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

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

Version: ~ [ linux-5.7 ] ~ [ linux-5.6.15 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.43 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.125 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.182 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.225 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.225 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.84 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ 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 // SPDX-License-Identifier: GPL-2.0
  2 /* Copyright (C) 2012-2018  B.A.T.M.A.N. contributors:
  3  *
  4  * Edo Monticelli, Antonio Quartulli
  5  *
  6  * This program is free software; you can redistribute it and/or
  7  * modify it under the terms of version 2 of the GNU General Public
  8  * License as published by the Free Software Foundation.
  9  *
 10  * This program is distributed in the hope that it will be useful, but
 11  * WITHOUT ANY WARRANTY; without even the implied warranty of
 12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
 13  * General Public License for more details.
 14  *
 15  * You should have received a copy of the GNU General Public License
 16  * along with this program; if not, see <http://www.gnu.org/licenses/>.
 17  */
 18 
 19 #include "tp_meter.h"
 20 #include "main.h"
 21 
 22 #include <linux/atomic.h>
 23 #include <linux/build_bug.h>
 24 #include <linux/byteorder/generic.h>
 25 #include <linux/cache.h>
 26 #include <linux/compiler.h>
 27 #include <linux/err.h>
 28 #include <linux/etherdevice.h>
 29 #include <linux/gfp.h>
 30 #include <linux/if_ether.h>
 31 #include <linux/init.h>
 32 #include <linux/jiffies.h>
 33 #include <linux/kernel.h>
 34 #include <linux/kref.h>
 35 #include <linux/kthread.h>
 36 #include <linux/list.h>
 37 #include <linux/netdevice.h>
 38 #include <linux/param.h>
 39 #include <linux/printk.h>
 40 #include <linux/random.h>
 41 #include <linux/rculist.h>
 42 #include <linux/rcupdate.h>
 43 #include <linux/sched.h>
 44 #include <linux/skbuff.h>
 45 #include <linux/slab.h>
 46 #include <linux/spinlock.h>
 47 #include <linux/stddef.h>
 48 #include <linux/string.h>
 49 #include <linux/timer.h>
 50 #include <linux/wait.h>
 51 #include <linux/workqueue.h>
 52 #include <uapi/linux/batadv_packet.h>
 53 #include <uapi/linux/batman_adv.h>
 54 
 55 #include "hard-interface.h"
 56 #include "log.h"
 57 #include "netlink.h"
 58 #include "originator.h"
 59 #include "send.h"
 60 
 61 /**
 62  * BATADV_TP_DEF_TEST_LENGTH - Default test length if not specified by the user
 63  *  in milliseconds
 64  */
 65 #define BATADV_TP_DEF_TEST_LENGTH 10000
 66 
 67 /**
 68  * BATADV_TP_AWND - Advertised window by the receiver (in bytes)
 69  */
 70 #define BATADV_TP_AWND 0x20000000
 71 
 72 /**
 73  * BATADV_TP_RECV_TIMEOUT - Receiver activity timeout. If the receiver does not
 74  *  get anything for such amount of milliseconds, the connection is killed
 75  */
 76 #define BATADV_TP_RECV_TIMEOUT 1000
 77 
 78 /**
 79  * BATADV_TP_MAX_RTO - Maximum sender timeout. If the sender RTO gets beyond
 80  * such amound of milliseconds, the receiver is considered unreachable and the
 81  * connection is killed
 82  */
 83 #define BATADV_TP_MAX_RTO 30000
 84 
 85 /**
 86  * BATADV_TP_FIRST_SEQ - First seqno of each session. The number is rather high
 87  *  in order to immediately trigger a wrap around (test purposes)
 88  */
 89 #define BATADV_TP_FIRST_SEQ ((u32)-1 - 2000)
 90 
 91 /**
 92  * BATADV_TP_PLEN - length of the payload (data after the batadv_unicast header)
 93  *  to simulate
 94  */
 95 #define BATADV_TP_PLEN (BATADV_TP_PACKET_LEN - ETH_HLEN - \
 96                         sizeof(struct batadv_unicast_packet))
 97 
 98 static u8 batadv_tp_prerandom[4096] __read_mostly;
 99 
100 /**
101  * batadv_tp_session_cookie() - generate session cookie based on session ids
102  * @session: TP session identifier
103  * @icmp_uid: icmp pseudo uid of the tp session
104  *
105  * Return: 32 bit tp_meter session cookie
106  */
107 static u32 batadv_tp_session_cookie(const u8 session[2], u8 icmp_uid)
108 {
109         u32 cookie;
110 
111         cookie = icmp_uid << 16;
112         cookie |= session[0] << 8;
113         cookie |= session[1];
114 
115         return cookie;
116 }
117 
118 /**
119  * batadv_tp_cwnd() - compute the new cwnd size
120  * @base: base cwnd size value
121  * @increment: the value to add to base to get the new size
122  * @min: minumim cwnd value (usually MSS)
123  *
124  * Return the new cwnd size and ensures it does not exceed the Advertised
125  * Receiver Window size. It is wrap around safe.
