~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/batman-adv/tp_meter.c

Version: ~ [ linux-5.12-rc1 ] ~ [ linux-5.11.2 ] ~ [ linux-5.10.19 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.101 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.177 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.222 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.258 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.258 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.18.140 ] ~ [ linux-3.16.85 ] ~ [ linux-3.14.79 ] ~ [ linux-3.12.74 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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                 tp_vars->reason = err;
841                 goto out;
842         }
843 
844         /* assume that all the hard_interfaces have a correctly
845          * configured MTU, so use the soft_iface MTU as MSS.
846          * This might not be true and in that case the fragmentation
847          * should be used.
848          * Now, try to send the packet as it is
849          */
850         payload_len = BATADV_TP_PLEN;
851         BUILD_BUG_ON(sizeof(struct batadv_icmp_tp_packet) > BATADV_TP_PLEN);
852 
853         batadv_tp_reset_sender_timer(tp_vars);
854 
855         /* queue the worker in charge of terminating the test */
856         queue_delayed_work(batadv_event_workqueue, &tp_vars->finish_work,
857                            msecs_to_jiffies(tp_vars->test_length));
858 
859         while (atomic_read(&tp_vars->sending) != 0) {
860                 if (unlikely(!batadv_tp_avail(tp_vars, payload_len))) {
861                         batadv_tp_wait_available(tp_vars, payload_len);
862                         continue;
863                 }
864 
865                 /* to emulate normal unicast traffic, add to the payload len
866                  * the size of the unicast header
867                  */
868                 packet_len = payload_len + sizeof(struct batadv_unicast_packet);
869 
870                 err = batadv_tp_send_msg(tp_vars, primary_if->net_dev->dev_addr,
871                                          orig_node, tp_vars->last_sent,
872                                          packet_len,
873                                          tp_vars->session, tp_vars->icmp_uid,
874                                          jiffies_to_msecs(jiffies));
875 
876                 /* something went wrong during the preparation/transmission */
877                 if (unlikely(err && err != BATADV_TP_REASON_CANT_SEND)) {
878                         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
879                                    "Meter: batadv_tp_send() cannot send packets (%d)\n",
880                                    err);
881                         /* ensure nobody else tries to stop the thread now */
882                         if (atomic_dec_and_test(&tp_vars->sending))
883                                 tp_vars->reason = err;
884                         break;
885                 }
886 
887                 /* right-shift the TWND */
888                 if (!err)
889                         tp_vars->last_sent += payload_len;
890 
891                 cond_resched();
892         }
893 
894 out:
895         if (likely(primary_if))
896                 batadv_hardif_put(primary_if);
897         if (likely(orig_node))
898                 batadv_orig_node_put(orig_node);
899 
900         batadv_tp_sender_end(bat_priv, tp_vars);
901         batadv_tp_sender_cleanup(bat_priv, tp_vars);
902 
903         batadv_tp_vars_put(tp_vars);
904 
905         do_exit(0);
906 }
907 
908 /**
909  * batadv_tp_start_kthread - start new thread which manages the tp meter sender
910  * @tp_vars: the private data of the current TP meter session
911  */
912 static void batadv_tp_start_kthread(struct batadv_tp_vars *tp_vars)
913 {
914         struct task_struct *kthread;
915         struct batadv_priv *bat_priv = tp_vars->bat_priv;
916         u32 session_cookie;
917 
918         kref_get(&tp_vars->refcount);
919         kthread = kthread_create(batadv_tp_send, tp_vars, "kbatadv_tp_meter");
920         if (IS_ERR(kthread)) {
921                 session_cookie = batadv_tp_session_cookie(tp_vars->session,
922                                                           tp_vars->icmp_uid);
