Loading...
1/* -*- mode: c; c-basic-offset: 8; -*-
2 *
3 * vim: noexpandtab sw=8 ts=8 sts=0:
4 *
5 * Copyright (C) 2004 Oracle. All rights reserved.
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public
18 * License along with this program; if not, write to the
19 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 * Boston, MA 021110-1307, USA.
21 *
22 * ----
23 *
24 * Callers for this were originally written against a very simple synchronus
25 * API. This implementation reflects those simple callers. Some day I'm sure
26 * we'll need to move to a more robust posting/callback mechanism.
27 *
28 * Transmit calls pass in kernel virtual addresses and block copying this into
29 * the socket's tx buffers via a usual blocking sendmsg. They'll block waiting
30 * for a failed socket to timeout. TX callers can also pass in a poniter to an
31 * 'int' which gets filled with an errno off the wire in response to the
32 * message they send.
33 *
34 * Handlers for unsolicited messages are registered. Each socket has a page
35 * that incoming data is copied into. First the header, then the data.
36 * Handlers are called from only one thread with a reference to this per-socket
37 * page. This page is destroyed after the handler call, so it can't be
38 * referenced beyond the call. Handlers may block but are discouraged from
39 * doing so.
40 *
41 * Any framing errors (bad magic, large payload lengths) close a connection.
42 *
43 * Our sock_container holds the state we associate with a socket. It's current
44 * framing state is held there as well as the refcounting we do around when it
45 * is safe to tear down the socket. The socket is only finally torn down from
46 * the container when the container loses all of its references -- so as long
47 * as you hold a ref on the container you can trust that the socket is valid
48 * for use with kernel socket APIs.
49 *
50 * Connections are initiated between a pair of nodes when the node with the
51 * higher node number gets a heartbeat callback which indicates that the lower
52 * numbered node has started heartbeating. The lower numbered node is passive
53 * and only accepts the connection if the higher numbered node is heartbeating.
54 */
55
56#include <linux/kernel.h>
57#include <linux/jiffies.h>
58#include <linux/slab.h>
59#include <linux/idr.h>
60#include <linux/kref.h>
61#include <linux/net.h>
62#include <linux/export.h>
63#include <linux/uaccess.h>
64#include <net/tcp.h>
65
66
67#include "heartbeat.h"
68#include "tcp.h"
69#include "nodemanager.h"
70#define MLOG_MASK_PREFIX ML_TCP
71#include "masklog.h"
72
73#include "tcp_internal.h"
74
75#define SC_NODEF_FMT "node %s (num %u) at %pI4:%u"
76
77/*
78 * In the following two log macros, the whitespace after the ',' just
79 * before ##args is intentional. Otherwise, gcc 2.95 will eat the
80 * previous token if args expands to nothing.
81 */
82#define msglog(hdr, fmt, args...) do { \
83 typeof(hdr) __hdr = (hdr); \
84 mlog(ML_MSG, "[mag %u len %u typ %u stat %d sys_stat %d " \
85 "key %08x num %u] " fmt, \
86 be16_to_cpu(__hdr->magic), be16_to_cpu(__hdr->data_len), \
87 be16_to_cpu(__hdr->msg_type), be32_to_cpu(__hdr->status), \
88 be32_to_cpu(__hdr->sys_status), be32_to_cpu(__hdr->key), \
89 be32_to_cpu(__hdr->msg_num) , ##args); \
90} while (0)
91
92#define sclog(sc, fmt, args...) do { \
93 typeof(sc) __sc = (sc); \
94 mlog(ML_SOCKET, "[sc %p refs %d sock %p node %u page %p " \
95 "pg_off %zu] " fmt, __sc, \
96 atomic_read(&__sc->sc_kref.refcount), __sc->sc_sock, \
97 __sc->sc_node->nd_num, __sc->sc_page, __sc->sc_page_off , \
98 ##args); \
99} while (0)
100
101static DEFINE_RWLOCK(r2net_handler_lock);
102static struct rb_root r2net_handler_tree = RB_ROOT;
103
104static struct r2net_node r2net_nodes[R2NM_MAX_NODES];
105
106/* XXX someday we'll need better accounting */
107static struct socket *r2net_listen_sock;
108
109/*
110 * listen work is only queued by the listening socket callbacks on the
111 * r2net_wq. teardown detaches the callbacks before destroying the workqueue.
112 * quorum work is queued as sock containers are shutdown.. stop_listening
113 * tears down all the node's sock containers, preventing future shutdowns
114 * and queued quorum work, before canceling delayed quorum work and
115 * destroying the work queue.
116 */
117static struct workqueue_struct *r2net_wq;
118static struct work_struct r2net_listen_work;
119
120static struct r2hb_callback_func r2net_hb_up, r2net_hb_down;
121#define R2NET_HB_PRI 0x1
122
123static struct r2net_handshake *r2net_hand;
124static struct r2net_msg *r2net_keep_req, *r2net_keep_resp;
125
126static int r2net_sys_err_translations[R2NET_ERR_MAX] = {
127 [R2NET_ERR_NONE] = 0,
128 [R2NET_ERR_NO_HNDLR] = -ENOPROTOOPT,
129 [R2NET_ERR_OVERFLOW] = -EOVERFLOW,
130 [R2NET_ERR_DIED] = -EHOSTDOWN,};
131
132/* can't quite avoid *all* internal declarations :/ */
133static void r2net_sc_connect_completed(struct work_struct *work);
134static void r2net_rx_until_empty(struct work_struct *work);
135static void r2net_shutdown_sc(struct work_struct *work);
136static void r2net_listen_data_ready(struct sock *sk, int bytes);
137static void r2net_sc_send_keep_req(struct work_struct *work);
138static void r2net_idle_timer(unsigned long data);
139static void r2net_sc_postpone_idle(struct r2net_sock_container *sc);
140static void r2net_sc_reset_idle_timer(struct r2net_sock_container *sc);
141
142#ifdef CONFIG_DEBUG_FS
143static void r2net_init_nst(struct r2net_send_tracking *nst, u32 msgtype,
144 u32 msgkey, struct task_struct *task, u8 node)
145{
146 INIT_LIST_HEAD(&nst->st_net_debug_item);
147 nst->st_task = task;
148 nst->st_msg_type = msgtype;
149 nst->st_msg_key = msgkey;
150 nst->st_node = node;
151}
152
153static inline void r2net_set_nst_sock_time(struct r2net_send_tracking *nst)
154{
155 nst->st_sock_time = ktime_get();
156}
157
158static inline void r2net_set_nst_send_time(struct r2net_send_tracking *nst)
159{
160 nst->st_send_time = ktime_get();
161}
162
163static inline void r2net_set_nst_status_time(struct r2net_send_tracking *nst)
164{
165 nst->st_status_time = ktime_get();
166}
167
168static inline void r2net_set_nst_sock_container(struct r2net_send_tracking *nst,
169 struct r2net_sock_container *sc)
170{
171 nst->st_sc = sc;
172}
173
174static inline void r2net_set_nst_msg_id(struct r2net_send_tracking *nst,
175 u32 msg_id)
176{
177 nst->st_id = msg_id;
178}
179
180static inline void r2net_set_sock_timer(struct r2net_sock_container *sc)
181{
182 sc->sc_tv_timer = ktime_get();
183}
184
185static inline void r2net_set_data_ready_time(struct r2net_sock_container *sc)
186{
187 sc->sc_tv_data_ready = ktime_get();
188}
189
190static inline void r2net_set_advance_start_time(struct r2net_sock_container *sc)
191{
192 sc->sc_tv_advance_start = ktime_get();
193}
194
195static inline void r2net_set_advance_stop_time(struct r2net_sock_container *sc)
196{
197 sc->sc_tv_advance_stop = ktime_get();
198}
199
200static inline void r2net_set_func_start_time(struct r2net_sock_container *sc)
201{
202 sc->sc_tv_func_start = ktime_get();
203}
204
205static inline void r2net_set_func_stop_time(struct r2net_sock_container *sc)
206{
207 sc->sc_tv_func_stop = ktime_get();
208}
209
210#else /* CONFIG_DEBUG_FS */
211# define r2net_init_nst(a, b, c, d, e)
212# define r2net_set_nst_sock_time(a)
213# define r2net_set_nst_send_time(a)
214# define r2net_set_nst_status_time(a)
215# define r2net_set_nst_sock_container(a, b)
216# define r2net_set_nst_msg_id(a, b)
217# define r2net_set_sock_timer(a)
218# define r2net_set_data_ready_time(a)
219# define r2net_set_advance_start_time(a)
220# define r2net_set_advance_stop_time(a)
221# define r2net_set_func_start_time(a)
222# define r2net_set_func_stop_time(a)
223#endif /* CONFIG_DEBUG_FS */
224
225#ifdef CONFIG_RAMSTER_FS_STATS
226static ktime_t r2net_get_func_run_time(struct r2net_sock_container *sc)
227{
228 return ktime_sub(sc->sc_tv_func_stop, sc->sc_tv_func_start);
229}
230
231static void r2net_update_send_stats(struct r2net_send_tracking *nst,
232 struct r2net_sock_container *sc)
233{
234 sc->sc_tv_status_total = ktime_add(sc->sc_tv_status_total,
235 ktime_sub(ktime_get(),
236 nst->st_status_time));
237 sc->sc_tv_send_total = ktime_add(sc->sc_tv_send_total,
238 ktime_sub(nst->st_status_time,
239 nst->st_send_time));
240 sc->sc_tv_acquiry_total = ktime_add(sc->sc_tv_acquiry_total,
241 ktime_sub(nst->st_send_time,
242 nst->st_sock_time));
243 sc->sc_send_count++;
244}
245
246static void r2net_update_recv_stats(struct r2net_sock_container *sc)
247{
248 sc->sc_tv_process_total = ktime_add(sc->sc_tv_process_total,
249 r2net_get_func_run_time(sc));
250 sc->sc_recv_count++;
251}
252
253#else
254
255# define r2net_update_send_stats(a, b)
256
257# define r2net_update_recv_stats(sc)
258
259#endif /* CONFIG_RAMSTER_FS_STATS */