126  * For details refer to Section 3.1 of RFC5681
127  *
128  * Return: new congestion window size in bytes
129  */
130 static u32 batadv_tp_cwnd(u32 base, u32 increment, u32 min)
131 {
132         u32 new_size = base + increment;
133 
134         /* check for wrap-around */
135         if (new_size < base)
136                 new_size = (u32)ULONG_MAX;
137 
138         new_size = min_t(u32, new_size, BATADV_TP_AWND);
139 
140         return max_t(u32, new_size, min);
141 }
142 
143 /**
144  * batadv_tp_updated_cwnd() - update the Congestion Windows
145  * @tp_vars: the private data of the current TP meter session
146  * @mss: maximum segment size of transmission
147  *
148  * 1) if the session is in Slow Start, the CWND has to be increased by 1
149  * MSS every unique received ACK
150  * 2) if the session is in Congestion Avoidance, the CWND has to be
151  * increased by MSS * MSS / CWND for every unique received ACK
152  */
153 static void batadv_tp_update_cwnd(struct batadv_tp_vars *tp_vars, u32 mss)
154 {
155         spin_lock_bh(&tp_vars->cwnd_lock);
156 
157         /* slow start... */
158         if (tp_vars->cwnd <= tp_vars->ss_threshold) {
159                 tp_vars->dec_cwnd = 0;
160                 tp_vars->cwnd = batadv_tp_cwnd(tp_vars->cwnd, mss, mss);
161                 spin_unlock_bh(&tp_vars->cwnd_lock);
162                 return;
163         }
164 
165         /* increment CWND at least of 1 (section 3.1 of RFC5681) */
166         tp_vars->dec_cwnd += max_t(u32, 1U << 3,
167                                    ((mss * mss) << 6) / (tp_vars->cwnd << 3));
168         if (tp_vars->dec_cwnd < (mss << 3)) {
169                 spin_unlock_bh(&tp_vars->cwnd_lock);
170                 return;
171         }
172 
173         tp_vars->cwnd = batadv_tp_cwnd(tp_vars->cwnd, mss, mss);
174         tp_vars->dec_cwnd = 0;
175 
176         spin_unlock_bh(&tp_vars->cwnd_lock);
177 }
178 
179 /**
180  * batadv_tp_update_rto() - calculate new retransmission timeout
181  * @tp_vars: the private data of the current TP meter session
182  * @new_rtt: new roundtrip time in msec
183  */
184 static void batadv_tp_update_rto(struct batadv_tp_vars *tp_vars,
185                                  u32 new_rtt)
186 {
187         long m = new_rtt;
188 
189         /* RTT update
190          * Details in Section 2.2 and 2.3 of RFC6298
191          *
192          * It's tricky to understand. Don't lose hair please.
193          * Inspired by tcp_rtt_estimator() tcp_input.c
194          */
195         if (tp_vars->srtt != 0) {
196                 m -= (tp_vars->srtt >> 3); /* m is now error in rtt est */
197                 tp_vars->srtt += m; /* rtt = 7/8 srtt + 1/8 new */
198                 if (m < 0)
199                         m = -m;
200 
201                 m -= (tp_vars->rttvar >> 2);
202                 tp_vars->rttvar += m; /* mdev ~= 3/4 rttvar + 1/4 new */
203         } else {
204                 /* first measure getting in */
205                 tp_vars->srtt = m << 3; /* take the measured time to be srtt */
206                 tp_vars->rttvar = m << 1; /* new_rtt / 2 */
207         }
208 
209         /* rto = srtt + 4 * rttvar.
210          * rttvar is scaled by 4, therefore doesn't need to be multiplied
211          */
212         tp_vars->rto = (tp_vars->srtt >> 3) + tp_vars->rttvar;
213 }
214 
215 /**
216  * batadv_tp_batctl_notify() - send client status result to client
217  * @reason: reason for tp meter session stop
218  * @dst: destination of tp_meter session
219  * @bat_priv: the bat priv with all the soft interface information
220  * @start_time: start of transmission in jiffies
221  * @total_sent: bytes acked to the receiver
222  * @cookie: cookie of tp_meter session
223  */
224 static void batadv_tp_batctl_notify(enum batadv_tp_meter_reason reason,
225                                     const u8 *dst, struct batadv_priv *bat_priv,
226                                     unsigned long start_time, u64 total_sent,
227                                     u32 cookie)
228 {
229         u32 test_time;
230         u8 result;
231         u32 total_bytes;
232 
233         if (!batadv_tp_is_error(reason)) {
234                 result = BATADV_TP_REASON_COMPLETE;
235                 test_time = jiffies_to_msecs(jiffies - start_time);
236                 total_bytes = total_sent;
237         } else {
238                 result = reason;
239                 test_time = 0;
240                 total_bytes = 0;
241         }
242 
243         batadv_netlink_tpmeter_notify(bat_priv, dst, result, test_time,
244                                       total_bytes, cookie);
245 }
246 
247 /**
248  * batadv_tp_batctl_error_notify() - send client error result to client
249  * @reason: reason for tp meter session stop
250  * @dst: destination of tp_meter session
251  * @bat_priv: the bat priv with all the soft interface information
252  * @cookie: cookie of tp_meter session
253  */
254 static void batadv_tp_batctl_error_notify(enum batadv_tp_meter_reason reason,
255                                           const u8 *dst,
256                                           struct batadv_priv *bat_priv,
257                                           u32 cookie)
258 {
259         batadv_tp_batctl_notify(reason, dst, bat_priv, 0, 0, cookie);
260 }
261 
262 /**
263  * batadv_tp_list_find() - find a tp_vars object in the global list
264  * @bat_priv: the bat priv with all the soft interface information
265  * @dst: the other endpoint MAC address to look for
266  *
267  * Look for a tp_vars object matching dst as end_point and return it after
268  * having incremented the refcounter. Return NULL is not found
269  *
270  * Return: matching tp_vars or NULL when no tp_vars with @dst was found
271  */
272 static struct batadv_tp_vars *batadv_tp_list_find(struct batadv_priv *bat_priv,
273                                                   const u8 *dst)
274 {
275         struct batadv_tp_vars *pos, *tp_vars = NULL;
276 
277         rcu_read_lock();
278         hlist_for_each_entry_rcu(pos, &bat_priv->tp_list, list) {
279                 if (!batadv_compare_eth(pos->other_end, dst))
280                         continue;
281 
282                 /* most of the time this function is invoked during the normal
283                  * process..it makes sens to pay more when the session is
284                  * finished and to speed the process up during the measurement
285                  */
286                 if (unlikely(!kref_get_unless_zero(&pos->refcount)))
287                         continue;
288 
289                 tp_vars = pos;
290                 break;
291         }
292         rcu_read_unlock();
293 
294         return tp_vars;
295 }
296 
297 /**
298  * batadv_tp_list_find_session() - find tp_vars session object in the global
299  *  list
300  * @bat_priv: the bat priv with all the soft interface information
301  * @dst: the other endpoint MAC address to look for
302  * @session: session identifier
303  *
304  * Look for a tp_vars object matching dst as end_point, session as tp meter
305  * session and return it after having incremented the refcounter. Return NULL
306  * is not found
307  *
308  * Return: matching tp_vars or NULL when no tp_vars was found
309  */
310 static struct batadv_tp_vars *
311 batadv_tp_list_find_session(struct batadv_priv *bat_priv, const u8 *dst,
312                             const u8 *session)
313 {
314         struct batadv_tp_vars *pos, *tp_vars = NULL;
315 
316         rcu_read_lock();
317         hlist_for_each_entry_rcu(pos, &bat_priv->tp_list, list) {
318                 if (!batadv_compare_eth(pos->other_end, dst))
319                         continue;
320 
321                 if (memcmp(pos->session, session, sizeof(pos->session)) != 0)
322                         continue;
323 
324                 /* most of the time this function is invoked during the normal
325                  * process..it makes sense to pay more when the session is
326                  * finished and to speed the process up during the measurement
327                  */
328                 if (unlikely(!kref_get_unless_zero(&pos->refcount)))
329                         continue;
330 
331                 tp_vars = pos;
332                 break;
333         }
334         rcu_read_unlock();
335 
336         return tp_vars;
337 }
338 
339 /**
340  * batadv_tp_vars_release() - release batadv_tp_vars from lists and queue for
341  *  free after rcu grace period
342  * @ref: kref pointer of the batadv_tp_vars
343  */
344 static void batadv_tp_vars_release(struct kref *ref)
345 {
346         struct batadv_tp_vars *tp_vars;
347         struct batadv_tp_unacked *un, *safe;
348 
349         tp_vars = container_of(ref, struct batadv_tp_vars, refcount);
350 
351         /* lock should not be needed because this object is now out of any
352          * context!
