Commit | Line | Data |
---|---|---|
33a3bb4a AQ |
1 | /* Copyright (C) 2012-2016 B.A.T.M.A.N. contributors: |
2 | * | |
3 | * Edo Monticelli, Antonio Quartulli | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or | |
6 | * modify it under the terms of version 2 of the GNU General Public | |
7 | * License as published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope that it will be useful, but | |
10 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
12 | * General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, see <http://www.gnu.org/licenses/>. | |
16 | */ | |
17 | ||
18 | #include "tp_meter.h" | |
19 | #include "main.h" | |
20 | ||
21 | #include <linux/atomic.h> | |
22 | #include <linux/bug.h> | |
23 | #include <linux/byteorder/generic.h> | |
24 | #include <linux/cache.h> | |
25 | #include <linux/compiler.h> | |
26 | #include <linux/device.h> | |
27 | #include <linux/etherdevice.h> | |
28 | #include <linux/fs.h> | |
29 | #include <linux/if_ether.h> | |
30 | #include <linux/jiffies.h> | |
31 | #include <linux/kernel.h> | |
32 | #include <linux/kref.h> | |
33 | #include <linux/kthread.h> | |
34 | #include <linux/list.h> | |
35 | #include <linux/netdevice.h> | |
36 | #include <linux/param.h> | |
37 | #include <linux/printk.h> | |
38 | #include <linux/random.h> | |
39 | #include <linux/rculist.h> | |
40 | #include <linux/rcupdate.h> | |
41 | #include <linux/sched.h> | |
42 | #include <linux/skbuff.h> | |
43 | #include <linux/slab.h> | |
44 | #include <linux/spinlock.h> | |
45 | #include <linux/stddef.h> | |
46 | #include <linux/string.h> | |
47 | #include <linux/timer.h> | |
48 | #include <linux/wait.h> | |
49 | #include <linux/workqueue.h> | |
50 | #include <uapi/linux/batman_adv.h> | |
51 | ||
52 | #include "hard-interface.h" | |
53 | #include "log.h" | |
54 | #include "netlink.h" | |
55 | #include "originator.h" | |
56 | #include "packet.h" | |
57 | #include "send.h" | |
58 | ||
59 | /** | |
60 | * BATADV_TP_DEF_TEST_LENGTH - Default test length if not specified by the user | |
61 | * in milliseconds | |
62 | */ | |
63 | #define BATADV_TP_DEF_TEST_LENGTH 10000 | |
64 | ||
65 | /** | |
66 | * BATADV_TP_AWND - Advertised window by the receiver (in bytes) | |
67 | */ | |
68 | #define BATADV_TP_AWND 0x20000000 | |
69 | ||
70 | /** | |
71 | * BATADV_TP_RECV_TIMEOUT - Receiver activity timeout. If the receiver does not | |
72 | * get anything for such amount of milliseconds, the connection is killed | |
73 | */ | |
74 | #define BATADV_TP_RECV_TIMEOUT 1000 | |
75 | ||
76 | /** | |
77 | * BATADV_TP_MAX_RTO - Maximum sender timeout. If the sender RTO gets beyond | |
78 | * such amound of milliseconds, the receiver is considered unreachable and the | |
79 | * connection is killed | |
80 | */ | |
81 | #define BATADV_TP_MAX_RTO 30000 | |
82 | ||
83 | /** | |
84 | * BATADV_TP_FIRST_SEQ - First seqno of each session. The number is rather high | |
85 | * in order to immediately trigger a wrap around (test purposes) | |
86 | */ | |
87 | #define BATADV_TP_FIRST_SEQ ((u32)-1 - 2000) | |
88 | ||
89 | /** | |
90 | * BATADV_TP_PLEN - length of the payload (data after the batadv_unicast header) | |
91 | * to simulate | |
92 | */ | |
93 | #define BATADV_TP_PLEN (BATADV_TP_PACKET_LEN - ETH_HLEN - \ | |
94 | sizeof(struct batadv_unicast_packet)) | |
95 | ||
96 | static u8 batadv_tp_prerandom[4096] __read_mostly; | |
97 | ||
98 | /** | |
99 | * batadv_tp_session_cookie - generate session cookie based on session ids | |
100 | * @session: TP session identifier | |
101 | * @icmp_uid: icmp pseudo uid of the tp session | |
102 | * | |
103 | * Return: 32 bit tp_meter session cookie | |
104 | */ | |
105 | static u32 batadv_tp_session_cookie(const u8 session[2], u8 icmp_uid) | |
106 | { | |
107 | u32 cookie; | |
108 | ||
109 | cookie = icmp_uid << 16; | |
110 | cookie |= session[0] << 8; | |
111 | cookie |= session[1]; | |
112 | ||
113 | return cookie; | |
114 | } | |
115 | ||
116 | /** | |
117 | * batadv_tp_cwnd - compute the new cwnd size | |
118 | * @base: base cwnd size value | |
119 | * @increment: the value to add to base to get the new size | |
120 | * @min: minumim cwnd value (usually MSS) | |
121 | * | |
122 | * Return the new cwnd size and ensures it does not exceed the Advertised | |
123 | * Receiver Window size. It is wrap around safe. | |
124 | * For details refer to Section 3.1 of RFC5681 | |
125 | * | |
126 | * Return: new congestion window size in bytes | |
127 | */ | |
128 | static u32 batadv_tp_cwnd(u32 base, u32 increment, u32 min) | |
129 | { | |
130 | u32 new_size = base + increment; | |
131 | ||
132 | /* check for wrap-around */ | |
133 | if (new_size < base) | |
134 | new_size = (u32)ULONG_MAX; | |
135 | ||
136 | new_size = min_t(u32, new_size, BATADV_TP_AWND); | |
137 | ||
138 | return max_t(u32, new_size, min); | |
139 | } | |
140 | ||
141 | /** | |
142 | * batadv_tp_updated_cwnd - update the Congestion Windows | |
143 | * @tp_vars: the private data of the current TP meter session | |
144 | * @mss: maximum segment size of transmission | |
145 | * | |
146 | * 1) if the session is in Slow Start, the CWND has to be increased by 1 | |
147 | * MSS every unique received ACK | |
148 | * 2) if the session is in Congestion Avoidance, the CWND has to be | |
149 | * increased by MSS * MSS / CWND for every unique received ACK | |
150 | */ | |
151 | static void batadv_tp_update_cwnd(struct batadv_tp_vars *tp_vars, u32 mss) | |
152 | { | |
153 | spin_lock_bh(&tp_vars->cwnd_lock); | |
154 | ||
155 | /* slow start... */ | |
156 | if (tp_vars->cwnd <= tp_vars->ss_threshold) { | |
157 | tp_vars->dec_cwnd = 0; | |
158 | tp_vars->cwnd = batadv_tp_cwnd(tp_vars->cwnd, mss, mss); | |
159 | spin_unlock_bh(&tp_vars->cwnd_lock); | |
160 | return; | |
161 | } | |
162 | ||
163 | /* increment CWND at least of 1 (section 3.1 of RFC5681) */ | |
164 | tp_vars->dec_cwnd += max_t(u32, 1U << 3, | |
165 | ((mss * mss) << 6) / (tp_vars->cwnd << 3)); | |
166 | if (tp_vars->dec_cwnd < (mss << 3)) { | |
167 | spin_unlock_bh(&tp_vars->cwnd_lock); | |
168 | return; | |
169 | } | |
170 | ||
171 | tp_vars->cwnd = batadv_tp_cwnd(tp_vars->cwnd, mss, mss); | |
172 | tp_vars->dec_cwnd = 0; | |
173 | ||
174 | spin_unlock_bh(&tp_vars->cwnd_lock); | |
175 | } | |
176 | ||
177 | /** | |
178 | * batadv_tp_update_rto - calculate new retransmission timeout | |
179 | * @tp_vars: the private data of the current TP meter session | |
180 | * @new_rtt: new roundtrip time in msec | |
181 | */ | |
182 | static void batadv_tp_update_rto(struct batadv_tp_vars *tp_vars, | |
183 | u32 new_rtt) | |
184 | { | |
185 | long m = new_rtt; | |
186 | ||
187 | /* RTT update | |
188 | * Details in Section 2.2 and 2.3 of RFC6298 | |
189 | * | |
190 | * It's tricky to understand. Don't lose hair please. | |
191 | * Inspired by tcp_rtt_estimator() tcp_input.c | |
192 | */ | |
193 | if (tp_vars->srtt != 0) { | |
194 | m -= (tp_vars->srtt >> 3); /* m is now error in rtt est */ | |
195 | tp_vars->srtt += m; /* rtt = 7/8 srtt + 1/8 new */ | |
196 | if (m < 0) | |
197 | m = -m; | |
198 | ||
199 | m -= (tp_vars->rttvar >> 2); | |