260
261static inline int r2net_reconnect_delay(void)
262{
263 return r2nm_single_cluster->cl_reconnect_delay_ms;
264}
265
266static inline int r2net_keepalive_delay(void)
267{
268 return r2nm_single_cluster->cl_keepalive_delay_ms;
269}
270
271static inline int r2net_idle_timeout(void)
272{
273 return r2nm_single_cluster->cl_idle_timeout_ms;
274}
275
276static inline int r2net_sys_err_to_errno(enum r2net_system_error err)
277{
278 int trans;
279 BUG_ON(err >= R2NET_ERR_MAX);
280 trans = r2net_sys_err_translations[err];
281
282 /* Just in case we mess up the translation table above */
283 BUG_ON(err != R2NET_ERR_NONE && trans == 0);
284 return trans;
285}
286
287struct r2net_node *r2net_nn_from_num(u8 node_num)
288{
289 BUG_ON(node_num >= ARRAY_SIZE(r2net_nodes));
290 return &r2net_nodes[node_num];
291}
292
293static u8 r2net_num_from_nn(struct r2net_node *nn)
294{
295 BUG_ON(nn == NULL);
296 return nn - r2net_nodes;
297}
298
299/* ------------------------------------------------------------ */
300
301static int r2net_prep_nsw(struct r2net_node *nn, struct r2net_status_wait *nsw)
302{
303 int ret = 0;
304
305 do {
306 if (!idr_pre_get(&nn->nn_status_idr, GFP_ATOMIC)) {
307 ret = -EAGAIN;
308 break;
309 }
310 spin_lock(&nn->nn_lock);
311 ret = idr_get_new(&nn->nn_status_idr, nsw, &nsw->ns_id);
312 if (ret == 0)
313 list_add_tail(&nsw->ns_node_item,
314 &nn->nn_status_list);
315 spin_unlock(&nn->nn_lock);
316 } while (ret == -EAGAIN);
317
318 if (ret == 0) {
319 init_waitqueue_head(&nsw->ns_wq);
320 nsw->ns_sys_status = R2NET_ERR_NONE;
321 nsw->ns_status = 0;
322 }
323
324 return ret;
325}
326
327static void r2net_complete_nsw_locked(struct r2net_node *nn,
328 struct r2net_status_wait *nsw,
329 enum r2net_system_error sys_status,
330 s32 status)
331{
332 assert_spin_locked(&nn->nn_lock);
333
334 if (!list_empty(&nsw->ns_node_item)) {
335 list_del_init(&nsw->ns_node_item);
336 nsw->ns_sys_status = sys_status;
337 nsw->ns_status = status;
338 idr_remove(&nn->nn_status_idr, nsw->ns_id);
339 wake_up(&nsw->ns_wq);
340 }
341}
342
343static void r2net_complete_nsw(struct r2net_node *nn,
344 struct r2net_status_wait *nsw,
345 u64 id, enum r2net_system_error sys_status,
346 s32 status)
347{
348 spin_lock(&nn->nn_lock);
349 if (nsw == NULL) {
350 if (id > INT_MAX)
351 goto out;
352
353 nsw = idr_find(&nn->nn_status_idr, id);
354 if (nsw == NULL)
355 goto out;
356 }
357
358 r2net_complete_nsw_locked(nn, nsw, sys_status, status);
359
360out:
361 spin_unlock(&nn->nn_lock);
362 return;
363}
364
365static void r2net_complete_nodes_nsw(struct r2net_node *nn)
366{
367 struct r2net_status_wait *nsw, *tmp;
368 unsigned int num_kills = 0;
369
370 assert_spin_locked(&nn->nn_lock);
371
372 list_for_each_entry_safe(nsw, tmp, &nn->nn_status_list, ns_node_item) {
373 r2net_complete_nsw_locked(nn, nsw, R2NET_ERR_DIED, 0);
374 num_kills++;
375 }
376
377 mlog(0, "completed %d messages for node %u\n", num_kills,
378 r2net_num_from_nn(nn));
379}
380
381static int r2net_nsw_completed(struct r2net_node *nn,
382 struct r2net_status_wait *nsw)
383{
384 int completed;
385 spin_lock(&nn->nn_lock);
386 completed = list_empty(&nsw->ns_node_item);
387 spin_unlock(&nn->nn_lock);
388 return completed;
389}
390
391/* ------------------------------------------------------------ */
392
393static void sc_kref_release(struct kref *kref)
394{
395 struct r2net_sock_container *sc = container_of(kref,
396 struct r2net_sock_container, sc_kref);
397 BUG_ON(timer_pending(&sc->sc_idle_timeout));
398
399 sclog(sc, "releasing\n");
400
401 if (sc->sc_sock) {
402 sock_release(sc->sc_sock);
403 sc->sc_sock = NULL;
404 }
405
406 r2nm_undepend_item(&sc->sc_node->nd_item);
407 r2nm_node_put(sc->sc_node);
408 sc->sc_node = NULL;
409
410 r2net_debug_del_sc(sc);
411 kfree(sc);
412}
413
414static void sc_put(struct r2net_sock_container *sc)
415{
416 sclog(sc, "put\n");
417 kref_put(&sc->sc_kref, sc_kref_release);
418}
419static void sc_get(struct r2net_sock_container *sc)
420{
421 sclog(sc, "get\n");
422 kref_get(&sc->sc_kref);
423}
424static struct r2net_sock_container *sc_alloc(struct r2nm_node *node)
425{
426 struct r2net_sock_container *sc, *ret = NULL;
427 struct page *page = NULL;
428 int status = 0;
429
430 page = alloc_page(GFP_NOFS);
431 sc = kzalloc(sizeof(*sc), GFP_NOFS);
432 if (sc == NULL || page == NULL)
433 goto out;
434
435 kref_init(&sc->sc_kref);
436 r2nm_node_get(node);
437 sc->sc_node = node;
438
439 /* pin the node item of the remote node */
440 status = r2nm_depend_item(&node->nd_item);
441 if (status) {
442 mlog_errno(status);
443 r2nm_node_put(node);
444 goto out;
445 }
446 INIT_WORK(&sc->sc_connect_work, r2net_sc_connect_completed);
447 INIT_WORK(&sc->sc_rx_work, r2net_rx_until_empty);
448 INIT_WORK(&sc->sc_shutdown_work, r2net_shutdown_sc);
449 INIT_DELAYED_WORK(&sc->sc_keepalive_work, r2net_sc_send_keep_req);
450
451 init_timer(&sc->sc_idle_timeout);
452 sc->sc_idle_timeout.function = r2net_idle_timer;
453 sc->sc_idle_timeout.data = (unsigned long)sc;
454
455 sclog(sc, "alloced\n");
456
457 ret = sc;
458 sc->sc_page = page;
459 r2net_debug_add_sc(sc);
460 sc = NULL;
461 page = NULL;
462
463out:
464 if (page)
465 __free_page(page);
466 kfree(sc);
467
468 return ret;
469}
470
471/* ------------------------------------------------------------ */
472
473static void r2net_sc_queue_work(struct r2net_sock_container *sc,
474 struct work_struct *work)
475{
476 sc_get(sc);
477 if (!queue_work(r2net_wq, work))
478 sc_put(sc);
479}
480static void r2net_sc_queue_delayed_work(struct r2net_sock_container *sc,
481 struct delayed_work *work,
482 int delay)
483{
484 sc_get(sc);
485 if (!queue_delayed_work(r2net_wq, work, delay))
486 sc_put(sc);
487}
488static void r2net_sc_cancel_delayed_work(struct r2net_sock_container *sc,
489 struct delayed_work *work)
490{
491 if (cancel_delayed_work(work))
492 sc_put(sc);
493}
494
495static atomic_t r2net_connected_peers = ATOMIC_INIT(0);
496
497int r2net_num_connected_peers(void)
498{
499 return atomic_read(&r2net_connected_peers);
500}
501
502static void r2net_set_nn_state(struct r2net_node *nn,
503 struct r2net_sock_container *sc,
504 unsigned valid, int err)
505{
506 int was_valid = nn->nn_sc_valid;
507 int was_err = nn->nn_persistent_error;
508 struct r2net_sock_container *old_sc = nn->nn_sc;
509
510 assert_spin_locked(&nn->nn_lock);
511
512 if (old_sc && !sc)
513 atomic_dec(&r2net_connected_peers);
514 else if (!old_sc && sc)
515 atomic_inc(&r2net_connected_peers);
516
517 /* the node num comparison and single connect/accept path should stop
518 * an non-null sc from being overwritten with another */
519 BUG_ON(sc && nn->nn_sc && nn->nn_sc != sc);
520 mlog_bug_on_msg(err && valid, "err %d valid %u\n", err, valid);
521 mlog_bug_on_msg(valid && !sc, "valid %u sc %p\n", valid, sc);
522
523 if (was_valid && !valid && err == 0)
524 err = -ENOTCONN;
525
526 mlog(ML_CONN, "node %u sc: %p -> %p, valid %u -> %u, err %d -> %d\n",
527 r2net_num_from_nn(nn), nn->nn_sc, sc, nn->nn_sc_valid, valid,
528 nn->nn_persistent_error, err);
529
530 nn->nn_sc = sc;
531 nn->nn_sc_valid = valid ? 1 : 0;
532 nn->nn_persistent_error = err;
533
534 /* mirrors r2net_tx_can_proceed() */
535 if (nn->nn_persistent_error || nn->nn_sc_valid)
536 wake_up(&nn->nn_sc_wq);
537
538 if (!was_err && nn->nn_persistent_error) {
539 queue_delayed_work(r2net_wq, &nn->nn_still_up,
540 msecs_to_jiffies(R2NET_QUORUM_DELAY_MS));
541 }
542
543 if (was_valid && !valid) {
544 printk(KERN_NOTICE "ramster: No longer connected to "
545 SC_NODEF_FMT "\n",
546 old_sc->sc_node->nd_name, old_sc->sc_node->nd_num,
547 &old_sc->sc_node->nd_ipv4_address,
548 ntohs(old_sc->sc_node->nd_ipv4_port));
549 r2net_complete_nodes_nsw(nn);
550 }
551
552 if (!was_valid && valid) {
553 cancel_delayed_work(&nn->nn_connect_expired);
554 printk(KERN_NOTICE "ramster: %s " SC_NODEF_FMT "\n",
555 r2nm_this_node() > sc->sc_node->nd_num ?
556 "Connected to" : "Accepted connection from",
557 sc->sc_node->nd_name, sc->sc_node->nd_num,
558 &sc->sc_node->nd_ipv4_address,
559 ntohs(sc->sc_node->nd_ipv4_port));
560 }
561
562 /* trigger the connecting worker func as long as we're not valid,
563 * it will back off if it shouldn't connect. This can be called
564 * from node config teardown and so needs to be careful about
565 * the work queue actually being up. */
566 if (!valid && r2net_wq) {
567 unsigned long delay;
568 /* delay if we're within a RECONNECT_DELAY of the
569 * last attempt */
570 delay = (nn->nn_last_connect_attempt +
571 msecs_to_jiffies(r2net_reconnect_delay()))
572 - jiffies;
573 if (delay > msecs_to_jiffies(r2net_reconnect_delay()))
574 delay = 0;
575 mlog(ML_CONN, "queueing conn attempt in %lu jiffies\n", delay);
576 queue_delayed_work(r2net_wq, &nn->nn_connect_work, delay);
577
578 /*
579 * Delay the expired work after idle timeout.
