 473 |
2006-10-30 Dmitry Butskoy <[email protected]> - 2.0.2 |
| 474 |
|
| 475 |
* version 2.0.2 |
| 476 |
|
| 477 |
* More accurate check_expired() routine. |
| 478 |
|
| 479 |
* Some minor fixes. |
| 480 |
|
| 481 |
* Add NOTES section to manual |
| 482 |
|
| 483 |
|
| 484 |
2006-10-20 Dmitry Butskoy <[email protected]> - 2.0.1 |
| 485 |
|
| 486 |
* version 2.0.1 |
| 487 |
|
| 488 |
* Now ops methods write send_time (as well as recv_time) |
| 489 |
|
| 490 |
* Use SO_TIMESTAMP to obtain msecs precisely |
| 491 |
|
| 492 |
* Complete manual |
| 493 |
|