200 | tp_vars->rttvar += m; /* mdev ~= 3/4 rttvar + 1/4 new */ | |
201 | } else { | |
202 | /* first measure getting in */ | |
203 | tp_vars->srtt = m << 3; /* take the measured time to be srtt */ | |
204 | tp_vars->rttvar = m << 1; /* new_rtt / 2 */ | |
205 | } | |
206 | ||
207 | /* rto = srtt + 4 * rttvar. | |
208 | * rttvar is scaled by 4, therefore doesn't need to be multiplied | |
209 | */ | |
210 | tp_vars->rto = (tp_vars->srtt >> 3) + tp_vars->rttvar; | |
211 | } | |
212 | ||
213 | /** | |
214 | * batadv_tp_batctl_notify - send client status result to client | |
215 | * @reason: reason for tp meter session stop | |
216 | * @dst: destination of tp_meter session | |
217 | * @bat_priv: the bat priv with all the soft interface information | |
218 | * @start_time: start of transmission in jiffies | |
219 | * @total_sent: bytes acked to the receiver | |
220 | * @cookie: cookie of tp_meter session | |
221 | */ | |
222 | static void batadv_tp_batctl_notify(enum batadv_tp_meter_reason reason, | |
223 | const u8 *dst, struct batadv_priv *bat_priv, | |
224 | unsigned long start_time, u64 total_sent, | |
225 | u32 cookie) | |
226 | { | |
227 | u32 test_time; | |
228 | u8 result; | |
229 | u32 total_bytes; | |
230 | ||
231 | if (!batadv_tp_is_error(reason)) { | |
232 | result = BATADV_TP_REASON_COMPLETE; | |
233 | test_time = jiffies_to_msecs(jiffies - start_time); | |
234 | total_bytes = total_sent; | |
235 | } else { | |
236 | result = reason; | |
237 | test_time = 0; | |
238 | total_bytes = 0; | |
239 | } | |
240 | ||
241 | batadv_netlink_tpmeter_notify(bat_priv, dst, result, test_time, | |
242 | total_bytes, cookie); | |
243 | } | |
244 | ||
245 | /** | |
246 | * batadv_tp_batctl_error_notify - send client error result to client | |
247 | * @reason: reason for tp meter session stop | |
248 | * @dst: destination of tp_meter session | |
249 | * @bat_priv: the bat priv with all the soft interface information | |
250 | * @cookie: cookie of tp_meter session | |
251 | */ | |
252 | static void batadv_tp_batctl_error_notify(enum batadv_tp_meter_reason reason, | |
253 | const u8 *dst, | |
254 | struct batadv_priv *bat_priv, | |
255 | u32 cookie) | |
256 | { | |
257 | batadv_tp_batctl_notify(reason, dst, bat_priv, 0, 0, cookie); | |
258 | } | |
259 | ||
260 | /** | |
261 | * batadv_tp_list_find - find a tp_vars object in the global list | |
262 | * @bat_priv: the bat priv with all the soft interface information | |
263 | * @dst: the other endpoint MAC address to look for | |
264 | * | |
265 | * Look for a tp_vars object matching dst as end_point and return it after | |
266 | * having incremented the refcounter. Return NULL is not found | |
267 | * | |
268 | * Return: matching tp_vars or NULL when no tp_vars with @dst was found | |
269 | */ | |
270 | static struct batadv_tp_vars *batadv_tp_list_find(struct batadv_priv *bat_priv, | |
271 | const u8 *dst) | |
272 | { | |
273 | struct batadv_tp_vars *pos, *tp_vars = NULL; | |
274 | ||
275 | rcu_read_lock(); | |
276 | hlist_for_each_entry_rcu(pos, &bat_priv->tp_list, list) { | |
277 | if (!batadv_compare_eth(pos->other_end, dst)) | |
278 | continue; | |
279 | ||
280 | /* most of the time this function is invoked during the normal | |
281 | * process..it makes sens to pay more when the session is | |
282 | * finished and to speed the process up during the measurement | |
283 | */ | |
284 | if (unlikely(!kref_get_unless_zero(&pos->refcount))) | |
285 | continue; | |
286 | ||
287 | tp_vars = pos; | |
288 | break; | |
289 | } | |
290 | rcu_read_unlock(); | |
291 | ||
292 | return tp_vars; | |
293 | } | |
294 | ||
295 | /** | |
296 | * batadv_tp_list_find_session - find tp_vars session object in the global list | |
297 | * @bat_priv: the bat priv with all the soft interface information | |
298 | * @dst: the other endpoint MAC address to look for | |
299 | * @session: session identifier | |
300 | * | |
301 | * Look for a tp_vars object matching dst as end_point, session as tp meter | |
302 | * session and return it after having incremented the refcounter. Return NULL | |
303 | * is not found | |
304 | * | |
305 | * Return: matching tp_vars or NULL when no tp_vars was found | |
306 | */ | |
307 | static struct batadv_tp_vars * | |
308 | batadv_tp_list_find_session(struct batadv_priv *bat_priv, const u8 *dst, | |
309 | const u8 *session) | |
310 | { | |
311 | struct batadv_tp_vars *pos, *tp_vars = NULL; | |
312 | ||
313 | rcu_read_lock(); | |
314 | hlist_for_each_entry_rcu(pos, &bat_priv->tp_list, list) { | |
315 | if (!batadv_compare_eth(pos->other_end, dst)) | |
316 | continue; | |
317 | ||
318 | if (memcmp(pos->session, session, sizeof(pos->session)) != 0) | |
319 | continue; | |
320 | ||
321 | /* most of the time this function is invoked during the normal | |
322 | * process..it makes sense to pay more when the session is | |
323 | * finished and to speed the process up during the measurement | |
324 | */ | |
325 | if (unlikely(!kref_get_unless_zero(&pos->refcount))) | |
326 | continue; | |
327 | ||
328 | tp_vars = pos; | |
329 | break; | |
330 | } | |
331 | rcu_read_unlock(); | |
332 | ||
333 | return tp_vars; | |
334 | } | |
335 | ||
336 | /** | |
337 | * batadv_tp_vars_release - release batadv_tp_vars from lists and queue for | |
338 | * free after rcu grace period | |
339 | * @ref: kref pointer of the batadv_tp_vars | |
340 | */ | |
341 | static void batadv_tp_vars_release(struct kref *ref) | |
342 | { | |
343 | struct batadv_tp_vars *tp_vars; | |
344 | struct batadv_tp_unacked *un, *safe; | |
345 | ||
346 | tp_vars = container_of(ref, struct batadv_tp_vars, refcount); | |
347 | ||
348 | /* lock should not be needed because this object is now out of any | |
349 | * context! | |
350 | */ | |
351 | spin_lock_bh(&tp_vars->unacked_lock); | |
352 | list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) { | |
353 | list_del(&un->list); | |
354 | kfree(un); | |
355 | } | |
356 | spin_unlock_bh(&tp_vars->unacked_lock); | |
357 | ||
358 | kfree_rcu(tp_vars, rcu); | |
359 | } | |
360 | ||
361 | /** | |
362 | * batadv_tp_vars_put - decrement the batadv_tp_vars refcounter and possibly | |
363 | * release it | |
364 | * @tp_vars: the private data of the current TP meter session to be free'd | |
365 | */ | |
366 | static void batadv_tp_vars_put(struct batadv_tp_vars *tp_vars) | |
367 | { | |
368 | kref_put(&tp_vars->refcount, batadv_tp_vars_release); | |
369 | } | |
370 | ||
371 | /** | |
372 | * batadv_tp_sender_cleanup - cleanup sender data and drop and timer | |
373 | * @bat_priv: the bat priv with all the soft interface information | |
374 | * @tp_vars: the private data of the current TP meter session to cleanup | |
375 | */ | |
376 | static void batadv_tp_sender_cleanup(struct batadv_priv *bat_priv, | |
377 | struct batadv_tp_vars *tp_vars) | |
378 | { | |
379 | cancel_delayed_work(&tp_vars->finish_work); | |
380 | ||
381 | spin_lock_bh(&tp_vars->bat_priv->tp_list_lock); | |
382 | hlist_del_rcu(&tp_vars->list); | |
383 | spin_unlock_bh(&tp_vars->bat_priv->tp_list_lock); | |
384 | ||
385 | /* drop list reference */ | |
