d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
3a56853369
freebsd-skq/sys/netinet/sctp_cc_functions.c

1544 lines
45 KiB
C
Raw Normal View History

- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*-
* Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* a) Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
*
* b) Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the distribution.
*
* c) Neither the name of Cisco Systems, Inc. nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
* THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
* THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <netinet/sctp_os.h>
#include <netinet/sctp_var.h>
#include <netinet/sctp_sysctl.h>
#include <netinet/sctp_pcb.h>
#include <netinet/sctp_header.h>
#include <netinet/sctputil.h>
#include <netinet/sctp_output.h>
#include <netinet/sctp_input.h>
#include <netinet/sctp_indata.h>
#include <netinet/sctp_uio.h>
#include <netinet/sctp_timer.h>
#include <netinet/sctp_auth.h>
#include <netinet/sctp_asconf.h>
#include <netinet/sctp_cc_functions.h>
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
void
sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
{
/*
* We take the max of the burst limit times a MTU or the
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
* INITIAL_CWND. We then limit this to 4 MTU's of sending. cwnd must
* be at least 2 MTU.
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
*/
net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
net->ssthresh = stcb->asoc.peers_rwnd;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
}
}
void
sctp_cwnd_update_after_fr(struct sctp_tcb *stcb,
struct sctp_association *asoc)
{
struct sctp_nets *net;
/*-
* CMT fast recovery code. Need to debug. ((sctp_cmt_on_off == 1) &&
* (net->fast_retran_loss_recovery == 0)))
*/
TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if ((asoc->fast_retran_loss_recovery == 0) || (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 1)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* out of a RFC2582 Fast recovery window? */
if (net->net_ack > 0) {
/*
* per section 7.2.3, are there any
* destinations that had a fast retransmit
* to them. If so what we need to do is
* adjust ssthresh and cwnd.
*/
struct sctp_tmit_chunk *lchk;
int old_cwnd = net->cwnd;
net->ssthresh = net->cwnd / 2;
if (net->ssthresh < (net->mtu * 2)) {
net->ssthresh = 2 * net->mtu;
}
net->cwnd = net->ssthresh;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd),
SCTP_CWND_LOG_FROM_FR);
}
lchk = TAILQ_FIRST(&asoc->send_queue);
net->partial_bytes_acked = 0;
/* Turn on fast recovery window */
asoc->fast_retran_loss_recovery = 1;
if (lchk == NULL) {
/* Mark end of the window */
asoc->fast_recovery_tsn = asoc->sending_seq - 1;
} else {
asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
}
/*
* CMT fast recovery -- per destination
* recovery variable.
*/
net->fast_retran_loss_recovery = 1;
if (lchk == NULL) {
/* Mark end of the window */
net->fast_recovery_tsn = asoc->sending_seq - 1;
} else {
net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
}
/*
* Disable Nonce Sum Checking and store the
* resync tsn
*/
asoc->nonce_sum_check = 0;
asoc->nonce_resync_tsn = asoc->fast_recovery_tsn + 1;
sctp_timer_stop(SCTP_TIMER_TYPE_SEND,
stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32);
sctp_timer_start(SCTP_TIMER_TYPE_SEND,
stcb->sctp_ep, stcb, net);
}
} else if (net->net_ack > 0) {
/*
* Mark a peg that we WOULD have done a cwnd
* reduction but RFC2582 prevented this action.
*/
SCTP_STAT_INCR(sctps_fastretransinrtt);
}
}
}
void
sctp_cwnd_update_after_sack(struct sctp_tcb *stcb,
struct sctp_association *asoc,
int accum_moved, int reneged_all, int will_exit)
{
struct sctp_nets *net;
/******************************/
/* update cwnd and Early FR */
/******************************/
TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
#ifdef JANA_CMT_FAST_RECOVERY
/*
* CMT fast recovery code. Need to debug.
*/
if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) {
if (compare_with_wrap(asoc->last_acked_seq,
net->fast_recovery_tsn, MAX_TSN) ||
(asoc->last_acked_seq == net->fast_recovery_tsn) ||
compare_with_wrap(net->pseudo_cumack, net->fast_recovery_tsn, MAX_TSN) ||
(net->pseudo_cumack == net->fast_recovery_tsn)) {
net->will_exit_fast_recovery = 1;
}
}
#endif
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_early_fr)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* So, first of all do we need to have a Early FR
* timer running?
*/
if (((TAILQ_FIRST(&asoc->sent_queue)) &&
(net->ref_count > 1) &&
(net->flight_size < net->cwnd)) ||
(reneged_all)) {
/*
* yes, so in this case stop it if its
* running, and then restart it. Reneging
* all is a special case where we want to
* run the Early FR timer and then force the
* last few unacked to be sent, causing us
* to illicit a sack with gaps to force out
* the others.
*/
if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) {
SCTP_STAT_INCR(sctps_earlyfrstpidsck2);
sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net,
SCTP_FROM_SCTP_INDATA + SCTP_LOC_20);
}
SCTP_STAT_INCR(sctps_earlyfrstrid);
sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net);
} else {
/* No, stop it if its running */
if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) {
SCTP_STAT_INCR(sctps_earlyfrstpidsck3);
sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net,
SCTP_FROM_SCTP_INDATA + SCTP_LOC_21);
}
}
}
/* if nothing was acked on this destination skip it */
if (net->net_ack == 0) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK);
}
continue;
}
if (net->net_ack2 > 0) {
/*
* Karn's rule applies to clearing error count, this
* is optional.
*/
net->error_count = 0;
if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) ==
SCTP_ADDR_NOT_REACHABLE) {
/* addr came good */
net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE;
net->dest_state |= SCTP_ADDR_REACHABLE;
sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb,
- Locking compatiability changes. This involves adding additional flags to many function calls. The flags only get used in BSD when we compile with lock testing. These flags allow apple to escape the "giant" lock it holds on the socket and have more fine-grained locking in the NKE. It also allows us to test (with witness) the locking used by apple via a compile switch (manually applied). Approved by: re@freebsd.org(B Mah)
2007-09-08 11:35:11 +00:00
SCTP_RECEIVED_SACK, (void *)net, SCTP_SO_NOT_LOCKED);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* now was it the primary? if so restore */
if (net->dest_state & SCTP_ADDR_WAS_PRIMARY) {
(void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net);
}
}
/*
* JRS 5/14/07 - If CMT PF is on and the destination
* is in PF state, set the destination to active
* state and set the cwnd to one or two MTU's based
* on whether PF1 or PF2 is being used.
*
* Should we stop any running T3 timer here?
*/
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) &&
SCTP_BASE_SYSCTL(sctp_cmt_pf) &&
((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
net->dest_state &= ~SCTP_ADDR_PF;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
net->cwnd = net->mtu * SCTP_BASE_SYSCTL(sctp_cmt_pf);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n",
net, net->cwnd);
/*
* Since the cwnd value is explicitly set,
* skip the code that updates the cwnd
* value.