353          */
354         spin_lock_bh(&tp_vars->unacked_lock);
355         list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) {
356                 list_del(&un->list);
357                 kfree(un);
358         }
359         spin_unlock_bh(&tp_vars->unacked_lock);
360 
361         kfree_rcu(tp_vars, rcu);
362 }
363 
364 /**
365  * batadv_tp_vars_put() - decrement the batadv_tp_vars refcounter and possibly
366  *  release it
367  * @tp_vars: the private data of the current TP meter session to be free'd
368  */
369 static void batadv_tp_vars_put(struct batadv_tp_vars *tp_vars)
370 {
371         kref_put(&tp_vars->refcount, batadv_tp_vars_release);
372 }
373 
374 /**
375  * batadv_tp_sender_cleanup() - cleanup sender data and drop and timer
376  * @bat_priv: the bat priv with all the soft interface information
377  * @tp_vars: the private data of the current TP meter session to cleanup
378  */
379 static void batadv_tp_sender_cleanup(struct batadv_priv *bat_priv,
380                                      struct batadv_tp_vars *tp_vars)
381 {
382         cancel_delayed_work(&tp_vars->finish_work);
383 
384         spin_lock_bh(&tp_vars->bat_priv->tp_list_lock);
385         hlist_del_rcu(&tp_vars->list);
386         spin_unlock_bh(&tp_vars->bat_priv->tp_list_lock);
387 
388         /* drop list reference */
389         batadv_tp_vars_put(tp_vars);
390 
391         atomic_dec(&tp_vars->bat_priv->tp_num);
392 
393         /* kill the timer and remove its reference */
394         del_timer_sync(&tp_vars->timer);
395         /* the worker might have rearmed itself therefore we kill it again. Note
396          * that if the worker should run again before invoking the following
397          * del_timer(), it would not re-arm itself once again because the status
398          * is OFF now
399          */
400         del_timer(&tp_vars->timer);
401         batadv_tp_vars_put(tp_vars);
402 }
403 
404 /**
405  * batadv_tp_sender_end() - print info about ended session and inform client
406  * @bat_priv: the bat priv with all the soft interface information
407  * @tp_vars: the private data of the current TP meter session
408  */
409 static void batadv_tp_sender_end(struct batadv_priv *bat_priv,
410                                  struct batadv_tp_vars *tp_vars)
411 {
412         u32 session_cookie;
413 
414         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
415                    "Test towards %pM finished..shutting down (reason=%d)\n",
416                    tp_vars->other_end, tp_vars->reason);
417 
418         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
419                    "Last timing stats: SRTT=%ums RTTVAR=%ums RTO=%ums\n",
420                    tp_vars->srtt >> 3, tp_vars->rttvar >> 2, tp_vars->rto);
421 
422         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
423                    "Final values: cwnd=%u ss_threshold=%u\n",
424                    tp_vars->cwnd, tp_vars->ss_threshold);
425 
426         session_cookie = batadv_tp_session_cookie(tp_vars->session,
427                                                   tp_vars->icmp_uid);
428 
429         batadv_tp_batctl_notify(tp_vars->reason,
430                                 tp_vars->other_end,
431                                 bat_priv,
432                                 tp_vars->start_time,
433                                 atomic64_read(&tp_vars->tot_sent),
434                                 session_cookie);
435 }
436 
437 /**
438  * batadv_tp_sender_shutdown() - let sender thread/timer stop gracefully
439  * @tp_vars: the private data of the current TP meter session
440  * @reason: reason for tp meter session stop
441  */
442 static void batadv_tp_sender_shutdown(struct batadv_tp_vars *tp_vars,
443                                       enum batadv_tp_meter_reason reason)
444 {
445         if (!atomic_dec_and_test(&tp_vars->sending))
446                 return;
447 
448         tp_vars->reason = reason;
449 }
450 
451 /**
452  * batadv_tp_sender_finish() - stop sender session after test_length was reached
453  * @work: delayed work reference of the related tp_vars
454  */
455 static void batadv_tp_sender_finish(struct work_struct *work)
456 {
457         struct delayed_work *delayed_work;
458         struct batadv_tp_vars *tp_vars;
459 
460         delayed_work = to_delayed_work(work);
461         tp_vars = container_of(delayed_work, struct batadv_tp_vars,
462                                finish_work);
463 
464         batadv_tp_sender_shutdown(tp_vars, BATADV_TP_REASON_COMPLETE);
465 }
466 
467 /**
468  * batadv_tp_reset_sender_timer() - reschedule the sender timer
469  * @tp_vars: the private TP meter data for this session
470  *
471  * Reschedule the timer using tp_vars->rto as delay
472  */
473 static void batadv_tp_reset_sender_timer(struct batadv_tp_vars *tp_vars)
474 {
475         /* most of the time this function is invoked while normal packet
476          * reception...
