summaryrefslogtreecommitdiff
path: root/usr/src
diff options
context:
space:
mode:
Diffstat (limited to 'usr/src')
-rw-r--r--usr/src/cmd/mdb/common/modules/sctp/sctp.c17
-rw-r--r--usr/src/uts/common/inet/ip.h2
-rw-r--r--usr/src/uts/common/inet/sctp/sctp.c10
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_asconf.c25
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_bind.c2
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_common.c317
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_conn.c9
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_cookie.c3
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_error.c7
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_heartbeat.c6
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_impl.h112
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_init.c12
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_input.c47
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_opt_data.c7
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_output.c461
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_shutdown.c38
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_snmp.c55
-rw-r--r--usr/src/uts/common/inet/sctp/sctp_timer.c8
18 files changed, 814 insertions, 324 deletions
diff --git a/usr/src/cmd/mdb/common/modules/sctp/sctp.c b/usr/src/cmd/mdb/common/modules/sctp/sctp.c
index 90d733bc04..83904321bb 100644
--- a/usr/src/cmd/mdb/common/modules/sctp/sctp.c
+++ b/usr/src/cmd/mdb/common/modules/sctp/sctp.c
@@ -2,9 +2,8 @@
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License"). You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
*
* CDDL HEADER END
*/
+
/*
- * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
@@ -569,8 +569,11 @@ show_sctp_flags(sctp_t *sctp)
mdb_printf("\tndelay\t\t\t%d\n", sctp->sctp_ndelay);
mdb_printf("\tcondemned\t\t%d\n", sctp->sctp_condemned);
mdb_printf("\tchk_fast_rexmit\t\t%d\n", sctp->sctp_chk_fast_rexmit);
+
mdb_printf("\tprsctp_aware\t\t%d\n", sctp->sctp_prsctp_aware);
mdb_printf("\tlinklocal\t\t%d\n", sctp->sctp_linklocal);
+ mdb_printf("\tmac_exempt\t\t%d\n", sctp->sctp_mac_exempt);
+ mdb_printf("\trexmitting\t\t%d\n", sctp->sctp_rexmitting);
mdb_printf("\trecvsndrcvinfo\t\t%d\n", sctp->sctp_recvsndrcvinfo);
mdb_printf("\trecvassocevnt\t\t%d\n", sctp->sctp_recvassocevnt);
@@ -832,11 +835,13 @@ sctp(uintptr_t addr, uint_t flags, int argc, const mdb_arg_t *argv)
"max_init_rxt\t%?d\tpa_max_rxt\t%?d\n"
"pp_max_rxt\t%?d\trto_max\t\t%?u\n"
"rto_min\t\t%?u\trto_initial\t%?u\n"
- "init_rto_max\t%?u\n", sctp.sctp_cookie_mp,
+ "init_rto_max\t%?u\n"
+ "rxt_nxttsn\t%?u\trxt_maxtsn\t%?u\n", sctp.sctp_cookie_mp,
sctp.sctp_strikes, sctp.sctp_max_init_rxt,
sctp.sctp_pa_max_rxt, sctp.sctp_pp_max_rxt,
sctp.sctp_rto_max, sctp.sctp_rto_min,
- sctp.sctp_rto_initial, sctp.sctp_init_rto_max);
+ sctp.sctp_rto_initial, sctp.sctp_init_rto_max,
+ sctp.sctp_rxt_nxttsn, sctp.sctp_rxt_maxtsn);
}
if (opts & MDB_SCTP_SHOW_CONN) {
diff --git a/usr/src/uts/common/inet/ip.h b/usr/src/uts/common/inet/ip.h
index 1618e95bdd..1f9eb56a10 100644
--- a/usr/src/uts/common/inet/ip.h
+++ b/usr/src/uts/common/inet/ip.h
@@ -83,6 +83,8 @@ extern "C" {
#define TCP_DEV_NAME "/dev/tcp"
#define TCP6_DEV_NAME "/dev/tcp6"
+#define SCTP_MOD_NAME "sctp"
+
/* Minor numbers */
#define IPV4_MINOR 0
#define IPV6_MINOR 1
diff --git a/usr/src/uts/common/inet/sctp/sctp.c b/usr/src/uts/common/inet/sctp/sctp.c
index 33820e4cb4..cc9c1345ad 100644
--- a/usr/src/uts/common/inet/sctp/sctp.c
+++ b/usr/src/uts/common/inet/sctp/sctp.c
@@ -324,7 +324,6 @@ int
sctp_disconnect(sctp_t *sctp)
{
int error = 0;
- sctp_faddr_t *fp;
dprint(3, ("sctp_disconnect %p, state %d\n", (void *)sctp,
sctp->sctp_state));
@@ -375,8 +374,7 @@ sctp_disconnect(sctp_t *sctp)
sctp_send_shutdown(sctp, 0);
/* Pass gathered wisdom to IP for keeping */
- for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next)
- sctp_faddr2ire(sctp, fp);
+ sctp_update_ire(sctp);
/*
* If lingering on close then wait until the shutdown
@@ -1298,9 +1296,10 @@ sctp_create(void *sctp_ulpd, sctp_t *parent, int family, int flags,
if (credp == NULL)
return (NULL);
- if ((sctp_connp = ipcl_conn_create(IPCL_SCTPCONN, sleep)) == NULL)
+ if ((sctp_connp = ipcl_conn_create(IPCL_SCTPCONN, sleep)) == NULL) {
+ SCTP_KSTAT(sctp_conn_create);
return (NULL);
-
+ }
sctp_connp->conn_ulp_labeled = is_system_labeled();
psctp = (sctp_t *)parent;
@@ -1698,6 +1697,7 @@ sctp_find_next_tq(sctp_t *sctp)
sctp->sctp_recvq_tq = tq;
return (B_TRUE);
}
+ SCTP_KSTAT(sctp_find_next_tq);
return (B_FALSE);
}
diff --git a/usr/src/uts/common/inet/sctp/sctp_asconf.c b/usr/src/uts/common/inet/sctp/sctp_asconf.c
index 6dfa20b156..a242564f14 100644
--- a/usr/src/uts/common/inet/sctp/sctp_asconf.c
+++ b/usr/src/uts/common/inet/sctp/sctp_asconf.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -416,6 +417,7 @@ sctp_input_asconf(sctp_t *sctp, sctp_chunk_hdr_t *ch, sctp_faddr_t *fp)
hmp = sctp_make_mp(sctp, fp, sizeof (*ach) + sizeof (*idp));
if (hmp == NULL) {
/* Let the peer retransmit */
+ SCTP_KSTAT(sctp_send_asconf_ack_failed);
return;
}
ach = (sctp_chunk_hdr_t *)hmp->b_wptr;
@@ -478,6 +480,7 @@ sctp_input_asconf(sctp_t *sctp, sctp_chunk_hdr_t *ch, sctp_faddr_t *fp)
alist = kmem_alloc(asize, KM_NOSLEEP);
if (alist == NULL) {
freeb(hmp);
+ SCTP_KSTAT(sctp_cl_assoc_change);
return;
}
}
@@ -488,6 +491,7 @@ sctp_input_asconf(sctp_t *sctp, sctp_chunk_hdr_t *ch, sctp_faddr_t *fp)
if (acount > 0)
kmem_free(alist, asize);
freeb(hmp);
+ SCTP_KSTAT(sctp_cl_assoc_change);
return;
}
}
@@ -927,6 +931,7 @@ sctp_wput_asconf(sctp_t *sctp, sctp_faddr_t *fp)
ipmp = sctp_make_mp(sctp, fp, 0);
if (ipmp == NULL) {
SCTP_FADDR_RC_TIMER_RESTART(sctp, fp, fp->rto);
+ SCTP_KSTAT(sctp_send_asconf_failed);
return;
}
mp = sctp->sctp_cxmit_list;
@@ -1175,7 +1180,7 @@ sctp_addip_req(sctp_t *sctp, sctp_parm_hdr_t *ph, uint32_t cid,
}
/* Add the new address */
mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
- err = sctp_add_faddr(sctp, &addr, KM_NOSLEEP);
+ err = sctp_add_faddr(sctp, &addr, KM_NOSLEEP, B_FALSE);
mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
if (err == ENOMEM) {
/* no memory */
@@ -1223,13 +1228,7 @@ sctp_addip_req(sctp_t *sctp, sctp_parm_hdr_t *ph, uint32_t cid,
sctp->sctp_primary = fp;
}
if (sctp->sctp_current == nfp) {
- sctp->sctp_current = fp;
- sctp->sctp_mss = fp->sfa_pmss;
- sctp_faddr2hdraddr(fp, sctp);
-
- if (!SCTP_IS_DETACHED(sctp)) {
- sctp_set_ulp_prop(sctp);
- }
+ sctp_set_faddr_current(sctp, fp);
}
if (sctp->sctp_lastdata == nfp) {
sctp->sctp_lastdata = fp;
@@ -1376,15 +1375,7 @@ sctp_setprim_req(sctp_t *sctp, sctp_parm_hdr_t *ph, uint32_t cid,
if (nfp->state != SCTP_FADDRS_ALIVE || nfp == sctp->sctp_current) {
return (NULL);
}
- sctp->sctp_current = nfp;
- sctp->sctp_mss = nfp->sfa_pmss;
-
- /* Reset the addrs in the composite header */
- sctp_faddr2hdraddr(nfp, sctp);
- if (!SCTP_IS_DETACHED(sctp)) {
- sctp_set_ulp_prop(sctp);
- }
-
+ sctp_set_faddr_current(sctp, nfp);
return (NULL);
}
diff --git a/usr/src/uts/common/inet/sctp/sctp_bind.c b/usr/src/uts/common/inet/sctp/sctp_bind.c
index 9933bf4996..2cbb7c21ee 100644
--- a/usr/src/uts/common/inet/sctp/sctp_bind.c
+++ b/usr/src/uts/common/inet/sctp/sctp_bind.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -313,6 +314,7 @@ sctp_bind_add(sctp_t *sctp, const void *addrs, uint32_t addrcnt,
ASSERT(size == 0);
if (!caller_hold_lock)
WAKE_SCTP(sctp);
+ SCTP_KSTAT(sctp_cl_check_addrs);
return (err);
}
ASSERT(addrlist != NULL);
diff --git a/usr/src/uts/common/inet/sctp/sctp_common.c b/usr/src/uts/common/inet/sctp/sctp_common.c
index 516df089eb..0caf1913ca 100644
--- a/usr/src/uts/common/inet/sctp/sctp_common.c
+++ b/usr/src/uts/common/inet/sctp/sctp_common.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -55,9 +56,9 @@
#include "sctp_addr.h"
static struct kmem_cache *sctp_kmem_faddr_cache;
-static void sctp_init_faddr(sctp_t *, sctp_faddr_t *, in6_addr_t *);
+static void sctp_init_faddr(sctp_t *, sctp_faddr_t *, in6_addr_t *, mblk_t *);
-/* Set the source address. Refer to comments in sctp_ire2faddr(). */
+/* Set the source address. Refer to comments in sctp_get_ire(). */
static void
set_saddr(sctp_t *sctp, sctp_faddr_t *fp)
{
@@ -82,15 +83,15 @@ set_saddr(sctp_t *sctp, sctp_faddr_t *fp)
* Call this function to update the cached IRE of a peer addr fp.