580 *
581 * We might have lots of failed connection attempts that run
582 * through here but we only cancel the connect_expired work when
583 * a connection attempt succeeds. So only the first enqueue of
584 * the connect_expired work will do anything. The rest will see
585 * that it's already queued and do nothing.
586 */
587 delay += msecs_to_jiffies(r2net_idle_timeout());
588 queue_delayed_work(r2net_wq, &nn->nn_connect_expired, delay);
589 }
590
591 /* keep track of the nn's sc ref for the caller */
592 if ((old_sc == NULL) && sc)
593 sc_get(sc);
594 if (old_sc && (old_sc != sc)) {
595 r2net_sc_queue_work(old_sc, &old_sc->sc_shutdown_work);
596 sc_put(old_sc);
597 }
598}
599
600/* see r2net_register_callbacks() */
601static void r2net_data_ready(struct sock *sk, int bytes)
602{
603 void (*ready)(struct sock *sk, int bytes);
604
605 read_lock(&sk->sk_callback_lock);
606 if (sk->sk_user_data) {
607 struct r2net_sock_container *sc = sk->sk_user_data;
608 sclog(sc, "data_ready hit\n");
609 r2net_set_data_ready_time(sc);
610 r2net_sc_queue_work(sc, &sc->sc_rx_work);
611 ready = sc->sc_data_ready;
612 } else {
613 ready = sk->sk_data_ready;
614 }
615 read_unlock(&sk->sk_callback_lock);
616
617 ready(sk, bytes);
618}
619
620/* see r2net_register_callbacks() */
621static void r2net_state_change(struct sock *sk)
622{
623 void (*state_change)(struct sock *sk);
624 struct r2net_sock_container *sc;
625
626 read_lock(&sk->sk_callback_lock);
627 sc = sk->sk_user_data;
628 if (sc == NULL) {
629 state_change = sk->sk_state_change;
630 goto out;
631 }
632
633 sclog(sc, "state_change to %d\n", sk->sk_state);
634
635 state_change = sc->sc_state_change;
636
637 switch (sk->sk_state) {
638
639 /* ignore connecting sockets as they make progress */
640 case TCP_SYN_SENT:
641 case TCP_SYN_RECV:
642 break;
643 case TCP_ESTABLISHED:
644 r2net_sc_queue_work(sc, &sc->sc_connect_work);
645 break;
646 default:
647 printk(KERN_INFO "ramster: Connection to "
648 SC_NODEF_FMT " shutdown, state %d\n",
649 sc->sc_node->nd_name, sc->sc_node->nd_num,
650 &sc->sc_node->nd_ipv4_address,
651 ntohs(sc->sc_node->nd_ipv4_port), sk->sk_state);
652 r2net_sc_queue_work(sc, &sc->sc_shutdown_work);
653 break;
654
655 }
656out:
657 read_unlock(&sk->sk_callback_lock);
658 state_change(sk);
659}
660
661/*
662 * we register callbacks so we can queue work on events before calling
663 * the original callbacks. our callbacks are careful to test user_data
664 * to discover when they've reaced with r2net_unregister_callbacks().
665 */
666static void r2net_register_callbacks(struct sock *sk,
667 struct r2net_sock_container *sc)
668{
669 write_lock_bh(&sk->sk_callback_lock);
670
671 /* accepted sockets inherit the old listen socket data ready */
672 if (sk->sk_data_ready == r2net_listen_data_ready) {
673 sk->sk_data_ready = sk->sk_user_data;
674 sk->sk_user_data = NULL;
675 }
676
677 BUG_ON(sk->sk_user_data != NULL);
678 sk->sk_user_data = sc;
679 sc_get(sc);
680
681 sc->sc_data_ready = sk->sk_data_ready;
682 sc->sc_state_change = sk->sk_state_change;
683 sk->sk_data_ready = r2net_data_ready;
684 sk->sk_state_change = r2net_state_change;
685
686 mutex_init(&sc->sc_send_lock);
687
688 write_unlock_bh(&sk->sk_callback_lock);
689}
690
691static int r2net_unregister_callbacks(struct sock *sk,
692 struct r2net_sock_container *sc)
693{
694 int ret = 0;
695
696 write_lock_bh(&sk->sk_callback_lock);
697 if (sk->sk_user_data == sc) {
698 ret = 1;
699 sk->sk_user_data = NULL;
700 sk->sk_data_ready = sc->sc_data_ready;
701 sk->sk_state_change = sc->sc_state_change;
702 }
703 write_unlock_bh(&sk->sk_callback_lock);
704
705 return ret;
706}
707
708/*
709 * this is a little helper that is called by callers who have seen a problem
710 * with an sc and want to detach it from the nn if someone already hasn't beat
711 * them to it. if an error is given then the shutdown will be persistent
712 * and pending transmits will be canceled.
713 */
714static void r2net_ensure_shutdown(struct r2net_node *nn,
715 struct r2net_sock_container *sc,
716 int err)
717{
718 spin_lock(&nn->nn_lock);
719 if (nn->nn_sc == sc)
720 r2net_set_nn_state(nn, NULL, 0, err);
721 spin_unlock(&nn->nn_lock);
722}
723
724/*
725 * This work queue function performs the blocking parts of socket shutdown. A
726 * few paths lead here. set_nn_state will trigger this callback if it sees an
727 * sc detached from the nn. state_change will also trigger this callback
728 * directly when it sees errors. In that case we need to call set_nn_state
729 * ourselves as state_change couldn't get the nn_lock and call set_nn_state
730 * itself.
731 */
732static void r2net_shutdown_sc(struct work_struct *work)
733{
734 struct r2net_sock_container *sc =
735 container_of(work, struct r2net_sock_container,
736 sc_shutdown_work);
737 struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
738
739 sclog(sc, "shutting down\n");
740
741 /* drop the callbacks ref and call shutdown only once */
742 if (r2net_unregister_callbacks(sc->sc_sock->sk, sc)) {
743 /* we shouldn't flush as we're in the thread, the
744 * races with pending sc work structs are harmless */
745 del_timer_sync(&sc->sc_idle_timeout);
746 r2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
747 sc_put(sc);
748 kernel_sock_shutdown(sc->sc_sock, SHUT_RDWR);
749 }
750
751 /* not fatal so failed connects before the other guy has our
752 * heartbeat can be retried */
753 r2net_ensure_shutdown(nn, sc, 0);
754 sc_put(sc);
755}
756
757/* ------------------------------------------------------------ */
758
759static int r2net_handler_cmp(struct r2net_msg_handler *nmh, u32 msg_type,
760 u32 key)
761{
762 int ret = memcmp(&nmh->nh_key, &key, sizeof(key));
763
764 if (ret == 0)
765 ret = memcmp(&nmh->nh_msg_type, &msg_type, sizeof(msg_type));
766
767 return ret;
768}
769
770static struct r2net_msg_handler *
771r2net_handler_tree_lookup(u32 msg_type, u32 key, struct rb_node ***ret_p,
772 struct rb_node **ret_parent)
773{
774 struct rb_node **p = &r2net_handler_tree.rb_node;
775 struct rb_node *parent = NULL;
776 struct r2net_msg_handler *nmh, *ret = NULL;
777 int cmp;
778
779 while (*p) {
780 parent = *p;
781 nmh = rb_entry(parent, struct r2net_msg_handler, nh_node);
782 cmp = r2net_handler_cmp(nmh, msg_type, key);
783
784 if (cmp < 0)
785 p = &(*p)->rb_left;
786 else if (cmp > 0)
787 p = &(*p)->rb_right;
788 else {
789 ret = nmh;
790 break;
791 }
792 }
793
794 if (ret_p != NULL)
795 *ret_p = p;
796 if (ret_parent != NULL)
797 *ret_parent = parent;
798
799 return ret;
800}
801
802static void r2net_handler_kref_release(struct kref *kref)
803{
804 struct r2net_msg_handler *nmh;
805 nmh = container_of(kref, struct r2net_msg_handler, nh_kref);
806
807 kfree(nmh);
808}
809
810static void r2net_handler_put(struct r2net_msg_handler *nmh)
811{
812 kref_put(&nmh->nh_kref, r2net_handler_kref_release);
813}
814
815/* max_len is protection for the handler func. incoming messages won't
816 * be given to the handler if their payload is longer than the max. */
817int r2net_register_handler(u32 msg_type, u32 key, u32 max_len,
818 r2net_msg_handler_func *func, void *data,
819 r2net_post_msg_handler_func *post_func,
820 struct list_head *unreg_list)
821{
822 struct r2net_msg_handler *nmh = NULL;
823 struct rb_node **p, *parent;
824 int ret = 0;
825
826 if (max_len > R2NET_MAX_PAYLOAD_BYTES) {
827 mlog(0, "max_len for message handler out of range: %u\n",
828 max_len);
829 ret = -EINVAL;
830 goto out;
831 }
832
833 if (!msg_type) {
834 mlog(0, "no message type provided: %u, %p\n", msg_type, func);
835 ret = -EINVAL;
836 goto out;
837
838 }
839 if (!func) {
840 mlog(0, "no message handler provided: %u, %p\n",
841 msg_type, func);
842 ret = -EINVAL;
843 goto out;
844 }
845
846 nmh = kzalloc(sizeof(struct r2net_msg_handler), GFP_NOFS);
847 if (nmh == NULL) {
848 ret = -ENOMEM;
849 goto out;
850 }
851
852 nmh->nh_func = func;
853 nmh->nh_func_data = data;
854 nmh->nh_post_func = post_func;
855 nmh->nh_msg_type = msg_type;
856 nmh->nh_max_len = max_len;
857 nmh->nh_key = key;
858 /* the tree and list get this ref.. they're both removed in
859 * unregister when this ref is dropped */
860 kref_init(&nmh->nh_kref);