*/
goto skip_cwnd_update;
}
}
#ifdef JANA_CMT_FAST_RECOVERY
/*
* CMT fast recovery code
*/
/*
* if (sctp_cmt_on_off == 1 &&
* net->fast_retran_loss_recovery &&
* net->will_exit_fast_recovery == 0) { @@@ Do something }
* else if (sctp_cmt_on_off == 0 &&
* asoc->fast_retran_loss_recovery && will_exit == 0) {
*/
#endif
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (asoc->fast_retran_loss_recovery && will_exit == 0 && SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* If we are in loss recovery we skip any cwnd
* update
*/
goto skip_cwnd_update;
}
/*
* CMT: CUC algorithm. Update cwnd if pseudo-cumack has
* moved.
*/
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (accum_moved || (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && net->new_pseudo_cumack)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* If the cumulative ack moved we can proceed */
if (net->cwnd <= net->ssthresh) {
/* We are in slow start */
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
if (net->flight_size + net->net_ack >= net->cwnd) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (net->net_ack > (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable))) {
net->cwnd += (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable));
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_FROM_SS);
}
} else {
net->cwnd += net->net_ack;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_FROM_SS);
}
}
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_NOADV_SS);
}
}
} else {
/* We are in congestion avoidance */
- Define changes in sctp.h - Bug in CA that does not get us incrementing the PBA properly which made us more conservative. - comment updated in sctp_input.c - memsets added before we log - added arg to hmac id's MFC after: 2 weeks
2008-05-20 09:51:36 +00:00
/*
* Add to pba
*/
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
net->partial_bytes_acked += net->net_ack;
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
- Define changes in sctp.h - Bug in CA that does not get us incrementing the PBA properly which made us more conservative. - comment updated in sctp_input.c - memsets added before we log - added arg to hmac id's MFC after: 2 weeks
2008-05-20 09:51:36 +00:00
if ((net->flight_size + net->net_ack >= net->cwnd) &&
(net->partial_bytes_acked >= net->cwnd)) {
net->partial_bytes_acked -= net->cwnd;
net->cwnd += net->mtu;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Define changes in sctp.h - Bug in CA that does not get us incrementing the PBA properly which made us more conservative. - comment updated in sctp_input.c - memsets added before we log - added arg to hmac id's MFC after: 2 weeks
2008-05-20 09:51:36 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_FROM_CA);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
}
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_NOADV_CA);
}
}
}
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_NO_CUMACK);
}
}
skip_cwnd_update:
/*
* NOW, according to Karn's rule do we need to restore the
* RTO timer back? Check our net_ack2. If not set then we
* have a ambiguity.. i.e. all data ack'd was sent to more
* than one place.
*/
if (net->net_ack2) {
/* restore any doubled timers */
net->RTO = ((net->lastsa >> 2) + net->lastsv) >> 1;
if (net->RTO < stcb->asoc.minrto) {
net->RTO = stcb->asoc.minrto;
}
if (net->RTO > stcb->asoc.maxrto) {
net->RTO = stcb->asoc.maxrto;
}
}
}
}
void
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
sctp_cwnd_update_after_timeout(struct sctp_tcb *stcb, struct sctp_nets *net)
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
{
int old_cwnd = net->cwnd;
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
net->ssthresh = max(net->cwnd / 2, 2 * net->mtu);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
net->cwnd = net->mtu;
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
net->partial_bytes_acked = 0;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX);
}
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
}
void
sctp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb, struct sctp_nets *net)
{
int old_cwnd = net->cwnd;
SCTP_STAT_INCR(sctps_ecnereducedcwnd);
net->ssthresh = net->cwnd / 2;
if (net->ssthresh < net->mtu) {
net->ssthresh = net->mtu;
/* here back off the timer as well, to slow us down */
net->RTO <<= 1;
}
net->cwnd = net->ssthresh;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT);
}
}
void
sctp_cwnd_update_after_packet_dropped(struct sctp_tcb *stcb,
struct sctp_nets *net, struct sctp_pktdrop_chunk *cp,
uint32_t * bottle_bw, uint32_t * on_queue)
{
uint32_t bw_avail;
int rtt, incr;
int old_cwnd = net->cwnd;
/* need real RTT for this calc */
rtt = ((net->lastsa >> 2) + net->lastsv) >> 1;
/* get bottle neck bw */
*bottle_bw = ntohl(cp->bottle_bw);
/* and whats on queue */
*on_queue = ntohl(cp->current_onq);
/*
* adjust the on-queue if our flight is more it could be that the
* router has not yet gotten data "in-flight" to it
*/
if (*on_queue < net->flight_size)
*on_queue = net->flight_size;
/* calculate the available space */
bw_avail = (*bottle_bw * rtt) / 1000;
if (bw_avail > *bottle_bw) {
/*
* Cap the growth to no more than the bottle neck. This can
* happen as RTT slides up due to queues. It also means if
* you have more than a 1 second RTT with a empty queue you
* will be limited to the bottle_bw per second no matter if
* other points have 1/2 the RTT and you could get more
* out...
*/
bw_avail = *bottle_bw;
}
if (*on_queue > bw_avail) {
/*
* No room for anything else don't allow anything else to be
* "added to the fire".
*/
int seg_inflight, seg_onqueue, my_portion;
net->partial_bytes_acked = 0;
/* how much are we over queue size? */
incr = *on_queue - bw_avail;
if (stcb->asoc.seen_a_sack_this_pkt) {
/*
* undo any cwnd adjustment that the sack might have
* made
*/
net->cwnd = net->prev_cwnd;
}
/* Now how much of that is mine? */
seg_inflight = net->flight_size / net->mtu;
seg_onqueue = *on_queue / net->mtu;
my_portion = (incr * seg_inflight) / seg_onqueue;
/* Have I made an adjustment already */
if (net->cwnd > net->flight_size) {
/*
* for this flight I made an adjustment we need to
* decrease the portion by a share our previous
* adjustment.
*/
int diff_adj;
diff_adj = net->cwnd - net->flight_size;
if (diff_adj > my_portion)
my_portion = 0;
else
my_portion -= diff_adj;
}
/*
* back down to the previous cwnd (assume we have had a sack
* before this packet). minus what ever portion of the
* overage is my fault.
*/
net->cwnd -= my_portion;
/* we will NOT back down more than 1 MTU */
if (net->cwnd <= net->mtu) {
net->cwnd = net->mtu;
}
/* force into CA */
net->ssthresh = net->cwnd - 1;
} else {
/*
* Take 1/4 of the space left or max burst up .. whichever
* is less.
*/
incr = min((bw_avail - *on_queue) >> 2,
stcb->asoc.max_burst * net->mtu);
net->cwnd += incr;
}
if (net->cwnd > bw_avail) {
/* We can't exceed the pipe size */
net->cwnd = bw_avail;
}
if (net->cwnd < net->mtu) {
/* We always have 1 MTU */
net->cwnd = net->mtu;
}
if (net->cwnd - old_cwnd != 0) {
/* log only changes */
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd),
SCTP_CWND_LOG_FROM_SAT);
}
}
}
void
sctp_cwnd_update_after_output(struct sctp_tcb *stcb,
struct sctp_nets *net, int burst_limit)
{
int old_cwnd = net->cwnd;
if (net->ssthresh < net->cwnd)
net->ssthresh = net->cwnd;
net->cwnd = (net->flight_size + (burst_limit * net->mtu));
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_BRST);
}
}
void
sctp_cwnd_update_after_fr_timer(struct sctp_inpcb *inp,
struct sctp_tcb *stcb, struct sctp_nets *net)
{
int old_cwnd = net->cwnd;
sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED);
/*
* make a small adjustment to cwnd and force to CA.