*/
void
-sctp_ire2faddr(sctp_t *sctp, sctp_faddr_t *fp)
+sctp_get_ire(sctp_t *sctp, sctp_faddr_t *fp)
{
- ire_t *ire;
- ipaddr_t addr4;
- in6_addr_t laddr;
+ ire_t *ire;
+ ipaddr_t addr4;
+ in6_addr_t laddr;
sctp_saddr_ipif_t *sp;
- uint_t ipif_seqid;
- int hdrlen;
- ts_label_t *tsl;
+ uint_t ipif_seqid;
+ int hdrlen;
+ ts_label_t *tsl;
/* Remove the previous cache IRE */
if ((ire = fp->ire) != NULL) {
@@ -141,7 +142,9 @@ sctp_ire2faddr(sctp_t *sctp, sctp_faddr_t *fp)
* it won't be used to send data.
*/
set_saddr(sctp, fp);
- goto set_current;
+ if (fp->state == SCTP_FADDRS_UNREACH)
+ return;
+ goto check_current;
}
ipif_seqid = ire->ire_ipif->ipif_seqid;
@@ -154,7 +157,7 @@ sctp_ire2faddr(sctp_t *sctp, sctp_faddr_t *fp)
SCTP_PRINTADDR(ire->ire_src_addr_v6)));
}
- /* make sure the laddr is part of this association */
+ /* Make sure the laddr is part of this association */
if ((sp = sctp_ipif_lookup(sctp, ipif_seqid)) != NULL &&
!sp->saddr_ipif_dontsrc) {
if (sp->saddr_ipif_unconfirmed == 1)
@@ -162,16 +165,38 @@ sctp_ire2faddr(sctp_t *sctp, sctp_faddr_t *fp)
fp->saddr = laddr;
} else {
dprint(2, ("ire2faddr: src addr is not part of assc\n"));
- /* set the src to the first saddr and hope for the best */
+
+ /*
+ * Set the src to the first saddr and hope for the best.
+ * Note that we will still do the ire caching below.
+ * Otherwise, whenever we send a packet, we need to do
+ * the ire lookup again and still may not get the correct
+ * source address. Note that this case should very seldomly
+ * happen. One scenario this can happen is an app
+ * explicitly bind() to an address. But that address is
+ * not the preferred source address to send to the peer.
+ */
set_saddr(sctp, fp);
+ if (fp->state == SCTP_FADDRS_UNREACH) {
+ IRE_REFRELE(ire);
+ return;
+ }
}
- /* Cache the IRE */
+ /*
+ * Note that ire_cache_lookup_*() returns an ire with the tracing
+ * bits enabled. This requires the thread holding the ire also
+ * do the IRE_REFRELE(). Thus we need to do IRE_REFHOLD_NOTR()
+ * and then IRE_REFRELE() the ire here to make the tracing bits
+ * work.
+ */
IRE_REFHOLD_NOTR(ire);
+ IRE_REFRELE(ire);
+
+ /* Cache the IRE */
fp->ire = ire;
if (fp->ire->ire_type == IRE_LOOPBACK && !sctp->sctp_loopback)
sctp->sctp_loopback = 1;
- IRE_REFRELE(ire);
/*
* Pull out RTO information for this faddr and use it if we don't
@@ -209,74 +234,77 @@ sctp_ire2faddr(sctp_t *sctp, sctp_faddr_t *fp)
}
}
-set_current:
- if (fp == sctp->sctp_current) {
- sctp_faddr2hdraddr(fp, sctp);
- sctp->sctp_mss = fp->sfa_pmss;
- if (!SCTP_IS_DETACHED(sctp)) {
- sctp_set_ulp_prop(sctp);
- }
- }
+check_current:
+ if (fp == sctp->sctp_current)
+ sctp_set_faddr_current(sctp, fp);
}
void
-sctp_faddr2ire(sctp_t *sctp, sctp_faddr_t *fp)
+sctp_update_ire(sctp_t *sctp)
{
- ire_t *ire;
+ ire_t *ire;
+ sctp_faddr_t *fp;
- if ((ire = fp->ire) == NULL) {
- return;
- }
-
- mutex_enter(&ire->ire_lock);
-
- /* If the cached IRE is going away, there is no point to update it. */
- if (ire->ire_marks & IRE_MARK_CONDEMNED) {
- mutex_exit(&ire->ire_lock);
- IRE_REFRELE_NOTR(ire);
- fp->ire = NULL;
- return;
- }
+ for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next) {
+ if ((ire = fp->ire) == NULL)
+ continue;
+ mutex_enter(&ire->ire_lock);
- /*
- * Only record the PMTU for this faddr if we actually have
- * done discovery. This prevents initialized default from
- * clobbering any real info that IP may have.
- */
- if (fp->pmtu_discovered) {
- if (fp->isv4) {
- ire->ire_max_frag = fp->sfa_pmss + sctp->sctp_hdr_len;
- } else {
- ire->ire_max_frag = fp->sfa_pmss + sctp->sctp_hdr6_len;
+ /*
+ * If the cached IRE is going away, there is no point to
+ * update it.
+ */
+ if (ire->ire_marks & IRE_MARK_CONDEMNED) {
+ mutex_exit(&ire->ire_lock);
+ IRE_REFRELE_NOTR(ire);
+ fp->ire = NULL;
+ continue;
}
- }
- if (sctp_rtt_updates != 0 && fp->rtt_updates >= sctp_rtt_updates) {
/*
- * If there is no old cached values, initialize them
- * conservatively. Set them to be (1.5 * new value).
- * This code copied from ip_ire_advise(). The cached
- * value is in ms.
+ * Only record the PMTU for this faddr if we actually have
+ * done discovery. This prevents initialized default from
+ * clobbering any real info that IP may have.
*/
- if (ire->ire_uinfo.iulp_rtt != 0) {
- ire->ire_uinfo.iulp_rtt = (ire->ire_uinfo.iulp_rtt +
- TICK_TO_MSEC(fp->srtt)) >> 1;
- } else {
- ire->ire_uinfo.iulp_rtt = TICK_TO_MSEC(fp->srtt +
- (fp->srtt >> 1));
+ if (fp->pmtu_discovered) {
+ if (fp->isv4) {
+ ire->ire_max_frag = fp->sfa_pmss +
+ sctp->sctp_hdr_len;
+ } else {
+ ire->ire_max_frag = fp->sfa_pmss +
+ sctp->sctp_hdr6_len;
+ }
}
- if (ire->ire_uinfo.iulp_rtt_sd != 0) {
- ire->ire_uinfo.iulp_rtt_sd =
- (ire->ire_uinfo.iulp_rtt_sd +
- TICK_TO_MSEC(fp->rttvar)) >> 1;
- } else {
- ire->ire_uinfo.iulp_rtt_sd = TICK_TO_MSEC(fp->rttvar +
- (fp->rttvar >> 1));
+
+ if (sctp_rtt_updates != 0 &&
+ fp->rtt_updates >= sctp_rtt_updates) {
+ /*
+ * If there is no old cached values, initialize them
+ * conservatively. Set them to be (1.5 * new value).
+ * This code copied from ip_ire_advise(). The cached
+ * value is in ms.
+ */
+ if (ire->ire_uinfo.iulp_rtt != 0) {
+ ire->ire_uinfo.iulp_rtt =
+ (ire->ire_uinfo.iulp_rtt +
+ TICK_TO_MSEC(fp->srtt)) >> 1;
+ } else {
+ ire->ire_uinfo.iulp_rtt =
+ TICK_TO_MSEC(fp->srtt + (fp->srtt >> 1));
+ }
+ if (ire->ire_uinfo.iulp_rtt_sd != 0) {
+ ire->ire_uinfo.iulp_rtt_sd =
+ (ire->ire_uinfo.iulp_rtt_sd +
+ TICK_TO_MSEC(fp->rttvar)) >> 1;
+ } else {
+ ire->ire_uinfo.iulp_rtt_sd =
+ TICK_TO_MSEC(fp->rttvar +
+ (fp->rttvar >> 1));
+ }
+ fp->rtt_updates = 0;
}
- fp->rtt_updates = 0;
+ mutex_exit(&ire->ire_lock);
}
-
- mutex_exit(&ire->ire_lock);
}
/*
@@ -301,7 +329,7 @@ sctp_make_mp(sctp_t *sctp, sctp_faddr_t *sendto, int trailer)
/* Try to look for another IRE again. */
if (fp->ire == NULL)
- sctp_ire2faddr(sctp, fp);
+ sctp_get_ire(sctp, fp);
/* There is no suitable source address to use, return. */
if (fp->state == SCTP_FADDRS_UNREACH)
@@ -462,13 +490,18 @@ sctp_compare_faddrsets(sctp_faddr_t *a1, sctp_faddr_t *a2)
}
/*
- * Caller must hold conn fanout lock.
+ * Returns 0 on success, -1 on memory allocation failure. If sleep
+ * is true, this function should never fail. The boolean parameter
+ * first decides whether the newly created faddr structure should be
+ * added at the beginning of the list or at the end.
+ *
+ * Note: caller must hold conn fanout lock.
*/
-static int
-sctp_add_faddr_entry(sctp_t *sctp, in6_addr_t *addr, int sleep,
- boolean_t first)
+int
+sctp_add_faddr(sctp_t *sctp, in6_addr_t *addr, int sleep, boolean_t first)
{
- sctp_faddr_t *faddr;
+ sctp_faddr_t *faddr;
+ mblk_t *timer_mp;
if (is_system_labeled()) {
ts_label_t *tsl;
@@ -521,8 +554,14 @@ sctp_add_faddr_entry(sctp_t *sctp, in6_addr_t *addr, int sleep,
if ((faddr = kmem_cache_alloc(sctp_kmem_faddr_cache, sleep)) == NULL)
return (ENOMEM);
+ timer_mp = sctp_timer_alloc((sctp), sctp_rexmit_timer);
+ if (timer_mp == NULL) {
+ kmem_cache_free(sctp_kmem_faddr_cache, faddr);
+ return (ENOMEM);
+ }
+ ((sctpt_t *)(timer_mp->b_rptr))->sctpt_faddr = faddr;
- sctp_init_faddr(sctp, faddr, addr);
+ sctp_init_faddr(sctp, faddr, addr, timer_mp);
ASSERT(faddr->next == NULL);
if (sctp->sctp_faddrs == NULL) {
@@ -542,36 +581,6 @@ sctp_add_faddr_entry(sctp_t *sctp, in6_addr_t *addr, int sleep,
return (0);
}
-/*
- * Add new address to end of list.
- * Returns 0 on success, or errno on failure:
- * ENOMEM - allocation failure; only for sleep==KM_NOSLEEP
- * EACCES - label is incompatible with caller or connection
- * (labeled [trusted] solaris only)
- * Caller must hold conn fanout lock.
- */
-int
-sctp_add_faddr(sctp_t *sctp, in6_addr_t *addr, int sleep)
-{
- dprint(4, ("add_faddr: %x:%x:%x:%x %d\n", SCTP_PRINTADDR(*addr),
- sleep));
-
- return (sctp_add_faddr_entry(sctp, addr, sleep, B_FALSE));
-}
-
-/*
- * Same as sctp_add_faddr above, but put new entry at front of list.
- * Caller must hold conn fanout lock.
- */
-int
-sctp_add_faddr_first(sctp_t *sctp, in6_addr_t *addr, int sleep)
-{
- dprint(4, ("add_faddr_first: %x:%x:%x:%x %d\n", SCTP_PRINTADDR(*addr),
- sleep));
-
- return (sctp_add_faddr_entry(sctp, addr, sleep, B_TRUE));
-}
-
sctp_faddr_t *
sctp_lookup_faddr(sctp_t *sctp, in6_addr_t *addr)
{
@@ -597,21 +606,16 @@ sctp_lookup_faddr_nosctp(sctp_faddr_t *fp, in6_addr_t *addr)
return (fp);
}
+/*
+ * To change the currently used peer address to the specified one.