923                 pr_err("batadv: cannot create tp meter kthread\n");
924                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_MEMORY_ERROR,
925                                               tp_vars->other_end,
926                                               bat_priv, session_cookie);
927 
928                 /* drop reserved reference for kthread */
929                 batadv_tp_vars_put(tp_vars);
930 
931                 /* cleanup of failed tp meter variables */
932                 batadv_tp_sender_cleanup(bat_priv, tp_vars);
933                 return;
934         }
935 
936         wake_up_process(kthread);
937 }
938 
939 /**
940  * batadv_tp_start - start a new tp meter session
941  * @bat_priv: the bat priv with all the soft interface information
942  * @dst: the receiver MAC address
943  * @test_length: test length in milliseconds
944  * @cookie: session cookie
945  */
946 void batadv_tp_start(struct batadv_priv *bat_priv, const u8 *dst,
947                      u32 test_length, u32 *cookie)
948 {
949         struct batadv_tp_vars *tp_vars;
950         u8 session_id[2];
951         u8 icmp_uid;
952         u32 session_cookie;
953 
954         get_random_bytes(session_id, sizeof(session_id));
955         get_random_bytes(&icmp_uid, 1);
956         session_cookie = batadv_tp_session_cookie(session_id, icmp_uid);
957         *cookie = session_cookie;
958 
959         /* look for an already existing test towards this node */
960         spin_lock_bh(&bat_priv->tp_list_lock);
961         tp_vars = batadv_tp_list_find(bat_priv, dst);
962         if (tp_vars) {
963                 spin_unlock_bh(&bat_priv->tp_list_lock);
964                 batadv_tp_vars_put(tp_vars);
965                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
966                            "Meter: test to or from the same node already ongoing, aborting\n");
967                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_ALREADY_ONGOING,
968                                               dst, bat_priv, session_cookie);
969                 return;
970         }
971 
972         if (!atomic_add_unless(&bat_priv->tp_num, 1, BATADV_TP_MAX_NUM)) {
973                 spin_unlock_bh(&bat_priv->tp_list_lock);
974                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
975                            "Meter: too many ongoing sessions, aborting (SEND)\n");
976                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_TOO_MANY, dst,
977                                               bat_priv, session_cookie);
978                 return;
979         }
980 
981         tp_vars = kmalloc(sizeof(*tp_vars), GFP_ATOMIC);
982         if (!tp_vars) {
983                 spin_unlock_bh(&bat_priv->tp_list_lock);
984                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
985                            "Meter: batadv_tp_start cannot allocate list elements\n");
986                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_MEMORY_ERROR,
987                                               dst, bat_priv, session_cookie);
988                 return;
989         }
990 
991         /* initialize tp_vars */
992         ether_addr_copy(tp_vars->other_end, dst);
993         kref_init(&tp_vars->refcount);
994         tp_vars->role = BATADV_TP_SENDER;
995         atomic_set(&tp_vars->sending, 1);
996         memcpy(tp_vars->session, session_id, sizeof(session_id));
997         tp_vars->icmp_uid = icmp_uid;
998 
999         tp_vars->last_sent = BATADV_TP_FIRST_SEQ;
1000         atomic_set(&tp_vars->last_acked, BATADV_TP_FIRST_SEQ);
1001         tp_vars->fast_recovery = false;
1002         tp_vars->recover = BATADV_TP_FIRST_SEQ;
1003 
1004         /* initialise the CWND to 3*MSS (Section 3.1 in RFC5681).
1005          * For batman-adv the MSS is the size of the payload received by the
1006          * soft_interface, hence its MTU
1007          */
1008         tp_vars->cwnd = BATADV_TP_PLEN * 3;
1009         /* at the beginning initialise the SS threshold to the biggest possible
1010          * window size, hence the AWND size
1011          */
1012         tp_vars->ss_threshold = BATADV_TP_AWND;
1013 
1014         /* RTO initial value is 3 seconds.