*/
if (net->cwnd > net->mtu)
/* drop down one MTU after sending */
net->cwnd -= net->mtu;
if (net->cwnd < net->ssthresh)
/* still in SS move to CA */
net->ssthresh = net->cwnd - 1;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
sctp_log_cwnd(stcb, net, (old_cwnd - net->cwnd), SCTP_CWND_LOG_FROM_FR);
}
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
}
struct sctp_hs_raise_drop {
int32_t cwnd;
int32_t increase;
int32_t drop_percent;
};
#define SCTP_HS_TABLE_SIZE 73
struct sctp_hs_raise_drop sctp_cwnd_adjust[SCTP_HS_TABLE_SIZE] = {
{38, 1, 50}, /* 0 */
{118, 2, 44}, /* 1 */
{221, 3, 41}, /* 2 */
{347, 4, 38}, /* 3 */
{495, 5, 37}, /* 4 */
{663, 6, 35}, /* 5 */
{851, 7, 34}, /* 6 */
{1058, 8, 33}, /* 7 */
{1284, 9, 32}, /* 8 */
{1529, 10, 31}, /* 9 */
{1793, 11, 30}, /* 10 */
{2076, 12, 29}, /* 11 */
{2378, 13, 28}, /* 12 */
{2699, 14, 28}, /* 13 */
{3039, 15, 27}, /* 14 */
{3399, 16, 27}, /* 15 */
{3778, 17, 26}, /* 16 */
{4177, 18, 26}, /* 17 */
{4596, 19, 25}, /* 18 */
{5036, 20, 25}, /* 19 */
{5497, 21, 24}, /* 20 */
{5979, 22, 24}, /* 21 */
{6483, 23, 23}, /* 22 */
{7009, 24, 23}, /* 23 */
{7558, 25, 22}, /* 24 */
{8130, 26, 22}, /* 25 */
{8726, 27, 22}, /* 26 */
{9346, 28, 21}, /* 27 */
{9991, 29, 21}, /* 28 */
{10661, 30, 21}, /* 29 */
{11358, 31, 20}, /* 30 */
{12082, 32, 20}, /* 31 */
{12834, 33, 20}, /* 32 */
{13614, 34, 19}, /* 33 */
{14424, 35, 19}, /* 34 */
{15265, 36, 19}, /* 35 */
{16137, 37, 19}, /* 36 */
{17042, 38, 18}, /* 37 */
{17981, 39, 18}, /* 38 */
{18955, 40, 18}, /* 39 */
{19965, 41, 17}, /* 40 */
{21013, 42, 17}, /* 41 */
{22101, 43, 17}, /* 42 */
{23230, 44, 17}, /* 43 */
{24402, 45, 16}, /* 44 */
{25618, 46, 16}, /* 45 */
{26881, 47, 16}, /* 46 */
{28193, 48, 16}, /* 47 */
{29557, 49, 15}, /* 48 */
{30975, 50, 15}, /* 49 */
{32450, 51, 15}, /* 50 */
{33986, 52, 15}, /* 51 */
{35586, 53, 14}, /* 52 */
{37253, 54, 14}, /* 53 */
{38992, 55, 14}, /* 54 */
{40808, 56, 14}, /* 55 */
{42707, 57, 13}, /* 56 */
{44694, 58, 13}, /* 57 */
{46776, 59, 13}, /* 58 */
{48961, 60, 13}, /* 59 */
{51258, 61, 13}, /* 60 */
{53677, 62, 12}, /* 61 */
{56230, 63, 12}, /* 62 */
{58932, 64, 12}, /* 63 */
{61799, 65, 12}, /* 64 */
{64851, 66, 11}, /* 65 */
{68113, 67, 11}, /* 66 */
{71617, 68, 11}, /* 67 */
{75401, 69, 10}, /* 68 */
{79517, 70, 10}, /* 69 */
{84035, 71, 10}, /* 70 */
{89053, 72, 10}, /* 71 */
{94717, 73, 9} /* 72 */
};
static void
sctp_hs_cwnd_increase(struct sctp_tcb *stcb, struct sctp_nets *net)
{
int cur_val, i, indx, incr;
cur_val = net->cwnd >> 10;
indx = SCTP_HS_TABLE_SIZE - 1;
#ifdef SCTP_DEBUG
printf("HS CC CAlled.\n");
#endif
if (cur_val < sctp_cwnd_adjust[0].cwnd) {
/* normal mode */
if (net->net_ack > net->mtu) {
net->cwnd += net->mtu;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu, SCTP_CWND_LOG_FROM_SS);
}
} else {
net->cwnd += net->net_ack;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack, SCTP_CWND_LOG_FROM_SS);
}
}
} else {
for (i = net->last_hs_used; i < SCTP_HS_TABLE_SIZE; i++) {
if (cur_val < sctp_cwnd_adjust[i].cwnd) {
indx = i;
break;
}
}
net->last_hs_used = indx;
incr = ((sctp_cwnd_adjust[indx].increase) << 10);
net->cwnd += incr;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, incr, SCTP_CWND_LOG_FROM_SS);
}
}
}
static void
sctp_hs_cwnd_decrease(struct sctp_tcb *stcb, struct sctp_nets *net)
{
int cur_val, i, indx;
int old_cwnd = net->cwnd;
cur_val = net->cwnd >> 10;
indx = net->last_hs_used;
if (cur_val < sctp_cwnd_adjust[0].cwnd) {
/* normal mode */
net->ssthresh = net->cwnd / 2;
if (net->ssthresh < (net->mtu * 2)) {
net->ssthresh = 2 * net->mtu;
}
net->cwnd = net->ssthresh;
} else {
/* drop by the proper amount */
net->ssthresh = net->cwnd - (int)((net->cwnd / 100) *
sctp_cwnd_adjust[net->last_hs_used].drop_percent);
net->cwnd = net->ssthresh;
/* now where are we */
indx = net->last_hs_used;
cur_val = net->cwnd >> 10;
/* reset where we are in the table */
if (cur_val < sctp_cwnd_adjust[0].cwnd) {
/* feel out of hs */
net->last_hs_used = 0;
} else {
for (i = indx; i >= 1; i--) {
if (cur_val > sctp_cwnd_adjust[i - 1].cwnd) {
break;
}
}
net->last_hs_used = indx;
}
}
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_FR);
}
}
void
sctp_hs_cwnd_update_after_fr(struct sctp_tcb *stcb,
struct sctp_association *asoc)
{
struct sctp_nets *net;
/*
* CMT fast recovery code. Need to debug. ((sctp_cmt_on_off == 1) &&
* (net->fast_retran_loss_recovery == 0)))
*/
TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if ((asoc->fast_retran_loss_recovery == 0) || (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 1)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* out of a RFC2582 Fast recovery window? */
if (net->net_ack > 0) {
/*
* per section 7.2.3, are there any
* destinations that had a fast retransmit
* to them. If so what we need to do is
* adjust ssthresh and cwnd.