+ */
void
-sctp_faddr2hdraddr(sctp_faddr_t *fp, sctp_t *sctp)
+sctp_set_faddr_current(sctp_t *sctp, sctp_faddr_t *fp)
{
+ /* Now setup the composite header. */
if (fp->isv4) {
IN6_V4MAPPED_TO_IPADDR(&fp->faddr,
sctp->sctp_ipha->ipha_dst);
- /* Must not allow unspec src addr if not bound to all */
- if (IN6_IS_ADDR_V4MAPPED_ANY(&fp->saddr) &&
- !sctp->sctp_bound_to_all) {
- /*
- * set the src to the first v4 saddr and hope
- * for the best
- */
- fp->saddr = sctp_get_valid_addr(sctp, B_FALSE);
- }
IN6_V4MAPPED_TO_IPADDR(&fp->saddr, sctp->sctp_ipha->ipha_src);
/* update don't fragment bit */
if (fp->df) {
@@ -622,17 +626,15 @@ sctp_faddr2hdraddr(sctp_faddr_t *fp, sctp_t *sctp)
}
} else {
sctp->sctp_ip6h->ip6_dst = fp->faddr;
- /* Must not allow unspec src addr if not bound to all */
- if (IN6_IS_ADDR_UNSPECIFIED(&fp->saddr) &&
- !sctp->sctp_bound_to_all) {
- /*
- * set the src to the first v6 saddr and hope
- * for the best
- */
- fp->saddr = sctp_get_valid_addr(sctp, B_TRUE);
- }
sctp->sctp_ip6h->ip6_src = fp->saddr;
}
+
+ sctp->sctp_current = fp;
+ sctp->sctp_mss = fp->sfa_pmss;
+
+ /* Update the uppper layer for the change. */
+ if (!SCTP_IS_DETACHED(sctp))
+ sctp_set_ulp_prop(sctp);
}
void
@@ -641,10 +643,8 @@ sctp_redo_faddr_srcs(sctp_t *sctp)
sctp_faddr_t *fp;
for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next) {
- sctp_ire2faddr(sctp, fp);
+ sctp_get_ire(sctp, fp);
}
-
- sctp_faddr2hdraddr(sctp->sctp_current, sctp);
}
void
@@ -661,20 +661,20 @@ sctp_faddr_alive(sctp_t *sctp, sctp_faddr_t *fp)
fp->state = SCTP_FADDRS_ALIVE;
sctp_intf_event(sctp, fp->faddr, SCTP_ADDR_AVAILABLE, 0);
- /* If this is the primary, switch back to it now */
+ /*
+ * If this is the primary, switch back to it now. And
+ * we probably want to reset the source addr used to reach
+ * it.
+ */
if (fp == sctp->sctp_primary) {
- sctp->sctp_current = fp;
- sctp->sctp_mss = fp->sfa_pmss;
- /* Reset the addrs in the composite header */
- sctp_faddr2hdraddr(fp, sctp);
- if (!SCTP_IS_DETACHED(sctp)) {
- sctp_set_ulp_prop(sctp);
- }
+ sctp_set_faddr_current(sctp, fp);
+ sctp_get_ire(sctp, fp);
+ return;
}
}
if (fp->ire == NULL) {
/* Should have a full IRE now */
- sctp_ire2faddr(sctp, fp);
+ sctp_get_ire(sctp, fp);
}
}
@@ -719,7 +719,7 @@ sctp_faddr_dead(sctp_t *sctp, sctp_faddr_t *fp, int newstate)
/* Find next alive faddr */
ofp = fp;
- for (fp = fp->next; fp; fp = fp->next) {
+ for (fp = fp->next; fp != NULL; fp = fp->next) {
if (fp->state == SCTP_FADDRS_ALIVE) {
break;
}
@@ -734,19 +734,19 @@ sctp_faddr_dead(sctp_t *sctp, sctp_faddr_t *fp, int newstate)
}
}
+ /*
+ * Find a new fp, so if the current faddr is dead, use the new fp
+ * as the current one.
+ */
if (fp != ofp) {
if (sctp->sctp_current == NULL) {
dprint(1, ("sctp_faddr_dead: failover->%x:%x:%x:%x\n",
SCTP_PRINTADDR(fp->faddr)));
- sctp->sctp_current = fp;
- sctp->sctp_mss = fp->sfa_pmss;
-
- /* Reset the addrs in the composite header */
- sctp_faddr2hdraddr(fp, sctp);
-
- if (!SCTP_IS_DETACHED(sctp)) {
- sctp_set_ulp_prop(sctp);
- }
+ /*
+ * Note that we don't need to reset the source addr
+ * of the new fp.
+ */
+ sctp_set_faddr_current(sctp, fp);
}
return (0);
}
@@ -1120,9 +1120,8 @@ sctp_v6_label(sctp_t *sctp)
/*
* XXX implement more sophisticated logic
*/
-/* ARGSUSED */
int
-sctp_set_hdraddrs(sctp_t *sctp, cred_t *cr)
+sctp_set_hdraddrs(sctp_t *sctp)
{
sctp_faddr_t *fp;
int gotv4 = 0;
@@ -1352,7 +1351,7 @@ sctp_get_addrparams(sctp_t *sctp, sctp_t *psctp, mblk_t *pkt,
fp = sctp_lookup_faddr(sctp, hdrsaddr);
if (fp == NULL) {
/* not included; add it now */
- err = sctp_add_faddr_first(sctp, hdrsaddr, KM_NOSLEEP);
+ err = sctp_add_faddr(sctp, hdrsaddr, KM_NOSLEEP, B_TRUE);
if (err != 0)
return (err);
@@ -1453,7 +1452,8 @@ sctp_get_addrparams(sctp_t *sctp, sctp_t *psctp, mblk_t *pkt,
goto next;
/* OK, add it to the faddr set */
- err = sctp_add_faddr(sctp, &addr, KM_NOSLEEP);
+ err = sctp_add_faddr(sctp, &addr, KM_NOSLEEP,
+ B_FALSE);
if (err != 0)
return (err);
}
@@ -1480,7 +1480,8 @@ sctp_get_addrparams(sctp_t *sctp, sctp_t *psctp, mblk_t *pkt,
goto next;
err = sctp_add_faddr(sctp,
- (in6_addr_t *)(ph + 1), KM_NOSLEEP);
+ (in6_addr_t *)(ph + 1), KM_NOSLEEP,
+ B_FALSE);
if (err != 0)
return (err);
}
@@ -1510,8 +1511,10 @@ next:
asize = sizeof (in6_addr_t) * sctp->sctp_nfaddrs;
alist = kmem_alloc(asize, KM_NOSLEEP);
- if (alist == NULL)
+ if (alist == NULL) {
+ SCTP_KSTAT(sctp_cl_assoc_change);
return (ENOMEM);
+ }
/*
* Just include the address the INIT was sent to in the
* delete list and send the entire faddr list. We could
@@ -1524,6 +1527,7 @@ next:
dlist = kmem_alloc(dsize, KM_NOSLEEP);
if (dlist == NULL) {
kmem_free(alist, asize);
+ SCTP_KSTAT(sctp_cl_assoc_change);
return (ENOMEM);
}
bcopy(&curaddr, dlist, sizeof (curaddr));
@@ -1771,7 +1775,8 @@ sctp_congest_reset(sctp_t *sctp)
}
static void
-sctp_init_faddr(sctp_t *sctp, sctp_faddr_t *fp, in6_addr_t *addr)
+sctp_init_faddr(sctp_t *sctp, sctp_faddr_t *fp, in6_addr_t *addr,
+ mblk_t *timer_mp)
{
bcopy(addr, &fp->faddr, sizeof (*addr));
if (IN6_IS_ADDR_V4MAPPED(addr)) {
@@ -1798,7 +1803,7 @@ sctp_init_faddr(sctp_t *sctp, sctp_faddr_t *fp, in6_addr_t *addr)
fp->pba = 0;
fp->acked = 0;
fp->lastactive = lbolt64;
- fp->timer_mp = NULL;
+ fp->timer_mp = timer_mp;
fp->hb_pending = B_FALSE;
fp->timer_running = 0;
fp->df = 1;
@@ -1812,7 +1817,7 @@ sctp_init_faddr(sctp_t *sctp, sctp_faddr_t *fp, in6_addr_t *addr)
sizeof (fp->hb_secret));
fp->hb_expiry = lbolt64;
- sctp_ire2faddr(sctp, fp);
+ sctp_get_ire(sctp, fp);
}
/*ARGSUSED*/
diff --git a/usr/src/uts/common/inet/sctp/sctp_conn.c b/usr/src/uts/common/inet/sctp/sctp_conn.c
index 66db656acf..31d5220202 100644
--- a/usr/src/uts/common/inet/sctp/sctp_conn.c
+++ b/usr/src/uts/common/inet/sctp/sctp_conn.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -93,7 +94,7 @@ sctp_accept_comm(sctp_t *listener, sctp_t *acceptor, mblk_t *cr_pkt,
cr = lconnp->conn_cred;
}
- if ((err = sctp_set_hdraddrs(acceptor, cr)) != 0)
+ if ((err = sctp_set_hdraddrs(acceptor)) != 0)
return (err);
if ((sctp_options & SCTP_PRSCTP_OPTION) &&
@@ -275,6 +276,7 @@ sctp_conn_request(sctp_t *sctp, mblk_t *mp, uint_t ifindex, uint_t ip_hdr_len,
kmem_free(flist, fsize);
sctp_close_eager(eager);
BUMP_MIB(&sctp_mib, sctpListenDrop);
+ SCTP_KSTAT(sctp_cl_connect);
return (NULL);
}
/* The clustering module frees these list */
@@ -477,7 +479,8 @@ sctp_connect(sctp_t *sctp, const struct sockaddr *dst, uint32_t addrlen)
* OK; set up the peer addr (this may grow after we get
* the INIT ACK from the peer with additional addresses).
*/
- if ((err = sctp_add_faddr(sctp, &dstaddr, sleep)) != 0) {
+ if ((err = sctp_add_faddr(sctp, &dstaddr, sleep,
+ B_FALSE)) != 0) {
mutex_exit(&tbf->tf_lock);
WAKE_SCTP(sctp);
return (err);
@@ -496,7 +499,7 @@ sctp_connect(sctp_t *sctp, const struct sockaddr *dst, uint32_t addrlen)
mutex_exit(&tbf->tf_lock);
/* initialize composite headers */
- if ((err = sctp_set_hdraddrs(sctp, NULL)) != 0) {
+ if ((err = sctp_set_hdraddrs(sctp)) != 0) {
sctp_conn_hash_remove(sctp);
WAKE_SCTP(sctp);
return (err);
diff --git a/usr/src/uts/common/inet/sctp/sctp_cookie.c b/usr/src/uts/common/inet/sctp/sctp_cookie.c
index 71fff0067a..e84e21eeec 100644
--- a/usr/src/uts/common/inet/sctp/sctp_cookie.c
+++ b/usr/src/uts/common/inet/sctp/sctp_cookie.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -744,6 +745,7 @@ sctp_send_cookie_ack(sctp_t *sctp)
camp = sctp_make_mp(sctp, NULL, sizeof (*cach));
if (camp == NULL) {
/* XXX should abort, but don't have the inmp anymore */
+ SCTP_KSTAT(sctp_send_cookie_ack_failed);
return;
}
@@ -1024,6 +1026,7 @@ sendcookie:
SCTP_FADDR_TIMER_RESTART(sctp, fp, fp->rto);
if (errmp != NULL)
freeb(errmp);
+ SCTP_KSTAT(sctp_send_cookie_failed);
return;
}
/*
diff --git a/usr/src/uts/common/inet/sctp/sctp_error.c b/usr/src/uts/common/inet/sctp/sctp_error.c
index 9b4cca78e3..e5ec99104d 100644
--- a/usr/src/uts/common/inet/sctp/sctp_error.c
+++ b/usr/src/uts/common/inet/sctp/sctp_error.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -102,8 +103,11 @@ sctp_user_abort(sctp_t *sctp, mblk_t *data, boolean_t tbit)
sctp_faddr_t *fp = sctp->sctp_current;
mp = sctp_make_mp(sctp, fp, 0);
- if (mp == NULL)
+ if (mp == NULL) {
+ SCTP_KSTAT(sctp_send_user_abort_failed);
return;
+ }
+
/*
* Create abort chunk.