1015          * Details in Section 2.1 of RFC6298
1016          */
1017         tp_vars->rto = 1000;
1018         tp_vars->srtt = 0;
1019         tp_vars->rttvar = 0;
1020 
1021         atomic64_set(&tp_vars->tot_sent, 0);
1022 
1023         kref_get(&tp_vars->refcount);
1024         setup_timer(&tp_vars->timer, batadv_tp_sender_timeout,
1025                     (unsigned long)tp_vars);
1026 
1027         tp_vars->bat_priv = bat_priv;
1028         tp_vars->start_time = jiffies;
1029 
1030         init_waitqueue_head(&tp_vars->more_bytes);
1031 
1032         spin_lock_init(&tp_vars->unacked_lock);
1033         INIT_LIST_HEAD(&tp_vars->unacked_list);
1034 
1035         spin_lock_init(&tp_vars->cwnd_lock);
1036 
1037         tp_vars->prerandom_offset = 0;
1038         spin_lock_init(&tp_vars->prerandom_lock);
1039 
1040         kref_get(&tp_vars->refcount);
1041         hlist_add_head_rcu(&tp_vars->list, &bat_priv->tp_list);
1042         spin_unlock_bh(&bat_priv->tp_list_lock);
1043 
1044         tp_vars->test_length = test_length;
1045         if (!tp_vars->test_length)
1046                 tp_vars->test_length = BATADV_TP_DEF_TEST_LENGTH;
1047 
1048         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1049                    "Meter: starting throughput meter towards %pM (length=%ums)\n",
1050                    dst, test_length);
1051 
1052         /* init work item for finished tp tests */
1053         INIT_DELAYED_WORK(&tp_vars->finish_work, batadv_tp_sender_finish);
1054 
1055         /* start tp kthread. This way the write() call issued from userspace can
1056          * happily return and avoid to block
1057          */
1058         batadv_tp_start_kthread(tp_vars);
1059 
1060         /* don't return reference to new tp_vars */
1061         batadv_tp_vars_put(tp_vars);
1062 }
1063 
1064 /**
1065  * batadv_tp_stop - stop currently running tp meter session
1066  * @bat_priv: the bat priv with all the soft interface information
1067  * @dst: the receiver MAC address
1068  * @return_value: reason for tp meter session stop
1069  */
1070 void batadv_tp_stop(struct batadv_priv *bat_priv, const u8 *dst,
1071                     u8 return_value)
1072 {
1073         struct batadv_orig_node *orig_node;
1074         struct batadv_tp_vars *tp_vars;
1075 
1076         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1077                    "Meter: stopping test towards %pM\n", dst);
1078 
1079         orig_node = batadv_orig_hash_find(bat_priv, dst);
1080         if (!orig_node)
1081                 return;
1082 
1083         tp_vars = batadv_tp_list_find(bat_priv, orig_node->orig);
1084         if (!tp_vars) {
1085                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1086                            "Meter: trying to interrupt an already over connection\n");
1087                 goto out;
1088         }
1089 
1090         batadv_tp_sender_shutdown(tp_vars, return_value);
1091         batadv_tp_vars_put(tp_vars);
1092 out:
1093         batadv_orig_node_put(orig_node);
1094 }
1095 
1096 /**
1097  * batadv_tp_reset_receiver_timer - reset the receiver shutdown timer
1098  * @tp_vars: the private data of the current TP meter session
1099  *
1100  * start the receiver shutdown timer or reset it if already started
1101  */
1102 static void batadv_tp_reset_receiver_timer(struct batadv_tp_vars *tp_vars)
1103 {
1104         mod_timer(&tp_vars->timer,
1105                   jiffies + msecs_to_jiffies(BATADV_TP_RECV_TIMEOUT));
1106 }
1107 
1108 /**
1109  * batadv_tp_receiver_shutdown - stop a tp meter receiver when timeout is
1110  *  reached without received ack
1111  * @arg: address of the related tp_vars
1112  */
1113 static void batadv_tp_receiver_shutdown(unsigned long arg)
1114 {
1115         struct batadv_tp_vars *tp_vars = (struct batadv_tp_vars *)arg;
1116         struct batadv_tp_unacked *un, *safe;