*/
struct sctp_tmit_chunk *lchk;
sctp_hs_cwnd_decrease(stcb, net);
lchk = TAILQ_FIRST(&asoc->send_queue);
net->partial_bytes_acked = 0;
/* Turn on fast recovery window */
asoc->fast_retran_loss_recovery = 1;
if (lchk == NULL) {
/* Mark end of the window */
asoc->fast_recovery_tsn = asoc->sending_seq - 1;
} else {
asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
}
/*
* CMT fast recovery -- per destination
* recovery variable.
*/
net->fast_retran_loss_recovery = 1;
if (lchk == NULL) {
/* Mark end of the window */
net->fast_recovery_tsn = asoc->sending_seq - 1;
} else {
net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
}
/*
* Disable Nonce Sum Checking and store the
* resync tsn
*/
asoc->nonce_sum_check = 0;
asoc->nonce_resync_tsn = asoc->fast_recovery_tsn + 1;
sctp_timer_stop(SCTP_TIMER_TYPE_SEND,
stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32);
sctp_timer_start(SCTP_TIMER_TYPE_SEND,
stcb->sctp_ep, stcb, net);
}
} else if (net->net_ack > 0) {
/*
* Mark a peg that we WOULD have done a cwnd
* reduction but RFC2582 prevented this action.
*/
SCTP_STAT_INCR(sctps_fastretransinrtt);
}
}
}
void
sctp_hs_cwnd_update_after_sack(struct sctp_tcb *stcb,
struct sctp_association *asoc,
int accum_moved, int reneged_all, int will_exit)
{
struct sctp_nets *net;
/******************************/
/* update cwnd and Early FR */
/******************************/
TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
#ifdef JANA_CMT_FAST_RECOVERY
/*
* CMT fast recovery code. Need to debug.
*/
if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) {
if (compare_with_wrap(asoc->last_acked_seq,
net->fast_recovery_tsn, MAX_TSN) ||
(asoc->last_acked_seq == net->fast_recovery_tsn) ||
compare_with_wrap(net->pseudo_cumack, net->fast_recovery_tsn, MAX_TSN) ||
(net->pseudo_cumack == net->fast_recovery_tsn)) {
net->will_exit_fast_recovery = 1;
}
}
#endif
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_early_fr)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* So, first of all do we need to have a Early FR
* timer running?
*/
if (((TAILQ_FIRST(&asoc->sent_queue)) &&
(net->ref_count > 1) &&
(net->flight_size < net->cwnd)) ||
(reneged_all)) {
/*
* yes, so in this case stop it if its
* running, and then restart it. Reneging
* all is a special case where we want to
* run the Early FR timer and then force the
* last few unacked to be sent, causing us
* to illicit a sack with gaps to force out
* the others.
*/
if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) {
SCTP_STAT_INCR(sctps_earlyfrstpidsck2);
sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net,
SCTP_FROM_SCTP_INDATA + SCTP_LOC_20);
}
SCTP_STAT_INCR(sctps_earlyfrstrid);
sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net);
} else {
/* No, stop it if its running */
if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) {
SCTP_STAT_INCR(sctps_earlyfrstpidsck3);
sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net,
SCTP_FROM_SCTP_INDATA + SCTP_LOC_21);
}
}
}
/* if nothing was acked on this destination skip it */
if (net->net_ack == 0) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK);
}
continue;
}
if (net->net_ack2 > 0) {
/*
* Karn's rule applies to clearing error count, this
* is optional.
*/
net->error_count = 0;
if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) ==
SCTP_ADDR_NOT_REACHABLE) {
/* addr came good */
net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE;
net->dest_state |= SCTP_ADDR_REACHABLE;
sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb,
- Locking compatiability changes. This involves adding additional flags to many function calls. The flags only get used in BSD when we compile with lock testing. These flags allow apple to escape the "giant" lock it holds on the socket and have more fine-grained locking in the NKE. It also allows us to test (with witness) the locking used by apple via a compile switch (manually applied). Approved by: re@freebsd.org(B Mah)
2007-09-08 11:35:11 +00:00
SCTP_RECEIVED_SACK, (void *)net, SCTP_SO_NOT_LOCKED);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* now was it the primary? if so restore */
if (net->dest_state & SCTP_ADDR_WAS_PRIMARY) {
(void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net);
}
}
/*
* JRS 5/14/07 - If CMT PF is on and the destination
* is in PF state, set the destination to active
* state and set the cwnd to one or two MTU's based
* on whether PF1 or PF2 is being used.
*
* Should we stop any running T3 timer here?
*/
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) &&
SCTP_BASE_SYSCTL(sctp_cmt_pf) &&
((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
net->dest_state &= ~SCTP_ADDR_PF;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
net->cwnd = net->mtu * SCTP_BASE_SYSCTL(sctp_cmt_pf);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n",
net, net->cwnd);
/*
* Since the cwnd value is explicitly set,
* skip the code that updates the cwnd
* value.
*/
goto skip_cwnd_update;
}
}
#ifdef JANA_CMT_FAST_RECOVERY
/*
* CMT fast recovery code
*/
/*
* if (sctp_cmt_on_off == 1 &&
* net->fast_retran_loss_recovery &&
* net->will_exit_fast_recovery == 0) { @@@ Do something }
* else if (sctp_cmt_on_off == 0 &&
* asoc->fast_retran_loss_recovery && will_exit == 0) {
*/
#endif
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (asoc->fast_retran_loss_recovery && will_exit == 0 && SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* If we are in loss recovery we skip any cwnd
* update
*/
goto skip_cwnd_update;
}
/*
* CMT: CUC algorithm. Update cwnd if pseudo-cumack has
* moved.
*/
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (accum_moved || (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && net->new_pseudo_cumack)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* If the cumulative ack moved we can proceed */
if (net->cwnd <= net->ssthresh) {
/* We are in slow start */
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
if (net->flight_size + net->net_ack >= net->cwnd) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_hs_cwnd_increase(stcb, net);
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_NOADV_SS);
}
}
} else {
/* We are in congestion avoidance */
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
net->partial_bytes_acked += net->net_ack;
if ((net->flight_size + net->net_ack >= net->cwnd) &&
(net->partial_bytes_acked >= net->cwnd)) {
net->partial_bytes_acked -= net->cwnd;
net->cwnd += net->mtu;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_FROM_CA);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
}
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_NOADV_CA);
}
}
}
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_NO_CUMACK);
}
}
skip_cwnd_update:
/*
* NOW, according to Karn's rule do we need to restore the
* RTO timer back? Check our net_ack2. If not set then we
* have a ambiguity.. i.e. all data ack'd was sent to more
* than one place.
*/
if (net->net_ack2) {
/* restore any doubled timers */
net->RTO = ((net->lastsa >> 2) + net->lastsv) >> 1;
if (net->RTO < stcb->asoc.minrto) {
net->RTO = stcb->asoc.minrto;
}
if (net->RTO > stcb->asoc.maxrto) {
net->RTO = stcb->asoc.maxrto;
}
}
}
}
/*
* H-TCP congestion control. The algorithm is detailed in:
* R.N.Shorten, D.J.Leith:
* "H-TCP: TCP for high-speed and long-distance networks"
* Proc. PFLDnet, Argonne, 2004.