386 | batadv_tp_vars_put(tp_vars); | |
387 | ||
388 | atomic_dec(&tp_vars->bat_priv->tp_num); | |
389 | ||
390 | /* kill the timer and remove its reference */ | |
391 | del_timer_sync(&tp_vars->timer); | |
392 | /* the worker might have rearmed itself therefore we kill it again. Note | |
393 | * that if the worker should run again before invoking the following | |
394 | * del_timer(), it would not re-arm itself once again because the status | |
395 | * is OFF now | |
396 | */ | |
397 | del_timer(&tp_vars->timer); | |
398 | batadv_tp_vars_put(tp_vars); | |
399 | } | |
400 | ||
401 | /** | |
402 | * batadv_tp_sender_end - print info about ended session and inform client | |
403 | * @bat_priv: the bat priv with all the soft interface information | |
404 | * @tp_vars: the private data of the current TP meter session | |
405 | */ | |
406 | static void batadv_tp_sender_end(struct batadv_priv *bat_priv, | |
407 | struct batadv_tp_vars *tp_vars) | |
408 | { | |
409 | u32 session_cookie; | |
410 | ||
411 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
412 | "Test towards %pM finished..shutting down (reason=%d)\n", | |
413 | tp_vars->other_end, tp_vars->reason); | |
414 | ||
415 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
416 | "Last timing stats: SRTT=%ums RTTVAR=%ums RTO=%ums\n", | |
417 | tp_vars->srtt >> 3, tp_vars->rttvar >> 2, tp_vars->rto); | |
418 | ||
419 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
420 | "Final values: cwnd=%u ss_threshold=%u\n", | |
421 | tp_vars->cwnd, tp_vars->ss_threshold); | |
422 | ||
423 | session_cookie = batadv_tp_session_cookie(tp_vars->session, | |
424 | tp_vars->icmp_uid); | |
425 | ||
426 | batadv_tp_batctl_notify(tp_vars->reason, | |
427 | tp_vars->other_end, | |
428 | bat_priv, | |
429 | tp_vars->start_time, | |
430 | atomic64_read(&tp_vars->tot_sent), | |
431 | session_cookie); | |
432 | } | |
433 | ||
434 | /** | |
435 | * batadv_tp_sender_shutdown - let sender thread/timer stop gracefully | |
436 | * @tp_vars: the private data of the current TP meter session | |
437 | * @reason: reason for tp meter session stop | |
438 | */ | |
439 | static void batadv_tp_sender_shutdown(struct batadv_tp_vars *tp_vars, | |
440 | enum batadv_tp_meter_reason reason) | |
441 | { | |
442 | if (!atomic_dec_and_test(&tp_vars->sending)) | |
443 | return; | |
444 | ||
445 | tp_vars->reason = reason; | |
446 | } | |
447 | ||
448 | /** | |
449 | * batadv_tp_sender_finish - stop sender session after test_length was reached | |
450 | * @work: delayed work reference of the related tp_vars | |
451 | */ | |
452 | static void batadv_tp_sender_finish(struct work_struct *work) | |
453 | { | |
454 | struct delayed_work *delayed_work; | |
455 | struct batadv_tp_vars *tp_vars; | |
456 | ||
457 | delayed_work = to_delayed_work(work); | |
458 | tp_vars = container_of(delayed_work, struct batadv_tp_vars, | |
459 | finish_work); | |
460 | ||
461 | batadv_tp_sender_shutdown(tp_vars, BATADV_TP_REASON_COMPLETE); | |
462 | } | |
463 | ||
464 | /** | |
465 | * batadv_tp_reset_sender_timer - reschedule the sender timer | |
466 | * @tp_vars: the private TP meter data for this session | |
467 | * | |
468 | * Reschedule the timer using tp_vars->rto as delay | |
469 | */ | |
470 | static void batadv_tp_reset_sender_timer(struct batadv_tp_vars *tp_vars) | |
471 | { | |
472 | /* most of the time this function is invoked while normal packet | |
473 | * reception... | |
474 | */ | |
475 | if (unlikely(atomic_read(&tp_vars->sending) == 0)) | |
476 | /* timer ref will be dropped in batadv_tp_sender_cleanup */ | |
477 | return; | |
478 | ||
479 | mod_timer(&tp_vars->timer, jiffies + msecs_to_jiffies(tp_vars->rto)); | |
480 | } | |
481 | ||
482 | /** | |
483 | * batadv_tp_sender_timeout - timer that fires in case of packet loss | |
484 | * @arg: address of the related tp_vars | |
485 | * | |
486 | * If fired it means that there was packet loss. | |
487 | * Switch to Slow Start, set the ss_threshold to half of the current cwnd and | |
488 | * reset the cwnd to 3*MSS | |
489 | */ | |
490 | static void batadv_tp_sender_timeout(unsigned long arg) | |
491 | { | |
492 | struct batadv_tp_vars *tp_vars = (struct batadv_tp_vars *)arg; | |
493 | struct batadv_priv *bat_priv = tp_vars->bat_priv; | |
494 | ||
495 | if (atomic_read(&tp_vars->sending) == 0) | |
496 | return; | |
497 | ||
498 | /* if the user waited long enough...shutdown the test */ | |
499 | if (unlikely(tp_vars->rto >= BATADV_TP_MAX_RTO)) { | |
500 | batadv_tp_sender_shutdown(tp_vars, | |
501 | BATADV_TP_REASON_DST_UNREACHABLE); | |
502 | return; | |
503 | } | |
504 | ||
505 | /* RTO exponential backoff | |
506 | * Details in Section 5.5 of RFC6298 | |
507 | */ | |
508 | tp_vars->rto <<= 1; | |
509 | ||
510 | spin_lock_bh(&tp_vars->cwnd_lock); | |
511 | ||
512 | tp_vars->ss_threshold = tp_vars->cwnd >> 1; | |
513 | if (tp_vars->ss_threshold < BATADV_TP_PLEN * 2) | |
514 | tp_vars->ss_threshold = BATADV_TP_PLEN * 2; | |
515 | ||
516 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
517 | "Meter: RTO fired during test towards %pM! cwnd=%u new ss_thr=%u, resetting last_sent to %u\n", | |
518 | tp_vars->other_end, tp_vars->cwnd, tp_vars->ss_threshold, | |
519 | atomic_read(&tp_vars->last_acked)); | |
520 | ||
521 | tp_vars->cwnd = BATADV_TP_PLEN * 3; | |
522 | ||
523 | spin_unlock_bh(&tp_vars->cwnd_lock); | |
524 | ||
525 | /* resend the non-ACKed packets.. */ | |
526 | tp_vars->last_sent = atomic_read(&tp_vars->last_acked); | |
527 | wake_up(&tp_vars->more_bytes); | |
528 | ||
529 | batadv_tp_reset_sender_timer(tp_vars); | |
530 | } | |
531 | ||
532 | /** | |
533 | * batadv_tp_fill_prerandom - Fill buffer with prefetched random bytes | |
534 | * @tp_vars: the private TP meter data for this session | |
535 | * @buf: Buffer to fill with bytes | |
536 | * @nbytes: amount of pseudorandom bytes | |
537 | */ | |
538 | static void batadv_tp_fill_prerandom(struct batadv_tp_vars *tp_vars, | |
539 | u8 *buf, size_t nbytes) | |
540 | { | |
541 | u32 local_offset; | |
542 | size_t bytes_inbuf; | |
543 | size_t to_copy; | |
544 | size_t pos = 0; | |
545 | ||
546 | spin_lock_bh(&tp_vars->prerandom_lock); | |
547 | local_offset = tp_vars->prerandom_offset; | |
548 | tp_vars->prerandom_offset += nbytes; | |
549 | tp_vars->prerandom_offset %= sizeof(batadv_tp_prerandom); | |
550 | spin_unlock_bh(&tp_vars->prerandom_lock); | |
551 | ||
552 | while (nbytes) { | |
553 | local_offset %= sizeof(batadv_tp_prerandom); | |
554 | bytes_inbuf = sizeof(batadv_tp_prerandom) - local_offset; | |
555 | to_copy = min(nbytes, bytes_inbuf); | |
556 | ||
557 | memcpy(&buf[pos], &batadv_tp_prerandom[local_offset], to_copy); | |
558 | pos += to_copy; | |
559 | nbytes -= to_copy; | |
560 | local_offset = 0; | |
561 | } | |
562 | } | |
563 | ||
564 | /** | |
565 | * batadv_tp_send_msg - send a single message | |
566 | * @tp_vars: the private TP meter data for this session | |
567 | * @src: source mac address | |
568 | * @orig_node: the originator of the destination | |
569 | * @seqno: sequence number of this packet | |