1117         struct batadv_priv *bat_priv;
1118 
1119         bat_priv = tp_vars->bat_priv;
1120 
1121         /* if there is recent activity rearm the timer */
1122         if (!batadv_has_timed_out(tp_vars->last_recv_time,
1123                                   BATADV_TP_RECV_TIMEOUT)) {
1124                 /* reset the receiver shutdown timer */
1125                 batadv_tp_reset_receiver_timer(tp_vars);
1126                 return;
1127         }
1128 
1129         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1130                    "Shutting down for inactivity (more than %dms) from %pM\n",
1131                    BATADV_TP_RECV_TIMEOUT, tp_vars->other_end);
1132 
1133         spin_lock_bh(&tp_vars->bat_priv->tp_list_lock);
1134         hlist_del_rcu(&tp_vars->list);
1135         spin_unlock_bh(&tp_vars->bat_priv->tp_list_lock);
1136 
1137         /* drop list reference */
1138         batadv_tp_vars_put(tp_vars);
1139 
1140         atomic_dec(&bat_priv->tp_num);
1141 
1142         spin_lock_bh(&tp_vars->unacked_lock);
1143         list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) {
1144                 list_del(&un->list);
1145                 kfree(un);
1146         }
1147         spin_unlock_bh(&tp_vars->unacked_lock);
1148 
1149         /* drop reference of timer */
1150         batadv_tp_vars_put(tp_vars);
1151 }
1152 
1153 /**
1154  * batadv_tp_send_ack - send an ACK packet
1155  * @bat_priv: the bat priv with all the soft interface information
1156  * @dst: the mac address of the destination originator
1157  * @seq: the sequence number to ACK
1158  * @timestamp: the timestamp to echo back in the ACK
1159  * @session: session identifier
1160  * @socket_index: local ICMP socket identifier
1161  *
1162  * Return: 0 on success, a positive integer representing the reason of the
1163  * failure otherwise
1164  */
1165 static int batadv_tp_send_ack(struct batadv_priv *bat_priv, const u8 *dst,
1166                               u32 seq, __be32 timestamp, const u8 *session,
1167                               int socket_index)
1168 {
1169         struct batadv_hard_iface *primary_if = NULL;
1170         struct batadv_orig_node *orig_node;
1171         struct batadv_icmp_tp_packet *icmp;
1172         struct sk_buff *skb;
1173         int r, ret;
1174 
1175         orig_node = batadv_orig_hash_find(bat_priv, dst);
1176         if (unlikely(!orig_node)) {
1177                 ret = BATADV_TP_REASON_DST_UNREACHABLE;
1178                 goto out;
1179         }
1180 
1181         primary_if = batadv_primary_if_get_selected(bat_priv);
1182         if (unlikely(!primary_if)) {
1183                 ret = BATADV_TP_REASON_DST_UNREACHABLE;
1184                 goto out;
1185         }
1186 
1187         skb = netdev_alloc_skb_ip_align(NULL, sizeof(*icmp) + ETH_HLEN);
1188         if (unlikely(!skb)) {
1189                 ret = BATADV_TP_REASON_MEMORY_ERROR;
1190                 goto out;
1191         }
1192 
1193         skb_reserve(skb, ETH_HLEN);
1194         icmp = (struct batadv_icmp_tp_packet *)skb_put(skb, sizeof(*icmp));
1195         icmp->packet_type = BATADV_ICMP;
1196         icmp->version = BATADV_COMPAT_VERSION;
1197         icmp->ttl = BATADV_TTL;
1198         icmp->msg_type = BATADV_TP;
1199         ether_addr_copy(icmp->dst, orig_node->orig);
1200         ether_addr_copy(icmp->orig, primary_if->net_dev->dev_addr);
1201         icmp->uid = socket_index;
1202 
1203         icmp->subtype = BATADV_TP_ACK;
1204         memcpy(icmp->session, session, sizeof(icmp->session));
1205         icmp->seqno = htonl(seq);
1206         icmp->timestamp = timestamp;
1207 
1208         /* send the ack */
1209         r = batadv_send_skb_to_orig(skb, orig_node, NULL);
1210         if (r == -1)
1211                 kfree_skb(skb);
1212 
1213         if (unlikely(r < 0) || (r == NET_XMIT_DROP)) {