477          */
478         if (unlikely(atomic_read(&tp_vars->sending) == 0))
479                 /* timer ref will be dropped in batadv_tp_sender_cleanup */
480                 return;
481 
482         mod_timer(&tp_vars->timer, jiffies + msecs_to_jiffies(tp_vars->rto));
483 }
484 
485 /**
486  * batadv_tp_sender_timeout() - timer that fires in case of packet loss
487  * @t: address to timer_list inside tp_vars
488  *
489  * If fired it means that there was packet loss.
490  * Switch to Slow Start, set the ss_threshold to half of the current cwnd and
491  * reset the cwnd to 3*MSS
492  */
493 static void batadv_tp_sender_timeout(struct timer_list *t)
494 {
495         struct batadv_tp_vars *tp_vars = from_timer(tp_vars, t, timer);
496         struct batadv_priv *bat_priv = tp_vars->bat_priv;
497 
498         if (atomic_read(&tp_vars->sending) == 0)
499                 return;
500 
501         /* if the user waited long enough...shutdown the test */
502         if (unlikely(tp_vars->rto >= BATADV_TP_MAX_RTO)) {
503                 batadv_tp_sender_shutdown(tp_vars,
504                                           BATADV_TP_REASON_DST_UNREACHABLE);
505                 return;
506         }
507 
508         /* RTO exponential backoff
509          * Details in Section 5.5 of RFC6298
510          */
511         tp_vars->rto <<= 1;
512 
513         spin_lock_bh(&tp_vars->cwnd_lock);
514 
515         tp_vars->ss_threshold = tp_vars->cwnd >> 1;
516         if (tp_vars->ss_threshold < BATADV_TP_PLEN * 2)
517                 tp_vars->ss_threshold = BATADV_TP_PLEN * 2;
518 
519         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
520                    "Meter: RTO fired during test towards %pM! cwnd=%u new ss_thr=%u, resetting last_sent to %u\n",
521                    tp_vars->other_end, tp_vars->cwnd, tp_vars->ss_threshold,
522                    atomic_read(&tp_vars->last_acked));
523 
524         tp_vars->cwnd = BATADV_TP_PLEN * 3;
525 
526         spin_unlock_bh(&tp_vars->cwnd_lock);
527 
528         /* resend the non-ACKed packets.. */
529         tp_vars->last_sent = atomic_read(&tp_vars->last_acked);
530         wake_up(&tp_vars->more_bytes);
531 
532         batadv_tp_reset_sender_timer(tp_vars);
533 }
534 
535 /**
536  * batadv_tp_fill_prerandom() - Fill buffer with prefetched random bytes
537  * @tp_vars: the private TP meter data for this session
538  * @buf: Buffer to fill with bytes
539  * @nbytes: amount of pseudorandom bytes
540  */
541 static void batadv_tp_fill_prerandom(struct batadv_tp_vars *tp_vars,
542                                      u8 *buf, size_t nbytes)
543 {
544         u32 local_offset;
545         size_t bytes_inbuf;
546         size_t to_copy;
547         size_t pos = 0;
548 
549         spin_lock_bh(&tp_vars->prerandom_lock);
550         local_offset = tp_vars->prerandom_offset;
551         tp_vars->prerandom_offset += nbytes;
552         tp_vars->prerandom_offset %= sizeof(batadv_tp_prerandom);
553         spin_unlock_bh(&tp_vars->prerandom_lock);
554 
555         while (nbytes) {
556                 local_offset %= sizeof(batadv_tp_prerandom);
557                 bytes_inbuf = sizeof(batadv_tp_prerandom) - local_offset;
558                 to_copy = min(nbytes, bytes_inbuf);
559 
560                 memcpy(&buf[pos], &batadv_tp_prerandom[local_offset], to_copy);
561                 pos += to_copy;
562                 nbytes -= to_copy;
563                 local_offset = 0;
564         }
565 }
566 
567 /**
568  * batadv_tp_send_msg() - send a single message
569  * @tp_vars: the private TP meter data for this session
570  * @src: source mac address
571  * @orig_node: the originator of the destination
572  * @seqno: sequence number of this packet
573  * @len: length of the entire packet
574  * @session: session identifier
575  * @uid: local ICMP "socket" index
576  * @timestamp: timestamp in jiffies which is replied in ack
577  *
578  * Create and send a single TP Meter message.