570 | * @len: length of the entire packet | |
571 | * @session: session identifier | |
572 | * @uid: local ICMP "socket" index | |
573 | * @timestamp: timestamp in jiffies which is replied in ack | |
574 | * | |
575 | * Create and send a single TP Meter message. | |
576 | * | |
577 | * Return: 0 on success, BATADV_TP_REASON_DST_UNREACHABLE if the destination is | |
578 | * not reachable, BATADV_TP_REASON_MEMORY_ERROR if the packet couldn't be | |
579 | * allocated | |
580 | */ | |
581 | static int batadv_tp_send_msg(struct batadv_tp_vars *tp_vars, const u8 *src, | |
582 | struct batadv_orig_node *orig_node, | |
583 | u32 seqno, size_t len, const u8 *session, | |
584 | int uid, u32 timestamp) | |
585 | { | |
586 | struct batadv_icmp_tp_packet *icmp; | |
587 | struct sk_buff *skb; | |
588 | int r; | |
589 | u8 *data; | |
590 | size_t data_len; | |
591 | ||
592 | skb = netdev_alloc_skb_ip_align(NULL, len + ETH_HLEN); | |
593 | if (unlikely(!skb)) | |
594 | return BATADV_TP_REASON_MEMORY_ERROR; | |
595 | ||
596 | skb_reserve(skb, ETH_HLEN); | |
597 | icmp = (struct batadv_icmp_tp_packet *)skb_put(skb, sizeof(*icmp)); | |
598 | ||
599 | /* fill the icmp header */ | |
600 | ether_addr_copy(icmp->dst, orig_node->orig); | |
601 | ether_addr_copy(icmp->orig, src); | |
602 | icmp->version = BATADV_COMPAT_VERSION; | |
603 | icmp->packet_type = BATADV_ICMP; | |
604 | icmp->ttl = BATADV_TTL; | |
605 | icmp->msg_type = BATADV_TP; | |
606 | icmp->uid = uid; | |
607 | ||
608 | icmp->subtype = BATADV_TP_MSG; | |
609 | memcpy(icmp->session, session, sizeof(icmp->session)); | |
610 | icmp->seqno = htonl(seqno); | |
611 | icmp->timestamp = htonl(timestamp); | |
612 | ||
613 | data_len = len - sizeof(*icmp); | |
614 | data = (u8 *)skb_put(skb, data_len); | |
615 | batadv_tp_fill_prerandom(tp_vars, data, data_len); | |
616 | ||
617 | r = batadv_send_skb_to_orig(skb, orig_node, NULL); | |
618 | if (r == -1) | |
619 | kfree_skb(skb); | |
620 | ||
621 | if (r == NET_XMIT_SUCCESS) | |
622 | return 0; | |
623 | ||
624 | return BATADV_TP_REASON_CANT_SEND; | |
625 | } | |
626 | ||
627 | /** | |
628 | * batadv_tp_recv_ack - ACK receiving function | |
629 | * @bat_priv: the bat priv with all the soft interface information | |
630 | * @skb: the buffer containing the received packet | |
631 | * | |
632 | * Process a received TP ACK packet | |
633 | */ | |
634 | static void batadv_tp_recv_ack(struct batadv_priv *bat_priv, | |
635 | const struct sk_buff *skb) | |
636 | { | |
637 | struct batadv_hard_iface *primary_if = NULL; | |
638 | struct batadv_orig_node *orig_node = NULL; | |
639 | const struct batadv_icmp_tp_packet *icmp; | |
640 | struct batadv_tp_vars *tp_vars; | |
641 | size_t packet_len, mss; | |
642 | u32 rtt, recv_ack, cwnd; | |
643 | unsigned char *dev_addr; | |
644 | ||
645 | packet_len = BATADV_TP_PLEN; | |
646 | mss = BATADV_TP_PLEN; | |
647 | packet_len += sizeof(struct batadv_unicast_packet); | |
648 | ||
649 | icmp = (struct batadv_icmp_tp_packet *)skb->data; | |
650 | ||
651 | /* find the tp_vars */ | |
652 | tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig, | |
653 | icmp->session); | |
654 | if (unlikely(!tp_vars)) | |
655 | return; | |
656 | ||
657 | if (unlikely(atomic_read(&tp_vars->sending) == 0)) | |
658 | goto out; | |
659 | ||
660 | /* old ACK? silently drop it.. */ | |
661 | if (batadv_seq_before(ntohl(icmp->seqno), | |
662 | (u32)atomic_read(&tp_vars->last_acked))) | |
663 | goto out; | |
664 | ||
665 | primary_if = batadv_primary_if_get_selected(bat_priv); | |
666 | if (unlikely(!primary_if)) | |
667 | goto out; | |
668 | ||
669 | orig_node = batadv_orig_hash_find(bat_priv, icmp->orig); | |
670 | if (unlikely(!orig_node)) | |
671 | goto out; | |
672 | ||
673 | /* update RTO with the new sampled RTT, if any */ | |
674 | rtt = jiffies_to_msecs(jiffies) - ntohl(icmp->timestamp); | |
675 | if (icmp->timestamp && rtt) | |
676 | batadv_tp_update_rto(tp_vars, rtt); | |
677 | ||
678 | /* ACK for new data... reset the timer */ | |
679 | batadv_tp_reset_sender_timer(tp_vars); | |
680 | ||
681 | recv_ack = ntohl(icmp->seqno); | |
682 | ||
683 | /* check if this ACK is a duplicate */ | |
684 | if (atomic_read(&tp_vars->last_acked) == recv_ack) { | |
685 | atomic_inc(&tp_vars->dup_acks); | |
686 | if (atomic_read(&tp_vars->dup_acks) != 3) | |
687 | goto out; | |
688 | ||
689 | if (recv_ack >= tp_vars->recover) | |
690 | goto out; | |
691 | ||
692 | /* if this is the third duplicate ACK do Fast Retransmit */ | |
693 | batadv_tp_send_msg(tp_vars, primary_if->net_dev->dev_addr, | |
694 | orig_node, recv_ack, packet_len, | |
695 | icmp->session, icmp->uid, | |
696 | jiffies_to_msecs(jiffies)); | |
697 | ||
698 | spin_lock_bh(&tp_vars->cwnd_lock); | |
699 | ||
700 | /* Fast Recovery */ | |
701 | tp_vars->fast_recovery = true; | |
702 | /* Set recover to the last outstanding seqno when Fast Recovery | |
703 | * is entered. RFC6582, Section 3.2, step 1 | |
704 | */ | |
705 | tp_vars->recover = tp_vars->last_sent; | |
706 | tp_vars->ss_threshold = tp_vars->cwnd >> 1; | |
707 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
708 | "Meter: Fast Recovery, (cur cwnd=%u) ss_thr=%u last_sent=%u recv_ack=%u\n", | |
709 | tp_vars->cwnd, tp_vars->ss_threshold, | |
710 | tp_vars->last_sent, recv_ack); | |
711 | tp_vars->cwnd = batadv_tp_cwnd(tp_vars->ss_threshold, 3 * mss, | |
712 | mss); | |
713 | tp_vars->dec_cwnd = 0; | |
714 | tp_vars->last_sent = recv_ack; | |
715 | ||
716 | spin_unlock_bh(&tp_vars->cwnd_lock); | |
717 | } else { | |
718 | /* count the acked data */ | |
719 | atomic64_add(recv_ack - atomic_read(&tp_vars->last_acked), | |
720 | &tp_vars->tot_sent); | |
721 | /* reset the duplicate ACKs counter */ | |
722 | atomic_set(&tp_vars->dup_acks, 0); | |
723 | ||
724 | if (tp_vars->fast_recovery) { | |
725 | /* partial ACK */ | |
726 | if (batadv_seq_before(recv_ack, tp_vars->recover)) { | |
727 | /* this is another hole in the window. React | |
728 | * immediately as specified by NewReno (see | |
729 | * Section 3.2 of RFC6582 for details) | |
730 | */ | |
731 | dev_addr = primary_if->net_dev->dev_addr; | |
732 | batadv_tp_send_msg(tp_vars, dev_addr, | |
733 | orig_node, recv_ack, | |
734 | packet_len, icmp->session, | |
735 | icmp->uid, | |
736 | jiffies_to_msecs(jiffies)); | |
737 | tp_vars->cwnd = batadv_tp_cwnd(tp_vars->cwnd, | |
738 | mss, mss); | |
739 | } else { | |
740 | tp_vars->fast_recovery = false; | |
741 | /* set cwnd to the value of ss_threshold at the | |
742 | * moment that Fast Recovery was entered. | |
743 | * RFC6582, Section 3.2, step 3 | |
744 | */ | |
745 | cwnd = batadv_tp_cwnd(tp_vars->ss_threshold, 0, | |
746 | mss); | |
747 | tp_vars->cwnd = cwnd; | |
748 | } | |
749 | goto move_twnd; | |
750 | } | |
751 | ||
752 | if (recv_ack - atomic_read(&tp_vars->last_acked) >= mss) | |
753 | batadv_tp_update_cwnd(tp_vars, mss); | |
754 | move_twnd: | |
755 | /* move the Transmit Window */ | |
756 | atomic_set(&tp_vars->last_acked, recv_ack); | |
757 | } | |
758 | ||
759 | wake_up(&tp_vars->more_bytes); | |
760 | out: | |