861 INIT_LIST_HEAD(&nmh->nh_unregister_item);
862
863 write_lock(&r2net_handler_lock);
864 if (r2net_handler_tree_lookup(msg_type, key, &p, &parent))
865 ret = -EEXIST;
866 else {
867 rb_link_node(&nmh->nh_node, parent, p);
868 rb_insert_color(&nmh->nh_node, &r2net_handler_tree);
869 list_add_tail(&nmh->nh_unregister_item, unreg_list);
870
871 mlog(ML_TCP, "registered handler func %p type %u key %08x\n",
872 func, msg_type, key);
873 /* we've had some trouble with handlers seemingly vanishing. */
874 mlog_bug_on_msg(r2net_handler_tree_lookup(msg_type, key, &p,
875 &parent) == NULL,
876 "couldn't find handler we *just* registered "
877 "for type %u key %08x\n", msg_type, key);
878 }
879 write_unlock(&r2net_handler_lock);
880 if (ret)
881 goto out;
882
883out:
884 if (ret)
885 kfree(nmh);
886
887 return ret;
888}
889EXPORT_SYMBOL_GPL(r2net_register_handler);
890
891void r2net_unregister_handler_list(struct list_head *list)
892{
893 struct r2net_msg_handler *nmh, *n;
894
895 write_lock(&r2net_handler_lock);
896 list_for_each_entry_safe(nmh, n, list, nh_unregister_item) {
897 mlog(ML_TCP, "unregistering handler func %p type %u key %08x\n",
898 nmh->nh_func, nmh->nh_msg_type, nmh->nh_key);
899 rb_erase(&nmh->nh_node, &r2net_handler_tree);
900 list_del_init(&nmh->nh_unregister_item);
901 kref_put(&nmh->nh_kref, r2net_handler_kref_release);
902 }
903 write_unlock(&r2net_handler_lock);
904}
905EXPORT_SYMBOL_GPL(r2net_unregister_handler_list);
906
907static struct r2net_msg_handler *r2net_handler_get(u32 msg_type, u32 key)
908{
909 struct r2net_msg_handler *nmh;
910
911 read_lock(&r2net_handler_lock);
912 nmh = r2net_handler_tree_lookup(msg_type, key, NULL, NULL);
913 if (nmh)
914 kref_get(&nmh->nh_kref);
915 read_unlock(&r2net_handler_lock);
916
917 return nmh;
918}
919
920/* ------------------------------------------------------------ */
921
922static int r2net_recv_tcp_msg(struct socket *sock, void *data, size_t len)
923{
924 int ret;
925 mm_segment_t oldfs;
926 struct kvec vec = {
927 .iov_len = len,
928 .iov_base = data,
929 };
930 struct msghdr msg = {
931 .msg_iovlen = 1,
932 .msg_iov = (struct iovec *)&vec,
933 .msg_flags = MSG_DONTWAIT,
934 };
935
936 oldfs = get_fs();
937 set_fs(get_ds());
938 ret = sock_recvmsg(sock, &msg, len, msg.msg_flags);
939 set_fs(oldfs);
940
941 return ret;
942}
943
944static int r2net_send_tcp_msg(struct socket *sock, struct kvec *vec,
945 size_t veclen, size_t total)
946{
947 int ret;
948 mm_segment_t oldfs;
949 struct msghdr msg = {
950 .msg_iov = (struct iovec *)vec,
951 .msg_iovlen = veclen,
952 };
953
954 if (sock == NULL) {
955 ret = -EINVAL;
956 goto out;
957 }
958
959 oldfs = get_fs();
960 set_fs(get_ds());
961 ret = sock_sendmsg(sock, &msg, total);
962 set_fs(oldfs);
963 if (ret != total) {
964 mlog(ML_ERROR, "sendmsg returned %d instead of %zu\n", ret,
965 total);
966 if (ret >= 0)
967 ret = -EPIPE; /* should be smarter, I bet */
968 goto out;
969 }
970
971 ret = 0;
972out:
973 if (ret < 0)
974 mlog(0, "returning error: %d\n", ret);
975 return ret;
976}
977
978static void r2net_sendpage(struct r2net_sock_container *sc,
979 void *kmalloced_virt,
980 size_t size)
981{
982 struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
983 ssize_t ret;
984
985 while (1) {
986 mutex_lock(&sc->sc_send_lock);
987 ret = sc->sc_sock->ops->sendpage(sc->sc_sock,
988 virt_to_page(kmalloced_virt),
989 (long)kmalloced_virt & ~PAGE_MASK,
990 size, MSG_DONTWAIT);
991 mutex_unlock(&sc->sc_send_lock);
992 if (ret == size)
993 break;
994 if (ret == (ssize_t)-EAGAIN) {
995 mlog(0, "sendpage of size %zu to " SC_NODEF_FMT
996 " returned EAGAIN\n", size, sc->sc_node->nd_name,
997 sc->sc_node->nd_num,
998 &sc->sc_node->nd_ipv4_address,
999 ntohs(sc->sc_node->nd_ipv4_port));
1000 cond_resched();
1001 continue;
1002 }
1003 mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT
1004 " failed with %zd\n", size, sc->sc_node->nd_name,
1005 sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
1006 ntohs(sc->sc_node->nd_ipv4_port), ret);
1007 r2net_ensure_shutdown(nn, sc, 0);
1008 break;
1009 }
1010}
1011
1012static void r2net_init_msg(struct r2net_msg *msg, u16 data_len,
1013 u16 msg_type, u32 key)
1014{
1015 memset(msg, 0, sizeof(struct r2net_msg));
1016 msg->magic = cpu_to_be16(R2NET_MSG_MAGIC);
1017 msg->data_len = cpu_to_be16(data_len);
1018 msg->msg_type = cpu_to_be16(msg_type);
1019 msg->sys_status = cpu_to_be32(R2NET_ERR_NONE);
1020 msg->status = 0;
1021 msg->key = cpu_to_be32(key);
1022}
1023
1024static int r2net_tx_can_proceed(struct r2net_node *nn,
1025 struct r2net_sock_container **sc_ret,
1026 int *error)
1027{
1028 int ret = 0;
1029
1030 spin_lock(&nn->nn_lock);
1031 if (nn->nn_persistent_error) {
1032 ret = 1;
1033 *sc_ret = NULL;
1034 *error = nn->nn_persistent_error;
1035 } else if (nn->nn_sc_valid) {
1036 kref_get(&nn->nn_sc->sc_kref);
1037
1038 ret = 1;
1039 *sc_ret = nn->nn_sc;
1040 *error = 0;
1041 }
1042 spin_unlock(&nn->nn_lock);
1043
1044 return ret;
1045}
1046
1047/* Get a map of all nodes to which this node is currently connected to */
1048void r2net_fill_node_map(unsigned long *map, unsigned bytes)
1049{
1050 struct r2net_sock_container *sc;
1051 int node, ret;
1052
1053 BUG_ON(bytes < (BITS_TO_LONGS(R2NM_MAX_NODES) * sizeof(unsigned long)));
1054
1055 memset(map, 0, bytes);
1056 for (node = 0; node < R2NM_MAX_NODES; ++node) {
1057 r2net_tx_can_proceed(r2net_nn_from_num(node), &sc, &ret);
1058 if (!ret) {
1059 set_bit(node, map);
1060 sc_put(sc);
1061 }
1062 }
1063}
1064EXPORT_SYMBOL_GPL(r2net_fill_node_map);
1065
1066int r2net_send_message_vec(u32 msg_type, u32 key, struct kvec *caller_vec,
1067 size_t caller_veclen, u8 target_node, int *status)
1068{
1069 int ret = 0;
1070 struct r2net_msg *msg = NULL;
1071 size_t veclen, caller_bytes = 0;
1072 struct kvec *vec = NULL;
1073 struct r2net_sock_container *sc = NULL;
1074 struct r2net_node *nn = r2net_nn_from_num(target_node);
1075 struct r2net_status_wait nsw = {
1076 .ns_node_item = LIST_HEAD_INIT(nsw.ns_node_item),
1077 };
1078 struct r2net_send_tracking nst;
1079
1080 /* this may be a general bug fix */
1081 init_waitqueue_head(&nsw.ns_wq);
1082
1083 r2net_init_nst(&nst, msg_type, key, current, target_node);
1084
1085 if (r2net_wq == NULL) {
1086 mlog(0, "attempt to tx without r2netd running\n");
1087 ret = -ESRCH;
1088 goto out;
1089 }
1090
1091 if (caller_veclen == 0) {
1092 mlog(0, "bad kvec array length\n");
1093 ret = -EINVAL;
1094 goto out;
1095 }
1096
1097 caller_bytes = iov_length((struct iovec *)caller_vec, caller_veclen);
1098 if (caller_bytes > R2NET_MAX_PAYLOAD_BYTES) {
1099 mlog(0, "total payload len %zu too large\n", caller_bytes);
1100 ret = -EINVAL;
1101 goto out;
1102 }
1103
1104 if (target_node == r2nm_this_node()) {
1105 ret = -ELOOP;
1106 goto out;
1107 }
1108
1109 r2net_debug_add_nst(&nst);
1110
1111 r2net_set_nst_sock_time(&nst);
1112
1113 wait_event(nn->nn_sc_wq, r2net_tx_can_proceed(nn, &sc, &ret));
1114 if (ret)
1115 goto out;
1116
1117 r2net_set_nst_sock_container(&nst, sc);
1118
1119 veclen = caller_veclen + 1;
1120 vec = kmalloc(sizeof(struct kvec) * veclen, GFP_ATOMIC);
1121 if (vec == NULL) {
1122 mlog(0, "failed to %zu element kvec!\n", veclen);
1123 ret = -ENOMEM;
1124 goto out;
1125 }
1126
1127 msg = kmalloc(sizeof(struct r2net_msg), GFP_ATOMIC);
1128 if (!msg) {
1129 mlog(0, "failed to allocate a r2net_msg!\n");
1130 ret = -ENOMEM;
1131 goto out;
1132 }
1133
1134 r2net_init_msg(msg, caller_bytes, msg_type, key);
1135
1136 vec[0].iov_len = sizeof(struct r2net_msg);
1137 vec[0].iov_base = msg;
1138 memcpy(&vec[1], caller_vec, caller_veclen * sizeof(struct kvec));
1139
1140 ret = r2net_prep_nsw(nn, &nsw);
1141 if (ret)
1142 goto out;
1143
1144 msg->msg_num = cpu_to_be32(nsw.ns_id);
1145 r2net_set_nst_msg_id(&nst, nsw.ns_id);
1146
1147 r2net_set_nst_send_time(&nst);
1148
1149 /* finally, convert the message header to network byte-order
1150 * and send */
1151 mutex_lock(&sc->sc_send_lock);
1152 ret = r2net_send_tcp_msg(sc->sc_sock, vec, veclen,
1153 sizeof(struct r2net_msg) + caller_bytes);
1154 mutex_unlock(&sc->sc_send_lock);
1155 msglog(msg, "sending returned %d\n", ret);
1156 if (ret < 0) {