*/
@@ -366,6 +370,7 @@ sctp_send_err(sctp_t *sctp, mblk_t *emp, sctp_faddr_t *dest)
} else {
sendmp = sctp_make_mp(sctp, dest, 0);
if (sendmp == NULL) {
+ SCTP_KSTAT(sctp_send_err_failed);
freemsg(emp);
return;
}
diff --git a/usr/src/uts/common/inet/sctp/sctp_heartbeat.c b/usr/src/uts/common/inet/sctp/sctp_heartbeat.c
index 8da1e7d9b8..652cef09a7 100644
--- a/usr/src/uts/common/inet/sctp/sctp_heartbeat.c
+++ b/usr/src/uts/common/inet/sctp/sctp_heartbeat.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -81,6 +82,7 @@ sctp_return_heartbeat(sctp_t *sctp, sctp_chunk_hdr_t *hbcp, mblk_t *mp)
/* Create an IP header, returning to the src addr from the heartbt */
smp = sctp_make_mp(sctp, fp, len);
if (smp == NULL) {
+ SCTP_KSTAT(sctp_return_hb_failed);
return;
}
@@ -126,8 +128,10 @@ sctp_send_heartbeat(sctp_t *sctp, sctp_faddr_t *fp)
sizeof (fp->hb_secret) +
sizeof (fp->faddr);
hbmp = sctp_make_mp(sctp, fp, hblen);
- if (hbmp == NULL)
+ if (hbmp == NULL) {
+ SCTP_KSTAT(sctp_send_hb_failed);
return;
+ }
cp = (sctp_chunk_hdr_t *)hbmp->b_wptr;
cp->sch_id = CHUNK_HEARTBEAT;
diff --git a/usr/src/uts/common/inet/sctp/sctp_impl.h b/usr/src/uts/common/inet/sctp/sctp_impl.h
index 42fff07569..5251bd5d34 100644
--- a/usr/src/uts/common/inet/sctp/sctp_impl.h
+++ b/usr/src/uts/common/inet/sctp/sctp_impl.h
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -62,36 +63,6 @@ typedef struct sctpt_s {
} sctpt_t;
/*
- * faddr timer mblks are not allocated until first use. This macro
- * will allocate the timer mblk if necessary, set the faddr, and then
- * start the timer.
- */
-#define SCTP_FADDR_TIMER_RESTART(sctp, fp, intvl) \
- if ((fp)->timer_mp == NULL) { \
- (fp)->timer_mp = sctp_timer_alloc((sctp), sctp_rexmit_timer); \
- } \
- if ((fp)->timer_mp != NULL) { \
- ((sctpt_t *)((fp)->timer_mp->b_rptr))->sctpt_faddr = fp; \
- dprint(3, ("faddr_timer_restart: fp=%p %x:%x:%x:%x %d\n", \
- (void *)(fp), SCTP_PRINTADDR((fp)->faddr), \
- (int)(intvl))); \
- sctp_timer((sctp), (fp)->timer_mp, (intvl)); \
- (fp)->timer_running = 1; \
- }
-
-#define SCTP_FADDR_TIMER_STOP(fp) \
- if ((fp)->timer_running && (fp)->timer_mp != NULL) { \
- sctp_timer_stop((fp)->timer_mp); \
- (fp)->timer_running = 0; \
- }
-
-#define SCTP_CALC_RXT(fp, max) \
-{ \
- if (((fp)->rto <<= 1) > (max)) \
- (fp)->rto = (max); \
-}
-
-/*
* Maximum number of duplicate TSNs we can report. This is currently
* static, and governs the size of the mblk used to hold the duplicate
* reports. The use of duplcate TSN reports is currently experimental,
@@ -184,6 +155,31 @@ extern sctpparam_t sctp_param_arr[];
extern sctpparam_t sctp_wroff_xtra_param;
#define sctp_wroff_xtra sctp_wroff_xtra_param.sctp_param_val
+/*
+ * Retransmission timer start and stop macro for a given faddr.
+ */
+#define SCTP_FADDR_TIMER_RESTART(sctp, fp, intvl) \
+{ \
+ dprint(3, ("faddr_timer_restart: fp=%p %x:%x:%x:%x %d\n", \
+ (void *)(fp), SCTP_PRINTADDR((fp)->faddr), (int)(intvl))); \
+ sctp_timer((sctp), (fp)->timer_mp, (intvl)); \
+ (fp)->timer_running = 1; \
+}
+
+#define SCTP_FADDR_TIMER_STOP(fp) \
+ ASSERT((fp)->timer_mp != NULL); \
+ if ((fp)->timer_running) { \
+ sctp_timer_stop((fp)->timer_mp); \
+ (fp)->timer_running = 0; \
+ }
+
+#define SCTP_CALC_RXT(fp, max) \
+{ \
+ if (((fp)->rto <<= 1) > (max)) \
+ (fp)->rto = (max); \
+}
+
+
#define SCTP_MAX_COMBINED_HEADER_LENGTH (60 + 12) /* Maxed out ip + sctp */
#define SCTP_MAX_IP_OPTIONS_LENGTH (60 - IP_SIMPLE_HDR_LENGTH)
#define SCTP_MAX_HDR_LENGTH 60
@@ -385,6 +381,40 @@ extern sin6_t sctp_sin6_null; /* Zero address for quick clears */
extern mib2_sctp_t sctp_mib; /* SNMP fixed size info */
+/* SCTP kstat */
+typedef struct sctp_kstat_s {
+ kstat_named_t sctp_add_faddr;
+ kstat_named_t sctp_add_timer;
+ kstat_named_t sctp_conn_create;
+ kstat_named_t sctp_find_next_tq;
+ kstat_named_t sctp_fr_add_hdr;
+ kstat_named_t sctp_fr_not_found;
+ kstat_named_t sctp_output_failed;
+ kstat_named_t sctp_rexmit_failed;
+ kstat_named_t sctp_send_init_failed;
+ kstat_named_t sctp_send_cookie_failed;
+ kstat_named_t sctp_send_cookie_ack_failed;
+ kstat_named_t sctp_send_err_failed;
+ kstat_named_t sctp_send_sack_failed;
+ kstat_named_t sctp_send_shutdown_failed;
+ kstat_named_t sctp_send_shutdown_ack_failed;
+ kstat_named_t sctp_send_shutdown_comp_failed;
+ kstat_named_t sctp_send_user_abort_failed;
+ kstat_named_t sctp_send_asconf_failed;
+ kstat_named_t sctp_send_asconf_ack_failed;
+ kstat_named_t sctp_send_ftsn_failed;
+ kstat_named_t sctp_send_hb_failed;
+ kstat_named_t sctp_return_hb_failed;
+ kstat_named_t sctp_ss_rexmit_failed;
+ kstat_named_t sctp_cl_connect;
+ kstat_named_t sctp_cl_assoc_change;
+ kstat_named_t sctp_cl_check_addrs;
+} sctp_kstat_t;
+
+extern sctp_kstat_t sctp_statistics;
+
+#define SCTP_KSTAT(x) (sctp_statistics.x.value.ui64++)
+
/*
* Object to represent database of options to search passed to
* {sock,tpi}optcom_req() interface routine to take care of option
@@ -744,7 +774,9 @@ typedef struct sctp_s {
sctp_prsctp_aware : 1, /* is peer PR-SCTP aware? */
sctp_linklocal : 1, /* is linklocal assoc. */
sctp_mac_exempt : 1, /* SO_MAC_EXEMPT */
- sctp_dummy : 5;
+ sctp_rexmitting : 1, /* SCTP is retransmitting */
+
+ sctp_dummy : 4;
} sctp_bits;
struct {
uint32_t
@@ -787,6 +819,7 @@ typedef struct sctp_s {
#define sctp_prsctp_aware sctp_bits.sctp_prsctp_aware
#define sctp_linklocal sctp_bits.sctp_linklocal
#define sctp_mac_exempt sctp_bits.sctp_mac_exempt
+#define sctp_rexmitting sctp_bits.sctp_rexmitting
#define sctp_recvsndrcvinfo sctp_events.sctp_recvsndrcvinfo
#define sctp_recvassocevnt sctp_events.sctp_recvassocevnt
@@ -903,6 +936,8 @@ typedef struct sctp_s {
uint_t sctp_v4label_len; /* length of cached v4 label */
uint_t sctp_v6label_len; /* length of cached v6 label */
+ uint32_t sctp_rxt_nxttsn; /* Next TSN to be rexmitted */
+ uint32_t sctp_rxt_maxtsn; /* Max TSN sent at time out */
} sctp_t;
extern list_t sctp_g_list; /* Head of SCTP instance data chain */
@@ -919,8 +954,7 @@ extern mblk_t *sctp_pad_mp;
extern void sctp_ack_timer(sctp_t *);
extern size_t sctp_adaption_code_param(sctp_t *, uchar_t *);
extern void sctp_adaption_event(sctp_t *);
-extern int sctp_add_faddr(sctp_t *, in6_addr_t *, int);
-extern int sctp_add_faddr_first(sctp_t *, in6_addr_t *, int);
+extern int sctp_add_faddr(sctp_t *, in6_addr_t *, int, boolean_t);
extern boolean_t sctp_add_ftsn_set(sctp_ftsn_set_t **, sctp_faddr_t *, mblk_t *,
uint_t *, uint32_t *);
extern boolean_t sctp_add_recvq(sctp_t *, mblk_t *, boolean_t);
@@ -974,8 +1008,6 @@ extern void sctp_faddr_alive(sctp_t *, sctp_faddr_t *);
extern int sctp_faddr_dead(sctp_t *, sctp_faddr_t *, int);
extern void sctp_faddr_fini(void);
extern void sctp_faddr_init(void);
-extern void sctp_faddr2hdraddr(sctp_faddr_t *, sctp_t *);
-extern void sctp_faddr2ire(sctp_t *, sctp_faddr_t *);
extern void sctp_fast_rexmit(sctp_t *);
extern void sctp_fill_sack(sctp_t *, unsigned char *, int);
extern void sctp_free_faddr_timers(sctp_t *);
@@ -990,6 +1022,7 @@ extern int sctp_get_addrlist(sctp_t *, const void *, uint32_t *,
uchar_t **, int *, size_t *);
extern int sctp_get_addrparams(sctp_t *, sctp_t *, mblk_t *,
sctp_chunk_hdr_t *, uint_t *);
+extern void sctp_get_ire(sctp_t *, sctp_faddr_t *);
extern void sctp_get_faddr_list(sctp_t *, uchar_t *, size_t);
extern mblk_t *sctp_get_first_sent(sctp_t *);
extern mblk_t *sctp_get_msg_to_send(sctp_t *, mblk_t **, mblk_t *, int *,
@@ -1014,7 +1047,6 @@ extern uint32_t sctp_init2vtag(sctp_chunk_hdr_t *);
extern void sctp_intf_event(sctp_t *, in6_addr_t, int, int);
extern void sctp_input_data(sctp_t *, mblk_t *, mblk_t *);
extern void sctp_instream_cleanup(sctp_t *, boolean_t);
-extern void sctp_ire2faddr(sctp_t *, sctp_faddr_t *);
extern int sctp_is_a_faddr_clean(sctp_t *);
extern void sctp_kstat_init(void);
@@ -1075,16 +1107,19 @@ extern void sctp_send_initack(sctp_t *, sctp_chunk_hdr_t *, mblk_t *);
extern void sctp_send_shutdown(sctp_t *, int);
extern void sctp_send_heartbeat(sctp_t *, sctp_faddr_t *);
extern void sctp_sendfail_event(sctp_t *, mblk_t *, int, boolean_t);
-extern int sctp_set_hdraddrs(sctp_t *, cred_t *);
+extern void sctp_set_faddr_current(sctp_t *, sctp_faddr_t *);
+extern int sctp_set_hdraddrs(sctp_t *);