579  *
580  * Return: 0 on success, BATADV_TP_REASON_DST_UNREACHABLE if the destination is
581  * not reachable, BATADV_TP_REASON_MEMORY_ERROR if the packet couldn't be
582  * allocated
583  */
584 static int batadv_tp_send_msg(struct batadv_tp_vars *tp_vars, const u8 *src,
585                               struct batadv_orig_node *orig_node,
586                               u32 seqno, size_t len, const u8 *session,
587                               int uid, u32 timestamp)
588 {
589         struct batadv_icmp_tp_packet *icmp;
590         struct sk_buff *skb;
591         int r;
592         u8 *data;
593         size_t data_len;
594 
595         skb = netdev_alloc_skb_ip_align(NULL, len + ETH_HLEN);
596         if (unlikely(!skb))
597                 return BATADV_TP_REASON_MEMORY_ERROR;
598 
599         skb_reserve(skb, ETH_HLEN);
600         icmp = skb_put(skb, sizeof(*icmp));
601 
602         /* fill the icmp header */
603         ether_addr_copy(icmp->dst, orig_node->orig);
604         ether_addr_copy(icmp->orig, src);
605         icmp->version = BATADV_COMPAT_VERSION;
606         icmp->packet_type = BATADV_ICMP;
607         icmp->ttl = BATADV_TTL;
608         icmp->msg_type = BATADV_TP;
609         icmp->uid = uid;
610 
611         icmp->subtype = BATADV_TP_MSG;
612         memcpy(icmp->session, session, sizeof(icmp->session));
613         icmp->seqno = htonl(seqno);
614         icmp->timestamp = htonl(timestamp);
615 
616         data_len = len - sizeof(*icmp);
617         data = skb_put(skb, data_len);
618         batadv_tp_fill_prerandom(tp_vars, data, data_len);
619 
620         r = batadv_send_skb_to_orig(skb, orig_node, NULL);
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: %s() cannot send packets (%d)\n",
880                                    __func__, 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
910  *  sender
911  * @tp_vars: the private data of the current TP meter session
912  */
913 static void batadv_tp_start_kthread(struct batadv_tp_vars *tp_vars)
914 {
915         struct task_struct *kthread;
916         struct batadv_priv *bat_priv = tp_vars->bat_priv;
917         u32 session_cookie;
918 
919         kref_get(&tp_vars->refcount);
920         kthread = kthread_create(batadv_tp_send, tp_vars, "kbatadv_tp_meter");
921         if (IS_ERR(kthread)) {
922                 session_cookie = batadv_tp_session_cookie(tp_vars->session,
923                                                           tp_vars->icmp_uid);
924                 pr_err("batadv: cannot create tp meter kthread\n");
925                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_MEMORY_ERROR,
926                                               tp_vars->other_end,
927                                               bat_priv, session_cookie);
928 
929                 /* drop reserved reference for kthread */
930                 batadv_tp_vars_put(tp_vars);
931 
932                 /* cleanup of failed tp meter variables */
933                 batadv_tp_sender_cleanup(bat_priv, tp_vars);
934                 return;
935         }
936 
937         wake_up_process(kthread);
938 }
939 
940 /**
941  * batadv_tp_start() - start a new tp meter session
942  * @bat_priv: the bat priv with all the soft interface information
943  * @dst: the receiver MAC address
944  * @test_length: test length in milliseconds
945  * @cookie: session cookie
946  */
947 void batadv_tp_start(struct batadv_priv *bat_priv, const u8 *dst,
948                      u32 test_length, u32 *cookie)
949 {
950         struct batadv_tp_vars *tp_vars;
951         u8 session_id[2];
952         u8 icmp_uid;
953         u32 session_cookie;
954 
955         get_random_bytes(session_id, sizeof(session_id));
956         get_random_bytes(&icmp_uid, 1);
957         session_cookie = batadv_tp_session_cookie(session_id, icmp_uid);
958         *cookie = session_cookie;
959 
960         /* look for an already existing test towards this node */
961         spin_lock_bh(&bat_priv->tp_list_lock);
962         tp_vars = batadv_tp_list_find(bat_priv, dst);
963         if (tp_vars) {
964                 spin_unlock_bh(&bat_priv->tp_list_lock);
965                 batadv_tp_vars_put(tp_vars);
966                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
967                            "Meter: test to or from the same node already ongoing, aborting\n");
968                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_ALREADY_ONGOING,
969                                               dst, bat_priv, session_cookie);
970                 return;
971         }
972 
973         if (!atomic_add_unless(&bat_priv->tp_num, 1, BATADV_TP_MAX_NUM)) {
974                 spin_unlock_bh(&bat_priv->tp_list_lock);
975                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
976                            "Meter: too many ongoing sessions, aborting (SEND)\n");
977                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_TOO_MANY, dst,
978                                               bat_priv, session_cookie);
979                 return;
980         }
981 
982         tp_vars = kmalloc(sizeof(*tp_vars), GFP_ATOMIC);
983         if (!tp_vars) {
984                 spin_unlock_bh(&bat_priv->tp_list_lock);
985                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
986                            "Meter: %s cannot allocate list elements\n",
987                            __func__);
988                 batadv_tp_batctl_error_notify(BATADV_TP_REASON_MEMORY_ERROR,
989                                               dst, bat_priv, session_cookie);
990                 return;
991         }
992 
993         /* initialize tp_vars */
994         ether_addr_copy(tp_vars->other_end, dst);
995         kref_init(&tp_vars->refcount);
996         tp_vars->role = BATADV_TP_SENDER;
997         atomic_set(&tp_vars->sending, 1);
998         memcpy(tp_vars->session, session_id, sizeof(session_id));
999         tp_vars->icmp_uid = icmp_uid;
1000 
1001         tp_vars->last_sent = BATADV_TP_FIRST_SEQ;
1002         atomic_set(&tp_vars->last_acked, BATADV_TP_FIRST_SEQ);
1003         tp_vars->fast_recovery = false;
1004         tp_vars->recover = BATADV_TP_FIRST_SEQ;
1005 
1006         /* initialise the CWND to 3*MSS (Section 3.1 in RFC5681).
1007          * For batman-adv the MSS is the size of the payload received by the
1008          * soft_interface, hence its MTU
1009          */
1010         tp_vars->cwnd = BATADV_TP_PLEN * 3;
1011         /* at the beginning initialise the SS threshold to the biggest possible
1012          * window size, hence the AWND size
1013          */
1014         tp_vars->ss_threshold = BATADV_TP_AWND;
1015 
1016         /* RTO initial value is 3 seconds.