1157 mlog(0, "error returned from r2net_send_tcp_msg=%d\n", ret);
1158 goto out;
1159 }
1160
1161 /* wait on other node's handler */
1162 r2net_set_nst_status_time(&nst);
1163 wait_event(nsw.ns_wq, r2net_nsw_completed(nn, &nsw));
1164
1165 r2net_update_send_stats(&nst, sc);
1166
1167 /* Note that we avoid overwriting the callers status return
1168 * variable if a system error was reported on the other
1169 * side. Callers beware. */
1170 ret = r2net_sys_err_to_errno(nsw.ns_sys_status);
1171 if (status && !ret)
1172 *status = nsw.ns_status;
1173
1174 mlog(0, "woken, returning system status %d, user status %d\n",
1175 ret, nsw.ns_status);
1176out:
1177 r2net_debug_del_nst(&nst); /* must be before dropping sc and node */
1178 if (sc)
1179 sc_put(sc);
1180 kfree(vec);
1181 kfree(msg);
1182 r2net_complete_nsw(nn, &nsw, 0, 0, 0);
1183 return ret;
1184}
1185EXPORT_SYMBOL_GPL(r2net_send_message_vec);
1186
1187int r2net_send_message(u32 msg_type, u32 key, void *data, u32 len,
1188 u8 target_node, int *status)
1189{
1190 struct kvec vec = {
1191 .iov_base = data,
1192 .iov_len = len,
1193 };
1194 return r2net_send_message_vec(msg_type, key, &vec, 1,
1195 target_node, status);
1196}
1197EXPORT_SYMBOL_GPL(r2net_send_message);
1198
1199static int r2net_send_status_magic(struct socket *sock, struct r2net_msg *hdr,
1200 enum r2net_system_error syserr, int err)
1201{
1202 struct kvec vec = {
1203 .iov_base = hdr,
1204 .iov_len = sizeof(struct r2net_msg),
1205 };
1206
1207 BUG_ON(syserr >= R2NET_ERR_MAX);
1208
1209 /* leave other fields intact from the incoming message, msg_num
1210 * in particular */
1211 hdr->sys_status = cpu_to_be32(syserr);
1212 hdr->status = cpu_to_be32(err);
1213 /* twiddle the magic */
1214 hdr->magic = cpu_to_be16(R2NET_MSG_STATUS_MAGIC);
1215 hdr->data_len = 0;
1216
1217 msglog(hdr, "about to send status magic %d\n", err);
1218 /* hdr has been in host byteorder this whole time */
1219 return r2net_send_tcp_msg(sock, &vec, 1, sizeof(struct r2net_msg));
1220}
1221
1222/*
1223 * "data magic" is a long version of "status magic" where the message
1224 * payload actually contains data to be passed in reply to certain messages
1225 */
1226static int r2net_send_data_magic(struct r2net_sock_container *sc,
1227 struct r2net_msg *hdr,
1228 void *data, size_t data_len,
1229 enum r2net_system_error syserr, int err)
1230{
1231 struct kvec vec[2];
1232 int ret;
1233
1234 vec[0].iov_base = hdr;
1235 vec[0].iov_len = sizeof(struct r2net_msg);
1236 vec[1].iov_base = data;
1237 vec[1].iov_len = data_len;
1238
1239 BUG_ON(syserr >= R2NET_ERR_MAX);
1240
1241 /* leave other fields intact from the incoming message, msg_num
1242 * in particular */
1243 hdr->sys_status = cpu_to_be32(syserr);
1244 hdr->status = cpu_to_be32(err);
1245 hdr->magic = cpu_to_be16(R2NET_MSG_DATA_MAGIC); /* twiddle magic */
1246 hdr->data_len = cpu_to_be16(data_len);
1247
1248 msglog(hdr, "about to send data magic %d\n", err);
1249 /* hdr has been in host byteorder this whole time */
1250 ret = r2net_send_tcp_msg(sc->sc_sock, vec, 2,
1251 sizeof(struct r2net_msg) + data_len);
1252 return ret;
1253}
1254
1255/*
1256 * called by a message handler to convert an otherwise normal reply
1257 * message into a "data magic" message
1258 */
1259void r2net_force_data_magic(struct r2net_msg *hdr, u16 msgtype, u32 msgkey)
1260{
1261 hdr->magic = cpu_to_be16(R2NET_MSG_DATA_MAGIC);
1262 hdr->msg_type = cpu_to_be16(msgtype);
1263 hdr->key = cpu_to_be32(msgkey);
1264}
1265
1266/* this returns -errno if the header was unknown or too large, etc.
1267 * after this is called the buffer us reused for the next message */
1268static int r2net_process_message(struct r2net_sock_container *sc,
1269 struct r2net_msg *hdr)
1270{
1271 struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
1272 int ret = 0, handler_status;
1273 enum r2net_system_error syserr;
1274 struct r2net_msg_handler *nmh = NULL;
1275 void *ret_data = NULL;
1276 int data_magic = 0;
1277
1278 msglog(hdr, "processing message\n");
1279
1280 r2net_sc_postpone_idle(sc);
1281
1282 switch (be16_to_cpu(hdr->magic)) {
1283
1284 case R2NET_MSG_STATUS_MAGIC:
1285 /* special type for returning message status */
1286 r2net_complete_nsw(nn, NULL, be32_to_cpu(hdr->msg_num),
1287 be32_to_cpu(hdr->sys_status),
1288 be32_to_cpu(hdr->status));
1289 goto out;
1290 case R2NET_MSG_KEEP_REQ_MAGIC:
1291 r2net_sendpage(sc, r2net_keep_resp, sizeof(*r2net_keep_resp));
1292 goto out;
1293 case R2NET_MSG_KEEP_RESP_MAGIC:
1294 goto out;
1295 case R2NET_MSG_MAGIC:
1296 break;
1297 case R2NET_MSG_DATA_MAGIC:
1298 /*
1299 * unlike a normal status magic, a data magic DOES
1300 * (MUST) have a handler, so the control flow is
1301 * a little funky here as a result
1302 */
1303 data_magic = 1;
1304 break;
1305 default:
1306 msglog(hdr, "bad magic\n");
1307 ret = -EINVAL;
1308 goto out;
1309 break;
1310 }
1311
1312 /* find a handler for it */
1313 handler_status = 0;
1314 nmh = r2net_handler_get(be16_to_cpu(hdr->msg_type),
1315 be32_to_cpu(hdr->key));
1316 if (!nmh) {
1317 mlog(ML_TCP, "couldn't find handler for type %u key %08x\n",
1318 be16_to_cpu(hdr->msg_type), be32_to_cpu(hdr->key));
1319 syserr = R2NET_ERR_NO_HNDLR;
1320 goto out_respond;
1321 }
1322
1323 syserr = R2NET_ERR_NONE;
1324
1325 if (be16_to_cpu(hdr->data_len) > nmh->nh_max_len)
1326 syserr = R2NET_ERR_OVERFLOW;
1327
1328 if (syserr != R2NET_ERR_NONE)
1329 goto out_respond;
1330
1331 r2net_set_func_start_time(sc);
1332 sc->sc_msg_key = be32_to_cpu(hdr->key);
1333 sc->sc_msg_type = be16_to_cpu(hdr->msg_type);
1334 handler_status = (nmh->nh_func)(hdr, sizeof(struct r2net_msg) +
1335 be16_to_cpu(hdr->data_len),
1336 nmh->nh_func_data, &ret_data);
1337 if (data_magic) {
1338 /*
1339 * handler handled data sent in reply to request
1340 * so complete the transaction
1341 */
1342 r2net_complete_nsw(nn, NULL, be32_to_cpu(hdr->msg_num),
1343 be32_to_cpu(hdr->sys_status), handler_status);
1344 goto out;
1345 }
1346 /*
1347 * handler changed magic to DATA_MAGIC to reply to request for data,
1348 * implies ret_data points to data to return and handler_status
1349 * is the number of bytes of data
1350 */
1351 if (be16_to_cpu(hdr->magic) == R2NET_MSG_DATA_MAGIC) {
1352 ret = r2net_send_data_magic(sc, hdr,
1353 ret_data, handler_status,
1354 syserr, 0);
1355 hdr = NULL;
1356 mlog(0, "sending data reply %d, syserr %d returned %d\n",
1357 handler_status, syserr, ret);
1358 r2net_set_func_stop_time(sc);
1359
1360 r2net_update_recv_stats(sc);
1361 goto out;
1362 }
1363 r2net_set_func_stop_time(sc);
1364
1365 r2net_update_recv_stats(sc);
1366
1367out_respond:
1368 /* this destroys the hdr, so don't use it after this */
1369 mutex_lock(&sc->sc_send_lock);
1370 ret = r2net_send_status_magic(sc->sc_sock, hdr, syserr,
1371 handler_status);
1372 mutex_unlock(&sc->sc_send_lock);
1373 hdr = NULL;
1374 mlog(0, "sending handler status %d, syserr %d returned %d\n",
1375 handler_status, syserr, ret);
1376
1377 if (nmh) {
1378 BUG_ON(ret_data != NULL && nmh->nh_post_func == NULL);
1379 if (nmh->nh_post_func)
1380 (nmh->nh_post_func)(handler_status, nmh->nh_func_data,
1381 ret_data);
1382 }
1383
1384out:
1385 if (nmh)
1386 r2net_handler_put(nmh);
1387 return ret;
1388}
1389
1390static int r2net_check_handshake(struct r2net_sock_container *sc)
1391{
1392 struct r2net_handshake *hand = page_address(sc->sc_page);
1393 struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
1394
1395 if (hand->protocol_version != cpu_to_be64(R2NET_PROTOCOL_VERSION)) {
1396 printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " Advertised net "
1397 "protocol version %llu but %llu is required. "
1398 "Disconnecting.\n", sc->sc_node->nd_name,
1399 sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
1400 ntohs(sc->sc_node->nd_ipv4_port),
1401 (unsigned long long)be64_to_cpu(hand->protocol_version),
1402 R2NET_PROTOCOL_VERSION);
1403
1404 /* don't bother reconnecting if its the wrong version. */
1405 r2net_ensure_shutdown(nn, sc, -ENOTCONN);
1406 return -1;
1407 }
1408
1409 /*
1410 * Ensure timeouts are consistent with other nodes, otherwise
1411 * we can end up with one node thinking that the other must be down,
1412 * but isn't. This can ultimately cause corruption.