1214                 ret = BATADV_TP_REASON_DST_UNREACHABLE;
1215                 goto out;
1216         }
1217         ret = 0;
1218 
1219 out:
1220         if (likely(orig_node))
1221                 batadv_orig_node_put(orig_node);
1222         if (likely(primary_if))
1223                 batadv_hardif_put(primary_if);
1224 
1225         return ret;
1226 }
1227 
1228 /**
1229  * batadv_tp_handle_out_of_order - store an out of order packet
1230  * @tp_vars: the private data of the current TP meter session
1231  * @skb: the buffer containing the received packet
1232  *
1233  * Store the out of order packet in the unacked list for late processing. This
1234  * packets are kept in this list so that they can be ACKed at once as soon as
1235  * all the previous packets have been received
1236  *
1237  * Return: true if the packed has been successfully processed, false otherwise
1238  */
1239 static bool batadv_tp_handle_out_of_order(struct batadv_tp_vars *tp_vars,
1240                                           const struct sk_buff *skb)
1241 {
1242         const struct batadv_icmp_tp_packet *icmp;
1243         struct batadv_tp_unacked *un, *new;
1244         u32 payload_len;
1245         bool added = false;
1246 
1247         new = kmalloc(sizeof(*new), GFP_ATOMIC);
1248         if (unlikely(!new))
1249                 return false;
1250 
1251         icmp = (struct batadv_icmp_tp_packet *)skb->data;
1252 
1253         new->seqno = ntohl(icmp->seqno);
1254         payload_len = skb->len - sizeof(struct batadv_unicast_packet);
1255         new->len = payload_len;
1256 
1257         spin_lock_bh(&tp_vars->unacked_lock);
1258         /* if the list is empty immediately attach this new object */
1259         if (list_empty(&tp_vars->unacked_list)) {
1260                 list_add(&new->list, &tp_vars->unacked_list);
1261                 goto out;
1262         }
1263 
1264         /* otherwise loop over the list and either drop the packet because this
1265          * is a duplicate or store it at the right position.
1266          *
1267          * The iteration is done in the reverse way because it is likely that
1268          * the last received packet (the one being processed now) has a bigger
1269          * seqno than all the others already stored.
1270          */
1271         list_for_each_entry_reverse(un, &tp_vars->unacked_list, list) {
1272                 /* check for duplicates */
1273                 if (new->seqno == un->seqno) {
1274                         if (new->len > un->len)
1275                                 un->len = new->len;
1276                         kfree(new);
1277                         added = true;
1278                         break;
1279                 }
1280 
1281                 /* look for the right position */
1282                 if (batadv_seq_before(new->seqno, un->seqno))
1283                         continue;
1284 
1285                 /* as soon as an entry having a bigger seqno is found, the new
1286                  * one is attached _after_ it. In this way the list is kept in
1287                  * ascending order
1288                  */
1289                 list_add_tail(&new->list, &un->list);
1290                 added = true;
1291                 break;
1292         }
1293 
1294         /* received packet with smallest seqno out of order; add it to front */
1295         if (!added)
1296                 list_add(&new->list, &tp_vars->unacked_list);
1297 
1298 out:
1299         spin_unlock_bh(&tp_vars->unacked_lock);
1300 
1301         return true;
1302 }
1303 
1304 /**
1305  * batadv_tp_ack_unordered - update number received bytes in current stream
1306  *  without gaps
1307  * @tp_vars: the private data of the current TP meter session
1308  */
1309 static void batadv_tp_ack_unordered(struct batadv_tp_vars *tp_vars)
1310 {
1311         struct batadv_tp_unacked *un, *safe;
1312         u32 to_ack;
1313 