761 | if (likely(primary_if)) | |
762 | batadv_hardif_put(primary_if); | |
763 | if (likely(orig_node)) | |
764 | batadv_orig_node_put(orig_node); | |
765 | if (likely(tp_vars)) | |
766 | batadv_tp_vars_put(tp_vars); | |
767 | } | |
768 | ||
769 | /** | |
770 | * batadv_tp_avail - check if congestion window is not full | |
771 | * @tp_vars: the private data of the current TP meter session | |
772 | * @payload_len: size of the payload of a single message | |
773 | * | |
774 | * Return: true when congestion window is not full, false otherwise | |
775 | */ | |
776 | static bool batadv_tp_avail(struct batadv_tp_vars *tp_vars, | |
777 | size_t payload_len) | |
778 | { | |
779 | u32 win_left, win_limit; | |
780 | ||
781 | win_limit = atomic_read(&tp_vars->last_acked) + tp_vars->cwnd; | |
782 | win_left = win_limit - tp_vars->last_sent; | |
783 | ||
784 | return win_left >= payload_len; | |
785 | } | |
786 | ||
787 | /** | |
788 | * batadv_tp_wait_available - wait until congestion window becomes free or | |
789 | * timeout is reached | |
790 | * @tp_vars: the private data of the current TP meter session | |
791 | * @plen: size of the payload of a single message | |
792 | * | |
793 | * Return: 0 if the condition evaluated to false after the timeout elapsed, | |
794 | * 1 if the condition evaluated to true after the timeout elapsed, the | |
795 | * remaining jiffies (at least 1) if the condition evaluated to true before | |
796 | * the timeout elapsed, or -ERESTARTSYS if it was interrupted by a signal. | |
797 | */ | |
798 | static int batadv_tp_wait_available(struct batadv_tp_vars *tp_vars, size_t plen) | |
799 | { | |
800 | int ret; | |
801 | ||
802 | ret = wait_event_interruptible_timeout(tp_vars->more_bytes, | |
803 | batadv_tp_avail(tp_vars, plen), | |
804 | HZ / 10); | |
805 | ||
806 | return ret; | |
807 | } | |
808 | ||
809 | /** | |
810 | * batadv_tp_send - main sending thread of a tp meter session | |
811 | * @arg: address of the related tp_vars | |
812 | * | |
813 | * Return: nothing, this function never returns | |
814 | */ | |
815 | static int batadv_tp_send(void *arg) | |
816 | { | |
817 | struct batadv_tp_vars *tp_vars = arg; | |
818 | struct batadv_priv *bat_priv = tp_vars->bat_priv; | |
819 | struct batadv_hard_iface *primary_if = NULL; | |
820 | struct batadv_orig_node *orig_node = NULL; | |
821 | size_t payload_len, packet_len; | |
822 | int err = 0; | |
823 | ||
824 | if (unlikely(tp_vars->role != BATADV_TP_SENDER)) { | |
825 | err = BATADV_TP_REASON_DST_UNREACHABLE; | |
826 | tp_vars->reason = err; | |
827 | goto out; | |
828 | } | |
829 | ||
830 | orig_node = batadv_orig_hash_find(bat_priv, tp_vars->other_end); | |
831 | if (unlikely(!orig_node)) { | |
832 | err = BATADV_TP_REASON_DST_UNREACHABLE; | |
833 | tp_vars->reason = err; | |
834 | goto out; | |
835 | } | |
836 | ||
837 | primary_if = batadv_primary_if_get_selected(bat_priv); | |
838 | if (unlikely(!primary_if)) { | |
839 | err = BATADV_TP_REASON_DST_UNREACHABLE; | |
840 | goto out; | |
841 | } | |
842 | ||
843 | /* assume that all the hard_interfaces have a correctly | |
844 | * configured MTU, so use the soft_iface MTU as MSS. | |
845 | * This might not be true and in that case the fragmentation | |
846 | * should be used. | |
847 | * Now, try to send the packet as it is | |
848 | */ | |
849 | payload_len = BATADV_TP_PLEN; | |
850 | BUILD_BUG_ON(sizeof(struct batadv_icmp_tp_packet) > BATADV_TP_PLEN); | |
851 | ||
852 | batadv_tp_reset_sender_timer(tp_vars); | |
853 | ||
854 | /* queue the worker in charge of terminating the test */ | |
855 | queue_delayed_work(batadv_event_workqueue, &tp_vars->finish_work, | |
856 | msecs_to_jiffies(tp_vars->test_length)); | |
857 | ||
858 | while (atomic_read(&tp_vars->sending) != 0) { | |
859 | if (unlikely(!batadv_tp_avail(tp_vars, payload_len))) { | |
860 | batadv_tp_wait_available(tp_vars, payload_len); | |
861 | continue; | |
862 | } | |
863 | ||
864 | /* to emulate normal unicast traffic, add to the payload len | |
865 | * the size of the unicast header | |
866 | */ | |
867 | packet_len = payload_len + sizeof(struct batadv_unicast_packet); | |
868 | ||
869 | err = batadv_tp_send_msg(tp_vars, primary_if->net_dev->dev_addr, | |
870 | orig_node, tp_vars->last_sent, | |
871 | packet_len, | |
872 | tp_vars->session, tp_vars->icmp_uid, | |
873 | jiffies_to_msecs(jiffies)); | |
874 | ||
875 | /* something went wrong during the preparation/transmission */ | |
876 | if (unlikely(err && err != BATADV_TP_REASON_CANT_SEND)) { | |
877 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
878 | "Meter: batadv_tp_send() cannot send packets (%d)\n", | |
879 | err); | |
880 | /* ensure nobody else tries to stop the thread now */ | |
881 | if (atomic_dec_and_test(&tp_vars->sending)) | |
882 | tp_vars->reason = err; | |
883 | break; | |
884 | } | |
885 | ||
886 | /* right-shift the TWND */ | |
887 | if (!err) | |
888 | tp_vars->last_sent += payload_len; | |
889 | ||
890 | cond_resched(); | |
891 | } | |
892 | ||
893 | out: | |
894 | if (likely(primary_if)) | |
895 | batadv_hardif_put(primary_if); | |
896 | if (likely(orig_node)) | |
897 | batadv_orig_node_put(orig_node); | |
898 | ||
899 | batadv_tp_sender_end(bat_priv, tp_vars); | |
900 | batadv_tp_sender_cleanup(bat_priv, tp_vars); | |
901 | ||
902 | batadv_tp_vars_put(tp_vars); | |
903 | ||
904 | do_exit(0); | |
905 | } | |
906 | ||
907 | /** | |
908 | * batadv_tp_start_kthread - start new thread which manages the tp meter sender | |
909 | * @tp_vars: the private data of the current TP meter session | |
910 | */ | |
911 | static void batadv_tp_start_kthread(struct batadv_tp_vars *tp_vars) | |
912 | { | |
913 | struct task_struct *kthread; | |
914 | struct batadv_priv *bat_priv = tp_vars->bat_priv; | |
915 | u32 session_cookie; | |
916 | ||
917 | kref_get(&tp_vars->refcount); | |
918 | kthread = kthread_create(batadv_tp_send, tp_vars, "kbatadv_tp_meter"); | |
919 | if (IS_ERR(kthread)) { | |
920 | session_cookie = batadv_tp_session_cookie(tp_vars->session, | |
921 | tp_vars->icmp_uid); | |
922 | pr_err("batadv: cannot create tp meter kthread\n"); | |
923 | batadv_tp_batctl_error_notify(BATADV_TP_REASON_MEMORY_ERROR, | |
924 | tp_vars->other_end, | |
925 | bat_priv, session_cookie); | |
926 | ||
927 | /* drop reserved reference for kthread */ | |
928 | batadv_tp_vars_put(tp_vars); | |
929 | ||
930 | /* cleanup of failed tp meter variables */ | |
931 | batadv_tp_sender_cleanup(bat_priv, tp_vars); | |
932 | return; | |
933 | } | |
934 | ||
935 | wake_up_process(kthread); | |
936 | } | |
937 | ||
938 | /** | |
939 | * batadv_tp_start - start a new tp meter session | |
940 | * @bat_priv: the bat priv with all the soft interface information | |
941 | * @dst: the receiver MAC address | |
942 | * @test_length: test length in milliseconds | |
943 | * @cookie: session cookie | |
944 | */ | |
945 | void batadv_tp_start(struct batadv_priv *bat_priv, const u8 *dst, | |
946 | u32 test_length, u32 *cookie) | |
947 | { | |
948 | struct batadv_tp_vars *tp_vars; | |
949 | u8 session_id[2]; | |
950 | u8 icmp_uid; | |