* http://www.hamilton.ie/net/htcp3.pdf
*/
static int use_rtt_scaling = 1;
static int use_bandwidth_switch = 1;
static inline int
between(uint32_t seq1, uint32_t seq2, uint32_t seq3)
{
return seq3 - seq2 >= seq1 - seq2;
}
static inline uint32_t
htcp_cong_time(struct htcp *ca)
{
- added pre-checks to the bindx call. - use proper tick gathering macro instead of ticks directly. - Placed reasonable boundaries on sets that a user can do that are converted to ticks from ms. - Fix CMT_PF to always check to be sure CMT is on. - Fix ticks use of CMT_PF. - put back code to allow asconfs to be queued while INITs are in flight and before the assoc is established. - During window probes, an ack'd packet might be left with the window probe mark on it causing it to be retransmitted. Change so that the flight decrease macro clears the window_probe mark. - Additional logging flight size/reading and ASOC LOG. This is only enabled if you manually insert things into opt_sctp.h since its a set of debug code only. - Found an interesting SMP race in the way data was appended which could cause a reader to lose a part of a message, had to reorder when we marked the message was complete to after the data was appended. - bug in ADD-IP for the subset bound socket case when the peer has only one address - fix ASCONF implicit success/error handling case - proper support of jails in Freebsd 6> - copy out the timeval for the 64 bit sparc world on cookie-echo alignment error crashes without this). Approved by: re(Ken Smith)
2007-07-17 20:58:26 +00:00
return sctp_get_tick_count() - ca->last_cong;
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
}
static inline uint32_t
htcp_ccount(struct htcp *ca)
{
return htcp_cong_time(ca) / ca->minRTT;
}
static inline void
htcp_reset(struct htcp *ca)
{
ca->undo_last_cong = ca->last_cong;
ca->undo_maxRTT = ca->maxRTT;
ca->undo_old_maxB = ca->old_maxB;
- added pre-checks to the bindx call. - use proper tick gathering macro instead of ticks directly. - Placed reasonable boundaries on sets that a user can do that are converted to ticks from ms. - Fix CMT_PF to always check to be sure CMT is on. - Fix ticks use of CMT_PF. - put back code to allow asconfs to be queued while INITs are in flight and before the assoc is established. - During window probes, an ack'd packet might be left with the window probe mark on it causing it to be retransmitted. Change so that the flight decrease macro clears the window_probe mark. - Additional logging flight size/reading and ASOC LOG. This is only enabled if you manually insert things into opt_sctp.h since its a set of debug code only. - Found an interesting SMP race in the way data was appended which could cause a reader to lose a part of a message, had to reorder when we marked the message was complete to after the data was appended. - bug in ADD-IP for the subset bound socket case when the peer has only one address - fix ASCONF implicit success/error handling case - proper support of jails in Freebsd 6> - copy out the timeval for the 64 bit sparc world on cookie-echo alignment error crashes without this). Approved by: re(Ken Smith)
2007-07-17 20:58:26 +00:00
ca->last_cong = sctp_get_tick_count();
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
}
#ifdef SCTP_NOT_USED
static uint32_t
htcp_cwnd_undo(struct sctp_tcb *stcb, struct sctp_nets *net)
{
net->htcp_ca.last_cong = net->htcp_ca.undo_last_cong;
net->htcp_ca.maxRTT = net->htcp_ca.undo_maxRTT;
net->htcp_ca.old_maxB = net->htcp_ca.undo_old_maxB;
return max(net->cwnd, ((net->ssthresh / net->mtu << 7) / net->htcp_ca.beta) * net->mtu);
}
#endif
static inline void
measure_rtt(struct sctp_tcb *stcb, struct sctp_nets *net)
{
uint32_t srtt = net->lastsa >> 3;
/* keep track of minimum RTT seen so far, minRTT is zero at first */
if (net->htcp_ca.minRTT > srtt || !net->htcp_ca.minRTT)
net->htcp_ca.minRTT = srtt;
/* max RTT */
if (net->fast_retran_ip == 0 && net->ssthresh < 0xFFFF && htcp_ccount(&net->htcp_ca) > 3) {
if (net->htcp_ca.maxRTT < net->htcp_ca.minRTT)
net->htcp_ca.maxRTT = net->htcp_ca.minRTT;
if (net->htcp_ca.maxRTT < srtt && srtt <= net->htcp_ca.maxRTT + MSEC_TO_TICKS(20))
net->htcp_ca.maxRTT = srtt;
}
}
static void
measure_achieved_throughput(struct sctp_tcb *stcb, struct sctp_nets *net)
{
- added pre-checks to the bindx call. - use proper tick gathering macro instead of ticks directly. - Placed reasonable boundaries on sets that a user can do that are converted to ticks from ms. - Fix CMT_PF to always check to be sure CMT is on. - Fix ticks use of CMT_PF. - put back code to allow asconfs to be queued while INITs are in flight and before the assoc is established. - During window probes, an ack'd packet might be left with the window probe mark on it causing it to be retransmitted. Change so that the flight decrease macro clears the window_probe mark. - Additional logging flight size/reading and ASOC LOG. This is only enabled if you manually insert things into opt_sctp.h since its a set of debug code only. - Found an interesting SMP race in the way data was appended which could cause a reader to lose a part of a message, had to reorder when we marked the message was complete to after the data was appended. - bug in ADD-IP for the subset bound socket case when the peer has only one address - fix ASCONF implicit success/error handling case - proper support of jails in Freebsd 6> - copy out the timeval for the 64 bit sparc world on cookie-echo alignment error crashes without this). Approved by: re(Ken Smith)
2007-07-17 20:58:26 +00:00
uint32_t now = sctp_get_tick_count();
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
if (net->fast_retran_ip == 0)
net->htcp_ca.bytes_acked = net->net_ack;
if (!use_bandwidth_switch)
return;
/* achieved throughput calculations */
/* JRS - not 100% sure of this statement */
if (net->fast_retran_ip == 1) {
net->htcp_ca.bytecount = 0;
net->htcp_ca.lasttime = now;
return;
}
net->htcp_ca.bytecount += net->net_ack;
if (net->htcp_ca.bytecount >= net->cwnd - ((net->htcp_ca.alpha >> 7 ? : 1) * net->mtu)
&& now - net->htcp_ca.lasttime >= net->htcp_ca.minRTT
&& net->htcp_ca.minRTT > 0) {
uint32_t cur_Bi = net->htcp_ca.bytecount / net->mtu * hz / (now - net->htcp_ca.lasttime);
if (htcp_ccount(&net->htcp_ca) <= 3) {
/* just after backoff */
net->htcp_ca.minB = net->htcp_ca.maxB = net->htcp_ca.Bi = cur_Bi;
} else {
net->htcp_ca.Bi = (3 * net->htcp_ca.Bi + cur_Bi) / 4;
if (net->htcp_ca.Bi > net->htcp_ca.maxB)
net->htcp_ca.maxB = net->htcp_ca.Bi;
if (net->htcp_ca.minB > net->htcp_ca.maxB)
net->htcp_ca.minB = net->htcp_ca.maxB;
}
net->htcp_ca.bytecount = 0;
net->htcp_ca.lasttime = now;
}
}
static inline void
htcp_beta_update(struct htcp *ca, uint32_t minRTT, uint32_t maxRTT)
{
if (use_bandwidth_switch) {
uint32_t maxB = ca->maxB;
uint32_t old_maxB = ca->old_maxB;
ca->old_maxB = ca->maxB;
if (!between(5 * maxB, 4 * old_maxB, 6 * old_maxB)) {
ca->beta = BETA_MIN;
ca->modeswitch = 0;
return;
}
}
if (ca->modeswitch && minRTT > (uint32_t) MSEC_TO_TICKS(10) && maxRTT) {
ca->beta = (minRTT << 7) / maxRTT;
if (ca->beta < BETA_MIN)
ca->beta = BETA_MIN;
else if (ca->beta > BETA_MAX)
ca->beta = BETA_MAX;
} else {
ca->beta = BETA_MIN;
ca->modeswitch = 1;
}
}
static inline void
htcp_alpha_update(struct htcp *ca)
{
uint32_t minRTT = ca->minRTT;
uint32_t factor = 1;
uint32_t diff = htcp_cong_time(ca);
if (diff > (uint32_t) hz) {
diff -= hz;
factor = 1 + (10 * diff + ((diff / 2) * (diff / 2) / hz)) / hz;
}
if (use_rtt_scaling && minRTT) {
uint32_t scale = (hz << 3) / (10 * minRTT);
scale = min(max(scale, 1U << 2), 10U << 3); /* clamping ratio to
* interval [0.5,10]<<3 */
factor = (factor << 3) / scale;
if (!factor)
factor = 1;
}
ca->alpha = 2 * factor * ((1 << 7) - ca->beta);
if (!ca->alpha)
ca->alpha = ALPHA_BASE;
}
/* After we have the rtt data to calculate beta, we'd still prefer to wait one
* rtt before we adjust our beta to ensure we are working from a consistent
* data.