extern void sctp_sets_init(void);
extern void sctp_sets_fini(void);
extern void sctp_shutdown_event(sctp_t *);
extern void sctp_stop_faddr_timers(sctp_t *);
-extern int sctp_shutdown_received(sctp_t *, sctp_chunk_hdr_t *, int, int);
+extern int sctp_shutdown_received(sctp_t *, sctp_chunk_hdr_t *, boolean_t,
+ boolean_t, sctp_faddr_t *);
extern void sctp_shutdown_complete(sctp_t *);
extern void sctp_set_if_mtu(sctp_t *);
extern void sctp_set_iplen(sctp_t *, mblk_t *);
extern void sctp_set_ulp_prop(sctp_t *);
+extern void sctp_ss_rexmit(sctp_t *);
extern size_t sctp_supaddr_param_len(sctp_t *);
extern size_t sctp_supaddr_param(sctp_t *, uchar_t *);
@@ -1095,6 +1130,7 @@ extern void sctp_timer_free(mblk_t *);
extern void sctp_timer_stop(mblk_t *);
extern void sctp_unlink_faddr(sctp_t *, sctp_faddr_t *);
+extern void sctp_update_ire(sctp_t *sctp);
extern in_port_t sctp_update_next_port(in_port_t, zone_t *zone);
extern void sctp_update_rtt(sctp_t *, sctp_faddr_t *, clock_t);
extern void sctp_user_abort(sctp_t *, mblk_t *, boolean_t);
diff --git a/usr/src/uts/common/inet/sctp/sctp_init.c b/usr/src/uts/common/inet/sctp/sctp_init.c
index f1d0c54b34..5585d42213 100644
--- a/usr/src/uts/common/inet/sctp/sctp_init.c
+++ b/usr/src/uts/common/inet/sctp/sctp_init.c
@@ -2,9 +2,8 @@
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License"). You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
*
* CDDL HEADER END
*/
+
/*
- * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
@@ -202,8 +202,10 @@ sctp_init_mp(sctp_t *sctp)
sctp->sctp_sctph6->sh_verf = 0;
mp = sctp_make_mp(sctp, NULL, initlen);
- if (mp == NULL)
+ if (mp == NULL) {
+ SCTP_KSTAT(sctp_send_init_failed);
return (NULL);
+ }
/* Lay in a new INIT chunk, starting with the chunk header */
chp = (sctp_chunk_hdr_t *)mp->b_wptr;
diff --git a/usr/src/uts/common/inet/sctp/sctp_input.c b/usr/src/uts/common/inet/sctp/sctp_input.c
index e17b8ff3ec..8b2218d862 100644
--- a/usr/src/uts/common/inet/sctp/sctp_input.c
+++ b/usr/src/uts/common/inet/sctp/sctp_input.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -1648,6 +1649,7 @@ checks_done:
(sizeof (sctp_sack_frag_t) * sctp->sctp_sack_gaps);
smp = sctp_make_mp(sctp, sendto, slen);
if (smp == NULL) {
+ SCTP_KSTAT(sctp_send_sack_failed);
return (NULL);
}
sch = (sctp_chunk_hdr_t *)smp->b_wptr;
@@ -1742,6 +1744,7 @@ sctp_check_abandoned_msg(sctp_t *sctp, mblk_t *meta)
if (head == NULL) {
sctp->sctp_adv_pap = adv_pap;
freemsg(nmp);
+ SCTP_KSTAT(sctp_send_ftsn_failed);
return (ENOMEM);
}
SCTP_MSG_SET_ABANDONED(meta);
@@ -1831,6 +1834,7 @@ sctp_cumack(sctp_t *sctp, uint32_t tsn, mblk_t **first_unacked)
}
if (SCTP_CHUNK_ISACKED(mp))
continue;
+ SCTP_CHUNK_SET_SACKCNT(mp, 0);
SCTP_CHUNK_ACKED(mp);
ASSERT(fp->suna >= chunklen);
fp->suna -= chunklen;
@@ -2780,7 +2784,7 @@ ret:
fp->cwnd = fp->suna + sctp_maxburst * fp->sfa_pmss;
}
fp->acked = 0;
- return (trysend);
+ goto check_ss_rxmit;
}
for (fp = sctp->sctp_faddrs; fp; fp = fp->next) {
if (cumack_forward && fp->acked && !fast_recovery &&
@@ -2807,6 +2811,31 @@ ret:
}
fp->acked = 0;
}
+check_ss_rxmit:
+ /*
+ * If this is a SACK following a timeout, check if there are
+ * still unacked chunks (sent before the timeout) that we can
+ * send.
+ */
+ if (sctp->sctp_rexmitting) {
+ if (SEQ_LT(sctp->sctp_lastack_rxd, sctp->sctp_rxt_maxtsn)) {
+ /*
+ * As we are in retransmission phase, we may get a
+ * SACK which indicates some new chunks are received
+ * but cum_tsn does not advance. During this
+ * phase, the other side advances cum_tsn only because
+ * it receives our retransmitted chunks. Only
+ * this signals that some chunks are still
+ * missing.
+ */
+ if (cumack_forward)
+ sctp_ss_rexmit(sctp);
+ } else {
+ sctp->sctp_rexmitting = B_FALSE;
+ sctp->sctp_rxt_nxttsn = sctp->sctp_ltsn;
+ sctp->sctp_rxt_maxtsn = sctp->sctp_ltsn;
+ }
+ }
return (trysend);
}
@@ -3551,7 +3580,7 @@ sctp_input_data(sctp_t *sctp, mblk_t *mp, mblk_t *ipsec_mp)
case CHUNK_SHUTDOWN:
sctp_shutdown_event(sctp);
trysend = sctp_shutdown_received(sctp, ch,
- 0, 0);
+ B_FALSE, B_FALSE, fp);
BUMP_LOCAL(sctp->sctp_ibchunks);
break;
case CHUNK_SHUTDOWN_ACK:
@@ -3906,10 +3935,7 @@ sctp_input_data(sctp_t *sctp, mblk_t *mp, mblk_t *ipsec_mp)
switch (ch->sch_id) {
case CHUNK_ABORT:
/* Pass gathered wisdom to IP for keeping */
- for (fp = sctp->sctp_faddrs; fp != NULL;
- fp = fp->next) {
- sctp_faddr2ire(sctp, fp);
- }
+ sctp_update_ire(sctp);
sctp_process_abort(sctp, ch, 0);
goto done;
case CHUNK_SHUTDOWN_COMPLETE:
@@ -3919,10 +3945,7 @@ sctp_input_data(sctp_t *sctp, mblk_t *mp, mblk_t *ipsec_mp)
NULL);
/* Pass gathered wisdom to IP for keeping */
- for (fp = sctp->sctp_faddrs; fp != NULL;
- fp = fp->next) {
- sctp_faddr2ire(sctp, fp);
- }
+ sctp_update_ire(sctp);
sctp_clean_death(sctp, 0);
goto done;
case CHUNK_SHUTDOWN_ACK:
@@ -3935,7 +3958,7 @@ sctp_input_data(sctp_t *sctp, mblk_t *mp, mblk_t *ipsec_mp)
goto done;
case CHUNK_COOKIE:
(void) sctp_shutdown_received(sctp, NULL,
- 1, 0);
+ B_TRUE, B_FALSE, fp);
BUMP_LOCAL(sctp->sctp_ibchunks);
break;
case CHUNK_HEARTBEAT:
@@ -3953,7 +3976,7 @@ sctp_input_data(sctp_t *sctp, mblk_t *mp, mblk_t *ipsec_mp)
switch (ch->sch_id) {
case CHUNK_SHUTDOWN:
trysend = sctp_shutdown_received(sctp, ch,
- 0, 0);
+ B_FALSE, B_FALSE, fp);
break;
case CHUNK_SACK:
trysend = sctp_got_sack(sctp, ch);
diff --git a/usr/src/uts/common/inet/sctp/sctp_opt_data.c b/usr/src/uts/common/inet/sctp/sctp_opt_data.c
index cd263faeca..89ffd6ae60 100644
--- a/usr/src/uts/common/inet/sctp/sctp_opt_data.c
+++ b/usr/src/uts/common/inet/sctp/sctp_opt_data.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -661,11 +662,7 @@ sctp_set_prim(sctp_t *sctp, const void *invalp, uint_t inlen)
if (fp->state != SCTP_FADDRS_ALIVE || fp == sctp->sctp_current) {
return (0);
}
- sctp->sctp_current = fp;
- sctp->sctp_mss = fp->sfa_pmss;
- /* Reset the addrs in the composite header */
- sctp_faddr2hdraddr(fp, sctp);
- sctp_set_ulp_prop(sctp);
+ sctp_set_faddr_current(sctp, fp);
return (0);
}
diff --git a/usr/src/uts/common/inet/sctp/sctp_output.c b/usr/src/uts/common/inet/sctp/sctp_output.c
index 8ded7748c9..a71c5d2f19 100644
--- a/usr/src/uts/common/inet/sctp/sctp_output.c
+++ b/usr/src/uts/common/inet/sctp/sctp_output.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -563,14 +564,14 @@ sctp_add_proto_hdr(sctp_t *sctp, sctp_faddr_t *fp, mblk_t *mp, int sacklen,
* A null fp->ire could mean that the address is 'down'. Similarly,
* it is possible that the address went down, we tried to send an
* heartbeat and ended up setting fp->saddr as unspec because we
- * didn't have any usable source address. In either case
- * sctp_ire2faddr() will try find an IRE, if available, and set
- * the source address, if needed. If we still don't have any
+ * didn't have any usable source address. In either case
+ * sctp_get_ire() will try find an IRE, if available, and set
+ * the source address, if needed. If we still don't have any
* usable source address, fp->state will be SCTP_FADDRS_UNREACH and
* we return EHOSTUNREACH.
*/
if (fp->ire == NULL || SCTP_IS_ADDR_UNSPEC(fp->isv4, fp->saddr)) {
- sctp_ire2faddr(sctp, fp);
+ sctp_get_ire(sctp, fp);
if (fp->state == SCTP_FADDRS_UNREACH) {
if (error != NULL)
*error = EHOSTUNREACH;
@@ -579,8 +580,10 @@ sctp_add_proto_hdr(sctp_t *sctp, sctp_faddr_t *fp, mblk_t *mp, int sacklen,
}
/* Copy in IP header. */
if ((mp->b_rptr - mp->b_datap->db_base) <
- (sctp_wroff_xtra + hdrlen + sacklen) || DB_REF(mp) > 2) {
+ (sctp_wroff_xtra + hdrlen + sacklen) || DB_REF(mp) > 2 ||
+ !IS_P2ALIGNED(DB_BASE(mp), sizeof (ire_t *))) {
mblk_t *nmp;
+
/*
* This can happen if IP headers are adjusted after
* data was moved into chunks, or during retransmission,
@@ -683,6 +686,8 @@ sctp_find_fast_rexmit_mblks(sctp_t *sctp, int *total, sctp_faddr_t **fp)
int msglen;
int extra;
sctp_msg_hdr_t *msg_hdr;
+ sctp_faddr_t *old_fp = NULL;
+ sctp_faddr_t *chunk_fp;
for (meta = sctp->sctp_xmit_head; meta != NULL; meta = meta->b_next) {
msg_hdr = (sctp_msg_hdr_t *)meta->b_rptr;
@@ -694,13 +699,23 @@ sctp_find_fast_rexmit_mblks(sctp_t *sctp, int *total, sctp_faddr_t **fp)
if (SCTP_CHUNK_WANT_REXMIT(mp)) {
/*
* Use the same peer address to do fast
- * retransmission.