951 | u32 session_cookie; | |
952 | ||
953 | get_random_bytes(session_id, sizeof(session_id)); | |
954 | get_random_bytes(&icmp_uid, 1); | |
955 | session_cookie = batadv_tp_session_cookie(session_id, icmp_uid); | |
956 | *cookie = session_cookie; | |
957 | ||
958 | /* look for an already existing test towards this node */ | |
959 | spin_lock_bh(&bat_priv->tp_list_lock); | |
960 | tp_vars = batadv_tp_list_find(bat_priv, dst); | |
961 | if (tp_vars) { | |
962 | spin_unlock_bh(&bat_priv->tp_list_lock); | |
963 | batadv_tp_vars_put(tp_vars); | |
964 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
965 | "Meter: test to or from the same node already ongoing, aborting\n"); | |
966 | batadv_tp_batctl_error_notify(BATADV_TP_REASON_ALREADY_ONGOING, | |
967 | dst, bat_priv, session_cookie); | |
968 | return; | |
969 | } | |
970 | ||
971 | if (!atomic_add_unless(&bat_priv->tp_num, 1, BATADV_TP_MAX_NUM)) { | |
972 | spin_unlock_bh(&bat_priv->tp_list_lock); | |
973 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
974 | "Meter: too many ongoing sessions, aborting (SEND)\n"); | |
975 | batadv_tp_batctl_error_notify(BATADV_TP_REASON_TOO_MANY, dst, | |
976 | bat_priv, session_cookie); | |
977 | return; | |
978 | } | |
979 | ||
980 | tp_vars = kmalloc(sizeof(*tp_vars), GFP_ATOMIC); | |
981 | if (!tp_vars) { | |
982 | spin_unlock_bh(&bat_priv->tp_list_lock); | |
983 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
984 | "Meter: batadv_tp_start cannot allocate list elements\n"); | |
985 | batadv_tp_batctl_error_notify(BATADV_TP_REASON_MEMORY_ERROR, | |
986 | dst, bat_priv, session_cookie); | |
987 | return; | |
988 | } | |
989 | ||
990 | /* initialize tp_vars */ | |
991 | ether_addr_copy(tp_vars->other_end, dst); | |
992 | kref_init(&tp_vars->refcount); | |
993 | tp_vars->role = BATADV_TP_SENDER; | |
994 | atomic_set(&tp_vars->sending, 1); | |
995 | memcpy(tp_vars->session, session_id, sizeof(session_id)); | |
996 | tp_vars->icmp_uid = icmp_uid; | |
997 | ||
998 | tp_vars->last_sent = BATADV_TP_FIRST_SEQ; | |
999 | atomic_set(&tp_vars->last_acked, BATADV_TP_FIRST_SEQ); | |
1000 | tp_vars->fast_recovery = false; | |
1001 | tp_vars->recover = BATADV_TP_FIRST_SEQ; | |
1002 | ||
1003 | /* initialise the CWND to 3*MSS (Section 3.1 in RFC5681). | |
1004 | * For batman-adv the MSS is the size of the payload received by the | |
1005 | * soft_interface, hence its MTU | |
1006 | */ | |
1007 | tp_vars->cwnd = BATADV_TP_PLEN * 3; | |
1008 | /* at the beginning initialise the SS threshold to the biggest possible | |
1009 | * window size, hence the AWND size | |
1010 | */ | |
1011 | tp_vars->ss_threshold = BATADV_TP_AWND; | |
1012 | ||
1013 | /* RTO initial value is 3 seconds. | |
1014 | * Details in Section 2.1 of RFC6298 | |
1015 | */ | |
1016 | tp_vars->rto = 1000; | |
1017 | tp_vars->srtt = 0; | |
1018 | tp_vars->rttvar = 0; | |
1019 | ||
1020 | atomic64_set(&tp_vars->tot_sent, 0); | |
1021 | ||
1022 | kref_get(&tp_vars->refcount); | |
1023 | setup_timer(&tp_vars->timer, batadv_tp_sender_timeout, | |
1024 | (unsigned long)tp_vars); | |
1025 | ||
1026 | tp_vars->bat_priv = bat_priv; | |
1027 | tp_vars->start_time = jiffies; | |
1028 | ||
1029 | init_waitqueue_head(&tp_vars->more_bytes); | |
1030 | ||
1031 | spin_lock_init(&tp_vars->unacked_lock); | |
1032 | INIT_LIST_HEAD(&tp_vars->unacked_list); | |
1033 | ||
1034 | spin_lock_init(&tp_vars->cwnd_lock); | |
1035 | ||
1036 | tp_vars->prerandom_offset = 0; | |
1037 | spin_lock_init(&tp_vars->prerandom_lock); | |
1038 | ||
1039 | kref_get(&tp_vars->refcount); | |
1040 | hlist_add_head_rcu(&tp_vars->list, &bat_priv->tp_list); | |
1041 | spin_unlock_bh(&bat_priv->tp_list_lock); | |
1042 | ||
1043 | tp_vars->test_length = test_length; | |
1044 | if (!tp_vars->test_length) | |
1045 | tp_vars->test_length = BATADV_TP_DEF_TEST_LENGTH; | |
1046 | ||
1047 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1048 | "Meter: starting throughput meter towards %pM (length=%ums)\n", | |
1049 | dst, test_length); | |
1050 | ||
1051 | /* init work item for finished tp tests */ | |
1052 | INIT_DELAYED_WORK(&tp_vars->finish_work, batadv_tp_sender_finish); | |
1053 | ||
1054 | /* start tp kthread. This way the write() call issued from userspace can | |
1055 | * happily return and avoid to block | |
1056 | */ | |
1057 | batadv_tp_start_kthread(tp_vars); | |
1058 | ||
1059 | /* don't return reference to new tp_vars */ | |
1060 | batadv_tp_vars_put(tp_vars); | |
1061 | } | |
1062 | ||
1063 | /** | |
1064 | * batadv_tp_stop - stop currently running tp meter session | |
1065 | * @bat_priv: the bat priv with all the soft interface information | |
1066 | * @dst: the receiver MAC address | |
1067 | * @return_value: reason for tp meter session stop | |
1068 | */ | |
1069 | void batadv_tp_stop(struct batadv_priv *bat_priv, const u8 *dst, | |
1070 | u8 return_value) | |
1071 | { | |
1072 | struct batadv_orig_node *orig_node; | |
1073 | struct batadv_tp_vars *tp_vars; | |
1074 | ||
1075 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1076 | "Meter: stopping test towards %pM\n", dst); | |
1077 | ||
1078 | orig_node = batadv_orig_hash_find(bat_priv, dst); | |
1079 | if (!orig_node) | |
1080 | return; | |
1081 | ||
1082 | tp_vars = batadv_tp_list_find(bat_priv, orig_node->orig); | |
1083 | if (!tp_vars) { | |
1084 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1085 | "Meter: trying to interrupt an already over connection\n"); | |
1086 | goto out; | |
1087 | } | |
1088 | ||
1089 | batadv_tp_sender_shutdown(tp_vars, return_value); | |
1090 | batadv_tp_vars_put(tp_vars); | |
1091 | out: | |
1092 | batadv_orig_node_put(orig_node); | |
1093 | } | |
1094 | ||
1095 | /** | |
1096 | * batadv_tp_reset_receiver_timer - reset the receiver shutdown timer | |
1097 | * @tp_vars: the private data of the current TP meter session | |
1098 | * | |
1099 | * start the receiver shutdown timer or reset it if already started | |
1100 | */ | |
1101 | static void batadv_tp_reset_receiver_timer(struct batadv_tp_vars *tp_vars) | |
1102 | { | |
1103 | mod_timer(&tp_vars->timer, | |
1104 | jiffies + msecs_to_jiffies(BATADV_TP_RECV_TIMEOUT)); | |
1105 | } | |
1106 | ||
1107 | /** | |
1108 | * batadv_tp_receiver_shutdown - stop a tp meter receiver when timeout is | |
1109 | * reached without received ack | |
1110 | * @arg: address of the related tp_vars | |
1111 | */ | |
1112 | static void batadv_tp_receiver_shutdown(unsigned long arg) | |
1113 | { | |
1114 | struct batadv_tp_vars *tp_vars = (struct batadv_tp_vars *)arg; | |
1115 | struct batadv_tp_unacked *un, *safe; | |
1116 | struct batadv_priv *bat_priv; | |
1117 | ||
1118 | bat_priv = tp_vars->bat_priv; | |
1119 | ||
1120 | /* if there is recent activity rearm the timer */ | |
1121 | if (!batadv_has_timed_out(tp_vars->last_recv_time, | |
1122 | BATADV_TP_RECV_TIMEOUT)) { | |
1123 | /* reset the receiver shutdown timer */ | |
1124 | batadv_tp_reset_receiver_timer(tp_vars); | |
1125 | return; | |
1126 | } | |
1127 | ||
1128 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1129 | "Shutting down for inactivity (more than %dms) from %pM\n", | |