*
* This function should be called when we hit a congestion event since only at
* that point do we really have a real sense of maxRTT (the queues en route
* were getting just too full now).
*/
static void
htcp_param_update(struct sctp_tcb *stcb, struct sctp_nets *net)
{
uint32_t minRTT = net->htcp_ca.minRTT;
uint32_t maxRTT = net->htcp_ca.maxRTT;
htcp_beta_update(&net->htcp_ca, minRTT, maxRTT);
htcp_alpha_update(&net->htcp_ca);
/*
* add slowly fading memory for maxRTT to accommodate routing
* changes etc
*/
if (minRTT > 0 && maxRTT > minRTT)
net->htcp_ca.maxRTT = minRTT + ((maxRTT - minRTT) * 95) / 100;
}
static uint32_t
htcp_recalc_ssthresh(struct sctp_tcb *stcb, struct sctp_nets *net)
{
htcp_param_update(stcb, net);
return max(((net->cwnd / net->mtu * net->htcp_ca.beta) >> 7) * net->mtu, 2U * net->mtu);
}
static void
htcp_cong_avoid(struct sctp_tcb *stcb, struct sctp_nets *net)
{
/*-
* How to handle these functions?
* if (!tcp_is_cwnd_limited(sk, in_flight)) RRS - good question.
* return;
*/
if (net->cwnd <= net->ssthresh) {
/* We are in slow start */
if (net->flight_size + net->net_ack >= net->cwnd) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (net->net_ack > (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable))) {
net->cwnd += (net->mtu * SCTP_BASE_SYSCTL(sctp_L2_abc_variable));
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_FROM_SS);
}
} else {
net->cwnd += net->net_ack;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_FROM_SS);
}
}
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_NOADV_SS);
}
}
} else {
measure_rtt(stcb, net);
/*
* In dangerous area, increase slowly. In theory this is
* net->cwnd += alpha / net->cwnd
*/
/* What is snd_cwnd_cnt?? */
if (((net->partial_bytes_acked / net->mtu * net->htcp_ca.alpha) >> 7) * net->mtu >= net->cwnd) {
/*-
* Does SCTP have a cwnd clamp?
* if (net->snd_cwnd < net->snd_cwnd_clamp) - Nope (RRS).
*/
net->cwnd += net->mtu;
net->partial_bytes_acked = 0;
htcp_alpha_update(&net->htcp_ca);
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_FROM_CA);
}
} else {
net->partial_bytes_acked += net->net_ack;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->net_ack,
SCTP_CWND_LOG_NOADV_CA);
}
}
net->htcp_ca.bytes_acked = net->mtu;
}
}
#ifdef SCTP_NOT_USED
/* Lower bound on congestion window. */
static uint32_t
htcp_min_cwnd(struct sctp_tcb *stcb, struct sctp_nets *net)
{
return net->ssthresh;
}
#endif
static void
htcp_init(struct sctp_tcb *stcb, struct sctp_nets *net)
{
memset(&net->htcp_ca, 0, sizeof(struct htcp));
net->htcp_ca.alpha = ALPHA_BASE;
net->htcp_ca.beta = BETA_MIN;
net->htcp_ca.bytes_acked = net->mtu;
- added pre-checks to the bindx call. - use proper tick gathering macro instead of ticks directly. - Placed reasonable boundaries on sets that a user can do that are converted to ticks from ms. - Fix CMT_PF to always check to be sure CMT is on. - Fix ticks use of CMT_PF. - put back code to allow asconfs to be queued while INITs are in flight and before the assoc is established. - During window probes, an ack'd packet might be left with the window probe mark on it causing it to be retransmitted. Change so that the flight decrease macro clears the window_probe mark. - Additional logging flight size/reading and ASOC LOG. This is only enabled if you manually insert things into opt_sctp.h since its a set of debug code only. - Found an interesting SMP race in the way data was appended which could cause a reader to lose a part of a message, had to reorder when we marked the message was complete to after the data was appended. - bug in ADD-IP for the subset bound socket case when the peer has only one address - fix ASCONF implicit success/error handling case - proper support of jails in Freebsd 6> - copy out the timeval for the 64 bit sparc world on cookie-echo alignment error crashes without this). Approved by: re(Ken Smith)
2007-07-17 20:58:26 +00:00
net->htcp_ca.last_cong = sctp_get_tick_count();
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
}
void
sctp_htcp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
{
/*
* We take the max of the burst limit times a MTU or the
* INITIAL_CWND. We then limit this to 4 MTU's of sending.
*/
net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
net->ssthresh = stcb->asoc.peers_rwnd;
htcp_init(stcb, net);
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & (SCTP_CWND_MONITOR_ENABLE | SCTP_CWND_LOGGING_ENABLE)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
}
}
void
sctp_htcp_cwnd_update_after_sack(struct sctp_tcb *stcb,
struct sctp_association *asoc,
int accum_moved, int reneged_all, int will_exit)
{
struct sctp_nets *net;
/******************************/
/* update cwnd and Early FR */
/******************************/
TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
#ifdef JANA_CMT_FAST_RECOVERY
/*
* CMT fast recovery code. Need to debug.