1314         /* go through the unacked packet list and possibly ACK them as
1315          * well
1316          */
1317         spin_lock_bh(&tp_vars->unacked_lock);
1318         list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) {
1319                 /* the list is ordered, therefore it is possible to stop as soon
1320                  * there is a gap between the last acked seqno and the seqno of
1321                  * the packet under inspection
1322                  */
1323                 if (batadv_seq_before(tp_vars->last_recv, un->seqno))
1324                         break;
1325 
1326                 to_ack = un->seqno + un->len - tp_vars->last_recv;
1327 
1328                 if (batadv_seq_before(tp_vars->last_recv, un->seqno + un->len))
1329                         tp_vars->last_recv += to_ack;
1330 
1331                 list_del(&un->list);
1332                 kfree(un);
1333         }
1334         spin_unlock_bh(&tp_vars->unacked_lock);
1335 }
1336 
1337 /**
1338  * batadv_tp_init_recv - return matching or create new receiver tp_vars
1339  * @bat_priv: the bat priv with all the soft interface information
1340  * @icmp: received icmp tp msg
1341  *
1342  * Return: corresponding tp_vars or NULL on errors
1343  */
1344 static struct batadv_tp_vars *
1345 batadv_tp_init_recv(struct batadv_priv *bat_priv,
1346                     const struct batadv_icmp_tp_packet *icmp)
1347 {
1348         struct batadv_tp_vars *tp_vars;
1349 
1350         spin_lock_bh(&bat_priv->tp_list_lock);
1351         tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig,
1352                                               icmp->session);
1353         if (tp_vars)
1354                 goto out_unlock;
1355 
1356         if (!atomic_add_unless(&bat_priv->tp_num, 1, BATADV_TP_MAX_NUM)) {
1357                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1358                            "Meter: too many ongoing sessions, aborting (RECV)\n");
1359                 goto out_unlock;
1360         }
1361 
1362         tp_vars = kmalloc(sizeof(*tp_vars), GFP_ATOMIC);
1363         if (!tp_vars)
1364                 goto out_unlock;
1365 
1366         ether_addr_copy(tp_vars->other_end, icmp->orig);
1367         tp_vars->role = BATADV_TP_RECEIVER;
1368         memcpy(tp_vars->session, icmp->session, sizeof(tp_vars->session));
1369         tp_vars->last_recv = BATADV_TP_FIRST_SEQ;
1370         tp_vars->bat_priv = bat_priv;
1371         kref_init(&tp_vars->refcount);
1372 
1373         spin_lock_init(&tp_vars->unacked_lock);
1374         INIT_LIST_HEAD(&tp_vars->unacked_list);
1375 
1376         kref_get(&tp_vars->refcount);
1377         hlist_add_head_rcu(&tp_vars->list, &bat_priv->tp_list);
1378 
1379         kref_get(&tp_vars->refcount);
1380         setup_timer(&tp_vars->timer, batadv_tp_receiver_shutdown,
1381                     (unsigned long)tp_vars);
1382 
1383         batadv_tp_reset_receiver_timer(tp_vars);
1384 
1385 out_unlock:
1386         spin_unlock_bh(&bat_priv->tp_list_lock);
1387 
1388         return tp_vars;
1389 }
1390 
1391 /**
1392  * batadv_tp_recv_msg - process a single data message
1393  * @bat_priv: the bat priv with all the soft interface information
1394  * @skb: the buffer containing the received packet
1395  *
1396  * Process a received TP MSG packet
1397  */
1398 static void batadv_tp_recv_msg(struct batadv_priv *bat_priv,
1399                                const struct sk_buff *skb)
1400 {
1401         const struct batadv_icmp_tp_packet *icmp;
1402         struct batadv_tp_vars *tp_vars;
1403         size_t packet_size;
1404         u32 seqno;
1405 
1406         icmp = (struct batadv_icmp_tp_packet *)skb->data;
1407 
1408         seqno = ntohl(icmp->seqno);
1409         /* check if this is the first seqno. This means that if the
1410          * first packet is lost, the tp meter does not work anymore!