1130 | BATADV_TP_RECV_TIMEOUT, tp_vars->other_end); | |
1131 | ||
1132 | spin_lock_bh(&tp_vars->bat_priv->tp_list_lock); | |
1133 | hlist_del_rcu(&tp_vars->list); | |
1134 | spin_unlock_bh(&tp_vars->bat_priv->tp_list_lock); | |
1135 | ||
1136 | /* drop list reference */ | |
1137 | batadv_tp_vars_put(tp_vars); | |
1138 | ||
1139 | atomic_dec(&bat_priv->tp_num); | |
1140 | ||
1141 | spin_lock_bh(&tp_vars->unacked_lock); | |
1142 | list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) { | |
1143 | list_del(&un->list); | |
1144 | kfree(un); | |
1145 | } | |
1146 | spin_unlock_bh(&tp_vars->unacked_lock); | |
1147 | ||
1148 | /* drop reference of timer */ | |
1149 | batadv_tp_vars_put(tp_vars); | |
1150 | } | |
1151 | ||
1152 | /** | |
1153 | * batadv_tp_send_ack - send an ACK packet | |
1154 | * @bat_priv: the bat priv with all the soft interface information | |
1155 | * @dst: the mac address of the destination originator | |
1156 | * @seq: the sequence number to ACK | |
1157 | * @timestamp: the timestamp to echo back in the ACK | |
1158 | * @session: session identifier | |
1159 | * @socket_index: local ICMP socket identifier | |
1160 | * | |
1161 | * Return: 0 on success, a positive integer representing the reason of the | |
1162 | * failure otherwise | |
1163 | */ | |
1164 | static int batadv_tp_send_ack(struct batadv_priv *bat_priv, const u8 *dst, | |
1165 | u32 seq, __be32 timestamp, const u8 *session, | |
1166 | int socket_index) | |
1167 | { | |
1168 | struct batadv_hard_iface *primary_if = NULL; | |
1169 | struct batadv_orig_node *orig_node; | |
1170 | struct batadv_icmp_tp_packet *icmp; | |
1171 | struct sk_buff *skb; | |
1172 | int r, ret; | |
1173 | ||
1174 | orig_node = batadv_orig_hash_find(bat_priv, dst); | |
1175 | if (unlikely(!orig_node)) { | |
1176 | ret = BATADV_TP_REASON_DST_UNREACHABLE; | |
1177 | goto out; | |
1178 | } | |
1179 | ||
1180 | primary_if = batadv_primary_if_get_selected(bat_priv); | |
1181 | if (unlikely(!primary_if)) { | |
1182 | ret = BATADV_TP_REASON_DST_UNREACHABLE; | |
1183 | goto out; | |
1184 | } | |
1185 | ||
1186 | skb = netdev_alloc_skb_ip_align(NULL, sizeof(*icmp) + ETH_HLEN); | |
1187 | if (unlikely(!skb)) { | |
1188 | ret = BATADV_TP_REASON_MEMORY_ERROR; | |
1189 | goto out; | |
1190 | } | |
1191 | ||
1192 | skb_reserve(skb, ETH_HLEN); | |
1193 | icmp = (struct batadv_icmp_tp_packet *)skb_put(skb, sizeof(*icmp)); | |
1194 | icmp->packet_type = BATADV_ICMP; | |
1195 | icmp->version = BATADV_COMPAT_VERSION; | |
1196 | icmp->ttl = BATADV_TTL; | |
1197 | icmp->msg_type = BATADV_TP; | |
1198 | ether_addr_copy(icmp->dst, orig_node->orig); | |
1199 | ether_addr_copy(icmp->orig, primary_if->net_dev->dev_addr); | |
1200 | icmp->uid = socket_index; | |
1201 | ||
1202 | icmp->subtype = BATADV_TP_ACK; | |
1203 | memcpy(icmp->session, session, sizeof(icmp->session)); | |
1204 | icmp->seqno = htonl(seq); | |
1205 | icmp->timestamp = timestamp; | |
1206 | ||
1207 | /* send the ack */ | |
1208 | r = batadv_send_skb_to_orig(skb, orig_node, NULL); | |
1209 | if (r == -1) | |
1210 | kfree_skb(skb); | |
1211 | ||
1212 | if (unlikely(r < 0) || (r == NET_XMIT_DROP)) { | |
1213 | ret = BATADV_TP_REASON_DST_UNREACHABLE; | |
1214 | goto out; | |
1215 | } | |
1216 | ret = 0; | |
1217 | ||
1218 | out: | |
1219 | if (likely(orig_node)) | |
1220 | batadv_orig_node_put(orig_node); | |
1221 | if (likely(primary_if)) | |
1222 | batadv_hardif_put(primary_if); | |
1223 | ||
1224 | return ret; | |
1225 | } | |
1226 | ||
1227 | /** | |
1228 | * batadv_tp_handle_out_of_order - store an out of order packet | |
1229 | * @tp_vars: the private data of the current TP meter session | |
1230 | * @skb: the buffer containing the received packet | |
1231 | * | |
1232 | * Store the out of order packet in the unacked list for late processing. This | |
1233 | * packets are kept in this list so that they can be ACKed at once as soon as | |
1234 | * all the previous packets have been received | |
1235 | * | |
1236 | * Return: true if the packed has been successfully processed, false otherwise | |
1237 | */ | |
1238 | static bool batadv_tp_handle_out_of_order(struct batadv_tp_vars *tp_vars, | |
1239 | const struct sk_buff *skb) | |
1240 | { | |
1241 | const struct batadv_icmp_tp_packet *icmp; | |
1242 | struct batadv_tp_unacked *un, *new; | |
1243 | u32 payload_len; | |
1244 | bool added = false; | |
1245 | ||
1246 | new = kmalloc(sizeof(*new), GFP_ATOMIC); | |
1247 | if (unlikely(!new)) | |
1248 | return false; | |
1249 | ||
1250 | icmp = (struct batadv_icmp_tp_packet *)skb->data; | |
1251 | ||
1252 | new->seqno = ntohl(icmp->seqno); | |
1253 | payload_len = skb->len - sizeof(struct batadv_unicast_packet); | |
1254 | new->len = payload_len; | |
1255 | ||
1256 | spin_lock_bh(&tp_vars->unacked_lock); | |
1257 | /* if the list is empty immediately attach this new object */ | |
1258 | if (list_empty(&tp_vars->unacked_list)) { | |
1259 | list_add(&new->list, &tp_vars->unacked_list); | |
1260 | goto out; | |
1261 | } | |
1262 | ||
1263 | /* otherwise loop over the list and either drop the packet because this | |
1264 | * is a duplicate or store it at the right position. | |
1265 | * | |
1266 | * The iteration is done in the reverse way because it is likely that | |
1267 | * the last received packet (the one being processed now) has a bigger | |
1268 | * seqno than all the others already stored. | |
1269 | */ | |
1270 | list_for_each_entry_reverse(un, &tp_vars->unacked_list, list) { | |
1271 | /* check for duplicates */ | |
1272 | if (new->seqno == un->seqno) { | |
1273 | if (new->len > un->len) | |
1274 | un->len = new->len; | |
1275 | kfree(new); | |
1276 | added = true; | |
1277 | break; | |
1278 | } | |
1279 | ||
1280 | /* look for the right position */ | |
1281 | if (batadv_seq_before(new->seqno, un->seqno)) | |
1282 | continue; | |
1283 | ||
1284 | /* as soon as an entry having a bigger seqno is found, the new | |
1285 | * one is attached _after_ it. In this way the list is kept in | |
1286 | * ascending order | |
1287 | */ | |
1288 | list_add_tail(&new->list, &un->list); | |
1289 | added = true; | |
1290 | break; | |
1291 | } | |
1292 | ||
1293 | /* received packet with smallest seqno out of order; add it to front */ | |
1294 | if (!added) | |
1295 | list_add(&new->list, &tp_vars->unacked_list); | |
1296 | ||
1297 | out: | |
1298 | spin_unlock_bh(&tp_vars->unacked_lock); | |
1299 | ||
1300 | return true; | |
1301 | } | |
1302 | ||
1303 | /** | |
1304 | * batadv_tp_ack_unordered - update number received bytes in current stream | |
1305 | * without gaps | |
1306 | * @tp_vars: the private data of the current TP meter session | |
1307 | */ | |
1308 | static void batadv_tp_ack_unordered(struct batadv_tp_vars *tp_vars) | |
1309 | { | |
1310 | struct batadv_tp_unacked *un, *safe; | |
1311 | u32 to_ack; | |
1312 | ||
1313 | /* go through the unacked packet list and possibly ACK them as | |
1314 | * well | |
1315 | */ | |
1316 | spin_lock_bh(&tp_vars->unacked_lock); | |
1317 | list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) { | |