+ * retransmission. If the original peer
+ * address is dead, switch to the current
+ * one. Record the old one so that we
+ * will pick the chunks sent to the old
+ * one for fast retransmission.
*/
+ chunk_fp = SCTP_CHUNK_DEST(mp);
if (*fp == NULL) {
- *fp = SCTP_CHUNK_DEST(mp);
- if ((*fp)->state != SCTP_FADDRS_ALIVE)
+ *fp = chunk_fp;
+ if ((*fp)->state != SCTP_FADDRS_ALIVE) {
+ old_fp = *fp;
*fp = sctp->sctp_current;
- } else if (*fp != SCTP_CHUNK_DEST(mp)) {
+ }
+ } else if (old_fp == NULL && *fp != chunk_fp) {
+ continue;
+ } else if (old_fp != NULL &&
+ old_fp != chunk_fp) {
continue;
}
@@ -730,6 +745,7 @@ sctp_find_fast_rexmit_mblks(sctp_t *sctp, int *total, sctp_faddr_t **fp)
}
}
BUMP_MIB(&sctp_mib, sctpOutFastRetrans);
+ BUMP_LOCAL(sctp->sctp_rxtchunks);
SCTP_CHUNK_CLEAR_REXMIT(mp);
if (start_mp == NULL) {
start_mp = nmp;
@@ -927,10 +943,13 @@ sctp_fast_rexmit(sctp_t *sctp)
ASSERT(sctp->sctp_xmit_head != NULL);
mp = sctp_find_fast_rexmit_mblks(sctp, &pktlen, &fp);
- if (mp == NULL)
+ if (mp == NULL) {
+ SCTP_KSTAT(sctp_fr_not_found);
return;
+ }
if ((head = sctp_add_proto_hdr(sctp, fp, mp, 0, NULL)) == NULL) {
freemsg(mp);
+ SCTP_KSTAT(sctp_fr_add_hdr);
return;
}
if ((pktlen > fp->sfa_pmss) && fp->isv4) {
@@ -1069,6 +1088,7 @@ sctp_output(sctp_t *sctp)
fp, chunklen, meta);
}
freemsg(nmp);
+ SCTP_KSTAT(sctp_output_failed);
goto unsent_data;
}
seglen += sacklen;
@@ -1086,7 +1106,6 @@ sctp_output(sctp_t *sctp)
fp->cwnd = sctp_slow_start_after_idle *
fp->sfa_pmss;
}
- fp->lastactive = now;
pathmax = fp->cwnd - fp->suna;
if (seglen + extra > pathmax) {
@@ -1115,9 +1134,11 @@ sctp_output(sctp_t *sctp)
fp, chunklen, meta);
}
freemsg(nmp);
+ SCTP_KSTAT(sctp_output_failed);
goto unsent_data;
}
}
+ fp->lastactive = now;
if (pathmax > fp->sfa_pmss)
pathmax = fp->sfa_pmss;
SCTP_CHUNK_SENT(sctp, mp, sdc, fp, chunklen, meta);
@@ -1127,6 +1148,7 @@ sctp_output(sctp_t *sctp)
if (sctp->sctp_out_time == 0) {
sctp->sctp_out_time = now;
sctp->sctp_rtt_tsn = sctp->sctp_ltsn - 1;
+ ASSERT(sctp->sctp_rtt_tsn == ntohl(sdc->sdh_tsn));
}
if (extra > 0) {
fill = sctp_get_padding(extra);
@@ -1460,6 +1482,7 @@ ftsn_done:
if (head == NULL) {
freemsg(*nmp);
*nmp = NULL;
+ SCTP_KSTAT(sctp_send_ftsn_failed);
return;
}
*seglen += sacklen;
@@ -1572,6 +1595,24 @@ sctp_check_adv_ack_pt(sctp_t *sctp, mblk_t *meta, mblk_t *mp)
sctp->sctp_adv_pap = tsn;
}
+
+/*
+ * Determine if we should bundle a data chunk with the chunk being
+ * retransmitted. We bundle if
+ *
+ * - the chunk is sent to the same destination and unack'ed.
+ *
+ * OR
+ *
+ * - the chunk is unsent, i.e. new data.
+ */
+#define SCTP_CHUNK_RX_CANBUNDLE(mp, fp) \
+ (!SCTP_CHUNK_ABANDONED((mp)) && \
+ ((SCTP_CHUNK_ISSENT((mp)) && (SCTP_CHUNK_DEST(mp) == (fp) && \
+ !SCTP_CHUNK_ISACKED(mp))) || \
+ (((mp)->b_flag & (SCTP_CHUNK_FLAG_REXMIT|SCTP_CHUNK_FLAG_SENT)) != \
+ SCTP_CHUNK_FLAG_SENT)))
+
/*
* Retransmit first segment which hasn't been acked with cumtsn or send
* a Forward TSN chunk, if appropriate.
@@ -1590,10 +1631,12 @@ sctp_rexmit(sctp_t *sctp, sctp_faddr_t *oldfp)
int extra;
sctp_data_hdr_t *sdc;
sctp_faddr_t *fp;
- int error;
uint32_t adv_pap = sctp->sctp_adv_pap;
boolean_t do_ftsn = B_FALSE;
boolean_t ftsn_check = B_TRUE;
+ uint32_t first_ua_tsn;
+ sctp_msg_hdr_t *mhdr;
+ uint32_t tot_wnd;
while (meta != NULL) {
for (mp = meta->b_cont; mp != NULL; mp = mp->b_next) {
@@ -1651,7 +1694,7 @@ sctp_rexmit(sctp_t *sctp, sctp_faddr_t *oldfp)
}
meta = meta->b_next;
if (meta != NULL && sctp->sctp_prsctp_aware) {
- sctp_msg_hdr_t *mhdr = (sctp_msg_hdr_t *)meta->b_rptr;
+ mhdr = (sctp_msg_hdr_t *)meta->b_rptr;
while (meta != NULL && (SCTP_IS_MSG_ABANDONED(meta) ||
SCTP_MSG_TO_BE_ABANDONED(meta, mhdr, sctp))) {
@@ -1689,6 +1732,7 @@ out:
ASSERT(fp != NULL);
sdc = (sctp_data_hdr_t *)mp->b_rptr;
+ first_ua_tsn = ntohl(sdc->sdh_tsn);
if (do_ftsn) {
sctp_make_ftsns(sctp, meta, mp, &nmp, fp, &seglen);
if (nmp == NULL) {
@@ -1696,10 +1740,16 @@ out:
goto restart_timer;
}
head = nmp;
- mp = NULL;
- meta = sctp->sctp_xmit_tail;
+ /*
+ * Move to the next unabandoned chunk. XXXCheck if meta will
+ * always be marked abandoned.
+ */
+ while (meta != NULL && SCTP_IS_MSG_ABANDONED(meta))
+ meta = meta->b_next;
if (meta != NULL)
- mp = meta->b_cont;
+ mp = mp->b_cont;
+ else
+ mp = NULL;
goto try_bundle;
}
seglen = ntohs(sdc->sdh_len);
@@ -1707,17 +1757,7 @@ out:
if ((extra = seglen & (SCTP_ALIGN - 1)) != 0)
extra = SCTP_ALIGN - extra;
- /*
- * Cancel RTT measurement if the retransmitted TSN is before the
- * TSN used for timimg.
- */
- if (sctp->sctp_out_time != 0 &&
- SEQ_GEQ(sctp->sctp_rtt_tsn, sdc->sdh_tsn)) {
- sctp->sctp_out_time = 0;
- }
- /* Clear the counter as the RTT calculation may be off. */
- fp->rtt_updates = 0;
-
+ /* Find out if we need to piggyback SACK. */
if (sctp->sctp_ftsn == sctp->sctp_lastacked + 1) {
sacklen = 0;
} else {
@@ -1728,10 +1768,39 @@ out:
/* piggybacked SACK doesn't fit */
sacklen = 0;
} else {
- fp = sctp->sctp_lastdata;
+ /*
+ * OK, we have room to send SACK back. But we
+ * should send it back to the last fp where we
+ * receive data from, unless sctp_lastdata equals
+ * oldfp, then we should probably not send it
+ * back to that fp. Also we should check that
+ * the fp is alive.
+ */
+ if (sctp->sctp_lastdata != oldfp &&
+ sctp->sctp_lastdata->state == SCTP_FADDRS_ALIVE) {
+ fp = sctp->sctp_lastdata;
+ }
}
}
+ /*
+ * Cancel RTT measurement if the retransmitted TSN is before the
+ * TSN used for timimg.
+ */
+ if (sctp->sctp_out_time != 0 &&
+ SEQ_GEQ(sctp->sctp_rtt_tsn, sdc->sdh_tsn)) {
+ sctp->sctp_out_time = 0;
+ }
+ /* Clear the counter as the RTT calculation may be off. */
+ fp->rtt_updates = 0;
+ oldfp->rtt_updates = 0;
+
+ /*
+ * After a timeout, we should change the current faddr so that
+ * new chunks will be sent to the alternate address.
+ */
+ sctp_set_faddr_current(sctp, fp);
+
nmp = dupmsg(mp);
if (nmp == NULL)
goto restart_timer;
@@ -1749,6 +1818,7 @@ out:
head = sctp_add_proto_hdr(sctp, fp, nmp, sacklen, NULL);
if (head == NULL) {
freemsg(nmp);
+ SCTP_KSTAT(sctp_rexmit_failed);
goto restart_timer;
}
seglen += sacklen;
@@ -1756,55 +1826,80 @@ out:
SCTP_CHUNK_SENT(sctp, mp, sdc, fp, chunklen, meta);
mp = mp->b_next;
+
+ /* Check how much more we can send. */
+ tot_wnd = MIN(fp->cwnd, sctp->sctp_frwnd);
+ /*
+ * If the number of outstanding bytes is more than what we are
+ * allowed to send, stop.