1017          * Details in Section 2.1 of RFC6298
1018          */
1019         tp_vars->rto = 1000;
1020         tp_vars->srtt = 0;
1021         tp_vars->rttvar = 0;
1022 
1023         atomic64_set(&tp_vars->tot_sent, 0);
1024 
1025         kref_get(&tp_vars->refcount);
1026         timer_setup(&tp_vars->timer, batadv_tp_sender_timeout, 0);
1027 
1028         tp_vars->bat_priv = bat_priv;
1029         tp_vars->start_time = jiffies;
1030 
1031         init_waitqueue_head(&tp_vars->more_bytes);
1032 
1033         spin_lock_init(&tp_vars->unacked_lock);
1034         INIT_LIST_HEAD(&tp_vars->unacked_list);
1035 
1036         spin_lock_init(&tp_vars->cwnd_lock);
1037 
1038         tp_vars->prerandom_offset = 0;
1039         spin_lock_init(&tp_vars->prerandom_lock);
1040 
1041         kref_get(&tp_vars->refcount);
1042         hlist_add_head_rcu(&tp_vars->list, &bat_priv->tp_list);
1043         spin_unlock_bh(&bat_priv->tp_list_lock);
1044 
1045         tp_vars->test_length = test_length;
1046         if (!tp_vars->test_length)
1047                 tp_vars->test_length = BATADV_TP_DEF_TEST_LENGTH;
1048 
1049         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1050                    "Meter: starting throughput meter towards %pM (length=%ums)\n",
1051                    dst, test_length);
1052 
1053         /* init work item for finished tp tests */
1054         INIT_DELAYED_WORK(&tp_vars->finish_work, batadv_tp_sender_finish);
1055 
1056         /* start tp kthread. This way the write() call issued from userspace can
1057          * happily return and avoid to block
1058          */
1059         batadv_tp_start_kthread(tp_vars);
1060 
1061         /* don't return reference to new tp_vars */
1062         batadv_tp_vars_put(tp_vars);
1063 }
1064 
1065 /**
1066  * batadv_tp_stop() - stop currently running tp meter session
1067  * @bat_priv: the bat priv with all the soft interface information
1068  * @dst: the receiver MAC address
1069  * @return_value: reason for tp meter session stop
1070  */
1071 void batadv_tp_stop(struct batadv_priv *bat_priv, const u8 *dst,
1072                     u8 return_value)
1073 {
1074         struct batadv_orig_node *orig_node;
1075         struct batadv_tp_vars *tp_vars;
1076 
1077         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1078                    "Meter: stopping test towards %pM\n", dst);
1079 
1080         orig_node = batadv_orig_hash_find(bat_priv, dst);
1081         if (!orig_node)
1082                 return;
1083 
1084         tp_vars = batadv_tp_list_find(bat_priv, orig_node->orig);
1085         if (!tp_vars) {
1086                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1087                            "Meter: trying to interrupt an already over connection\n");
1088                 goto out;
1089         }
1090 
1091         batadv_tp_sender_shutdown(tp_vars, return_value);
1092         batadv_tp_vars_put(tp_vars);
1093 out:
1094         batadv_orig_node_put(orig_node);
1095 }
1096 
1097 /**
1098  * batadv_tp_reset_receiver_timer() - reset the receiver shutdown timer
1099  * @tp_vars: the private data of the current TP meter session
1100  *
1101  * start the receiver shutdown timer or reset it if already started
1102  */
1103 static void batadv_tp_reset_receiver_timer(struct batadv_tp_vars *tp_vars)
1104 {
1105         mod_timer(&tp_vars->timer,
1106                   jiffies + msecs_to_jiffies(BATADV_TP_RECV_TIMEOUT));
1107 }
1108 
1109 /**
1110  * batadv_tp_receiver_shutdown() - stop a tp meter receiver when timeout is
1111  *  reached without received ack
1112  * @t: address to timer_list inside tp_vars
1113  */
1114 static void batadv_tp_receiver_shutdown(struct timer_list *t)
1115 {
1116         struct batadv_tp_vars *tp_vars = from_timer(tp_vars, t, timer);
1117         struct batadv_tp_unacked *un, *safe;
1118         struct batadv_priv *bat_priv;
1119 
1120         bat_priv = tp_vars->bat_priv;
1121 
1122         /* if there is recent activity rearm the timer */
1123         if (!batadv_has_timed_out(tp_vars->last_recv_time,
1124                                   BATADV_TP_RECV_TIMEOUT)) {
1125                 /* reset the receiver shutdown timer */
1126                 batadv_tp_reset_receiver_timer(tp_vars);
1127                 return;
1128         }
1129 
1130         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1131                    "Shutting down for inactivity (more than %dms) from %pM\n",
1132                    BATADV_TP_RECV_TIMEOUT, tp_vars->other_end);
1133 
1134         spin_lock_bh(&tp_vars->bat_priv->tp_list_lock);
1135         hlist_del_rcu(&tp_vars->list);
1136         spin_unlock_bh(&tp_vars->bat_priv->tp_list_lock);
1137 
1138         /* drop list reference */
1139         batadv_tp_vars_put(tp_vars);
1140 
1141         atomic_dec(&bat_priv->tp_num);
1142 
1143         spin_lock_bh(&tp_vars->unacked_lock);
1144         list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) {
1145                 list_del(&un->list);
1146                 kfree(un);
1147         }
1148         spin_unlock_bh(&tp_vars->unacked_lock);
1149 
1150         /* drop reference of timer */
1151         batadv_tp_vars_put(tp_vars);
1152 }
1153 
1154 /**
1155  * batadv_tp_send_ack() - send an ACK packet
1156  * @bat_priv: the bat priv with all the soft interface information
1157  * @dst: the mac address of the destination originator
1158  * @seq: the sequence number to ACK
1159  * @timestamp: the timestamp to echo back in the ACK
1160  * @session: session identifier
1161  * @socket_index: local ICMP socket identifier
1162  *
1163  * Return: 0 on success, a positive integer representing the reason of the
1164  * failure otherwise
1165  */
1166 static int batadv_tp_send_ack(struct batadv_priv *bat_priv, const u8 *dst,
1167                               u32 seq, __be32 timestamp, const u8 *session,
1168                               int socket_index)
1169 {
1170         struct batadv_hard_iface *primary_if = NULL;
1171         struct batadv_orig_node *orig_node;
1172         struct batadv_icmp_tp_packet *icmp;
1173         struct sk_buff *skb;
1174         int r, ret;
1175 
1176         orig_node = batadv_orig_hash_find(bat_priv, dst);
1177         if (unlikely(!orig_node)) {
1178                 ret = BATADV_TP_REASON_DST_UNREACHABLE;
1179                 goto out;
1180         }
1181 
1182         primary_if = batadv_primary_if_get_selected(bat_priv);
1183         if (unlikely(!primary_if)) {