1413 */
1414 if (be32_to_cpu(hand->r2net_idle_timeout_ms) !=
1415 r2net_idle_timeout()) {
1416 printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " uses a network "
1417 "idle timeout of %u ms, but we use %u ms locally. "
1418 "Disconnecting.\n", sc->sc_node->nd_name,
1419 sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
1420 ntohs(sc->sc_node->nd_ipv4_port),
1421 be32_to_cpu(hand->r2net_idle_timeout_ms),
1422 r2net_idle_timeout());
1423 r2net_ensure_shutdown(nn, sc, -ENOTCONN);
1424 return -1;
1425 }
1426
1427 if (be32_to_cpu(hand->r2net_keepalive_delay_ms) !=
1428 r2net_keepalive_delay()) {
1429 printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " uses a keepalive "
1430 "delay of %u ms, but we use %u ms locally. "
1431 "Disconnecting.\n", sc->sc_node->nd_name,
1432 sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
1433 ntohs(sc->sc_node->nd_ipv4_port),
1434 be32_to_cpu(hand->r2net_keepalive_delay_ms),
1435 r2net_keepalive_delay());
1436 r2net_ensure_shutdown(nn, sc, -ENOTCONN);
1437 return -1;
1438 }
1439
1440 if (be32_to_cpu(hand->r2hb_heartbeat_timeout_ms) !=
1441 R2HB_MAX_WRITE_TIMEOUT_MS) {
1442 printk(KERN_NOTICE "ramster: " SC_NODEF_FMT " uses a heartbeat "
1443 "timeout of %u ms, but we use %u ms locally. "
1444 "Disconnecting.\n", sc->sc_node->nd_name,
1445 sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
1446 ntohs(sc->sc_node->nd_ipv4_port),
1447 be32_to_cpu(hand->r2hb_heartbeat_timeout_ms),
1448 R2HB_MAX_WRITE_TIMEOUT_MS);
1449 r2net_ensure_shutdown(nn, sc, -ENOTCONN);
1450 return -1;
1451 }
1452
1453 sc->sc_handshake_ok = 1;
1454
1455 spin_lock(&nn->nn_lock);
1456 /* set valid and queue the idle timers only if it hasn't been
1457 * shut down already */
1458 if (nn->nn_sc == sc) {
1459 r2net_sc_reset_idle_timer(sc);
1460 atomic_set(&nn->nn_timeout, 0);
1461 r2net_set_nn_state(nn, sc, 1, 0);
1462 }
1463 spin_unlock(&nn->nn_lock);
1464
1465 /* shift everything up as though it wasn't there */
1466 sc->sc_page_off -= sizeof(struct r2net_handshake);
1467 if (sc->sc_page_off)
1468 memmove(hand, hand + 1, sc->sc_page_off);
1469
1470 return 0;
1471}
1472
1473/* this demuxes the queued rx bytes into header or payload bits and calls
1474 * handlers as each full message is read off the socket. it returns -error,
1475 * == 0 eof, or > 0 for progress made.*/
1476static int r2net_advance_rx(struct r2net_sock_container *sc)
1477{
1478 struct r2net_msg *hdr;
1479 int ret = 0;
1480 void *data;
1481 size_t datalen;
1482
1483 sclog(sc, "receiving\n");
1484 r2net_set_advance_start_time(sc);
1485
1486 if (unlikely(sc->sc_handshake_ok == 0)) {
1487 if (sc->sc_page_off < sizeof(struct r2net_handshake)) {
1488 data = page_address(sc->sc_page) + sc->sc_page_off;
1489 datalen = sizeof(struct r2net_handshake) -
1490 sc->sc_page_off;
1491 ret = r2net_recv_tcp_msg(sc->sc_sock, data, datalen);
1492 if (ret > 0)
1493 sc->sc_page_off += ret;
1494 }
1495
1496 if (sc->sc_page_off == sizeof(struct r2net_handshake)) {
1497 r2net_check_handshake(sc);
1498 if (unlikely(sc->sc_handshake_ok == 0))
1499 ret = -EPROTO;
1500 }
1501 goto out;
1502 }
1503
1504 /* do we need more header? */
1505 if (sc->sc_page_off < sizeof(struct r2net_msg)) {
1506 data = page_address(sc->sc_page) + sc->sc_page_off;
1507 datalen = sizeof(struct r2net_msg) - sc->sc_page_off;
1508 ret = r2net_recv_tcp_msg(sc->sc_sock, data, datalen);
1509 if (ret > 0) {
1510 sc->sc_page_off += ret;
1511 /* only swab incoming here.. we can
1512 * only get here once as we cross from
1513 * being under to over */
1514 if (sc->sc_page_off == sizeof(struct r2net_msg)) {
1515 hdr = page_address(sc->sc_page);
1516 if (be16_to_cpu(hdr->data_len) >
1517 R2NET_MAX_PAYLOAD_BYTES)
1518 ret = -EOVERFLOW;
1519 }
1520 }
1521 if (ret <= 0)
1522 goto out;
1523 }
1524
1525 if (sc->sc_page_off < sizeof(struct r2net_msg)) {
1526 /* oof, still don't have a header */
1527 goto out;
1528 }
1529
1530 /* this was swabbed above when we first read it */
1531 hdr = page_address(sc->sc_page);
1532
1533 msglog(hdr, "at page_off %zu\n", sc->sc_page_off);
1534
1535 /* do we need more payload? */
1536 if (sc->sc_page_off - sizeof(struct r2net_msg) <
1537 be16_to_cpu(hdr->data_len)) {
1538 /* need more payload */
1539 data = page_address(sc->sc_page) + sc->sc_page_off;
1540 datalen = (sizeof(struct r2net_msg) +
1541 be16_to_cpu(hdr->data_len)) -
1542 sc->sc_page_off;
1543 ret = r2net_recv_tcp_msg(sc->sc_sock, data, datalen);
1544 if (ret > 0)
1545 sc->sc_page_off += ret;
1546 if (ret <= 0)
1547 goto out;
1548 }
1549
1550 if (sc->sc_page_off - sizeof(struct r2net_msg) ==
1551 be16_to_cpu(hdr->data_len)) {
1552 /* we can only get here once, the first time we read
1553 * the payload.. so set ret to progress if the handler
1554 * works out. after calling this the message is toast */
1555 ret = r2net_process_message(sc, hdr);
1556 if (ret == 0)
1557 ret = 1;
1558 sc->sc_page_off = 0;
1559 }
1560
1561out:
1562 sclog(sc, "ret = %d\n", ret);
1563 r2net_set_advance_stop_time(sc);
1564 return ret;
1565}
1566
1567/* this work func is triggerd by data ready. it reads until it can read no
1568 * more. it interprets 0, eof, as fatal. if data_ready hits while we're doing
1569 * our work the work struct will be marked and we'll be called again. */
1570static void r2net_rx_until_empty(struct work_struct *work)
1571{
1572 struct r2net_sock_container *sc =
1573 container_of(work, struct r2net_sock_container, sc_rx_work);
1574 int ret;
1575
1576 do {
1577 ret = r2net_advance_rx(sc);
1578 } while (ret > 0);
1579
1580 if (ret <= 0 && ret != -EAGAIN) {
1581 struct r2net_node *nn = r2net_nn_from_num(sc->sc_node->nd_num);
1582 sclog(sc, "saw error %d, closing\n", ret);
1583 /* not permanent so read failed handshake can retry */
1584 r2net_ensure_shutdown(nn, sc, 0);
1585 }
1586
1587 sc_put(sc);
1588}
1589
1590static int r2net_set_nodelay(struct socket *sock)
1591{
1592 int ret, val = 1;
1593 mm_segment_t oldfs;
1594
1595 oldfs = get_fs();
1596 set_fs(KERNEL_DS);
1597
1598 /*
1599 * Dear unsuspecting programmer,
1600 *
1601 * Don't use sock_setsockopt() for SOL_TCP. It doesn't check its level
1602 * argument and assumes SOL_SOCKET so, say, your TCP_NODELAY will
1603 * silently turn into SO_DEBUG.
1604 *
1605 * Yours,
1606 * Keeper of hilariously fragile interfaces.
1607 */
1608 ret = sock->ops->setsockopt(sock, SOL_TCP, TCP_NODELAY,
1609 (char __user *)&val, sizeof(val));
1610
1611 set_fs(oldfs);
1612 return ret;
1613}
1614
1615static void r2net_initialize_handshake(void)
1616{
1617 r2net_hand->r2hb_heartbeat_timeout_ms = cpu_to_be32(
1618 R2HB_MAX_WRITE_TIMEOUT_MS);
1619 r2net_hand->r2net_idle_timeout_ms = cpu_to_be32(r2net_idle_timeout());
1620 r2net_hand->r2net_keepalive_delay_ms = cpu_to_be32(
1621 r2net_keepalive_delay());
1622 r2net_hand->r2net_reconnect_delay_ms = cpu_to_be32(
1623 r2net_reconnect_delay());
1624}
1625
1626/* ------------------------------------------------------------ */
1627
1628/* called when a connect completes and after a sock is accepted. the
1629 * rx path will see the response and mark the sc valid */
1630static void r2net_sc_connect_completed(struct work_struct *work)
1631{
1632 struct r2net_sock_container *sc =
1633 container_of(work, struct r2net_sock_container,
1634 sc_connect_work);
1635
1636 mlog(ML_MSG, "sc sending handshake with ver %llu id %llx\n",
1637 (unsigned long long)R2NET_PROTOCOL_VERSION,
1638 (unsigned long long)be64_to_cpu(r2net_hand->connector_id));
1639
1640 r2net_initialize_handshake();
1641 r2net_sendpage(sc, r2net_hand, sizeof(*r2net_hand));
1642 sc_put(sc);
1643}
1644
1645/* this is called as a work_struct func. */
1646static void r2net_sc_send_keep_req(struct work_struct *work)
1647{
1648 struct r2net_sock_container *sc =
1649 container_of(work, struct r2net_sock_container,
1650 sc_keepalive_work.work);
1651
1652 r2net_sendpage(sc, r2net_keep_req, sizeof(*r2net_keep_req));
1653 sc_put(sc);
1654}
1655
1656/* socket shutdown does a del_timer_sync against this as it tears down.
1657 * we can't start this timer until we've got to the point in sc buildup
1658 * where shutdown is going to be involved */
1659static void r2net_idle_timer(unsigned long data)
1660{
1661 struct r2net_sock_container *sc = (struct r2net_sock_container *)data;
1662#ifdef CONFIG_DEBUG_FS
1663 unsigned long msecs = ktime_to_ms(ktime_get()) -
1664 ktime_to_ms(sc->sc_tv_timer);
1665#else
1666 unsigned long msecs = r2net_idle_timeout();
1667#endif
1668
1669 printk(KERN_NOTICE "ramster: Connection to " SC_NODEF_FMT " has been "
1670 "idle for %lu.%lu secs, shutting it down.\n",
1671 sc->sc_node->nd_name, sc->sc_node->nd_num,
1672 &sc->sc_node->nd_ipv4_address, ntohs(sc->sc_node->nd_ipv4_port),
1673 msecs / 1000, msecs % 1000);
1674
1675 /*
1676 * Initialize the nn_timeout so that the next connection attempt
1677 * will continue in r2net_start_connect.