+ */
+ if (tot_wnd <= chunklen || tot_wnd < fp->suna + chunklen)
+ goto done_bundle;
+ else
+ tot_wnd -= chunklen;
+
try_bundle:
while (seglen < fp->sfa_pmss) {
int32_t new_len;
+ /* Go through the list to find more chunks to be bundled. */
while (mp != NULL) {
- if (SCTP_CHUNK_CANSEND(mp))
+ /* Check if the chunk can be bundled. */
+ if (SCTP_CHUNK_RX_CANBUNDLE(mp, oldfp))
break;
mp = mp->b_next;
}
+ /* Go to the next message. */
if (mp == NULL) {
- meta = sctp_get_msg_to_send(sctp, &mp, meta->b_next,
- &error, 0, 0, oldfp);
- if (error != 0 || meta == NULL)
- break;
- ASSERT(mp != NULL);
- sctp->sctp_xmit_tail = meta;
+ for (meta = meta->b_next; meta != NULL;
+ meta = meta->b_next) {
+ mhdr = (sctp_msg_hdr_t *)meta->b_rptr;
+
+ if (SCTP_IS_MSG_ABANDONED(meta) ||
+ SCTP_MSG_TO_BE_ABANDONED(meta, mhdr,
+ sctp)) {
+ continue;
+ }
+
+ mp = meta->b_cont;
+ goto try_bundle;
+ }
+ /* No more chunk to be bundled. */
+ break;
}
+
sdc = (sctp_data_hdr_t *)mp->b_rptr;
- chunklen = ntohs(sdc->sdh_len) - sizeof (*sdc);
- new_len = seglen + ntohs(sdc->sdh_len);
+ new_len = ntohs(sdc->sdh_len);
+ chunklen = new_len - sizeof (*sdc);
+ if (chunklen > tot_wnd)
+ break;
- if (seglen & (SCTP_ALIGN - 1)) {
- extra = SCTP_ALIGN - (seglen & (SCTP_ALIGN - 1));
+ if ((extra = new_len & (SCTP_ALIGN - 1)) != 0)
+ extra = SCTP_ALIGN - extra;
+ if ((new_len = seglen + new_len + extra) > fp->sfa_pmss)
+ break;
+ if ((nmp = dupmsg(mp)) == NULL)
+ break;
- if (new_len + extra > fp->sfa_pmss) {
- break;
- }
+ if (extra > 0) {
fill = sctp_get_padding(extra);
if (fill != NULL) {
- new_len += extra;
- linkb(head, fill);
+ linkb(nmp, fill);
} else {
- break;
- }
- } else {
- if (new_len > fp->sfa_pmss) {
+ freemsg(nmp);
break;
}
}
- if ((nmp = dupmsg(mp)) == NULL) {
- break;
- }
- seglen = new_len;
+ linkb(head, nmp);
SCTP_CHUNK_CLEAR_FLAGS(nmp);
SCTP_CHUNK_SENT(sctp, mp, sdc, fp, chunklen, meta);
- linkb(head, nmp);
+
+ seglen = new_len;
+ tot_wnd -= chunklen;
mp = mp->b_next;
}
+done_bundle:
if ((seglen > fp->sfa_pmss) && fp->isv4) {
ipha_t *iph = (ipha_t *)head->b_rptr;
@@ -1820,20 +1915,38 @@ try_bundle:
seglen, ntohl(sdc->sdh_tsn), ntohs(sdc->sdh_ssn),
(void *)fp, sctp->sctp_frwnd, sctp->sctp_lastack_rxd));
+ sctp->sctp_rexmitting = B_TRUE;
+ sctp->sctp_rxt_nxttsn = first_ua_tsn;
+ sctp->sctp_rxt_maxtsn = sctp->sctp_ltsn - 1;
sctp_set_iplen(sctp, head);
sctp_add_sendq(sctp, head);
/*
- * Restart timer with exponential backoff
+ * Restart the oldfp timer with exponential backoff and
+ * the new fp timer for the retransmitted chunks.
*/
restart_timer:
oldfp->strikes++;
sctp->sctp_strikes++;
SCTP_CALC_RXT(oldfp, sctp->sctp_rto_max);
- SCTP_FADDR_TIMER_RESTART(sctp, fp, fp->rto);
if (oldfp->suna != 0)
SCTP_FADDR_TIMER_RESTART(sctp, oldfp, oldfp->rto);
sctp->sctp_active = lbolt64;
+
+ /*
+ * Should we restart the timer of the new fp? If there is
+ * outstanding data to the new fp, the timer should be
+ * running already. So restarting it means that the timer
+ * will fire later for those outstanding data. But if
+ * we don't restart it, the timer will fire too early for the
+ * just retransmitted chunks to the new fp. The reason is that we
+ * don't keep a timestamp on when a chunk is retransmitted.
+ * So when the timer fires, it will just search for the
+ * chunk with the earliest TSN sent to new fp. This probably
+ * is the chunk we just retransmitted. So for now, let's
+ * be conservative and restart the timer of the new fp.
+ */
+ SCTP_FADDR_TIMER_RESTART(sctp, fp, fp->rto);
}
/*
@@ -1885,3 +1998,241 @@ sctp_wput(queue_t *q, mblk_t *mp)
return;
}
}
+
+/*
+ * This function is called by sctp_ss_rexmit() to create a packet
+ * to be retransmitted to the given fp. The given meta and mp
+ * parameters are respectively the sctp_msg_hdr_t and the mblk of the
+ * first chunk to be retransmitted.
+ */
+static mblk_t *
+sctp_rexmit_packet(sctp_t *sctp, mblk_t **meta, mblk_t **mp, sctp_faddr_t *fp,
+ uint_t *packet_len)
+{
+ uint32_t seglen = 0;
+ uint16_t chunklen;
+ int extra;
+ mblk_t *nmp;
+ mblk_t *head;
+ mblk_t *fill;
+ sctp_data_hdr_t *sdc;
+ sctp_msg_hdr_t *mhdr;
+
+ sdc = (sctp_data_hdr_t *)(*mp)->b_rptr;
+ seglen = ntohs(sdc->sdh_len);
+ chunklen = seglen - sizeof (*sdc);
+ if ((extra = seglen & (SCTP_ALIGN - 1)) != 0)
+ extra = SCTP_ALIGN - extra;
+
+ nmp = dupmsg(*mp);
+ if (nmp == NULL)
+ return (NULL);
+ if (extra > 0) {
+ fill = sctp_get_padding(extra);
+ if (fill != NULL) {
+ linkb(nmp, fill);
+ seglen += extra;
+ } else {
+ freemsg(nmp);
+ return (NULL);
+ }
+ }
+ SCTP_CHUNK_CLEAR_FLAGS(nmp);
+ head = sctp_add_proto_hdr(sctp, fp, nmp, 0, NULL);
+ if (head == NULL) {
+ freemsg(nmp);
+ return (NULL);
+ }
+ SCTP_CHUNK_SENT(sctp, *mp, sdc, fp, chunklen, *meta);
+ sctp->sctp_rxt_nxttsn = ntohl(sdc->sdh_tsn);
+ *mp = (*mp)->b_next;
+
+try_bundle:
+ while (seglen < fp->sfa_pmss) {
+ int32_t new_len;
+
+ /*
+ * Go through the list to find more chunks to be bundled.
+ * We should only retransmit sent by unack'ed chunks. Since
+ * they were sent before, the peer's receive window should
+ * be able to receive them.
+ */
+ while (*mp != NULL) {
+ /* Check if the chunk can be bundled. */
+ if (SCTP_CHUNK_ISSENT(*mp) && !SCTP_CHUNK_ISACKED(*mp))
+ break;
+ *mp = (*mp)->b_next;
+ }
+ /* Go to the next message. */
+ if (*mp == NULL) {
+ for (*meta = (*meta)->b_next; *meta != NULL;
+ *meta = (*meta)->b_next) {
+ mhdr = (sctp_msg_hdr_t *)(*meta)->b_rptr;
+
+ if (SCTP_IS_MSG_ABANDONED(*meta) ||
+ SCTP_MSG_TO_BE_ABANDONED(*meta, mhdr,
+ sctp)) {
+ continue;
+ }
+
+ *mp = (*meta)->b_cont;
+ goto try_bundle;
+ }
+ /* No more chunk to be bundled. */
+ break;
+ }
+
+ sdc = (sctp_data_hdr_t *)(*mp)->b_rptr;
+ /* Don't bundle chunks beyond sctp_rxt_maxtsn. */
+ if (SEQ_GT(ntohl(sdc->sdh_tsn), sctp->sctp_rxt_maxtsn))
+ break;
+ new_len = ntohs(sdc->sdh_len);
+ chunklen = new_len - sizeof (*sdc);
+
+ if ((extra = new_len & (SCTP_ALIGN - 1)) != 0)
+ extra = SCTP_ALIGN - extra;
+ if ((new_len = seglen + new_len + extra) > fp->sfa_pmss)
+ break;
+ if ((nmp = dupmsg(*mp)) == NULL)
+ break;
+
+ if (extra > 0) {
+ fill = sctp_get_padding(extra);
+ if (fill != NULL) {
+ linkb(nmp, fill);
+ } else {
+ freemsg(nmp);
+ break;
+ }
+ }
+ linkb(head, nmp);
+
+ SCTP_CHUNK_CLEAR_FLAGS(nmp);
+ SCTP_CHUNK_SENT(sctp, *mp, sdc, fp, chunklen, *meta);
+ sctp->sctp_rxt_nxttsn = ntohl(sdc->sdh_tsn);
+
+ seglen = new_len;
+ *mp = (*mp)->b_next;
+ }
+ *packet_len = seglen;
+ return (head);
+}
+
+/*
+ * sctp_ss_rexmit() is called when we get a SACK after a timeout which
+ * advances the cum_tsn but the cum_tsn is still less than what we have sent
+ * (sctp_rxt_maxtsn) at the time of the timeout. This SACK is a "partial"
+ * SACK. We retransmit unacked chunks without having to wait for another
+ * timeout. The rationale is that the SACK should not be "partial" if all the
+ * lost chunks have been retransmitted. Since the SACK is "partial,"
+ * the chunks between the cum_tsn and the sctp_rxt_maxtsn should still
+ * be missing. It is better for us to retransmit them now instead
+ * of waiting for a timeout.
+ */
+void
+sctp_ss_rexmit(sctp_t *sctp)
+{
+ mblk_t *meta;
+ mblk_t *mp;
+ mblk_t *pkt;
+ sctp_faddr_t *fp;
+ uint_t pkt_len;
+ uint32_t tot_wnd;
+ sctp_data_hdr_t *sdc;
+ int burst;
+
+ /*
+ * If the last cum ack is smaller than what we have just
+ * retransmitted, simply return.
+ */
+ if (SEQ_GEQ(sctp->sctp_lastack_rxd, sctp->sctp_rxt_nxttsn))
+ sctp->sctp_rxt_nxttsn = sctp->sctp_lastack_rxd + 1;
+ else
+ return;
+
+ ASSERT(SEQ_LEQ(sctp->sctp_rxt_nxttsn, sctp->sctp_rxt_maxtsn));
+
+ /*
+ * After a timer fires, sctp_current should be set to the new
+ * fp where the retransmitted chunks are sent.
+ */
+ fp = sctp->sctp_current;
+
+ /*
+ * Since we are retransmitting, we can only use cwnd to determine
+ * how much we can send as we were allowed to send those chunks
+ * previously.
+ */
+ tot_wnd = fp->cwnd;
+ /* So we have sent more than we can, just return. */
+ if (tot_wnd < fp->suna || tot_wnd - fp->suna < fp->sfa_pmss)
+ return;
+ else
+ tot_wnd -= fp->suna;
+
+ /* Find the first unack'ed chunk */
+ for (meta = sctp->sctp_xmit_head; meta != NULL; meta = meta->b_next) {
+ sctp_msg_hdr_t *mhdr = (sctp_msg_hdr_t *)meta->b_rptr;
+
+ if (SCTP_IS_MSG_ABANDONED(meta) ||
+ SCTP_MSG_TO_BE_ABANDONED(meta, mhdr, sctp)) {
+ continue;
+ }
+
+ for (mp = meta->b_cont; mp != NULL; mp = mp->b_next) {
+ /* Again, this may not be possible */
+ if (!SCTP_CHUNK_ISSENT(mp))
+ return;
+ sdc = (sctp_data_hdr_t *)mp->b_rptr;
+ if (ntohl(sdc->sdh_tsn) == sctp->sctp_rxt_nxttsn)
+ goto found_msg;
+ }
+ }
+
+ /* Everything is abandoned... */
+ return;
+
+found_msg:
+ if (!fp->timer_running)
+ SCTP_FADDR_TIMER_RESTART(sctp, fp, fp->rto);
+ pkt = sctp_rexmit_packet(sctp, &meta, &mp, fp, &pkt_len);
+ if (pkt == NULL) {
+ SCTP_KSTAT(sctp_ss_rexmit_failed);
+ return;
+ }
+ if ((pkt_len > fp->sfa_pmss) && fp->isv4) {
+ ipha_t *iph = (ipha_t *)pkt->b_rptr;
+
+ /*
+ * Path MTU is different from path we thought it would
+ * be when we created chunks, or IP headers have grown.