1318 | /* the list is ordered, therefore it is possible to stop as soon | |
1319 | * there is a gap between the last acked seqno and the seqno of | |
1320 | * the packet under inspection | |
1321 | */ | |
1322 | if (batadv_seq_before(tp_vars->last_recv, un->seqno)) | |
1323 | break; | |
1324 | ||
1325 | to_ack = un->seqno + un->len - tp_vars->last_recv; | |
1326 | ||
1327 | if (batadv_seq_before(tp_vars->last_recv, un->seqno + un->len)) | |
1328 | tp_vars->last_recv += to_ack; | |
1329 | ||
1330 | list_del(&un->list); | |
1331 | kfree(un); | |
1332 | } | |
1333 | spin_unlock_bh(&tp_vars->unacked_lock); | |
1334 | } | |
1335 | ||
1336 | /** | |
1337 | * batadv_tp_init_recv - return matching or create new receiver tp_vars | |
1338 | * @bat_priv: the bat priv with all the soft interface information | |
1339 | * @icmp: received icmp tp msg | |
1340 | * | |
1341 | * Return: corresponding tp_vars or NULL on errors | |
1342 | */ | |
1343 | static struct batadv_tp_vars * | |
1344 | batadv_tp_init_recv(struct batadv_priv *bat_priv, | |
1345 | const struct batadv_icmp_tp_packet *icmp) | |
1346 | { | |
1347 | struct batadv_tp_vars *tp_vars; | |
1348 | ||
1349 | spin_lock_bh(&bat_priv->tp_list_lock); | |
1350 | tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig, | |
1351 | icmp->session); | |
1352 | if (tp_vars) | |
1353 | goto out_unlock; | |
1354 | ||
1355 | if (!atomic_add_unless(&bat_priv->tp_num, 1, BATADV_TP_MAX_NUM)) { | |
1356 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1357 | "Meter: too many ongoing sessions, aborting (RECV)\n"); | |
1358 | goto out_unlock; | |
1359 | } | |
1360 | ||
1361 | tp_vars = kmalloc(sizeof(*tp_vars), GFP_ATOMIC); | |
1362 | if (!tp_vars) | |
1363 | goto out_unlock; | |
1364 | ||
1365 | ether_addr_copy(tp_vars->other_end, icmp->orig); | |
1366 | tp_vars->role = BATADV_TP_RECEIVER; | |
1367 | memcpy(tp_vars->session, icmp->session, sizeof(tp_vars->session)); | |
1368 | tp_vars->last_recv = BATADV_TP_FIRST_SEQ; | |
1369 | tp_vars->bat_priv = bat_priv; | |
1370 | kref_init(&tp_vars->refcount); | |
1371 | ||
1372 | spin_lock_init(&tp_vars->unacked_lock); | |
1373 | INIT_LIST_HEAD(&tp_vars->unacked_list); | |
1374 | ||
1375 | kref_get(&tp_vars->refcount); | |
1376 | hlist_add_head_rcu(&tp_vars->list, &bat_priv->tp_list); | |
1377 | ||
1378 | kref_get(&tp_vars->refcount); | |
1379 | setup_timer(&tp_vars->timer, batadv_tp_receiver_shutdown, | |
1380 | (unsigned long)tp_vars); | |
1381 | ||
1382 | batadv_tp_reset_receiver_timer(tp_vars); | |
1383 | ||
1384 | out_unlock: | |
1385 | spin_unlock_bh(&bat_priv->tp_list_lock); | |
1386 | ||
1387 | return tp_vars; | |
1388 | } | |
1389 | ||
1390 | /** | |
1391 | * batadv_tp_recv_msg - process a single data message | |
1392 | * @bat_priv: the bat priv with all the soft interface information | |
1393 | * @skb: the buffer containing the received packet | |
1394 | * | |
1395 | * Process a received TP MSG packet | |
1396 | */ | |
1397 | static void batadv_tp_recv_msg(struct batadv_priv *bat_priv, | |
1398 | const struct sk_buff *skb) | |
1399 | { | |
1400 | const struct batadv_icmp_tp_packet *icmp; | |
1401 | struct batadv_tp_vars *tp_vars; | |
1402 | size_t packet_size; | |
1403 | u32 seqno; | |
1404 | ||
1405 | icmp = (struct batadv_icmp_tp_packet *)skb->data; | |
1406 | ||
1407 | seqno = ntohl(icmp->seqno); | |
1408 | /* check if this is the first seqno. This means that if the | |
1409 | * first packet is lost, the tp meter does not work anymore! | |
1410 | */ | |
1411 | if (seqno == BATADV_TP_FIRST_SEQ) { | |
1412 | tp_vars = batadv_tp_init_recv(bat_priv, icmp); | |
1413 | if (!tp_vars) { | |
1414 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1415 | "Meter: seqno != BATADV_TP_FIRST_SEQ cannot initiate connection\n"); | |
1416 | goto out; | |
1417 | } | |
1418 | } else { | |
1419 | tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig, | |
1420 | icmp->session); | |
1421 | if (!tp_vars) { | |
1422 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1423 | "Unexpected packet from %pM!\n", | |
1424 | icmp->orig); | |
1425 | goto out; | |
1426 | } | |
1427 | } | |
1428 | ||
1429 | if (unlikely(tp_vars->role != BATADV_TP_RECEIVER)) { | |
1430 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1431 | "Meter: dropping packet: not expected (role=%u)\n", | |
1432 | tp_vars->role); | |
1433 | goto out; | |
1434 | } | |
1435 | ||
1436 | tp_vars->last_recv_time = jiffies; | |
1437 | ||
1438 | /* if the packet is a duplicate, it may be the case that an ACK has been | |
1439 | * lost. Resend the ACK | |
1440 | */ | |
1441 | if (batadv_seq_before(seqno, tp_vars->last_recv)) | |
1442 | goto send_ack; | |
1443 | ||
1444 | /* if the packet is out of order enqueue it */ | |
1445 | if (ntohl(icmp->seqno) != tp_vars->last_recv) { | |
1446 | /* exit immediately (and do not send any ACK) if the packet has | |
1447 | * not been enqueued correctly | |
1448 | */ | |
1449 | if (!batadv_tp_handle_out_of_order(tp_vars, skb)) | |
1450 | goto out; | |
1451 | ||
1452 | /* send a duplicate ACK */ | |
1453 | goto send_ack; | |
1454 | } | |
1455 | ||
1456 | /* if everything was fine count the ACKed bytes */ | |
1457 | packet_size = skb->len - sizeof(struct batadv_unicast_packet); | |
1458 | tp_vars->last_recv += packet_size; | |
1459 | ||
1460 | /* check if this ordered message filled a gap.... */ | |
1461 | batadv_tp_ack_unordered(tp_vars); | |
1462 | ||
1463 | send_ack: | |
1464 | /* send the ACK. If the received packet was out of order, the ACK that | |
1465 | * is going to be sent is a duplicate (the sender will count them and | |
1466 | * possibly enter Fast Retransmit as soon as it has reached 3) | |
1467 | */ | |
1468 | batadv_tp_send_ack(bat_priv, icmp->orig, tp_vars->last_recv, | |
1469 | icmp->timestamp, icmp->session, icmp->uid); | |
1470 | out: | |
1471 | if (likely(tp_vars)) | |
1472 | batadv_tp_vars_put(tp_vars); | |
1473 | } | |
1474 | ||
1475 | /** | |
1476 | * batadv_tp_meter_recv - main TP Meter receiving function | |
1477 | * @bat_priv: the bat priv with all the soft interface information | |
1478 | * @skb: the buffer containing the received packet | |
1479 | */ | |
1480 | void batadv_tp_meter_recv(struct batadv_priv *bat_priv, struct sk_buff *skb) | |
1481 | { | |
1482 | struct batadv_icmp_tp_packet *icmp; | |
1483 | ||
1484 | icmp = (struct batadv_icmp_tp_packet *)skb->data; | |
1485 | ||
1486 | switch (icmp->subtype) { | |
1487 | case BATADV_TP_MSG: | |
1488 | batadv_tp_recv_msg(bat_priv, skb); | |
1489 | break; | |
1490 | case BATADV_TP_ACK: | |
1491 | batadv_tp_recv_ack(bat_priv, skb); | |
1492 | break; | |
1493 | default: | |
1494 | batadv_dbg(BATADV_DBG_TP_METER, bat_priv, | |
1495 | "Received unknown TP Metric packet type %u\n", | |
1496 | icmp->subtype); | |
1497 | } | |
1498 | consume_skb(skb); | |
1499 | } | |
1500 | ||
1501 | /** | |
1502 | * batadv_tp_meter_init - initialize global tp_meter structures | |
1503 | */ | |
1504 | void batadv_tp_meter_init(void) | |
1505 | { | |
1506 | get_random_bytes(batadv_tp_prerandom, sizeof(batadv_tp_prerandom)); | |
1507 | } |