*/
if (net->fast_retran_loss_recovery && net->new_pseudo_cumack) {
if (compare_with_wrap(asoc->last_acked_seq,
net->fast_recovery_tsn, MAX_TSN) ||
(asoc->last_acked_seq == net->fast_recovery_tsn) ||
compare_with_wrap(net->pseudo_cumack, net->fast_recovery_tsn, MAX_TSN) ||
(net->pseudo_cumack == net->fast_recovery_tsn)) {
net->will_exit_fast_recovery = 1;
}
}
#endif
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_early_fr)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* So, first of all do we need to have a Early FR
* timer running?
*/
if (((TAILQ_FIRST(&asoc->sent_queue)) &&
(net->ref_count > 1) &&
(net->flight_size < net->cwnd)) ||
(reneged_all)) {
/*
* yes, so in this case stop it if its
* running, and then restart it. Reneging
* all is a special case where we want to
* run the Early FR timer and then force the
* last few unacked to be sent, causing us
* to illicit a sack with gaps to force out
* the others.
*/
if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) {
SCTP_STAT_INCR(sctps_earlyfrstpidsck2);
sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net,
SCTP_FROM_SCTP_INDATA + SCTP_LOC_20);
}
SCTP_STAT_INCR(sctps_earlyfrstrid);
sctp_timer_start(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net);
} else {
/* No, stop it if its running */
if (SCTP_OS_TIMER_PENDING(&net->fr_timer.timer)) {
SCTP_STAT_INCR(sctps_earlyfrstpidsck3);
sctp_timer_stop(SCTP_TIMER_TYPE_EARLYFR, stcb->sctp_ep, stcb, net,
SCTP_FROM_SCTP_INDATA + SCTP_LOC_21);
}
}
}
/* if nothing was acked on this destination skip it */
if (net->net_ack == 0) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, 0, SCTP_CWND_LOG_FROM_SACK);
}
continue;
}
if (net->net_ack2 > 0) {
/*
* Karn's rule applies to clearing error count, this
* is optional.
*/
net->error_count = 0;
if ((net->dest_state & SCTP_ADDR_NOT_REACHABLE) ==
SCTP_ADDR_NOT_REACHABLE) {
/* addr came good */
net->dest_state &= ~SCTP_ADDR_NOT_REACHABLE;
net->dest_state |= SCTP_ADDR_REACHABLE;
sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_UP, stcb,
- Locking compatiability changes. This involves adding additional flags to many function calls. The flags only get used in BSD when we compile with lock testing. These flags allow apple to escape the "giant" lock it holds on the socket and have more fine-grained locking in the NKE. It also allows us to test (with witness) the locking used by apple via a compile switch (manually applied). Approved by: re@freebsd.org(B Mah)
2007-09-08 11:35:11 +00:00
SCTP_RECEIVED_SACK, (void *)net, SCTP_SO_NOT_LOCKED);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* now was it the primary? if so restore */
if (net->dest_state & SCTP_ADDR_WAS_PRIMARY) {
(void)sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net);
}
}
/*
* JRS 5/14/07 - If CMT PF is on and the destination
* is in PF state, set the destination to active
* state and set the cwnd to one or two MTU's based
* on whether PF1 or PF2 is being used.
*
* Should we stop any running T3 timer here?
*/
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_cmt_on_off) &&
SCTP_BASE_SYSCTL(sctp_cmt_pf) &&
((net->dest_state & SCTP_ADDR_PF) == SCTP_ADDR_PF)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
net->dest_state &= ~SCTP_ADDR_PF;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
net->cwnd = net->mtu * SCTP_BASE_SYSCTL(sctp_cmt_pf);
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
SCTPDBG(SCTP_DEBUG_INDATA1, "Destination %p moved from PF to reachable with cwnd %d.\n",
net, net->cwnd);
/*
* Since the cwnd value is explicitly set,
* skip the code that updates the cwnd
* value.
*/
goto skip_cwnd_update;
}
}
#ifdef JANA_CMT_FAST_RECOVERY
/*
* CMT fast recovery code
*/
/*
* if (sctp_cmt_on_off == 1 &&
* net->fast_retran_loss_recovery &&
* net->will_exit_fast_recovery == 0) { @@@ Do something }
* else if (sctp_cmt_on_off == 0 &&
* asoc->fast_retran_loss_recovery && will_exit == 0) {
*/
#endif
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (asoc->fast_retran_loss_recovery && will_exit == 0 && SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 0) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* If we are in loss recovery we skip any cwnd
* update
*/
goto skip_cwnd_update;
}
/*
* CMT: CUC algorithm. Update cwnd if pseudo-cumack has
* moved.
*/
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (accum_moved || (SCTP_BASE_SYSCTL(sctp_cmt_on_off) && net->new_pseudo_cumack)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
htcp_cong_avoid(stcb, net);
measure_achieved_throughput(stcb, net);
} else {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_LOGGING_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->mtu,
SCTP_CWND_LOG_NO_CUMACK);
}
}
skip_cwnd_update:
/*
* NOW, according to Karn's rule do we need to restore the
* RTO timer back? Check our net_ack2. If not set then we
* have a ambiguity.. i.e. all data ack'd was sent to more
* than one place.
*/
if (net->net_ack2) {
/* restore any doubled timers */
net->RTO = ((net->lastsa >> 2) + net->lastsv) >> 1;
if (net->RTO < stcb->asoc.minrto) {
net->RTO = stcb->asoc.minrto;
}
if (net->RTO > stcb->asoc.maxrto) {
net->RTO = stcb->asoc.maxrto;
}
}
}
}
void
sctp_htcp_cwnd_update_after_fr(struct sctp_tcb *stcb,
struct sctp_association *asoc)
{
struct sctp_nets *net;
/*
* CMT fast recovery code. Need to debug. ((sctp_cmt_on_off == 1) &&
* (net->fast_retran_loss_recovery == 0)))
*/
TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if ((asoc->fast_retran_loss_recovery == 0) || (SCTP_BASE_SYSCTL(sctp_cmt_on_off) == 1)) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/* out of a RFC2582 Fast recovery window? */
if (net->net_ack > 0) {
/*
* per section 7.2.3, are there any
* destinations that had a fast retransmit
* to them. If so what we need to do is
* adjust ssthresh and cwnd.
*/
struct sctp_tmit_chunk *lchk;
int old_cwnd = net->cwnd;
/* JRS - reset as if state were changed */
htcp_reset(&net->htcp_ca);
net->ssthresh = htcp_recalc_ssthresh(stcb, net);
net->cwnd = net->ssthresh;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd),
SCTP_CWND_LOG_FROM_FR);
}
lchk = TAILQ_FIRST(&asoc->send_queue);
net->partial_bytes_acked = 0;
/* Turn on fast recovery window */
asoc->fast_retran_loss_recovery = 1;
if (lchk == NULL) {
/* Mark end of the window */
asoc->fast_recovery_tsn = asoc->sending_seq - 1;
} else {
asoc->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
}
/*
* CMT fast recovery -- per destination
* recovery variable.