+ * Need to clear the DF bit.
+ */
+ iph->ipha_fragment_offset_and_flags = 0;
+ }
+ sctp_set_iplen(sctp, pkt);
+ sctp_add_sendq(sctp, pkt);
+
+ /* Check and see if there is more chunk to be retransmitted. */
+ if (tot_wnd <= pkt_len || tot_wnd - pkt_len < fp->sfa_pmss ||
+ meta == NULL)
+ return;
+ if (mp == NULL)
+ meta = meta->b_next;
+ if (meta == NULL)
+ return;
+
+ /* Retransmit another packet if the window allows. */
+ for (tot_wnd -= pkt_len, burst = sctp_maxburst - 1;
+ meta != NULL && burst > 0; meta = meta->b_next, burst--) {
+ if (mp == NULL)
+ mp = meta->b_cont;
+ for (; mp != NULL; mp = mp->b_next) {
+ /* Again, this may not be possible */
+ if (!SCTP_CHUNK_ISSENT(mp))
+ return;
+ if (!SCTP_CHUNK_ISACKED(mp))
+ goto found_msg;
+ }
+ }
+}
diff --git a/usr/src/uts/common/inet/sctp/sctp_shutdown.c b/usr/src/uts/common/inet/sctp/sctp_shutdown.c
index 0fcf98b959..8b52360849 100644
--- a/usr/src/uts/common/inet/sctp/sctp_shutdown.c
+++ b/usr/src/uts/common/inet/sctp/sctp_shutdown.c
@@ -2,9 +2,8 @@
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
- * Common Development and Distribution License, Version 1.0 only
- * (the "License"). You may not use this file except in compliance
- * with the License.
+ * Common Development and Distribution License (the "License").
+ * You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
@@ -19,8 +18,9 @@
*
* CDDL HEADER END
*/
+
/*
- * Copyright 2004 Sun Microsystems, Inc. All rights reserved.
+ * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
@@ -110,6 +110,7 @@ sctp_send_shutdown(sctp_t *sctp, int rexmit)
sendmp = sctp_make_mp(sctp, fp,
sizeof (*sch) + sizeof (*ctsn));
if (sendmp == NULL) {
+ SCTP_KSTAT(sctp_send_shutdown_failed);
goto done;
}
sch = (sctp_chunk_hdr_t *)sendmp->b_wptr;
@@ -142,14 +143,13 @@ done:
}
int
-sctp_shutdown_received(sctp_t *sctp, sctp_chunk_hdr_t *sch, int crwsd,
- int rexmit)
+sctp_shutdown_received(sctp_t *sctp, sctp_chunk_hdr_t *sch, boolean_t crwsd,
+ boolean_t rexmit, sctp_faddr_t *fp)
{
mblk_t *samp;
sctp_chunk_hdr_t *sach;
uint32_t *tsn;
int trysend = 0;
- sctp_faddr_t *fp;
if (sctp->sctp_state != SCTPS_SHUTDOWN_ACK_SENT)
sctp->sctp_state = SCTPS_SHUTDOWN_RECEIVED;
@@ -172,15 +172,20 @@ sctp_shutdown_received(sctp_t *sctp, sctp_chunk_hdr_t *sch, int crwsd,
if (sctp->sctp_xmit_head != NULL || sctp->sctp_xmit_unsent != NULL)
return (1);
- /* rotate faddrs if we are retransmitting */
- if (!rexmit)
- fp = sctp->sctp_current;
- else
- fp = sctp_rotate_faddr(sctp, sctp->sctp_shutdown_faddr);
+ if (fp == NULL) {
+ /* rotate faddrs if we are retransmitting */
+ if (!rexmit)
+ fp = sctp->sctp_current;
+ else
+ fp = sctp_rotate_faddr(sctp, sctp->sctp_shutdown_faddr);
+ }
+ sctp->sctp_shutdown_faddr = fp;
samp = sctp_make_mp(sctp, fp, sizeof (*sach));
- if (samp == NULL)
+ if (samp == NULL) {
+ SCTP_KSTAT(sctp_send_shutdown_ack_failed);
goto dotimer;
+ }
sach = (sctp_chunk_hdr_t *)samp->b_wptr;
sach->sch_id = CHUNK_SHUTDOWN_ACK;
@@ -226,6 +231,7 @@ sctp_shutdown_complete(sctp_t *sctp)
scmp = sctp_make_mp(sctp, NULL, sizeof (*scch));
if (scmp == NULL) {
/* XXX use timer approach */
+ SCTP_KSTAT(sctp_send_shutdown_comp_failed);
return;
}
@@ -271,11 +277,13 @@ sctp_ootb_shutdown_ack(sctp_t *gsctp, mblk_t *inmp, uint_t ip_hdr_len)
/*
* Check to see if we can reuse the incoming mblk. There should
- * not be other reference. Since this packet comes from below,
+ * not be other reference and the db_base of the mblk should be
+ * properly aligned. Since this packet comes from below,
* there should be enough header space to fill in what the lower
* layers want to add. And we will not stash anything there.
*/
- if (DB_REF(inmp) != 1) {
+ if (!IS_P2ALIGNED(DB_BASE(inmp), sizeof (ire_t *)) ||
+ DB_REF(inmp) != 1) {
mp1 = allocb(MBLKL(inmp) + sctp_wroff_xtra, BPRI_MED);
if (mp1 == NULL) {
freeb(inmp);
diff --git a/usr/src/uts/common/inet/sctp/sctp_snmp.c b/usr/src/uts/common/inet/sctp/sctp_snmp.c
index f078fb4c20..d42cce5d81 100644
--- a/usr/src/uts/common/inet/sctp/sctp_snmp.c
+++ b/usr/src/uts/common/inet/sctp/sctp_snmp.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -46,10 +47,47 @@
#include "sctp_addr.h"
mib2_sctp_t sctp_mib;
-kstat_t *sctp_mibkp; /* kstat exporting sctp_mib data */
+static kstat_t *sctp_mibkp; /* kstat exporting sctp_mib data */
+static kstat_t *sctp_kstat; /* kstat exporting general sctp stats */
static int sctp_snmp_state(sctp_t *sctp);
+/*
+ * The following kstats are for debugging purposes. They keep
+ * track of problems which should not happen normally. But in
+ * those cases which they do happen, these kstats would be handy
+ * for engineers to diagnose the problems. They are not intended
+ * to be consumed by customers.
+ */
+sctp_kstat_t sctp_statistics = {
+ { "sctp_add_faddr", KSTAT_DATA_UINT64 },
+ { "sctp_add_timer", KSTAT_DATA_UINT64 },
+ { "sctp_conn_create", KSTAT_DATA_UINT64 },
+ { "sctp_find_next_tq", KSTAT_DATA_UINT64 },
+ { "sctp_fr_add_hdr", KSTAT_DATA_UINT64 },
+ { "sctp_fr_not_found", KSTAT_DATA_UINT64 },
+ { "sctp_output_failed", KSTAT_DATA_UINT64 },
+ { "sctp_rexmit_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_init_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_cookie_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_cookie_ack_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_err_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_sack_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_shutdown_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_shutdown_ack_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_shutdown_comp_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_user_abort_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_asconf_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_asconf_ack_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_ftsn_failed", KSTAT_DATA_UINT64 },
+ { "sctp_send_hb_failed", KSTAT_DATA_UINT64 },
+ { "sctp_return_hb_failed", KSTAT_DATA_UINT64 },
+ { "sctp_ss_rexmit_failed", KSTAT_DATA_UINT64 },
+ { "sctp_cl_connect", KSTAT_DATA_UINT64 },
+ { "sctp_cl_assoc_change", KSTAT_DATA_UINT64 },
+ { "sctp_cl_check_addrs", KSTAT_DATA_UINT64 },
+};
+
static int
sctp_kstat_update(kstat_t *kp, int rw)
{
@@ -261,8 +299,8 @@ sctp_kstat_init(void)
{ "sctpInClosed", KSTAT_DATA_INT32, 0 }
};
- sctp_mibkp = kstat_create("sctp", 0, "sctp", "mib2", KSTAT_TYPE_NAMED,
- NUM_OF_FIELDS(sctp_named_kstat_t), 0);
+ sctp_mibkp = kstat_create(SCTP_MOD_NAME, 0, "sctp", "mib2",
+ KSTAT_TYPE_NAMED, NUM_OF_FIELDS(sctp_named_kstat_t), 0);
if (sctp_mibkp == NULL)
return;
@@ -276,6 +314,13 @@ sctp_kstat_init(void)
sctp_mibkp->ks_update = sctp_kstat_update;
kstat_install(sctp_mibkp);
+
+ if ((sctp_kstat = kstat_create(SCTP_MOD_NAME, 0, "sctpstat",
+ "net", KSTAT_TYPE_NAMED, NUM_OF_FIELDS(sctp_statistics),
+ KSTAT_FLAG_VIRTUAL)) != NULL) {
+ sctp_kstat->ks_data = &sctp_statistics;
+ kstat_install(sctp_kstat);
+ }
}
void
@@ -285,6 +330,10 @@ sctp_kstat_fini(void)
kstat_delete(sctp_mibkp);
sctp_mibkp = NULL;
}
+ if (sctp_kstat != NULL) {
+ kstat_delete(sctp_kstat);
+ sctp_kstat = NULL;
+ }
}
/*
diff --git a/usr/src/uts/common/inet/sctp/sctp_timer.c b/usr/src/uts/common/inet/sctp/sctp_timer.c
index 0b2ae2a7fe..fb19313897 100644
--- a/usr/src/uts/common/inet/sctp/sctp_timer.c
+++ b/usr/src/uts/common/inet/sctp/sctp_timer.c
@@ -18,6 +18,7 @@
*
* CDDL HEADER END
*/
+
/*
* Copyright 2006 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
@@ -166,6 +167,7 @@ sctp_timer_alloc(sctp_t *sctp, pfv_t func)
sctpt->sctpt_pfv = func;
return (mp);
}
+ SCTP_KSTAT(sctp_add_timer);
return (NULL);
}
@@ -578,7 +580,8 @@ sctp_rexmit_timer(sctp_t *sctp, sctp_faddr_t *fp)
case SCTPS_SHUTDOWN_PENDING:
case SCTPS_SHUTDOWN_RECEIVED:
if (sctp->sctp_state == SCTPS_SHUTDOWN_RECEIVED) {
- (void) sctp_shutdown_received(sctp, NULL, 0, 1);
+ (void) sctp_shutdown_received(sctp, NULL, B_FALSE,
+ B_TRUE, NULL);
}
if (sctp->sctp_xmit_head == NULL &&
@@ -645,7 +648,8 @@ rxmit_init:
ASSERT(sctp->sctp_xmit_unsent == NULL);
BUMP_LOCAL(sctp->sctp_T2expire);
- (void) sctp_shutdown_received(sctp, NULL, 0, 1);
+ (void) sctp_shutdown_received(sctp, NULL, B_FALSE, B_TRUE,
+ NULL);
BUMP_MIB(&sctp_mib, sctpTimRetrans);
break;
default:
generated by cgit v1.2.3 (git 2.39.1) at 2025-04-29 15:14:48 +0000