1678 */
1679 /* Avoid spurious shutdowns... not sure if this is still necessary */
1680 pr_err("ramster_idle_timer, skipping shutdown work\n");
1681#if 0
1682 /* old code used to do these two lines */
1683 atomic_set(&nn->nn_timeout, 1);
1684 r2net_sc_queue_work(sc, &sc->sc_shutdown_work);
1685#endif
1686}
1687
1688static void r2net_sc_reset_idle_timer(struct r2net_sock_container *sc)
1689{
1690 r2net_sc_cancel_delayed_work(sc, &sc->sc_keepalive_work);
1691 r2net_sc_queue_delayed_work(sc, &sc->sc_keepalive_work,
1692 msecs_to_jiffies(r2net_keepalive_delay()));
1693 r2net_set_sock_timer(sc);
1694 mod_timer(&sc->sc_idle_timeout,
1695 jiffies + msecs_to_jiffies(r2net_idle_timeout()));
1696}
1697
1698static void r2net_sc_postpone_idle(struct r2net_sock_container *sc)
1699{
1700 /* Only push out an existing timer */
1701 if (timer_pending(&sc->sc_idle_timeout))
1702 r2net_sc_reset_idle_timer(sc);
1703}
1704
1705/* this work func is kicked whenever a path sets the nn state which doesn't
1706 * have valid set. This includes seeing hb come up, losing a connection,
1707 * having a connect attempt fail, etc. This centralizes the logic which decides
1708 * if a connect attempt should be made or if we should give up and all future
1709 * transmit attempts should fail */
1710static void r2net_start_connect(struct work_struct *work)
1711{
1712 struct r2net_node *nn =
1713 container_of(work, struct r2net_node, nn_connect_work.work);
1714 struct r2net_sock_container *sc = NULL;
1715 struct r2nm_node *node = NULL, *mynode = NULL;
1716 struct socket *sock = NULL;
1717 struct sockaddr_in myaddr = {0, }, remoteaddr = {0, };
1718 int ret = 0, stop;
1719 unsigned int timeout;
1720
1721 /* if we're greater we initiate tx, otherwise we accept */
1722 if (r2nm_this_node() <= r2net_num_from_nn(nn))
1723 goto out;
1724
1725 /* watch for racing with tearing a node down */
1726 node = r2nm_get_node_by_num(r2net_num_from_nn(nn));
1727 if (node == NULL) {
1728 ret = 0;
1729 goto out;
1730 }
1731
1732 mynode = r2nm_get_node_by_num(r2nm_this_node());
1733 if (mynode == NULL) {
1734 ret = 0;
1735 goto out;
1736 }
1737
1738 spin_lock(&nn->nn_lock);
1739 /*
1740 * see if we already have one pending or have given up.
1741 * For nn_timeout, it is set when we close the connection
1742 * because of the idle time out. So it means that we have
1743 * at least connected to that node successfully once,
1744 * now try to connect to it again.
1745 */
1746 timeout = atomic_read(&nn->nn_timeout);
1747 stop = (nn->nn_sc ||
1748 (nn->nn_persistent_error &&
1749 (nn->nn_persistent_error != -ENOTCONN || timeout == 0)));
1750 spin_unlock(&nn->nn_lock);
1751 if (stop)
1752 goto out;
1753
1754 nn->nn_last_connect_attempt = jiffies;
1755
1756 sc = sc_alloc(node);
1757 if (sc == NULL) {
1758 mlog(0, "couldn't allocate sc\n");
1759 ret = -ENOMEM;
1760 goto out;
1761 }
1762
1763 ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
1764 if (ret < 0) {
1765 mlog(0, "can't create socket: %d\n", ret);
1766 goto out;
1767 }
1768 sc->sc_sock = sock; /* freed by sc_kref_release */
1769
1770 sock->sk->sk_allocation = GFP_ATOMIC;
1771
1772 myaddr.sin_family = AF_INET;
1773 myaddr.sin_addr.s_addr = mynode->nd_ipv4_address;
1774 myaddr.sin_port = htons(0); /* any port */
1775
1776 ret = sock->ops->bind(sock, (struct sockaddr *)&myaddr,
1777 sizeof(myaddr));
1778 if (ret) {
1779 mlog(ML_ERROR, "bind failed with %d at address %pI4\n",
1780 ret, &mynode->nd_ipv4_address);
1781 goto out;
1782 }
1783
1784 ret = r2net_set_nodelay(sc->sc_sock);
1785 if (ret) {
1786 mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
1787 goto out;
1788 }
1789
1790 r2net_register_callbacks(sc->sc_sock->sk, sc);
1791
1792 spin_lock(&nn->nn_lock);
1793 /* handshake completion will set nn->nn_sc_valid */
1794 r2net_set_nn_state(nn, sc, 0, 0);
1795 spin_unlock(&nn->nn_lock);
1796
1797 remoteaddr.sin_family = AF_INET;
1798 remoteaddr.sin_addr.s_addr = node->nd_ipv4_address;
1799 remoteaddr.sin_port = node->nd_ipv4_port;
1800
1801 ret = sc->sc_sock->ops->connect(sc->sc_sock,
1802 (struct sockaddr *)&remoteaddr,
1803 sizeof(remoteaddr),
1804 O_NONBLOCK);
1805 if (ret == -EINPROGRESS)
1806 ret = 0;
1807
1808out:
1809 if (ret) {
1810 printk(KERN_NOTICE "ramster: Connect attempt to " SC_NODEF_FMT
1811 " failed with errno %d\n", sc->sc_node->nd_name,
1812 sc->sc_node->nd_num, &sc->sc_node->nd_ipv4_address,
1813 ntohs(sc->sc_node->nd_ipv4_port), ret);
1814 /* 0 err so that another will be queued and attempted
1815 * from set_nn_state */
1816 if (sc)
1817 r2net_ensure_shutdown(nn, sc, 0);
1818 }
1819 if (sc)
1820 sc_put(sc);
1821 if (node)
1822 r2nm_node_put(node);
1823 if (mynode)
1824 r2nm_node_put(mynode);
1825
1826 return;
1827}
1828
1829static void r2net_connect_expired(struct work_struct *work)
1830{
1831 struct r2net_node *nn =
1832 container_of(work, struct r2net_node, nn_connect_expired.work);
1833
1834 spin_lock(&nn->nn_lock);
1835 if (!nn->nn_sc_valid) {
1836 printk(KERN_NOTICE "ramster: No connection established with "
1837 "node %u after %u.%u seconds, giving up.\n",
1838 r2net_num_from_nn(nn),
1839 r2net_idle_timeout() / 1000,
1840 r2net_idle_timeout() % 1000);
1841
1842 r2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
1843 }
1844 spin_unlock(&nn->nn_lock);
1845}
1846
1847static void r2net_still_up(struct work_struct *work)
1848{
1849}
1850
1851/* ------------------------------------------------------------ */
1852
1853void r2net_disconnect_node(struct r2nm_node *node)
1854{
1855 struct r2net_node *nn = r2net_nn_from_num(node->nd_num);
1856
1857 /* don't reconnect until it's heartbeating again */
1858 spin_lock(&nn->nn_lock);
1859 atomic_set(&nn->nn_timeout, 0);
1860 r2net_set_nn_state(nn, NULL, 0, -ENOTCONN);
1861 spin_unlock(&nn->nn_lock);
1862
1863 if (r2net_wq) {
1864 cancel_delayed_work(&nn->nn_connect_expired);
1865 cancel_delayed_work(&nn->nn_connect_work);
1866 cancel_delayed_work(&nn->nn_still_up);
1867 flush_workqueue(r2net_wq);
1868 }
1869}
1870
1871static void r2net_hb_node_down_cb(struct r2nm_node *node, int node_num,
1872 void *data)
1873{
1874 if (!node)
1875 return;
1876
1877 if (node_num != r2nm_this_node())
1878 r2net_disconnect_node(node);
1879
1880 BUG_ON(atomic_read(&r2net_connected_peers) < 0);
1881}
1882
1883static void r2net_hb_node_up_cb(struct r2nm_node *node, int node_num,
1884 void *data)
1885{
1886 struct r2net_node *nn = r2net_nn_from_num(node_num);
1887
1888 BUG_ON(!node);
1889
1890 /* ensure an immediate connect attempt */
1891 nn->nn_last_connect_attempt = jiffies -
1892 (msecs_to_jiffies(r2net_reconnect_delay()) + 1);
1893
1894 if (node_num != r2nm_this_node()) {
1895 /* believe it or not, accept and node hearbeating testing
1896 * can succeed for this node before we got here.. so
1897 * only use set_nn_state to clear the persistent error
1898 * if that hasn't already happened */
1899 spin_lock(&nn->nn_lock);
1900 atomic_set(&nn->nn_timeout, 0);
1901 if (nn->nn_persistent_error)
1902 r2net_set_nn_state(nn, NULL, 0, 0);
1903 spin_unlock(&nn->nn_lock);
1904 }
1905}
1906
1907void r2net_unregister_hb_callbacks(void)
1908{
1909 r2hb_unregister_callback(NULL, &r2net_hb_up);
1910 r2hb_unregister_callback(NULL, &r2net_hb_down);
1911}
1912
1913int r2net_register_hb_callbacks(void)
1914{
1915 int ret;
1916
1917 r2hb_setup_callback(&r2net_hb_down, R2HB_NODE_DOWN_CB,
1918 r2net_hb_node_down_cb, NULL, R2NET_HB_PRI);
1919 r2hb_setup_callback(&r2net_hb_up, R2HB_NODE_UP_CB,
1920 r2net_hb_node_up_cb, NULL, R2NET_HB_PRI);
1921
1922 ret = r2hb_register_callback(NULL, &r2net_hb_up);
1923 if (ret == 0)
1924 ret = r2hb_register_callback(NULL, &r2net_hb_down);
1925
1926 if (ret)
1927 r2net_unregister_hb_callbacks();
1928
1929 return ret;
1930}
1931
1932/* ------------------------------------------------------------ */
1933
1934static int r2net_accept_one(struct socket *sock)
1935{
1936 int ret, slen;
1937 struct sockaddr_in sin;
1938 struct socket *new_sock = NULL;
1939 struct r2nm_node *node = NULL;
1940 struct r2nm_node *local_node = NULL;
1941 struct r2net_sock_container *sc = NULL;
1942 struct r2net_node *nn;
1943
1944 BUG_ON(sock == NULL);
1945 ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type,
1946 sock->sk->sk_protocol, &new_sock);
1947 if (ret)
1948 goto out;
1949
1950 new_sock->type = sock->type;
1951 new_sock->ops = sock->ops;
1952 ret = sock->ops->accept(sock, new_sock, O_NONBLOCK);
1953 if (ret < 0)
1954 goto out;
1955
1956 new_sock->sk->sk_allocation = GFP_ATOMIC;
1957
1958 ret = r2net_set_nodelay(new_sock);
1959 if (ret) {
1960 mlog(ML_ERROR, "setting TCP_NODELAY failed with %d\n", ret);
1961 goto out;
1962 }
1963
1964 slen = sizeof(sin);
1965 ret = new_sock->ops->getname(new_sock, (struct sockaddr *) &sin,
1966 &slen, 1);
1967 if (ret < 0)
1968 goto out;
1969
1970 node = r2nm_get_node_by_ip(sin.sin_addr.s_addr);
1971 if (node == NULL) {
1972 printk(KERN_NOTICE "ramster: Attempt to connect from unknown "
1973 "node at %pI4:%d\n", &sin.sin_addr.s_addr,
1974 ntohs(sin.sin_port));
1975 ret = -EINVAL;
1976 goto out;
1977 }
1978
1979 if (r2nm_this_node() >= node->nd_num) {
1980 local_node = r2nm_get_node_by_num(r2nm_this_node());
1981 printk(KERN_NOTICE "ramster: Unexpected connect attempt seen "
1982 "at node '%s' (%u, %pI4:%d) from node '%s' (%u, "
1983 "%pI4:%d)\n", local_node->nd_name, local_node->nd_num,
1984 &(local_node->nd_ipv4_address),
1985 ntohs(local_node->nd_ipv4_port), node->nd_name,
1986 node->nd_num, &sin.sin_addr.s_addr, ntohs(sin.sin_port));
1987 ret = -EINVAL;
1988 goto out;
1989 }
1990
1991 /* this happens all the time when the other node sees our heartbeat
1992 * and tries to connect before we see their heartbeat */
1993 if (!r2hb_check_node_heartbeating_from_callback(node->nd_num)) {
1994 mlog(ML_CONN, "attempt to connect from node '%s' at "
1995 "%pI4:%d but it isn't heartbeating\n",
1996 node->nd_name, &sin.sin_addr.s_addr,
1997 ntohs(sin.sin_port));
1998 ret = -EINVAL;
1999 goto out;
2000 }
2001
2002 nn = r2net_nn_from_num(node->nd_num);
2003
2004 spin_lock(&nn->nn_lock);
2005 if (nn->nn_sc)
2006 ret = -EBUSY;
2007 else
2008 ret = 0;
2009 spin_unlock(&nn->nn_lock);
2010 if (ret) {
2011 printk(KERN_NOTICE "ramster: Attempt to connect from node '%s' "
2012 "at %pI4:%d but it already has an open connection\n",
2013 node->nd_name, &sin.sin_addr.s_addr,
2014 ntohs(sin.sin_port));
2015 goto out;
2016 }
2017
2018 sc = sc_alloc(node);
2019 if (sc == NULL) {
2020 ret = -ENOMEM;
2021 goto out;
2022 }
2023
2024 sc->sc_sock = new_sock;
2025 new_sock = NULL;
2026
2027 spin_lock(&nn->nn_lock);
2028 atomic_set(&nn->nn_timeout, 0);
2029 r2net_set_nn_state(nn, sc, 0, 0);
2030 spin_unlock(&nn->nn_lock);
2031
2032 r2net_register_callbacks(sc->sc_sock->sk, sc);
2033 r2net_sc_queue_work(sc, &sc->sc_rx_work);
2034
2035 r2net_initialize_handshake();
2036 r2net_sendpage(sc, r2net_hand, sizeof(*r2net_hand));
2037
2038out:
2039 if (new_sock)
2040 sock_release(new_sock);
2041 if (node)
2042 r2nm_node_put(node);
2043 if (local_node)
2044 r2nm_node_put(local_node);
2045 if (sc)
2046 sc_put(sc);
2047 return ret;
2048}
2049
2050static void r2net_accept_many(struct work_struct *work)
2051{
2052 struct socket *sock = r2net_listen_sock;
2053 while (r2net_accept_one(sock) == 0)
2054 cond_resched();
2055}
2056
2057static void r2net_listen_data_ready(struct sock *sk, int bytes)
2058{
2059 void (*ready)(struct sock *sk, int bytes);
2060
2061 read_lock(&sk->sk_callback_lock);
2062 ready = sk->sk_user_data;
2063 if (ready == NULL) { /* check for teardown race */
2064 ready = sk->sk_data_ready;
2065 goto out;
2066 }
2067
2068 /* ->sk_data_ready is also called for a newly established child socket
2069 * before it has been accepted and the acceptor has set up their
2070 * data_ready.. we only want to queue listen work for our listening
2071 * socket */
2072 if (sk->sk_state == TCP_LISTEN) {
2073 mlog(ML_TCP, "bytes: %d\n", bytes);
2074 queue_work(r2net_wq, &r2net_listen_work);
2075 }
2076
2077out:
2078 read_unlock(&sk->sk_callback_lock);
2079 ready(sk, bytes);
2080}
2081
2082static int r2net_open_listening_sock(__be32 addr, __be16 port)
2083{
2084 struct socket *sock = NULL;
2085 int ret;
2086 struct sockaddr_in sin = {
2087 .sin_family = PF_INET,
2088 .sin_addr = { .s_addr = addr },
2089 .sin_port = port,
2090 };
2091
2092 ret = sock_create(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock);
2093 if (ret < 0) {
2094 printk(KERN_ERR "ramster: Error %d while creating socket\n",
2095 ret);
2096 goto out;
2097 }
2098
2099 sock->sk->sk_allocation = GFP_ATOMIC;
2100
2101 write_lock_bh(&sock->sk->sk_callback_lock);
2102 sock->sk->sk_user_data = sock->sk->sk_data_ready;
2103 sock->sk->sk_data_ready = r2net_listen_data_ready;
2104 write_unlock_bh(&sock->sk->sk_callback_lock);
2105
2106 r2net_listen_sock = sock;
2107 INIT_WORK(&r2net_listen_work, r2net_accept_many);
2108
2109 sock->sk->sk_reuse = SK_CAN_REUSE;
2110 ret = sock->ops->bind(sock, (struct sockaddr *)&sin, sizeof(sin));
2111 if (ret < 0) {
2112 printk(KERN_ERR "ramster: Error %d while binding socket at "
2113 "%pI4:%u\n", ret, &addr, ntohs(port));
2114 goto out;
2115 }
2116
2117 ret = sock->ops->listen(sock, 64);
2118 if (ret < 0)
2119 printk(KERN_ERR "ramster: Error %d while listening on %pI4:%u\n",
2120 ret, &addr, ntohs(port));
2121
2122out:
2123 if (ret) {
2124 r2net_listen_sock = NULL;
2125 if (sock)
2126 sock_release(sock);
2127 }
2128 return ret;
2129}
2130
2131/*
2132 * called from node manager when we should bring up our network listening
2133 * socket. node manager handles all the serialization to only call this
2134 * once and to match it with r2net_stop_listening(). note,
2135 * r2nm_this_node() doesn't work yet as we're being called while it
2136 * is being set up.
2137 */
2138int r2net_start_listening(struct r2nm_node *node)
2139{
2140 int ret = 0;
2141
2142 BUG_ON(r2net_wq != NULL);
2143 BUG_ON(r2net_listen_sock != NULL);
2144
2145 mlog(ML_KTHREAD, "starting r2net thread...\n");
2146 r2net_wq = create_singlethread_workqueue("r2net");
2147 if (r2net_wq == NULL) {
2148 mlog(ML_ERROR, "unable to launch r2net thread\n");
2149 return -ENOMEM; /* ? */
2150 }
2151
2152 ret = r2net_open_listening_sock(node->nd_ipv4_address,
2153 node->nd_ipv4_port);
2154 if (ret) {
2155 destroy_workqueue(r2net_wq);
2156 r2net_wq = NULL;
2157 }
2158
2159 return ret;
2160}
2161
2162/* again, r2nm_this_node() doesn't work here as we're involved in
2163 * tearing it down */
2164void r2net_stop_listening(struct r2nm_node *node)
2165{
2166 struct socket *sock = r2net_listen_sock;
2167 size_t i;
2168
2169 BUG_ON(r2net_wq == NULL);
2170 BUG_ON(r2net_listen_sock == NULL);
2171
2172 /* stop the listening socket from generating work */
2173 write_lock_bh(&sock->sk->sk_callback_lock);
2174 sock->sk->sk_data_ready = sock->sk->sk_user_data;
2175 sock->sk->sk_user_data = NULL;
2176 write_unlock_bh(&sock->sk->sk_callback_lock);
2177
2178 for (i = 0; i < ARRAY_SIZE(r2net_nodes); i++) {
2179 struct r2nm_node *node = r2nm_get_node_by_num(i);
2180 if (node) {
2181 r2net_disconnect_node(node);
2182 r2nm_node_put(node);
2183 }
2184 }
2185
2186 /* finish all work and tear down the work queue */
2187 mlog(ML_KTHREAD, "waiting for r2net thread to exit....\n");
2188 destroy_workqueue(r2net_wq);
2189 r2net_wq = NULL;
2190
2191 sock_release(r2net_listen_sock);
2192 r2net_listen_sock = NULL;
2193}
2194
2195void r2net_hb_node_up_manual(int node_num)
2196{
2197 struct r2nm_node dummy;
2198 if (r2nm_single_cluster == NULL)
2199 pr_err("ramster: cluster not alive, node_up_manual ignored\n");
2200 else {
2201 r2hb_manual_set_node_heartbeating(node_num);
2202 r2net_hb_node_up_cb(&dummy, node_num, NULL);
2203 }
2204}
2205
2206/* ------------------------------------------------------------ */
2207
2208int r2net_init(void)
2209{
2210 unsigned long i;
2211
2212 if (r2net_debugfs_init())
2213 return -ENOMEM;
2214
2215 r2net_hand = kzalloc(sizeof(struct r2net_handshake), GFP_KERNEL);
2216 r2net_keep_req = kzalloc(sizeof(struct r2net_msg), GFP_KERNEL);
2217 r2net_keep_resp = kzalloc(sizeof(struct r2net_msg), GFP_KERNEL);
2218 if (!r2net_hand || !r2net_keep_req || !r2net_keep_resp) {
2219 kfree(r2net_hand);
2220 kfree(r2net_keep_req);
2221 kfree(r2net_keep_resp);
2222 return -ENOMEM;
2223 }
2224
2225 r2net_hand->protocol_version = cpu_to_be64(R2NET_PROTOCOL_VERSION);
2226 r2net_hand->connector_id = cpu_to_be64(1);
2227
2228 r2net_keep_req->magic = cpu_to_be16(R2NET_MSG_KEEP_REQ_MAGIC);
2229 r2net_keep_resp->magic = cpu_to_be16(R2NET_MSG_KEEP_RESP_MAGIC);
2230
2231 for (i = 0; i < ARRAY_SIZE(r2net_nodes); i++) {
2232 struct r2net_node *nn = r2net_nn_from_num(i);
2233
2234 atomic_set(&nn->nn_timeout, 0);
2235 spin_lock_init(&nn->nn_lock);
2236 INIT_DELAYED_WORK(&nn->nn_connect_work, r2net_start_connect);
2237 INIT_DELAYED_WORK(&nn->nn_connect_expired,
2238 r2net_connect_expired);
2239 INIT_DELAYED_WORK(&nn->nn_still_up, r2net_still_up);
2240 /* until we see hb from a node we'll return einval */
2241 nn->nn_persistent_error = -ENOTCONN;
2242 init_waitqueue_head(&nn->nn_sc_wq);
2243 idr_init(&nn->nn_status_idr);
2244 INIT_LIST_HEAD(&nn->nn_status_list);
2245 }
2246
2247 return 0;
2248}
2249
2250void r2net_exit(void)
2251{
2252 kfree(r2net_hand);
2253 kfree(r2net_keep_req);
2254 kfree(r2net_keep_resp);
2255 r2net_debugfs_exit();
2256}