1184                 ret = BATADV_TP_REASON_DST_UNREACHABLE;
1185                 goto out;
1186         }
1187 
1188         skb = netdev_alloc_skb_ip_align(NULL, sizeof(*icmp) + ETH_HLEN);
1189         if (unlikely(!skb)) {
1190                 ret = BATADV_TP_REASON_MEMORY_ERROR;
1191                 goto out;
1192         }
1193 
1194         skb_reserve(skb, ETH_HLEN);
1195         icmp = skb_put(skb, sizeof(*icmp));
1196         icmp->packet_type = BATADV_ICMP;
1197         icmp->version = BATADV_COMPAT_VERSION;
1198         icmp->ttl = BATADV_TTL;
1199         icmp->msg_type = BATADV_TP;
1200         ether_addr_copy(icmp->dst, orig_node->orig);
1201         ether_addr_copy(icmp->orig, primary_if->net_dev->dev_addr);
1202         icmp->uid = socket_index;
1203 
1204         icmp->subtype = BATADV_TP_ACK;
1205         memcpy(icmp->session, session, sizeof(icmp->session));
1206         icmp->seqno = htonl(seq);
1207         icmp->timestamp = timestamp;
1208 
1209         /* send the ack */
1210         r = batadv_send_skb_to_orig(skb, orig_node, NULL);
1211         if (unlikely(r < 0) || r == NET_XMIT_DROP) {
1212                 ret = BATADV_TP_REASON_DST_UNREACHABLE;
1213                 goto out;
1214         }
1215         ret = 0;
1216 
1217 out:
1218         if (likely(orig_node))
1219                 batadv_orig_node_put(orig_node);
1220         if (likely(primary_if))
1221                 batadv_hardif_put(primary_if);
1222 
1223         return ret;
1224 }
1225 
1226 /**
1227  * batadv_tp_handle_out_of_order() - store an out of order packet
1228  * @tp_vars: the private data of the current TP meter session
1229  * @skb: the buffer containing the received packet
1230  *
1231  * Store the out of order packet in the unacked list for late processing. This
1232  * packets are kept in this list so that they can be ACKed at once as soon as
1233  * all the previous packets have been received
1234  *
1235  * Return: true if the packed has been successfully processed, false otherwise
1236  */
1237 static bool batadv_tp_handle_out_of_order(struct batadv_tp_vars *tp_vars,
1238                                           const struct sk_buff *skb)
1239 {
1240         const struct batadv_icmp_tp_packet *icmp;
1241         struct batadv_tp_unacked *un, *new;
1242         u32 payload_len;
1243         bool added = false;
1244 
1245         new = kmalloc(sizeof(*new), GFP_ATOMIC);
1246         if (unlikely(!new))
1247                 return false;
1248 
1249         icmp = (struct batadv_icmp_tp_packet *)skb->data;
1250 
1251         new->seqno = ntohl(icmp->seqno);
1252         payload_len = skb->len - sizeof(struct batadv_unicast_packet);
1253         new->len = payload_len;
1254 
1255         spin_lock_bh(&tp_vars->unacked_lock);
1256         /* if the list is empty immediately attach this new object */
1257         if (list_empty(&tp_vars->unacked_list)) {
1258                 list_add(&new->list, &tp_vars->unacked_list);
1259                 goto out;
1260         }
1261 
1262         /* otherwise loop over the list and either drop the packet because this
1263          * is a duplicate or store it at the right position.
1264          *
1265          * The iteration is done in the reverse way because it is likely that
1266          * the last received packet (the one being processed now) has a bigger
1267          * seqno than all the others already stored.
1268          */
1269         list_for_each_entry_reverse(un, &tp_vars->unacked_list, list) {
1270                 /* check for duplicates */
1271                 if (new->seqno == un->seqno) {
1272                         if (new->len > un->len)
1273                                 un->len = new->len;
1274                         kfree(new);
1275                         added = true;
1276                         break;
1277                 }
1278 
1279                 /* look for the right position */
1280                 if (batadv_seq_before(new->seqno, un->seqno))
1281                         continue;
1282 
1283                 /* as soon as an entry having a bigger seqno is found, the new
1284                  * one is attached _after_ it. In this way the list is kept in
1285                  * ascending order
1286                  */
1287                 list_add_tail(&new->list, &un->list);
1288                 added = true;
1289                 break;
1290         }
1291 
1292         /* received packet with smallest seqno out of order; add it to front */
1293         if (!added)
1294                 list_add(&new->list, &tp_vars->unacked_list);
1295 
1296 out:
1297         spin_unlock_bh(&tp_vars->unacked_lock);
1298 
1299         return true;
1300 }
1301 
1302 /**
1303  * batadv_tp_ack_unordered() - update number received bytes in current stream
1304  *  without gaps
1305  * @tp_vars: the private data of the current TP meter session
1306  */
1307 static void batadv_tp_ack_unordered(struct batadv_tp_vars *tp_vars)
1308 {
1309         struct batadv_tp_unacked *un, *safe;
1310         u32 to_ack;
1311 
1312         /* go through the unacked packet list and possibly ACK them as
1313          * well
1314          */
1315         spin_lock_bh(&tp_vars->unacked_lock);
1316         list_for_each_entry_safe(un, safe, &tp_vars->unacked_list, list) {
1317                 /* the list is ordered, therefore it is possible to stop as soon
1318                  * there is a gap between the last acked seqno and the seqno of
1319                  * the packet under inspection
1320                  */
1321                 if (batadv_seq_before(tp_vars->last_recv, un->seqno))
1322                         break;
1323 
1324                 to_ack = un->seqno + un->len - tp_vars->last_recv;
1325 
1326                 if (batadv_seq_before(tp_vars->last_recv, un->seqno + un->len))
1327                         tp_vars->last_recv += to_ack;
1328 
1329                 list_del(&un->list);
1330                 kfree(un);
1331         }
1332         spin_unlock_bh(&tp_vars->unacked_lock);
1333 }
1334 
1335 /**
1336  * batadv_tp_init_recv() - return matching or create new receiver tp_vars
1337  * @bat_priv: the bat priv with all the soft interface information
1338  * @icmp: received icmp tp msg
1339  *
1340  * Return: corresponding tp_vars or NULL on errors
1341  */
1342 static struct batadv_tp_vars *
1343 batadv_tp_init_recv(struct batadv_priv *bat_priv,
1344                     const struct batadv_icmp_tp_packet *icmp)
1345 {
1346         struct batadv_tp_vars *tp_vars;
1347 