1411          */
1412         if (seqno == BATADV_TP_FIRST_SEQ) {
1413                 tp_vars = batadv_tp_init_recv(bat_priv, icmp);
1414                 if (!tp_vars) {
1415                         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1416                                    "Meter: seqno != BATADV_TP_FIRST_SEQ cannot initiate connection\n");
1417                         goto out;
1418                 }
1419         } else {
1420                 tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig,
1421                                                       icmp->session);
1422                 if (!tp_vars) {
1423                         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1424                                    "Unexpected packet from %pM!\n",
1425                                    icmp->orig);
1426                         goto out;
1427                 }
1428         }
1429 
1430         if (unlikely(tp_vars->role != BATADV_TP_RECEIVER)) {
1431                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1432                            "Meter: dropping packet: not expected (role=%u)\n",
1433                            tp_vars->role);
1434                 goto out;
1435         }
1436 
1437         tp_vars->last_recv_time = jiffies;
1438 
1439         /* if the packet is a duplicate, it may be the case that an ACK has been
1440          * lost. Resend the ACK
1441          */
1442         if (batadv_seq_before(seqno, tp_vars->last_recv))
1443                 goto send_ack;
1444 
1445         /* if the packet is out of order enqueue it */
1446         if (ntohl(icmp->seqno) != tp_vars->last_recv) {
1447                 /* exit immediately (and do not send any ACK) if the packet has
1448                  * not been enqueued correctly
1449                  */
1450                 if (!batadv_tp_handle_out_of_order(tp_vars, skb))
1451                         goto out;
1452 
1453                 /* send a duplicate ACK */
1454                 goto send_ack;
1455         }
1456 
1457         /* if everything was fine count the ACKed bytes */
1458         packet_size = skb->len - sizeof(struct batadv_unicast_packet);
1459         tp_vars->last_recv += packet_size;
1460 
1461         /* check if this ordered message filled a gap.... */
1462         batadv_tp_ack_unordered(tp_vars);
1463 
1464 send_ack:
1465         /* send the ACK. If the received packet was out of order, the ACK that
1466          * is going to be sent is a duplicate (the sender will count them and
1467          * possibly enter Fast Retransmit as soon as it has reached 3)
1468          */
1469         batadv_tp_send_ack(bat_priv, icmp->orig, tp_vars->last_recv,
1470                            icmp->timestamp, icmp->session, icmp->uid);
1471 out:
1472         if (likely(tp_vars))
1473                 batadv_tp_vars_put(tp_vars);
1474 }
1475 
1476 /**
1477  * batadv_tp_meter_recv - main TP Meter receiving function
1478  * @bat_priv: the bat priv with all the soft interface information
1479  * @skb: the buffer containing the received packet
1480  */
1481 void batadv_tp_meter_recv(struct batadv_priv *bat_priv, struct sk_buff *skb)
1482 {
1483         struct batadv_icmp_tp_packet *icmp;
1484 
1485         icmp = (struct batadv_icmp_tp_packet *)skb->data;
1486 
1487         switch (icmp->subtype) {
1488         case BATADV_TP_MSG:
1489                 batadv_tp_recv_msg(bat_priv, skb);
1490                 break;
1491         case BATADV_TP_ACK:
1492                 batadv_tp_recv_ack(bat_priv, skb);
1493                 break;
1494         default:
1495                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1496                            "Received unknown TP Metric packet type %u\n",
1497                            icmp->subtype);
1498         }
1499         consume_skb(skb);
1500 }
1501 
1502 /**
1503  * batadv_tp_meter_init - initialize global tp_meter structures
1504  */
1505 void batadv_tp_meter_init(void)
1506 {
1507         get_random_bytes(batadv_tp_prerandom, sizeof(batadv_tp_prerandom));
1508 }
1509 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp