1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2001, 2002 Cluster File Systems, Inc.
5 * Author: Zach Brown <zab@zabbo.net>
6 * Author: Peter J. Braam <braam@clusterfs.com>
7 * Author: Phil Schwan <phil@clusterfs.com>
8 * Author: Eric Barton <eric@bartonsoftware.com>
10 * This file is part of Portals, http://www.sf.net/projects/sandiaportals/
12 * Portals is free software; you can redistribute it and/or
13 * modify it under the terms of version 2 of the GNU General Public
14 * License as published by the Free Software Foundation.
16 * Portals is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 * GNU General Public License for more details.
21 * You should have received a copy of the GNU General Public License
22 * along with Portals; if not, write to the Free Software
23 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
29 * LIB functions follow
33 ksocknal_read(nal_cb_t *nal, void *private, void *dst_addr,
34 user_ptr src_addr, size_t len)
36 CDEBUG(D_NET, LPX64": reading %ld bytes from %p -> %p\n",
37 nal->ni.nid, (long)len, src_addr, dst_addr);
39 memcpy( dst_addr, src_addr, len );
44 ksocknal_write(nal_cb_t *nal, void *private, user_ptr dst_addr,
45 void *src_addr, size_t len)
47 CDEBUG(D_NET, LPX64": writing %ld bytes from %p -> %p\n",
48 nal->ni.nid, (long)len, src_addr, dst_addr);
50 memcpy( dst_addr, src_addr, len );
55 ksocknal_callback (nal_cb_t * nal, void *private, lib_eq_t *eq,
58 CDEBUG(D_NET, LPX64": callback eq %p ev %p\n",
61 if (eq->event_callback != NULL)
62 eq->event_callback(ev);
68 ksocknal_malloc(nal_cb_t *nal, size_t len)
72 PORTAL_ALLOC(buf, len);
81 ksocknal_free(nal_cb_t *nal, void *buf, size_t len)
83 PORTAL_FREE(buf, len);
87 ksocknal_printf(nal_cb_t *nal, const char *fmt, ...)
93 vsnprintf (msg, sizeof (msg), fmt, ap); /* sprint safely */
96 msg[sizeof (msg) - 1] = 0; /* ensure terminated */
98 CDEBUG (D_NET, "%s", msg);
102 ksocknal_cli(nal_cb_t *nal, unsigned long *flags)
104 ksock_nal_data_t *data = nal->nal_data;
106 spin_lock(&data->ksnd_nal_cb_lock);
110 ksocknal_sti(nal_cb_t *nal, unsigned long *flags)
112 ksock_nal_data_t *data;
113 data = nal->nal_data;
115 spin_unlock(&data->ksnd_nal_cb_lock);
119 ksocknal_dist(nal_cb_t *nal, ptl_nid_t nid, unsigned long *dist)
121 /* I would guess that if ksocknal_get_peer (nid) == NULL,
122 and we're not routing, then 'nid' is very distant :) */
123 if ( nal->ni.nid == nid ) {
133 ksocknal_free_ltx (ksock_ltx_t *ltx)
135 atomic_dec(&ksocknal_data.ksnd_nactive_ltxs);
136 PORTAL_FREE(ltx, ltx->ltx_desc_size);
141 ksocknal_kvaddr_to_page (unsigned long vaddr)
145 if (vaddr >= VMALLOC_START &&
147 page = vmalloc_to_page ((void *)vaddr);
149 else if (vaddr >= PKMAP_BASE &&
150 vaddr < (PKMAP_BASE + LAST_PKMAP * PAGE_SIZE))
151 page = vmalloc_to_page ((void *)vaddr);
152 /* in 2.4 ^ just walks the page tables */
155 page = virt_to_page (vaddr);
166 ksocknal_send_iov (ksock_conn_t *conn, ksock_tx_t *tx)
168 struct socket *sock = conn->ksnc_sock;
169 struct iovec *iov = tx->tx_iov;
170 int fragsize = iov->iov_len;
171 unsigned long vaddr = (unsigned long)iov->iov_base;
172 int more = (tx->tx_niov > 1) ||
173 (tx->tx_nkiov > 0) ||
174 (!list_empty (&conn->ksnc_tx_queue));
176 int offset = vaddr & (PAGE_SIZE - 1);
177 int zcsize = MIN (fragsize, PAGE_SIZE - offset);
182 /* NB we can't trust socket ops to either consume our iovs
183 * or leave them alone, so we only send 1 frag at a time. */
184 LASSERT (fragsize <= tx->tx_resid);
185 LASSERT (tx->tx_niov > 0);
188 if (zcsize >= ksocknal_data.ksnd_zc_min_frag &&
189 (sock->sk->route_caps & NETIF_F_SG) &&
190 (sock->sk->route_caps & (NETIF_F_IP_CSUM | NETIF_F_NO_CSUM | NETIF_F_HW_CSUM)) &&
191 (page = ksocknal_kvaddr_to_page (vaddr)) != NULL) {
193 CDEBUG(D_NET, "vaddr %p, page %p->%p + offset %x for %d\n",
194 (void *)vaddr, page, page_address(page), offset, zcsize);
196 if (fragsize > zcsize) {
201 rc = tcp_sendpage_zccd(sock, page, offset, zcsize,
202 more ? (MSG_DONTWAIT | MSG_MORE) : MSG_DONTWAIT,
207 /* NB don't pass tx's iov; sendmsg may or may not update it */
208 struct iovec fragiov = { .iov_base = (void *)vaddr,
209 .iov_len = fragsize};
210 struct msghdr msg = {
217 .msg_flags = more ? (MSG_DONTWAIT | MSG_MORE) : MSG_DONTWAIT
219 mm_segment_t oldmm = get_fs();
222 rc = sock_sendmsg(sock, &msg, fragsize);
229 if (rc < iov->iov_len) {
230 /* didn't send whole iov entry... */
231 iov->iov_base = (void *)(vaddr + rc);
243 ksocknal_send_kiov (ksock_conn_t *conn, ksock_tx_t *tx)
245 struct socket *sock = conn->ksnc_sock;
246 ptl_kiov_t *kiov = tx->tx_kiov;
247 int fragsize = kiov->kiov_len;
248 struct page *page = kiov->kiov_page;
249 int offset = kiov->kiov_offset;
250 int more = (tx->tx_nkiov > 1) ||
251 (!list_empty (&conn->ksnc_tx_queue));
254 /* NB we can't trust socket ops to either consume our iovs
255 * or leave them alone, so we only send 1 frag at a time. */
256 LASSERT (fragsize <= tx->tx_resid);
257 LASSERT (offset + fragsize <= PAGE_SIZE);
258 LASSERT (tx->tx_niov == 0);
259 LASSERT (tx->tx_nkiov > 0);
262 if (fragsize >= ksocknal_data.ksnd_zc_min_frag &&
263 (sock->sk->route_caps & NETIF_F_SG) &&
264 (sock->sk->route_caps & (NETIF_F_IP_CSUM | NETIF_F_NO_CSUM | NETIF_F_HW_CSUM))) {
266 CDEBUG(D_NET, "page %p + offset %x for %d\n",
267 page, offset, fragsize);
269 rc = tcp_sendpage_zccd(sock, page, offset, fragsize,
270 more ? (MSG_DONTWAIT | MSG_MORE) : MSG_DONTWAIT,
275 char *addr = ((char *)kmap (page)) + offset;
276 struct iovec fragiov = {.iov_base = addr,
277 .iov_len = fragsize};
278 struct msghdr msg = {
285 .msg_flags = more ? (MSG_DONTWAIT | MSG_MORE) : MSG_DONTWAIT
287 mm_segment_t oldmm = get_fs();
290 rc = sock_sendmsg(sock, &msg, fragsize);
300 kiov->kiov_offset = offset + rc;
301 kiov->kiov_len = fragsize - rc;
312 ksocknal_transmit (ksock_conn_t *conn, ksock_tx_t *tx)
316 if (ksocknal_data.ksnd_stall_tx != 0) {
317 set_current_state (TASK_UNINTERRUPTIBLE);
318 schedule_timeout (ksocknal_data.ksnd_stall_tx * HZ);
321 LASSERT (tx->tx_resid != 0);
323 rc = ksocknal_getconnsock (conn);
325 LASSERT (conn->ksnc_closing);
330 if (ksocknal_data.ksnd_enomem_tx > 0) {
332 ksocknal_data.ksnd_enomem_tx--;
334 } else if (tx->tx_niov != 0) {
335 rc = ksocknal_send_iov (conn, tx);
337 rc = ksocknal_send_kiov (conn, tx);
341 /* Didn't write anything.
343 * NB: rc == 0 and rc == -EAGAIN both mean try
344 * again later (linux stack returns -EAGAIN for
345 * this, but Adaptech TOE returns 0).
347 * Also, sends never fail with -ENOMEM, just
348 * -EAGAIN, but with the added bonus that we can't
349 * expect write_space() to call us back to tell us
350 * when to try sending again. We use the
351 * SOCK_NOSPACE flag to diagnose... */
353 LASSERT(rc != -ENOMEM);
355 if (rc == 0 || rc == -EAGAIN) {
356 if (test_bit(SOCK_NOSPACE,
357 &conn->ksnc_sock->flags)) {
363 if ((counter & (-counter)) == counter)
364 CWARN("%d ENOMEM tx %p\n",
374 /* Consider the connection alive since we managed to chuck
375 * more data into it. Really, we'd like to consider it
376 * alive only when the peer ACKs something, but
377 * write_space() only gets called back while SOCK_NOSPACE
378 * is set. Instead, we presume peer death has occurred if
379 * the socket doesn't drain within a timout */
380 conn->ksnc_tx_deadline = jiffies +
381 ksocknal_data.ksnd_io_timeout * HZ;
382 conn->ksnc_peer->ksnp_last_alive = jiffies;
384 } while (tx->tx_resid != 0);
386 ksocknal_putconnsock (conn);
391 ksocknal_eager_ack (ksock_conn_t *conn)
394 mm_segment_t oldmm = get_fs();
395 struct socket *sock = conn->ksnc_sock;
397 /* Remind the socket to ACK eagerly. If I don't, the socket might
398 * think I'm about to send something it could piggy-back the ACK
399 * on, introducing delay in completing zero-copy sends in my
403 sock->ops->setsockopt (sock, SOL_TCP, TCP_QUICKACK,
404 (char *)&opt, sizeof (opt));
409 ksocknal_recv_iov (ksock_conn_t *conn)
411 struct iovec *iov = conn->ksnc_rx_iov;
412 int fragsize = iov->iov_len;
413 unsigned long vaddr = (unsigned long)iov->iov_base;
414 struct iovec fragiov = { .iov_base = (void *)vaddr,
415 .iov_len = fragsize};
416 struct msghdr msg = {
425 mm_segment_t oldmm = get_fs();
428 /* NB we can't trust socket ops to either consume our iovs
429 * or leave them alone, so we only receive 1 frag at a time. */
430 LASSERT (conn->ksnc_rx_niov > 0);
431 LASSERT (fragsize <= conn->ksnc_rx_nob_wanted);
434 rc = sock_recvmsg (conn->ksnc_sock, &msg, fragsize, MSG_DONTWAIT);
435 /* NB this is just a boolean............................^ */
441 /* received something... */
442 conn->ksnc_peer->ksnp_last_alive = jiffies;
443 conn->ksnc_rx_deadline = jiffies +
444 ksocknal_data.ksnd_io_timeout * HZ;
445 mb(); /* order with setting rx_started */
446 conn->ksnc_rx_started = 1;
448 conn->ksnc_rx_nob_wanted -= rc;
449 conn->ksnc_rx_nob_left -= rc;
452 iov->iov_base = (void *)(vaddr + rc);
453 iov->iov_len = fragsize - rc;
458 conn->ksnc_rx_niov--;
463 ksocknal_recv_kiov (ksock_conn_t *conn)
465 ptl_kiov_t *kiov = conn->ksnc_rx_kiov;
466 struct page *page = kiov->kiov_page;
467 int offset = kiov->kiov_offset;
468 int fragsize = kiov->kiov_len;
469 unsigned long vaddr = ((unsigned long)kmap (page)) + offset;
470 struct iovec fragiov = { .iov_base = (void *)vaddr,
471 .iov_len = fragsize};
472 struct msghdr msg = {
481 mm_segment_t oldmm = get_fs();
484 /* NB we can't trust socket ops to either consume our iovs
485 * or leave them alone, so we only receive 1 frag at a time. */
486 LASSERT (fragsize <= conn->ksnc_rx_nob_wanted);
487 LASSERT (conn->ksnc_rx_nkiov > 0);
488 LASSERT (offset + fragsize <= PAGE_SIZE);
491 rc = sock_recvmsg (conn->ksnc_sock, &msg, fragsize, MSG_DONTWAIT);
492 /* NB this is just a boolean............................^ */
500 /* received something... */
501 conn->ksnc_peer->ksnp_last_alive = jiffies;
502 conn->ksnc_rx_deadline = jiffies +
503 ksocknal_data.ksnd_io_timeout * HZ;
504 mb(); /* order with setting rx_started */
505 conn->ksnc_rx_started = 1;
507 conn->ksnc_rx_nob_wanted -= rc;
508 conn->ksnc_rx_nob_left -= rc;
511 kiov->kiov_offset = offset + rc;
512 kiov->kiov_len = fragsize - rc;
516 conn->ksnc_rx_kiov++;
517 conn->ksnc_rx_nkiov--;
522 ksocknal_receive (ksock_conn_t *conn)
524 /* Return 1 on success, 0 on EOF, < 0 on error.
525 * Caller checks ksnc_rx_nob_wanted to determine
526 * progress/completion. */
530 if (ksocknal_data.ksnd_stall_rx != 0) {
531 set_current_state (TASK_UNINTERRUPTIBLE);
532 schedule_timeout (ksocknal_data.ksnd_stall_rx * HZ);
535 rc = ksocknal_getconnsock (conn);
537 LASSERT (conn->ksnc_closing);
542 if (conn->ksnc_rx_niov != 0)
543 rc = ksocknal_recv_iov (conn);
545 rc = ksocknal_recv_kiov (conn);
548 /* error/EOF or partial receive */
551 } else if (rc == 0 && conn->ksnc_rx_started) {
552 /* EOF in the middle of a message */
558 /* Completed a fragment */
560 if (conn->ksnc_rx_nob_wanted == 0) {
561 /* Completed a message segment (header or payload) */
562 if ((ksocknal_data.ksnd_eager_ack & conn->ksnc_type) != 0 &&
563 (conn->ksnc_rx_state == SOCKNAL_RX_BODY ||
564 conn->ksnc_rx_state == SOCKNAL_RX_BODY_FWD)) {
565 /* Remind the socket to ack eagerly... */
566 ksocknal_eager_ack(conn);
573 ksocknal_putconnsock (conn);
579 ksocknal_zc_callback (zccd_t *zcd)
581 ksock_tx_t *tx = KSOCK_ZCCD_2_TX(zcd);
582 ksock_sched_t *sched = tx->tx_conn->ksnc_scheduler;
586 /* Schedule tx for cleanup (can't do it now due to lock conflicts) */
588 spin_lock_irqsave (&sched->kss_lock, flags);
590 list_add_tail (&tx->tx_list, &sched->kss_zctxdone_list);
591 wake_up (&sched->kss_waitq);
593 spin_unlock_irqrestore (&sched->kss_lock, flags);
599 ksocknal_tx_done (ksock_tx_t *tx, int asynch)
604 if (tx->tx_conn != NULL) {
605 /* This tx got queued on a conn; do the accounting... */
606 atomic_sub (tx->tx_nob, &tx->tx_conn->ksnc_tx_nob);
608 /* zero copy completion isn't always from
609 * process_transmit() so it needs to keep a ref on
612 ksocknal_put_conn (tx->tx_conn);
618 if (tx->tx_isfwd) { /* was a forwarded packet? */
619 kpr_fwd_done (&ksocknal_data.ksnd_router,
620 KSOCK_TX_2_KPR_FWD_DESC (tx), 0);
626 ltx = KSOCK_TX_2_KSOCK_LTX (tx);
628 lib_finalize (&ksocknal_lib, ltx->ltx_private, ltx->ltx_cookie);
630 ksocknal_free_ltx (ltx);
635 ksocknal_tx_launched (ksock_tx_t *tx)
638 if (atomic_read (&tx->tx_zccd.zccd_count) != 1) {
639 ksock_conn_t *conn = tx->tx_conn;
641 /* zccd skbufs are still in-flight. First take a ref on
642 * conn, so it hangs about for ksocknal_tx_done... */
643 atomic_inc (&conn->ksnc_refcount);
645 /* ...then drop the initial ref on zccd, so the zero copy
646 * callback can occur */
647 zccd_put (&tx->tx_zccd);
651 /* Any zero-copy-ness (if any) has completed; I can complete the
652 * transmit now, avoiding an extra schedule */
653 ksocknal_tx_done (tx, 0);
657 ksocknal_process_transmit (ksock_conn_t *conn, ksock_tx_t *tx)
662 rc = ksocknal_transmit (conn, tx);
664 CDEBUG (D_NET, "send(%d) %d\n", tx->tx_resid, rc);
666 if (tx->tx_resid == 0) {
667 /* Sent everything OK */
670 ksocknal_tx_launched (tx);
678 /* Queue on ksnd_enomem_conns for retry after a timeout */
679 spin_lock_irqsave(&ksocknal_data.ksnd_reaper_lock, flags);
681 /* enomem list takes over scheduler's ref... */
682 LASSERT (conn->ksnc_tx_scheduled);
683 list_add_tail(&conn->ksnc_tx_list,
684 &ksocknal_data.ksnd_enomem_conns);
685 if (!time_after_eq(jiffies + SOCKNAL_ENOMEM_RETRY,
686 ksocknal_data.ksnd_reaper_waketime))
687 wake_up (&ksocknal_data.ksnd_reaper_waitq);
689 spin_unlock_irqrestore(&ksocknal_data.ksnd_reaper_lock, flags);
696 if (!conn->ksnc_closing)
697 CERROR ("[%p] Error %d on write to "LPX64
698 " ip %d.%d.%d.%d:%d\n",conn, rc,
699 conn->ksnc_peer->ksnp_nid,
700 HIPQUAD(conn->ksnc_ipaddr),
703 ksocknal_close_conn_and_siblings (conn, rc);
704 ksocknal_tx_launched (tx);
710 ksocknal_launch_autoconnect_locked (ksock_route_t *route)
714 /* called holding write lock on ksnd_global_lock */
716 LASSERT (!route->ksnr_deleted);
717 LASSERT ((route->ksnr_connected & (1 << SOCKNAL_CONN_ANY)) == 0);
718 LASSERT ((route->ksnr_connected & KSNR_TYPED_ROUTES) != KSNR_TYPED_ROUTES);
719 LASSERT (!route->ksnr_connecting);
721 if (ksocknal_data.ksnd_typed_conns)
722 route->ksnr_connecting =
723 KSNR_TYPED_ROUTES & ~route->ksnr_connected;
725 route->ksnr_connecting = (1 << SOCKNAL_CONN_ANY);
727 atomic_inc (&route->ksnr_refcount); /* extra ref for asynchd */
729 spin_lock_irqsave (&ksocknal_data.ksnd_autoconnectd_lock, flags);
731 list_add_tail (&route->ksnr_connect_list,
732 &ksocknal_data.ksnd_autoconnectd_routes);
733 wake_up (&ksocknal_data.ksnd_autoconnectd_waitq);
735 spin_unlock_irqrestore (&ksocknal_data.ksnd_autoconnectd_lock, flags);
739 ksocknal_find_target_peer_locked (ksock_tx_t *tx, ptl_nid_t nid)
741 char ipbuf[PTL_NALFMT_SIZE];
742 ptl_nid_t target_nid;
744 ksock_peer_t *peer = ksocknal_find_peer_locked (nid);
750 CERROR ("Can't send packet to "LPX64
751 " %s: routed target is not a peer\n",
752 nid, portals_nid2str(SOCKNAL, nid, ipbuf));
756 rc = kpr_lookup (&ksocknal_data.ksnd_router, nid, tx->tx_nob,
759 CERROR ("Can't route to "LPX64" %s: router error %d\n",
760 nid, portals_nid2str(SOCKNAL, nid, ipbuf), rc);
764 peer = ksocknal_find_peer_locked (target_nid);
768 CERROR ("Can't send packet to "LPX64" %s: no peer entry\n",
769 target_nid, portals_nid2str(SOCKNAL, target_nid, ipbuf));
774 ksocknal_find_conn_locked (ksock_tx_t *tx, ksock_peer_t *peer)
776 struct list_head *tmp;
777 ksock_conn_t *typed = NULL;
779 ksock_conn_t *fallback = NULL;
782 /* Find the conn with the shortest tx queue */
783 list_for_each (tmp, &peer->ksnp_conns) {
784 ksock_conn_t *c = list_entry(tmp, ksock_conn_t, ksnc_list);
785 int nob = atomic_read(&c->ksnc_tx_nob);
787 LASSERT (!c->ksnc_closing);
789 if (fallback == NULL || nob < fnob) {
794 if (!ksocknal_data.ksnd_typed_conns)
797 switch (c->ksnc_type) {
800 case SOCKNAL_CONN_ANY:
802 case SOCKNAL_CONN_BULK_IN:
804 case SOCKNAL_CONN_BULK_OUT:
805 if (tx->tx_nob < ksocknal_data.ksnd_min_bulk)
808 case SOCKNAL_CONN_CONTROL:
809 if (tx->tx_nob >= ksocknal_data.ksnd_min_bulk)
814 if (typed == NULL || nob < tnob) {
820 /* prefer the typed selection */
821 return ((typed != NULL) ? typed : fallback);
825 ksocknal_queue_tx_locked (ksock_tx_t *tx, ksock_conn_t *conn)
828 ksock_sched_t *sched = conn->ksnc_scheduler;
830 /* called holding global lock (read or irq-write) and caller may
831 * not have dropped this lock between finding conn and calling me,
832 * so we don't need the {get,put}connsock dance to deref
834 LASSERT(!conn->ksnc_closing);
835 LASSERT(tx->tx_resid == tx->tx_nob);
837 CDEBUG (D_NET, "Sending to "LPX64" ip %d.%d.%d.%d:%d\n",
838 conn->ksnc_peer->ksnp_nid,
839 HIPQUAD(conn->ksnc_ipaddr),
842 atomic_add (tx->tx_nob, &conn->ksnc_tx_nob);
846 zccd_init (&tx->tx_zccd, ksocknal_zc_callback);
847 /* NB this sets 1 ref on zccd, so the callback can only occur after
848 * I've released this ref. */
850 spin_lock_irqsave (&sched->kss_lock, flags);
852 conn->ksnc_tx_deadline = jiffies +
853 ksocknal_data.ksnd_io_timeout * HZ;
854 mb(); /* order with list_add_tail */
856 list_add_tail (&tx->tx_list, &conn->ksnc_tx_queue);
858 if (conn->ksnc_tx_ready && /* able to send */
859 !conn->ksnc_tx_scheduled) { /* not scheduled to send */
860 /* +1 ref for scheduler */
861 atomic_inc (&conn->ksnc_refcount);
862 list_add_tail (&conn->ksnc_tx_list,
863 &sched->kss_tx_conns);
864 conn->ksnc_tx_scheduled = 1;
865 wake_up (&sched->kss_waitq);
868 spin_unlock_irqrestore (&sched->kss_lock, flags);
872 ksocknal_find_connectable_route_locked (ksock_peer_t *peer)
874 struct list_head *tmp;
875 ksock_route_t *route;
876 ksock_route_t *candidate = NULL;
880 list_for_each (tmp, &peer->ksnp_routes) {
881 route = list_entry (tmp, ksock_route_t, ksnr_list);
882 bits = route->ksnr_connected;
884 if ((bits & KSNR_TYPED_ROUTES) == KSNR_TYPED_ROUTES ||
885 (bits & (1 << SOCKNAL_CONN_ANY)) != 0 ||
886 route->ksnr_connecting != 0) {
887 /* All typed connections have been established, or
888 * an untyped connection has been established, or
889 * connections are currently being established */
894 /* too soon to retry this guy? */
895 if (!time_after_eq (jiffies, route->ksnr_timeout))
898 /* always do eager routes */
899 if (route->ksnr_eager)
902 if (candidate == NULL) {
903 /* If we don't find any other route that is fully
904 * connected or connecting, the first connectable
905 * route is returned. If it fails to connect, it
906 * will get placed at the end of the list */
911 return (found ? NULL : candidate);
915 ksocknal_find_connecting_route_locked (ksock_peer_t *peer)
917 struct list_head *tmp;
918 ksock_route_t *route;
920 list_for_each (tmp, &peer->ksnp_routes) {
921 route = list_entry (tmp, ksock_route_t, ksnr_list);
923 if (route->ksnr_connecting != 0)
931 ksocknal_launch_packet (ksock_tx_t *tx, ptl_nid_t nid)
936 ksock_route_t *route;
939 /* Ensure the frags we've been given EXACTLY match the number of
940 * bytes we want to send. Many TCP/IP stacks disregard any total
941 * size parameters passed to them and just look at the frags.
943 * We always expect at least 1 mapped fragment containing the
944 * complete portals header. */
945 LASSERT (lib_iov_nob (tx->tx_niov, tx->tx_iov) +
946 lib_kiov_nob (tx->tx_nkiov, tx->tx_kiov) == tx->tx_nob);
947 LASSERT (tx->tx_niov >= 1);
948 LASSERT (tx->tx_iov[0].iov_len >= sizeof (ptl_hdr_t));
950 CDEBUG (D_NET, "packet %p type %d, nob %d niov %d nkiov %d\n",
951 tx, ((ptl_hdr_t *)tx->tx_iov[0].iov_base)->type,
952 tx->tx_nob, tx->tx_niov, tx->tx_nkiov);
954 tx->tx_conn = NULL; /* only set when assigned a conn */
955 tx->tx_resid = tx->tx_nob;
956 tx->tx_hdr = (ptl_hdr_t *)tx->tx_iov[0].iov_base;
958 g_lock = &ksocknal_data.ksnd_global_lock;
961 peer = ksocknal_find_target_peer_locked (tx, nid);
963 read_unlock (g_lock);
964 return (-EHOSTUNREACH);
967 if (ksocknal_find_connectable_route_locked(peer) == NULL) {
968 conn = ksocknal_find_conn_locked (tx, peer);
970 /* I've got no autoconnect routes that need to be
971 * connecting and I do have an actual connection... */
972 ksocknal_queue_tx_locked (tx, conn);
973 read_unlock (g_lock);
978 /* Making one or more connections; I'll need a write lock... */
980 atomic_inc (&peer->ksnp_refcount); /* +1 ref for me while I unlock */
981 read_unlock (g_lock);
982 write_lock_irqsave (g_lock, flags);
984 if (peer->ksnp_closing) { /* peer deleted as I blocked! */
985 write_unlock_irqrestore (g_lock, flags);
986 ksocknal_put_peer (peer);
987 return (-EHOSTUNREACH);
989 ksocknal_put_peer (peer); /* drop ref I got above */
992 /* launch any/all autoconnections that need it */
993 route = ksocknal_find_connectable_route_locked (peer);
997 ksocknal_launch_autoconnect_locked (route);
1000 conn = ksocknal_find_conn_locked (tx, peer);
1002 /* Connection exists; queue message on it */
1003 ksocknal_queue_tx_locked (tx, conn);
1004 write_unlock_irqrestore (g_lock, flags);
1008 route = ksocknal_find_connecting_route_locked (peer);
1009 if (route != NULL) {
1010 /* At least 1 connection is being established; queue the
1012 list_add_tail (&tx->tx_list, &peer->ksnp_tx_queue);
1013 write_unlock_irqrestore (g_lock, flags);
1017 write_unlock_irqrestore (g_lock, flags);
1018 return (-EHOSTUNREACH);
1022 ksocknal_sendmsg(nal_cb_t *nal,
1029 unsigned int payload_niov,
1030 struct iovec *payload_iov,
1031 ptl_kiov_t *payload_kiov,
1038 /* NB 'private' is different depending on what we're sending.
1039 * Just ignore it... */
1041 CDEBUG(D_NET, "sending "LPSZ" bytes in %d frags to nid:"LPX64
1042 " pid %d\n", payload_nob, payload_niov, nid , pid);
1044 LASSERT (payload_nob == 0 || payload_niov > 0);
1045 LASSERT (payload_niov <= PTL_MD_MAX_IOV);
1047 /* It must be OK to kmap() if required */
1048 LASSERT (payload_kiov == NULL || !in_interrupt ());
1049 /* payload is either all vaddrs or all pages */
1050 LASSERT (!(payload_kiov != NULL && payload_iov != NULL));
1052 if (payload_iov != NULL)
1053 desc_size = offsetof(ksock_ltx_t, ltx_iov[1 + payload_niov]);
1055 desc_size = offsetof(ksock_ltx_t, ltx_kiov[payload_niov]);
1057 if (in_interrupt() ||
1058 type == PTL_MSG_ACK ||
1059 type == PTL_MSG_REPLY) {
1060 /* Can't block if in interrupt or responding to an incoming
1062 PORTAL_ALLOC_ATOMIC(ltx, desc_size);
1064 PORTAL_ALLOC(ltx, desc_size);
1068 CERROR("Can't allocate tx desc type %d size %d %s\n",
1069 type, desc_size, in_interrupt() ? "(intr)" : "");
1070 return (PTL_NOSPACE);
1073 atomic_inc(&ksocknal_data.ksnd_nactive_ltxs);
1075 ltx->ltx_desc_size = desc_size;
1077 /* We always have 1 mapped frag for the header */
1078 ltx->ltx_tx.tx_iov = ltx->ltx_iov;
1079 ltx->ltx_iov[0].iov_base = <x->ltx_hdr;
1080 ltx->ltx_iov[0].iov_len = sizeof(*hdr);
1081 ltx->ltx_hdr = *hdr;
1083 ltx->ltx_private = private;
1084 ltx->ltx_cookie = cookie;
1086 ltx->ltx_tx.tx_isfwd = 0;
1087 ltx->ltx_tx.tx_nob = sizeof (*hdr) + payload_nob;
1089 if (payload_iov != NULL) {
1090 /* payload is all mapped */
1091 ltx->ltx_tx.tx_kiov = NULL;
1092 ltx->ltx_tx.tx_nkiov = 0;
1094 ltx->ltx_tx.tx_niov = 1 + payload_niov;
1096 memcpy(ltx->ltx_iov + 1, payload_iov,
1097 payload_niov * sizeof (*payload_iov));
1100 /* payload is all pages */
1101 ltx->ltx_tx.tx_kiov = ltx->ltx_kiov;
1102 ltx->ltx_tx.tx_nkiov = payload_niov;
1104 ltx->ltx_tx.tx_niov = 1;
1106 memcpy(ltx->ltx_kiov, payload_kiov,
1107 payload_niov * sizeof (*payload_kiov));
1110 rc = ksocknal_launch_packet(<x->ltx_tx, nid);
1114 ksocknal_free_ltx(ltx);
1119 ksocknal_send (nal_cb_t *nal, void *private, lib_msg_t *cookie,
1120 ptl_hdr_t *hdr, int type, ptl_nid_t nid, ptl_pid_t pid,
1121 unsigned int payload_niov, struct iovec *payload_iov,
1124 return (ksocknal_sendmsg(nal, private, cookie,
1125 hdr, type, nid, pid,
1126 payload_niov, payload_iov, NULL,
1131 ksocknal_send_pages (nal_cb_t *nal, void *private, lib_msg_t *cookie,
1132 ptl_hdr_t *hdr, int type, ptl_nid_t nid, ptl_pid_t pid,
1133 unsigned int payload_niov, ptl_kiov_t *payload_kiov,
1136 return (ksocknal_sendmsg(nal, private, cookie,
1137 hdr, type, nid, pid,
1138 payload_niov, NULL, payload_kiov,
1143 ksocknal_fwd_packet (void *arg, kpr_fwd_desc_t *fwd)
1145 ptl_nid_t nid = fwd->kprfd_gateway_nid;
1146 ksock_tx_t *tx = (ksock_tx_t *)&fwd->kprfd_scratch;
1149 CDEBUG (D_NET, "Forwarding [%p] -> "LPX64" ("LPX64"))\n", fwd,
1150 fwd->kprfd_gateway_nid, fwd->kprfd_target_nid);
1152 /* I'm the gateway; must be the last hop */
1153 if (nid == ksocknal_lib.ni.nid)
1154 nid = fwd->kprfd_target_nid;
1156 tx->tx_isfwd = 1; /* This is a forwarding packet */
1157 tx->tx_nob = fwd->kprfd_nob;
1158 tx->tx_niov = fwd->kprfd_niov;
1159 tx->tx_iov = fwd->kprfd_iov;
1163 rc = ksocknal_launch_packet (tx, nid);
1165 kpr_fwd_done (&ksocknal_data.ksnd_router, fwd, rc);
1169 ksocknal_thread_start (int (*fn)(void *arg), void *arg)
1171 long pid = kernel_thread (fn, arg, 0);
1176 atomic_inc (&ksocknal_data.ksnd_nthreads);
1181 ksocknal_thread_fini (void)
1183 atomic_dec (&ksocknal_data.ksnd_nthreads);
1187 ksocknal_fmb_callback (void *arg, int error)
1189 ksock_fmb_t *fmb = (ksock_fmb_t *)arg;
1190 ksock_fmb_pool_t *fmp = fmb->fmb_pool;
1191 ptl_hdr_t *hdr = (ptl_hdr_t *) page_address(fmb->fmb_pages[0]);
1192 ksock_conn_t *conn = NULL;
1193 ksock_sched_t *sched;
1194 unsigned long flags;
1195 char ipbuf[PTL_NALFMT_SIZE];
1198 CERROR("Failed to route packet from "
1199 LPX64" %s to "LPX64" %s: %d\n",
1200 NTOH__u64(hdr->src_nid),
1201 portals_nid2str(SOCKNAL, NTOH__u64(hdr->src_nid), ipbuf),
1202 NTOH__u64(hdr->dest_nid),
1203 portals_nid2str(SOCKNAL, NTOH__u64(hdr->dest_nid), ipbuf),
1206 CDEBUG (D_NET, "routed packet from "LPX64" to "LPX64": OK\n",
1207 NTOH__u64 (hdr->src_nid), NTOH__u64 (hdr->dest_nid));
1209 /* drop peer ref taken on init */
1210 ksocknal_put_peer (fmb->fmb_peer);
1212 spin_lock_irqsave (&fmp->fmp_lock, flags);
1214 list_add (&fmb->fmb_list, &fmp->fmp_idle_fmbs);
1215 fmp->fmp_nactive_fmbs--;
1217 if (!list_empty (&fmp->fmp_blocked_conns)) {
1218 conn = list_entry (fmb->fmb_pool->fmp_blocked_conns.next,
1219 ksock_conn_t, ksnc_rx_list);
1220 list_del (&conn->ksnc_rx_list);
1223 spin_unlock_irqrestore (&fmp->fmp_lock, flags);
1228 CDEBUG (D_NET, "Scheduling conn %p\n", conn);
1229 LASSERT (conn->ksnc_rx_scheduled);
1230 LASSERT (conn->ksnc_rx_state == SOCKNAL_RX_FMB_SLEEP);
1232 conn->ksnc_rx_state = SOCKNAL_RX_GET_FMB;
1234 sched = conn->ksnc_scheduler;
1236 spin_lock_irqsave (&sched->kss_lock, flags);
1238 list_add_tail (&conn->ksnc_rx_list, &sched->kss_rx_conns);
1239 wake_up (&sched->kss_waitq);
1241 spin_unlock_irqrestore (&sched->kss_lock, flags);
1245 ksocknal_get_idle_fmb (ksock_conn_t *conn)
1247 int payload_nob = conn->ksnc_rx_nob_left;
1248 int packet_nob = sizeof (ptl_hdr_t) + payload_nob;
1249 unsigned long flags;
1250 ksock_fmb_pool_t *pool;
1253 LASSERT (conn->ksnc_rx_state == SOCKNAL_RX_GET_FMB);
1254 LASSERT (kpr_routing(&ksocknal_data.ksnd_router));
1256 if (packet_nob <= SOCKNAL_SMALL_FWD_PAGES * PAGE_SIZE)
1257 pool = &ksocknal_data.ksnd_small_fmp;
1259 pool = &ksocknal_data.ksnd_large_fmp;
1261 spin_lock_irqsave (&pool->fmp_lock, flags);
1263 if (!list_empty (&pool->fmp_idle_fmbs)) {
1264 fmb = list_entry(pool->fmp_idle_fmbs.next,
1265 ksock_fmb_t, fmb_list);
1266 list_del (&fmb->fmb_list);
1267 pool->fmp_nactive_fmbs++;
1268 spin_unlock_irqrestore (&pool->fmp_lock, flags);
1273 /* deschedule until fmb free */
1275 conn->ksnc_rx_state = SOCKNAL_RX_FMB_SLEEP;
1277 list_add_tail (&conn->ksnc_rx_list,
1278 &pool->fmp_blocked_conns);
1280 spin_unlock_irqrestore (&pool->fmp_lock, flags);
1285 ksocknal_init_fmb (ksock_conn_t *conn, ksock_fmb_t *fmb)
1287 int payload_nob = conn->ksnc_rx_nob_left;
1288 int packet_nob = sizeof (ptl_hdr_t) + payload_nob;
1289 ptl_nid_t dest_nid = NTOH__u64 (conn->ksnc_hdr.dest_nid);
1290 int niov; /* at least the header */
1293 LASSERT (conn->ksnc_rx_scheduled);
1294 LASSERT (conn->ksnc_rx_state == SOCKNAL_RX_GET_FMB);
1295 LASSERT (conn->ksnc_rx_nob_wanted == conn->ksnc_rx_nob_left);
1296 LASSERT (payload_nob >= 0);
1297 LASSERT (packet_nob <= fmb->fmb_npages * PAGE_SIZE);
1298 LASSERT (sizeof (ptl_hdr_t) < PAGE_SIZE);
1300 /* Got a forwarding buffer; copy the header we just read into the
1301 * forwarding buffer. If there's payload, start reading reading it
1302 * into the buffer, otherwise the forwarding buffer can be kicked
1305 * NB fmb->fmb_iov spans the WHOLE packet.
1306 * conn->ksnc_rx_iov spans just the payload.
1308 fmb->fmb_iov[0].iov_base = page_address (fmb->fmb_pages[0]);
1311 memcpy (fmb->fmb_iov[0].iov_base, &conn->ksnc_hdr, sizeof (ptl_hdr_t));
1313 /* Take a ref on the conn's peer to prevent module unload before
1314 * forwarding completes. NB we ref peer and not conn since because
1315 * all refs on conn after it has been closed must remove themselves
1317 fmb->fmb_peer = conn->ksnc_peer;
1318 atomic_inc (&conn->ksnc_peer->ksnp_refcount);
1320 if (payload_nob == 0) { /* got complete packet already */
1321 CDEBUG (D_NET, "%p "LPX64"->"LPX64" %d fwd_start (immediate)\n",
1322 conn, NTOH__u64 (conn->ksnc_hdr.src_nid),
1323 dest_nid, packet_nob);
1325 fmb->fmb_iov[0].iov_len = sizeof (ptl_hdr_t);
1327 kpr_fwd_init (&fmb->fmb_fwd, dest_nid,
1328 packet_nob, 1, fmb->fmb_iov,
1329 ksocknal_fmb_callback, fmb);
1331 /* forward it now */
1332 kpr_fwd_start (&ksocknal_data.ksnd_router, &fmb->fmb_fwd);
1334 ksocknal_new_packet (conn, 0); /* on to next packet */
1339 if (packet_nob <= PAGE_SIZE) { /* whole packet fits in first page */
1340 fmb->fmb_iov[0].iov_len = packet_nob;
1342 fmb->fmb_iov[0].iov_len = PAGE_SIZE;
1343 nob = packet_nob - PAGE_SIZE;
1346 LASSERT (niov < fmb->fmb_npages);
1347 fmb->fmb_iov[niov].iov_base =
1348 page_address (fmb->fmb_pages[niov]);
1349 fmb->fmb_iov[niov].iov_len = MIN (PAGE_SIZE, nob);
1355 kpr_fwd_init (&fmb->fmb_fwd, dest_nid,
1356 packet_nob, niov, fmb->fmb_iov,
1357 ksocknal_fmb_callback, fmb);
1359 conn->ksnc_cookie = fmb; /* stash fmb for later */
1360 conn->ksnc_rx_state = SOCKNAL_RX_BODY_FWD; /* read in the payload */
1362 /* payload is desc's iov-ed buffer, but skipping the hdr */
1363 LASSERT (niov <= sizeof (conn->ksnc_rx_iov_space) /
1364 sizeof (struct iovec));
1366 conn->ksnc_rx_iov = (struct iovec *)&conn->ksnc_rx_iov_space;
1367 conn->ksnc_rx_iov[0].iov_base =
1368 (void *)(((unsigned long)fmb->fmb_iov[0].iov_base) +
1369 sizeof (ptl_hdr_t));
1370 conn->ksnc_rx_iov[0].iov_len =
1371 fmb->fmb_iov[0].iov_len - sizeof (ptl_hdr_t);
1374 memcpy(&conn->ksnc_rx_iov[1], &fmb->fmb_iov[1],
1375 (niov - 1) * sizeof (struct iovec));
1377 conn->ksnc_rx_niov = niov;
1379 CDEBUG (D_NET, "%p "LPX64"->"LPX64" %d reading body\n", conn,
1380 NTOH__u64 (conn->ksnc_hdr.src_nid), dest_nid, payload_nob);
1385 ksocknal_fwd_parse (ksock_conn_t *conn)
1388 ptl_nid_t dest_nid = NTOH__u64 (conn->ksnc_hdr.dest_nid);
1389 ptl_nid_t src_nid = NTOH__u64 (conn->ksnc_hdr.src_nid);
1390 int body_len = NTOH__u32 (conn->ksnc_hdr.payload_length);
1391 char str[PTL_NALFMT_SIZE];
1393 CDEBUG (D_NET, "%p "LPX64"->"LPX64" %d parsing header\n", conn,
1394 src_nid, dest_nid, conn->ksnc_rx_nob_left);
1396 LASSERT (conn->ksnc_rx_state == SOCKNAL_RX_HEADER);
1397 LASSERT (conn->ksnc_rx_scheduled);
1399 if (body_len < 0) { /* length corrupt (overflow) */
1400 CERROR("dropping packet from "LPX64" (%s) for "LPX64" (%s): "
1401 "packet size %d illegal\n",
1402 src_nid, portals_nid2str(TCPNAL, src_nid, str),
1403 dest_nid, portals_nid2str(TCPNAL, dest_nid, str),
1406 ksocknal_new_packet (conn, 0); /* on to new packet */
1410 if (!kpr_routing(&ksocknal_data.ksnd_router)) { /* not forwarding */
1411 CERROR("dropping packet from "LPX64" (%s) for "LPX64
1412 " (%s): not forwarding\n",
1413 src_nid, portals_nid2str(TCPNAL, src_nid, str),
1414 dest_nid, portals_nid2str(TCPNAL, dest_nid, str));
1415 /* on to new packet (skip this one's body) */
1416 ksocknal_new_packet (conn, body_len);
1420 if (body_len > PTL_MTU) { /* too big to forward */
1421 CERROR ("dropping packet from "LPX64" (%s) for "LPX64
1422 "(%s): packet size %d too big\n",
1423 src_nid, portals_nid2str(TCPNAL, src_nid, str),
1424 dest_nid, portals_nid2str(TCPNAL, dest_nid, str),
1426 /* on to new packet (skip this one's body) */
1427 ksocknal_new_packet (conn, body_len);
1431 /* should have gone direct */
1432 peer = ksocknal_get_peer (conn->ksnc_hdr.dest_nid);
1434 CERROR ("dropping packet from "LPX64" (%s) for "LPX64
1435 "(%s): target is a peer\n",
1436 src_nid, portals_nid2str(TCPNAL, src_nid, str),
1437 dest_nid, portals_nid2str(TCPNAL, dest_nid, str));
1438 ksocknal_put_peer (peer); /* drop ref from get above */
1440 /* on to next packet (skip this one's body) */
1441 ksocknal_new_packet (conn, body_len);
1445 conn->ksnc_rx_state = SOCKNAL_RX_GET_FMB; /* Getting FMB now */
1446 conn->ksnc_rx_nob_left = body_len; /* stash packet size */
1447 conn->ksnc_rx_nob_wanted = body_len; /* (no slop) */
1451 ksocknal_new_packet (ksock_conn_t *conn, int nob_to_skip)
1453 static char ksocknal_slop_buffer[4096];
1459 if (nob_to_skip == 0) { /* right at next packet boundary now */
1460 conn->ksnc_rx_started = 0;
1461 mb (); /* racing with timeout thread */
1463 conn->ksnc_rx_state = SOCKNAL_RX_HEADER;
1464 conn->ksnc_rx_nob_wanted = sizeof (ptl_hdr_t);
1465 conn->ksnc_rx_nob_left = sizeof (ptl_hdr_t);
1467 conn->ksnc_rx_iov = (struct iovec *)&conn->ksnc_rx_iov_space;
1468 conn->ksnc_rx_iov[0].iov_base = (char *)&conn->ksnc_hdr;
1469 conn->ksnc_rx_iov[0].iov_len = sizeof (ptl_hdr_t);
1470 conn->ksnc_rx_niov = 1;
1472 conn->ksnc_rx_kiov = NULL;
1473 conn->ksnc_rx_nkiov = 0;
1477 /* Set up to skip as much a possible now. If there's more left
1478 * (ran out of iov entries) we'll get called again */
1480 conn->ksnc_rx_state = SOCKNAL_RX_SLOP;
1481 conn->ksnc_rx_nob_left = nob_to_skip;
1482 conn->ksnc_rx_iov = (struct iovec *)&conn->ksnc_rx_iov_space;
1487 nob = MIN (nob_to_skip, sizeof (ksocknal_slop_buffer));
1489 conn->ksnc_rx_iov[niov].iov_base = ksocknal_slop_buffer;
1490 conn->ksnc_rx_iov[niov].iov_len = nob;
1495 } while (nob_to_skip != 0 && /* mustn't overflow conn's rx iov */
1496 niov < sizeof(conn->ksnc_rx_iov_space) / sizeof (struct iovec));
1498 conn->ksnc_rx_niov = niov;
1499 conn->ksnc_rx_kiov = NULL;
1500 conn->ksnc_rx_nkiov = 0;
1501 conn->ksnc_rx_nob_wanted = skipped;
1506 ksocknal_process_receive (ksock_conn_t *conn)
1511 LASSERT (atomic_read (&conn->ksnc_refcount) > 0);
1513 /* doesn't need a forwarding buffer */
1514 if (conn->ksnc_rx_state != SOCKNAL_RX_GET_FMB)
1518 fmb = ksocknal_get_idle_fmb (conn);
1520 /* conn descheduled waiting for idle fmb */
1524 if (ksocknal_init_fmb (conn, fmb)) {
1525 /* packet forwarded */
1530 /* NB: sched lock NOT held */
1531 LASSERT (conn->ksnc_rx_state == SOCKNAL_RX_HEADER ||
1532 conn->ksnc_rx_state == SOCKNAL_RX_BODY ||
1533 conn->ksnc_rx_state == SOCKNAL_RX_BODY_FWD ||
1534 conn->ksnc_rx_state == SOCKNAL_RX_SLOP);
1536 LASSERT (conn->ksnc_rx_nob_wanted > 0);
1538 rc = ksocknal_receive(conn);
1541 LASSERT (rc != -EAGAIN);
1544 CWARN ("[%p] EOF from "LPX64" ip %d.%d.%d.%d:%d\n",
1545 conn, conn->ksnc_peer->ksnp_nid,
1546 HIPQUAD(conn->ksnc_ipaddr),
1548 else if (!conn->ksnc_closing)
1549 CERROR ("[%p] Error %d on read from "LPX64
1550 " ip %d.%d.%d.%d:%d\n",
1551 conn, rc, conn->ksnc_peer->ksnp_nid,
1552 HIPQUAD(conn->ksnc_ipaddr),
1555 ksocknal_close_conn_and_siblings (conn, rc);
1556 return (rc == 0 ? -ESHUTDOWN : rc);
1559 if (conn->ksnc_rx_nob_wanted != 0) {
1564 switch (conn->ksnc_rx_state) {
1565 case SOCKNAL_RX_HEADER:
1566 if (conn->ksnc_hdr.type != HTON__u32(PTL_MSG_HELLO) &&
1567 NTOH__u64(conn->ksnc_hdr.dest_nid) != ksocknal_lib.ni.nid) {
1568 /* This packet isn't for me */
1569 ksocknal_fwd_parse (conn);
1570 switch (conn->ksnc_rx_state) {
1571 case SOCKNAL_RX_HEADER: /* skipped (zero payload) */
1572 return (0); /* => come back later */
1573 case SOCKNAL_RX_SLOP: /* skipping packet's body */
1574 goto try_read; /* => go read it */
1575 case SOCKNAL_RX_GET_FMB: /* forwarding */
1576 goto get_fmb; /* => go get a fwd msg buffer */
1583 /* sets wanted_len, iovs etc */
1584 lib_parse(&ksocknal_lib, &conn->ksnc_hdr, conn);
1586 if (conn->ksnc_rx_nob_wanted != 0) { /* need to get payload? */
1587 conn->ksnc_rx_state = SOCKNAL_RX_BODY;
1588 goto try_read; /* go read the payload */
1590 /* Fall through (completed packet for me) */
1592 case SOCKNAL_RX_BODY:
1593 /* payload all received */
1594 lib_finalize(&ksocknal_lib, NULL, conn->ksnc_cookie);
1597 case SOCKNAL_RX_SLOP:
1598 /* starting new packet? */
1599 if (ksocknal_new_packet (conn, conn->ksnc_rx_nob_left))
1600 return (0); /* come back later */
1601 goto try_read; /* try to finish reading slop now */
1603 case SOCKNAL_RX_BODY_FWD:
1604 /* payload all received */
1605 CDEBUG (D_NET, "%p "LPX64"->"LPX64" %d fwd_start (got body)\n",
1606 conn, NTOH__u64 (conn->ksnc_hdr.src_nid),
1607 NTOH__u64 (conn->ksnc_hdr.dest_nid),
1608 conn->ksnc_rx_nob_left);
1610 /* forward the packet. NB ksocknal_init_fmb() put fmb into
1611 * conn->ksnc_cookie */
1612 fmb = (ksock_fmb_t *)conn->ksnc_cookie;
1613 kpr_fwd_start (&ksocknal_data.ksnd_router, &fmb->fmb_fwd);
1615 /* no slop in forwarded packets */
1616 LASSERT (conn->ksnc_rx_nob_left == 0);
1618 ksocknal_new_packet (conn, 0); /* on to next packet */
1619 return (0); /* (later) */
1627 return (-EINVAL); /* keep gcc happy */
1631 ksocknal_recv (nal_cb_t *nal, void *private, lib_msg_t *msg,
1632 unsigned int niov, struct iovec *iov, size_t mlen, size_t rlen)
1634 ksock_conn_t *conn = (ksock_conn_t *)private;
1636 LASSERT (mlen <= rlen);
1637 LASSERT (niov <= PTL_MD_MAX_IOV);
1639 conn->ksnc_cookie = msg;
1640 conn->ksnc_rx_nob_wanted = mlen;
1641 conn->ksnc_rx_nob_left = rlen;
1643 conn->ksnc_rx_nkiov = 0;
1644 conn->ksnc_rx_kiov = NULL;
1645 conn->ksnc_rx_niov = niov;
1646 conn->ksnc_rx_iov = conn->ksnc_rx_iov_space.iov;
1647 memcpy (conn->ksnc_rx_iov, iov, niov * sizeof (*iov));
1650 lib_iov_nob (conn->ksnc_rx_niov, conn->ksnc_rx_iov) +
1651 lib_kiov_nob (conn->ksnc_rx_nkiov, conn->ksnc_rx_kiov));
1657 ksocknal_recv_pages (nal_cb_t *nal, void *private, lib_msg_t *msg,
1658 unsigned int niov, ptl_kiov_t *kiov, size_t mlen, size_t rlen)
1660 ksock_conn_t *conn = (ksock_conn_t *)private;
1662 LASSERT (mlen <= rlen);
1663 LASSERT (niov <= PTL_MD_MAX_IOV);
1665 conn->ksnc_cookie = msg;
1666 conn->ksnc_rx_nob_wanted = mlen;
1667 conn->ksnc_rx_nob_left = rlen;
1669 conn->ksnc_rx_niov = 0;
1670 conn->ksnc_rx_iov = NULL;
1671 conn->ksnc_rx_nkiov = niov;
1672 conn->ksnc_rx_kiov = conn->ksnc_rx_iov_space.kiov;
1673 memcpy (conn->ksnc_rx_kiov, kiov, niov * sizeof (*kiov));
1676 lib_iov_nob (conn->ksnc_rx_niov, conn->ksnc_rx_iov) +
1677 lib_kiov_nob (conn->ksnc_rx_nkiov, conn->ksnc_rx_kiov));
1682 int ksocknal_scheduler (void *arg)
1684 ksock_sched_t *sched = (ksock_sched_t *)arg;
1687 unsigned long flags;
1690 int id = sched - ksocknal_data.ksnd_schedulers;
1693 snprintf (name, sizeof (name),"ksocknald_%02d", id);
1694 kportal_daemonize (name);
1695 kportal_blockallsigs ();
1697 #if (CONFIG_SMP && CPU_AFFINITY)
1698 if ((cpu_online_map & (1 << id)) != 0) {
1700 current->cpus_allowed = (1 << id);
1702 set_cpus_allowed (current, 1<<id);
1705 CERROR ("Can't set CPU affinity for %s\n", name);
1707 #endif /* CONFIG_SMP && CPU_AFFINITY */
1709 spin_lock_irqsave (&sched->kss_lock, flags);
1711 while (!ksocknal_data.ksnd_shuttingdown) {
1712 int did_something = 0;
1714 /* Ensure I progress everything semi-fairly */
1716 if (!list_empty (&sched->kss_rx_conns)) {
1717 conn = list_entry(sched->kss_rx_conns.next,
1718 ksock_conn_t, ksnc_rx_list);
1719 list_del(&conn->ksnc_rx_list);
1721 LASSERT(conn->ksnc_rx_scheduled);
1722 LASSERT(conn->ksnc_rx_ready);
1724 /* clear rx_ready in case receive isn't complete.
1725 * Do it BEFORE we call process_recv, since
1726 * data_ready can set it any time after we release
1728 conn->ksnc_rx_ready = 0;
1729 spin_unlock_irqrestore(&sched->kss_lock, flags);
1731 rc = ksocknal_process_receive(conn);
1733 spin_lock_irqsave(&sched->kss_lock, flags);
1735 /* I'm the only one that can clear this flag */
1736 LASSERT(conn->ksnc_rx_scheduled);
1738 /* Did process_receive get everything it wanted? */
1740 conn->ksnc_rx_ready = 1;
1742 if (conn->ksnc_rx_state == SOCKNAL_RX_FMB_SLEEP ||
1743 conn->ksnc_rx_state == SOCKNAL_RX_GET_FMB) {
1744 /* Conn blocked for a forwarding buffer.
1745 * It will get queued for my attention when
1746 * one becomes available (and it might just
1747 * already have been!). Meanwhile my ref
1748 * on it stays put. */
1749 } else if (conn->ksnc_rx_ready) {
1750 /* reschedule for rx */
1751 list_add_tail (&conn->ksnc_rx_list,
1752 &sched->kss_rx_conns);
1754 conn->ksnc_rx_scheduled = 0;
1756 ksocknal_put_conn(conn);
1762 if (!list_empty (&sched->kss_tx_conns)) {
1763 conn = list_entry(sched->kss_tx_conns.next,
1764 ksock_conn_t, ksnc_tx_list);
1765 list_del (&conn->ksnc_tx_list);
1767 LASSERT(conn->ksnc_tx_scheduled);
1768 LASSERT(conn->ksnc_tx_ready);
1769 LASSERT(!list_empty(&conn->ksnc_tx_queue));
1771 tx = list_entry(conn->ksnc_tx_queue.next,
1772 ksock_tx_t, tx_list);
1773 /* dequeue now so empty list => more to send */
1774 list_del(&tx->tx_list);
1776 /* Clear tx_ready in case send isn't complete. Do
1777 * it BEFORE we call process_transmit, since
1778 * write_space can set it any time after we release
1780 conn->ksnc_tx_ready = 0;
1781 spin_unlock_irqrestore (&sched->kss_lock, flags);
1783 rc = ksocknal_process_transmit(conn, tx);
1785 spin_lock_irqsave (&sched->kss_lock, flags);
1787 if (rc == -ENOMEM || rc == -EAGAIN) {
1788 /* Incomplete send: replace tx on HEAD of tx_queue */
1789 list_add (&tx->tx_list, &conn->ksnc_tx_queue);
1791 /* Complete send; assume space for more */
1792 conn->ksnc_tx_ready = 1;
1795 if (rc == -ENOMEM) {
1796 /* Do nothing; after a short timeout, this
1797 * conn will be reposted on kss_tx_conns. */
1798 } else if (conn->ksnc_tx_ready &&
1799 !list_empty (&conn->ksnc_tx_queue)) {
1800 /* reschedule for tx */
1801 list_add_tail (&conn->ksnc_tx_list,
1802 &sched->kss_tx_conns);
1804 conn->ksnc_tx_scheduled = 0;
1806 ksocknal_put_conn (conn);
1812 if (!list_empty (&sched->kss_zctxdone_list)) {
1814 list_entry(sched->kss_zctxdone_list.next,
1815 ksock_tx_t, tx_list);
1818 list_del (&tx->tx_list);
1819 spin_unlock_irqrestore (&sched->kss_lock, flags);
1821 ksocknal_tx_done (tx, 1);
1823 spin_lock_irqsave (&sched->kss_lock, flags);
1826 if (!did_something || /* nothing to do */
1827 ++nloops == SOCKNAL_RESCHED) { /* hogging CPU? */
1828 spin_unlock_irqrestore (&sched->kss_lock, flags);
1832 if (!did_something) { /* wait for something to do */
1834 rc = wait_event_interruptible (sched->kss_waitq,
1835 ksocknal_data.ksnd_shuttingdown ||
1836 !list_empty(&sched->kss_rx_conns) ||
1837 !list_empty(&sched->kss_tx_conns) ||
1838 !list_empty(&sched->kss_zctxdone_list));
1840 rc = wait_event_interruptible (sched->kss_waitq,
1841 ksocknal_data.ksnd_shuttingdown ||
1842 !list_empty(&sched->kss_rx_conns) ||
1843 !list_empty(&sched->kss_tx_conns));
1849 spin_lock_irqsave (&sched->kss_lock, flags);
1853 spin_unlock_irqrestore (&sched->kss_lock, flags);
1854 ksocknal_thread_fini ();
1859 ksocknal_data_ready (struct sock *sk, int n)
1861 unsigned long flags;
1863 ksock_sched_t *sched;
1866 /* interleave correctly with closing sockets... */
1867 read_lock (&ksocknal_data.ksnd_global_lock);
1869 conn = sk->sk_user_data;
1870 if (conn == NULL) { /* raced with ksocknal_terminate_conn */
1871 LASSERT (sk->sk_data_ready != &ksocknal_data_ready);
1872 sk->sk_data_ready (sk, n);
1874 sched = conn->ksnc_scheduler;
1876 spin_lock_irqsave (&sched->kss_lock, flags);
1878 conn->ksnc_rx_ready = 1;
1880 if (!conn->ksnc_rx_scheduled) { /* not being progressed */
1881 list_add_tail(&conn->ksnc_rx_list,
1882 &sched->kss_rx_conns);
1883 conn->ksnc_rx_scheduled = 1;
1884 /* extra ref for scheduler */
1885 atomic_inc (&conn->ksnc_refcount);
1887 wake_up (&sched->kss_waitq);
1890 spin_unlock_irqrestore (&sched->kss_lock, flags);
1893 read_unlock (&ksocknal_data.ksnd_global_lock);
1899 ksocknal_write_space (struct sock *sk)
1901 unsigned long flags;
1903 ksock_sched_t *sched;
1905 /* interleave correctly with closing sockets... */
1906 read_lock (&ksocknal_data.ksnd_global_lock);
1908 conn = sk->sk_user_data;
1910 CDEBUG(D_NET, "sk %p wspace %d low water %d conn %p%s%s%s\n",
1911 sk, tcp_wspace(sk), SOCKNAL_TX_LOW_WATER(sk), conn,
1912 (conn == NULL) ? "" : (conn->ksnc_tx_ready ?
1913 " ready" : " blocked"),
1914 (conn == NULL) ? "" : (conn->ksnc_tx_scheduled ?
1915 " scheduled" : " idle"),
1916 (conn == NULL) ? "" : (list_empty (&conn->ksnc_tx_queue) ?
1917 " empty" : " queued"));
1919 if (conn == NULL) { /* raced with ksocknal_terminate_conn */
1920 LASSERT (sk->sk_write_space != &ksocknal_write_space);
1921 sk->sk_write_space (sk);
1923 read_unlock (&ksocknal_data.ksnd_global_lock);
1927 if (tcp_wspace(sk) >= SOCKNAL_TX_LOW_WATER(sk)) { /* got enough space */
1928 clear_bit (SOCK_NOSPACE, &sk->sk_socket->flags);
1930 sched = conn->ksnc_scheduler;
1932 spin_lock_irqsave (&sched->kss_lock, flags);
1934 conn->ksnc_tx_ready = 1;
1936 if (!conn->ksnc_tx_scheduled && // not being progressed
1937 !list_empty(&conn->ksnc_tx_queue)){//packets to send
1938 list_add_tail (&conn->ksnc_tx_list,
1939 &sched->kss_tx_conns);
1940 conn->ksnc_tx_scheduled = 1;
1941 /* extra ref for scheduler */
1942 atomic_inc (&conn->ksnc_refcount);
1944 wake_up (&sched->kss_waitq);
1947 spin_unlock_irqrestore (&sched->kss_lock, flags);
1950 read_unlock (&ksocknal_data.ksnd_global_lock);
1954 ksocknal_sock_write (struct socket *sock, void *buffer, int nob)
1957 mm_segment_t oldmm = get_fs();
1960 struct iovec iov = {
1964 struct msghdr msg = {
1969 .msg_control = NULL,
1970 .msg_controllen = 0,
1975 rc = sock_sendmsg (sock, &msg, iov.iov_len);
1982 CERROR ("Unexpected zero rc\n");
1983 return (-ECONNABORTED);
1986 buffer = ((char *)buffer) + rc;
1994 ksocknal_sock_read (struct socket *sock, void *buffer, int nob)
1997 mm_segment_t oldmm = get_fs();
2000 struct iovec iov = {
2004 struct msghdr msg = {
2009 .msg_control = NULL,
2010 .msg_controllen = 0,
2015 rc = sock_recvmsg (sock, &msg, iov.iov_len, 0);
2022 return (-ECONNABORTED);
2024 buffer = ((char *)buffer) + rc;
2032 ksocknal_hello (struct socket *sock, ptl_nid_t *nid, int *type, __u64 *incarnation)
2036 ptl_magicversion_t *hmv = (ptl_magicversion_t *)&hdr.dest_nid;
2037 char ipbuf[PTL_NALFMT_SIZE];
2039 LASSERT (sizeof (*hmv) == sizeof (hdr.dest_nid));
2041 memset (&hdr, 0, sizeof (hdr));
2042 hmv->magic = __cpu_to_le32 (PORTALS_PROTO_MAGIC);
2043 hmv->version_major = __cpu_to_le32 (PORTALS_PROTO_VERSION_MAJOR);
2044 hmv->version_minor = __cpu_to_le32 (PORTALS_PROTO_VERSION_MINOR);
2046 hdr.src_nid = __cpu_to_le64 (ksocknal_lib.ni.nid);
2047 hdr.type = __cpu_to_le32 (PTL_MSG_HELLO);
2049 hdr.msg.hello.type = __cpu_to_le32 (*type);
2050 hdr.msg.hello.incarnation =
2051 __cpu_to_le64 (ksocknal_data.ksnd_incarnation);
2053 /* Assume sufficient socket buffering for this message */
2054 rc = ksocknal_sock_write (sock, &hdr, sizeof (hdr));
2056 CERROR ("Error %d sending HELLO to "LPX64" %s\n",
2057 rc, *nid, portals_nid2str(SOCKNAL, *nid, ipbuf));
2061 rc = ksocknal_sock_read (sock, hmv, sizeof (*hmv));
2063 CERROR ("Error %d reading HELLO from "LPX64" %s\n",
2064 rc, *nid, portals_nid2str(SOCKNAL, *nid, ipbuf));
2068 if (hmv->magic != __le32_to_cpu (PORTALS_PROTO_MAGIC)) {
2069 CERROR ("Bad magic %#08x (%#08x expected) from "LPX64" %s\n",
2070 __cpu_to_le32 (hmv->magic), PORTALS_PROTO_MAGIC, *nid,
2071 portals_nid2str(SOCKNAL, *nid, ipbuf));
2075 if (hmv->version_major != __cpu_to_le16 (PORTALS_PROTO_VERSION_MAJOR) ||
2076 hmv->version_minor != __cpu_to_le16 (PORTALS_PROTO_VERSION_MINOR)) {
2077 CERROR ("Incompatible protocol version %d.%d (%d.%d expected)"
2078 " from "LPX64" %s\n",
2079 __le16_to_cpu (hmv->version_major),
2080 __le16_to_cpu (hmv->version_minor),
2081 PORTALS_PROTO_VERSION_MAJOR,
2082 PORTALS_PROTO_VERSION_MINOR,
2083 *nid, portals_nid2str(SOCKNAL, *nid, ipbuf));
2087 #if (PORTALS_PROTO_VERSION_MAJOR != 0)
2088 # error "This code only understands protocol version 0.x"
2090 /* version 0 sends magic/version as the dest_nid of a 'hello' header,
2091 * so read the rest of it in now... */
2093 rc = ksocknal_sock_read (sock, hmv + 1, sizeof (hdr) - sizeof (*hmv));
2095 CERROR ("Error %d reading rest of HELLO hdr from "LPX64" %s\n",
2096 rc, *nid, portals_nid2str(SOCKNAL, *nid, ipbuf));
2100 /* ...and check we got what we expected */
2101 if (hdr.type != __cpu_to_le32 (PTL_MSG_HELLO) ||
2102 hdr.payload_length != __cpu_to_le32 (0)) {
2103 CERROR ("Expecting a HELLO hdr with 0 payload,"
2104 " but got type %d with %d payload from "LPX64" %s\n",
2105 __le32_to_cpu (hdr.type),
2106 __le32_to_cpu (hdr.payload_length), *nid,
2107 portals_nid2str(SOCKNAL, *nid, ipbuf));
2111 if (__le64_to_cpu(hdr.src_nid) == PTL_NID_ANY) {
2112 CERROR("Expecting a HELLO hdr with a NID, but got PTL_NID_ANY\n");
2116 if (*nid == PTL_NID_ANY) { /* don't know peer's nid yet */
2117 *nid = __le64_to_cpu(hdr.src_nid);
2118 } else if (*nid != __le64_to_cpu (hdr.src_nid)) {
2119 CERROR ("Connected to nid "LPX64" %s, but expecting "LPX64" %s\n",
2120 __le64_to_cpu (hdr.src_nid),
2121 portals_nid2str(SOCKNAL,
2122 __le64_to_cpu(hdr.src_nid),
2124 *nid, portals_nid2str(SOCKNAL, *nid, ipbuf));
2128 if (*type == SOCKNAL_CONN_NONE) {
2129 /* I've accepted this connection; peer determines type */
2130 *type = __le32_to_cpu(hdr.msg.hello.type);
2132 case SOCKNAL_CONN_ANY:
2133 case SOCKNAL_CONN_CONTROL:
2135 case SOCKNAL_CONN_BULK_IN:
2136 *type = SOCKNAL_CONN_BULK_OUT;
2138 case SOCKNAL_CONN_BULK_OUT:
2139 *type = SOCKNAL_CONN_BULK_IN;
2142 CERROR ("Unexpected type %d from "LPX64" %s\n",
2144 portals_nid2str(SOCKNAL, *nid, ipbuf));
2147 } else if (__le32_to_cpu(hdr.msg.hello.type) != SOCKNAL_CONN_NONE) {
2148 CERROR ("Mismatched types: me %d "LPX64" %s %d\n",
2149 *type, *nid, portals_nid2str(SOCKNAL, *nid, ipbuf),
2150 __le32_to_cpu(hdr.msg.hello.type));
2154 *incarnation = __le64_to_cpu(hdr.msg.hello.incarnation);
2160 ksocknal_setup_sock (struct socket *sock)
2162 mm_segment_t oldmm = get_fs ();
2165 struct linger linger;
2167 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
2168 sock->sk->sk_allocation = GFP_NOFS;
2170 sock->sk->allocation = GFP_NOFS;
2173 /* Ensure this socket aborts active sends immediately when we close
2177 linger.l_linger = 0;
2180 rc = sock_setsockopt (sock, SOL_SOCKET, SO_LINGER,
2181 (char *)&linger, sizeof (linger));
2184 CERROR ("Can't set SO_LINGER: %d\n", rc);
2190 rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_LINGER2,
2191 (char *)&option, sizeof (option));
2194 CERROR ("Can't set SO_LINGER2: %d\n", rc);
2198 #if SOCKNAL_USE_KEEPALIVES
2199 /* Keepalives: If 3/4 of the timeout elapses, start probing every
2200 * second until the timeout elapses. */
2202 option = (ksocknal_data.ksnd_io_timeout * 3) / 4;
2204 rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_KEEPIDLE,
2205 (char *)&option, sizeof (option));
2208 CERROR ("Can't set TCP_KEEPIDLE: %d\n", rc);
2214 rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_KEEPINTVL,
2215 (char *)&option, sizeof (option));
2218 CERROR ("Can't set TCP_KEEPINTVL: %d\n", rc);
2222 option = ksocknal_data.ksnd_io_timeout / 4;
2224 rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_KEEPCNT,
2225 (char *)&option, sizeof (option));
2228 CERROR ("Can't set TCP_KEEPINTVL: %d\n", rc);
2234 rc = sock_setsockopt (sock, SOL_SOCKET, SO_KEEPALIVE,
2235 (char *)&option, sizeof (option));
2238 CERROR ("Can't set SO_KEEPALIVE: %d\n", rc);
2246 ksocknal_connect_peer (ksock_route_t *route, int type)
2248 struct sockaddr_in peer_addr;
2249 mm_segment_t oldmm = get_fs();
2252 struct socket *sock;
2254 char ipbuf[PTL_NALFMT_SIZE];
2256 rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock);
2258 CERROR ("Can't create autoconnect socket: %d\n", rc);
2262 /* Ugh; have to map_fd for compatibility with sockets passed in
2263 * from userspace. And we actually need the sock->file refcounting
2264 * that this gives you :) */
2266 fd = sock_map_fd (sock);
2268 sock_release (sock);
2269 CERROR ("sock_map_fd error %d\n", fd);
2273 /* NB the fd now owns the ref on sock->file */
2274 LASSERT (sock->file != NULL);
2275 LASSERT (file_count(sock->file) == 1);
2277 /* Set the socket timeouts, so our connection attempt completes in
2279 tv.tv_sec = ksocknal_data.ksnd_io_timeout;
2283 rc = sock_setsockopt (sock, SOL_SOCKET, SO_SNDTIMEO,
2284 (char *)&tv, sizeof (tv));
2287 CERROR ("Can't set send timeout %d: %d\n",
2288 ksocknal_data.ksnd_io_timeout, rc);
2293 rc = sock_setsockopt (sock, SOL_SOCKET, SO_RCVTIMEO,
2294 (char *)&tv, sizeof (tv));
2297 CERROR ("Can't set receive timeout %d: %d\n",
2298 ksocknal_data.ksnd_io_timeout, rc);
2306 rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_NODELAY,
2307 (char *)&option, sizeof (option));
2310 CERROR ("Can't disable nagle: %d\n", rc);
2315 if (route->ksnr_buffer_size != 0) {
2316 int option = route->ksnr_buffer_size;
2319 rc = sock_setsockopt (sock, SOL_SOCKET, SO_SNDBUF,
2320 (char *)&option, sizeof (option));
2323 CERROR ("Can't set send buffer %d: %d\n",
2324 route->ksnr_buffer_size, rc);
2329 rc = sock_setsockopt (sock, SOL_SOCKET, SO_RCVBUF,
2330 (char *)&option, sizeof (option));
2333 CERROR ("Can't set receive buffer %d: %d\n",
2334 route->ksnr_buffer_size, rc);
2339 memset (&peer_addr, 0, sizeof (peer_addr));
2340 peer_addr.sin_family = AF_INET;
2341 peer_addr.sin_port = htons (route->ksnr_port);
2342 peer_addr.sin_addr.s_addr = htonl (route->ksnr_ipaddr);
2344 rc = sock->ops->connect (sock, (struct sockaddr *)&peer_addr,
2345 sizeof (peer_addr), sock->file->f_flags);
2347 CERROR ("Error %d connecting to "LPX64" %s\n", rc,
2348 route->ksnr_peer->ksnp_nid,
2349 portals_nid2str(SOCKNAL,
2350 route->ksnr_peer->ksnp_nid,
2355 rc = ksocknal_create_conn (route, sock, route->ksnr_irq_affinity, type);
2357 /* Take an extra ref on sock->file to compensate for the
2358 * upcoming close which will lose fd's ref on it. */
2359 get_file (sock->file);
2368 ksocknal_autoconnect (ksock_route_t *route)
2370 LIST_HEAD (zombies);
2373 unsigned long flags;
2378 for (type = 0; type < SOCKNAL_CONN_NTYPES; type++)
2379 if ((route->ksnr_connecting & (1 << type)) != 0)
2381 LASSERT (type < SOCKNAL_CONN_NTYPES);
2383 rc = ksocknal_connect_peer (route, type);
2388 /* successfully autoconnected: create_conn did the
2389 * route/conn binding and scheduled any blocked packets */
2391 if (route->ksnr_connecting == 0) {
2392 /* No more connections required */
2397 /* Connection attempt failed */
2399 write_lock_irqsave (&ksocknal_data.ksnd_global_lock, flags);
2401 peer = route->ksnr_peer;
2402 route->ksnr_connecting = 0;
2404 /* This is a retry rather than a new connection */
2405 LASSERT (route->ksnr_retry_interval != 0);
2406 route->ksnr_timeout = jiffies + route->ksnr_retry_interval;
2407 route->ksnr_retry_interval = MIN (route->ksnr_retry_interval * 2,
2408 SOCKNAL_MAX_RECONNECT_INTERVAL);
2410 if (!list_empty (&peer->ksnp_tx_queue) &&
2411 ksocknal_find_connecting_route_locked (peer) == NULL) {
2412 LASSERT (list_empty (&peer->ksnp_conns));
2414 /* None of the connections that the blocked packets are
2415 * waiting for have been successful. Complete them now... */
2417 tx = list_entry (peer->ksnp_tx_queue.next,
2418 ksock_tx_t, tx_list);
2419 list_del (&tx->tx_list);
2420 list_add_tail (&tx->tx_list, &zombies);
2421 } while (!list_empty (&peer->ksnp_tx_queue));
2424 /* make this route least-favourite for re-selection */
2425 if (!route->ksnr_deleted) {
2426 list_del(&route->ksnr_list);
2427 list_add_tail(&route->ksnr_list, &peer->ksnp_routes);
2430 write_unlock_irqrestore (&ksocknal_data.ksnd_global_lock, flags);
2432 while (!list_empty (&zombies)) {
2433 char ipbuf[PTL_NALFMT_SIZE];
2434 tx = list_entry (zombies.next, ksock_tx_t, tx_list);
2436 CERROR ("Deleting packet type %d len %d ("LPX64" %s->"LPX64" %s)\n",
2437 NTOH__u32 (tx->tx_hdr->type),
2438 NTOH__u32 (tx->tx_hdr->payload_length),
2439 NTOH__u64 (tx->tx_hdr->src_nid),
2440 portals_nid2str(SOCKNAL,
2441 NTOH__u64(tx->tx_hdr->src_nid),
2443 NTOH__u64 (tx->tx_hdr->dest_nid),
2444 portals_nid2str(SOCKNAL,
2445 NTOH__u64(tx->tx_hdr->src_nid),
2448 list_del (&tx->tx_list);
2450 ksocknal_tx_done (tx, 0);
2455 ksocknal_autoconnectd (void *arg)
2457 long id = (long)arg;
2459 unsigned long flags;
2460 ksock_route_t *route;
2463 snprintf (name, sizeof (name), "ksocknal_ad%02ld", id);
2464 kportal_daemonize (name);
2465 kportal_blockallsigs ();
2467 spin_lock_irqsave (&ksocknal_data.ksnd_autoconnectd_lock, flags);
2469 while (!ksocknal_data.ksnd_shuttingdown) {
2471 if (!list_empty (&ksocknal_data.ksnd_autoconnectd_routes)) {
2472 route = list_entry (ksocknal_data.ksnd_autoconnectd_routes.next,
2473 ksock_route_t, ksnr_connect_list);
2475 list_del (&route->ksnr_connect_list);
2476 spin_unlock_irqrestore (&ksocknal_data.ksnd_autoconnectd_lock, flags);
2478 ksocknal_autoconnect (route);
2479 ksocknal_put_route (route);
2481 spin_lock_irqsave (&ksocknal_data.ksnd_autoconnectd_lock, flags);
2485 spin_unlock_irqrestore (&ksocknal_data.ksnd_autoconnectd_lock, flags);
2487 rc = wait_event_interruptible (ksocknal_data.ksnd_autoconnectd_waitq,
2488 ksocknal_data.ksnd_shuttingdown ||
2489 !list_empty (&ksocknal_data.ksnd_autoconnectd_routes));
2491 spin_lock_irqsave (&ksocknal_data.ksnd_autoconnectd_lock, flags);
2494 spin_unlock_irqrestore (&ksocknal_data.ksnd_autoconnectd_lock, flags);
2496 ksocknal_thread_fini ();
2501 ksocknal_find_timed_out_conn (ksock_peer_t *peer)
2503 /* We're called with a shared lock on ksnd_global_lock */
2505 struct list_head *ctmp;
2506 ksock_sched_t *sched;
2508 list_for_each (ctmp, &peer->ksnp_conns) {
2509 conn = list_entry (ctmp, ksock_conn_t, ksnc_list);
2510 sched = conn->ksnc_scheduler;
2512 /* Don't need the {get,put}connsock dance to deref ksnc_sock... */
2513 LASSERT (!conn->ksnc_closing);
2515 if (conn->ksnc_rx_started &&
2516 time_after_eq (jiffies, conn->ksnc_rx_deadline)) {
2517 /* Timed out incomplete incoming message */
2518 atomic_inc (&conn->ksnc_refcount);
2519 CERROR ("Timed out RX from "LPX64" %p %d.%d.%d.%d\n",
2520 peer->ksnp_nid, conn, HIPQUAD(conn->ksnc_ipaddr));
2524 if ((!list_empty (&conn->ksnc_tx_queue) ||
2525 conn->ksnc_sock->sk->sk_wmem_queued != 0) &&
2526 time_after_eq (jiffies, conn->ksnc_tx_deadline)) {
2527 /* Timed out messages queued for sending, or
2528 * messages buffered in the socket's send buffer */
2529 atomic_inc (&conn->ksnc_refcount);
2530 CERROR ("Timed out TX to "LPX64" %s%d %p %d.%d.%d.%d\n",
2532 list_empty (&conn->ksnc_tx_queue) ? "" : "Q ",
2533 conn->ksnc_sock->sk->sk_wmem_queued, conn,
2534 HIPQUAD(conn->ksnc_ipaddr));
2543 ksocknal_check_peer_timeouts (int idx)
2545 struct list_head *peers = &ksocknal_data.ksnd_peers[idx];
2546 struct list_head *ptmp;
2551 /* NB. We expect to have a look at all the peers and not find any
2552 * connections to time out, so we just use a shared lock while we
2554 read_lock (&ksocknal_data.ksnd_global_lock);
2556 list_for_each (ptmp, peers) {
2557 peer = list_entry (ptmp, ksock_peer_t, ksnp_list);
2558 conn = ksocknal_find_timed_out_conn (peer);
2561 read_unlock (&ksocknal_data.ksnd_global_lock);
2563 CERROR ("Timeout out conn->"LPX64" ip %d.%d.%d.%d:%d\n",
2565 HIPQUAD(conn->ksnc_ipaddr),
2567 ksocknal_close_conn_and_siblings (conn, -ETIMEDOUT);
2569 /* NB we won't find this one again, but we can't
2570 * just proceed with the next peer, since we dropped
2571 * ksnd_global_lock and it might be dead already! */
2572 ksocknal_put_conn (conn);
2577 read_unlock (&ksocknal_data.ksnd_global_lock);
2581 ksocknal_reaper (void *arg)
2584 unsigned long flags;
2586 ksock_sched_t *sched;
2587 struct list_head enomem_conns;
2592 unsigned long deadline = jiffies;
2594 kportal_daemonize ("ksocknal_reaper");
2595 kportal_blockallsigs ();
2597 INIT_LIST_HEAD(&enomem_conns);
2598 init_waitqueue_entry (&wait, current);
2600 spin_lock_irqsave (&ksocknal_data.ksnd_reaper_lock, flags);
2602 while (!ksocknal_data.ksnd_shuttingdown) {
2604 if (!list_empty (&ksocknal_data.ksnd_deathrow_conns)) {
2605 conn = list_entry (ksocknal_data.ksnd_deathrow_conns.next,
2606 ksock_conn_t, ksnc_list);
2607 list_del (&conn->ksnc_list);
2609 spin_unlock_irqrestore (&ksocknal_data.ksnd_reaper_lock, flags);
2611 ksocknal_terminate_conn (conn);
2612 ksocknal_put_conn (conn);
2614 spin_lock_irqsave (&ksocknal_data.ksnd_reaper_lock, flags);
2618 if (!list_empty (&ksocknal_data.ksnd_zombie_conns)) {
2619 conn = list_entry (ksocknal_data.ksnd_zombie_conns.next,
2620 ksock_conn_t, ksnc_list);
2621 list_del (&conn->ksnc_list);
2623 spin_unlock_irqrestore (&ksocknal_data.ksnd_reaper_lock, flags);
2625 ksocknal_destroy_conn (conn);
2627 spin_lock_irqsave (&ksocknal_data.ksnd_reaper_lock, flags);
2631 if (!list_empty (&ksocknal_data.ksnd_enomem_conns)) {
2632 list_add(&enomem_conns, &ksocknal_data.ksnd_enomem_conns);
2633 list_del_init(&ksocknal_data.ksnd_enomem_conns);
2636 spin_unlock_irqrestore (&ksocknal_data.ksnd_reaper_lock, flags);
2638 /* reschedule all the connections that stalled with ENOMEM... */
2640 while (!list_empty (&enomem_conns)) {
2641 conn = list_entry (enomem_conns.next,
2642 ksock_conn_t, ksnc_tx_list);
2643 list_del (&conn->ksnc_tx_list);
2645 sched = conn->ksnc_scheduler;
2647 spin_lock_irqsave (&sched->kss_lock, flags);
2649 LASSERT (conn->ksnc_tx_scheduled);
2650 conn->ksnc_tx_ready = 1;
2651 list_add_tail (&conn->ksnc_tx_list, &sched->kss_tx_conns);
2652 wake_up (&sched->kss_waitq);
2654 spin_unlock_irqrestore (&sched->kss_lock, flags);
2658 /* careful with the jiffy wrap... */
2659 while ((timeout = (int)(deadline - jiffies)) <= 0) {
2662 int chunk = ksocknal_data.ksnd_peer_hash_size;
2664 /* Time to check for timeouts on a few more peers: I do
2665 * checks every 'p' seconds on a proportion of the peer
2666 * table and I need to check every connection 'n' times
2667 * within a timeout interval, to ensure I detect a
2668 * timeout on any connection within (n+1)/n times the
2669 * timeout interval. */
2671 if (ksocknal_data.ksnd_io_timeout > n * p)
2672 chunk = (chunk * n * p) /
2673 ksocknal_data.ksnd_io_timeout;
2677 for (i = 0; i < chunk; i++) {
2678 ksocknal_check_peer_timeouts (peer_index);
2679 peer_index = (peer_index + 1) %
2680 ksocknal_data.ksnd_peer_hash_size;
2686 if (nenomem_conns != 0) {
2687 /* Reduce my timeout if I rescheduled ENOMEM conns.
2688 * This also prevents me getting woken immediately
2689 * if any go back on my enomem list. */
2690 timeout = SOCKNAL_ENOMEM_RETRY;
2692 ksocknal_data.ksnd_reaper_waketime = jiffies + timeout;
2694 add_wait_queue (&ksocknal_data.ksnd_reaper_waitq, &wait);
2695 set_current_state (TASK_INTERRUPTIBLE);
2697 if (!ksocknal_data.ksnd_shuttingdown &&
2698 list_empty (&ksocknal_data.ksnd_deathrow_conns) &&
2699 list_empty (&ksocknal_data.ksnd_zombie_conns))
2700 schedule_timeout (timeout);
2702 set_current_state (TASK_RUNNING);
2703 remove_wait_queue (&ksocknal_data.ksnd_reaper_waitq, &wait);
2705 spin_lock_irqsave (&ksocknal_data.ksnd_reaper_lock, flags);
2708 spin_unlock_irqrestore (&ksocknal_data.ksnd_reaper_lock, flags);
2710 ksocknal_thread_fini ();
2714 nal_cb_t ksocknal_lib = {
2715 nal_data: &ksocknal_data, /* NAL private data */
2716 cb_send: ksocknal_send,
2717 cb_send_pages: ksocknal_send_pages,
2718 cb_recv: ksocknal_recv,
2719 cb_recv_pages: ksocknal_recv_pages,
2720 cb_read: ksocknal_read,
2721 cb_write: ksocknal_write,
2722 cb_callback: ksocknal_callback,
2723 cb_malloc: ksocknal_malloc,
2724 cb_free: ksocknal_free,
2725 cb_printf: ksocknal_printf,
2726 cb_cli: ksocknal_cli,
2727 cb_sti: ksocknal_sti,
2728 cb_dist: ksocknal_dist