1348         spin_lock_bh(&bat_priv->tp_list_lock);
1349         tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig,
1350                                               icmp->session);
1351         if (tp_vars)
1352                 goto out_unlock;
1353 
1354         if (!atomic_add_unless(&bat_priv->tp_num, 1, BATADV_TP_MAX_NUM)) {
1355                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1356                            "Meter: too many ongoing sessions, aborting (RECV)\n");
1357                 goto out_unlock;
1358         }
1359 
1360         tp_vars = kmalloc(sizeof(*tp_vars), GFP_ATOMIC);
1361         if (!tp_vars)
1362                 goto out_unlock;
1363 
1364         ether_addr_copy(tp_vars->other_end, icmp->orig);
1365         tp_vars->role = BATADV_TP_RECEIVER;
1366         memcpy(tp_vars->session, icmp->session, sizeof(tp_vars->session));
1367         tp_vars->last_recv = BATADV_TP_FIRST_SEQ;
1368         tp_vars->bat_priv = bat_priv;
1369         kref_init(&tp_vars->refcount);
1370 
1371         spin_lock_init(&tp_vars->unacked_lock);
1372         INIT_LIST_HEAD(&tp_vars->unacked_list);
1373 
1374         kref_get(&tp_vars->refcount);
1375         hlist_add_head_rcu(&tp_vars->list, &bat_priv->tp_list);
1376 
1377         kref_get(&tp_vars->refcount);
1378         timer_setup(&tp_vars->timer, batadv_tp_receiver_shutdown, 0);
1379 
1380         batadv_tp_reset_receiver_timer(tp_vars);
1381 
1382 out_unlock:
1383         spin_unlock_bh(&bat_priv->tp_list_lock);
1384 
1385         return tp_vars;
1386 }
1387 
1388 /**
1389  * batadv_tp_recv_msg() - process a single data message
1390  * @bat_priv: the bat priv with all the soft interface information
1391  * @skb: the buffer containing the received packet
1392  *
1393  * Process a received TP MSG packet
1394  */
1395 static void batadv_tp_recv_msg(struct batadv_priv *bat_priv,
1396                                const struct sk_buff *skb)
1397 {
1398         const struct batadv_icmp_tp_packet *icmp;
1399         struct batadv_tp_vars *tp_vars;
1400         size_t packet_size;
1401         u32 seqno;
1402 
1403         icmp = (struct batadv_icmp_tp_packet *)skb->data;
1404 
1405         seqno = ntohl(icmp->seqno);
1406         /* check if this is the first seqno. This means that if the
1407          * first packet is lost, the tp meter does not work anymore!
1408          */
1409         if (seqno == BATADV_TP_FIRST_SEQ) {
1410                 tp_vars = batadv_tp_init_recv(bat_priv, icmp);
1411                 if (!tp_vars) {
1412                         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1413                                    "Meter: seqno != BATADV_TP_FIRST_SEQ cannot initiate connection\n");
1414                         goto out;
1415                 }
1416         } else {
1417                 tp_vars = batadv_tp_list_find_session(bat_priv, icmp->orig,
1418                                                       icmp->session);
1419                 if (!tp_vars) {
1420                         batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1421                                    "Unexpected packet from %pM!\n",
1422                                    icmp->orig);
1423                         goto out;
1424                 }
1425         }
1426 
1427         if (unlikely(tp_vars->role != BATADV_TP_RECEIVER)) {
1428                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1429                            "Meter: dropping packet: not expected (role=%u)\n",
1430                            tp_vars->role);
1431                 goto out;
1432         }
1433 
1434         tp_vars->last_recv_time = jiffies;
1435 
1436         /* if the packet is a duplicate, it may be the case that an ACK has been
1437          * lost. Resend the ACK
1438          */
1439         if (batadv_seq_before(seqno, tp_vars->last_recv))
1440                 goto send_ack;
1441 
1442         /* if the packet is out of order enqueue it */
1443         if (ntohl(icmp->seqno) != tp_vars->last_recv) {
1444                 /* exit immediately (and do not send any ACK) if the packet has
1445                  * not been enqueued correctly
1446                  */
1447                 if (!batadv_tp_handle_out_of_order(tp_vars, skb))
1448                         goto out;
1449 
1450                 /* send a duplicate ACK */
1451                 goto send_ack;
1452         }
1453 
1454         /* if everything was fine count the ACKed bytes */
1455         packet_size = skb->len - sizeof(struct batadv_unicast_packet);
1456         tp_vars->last_recv += packet_size;
1457 
1458         /* check if this ordered message filled a gap.... */
1459         batadv_tp_ack_unordered(tp_vars);
1460 
1461 send_ack:
1462         /* send the ACK. If the received packet was out of order, the ACK that
1463          * is going to be sent is a duplicate (the sender will count them and
1464          * possibly enter Fast Retransmit as soon as it has reached 3)
1465          */
1466         batadv_tp_send_ack(bat_priv, icmp->orig, tp_vars->last_recv,
1467                            icmp->timestamp, icmp->session, icmp->uid);
1468 out:
1469         if (likely(tp_vars))
1470                 batadv_tp_vars_put(tp_vars);
1471 }
1472 
1473 /**
1474  * batadv_tp_meter_recv() - main TP Meter receiving function
1475  * @bat_priv: the bat priv with all the soft interface information
1476  * @skb: the buffer containing the received packet
1477  */
1478 void batadv_tp_meter_recv(struct batadv_priv *bat_priv, struct sk_buff *skb)
1479 {
1480         struct batadv_icmp_tp_packet *icmp;
1481 
1482         icmp = (struct batadv_icmp_tp_packet *)skb->data;
1483 
1484         switch (icmp->subtype) {
1485         case BATADV_TP_MSG:
1486                 batadv_tp_recv_msg(bat_priv, skb);
1487                 break;
1488         case BATADV_TP_ACK:
1489                 batadv_tp_recv_ack(bat_priv, skb);
1490                 break;
1491         default:
1492                 batadv_dbg(BATADV_DBG_TP_METER, bat_priv,
1493                            "Received unknown TP Metric packet type %u\n",
1494                            icmp->subtype);
1495         }
1496         consume_skb(skb);
1497 }
1498 
1499 /**
1500  * batadv_tp_meter_init() - initialize global tp_meter structures
1501  */
1502 void __init batadv_tp_meter_init(void)
1503 {
1504         get_random_bytes(batadv_tp_prerandom, sizeof(batadv_tp_prerandom));
1505 }
1506 

~ [ 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