*/
net->fast_retran_loss_recovery = 1;
if (lchk == NULL) {
/* Mark end of the window */
net->fast_recovery_tsn = asoc->sending_seq - 1;
} else {
net->fast_recovery_tsn = lchk->rec.data.TSN_seq - 1;
}
/*
* Disable Nonce Sum Checking and store the
* resync tsn
*/
asoc->nonce_sum_check = 0;
asoc->nonce_resync_tsn = asoc->fast_recovery_tsn + 1;
sctp_timer_stop(SCTP_TIMER_TYPE_SEND,
stcb->sctp_ep, stcb, net, SCTP_FROM_SCTP_INDATA + SCTP_LOC_32);
sctp_timer_start(SCTP_TIMER_TYPE_SEND,
stcb->sctp_ep, stcb, net);
}
} else if (net->net_ack > 0) {
/*
* Mark a peg that we WOULD have done a cwnd
* reduction but RFC2582 prevented this action.
*/
SCTP_STAT_INCR(sctps_fastretransinrtt);
}
}
}
void
sctp_htcp_cwnd_update_after_timeout(struct sctp_tcb *stcb,
struct sctp_nets *net)
{
int old_cwnd = net->cwnd;
/* JRS - reset as if the state were being changed to timeout */
htcp_reset(&net->htcp_ca);
net->ssthresh = htcp_recalc_ssthresh(stcb, net);
net->cwnd = net->mtu;
- Adds support for the multi-asconf (From Kozuka-san) - Adds some prepwork (Not all yet) for vimage in particular support the delete the sctppcbinfo.xx structs. There is still a leak in here if it were to be called plus we stil need the regrouping (From Me and Michael Tuexen) - Adds support for UDP tunneling. For BSD there is no socket yet setup so its disabled, but major argument changes are in here to emcompass the passing of the port number (zero when you don't have a udp tunnel, the default for BSD). Will add some hooks in UDP here shortly (discussed with Robert) that will allow easy tunneling. (Mainly from Peter Lei and Michael Tuexen with some BSD work from me :-D) - Some ease for windows, evidently leave is reserved by their compile move label leave: -> out: MFC after: 1 week
2008-05-20 13:47:46 +00:00
net->partial_bytes_acked = 0;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, net->cwnd - old_cwnd, SCTP_CWND_LOG_FROM_RTX);
}
}
void
sctp_htcp_cwnd_update_after_fr_timer(struct sctp_inpcb *inp,
struct sctp_tcb *stcb, struct sctp_nets *net)
{
int old_cwnd;
old_cwnd = net->cwnd;
- Locking compatiability changes. This involves adding additional flags to many function calls. The flags only get used in BSD when we compile with lock testing. These flags allow apple to escape the "giant" lock it holds on the socket and have more fine-grained locking in the NKE. It also allows us to test (with witness) the locking used by apple via a compile switch (manually applied). Approved by: re@freebsd.org(B Mah)
2007-09-08 11:35:11 +00:00
sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_EARLY_FR_TMR, SCTP_SO_NOT_LOCKED);
- added pre-checks to the bindx call. - use proper tick gathering macro instead of ticks directly. - Placed reasonable boundaries on sets that a user can do that are converted to ticks from ms. - Fix CMT_PF to always check to be sure CMT is on. - Fix ticks use of CMT_PF. - put back code to allow asconfs to be queued while INITs are in flight and before the assoc is established. - During window probes, an ack'd packet might be left with the window probe mark on it causing it to be retransmitted. Change so that the flight decrease macro clears the window_probe mark. - Additional logging flight size/reading and ASOC LOG. This is only enabled if you manually insert things into opt_sctp.h since its a set of debug code only. - Found an interesting SMP race in the way data was appended which could cause a reader to lose a part of a message, had to reorder when we marked the message was complete to after the data was appended. - bug in ADD-IP for the subset bound socket case when the peer has only one address - fix ASCONF implicit success/error handling case - proper support of jails in Freebsd 6> - copy out the timeval for the 64 bit sparc world on cookie-echo alignment error crashes without this). Approved by: re(Ken Smith)
2007-07-17 20:58:26 +00:00
net->htcp_ca.last_cong = sctp_get_tick_count();
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
/*
* make a small adjustment to cwnd and force to CA.
*/
if (net->cwnd > net->mtu)
/* drop down one MTU after sending */
net->cwnd -= net->mtu;
if (net->cwnd < net->ssthresh)
/* still in SS move to CA */
net->ssthresh = net->cwnd - 1;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, (old_cwnd - net->cwnd), SCTP_CWND_LOG_FROM_FR);
}
}
void
sctp_htcp_cwnd_update_after_ecn_echo(struct sctp_tcb *stcb,
struct sctp_nets *net)
{
int old_cwnd;
old_cwnd = net->cwnd;
/* JRS - reset hctp as if state changed */
htcp_reset(&net->htcp_ca);
SCTP_STAT_INCR(sctps_ecnereducedcwnd);
net->ssthresh = htcp_recalc_ssthresh(stcb, net);
if (net->ssthresh < net->mtu) {
net->ssthresh = net->mtu;
/* here back off the timer as well, to slow us down */
net->RTO <<= 1;
}
net->cwnd = net->ssthresh;
- Macro-izes the packed declaration in all headers. - Vimage prep - these are major restructures to move all global variables to be accessed via a macro or two. The variables all go into a single structure. - Asconf address addition tweaks (add_or_del Interfaces) - Fix rwnd calcualtion to be more conservative. - Support SACK_IMMEDIATE flag to skip delayed sack by demand of peer. - Comment updates in the sack mapping calculations - Invarients panic added. - Pre-support for UDP tunneling (we can do this on MAC but will need added support from UDP to get a "pipe" of UDP packets in. - clear trace buffer sysctl added when local tracing on. Note the majority of this huge patch is all the vimage prep stuff :-)
2008-06-14 07:58:05 +00:00
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_CWND_MONITOR_ENABLE) {
- Modular congestion control, with RFC2581 being the default. - CMT_PF states added (w/sysctl to turn the PF version on) - sctp_input.c had a missing incr of cookie case when the auth was bad. This meant a free was called without an increment to refcnt, added increment like rest of code. - There was a case, unlikely, when the scope of the destination changed (this is a TSNH case). In that case, it would not free the alloc'ed asoc (in sctp_input.c). - When listed addresses found a colliding cookie/Init, then the collided upon tcb was not unlocked in sctp_pcb.c - Add error checking on arguments of sctp_sendx(3) to prevent it from referencing a NULL pointer. - Fix an error return of sctp_sendx(3), it was returing ENOMEM not -1. - Get assoc id was changed to use the sanctified socket api method for getting a assoc id (PEER_ADDR_INFO instead of PEER_ADDR_PARAMS). - Fix it so a peeled off socket will get a proper error return if it trys to send to a different address then it is connected to. - Fix so that select_a_stream can avoid an endless loop that could hang a caller. - time_entered (state set time) was not being set in all cases to the time we went established. Approved by: re(ken smith)
2007-07-14 09:36:28 +00:00
sctp_log_cwnd(stcb, net, (net->cwnd - old_cwnd), SCTP_CWND_LOG_FROM_SAT);
}
}
Reference in New Issue Copy Permalink